Dynamics of the Control of Body Pattern in the Development of Xenopus Laevis 1

J. Embryol. exp. Morph. 88, 85-112 (1985) 85 Printed in Great Britain © The Company of Biologists Limited 1985

Dynamics of the control of body pattern in the development of Xenopus laevis 1. Timing and pattern in the development of dorsoanterior and posterior blastomere pairs, isolated at the 4-cell stage

JONATHAN COOKE AND JOHN A. WEBBER National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7, U.K

SUMMARY Xenopus embryos have been selected in which the second cleavage is occuring in a frontal plane, i.e one tending to lie at right angles to the prospective plane of bilateral symmetry for the body pattern. Some of these have been used to deduce a map of the disposition of materials for the normal mesodermal pattern (the normal 'fate map') by injecting blastomeres to found fluorescently marked clones from 4- to 32-cell stages. Other such 4-cell embryos have been separated into two isolates across this second cleavage; in fate-map terms, prospective dorsoanterior and posterior isolates. These have been allowed to develop to control axial larval stages, with examination of the time schedule of their gastrulation movements in relation to cofertilized whole controls. The patterns of mesoderm produced have been examined and interpreted in the light of quantitative knowledge about the normal pattern, and our current understanding of the map. A meaningful fate map exists for the egg material even at this early, essentially acellular stage, and it differs appreciably from what might have been expected in view of that traditionally shown for early gastrula stages. The patterns developed in the isolates show that at least in many eggs, widespread information that positively specifies material as to its body position is available from at most 1 h after the events that give rise to bilateral symmetry upon fertilization. This information usually leads to a mosaic development of the appropriate mesodermal part-pattern in dorsoanterior isolates, and frequently allows development that approximates to this in the reciprocal posterior part. Regulation, i.e. the replacement of removed information to specify a development more complete than the normal contribution in isolates, is not observed. The results suggest a revision of former claims for regulative ability in at least this amphibian embryo. They also imply that systems for ascribing position value (positional information) to early embryonic tissue can be diverse in dynamics, even among embryos whose body plans are obviously homologous as are those of vertebrates.

INTRODUCTION This is the first of a series of papers which will report a re-appraisal of pattern formation in the development of an anuran amphibian, using the contemporary techniques of cell lineage marking and the quantitative measurement of anatomy at a particular early larval stage. These techniques have been used alongside various traditional experimental manipulations, performed on newly fertilized

Key words: egg organization, pattern specification, gastrulation, fate mapping, non-equivalence, Xenopus laevis. 86 J. COOKE AND J. A. WEBBER eggs and early embryos of the African Clawed Frog, Xenopus laevis. An overall attempt to introduce these studies in terms of the inherited beliefs and concepts of experimental embryology would result, rather, in a lengthy review. The reader is referred instead to reviews by Holtfreter & Hamburger (1955), Nieuwkoop (1973) and Gerhart (1980, especially relevant for the events before first cleavage). In each of the present papers, revisions of particular previously held beliefs that are indicated by results to be described will be emphasized in the appropriate discussion sections. Attempts will be made at appropriate points to inter-relate the findings as a whole. This and the next paper deal with the traditional separation of the fertilized egg into defined fragments at very early cleavage stages, and their development in isolation. The results are partly at variance with the traditional reports, and resemble those recently documented by Kageura & Yamana (1983). They may help in understanding the nature of the information that underlies development of the body plan in embryos of this general type (Wolpert, 1971; Cooke, 1983a, b). Spemann and his students, working with urodele embryos, recorded that a special limited region of the egg could be defined, at least from soon after fertilization (Spemann, 1902, 1938). Constrictions were made, at times up to the onset of gastrulation, to produce developing fragments each reasonably balanced for constituents in the animal-vegetal axis. Any such fragment that included the region destined to become the dorsal midline in the normal course of embryo- genesis, could apparently develop an essentially complete, balanced body pattern though of reduced tissue dimensions (i.e regulation). Fragments lacking this region revealed, by their retention of radial symmetry and lack of dorsal axial histogenesis, a profound loss of developmental information. The crucial 'pre-localized' region was originally believed to be quite limited, subtending some 60° of arc in the egg, but later experiments of similar sorts revealed that the situation was more complex (Brice, 1958; Dollander, 1950; Fankhauser, 1948) and that a wider region was capable of organizing at least a significant partial degree of development. The extent to which the experimental patterns were analysed histologically and quantitatively, to see if they accorded with their 'gestalt' appearance, is also unclear. This work has nevertheless been the classic evidence for the idea that prolonged interactions or signalling processes across tissue, extending through blastula and into gastrula phases (104 cells), set up the body pattern by progressive regionalization during the normal development. This development has in its turn been treated as a model system in which to study early pattern formation in vertebrates generally. The mechanism has been thought of sometimes as a hierarchy or succession of specific inductions (Spemann, 1938; Nieuwkoop, 1973; Slack, 1983; see Cooke, 1983a) and sometimes as the setting up of a physiological gradient interpreted as positional information of some kind (Cooke, 1912a, 1973; Wolpert, 1971). We have re-investigated the states of specification for patterned development across the egg by capitalizing upon the recently expanded knowledge of the immediate postfertilization events in anuran amphibians, particularly Xenopus Pattern in Xenopus blastomere pairs 87 (Gerhart et al. 1981; Kirschner, Gerhart, Hara & Ubbels, 1980). By appropriate early manipulation and selection of eggs, preparations can be made in which the two pairs of blastomeres separated at the 4-cell stage will constitute, respectively, that part of the egg surrounding the presumptive dorsal midline and the remainder. It is known that the system organizing the future pattern is first orientated during the precleavage period by reorganization of plasms within the egg. This normally occurs in relation to the position of sperm entry which can be used as a marker. At second cleavage (orientated vertically like the first but orthogonal to it), we have selected eggs with a bilaterally symmetrical cleavage pattern lying orthogonal to the presumptive plane of symmetry of the embryo. Some were then used to fill particular bilateral pairs of blastomeres with a cell- lineage tracer (Gimlich & Cooke, 1983), and thus trace their relative contributions to mesodermal body pattern in normal development. Those contributions turn out to be quite characteristic and constant, so that a fate-mapping exercise has been performed, and the deduced map for mesoderm differs from what might be expected in view of previously published fate maps for amphibian early gastrula stages. The remaining eggs were used to perform separations into the two presumed bilaterally symmetrical pairs of blastomeres, followed by their devel- opment as isolates. The partial body plans developed by these reciprocal isolates were observed and representative samples examined histologically. Their meso- dermal component was studied in relation to our understanding of the fate map for whole development, and our knowledge of certain quantitative regularities in the normal pattern (Webber & Cooke, in preparation). We conclude that, by onset of the 4-cell stage, the informational system across the egg is frequently in such a condition as to lead to stable retention, by the separated parts, of the ability to develop in a near-mosaic way their reciprocal contributions to the normal whole mesoderm. Regulative behaviour - that is, production by a fragment of a greater part of the pattern than its presumptive contribution in the intact embryo - was not observed. Mosaic behaviour is not the invariable outcome however, and the whole spectrum of results is discussed in terms of the possible nature of the primary system of spatial information in the egg. Our understanding of the flow of developmental information that sets up and co-ordinates the principal layers of the body structure (germ layers) is also discussed in relation to the present results, in order to justify our concentration upon mesoderm as the most direct expression of the original system coding for 'body position'. We stress that the results reveal an early specification for position within the body plan (see Slack, 1983 for a defined usage of this term), and in no sense a set of determinations for tissue types. The behaviour which is described indeed precedes by several hours the origin of the mesoderm-forming tissue by cell interactions in the animal-vegetal dimension of the embryo (Sudarwati & Nieuwkoop, 1971). There is no incompatibility between a demonstration of specification, however early, and other findings that the normal fates of moderate-sized fragments of the material can be altered at much later stages by their experimental re-positioning 88 J. COOKE AND J. A. WEBBER within the embryo or in culture (Smith & Slack, 1983; Forman & Slack, 1980). But when the present results are related to those for other types of miniature embryonic body, as in the discussion of paper II in this series, they cause us to question previous interpretations that true regulation of pattern can occur in this system (Cooke, 1975, 1979, 1981, 1982).

MATERIALS AND METHODS Preparation of isolated blastomere pairs Females were injected with 350i.u. of human chorionic gonadotrophin (TregnyF Organon Ltd) and held at 20°C overnight. Batches of 1-200 eggs were stripped from them at intervals, between 15 and 20 h after injection, into wet glass Petri dishes. Each batch was immediately covered with fresh testis suspension in 30 % Ringer. Full-strength Ringer in this work is 0-1 M- NaCl, 2mM-KCl, 2mM-CaCl2 and lmM-MgCl2, buffered to pH7-3 with NaHCO?. Newly fertilized eggs were flooded with 20 % Ringer within lOmin. Within 30min after fertilization at 20 °C, eggs were de-jellied by standing for 10 min and then gentle agitation in 2 % cysteine HC1 in distilled water, brought to pH7-9 with NaOH. They were then washed and allowed to fall, through a layer of 5 % Ficoll (Sigma - type 400) in 20 % Ringer, onto a perspex grid of wells, each of about 70 % of typical egg diameter. Examples with prominent sperm entry points well down from the animal pole were rapidly selected and aligned under the dissecting microscope, so that animal poles faced upward and sperm entry meridians (see Fig. 1A) were all aligned as a record of the presumptive midline for pattern formation. Such eggs rapidly became pinned against any free-floating rotations in gravity, due to collapse of the perivitelline space by the Ficoll. Each egg was then manually rotated by some 30° in the sperm entry meridian so as to lower the s.e.p. in gravity, as this is understood to increase the accuracy with which the planes of sperm entry and bilateral symmetry co-incide by re-inforcing the shifts of egg plasm that are organized in relation to the spreading sperm aster during the remaining hour before first cleavage. In a pilot batch of 50 eggs treated in this way, over 80 % developed their dorsal midlines within 20° of the position opposite sperm entry, while none showed a deviation of as much as 45°. In each co-fertilized batch, when the second vertical cleavage furrow had progressed some halfway around the eggs from their animal poles (some lh and 45min after selection and orientation), a set of up to ten was re-selected in which the first cleavage had co-incided most closely with that plane originally established by sperm entry. Such eggs were demembranated with finely ground forceps while still in Ficoll, and immediately transferred to loosely fitting wells in 2% agarose under 66% Ringer which had been buffered to pH6-4 with 20mM-Na phosphates. Over the following 20min the products of the new cleavage plane could be separated to give two bilaterally symmetrical but different pairs (see Fig. 1C), by gentle blowing of the solution from a fine pipette against the new membrane in the deepening furrows. As the eggs stayed essentially vertically orientated in their wells throughout this process, undue distortion or tipping of the material in either of the separating pairs was minimized. Examples in which such noticeable distortion or tipping occurred were discarded. Some sperm-entry side isolates were made with even less disturbance, by awaiting the completion of cleavage, cutting away the material of the other pair with tungsten knives and then gently blowing off the residual membrane and cytoplasm. Neither procedure for creating isolates involved tipping for more than a minute at any point, whereas tipping for many minutes is believed to be required for artificial dorsalization or significant activation of pattern formation. Isolates were placed, as reciprocal pairs whenever possible, in smaller wells to continue development without the excessive flattening or monolayering of cleavage products that occurs on a plane surface. The solution was changed gradually to 20 % Ringer, pH 7-0-7-3 during early blastula stages when abnormal 'internal' type cell surface was no longer visible (see Observations). Controls within each synchronous batch were eggs (not necessarily so highly selected for direction of first cleavage) that had been demembranated and processed along with the experimental individuals until only the spindle mid-bodies appeared to join the blastomeres Pattern in Xenopus blastomere pairs 89

Fig. 1. Preparation of dorsoanterior and posterior isolates. (A) View from the animal pole at the 4-cell stage emphasizing the second cleavage plane in which the separation is made. Third cleavage is horizontal, thus in the plane of the page, and the normal partitioning of the egg by the vertical fourth cleavage planes is indicated dashed (see Fig. 2, the 16-cell stage). Sperm entry and dorsoanterior meridians are indicated by bracket and arc, respectively. Scale bar equals 0-75 mm approx. (B) A pair of isolates and a control in their wells at the morula stage. Positions of cleavage planes usually deviate from those that would normally have occurred, in control as well as exper- imental eggs after treatment. Areas of transient exposure of new, 'inner' type mem- brane are shown stippled. from each pair, but which were then left in full-size wells to re-adhere and develop as whole embryos. Early observation had revealed that the process of blowing for several minutes upon new cell surface to separate blastomeres causes, in itself, a relative delay of some 20-25 min in the earliest dorsal lip appearance many hours later. Embryos treated as above were therefore the correct controls for the following comparative observations.

Observation of comparative time courses of gastrulation Development of each matched set of embryos and isolates was slowed by cooling to 15 °C overnight, commencing at control 64-cell stage. Fifteen hours later, on return to 20°C, the late blastulae could be inverted in their wells in time for the onset and progression of the external pattern of gastrulation to be monitored by observation and drawing every 20 min (see Cooke 90 J. COOKE AND J. A. WEBBER 1972ft; Nieuwkoop & Florschutz, 1950; Keller, 1976). Under the conditions of equivalent treatment given here, co-fertilized whole embryos commence dorsal lip formation (stage 10 - Nieuwkoop & Faber, 1967) within at most 20 min of each other, and the later time of mid-ventral completion of the blastoporal ring is only a little more variable.

Fate mapping by lineage tracing Embryos to be used for this work were prepared and individually selected as for the blastomere separations. Single blastomeres or more usually bilateral pairs, corresponding to the isolates, were injected with fluoresceinated lysine-dextran (FLDX) as a lineage-linked tracer (see Gimlich & Cooke, 1983, for injection concentration and procedures for fixation, embedding, sectioning and visualization of the descendant cells within the structure of the tailbud larva). Additional cases were also analysed where bilateral pairs of blastomeres contributing to mesoderm from 8-, 16- or 32-cell stages had been filled. This extended our understanding of the geographical distribution of the presumptive regions within the egg, to enable the deduction of the fate maps of Fig. 2. Reconstruction of the contributions of blastomeres to the mesodermal layer was carried out by sampling the larval transverse section at twelve equal intervals between anterior notochord tip and tailbud, and recording the FLDX- filled domain on a form containing a standard series of sectional profiles of the normal stage-30 body. Somite numbers for filled domains were checked by horizontal sectioning.

Quantitative comparison of the mesodermal part-patterns produced by isolates with those of whole siblings Some representative reciprocal pairs of isolates, and synchronous controls, were fixed for quantitative analysis (see Cooke, 1979, 1981, 1983b for details of fixation, staining with Feulgen/light green/Orange G, serial sectioning at 1 pan and cell counting from regularly distributed sections). The standard stage chosen for examination of pattern throughout this

Fig. 2. Egg regions contributing to mesoderm from early cleavage stages. (A) Distribution of materials contributing to mesodermal pattern, superimposed on a lateral view of the idealized 16-cell partitioning of the egg. The second cleavage plane is emphasized in this and the following diagram. The future left side of the body is represented (cp. 2B), as the involution movements of gastrulation turn the mesoderm inside out while extending it to give the long cylinder of the larval mesodermal mantle. Mesoderm is never at the surface, and should be thought of as deriving from the inner parts of marginal blastomeres by the tangential divisions that begin at blastula stages. Cleavage lines are continued off the egg outline to distinguish them from fate map lines. (B) Formalized plan of the stage-30 larval mesoderm. Its partitioning by the second (2-4 cell) cleavage plane is indicated by the heavy line, while lighter lines indicate typical partitioning by the fourth (8-16 cell) planes that establish four sectors on each side of eggs. The somitogenic columns are considered as simple strips of para- chordal mesoderm, as they are in reality before myocoel formation. Precise con- tributions to mesodermal pattern made by marked clones vary according to precise positions of planes in individuals, and in particular the number and extent of somite segment contributions lying between planes, though truly represented, is an average from the studied material. The coherence and order of the mapping, however, diminish only where very early-founded marked clones participate in tail mesoderm, whose pattern is probably the latest specified and where much growth then occurs before morphogenesis. (C) A large marked clone extending from ventral mesoderm into primitively lateral somite domain, before and after the dorsal convergence and cavitation that accompany somite formation. In this way it can be understood how a 'posterior' blastomere makes a pair of discrete contributions to each somite as seen in Fig. 3, but also how the size of those contributions increases for successively more posterior somites as shown in Fig. 2A,B. Neural plate omitted from left diagram, s, somite segment; n, notochord; pn, pronephros; Ip, lateral plate; pc, prechordal. Pattern in Xenopus blastomere pairs 91 work, as in previous studies, was the late tailbud, stage 29-30. For most of the work in this and the subsequent paper on 2-cell separations the previously standard, transverse sectional series was used for cell counting. Certain experimentals and controls were sectioned horizontally however, to aid in comparison of the real size of notochord cell populations. Somite numbers segmented at the stage of fixation were usually directly countable in oblique lighting under the dissecting microscope.

D

pc pn

B 92 J. COOKE AND J. A. WEBBER

OBSERVATIONS The fate map by lineage tracing As compared with the traditional in situ vital staining experiment, the use of a satisfactory injected lineage tracer provides superior information regarding normal contributions to the body, and degrees of intermingling of descendant cells, from each part of the early cleaving egg. Single cells descended from labelled blastomeres can be distinguished unequivocally from neighbours in the early larva where multiplication to the order of 105 cells has occurred with as yet little true growth. There are interesting differences in the degree of cell mixing occur- ring during morphogenesis, among cells destined for the different germ layers (Smallcombe & Cooke, in preparation), but here we are concerned only with normal contributions to mesoderm. The first noteworthy aspect of the results is that it is indeed meaningful to speak of a fate map at pre- or early cleavage stages. This is not the case for mammalian development, and very unlikely to be so for that of other amniotes or for teleosts. But the relative constancy of the observed contributions from particular blastomeres in our selected cleavage pattern is evidence only that the materials of the egg are used in a consistent way relative to the 'reference' plane of bilateral symmetry and to gravity. In individuals with various other perfectly viable cleavage patterns there are correspondingly different sets of clonal contributions to an identical body pattern. There is thus no suggestion, in development of this type, of the kind of fixed lineage or cleavage programme that builds up the body in certain invertebrate embryos (e.g. Sulston, Schirenberg, White & Thomson, 1983). Fig. 2A, based on some 40 embryos with blastomeres filled at 4- to 32-cell stages, shows the distribution of materials contributing to the mesodermal pattern superimposed upon the cleavage planes of the ideal 16-cell stage as seen from the future left side (the movements of gastrulation invert the head-tail axis of the embryo by turning the mesodermal material 'inside-out'). 2nd and 3rd cleavage furrows have usually been so placed as to cause this stage to consist of a somewhat smaller dorsoanterior and larger posterior octet, and a distinctly smaller animal than vegetal octet. The vertical second cleavage is most relevant as this gives rise to the isolates whose developmental capacities are studied here. It separates off a sector containing the precursor of the classical 'organization centre' (Spemann, 1938; Nieuwkoop, 1973) from the remainder of the egg. Figs 2B and 2C display principally the partitioning of the mesodermal pattern of the tailbud larva by this cleavage, although the contributions of the four sectors between sperm-entry and mid-dorsal meridians, segregated at the later 16-cell stage, are also indicated in Fig. 2B. It can be seen that cellularization, epiboly, gastrulation and dorsal convergence successively bring about a great geometrical transformation as between the egg/morula and the larva. Great coherence of neighbour relationships among cellular relatives is nevertheless preserved during recruitment, migration and Pattern in Xenopus blastomere pairs 93

3A B-'

Fig. 3. Fluorescently marked clones in transverse sections of the larval structure, as used in fate-mapping. (A) T.S. of somite, notochord and epidermis at mid-trunk level (Somite 10-12), after filling a posterior 4-cell blastomere. Note the unfilled notochord and adjacent domain in the somite, with relatively little intermixture of descendant cells. The more brightly fluorescing profiles in somite are sections through nuclei, which also show up in the adjacent yolky endoderm cells derived from the same blastomere. (B) T.S. through postblastoporal tailbud region after filling both posterior 4-cell blastomeres, but in a case where the first cleavage had deviated appreciably from the sagittal plane for pattern. Thus the unlabelled juxtachordal somite contribution from the dorsoanterior egg region, normal for posterior somites, is much reduced to the right of the notochord and unusually expanded on the left. From this we can deduce that one posterolateral boundary of the notochord anlage reached almost to the cellular domain deriving from the posterior side of the second cleavage. morphogenesis throughout the mesoderm, diminishing only in the posterior tailbud where morphogenesis follows a great deal of growth from a very small region of the original material. Sectional appearances of particular fluorescently marked domains in larvae are shown in Fig. 3. In undisturbed development the majority of mesodermal cells descended from an individual blastomere, from the 4-cell even up to the 64-cell stage, are in contact with one another and form essentially a solid patch within the larval tissue. Such patches show no evidence, however, of relating to boundaries between anatomical domains or structures. The patch deriving from a single 'dorsal' 32-cell blastomere of the second tier from the 94 J. COOKE AND J. A. WEBBER animal pole, for instance, may contribute to prechordal plate, to notochord, and to every somite on one side of the body. The distances across which cell inter- mingling has given rise to peripheral clusters and isolated members surrounding main clonal patches are small in relation to the principal dimension of a patch originating at, say, the 8-cell stage. The fate map of Fig. 2, essentially from the egg stage, is strikingly similar to that ascertained by a new labelled-grafting technique at the onset of gastrulation (Webber & Cooke, in preparation). These maps differ in some important respects from previously published and 'textbook' interpretations for the respective stages of anuran amphibian development (e.g. Nakamura & Kishyama, 1971). The hori- zontal dimension around the pregastrular marginal zone is traditionally thought of as a 'dorsal-to-ventral' dimension for future pattern, but Fig. 2 shows that the mapping of larval body position onto this dimension, from the meridian of the presumptive dorsal lip to that of sperm entry, is in many respects a head-to-tail one. The lateral plate territory is distributed in this way as a belt around the vegetal marginal zone, and above it, the 'average' final body position for material contributed to somites passes from the head towards the tail end of the series as one passes from what will be the first-invaginating region (Keller, 1976) towards the sperm-entry meridian. Considerable numbers of posterior somites are derived principally from material in the 'ventral' 90° sector segregated at the 16-cell stage, and considerably more than the posterior 50% of the total somite cells are descended from the sperm-entry side of the second cleavage, i.e. the material of our 'posterior' isolates. The blastomeres opposite the original s.e.p. at the 4-cell stage produce the entire notochord, and make some contribution to all the somites, but quan- titatively their derivatives are overwhelmingly anterior. They will thus be referred to as the dorsoanterior pair (or isolate). A specific profile of relative tissue mass per somite, in different regions of the somite series, characterizes the normal body (Webber & Cooke, in preparation), and the differential contributions to this from the various sectors of the early material are also characteristic. But as seen from the animal pole aspect, the derivation of each somite is so widespread that material for most individual segments of the series is gathered from around much of the equatorial region on each side. Only the anterior 4-8 somites (dependent upon the degree of eccentricity of second cleavage) are entirely derived from the dorsoanterior pair, whereas even tail somites behind the 30th receive a juxta- notochordal contribution from those same blastomeres. In passing away from the sperm entry meridian around the belt of the somite territory one encounters material that will contribute to the lateral edges of increasingly anterior somite segments, reaching on average somite 6 in crossing from the original posterior to the dorsoanterior side of the second cleavage plane. Comparison of the schematic larval body plan of Fig. 2B with the more realistic block diagrams of sectors of the dorsal axis in Fig. 2C shows how the somite material becomes arranged. A clonal patch in the primitive mesodermal mantle which reaches only to the lateral edge of any particular somite territory (i.e. that Pattern in Xenopus blastomere pairs 95 furthest from the notochord anlage within the original cell sheet), comes to reside in two populations at dorsal and ventral 'wings' of that somite's cross section (see also Fig. 3). This is due to the later morphogenesis (Hamilton, 1969) in which a split, the derivative of the primitive myocoel, appears in the somitogenic columns during dorsal convergence movements. In the sample of 4-cell reciprocal fills used to establish these maps, the restricted territory of the pronephros bulb lies athwart the second cleavage plane, with its material often coming almost entirely from the posterior and never so much as half from the dorsoanterior cell pair. Gill and other head mesenchymes which push ventrally in the larval head, are of neural crest, and not primary mesodermal origin. This has been confirmed in embryos where animal cap blastomeres were filled at 32-cell stage, and primitive prechordal true mesoderm (unlabelled) could be distinguished from crest-derived mesenchymes (labelled and adjacent to labelled brain) that were intimately associated with them (Gimlith & Cooke, unpublished).

Patterns of mesoderm developed by dorsoanterior and posterior isolates Development in isolation of dorsoanterior cell pairs was relatively invariant, provided that good early restoration of the blastula configuration had occurred. 47 such successful isolates were observed. The mesodermal patterns achieved corresponded, qualitatively and quantitatively, to the normal contribution of these blastomeres to the whole pattern. There was full prechordal head structure and a notochord whose size (cell population) was indistinguishable from that in within- set controls. The series of somite segments was usually numerically complete, but its major, posterior part was represented only by small, shallow blocks of cells alongside the notochord (cp. Figs 2 and 3). The lateral plate tissue was substantial but entirely anteriorly positioned. The yolky endoderm took up a shape and distribution conforming to the partial mesodermal pattern - and indeed to the normal endodermal contribution from this blastomere pair - giving the charac- teristic 'pigeon-chested' configuration with very slim trunk and small tailbud with near-terminal blastopore, illustrated in Fig. 4B. Histological analysis of this body form confirmed that the anterior 4-6 somites, relatively few-celled and shallow in the complete larva and derived entirely from the relevant cell pair, retained approximately their normal construction. The remainder of the complement however, normally of somites progressively more massive in cross section, con- sisted instead of a set of reduced cell populations, comparable at most to those normally contributed to each segment by the blastomeres in question. This body form can be understood by imagining the dorsoanterior mesodermal contribution (Fig. 2) in isolation, with appropriate co-ordination of neurectoderm and yolky endoderm, and from comparing Fig. 5A and 5B. Only three of the nine such isolates examined histologically had developed appreciable pro-nephroi. Those without pro-nephric formations included four examples of a version of the isolated development showing a more extreme 'jug handle' body form in which the 96 J. COOKE AND J. A. WEBBER deep and very fully developed head and chest region, of near normal size, is followed only by a dorsally curling appendage lacking endoderm or fin and with entirely notochord and somite mesoderm, ending in an entirely terminal blastopore (Fig. 4C). Quantitative comparison of these particular bodies with the

Fig. 4. Partial body forms developing from isolates made across the second cleavage plane. (A) Typical control development at the stage used for examination of pattern. (B) Less extreme and (C) more extreme (commoner) dorsoanterior part patterns. B probably represents mosaic and C, a slightly more restricted than mosaic development of the normal blastomere contributions to the mesodermal pattern. (D) The fully axial development that can frequently be seen from isolates reciprocal to those giving B or C. Such a body represents mosaic development of the blastomere contributions to mesoderm, though with a nervous system part-pattern derived from quite other precursor cells than normal. (E) Significantly axial development of the posterior isolate, but containing a lower percentage of somite cells, and contributions to fewer segments of the plan, than the normal contribution. (F) Development corresponding to the 'belly-piece' of Spemann, containing lateral plate and blood-island mesoderm with little or no somite and no neural induction, fb, forebrain; ev, hindbrain level and ear vesicle; eg, cement gland; g, gill structure; e, eyecup; pn, pronephros; bp, proctodaeum (original blastopore). Somite segment numbers indicated approximately in relation to other structure. Scale bar represents 1 mm approx. Pattern in Xenopus blastomere pairs 97

Fig. 5. Internal pattern in development of isolates. (A) Trunk level in typical control development. Note the large yolky endoderm profile and deep somite segments. (B) Trunk level of pattern as in Fig. 4B, with slender yolky mass and small, shallow somites in relation to the full-sized notochord. (C) T.S., passing to horizontal section posteriorly, of extreme body form as in Fig. 4C. Note the abrupt posterior termination of the massive, fully differentiated notochord and disorganized somite. (D) Anterior level of a fully axially developed posterior isolate as in Fig. 4D. Note the well-formed pro-nephroi and small anterior-most somite mass, but absence of notochord. (E) Trunk level of body shown in (D). Note the more massive somite profile at the more posterior axial position. (F) Tailbud region of an isolate showing a small notochord structure absent from all more anterior levels. See text for probable explanation of such development. Scale bar represents 0-5 mm. s, somite; sp, spinal cord; y, yolky endoderm; n, notochord. Other labelling as in Fig. 4. 98 J. COOKE AND J. A. WEBBER normally extended pattern was difficult because of the variable plane of section involved, but careful comparison with both transversely and horizontally sec- tioned material suggested that as in dorsoanterior isolates generally, their noto- chord cell populations were not distinguishable from those produced in whole embryos. In the most extreme cases the segmentation in the rear somite material, which was well histodifferentiated and ensheathed the massive notochord, was unclear. Both differentiated notochord and somite masses terminated abruptly near the blastopore without the posterior immature or predifferentiated zone seen in normal or other experimental bodies (Fig. 5C). More than half of all these isolates (26/47) developed part-body patterns tending towards the more extreme version associated with the foregoing internal features. The numerical proportion of cells in lateral plate of dorsoanterior isolates was little, if any, less than that normal for controls (30-42%, mean 38-5% control 36-48 %, mean 44-2 %). But this was almost confined to levels opposite the first few somites, a distribution understandable in terms of a near-normal contribution according to the fate map (Fig. 2B). Pattern in the posterior isolates was much more profoundly variable, even among those with normal blastocoel closure after isolation. 48 such isolates were observed, and their mesodermal morphology varied between that suggesting once more a mosaic part-pattern according to the normal contribution, and the radially symmetrical cylinder of lateral plate differentiation which is the classical 'bauchstucke' (belly-piece) of Spemann. 37 examples were members of successful reciprocal pairs of isolates, and no correlation was apparent between their modes of development within the spectrum and the classification (extreme 'jug handle' configuration or simple mosaic dorsoanterior development) of their other halves. 16 of the most highly organized 'posterior' bodies displayed a pattern ap- proximating closely to the reciprocal of the typical dorsoanterior isolate pattern in terms of mesoderm. These contained well-formed pro-nephroi, slightly fewer than the normal body complement of well-developed somite segments, and about 50 % of their mesoderm cells as lateral plate. In the total absence of notochord, somite bodies were bridged across the midline and the pronephroi were situated at the anterior end of the series (Figs 4D, 5D,E). Somite number was three or four less than that formed in synchronously fixed controls, and the tailbud was of control size and developmental stage. The presence of a pair of ear vesicles flanking the widened, hindbrain-like anterior end of the induced neural tube, and just ahead of the first clearly segmented mesoderm, was an additional marker for pattern as judged by the field of inductive influence in mesoderm. 12 isolates were classified as radially symmetrical lateral plate only. The remaining 20 formed a range of intermediate morphologies in which there was no pro-nephric or ear vesicle development, and where a progressive diminution in the proportion of mesoderm as somite cells in relation to lateral plate was accompanied by production of progressively smaller numbers of segments, always followed by a terminal predifferentiated zone behind blastopore level and corresponding to tailbud (Fig. 4E). Pattern in Xenopus blastomere pairs 99 Fig. 6 shows how, in embryos of this series, the profiles of somite sizes (maximum cell counts in T.S.) against position in the body are interestingly related to the profile seen in the whole body pattern. As already mentioned, somites from different positions in the normal series incorporate characteristic relative amounts of the total somite tissue, as well as drawing characteristic proportions of their material from different sectors of the embryonic somite territory. The spatial patterns of somite contributions seen in these posterior part-bodies are reminiscent of the specific contributions to be expected from variously restricted 'posterior' (near the s.e.p.) egg sectors to the complete development (see Fig. 2B). The complete absence of notochord appeared not to affect axial elongation as such.

100-

80-

60-

40-

20-

12 16 20 23 Segment no.

Fig. 6. Relative sizes and numbers of somite contributions in development of posterior isolates. Horizontal bars represent the largest nuclear counts in each segment (back to somite 26) in three control embryos. From this the shape of the normal profile whereby somite cell populations differ along the series can be seen. Normal somitogenic columns have a particular shape, being at their most substantial between the positions of segments 15 and 20. The four histogram outlines (two heavier and two finer lined) show the same measurements in two fully axial (A,B) and two less complete (C,D) posterior isolates. If the ear vesicles (where present, in cases A and B) and largest somites seen are lined up in the plots as shown, the experimental profiles resemble quantitatively the sets of contributions to somite morphology that would be expected from sub-regions of the egg extending outwards from the meridian of sperm entry, in cases (A,B) as far as the second cleavage plane used to make the isolates. This can be understood by reference to Figs 2A,B, 3, 7A,B. Maximum somite numbers in posterior isolates are thus three or four less than in the complete pattern, and in those cases are accompanied by a terminal ear vesicle and well-developed pronephros. 100 J. COOKE AND J. A. WEBBER The qualitative and numerical observations are all consistent with the following anatomical interpretation. The part-pattern represented in Figs 4B and 5B prob- ably corresponds most nearly with mosaic dorsoanterior contribution to meso- derm, in relation to the normal map. The extreme 'jug-handle' form of Figs 4C and 5C probably corresponds with a pattern contribution slightly more restricted, at its posterior boundary, than that which the blastomeres would normally make. The posterior somite and notochord zone which is scheduled still to be pre- differentiated at stage 30 is absent. The most highly axial morphology achieved by the posterior isolates also represents a mosaic contribution to mesoderm according to fate, whereas the rest of the series would represent production of part-patterns corresponding to the normal contributions from successively more restricted 'posterior' sectors of the egg. In the limiting case, information for somite tissue participating in all parts of the segment series is lost, or has never been acquired, by the isolate. The highest grade posterior isolates achieved a pattern of somite contributions corresponding at least to the 'highest' encroachment into anterior somites ever seen after a 4-cell embryo had been symmetrically filled in the posterior blastomeres. In three additional isolates designated as posterior, the body that was developed included a very short, thin but unmistakable notochord region in the tailbud (less than 1 % of the cells), dividing somites that were elsewhere bridged across as usual (Fig. 5F). Other features were as in the most highly axial notochordless part- patterns. The reciprocals of these cases were unfortunately not survivors, but they could perhaps be assumed to be examples where the initial cleavage plane had in fact deviated sufficiently from the midline of pattern that they represented intermediate separations in the series between those dealt with in this paper and the lateral halves described in the next (see Fig. 1 of this paper and paper II discussion).

Early development after isolations, and the schedules of gastrulation in relation to subsequent mesodermal patterns The abnormal zone of naked new membrane, caused by blastomere separation, appeared to be removed from the embryo surface by a bilaterally symmetrical inward rotation of blastomeres around the site, so that only externally normal- appearing cell junctions were apparent by around control 64-cell stages. This re- arrangement was aided by a deviation from their presumptive positions of the cleavage planes in cell cycles immediately following separation, such that more cells occupying a single tier were characteristically produced before a horizontal cleavage set in. The only major redistribution of material that was apparent, however, was the final suturing together of what would have been two lateral regions at a new (midline) position in each half embryo (Fig. IB). Only cases in which this occurred promptly were studied further in detail. Individuals in which an internal, blastocoel-like surface was exposed to the medium for much of the blastula period sometimes eventually healed, but always gastrulated on altogether late, extended time courses. If of dorsoanterior presumptive fate, such isolates Pattern in Xenopus blastomere pairs 101 finally formed externally 'balanced'-looking miniature bodies, though with varying degrees of spina bifida. These proved to have disproportionately small or absent notochords and head parts, however. Presumptive posterior such isolates invariably gave non-axial, 'bauchstucke' forms. This effect of early prolonged healing may be important and interesting in relation to the slight delaying effect upon gastrulation of the original method of separations, and to possible interpretations of the discrepancy between the morphogenetic results in this paper and in previous ones (see Discussion). Dorsoanterior cell pairs, thus isolates, are typically smaller than their posterior partners as mentioned earlier, though the tendency varies greatly between egg batches. A study of absolute cell numbers in isolates remains to be done using proper techniques, but counting around the diameters and marginal zones of several reciprocal isolates at control stage 10 revealed a smaller surface cell number in the dorsoanterior partners, which still look smaller as gastrulae. In larval pattern analysis (see previous subsection) the mesodermal cell counts of posterior but axial part-patterns slightly exceeded 50 % of the within-set controls, in conformity with expectation from the normal contributions. Dorsoanterior part-patterns, while appearing to contain less than 50% control cell numbers, were too different geometrically from the control plan for safe comparison to be made. There is thus a strong indication that the two types of isolate enter gastrulation with different cell numbers, but no indication that this represents a deviation from their role in normal development. The causes of variation in numbers of cells produced, from eggs and early blastomeres of different sizes, will be discussed in paper II of this series. Observations of cell numbers throughout the present work, however, are in line with previous findings that surgical interruption of tissues and changes of configuration do not of themselves affect the cell cycle schedule before pattern establishment is complete in this experimental system (Cooke, 1979, 1981). Dorsoanterior isolates initiate the external signs of gastrulation within 10 or 15min of their entire controls, i.e synchronously with them. Observation of isolates from egg batches in which the pigment gradation across the animal hemisphere was pronounced enough to act as a marker have not indicated that the dorsal lip forms on a meridian differing from its presumptive one in the egg of origin. The subsequent lateral spreading and fusion of the lines of surface bottle cell activity, to give a ring-shaped blastopore, completes itself precociously in these isolates however. They thus show the ring blastopore characterizing normal stage 11 or 11+ while controls are still in the spreading lip phase 10+ or 10!. Surface cell counting suggests, on some occasions, that this precocity is simply because of the smaller tissue extent around the marginal zone in isolates, with the spread of new cellular activity occurring at control rates on a 'per cell' basis. This would represent a mosaic partial time course of gastrulation, with all material behaving according to its local presumptive schedule. In other instances, however, the initial lip is broad and rapidly spreading in relation to the control con- figuration, so that progression to give a ring occurs on a more rapid time course per 102 J. COOKE AND J. A. WEBBER cell than is occurring in controls. This represents departure from presumptive schedule to give true precocity of cellular activity in the more 'lateral' material. Blastopore closure (stages, Hi-13 equivalent appearances) continues on a relatively precocious schedule in dorsoanterior isolates and is frequently followed by an abnormal pear-shaped appearance, where the closed blastopore becomes produced into a nipple-like structure that appears to give rise to much of the slender trunk region that is later found. Those batches of isolates in which the truly precocious spread of gastrulation followed by the nipple-like blastopore was most pronounced, also gave rise to bodies that suggested the most restricted anterodorsal part-pattern of mesoderm in relation to the fate map (see previous subsection). They also included the smallest isolates, in some cases because of relatively great eccentricity of the second cleavage, and in others because of small original egg size. Posterior isolates pursued more variable schedules of spreading gastrulation, correlated with their more variable performance in pattern formation. Of 30 classified finally as showing pronounced axial elongation, somite segment formation and neural induction, 28 had gastrulated on a spreading schedule beginning 30-40min after control stage 10 (i.e. at control lateral lip stages), but competing the blastoporal ring no earlier than controls. Of 18 classified as closest to radially symmetrical, non-axial differentiation, 15 had shown essentially radially symmetrical or patchy, rapidly fusing zones of onset of marginal cell 'bottling', commencing 60-90 min after control stage 10 and at most 20 min before control blastoporal ring completion. Such a gastrulation pattern is reminiscent of that seen in those whole embryos which, having received u.v. irradiation to the vegetal hemisphere at a precleavage stage, will produce similarly non-axial patterns (Scharf & Gerhart, 1983; Cooke, 1985). On a crude binary classification, therefore, only five examples of morphology would not have been predicted from the time schedules of new cell activity at gastrulation that preceded them.

DISCUSSION The results call into question various earlier assertions and ways of thinking about anuran amphibian development, including some of the first author's. These previous interpretations of data, on apparent pattern regulation, will be dealt with first. We can then begin thinking anew about the primary spatial information system in this development.

Previous interpretations When large sectors surrounding the meridian of sperm entry are removed from mid-blastulae, and new cell contacts finally result in normal-looking but size- reduced late blastulae, these often give rise to small bodies that look qualitatively complete, and quantitatively balanced as to tissue types. (Cooke, 1981, 1982). Such bodies also show a number of small somite segments that is correct for developmental age (Cooke, 1975). This result remains obtainable in our Pattern in Xenopus blastomere pairs 103 laboratory after blastula operations, even when it is most unlikely that presumptive notochord territory has been removed. It is now clear that on simple fate-map grounds (Fig. 2), there is no need to postulate regulative events to explain constancy in the tissue proportion of somite to lateral plate, or in somite number after these ablations. What had been believed to be a 'ventral' ablated sector, normally contributing only to lateral plate mesoderm, in fact contributes to a great majority of somites and substantially so to many of them. It contains, if anything, proportionately more somite than lateral plate presumptive territory, and probably contributes also to pro-nephros. We can also now see that the dorsoanterior sector, left in blastulae after ablations, is due to contribute to the entire somite series, though its normal contribution is a partial one for all but the first few segments which are in any case reduced in cross section. The present work tells us, moreover, that by the earliest time the system can be partitioned in a plane that segregates out a contribution to only part of the somite file - namely the posterior isolate at the 4-cell stage, information leading to the development of appropriate (complete or incomplete) numbers of segments in isolation is often present. This observation, though startling from the standpoint of dynamic theories of development, simplifies the interpretation of the small embryos previously constructed. They represent largely a 'mosaic' development of part-patterns of mesoderm. Only the previously observed 'down-regulation' or reduction in size of the notochord territory, not the normal result in the present work, needs explanation. We have no definitive such explanation but clues may be offered by the following observations. Early abnormal exposure of inner-type or new blastomere membrane to the medium, as in the separation of blastomeres itself, delays measurably the normally precise tempo of development, including the onset of visibly new cell activities at gastrulation. Such delay becomes steeply more exaggerated in relation to increasing amounts of time spent with blastocoelic surface abnormally exposed or cell contacts relaxed, as in the unsuccessful reconstructions of morulae after blastomere separations. Typical blastulae resulting from previously published series of surgical excisions, where healing was often prolonged but subsequent gastrulation not monitored, would be expected on present evidence to have shown significantly retarded onsets of gastrulation. The present isolations, performed at the outset of development, have shown clearly that developmental delay beyond a certain rather strict threshold, occasioned by some physiological trauma in certain cases, is the only circumstance that will cause departure from fate in the patterns of mesoderm developed by reducing the territory achieving notochord status (see results in this and in the following paper). Such an effect is seen along with otherwise normal and healthy differentiation of the tissue pattern that does develop, and tends to cause uninterrupted but slender notochords rather than any deletion of particular regions of the rod-shaped structure. It is thus not strictly linked to pattern but to notochord as a tissue type (see later discussions and comparison with other types of notochord deficiency). We suggest therefore that the small notochords of previous miniature Xenopus 104 J. COOKE AND J. A. WEBBER larvae have been fortuitous effects of slight physiological damage upon the embryo's performance, though a systematic study of the morphogenetic effects of prolonged blastocoel opening is needed. No good evidence remains for specific regulative capacity as to whole body pattern in this embryo, from cleavage onwards. The range of results for frontal 4-cell separations in Xenopus, at least from external appearance, approximates closely those reported by Kageura & Yamana (1983). It is now hard to evaluate the classical claim for a different spectrum of results, and for true regulation in 'dorsal' fragments, in Urodele development. The two forms are possibly more deeply distinct, phylogenetically, than their post- embryonic anatomies would suggest (Nieuwkoop & Sutasurya, 1976). The tra- ditionally investigated species are also at the opposite end of a spectrum from Xenopus as regards egg size and the rapidity of development. There is therefore room for speculation that strategies for the spatial organization of early devel- opment are evolutionarily labile across even smaller phylogenetic gaps than that between amphibians as a whole and amniotes (where specification of the body plan is relatively delayed, in relation to an indefinite pattern of cleavage and early cell mingling). The original urodele work may on the other hand have been flawed, by inadequate numbers of examples and techniques for normal fate mapping and for reconstruction of the internal mesodermal patterns in isolates. The distribution of normal contributions to the mesodermal plan, around the marginal zone in the urodele pregastrula, may be somewhat different from that in anurans but is unlikely to be so divergent as to account for the different reported results on fate map grounds alone. The assertion that the Xenopus body plan lacks proportion-regulating mechanisms, from cleavage stages, obviously demands re-examination of claims to later regulative interactions continuing into gastrula stages (e.g. Cooke, 1982, 1983a,b). A future paper in this series following marked clones in embryos where pattern twinning has been caused by dorsal lip grafting, will offer such a re- examination. It appears at the present time that certain interactions that tend to harmonize proportions can operate within qualitatively 'complete' embryos as a fine-tuning principle, whereas the present work concerns the inability of part- embryos to re-establish missing boundaries to pattern.

The stability of a system specifying body position The present results reveal that in many Xenopus eggs, very shortly after the events of the precleavage interval, a configuration of stable information specifying body position for mesodermal contributions is disposed around the system approximately in register with the presumptive fate map. We can assume that this information is in the vegetal-equatorial part of the egg (see later discussion). It is laid out along a dimension at right angles to the animal-vegetal axis, in which cell interactions will subsequently specify the definitive germ layers that contribute to all the body regions in a co-ordinated way. We first explain why the results tell us Pattern in Xenopus blastomere pairs 105 this, and then begin to ask what the nature of the information for body position might be. The abnormal boundary of the fullest or most axial mesoderms made by posterior isolates, i.e. their point of anterodorsal truncation, corresponds closely with the partition made in the normal fate map by the cleavage plane. Frequently, then, reciprocal isolates across this plane develop so as to suggest that further communication between their descendant endomesodermal cell populations would have been minimal and unnecessary during normal embryogenesis. The dorsoanterior isolate always does something very close to this, so that the impression remains that the dorsoanterior extreme or boundary of pattern is in some sense a stable reference region in the egg. The deviations from mosaic mesodermal part-patterns that occur in many posterior isolate bodies represent losses to varying degrees of their full spread of information, or else failure to acquire it, rather than attempts at regulative acquisition of more complete sets of information. The system's behaviour can be illustrated graphically, as in Fig. 7, by symbolizing the information as levels on a quantitative gradation of state in the material. It formally resembles the behaviour of the ligatured insect blastoderm described by Sander (1975). But the information levels here appear to be very much more stable at early stages in relation to at least nuclear cleavage (the insect blastoderm is multinucleate, but sometimes a syncytium, at time of ligature). Loss of the correct gradation of information across material, when the boundary regions are experimentally isolated from each other, can be slow or absent. The morphology of at least two other types of experimental embryo bears out the contention that all levels in the informational system specifying body position can remain autonomously stable from early times. Precleavage u.v. irradiation attenuates, in a dose-dependent fashion, the intracellular movements that some- how set up the information. The various low grades of this phenomenon result in partial patterns bounded by normal tail formation but truncated anterodorsally at a continuous series of levels in the normal fate map. Similarly, the minor group of isolates, in which the egg has been transected obliquely in relation to the pattern axis, show how detailed the adherence to specific coding for body position can be around the egg substance. Axial but headless bodies, with partial notochords confined to specifically posterior positions, are almost certainly the results of such oblique partitioning. In all these cases, deprivation of an upper sector of the normal complement of information levels occurred during or immediately after the initial setting up of that information (Gerhart et al. 1981; Kirschner et al. 1980; Scharf & Gerhart, 1983). Even posterior cell pairs isolated without any tipping, by sacrificing the reciprocal pair, can develop the most fully mosaic part-patterns. It is unlikely that such blastomeres have undergone a second, artificial activation of pattern formation along the lines of that which can be induced in whole, u.v.-treated eggs by tipping with respect to gravity. The rather striking co-incidence of the second cleavage partition in the normal fate map, with the boundaries of the most complete part-patterns, is further evidence that we are studying information 106 J. COOKE AND J. A. WEBBER

Fig. 7. Some properties of the system specifying body position at cleavage stages. (A) A spatial distribution of a graded property that specifies position value according to the normal fates, in vegetal-marginal material, must exist in many individual eggs by the 4- cell stage. This is because the developments implied by such a system are adhered to quite precisely if the egg is partitioned (dotted line) at that time, and the boundary between the reciprocal part-patterns produced corresponds with the line taken by the cleavage plane through the normal fate map. (B) There may be a variable loss of developmental information evident after isolated development, corresponding to degrees of relaxation of the spatial gradation as shown. This is particularly seen in posterior isolates. Regulation - a tendency to restore complete gradations of infor- mation in one or both isolated sectors - is never observed and thus not shown. (C) Variation in the morphogenetic results of isolate development may be due to variation in intensity of the initial, rapid mechanical events that set up asymmetry and a graded profile of activation or position value. Thus in some eggs at the 4-cell stage the information may be in a restricted distribution not corresponding to the normal pattern contributions, so that a 'posterior' isolate will be permanently deprived, to a varying extent, of its normal spread of information. In normal development of such eggs (and in the later development of the dorsoanterior isolate), slow interactions corresponding to a spread of activation might normalize the position value profile in material by the time this is used to effect gastrulation and differentiation. already disposed across the material at the time of separations. A further study of isolates, from u.v.-treated eggs which will not otherwise form axes, is nevertheless desirable. Such stability on the one hand and lack of regulative dynamics on the other, in an informational system, contrasts with the behaviour to be expected of a positional gradient set up and maintained by interactions obeying diffusion-like laws, i.e. the reaction-diffusion, source-sink or fixed morphogen source models that have usually been linked with the idea of positional information (Wolpert, 1971; Gierer & Meinhardt, 1972; Gierer, 1981). The information used by the present isolates must be set up across essentially non-cellularized material within at most an hour, and two cell cycles, after the significant events of the mid precleavage interval. It would be simplest to assume that such information is purely structural and given at the time of symmetrization movements, though this is not necessarily so (cp. certain ctenophore and coelenterate embryos, Freeman, 1983). But it is often stable enough to direct development of part-patterns, when we know that such patterns will not be fixated as irrevocable cell commitments for another 10 or 15 h (Smith & Slack, 1983; Heaysman, Wylie, Hausen & Smith, 1984). How might we explain the sequence of results in posterior isolates that deviate, by loss of information, from the mosaic part-pattern? There may be partial loss of Pattern in Xenopus blastomere pairs 107 originally existing information, during blastula/gastrula stages (i.e. relaxation of the gradient in state as in Fig. 7B) or there may be a variability in the configuration of information achieved at the stage of isolation, with some eggs having a reliably polarized but still inappropriately restricted distribution of the state we will call 'activation' (see Fig. 7C). Such eggs would achieve the fully distributed activation profile, and thus the normally proportioned development that they undoubtedly give when left intact, by some slower 'back-up' mode of signalling which perhaps more resembles a traditional morphogen gradient system. Certainly, eggs vary within and especially between batches as to the apparent vigour of the structural movements after sperm entry. It would be of interest to carry out frontal isolations on individuals with records of the apparent extent of their early movements of symmetrization. But on the evidence of these experiments, more than 50% of Xenopus eggs would not require very much interaction along the presumptive axis after the 4-cell stage in normal development. This indicates that while one may choose to call the unknown variable specifying this pattern positional information, the means of acquiring it, and its initial molecular nature, can be different from those most recent workers on pattern have been considering. According to those models, large isolated parts from the system should show dichotomous behaviour, tending either to achieve small-scale patterns more complete than their fate-map contribution (i.e regulation) or to lose pattern information entirely. The earlier they are isolated in relation to the time of overt expression or differentiation of pattern, the less mosaic should their behaviour be. Observation of the real time course of gastrulation is limited by a) the variable smoothness of the surface progress of lip formation in different batches, even though stage-10 lip formation is very well timed as stated, and b) the imperfect relation between spread of bottle-cell formation and epibolic pigment gathering at the marginal surface, and the precise stage of true mesoderm involution under- lying it (Keller, 1976). With these provisos however, the observations strongly support the idea that a variable intimately linked with the signal that codes for position within the body is the specific programmed rate of local development. Thus, exactly timed onsets for mechanical contribution to gastrulation are prob- ably set within mesoderm cells induced at particular meridians of the egg, which are to construct particular levels of the pattern (see Fig. 2A). A consideration of what is understood about the early re-organization of plasms in the egg, which must go far to set up the organization revealed in these experiments, is a necessary prelude to thought about its molecular nature (Gerhart et al. 1981; Kirschner et al. 1980; Neff, Wakahara, Jurand & Malacinski, 1984; Elinson, 1984). Such consideration will be postponed for the integrated discussion of the present results and those of separations at the 2-cell stage, in Paper II.

The flow of information that sets up the germ layers How can separations at such early stages result in mosaic behaviour even though territories for endoderm, mesoderm and neurectoderm have not yet been 108 J. COOKE AND J. A. WEBBER established? Experiments, on the origin of the mesoderm have produced partially conflicting results so that the sequence of information flow in normal development is unclear (Nieuwkoop, 1973; Nakamura et al. 1970; Gimlich & Gerhart, 1984). Mesoderm is probably induced as a zone in the marginal region of the animal cap, by information from the vegetal yolky region that will become the endoderm. This information is still available at late blastula stage, but its normal spread may occur much earlier. The graded information for body position, with which this paper is concerned, may be present in those parts of the egg that will become both endoderm and mesoderm. But even if it is originally confined to the presumptive endoderm, there is some evidence that what is induced on each meridian in the blastular marginal zone is not just mesoderm, but mesoderm that is specified for development in co-ordination with its subjacent endoderm as a general body region (Nakamura, Takasaki & Mizohata, 1970; Holtfreter & Hamburger, 1955). If development is viewed as a series of often transitory specified states linked to territories (Slack, 1982, 1983), then the simple germ layer mesoderm, without further specification, may never exist. Mesodermal territories may from their inception share elements of positional coding with the endodermal territories that induce them. Some such assumption makes it easier to understand the present results, and also the observed preservation of spatial contiguity in the normal body plan by the endodermal and the mesodermal derivitives of large (early) clones. The other view, favoured by Nieuwkoop (1973), is that the endodermal inducing influence merely establishes a polarity by positioning the mesodermal organizer or presumptive dorsal lip at one particular meridian. Pattern specification within mesoderm is then assumed to occur, with a prolonged time course, mainly by interactions around the marginal zone. But in this case, 4-cell-stage isolates separated from the organizer-inducing region across the frontal plane could have no basis for observing mosaic development according to presumptive fate for mesoderm. They should be without sufficiently detailed information so that they either lose polarity altogether or regulate to give new pattern. We assume that primary information for the body plan is present relatively early in both endoderm and (probably by patterned induction) in its accompanying mesoderm. We have assayed pattern by inspecting mesoderm simply because, of these two germ layers, it makes the greatly superior display of that pattern. The featureless endoderm, although primary in causal terms, displays its pattern only by gross architecture in the tailbud larva. We have noted that in dorsoanterior isolates particularly but also in axially developing posterior ones, endoderm takes up a largely mosaic 'shape' according to the normal endodermal contributions of the blastomeres in the whole body, rather than disposing itself along the mesoderm according to simple mechanical constraints. The final phase of the informational flow that co-ordinates the germ layers is the induction of the neurectodermal pattern with information from mesoderm. This probably occurs only during gastrulation, as the appropriate spatial relationships between patterned mesoderm and the neurectodermal layer are only then established. Unlike the earlier one from endodermal to mesodermal zone, this Pattern in Xenopus blastomere pairs 109 passage of inductive information involves pronounced geometrical re-organization through morphogenetic movement. This difference is revealed in the development of the present frontal isolates. The neurectodermal patterns, of those posterior isolates whose mes-endoderms have developed mosaically according to fate, are found to have involved massive departure from their normal presumptive contributions by the ectodermal descendants of the original blastomeres. It was recorded earlier that well-developed CNS pattern consisting of spinal cord, hindbrain and ear vesicle characterizes such isolates. Examination of whole larvae whose posterior cell pairs have been filled with FLDX reveals, on the other hand, that these founder cells contribute almost entirely to the epidermal ectoderm and to only very restricted parts of. the normal nervous system, namely a little to forebrain (precisely what the isolate patterns lack) and to posterior or very lateral spinal neural plate. We presume that the alteration of the course of development, in the neurectodermal layer alone, is brought about by the enforced new spatial pattern of contact between the mesodermal part in isolates, which embodies considerable axial inductive information according to fate, and their ectodermal part which would normally have been largely excluded from such contact with inducing mesoderm and so would have become epidermis. This observation extends the recent experimental confirmations of Spemann's original view of the inductive origin of the CNS, that have been possible using lineage tracing techniques (Gimlich & Cooke, 1983; Jacobson, 1984).

Non-equivalence. The nature of the body pattern A final feature of the morphological results to which attention must be drawn is the way in which they support the notions of non-equivalence (Lewis & Wolpert, 1976) and of a 'second anatomy' underlying development (Slack, 1982, 1983). A straightforward consideration of cellular differentiation as studied by the cytologist, and of the mechanics of cell adhesion and locomotion, might suggest that early development is a matter of arranging that a simple pattern of territories of certain shapes and sizes be formed within a cell sheet, with the material in each territory embarking on one of the courses of differentiation that characterize the later body. If each such course of differentiation were to be accompanied by a particular set of earlier mechanical properties in the constituent cells, then the observed normal correlation of morphogenetic movements and the pattern of differentiation in the whole embryo might be expected to follow as a sort of mechanical resultant of the local forces. The two above-mentioned ideas contrast with this in proposing that a series of early decisions establishes a spatial pattern of regions, but that the molecular coding system whereby the regions are marked out is prior to, and independent of decisions underlying the development of the recognized cellular differentiations or even anatomical structures. Each region is indeed set to contribute to the final pattern in a defined way, but defined only by position within the plan of the body. This may include contribution to all or any of the differentiations and structures listed by the anatomist. The corollary is that each gross element of the body, as defined by histodifferentiation, is composed of 110 J. COOKE AND J. A. WEBBER contributions from diverse territories which were primarily specified as to the body region each was to be responsible for. In the current Xenopus experiments this shows up in the behaviour of somite and notochord. Their material (or the material that induces them) appears to be regionalized around the meridians of the egg at early cleavage stages, so as to set up quantitatively specific contributions to the different geographical parts of the final segment series and of the simple elongate notochord. The quantitative aspects of posterior part-patterns of somite segments in relation to the complete pattern, shown in Fig. 6, should be compared with the fate map as revealed in Figs 2 and 3. This shows how the preservation of partial information, leading to formation of particular posterior subsets of segments in each isolate, tends to be linked with patterns of quantitative contribution that follow the normal fate of material with that information in the complete development. Similarly the patterns formed by the minority of oblique isolates, transitional between the main results of this and of the following paper (Cooke & Webber, 1985), show the non- equivalence of regions even within the relatively restricted early notochord territory. In these isolates and also in minimally affected u.v.-irradiated embryos described in Paper III (Cooke, 1985), the fate map has been partitioned or reduced so as to include only a specific minor fragment (usually relatively lateral in the egg) of that territory. Partial notochords in restricted, posterior axial positions are produced, rather than the abnormally slender rods occupying the whole available space in the axis which a purely mechanical model of morphogenesis would predict. Such embryos, made by the earliest manipulations, are striking evidence of mosaicism and stability for something which we must call body-position value. The complete notochord anlage presumably contains an almost complete set of contributions from the belt-like equatorial zones, in the early embryo's meso- dermal map (see Fig. 2A), which become coded for successive anteroposterior axial positions.

This paper is dedicated to the memory of June Colville, skilled histologist for much of the authors' work, who died on 30th August 1984.

REFERENCES BRICE, M. C. (1958). A re-analysis of the consequences of frontal and saggital constrictions of newt blastulae and gastrulae. Archs BioL, Liege 69, 371-428. COOKE, J. (1972a). Properties of the primary organisation field in the embryo of Xenopus laevis. II. Positional information for axial organisation in embryos with two head organisers. J. Embryol. exp. Morph. 28, 27-46. COOKE, J. (19726). Properties of the primary organisation field in the embryo of Xenopus laevis. I. Autonomy of cell behaviour at the site of initial organiser formation. /. Embryol. exp. Morph. 28, 13-26. COOKE, J. (1975). Control of somite number during development of a vertebrate, Xenopus laevis. Nature 254, 196-199. COOKE, J. (1979). Cell number in relation to primary pattern formation in the embryo of Xenopus laevis. I. The cell cycle during new pattern formation in response to implanted organisers. J. Embryol. exp. Morph. 51, 165-182. Pattern in Xenopus blastomere pairs 111 COOKE, J. (1981). Scale of body pattern adjusts to available cell number in amphibian embryos. Nature 290, 775-778. COOKE, J. (1982). The relation between scale and the completeness of pattern in vertebrate embryogenesis: models and experiments. Amer. Zool. 22, 91-104. COOKE, J. (1983a). On the origin of the body pattern in vertebrate embryogenesis. Rivista di Biologia 76, 603-638. COOKE, J. (1983fc). Evidence for specific feedback signals underlying pattern control during vertebrate embryogenesis. /. Embryol. exp. Morph. 76, 95-114. COOKE, J. (1985). Dynamics of the control of body pattern in the development of Xenopus laevis. III. Timing and pattern after u.v.-irradiation of the egg and after excision of presumptive head endo-mesoderm. /. Embryol. exp. Morph. 88, 135-150. COOKE, J. & WEBBER, J. A. (1985). Dynamics of the control of body pattern in the development of Xenopus laevis. II. Timing and pattern in the development of presumptive lateral halves isolated at the 2-cell stage. /. Embryol. exp. Morph. 88, 113-133. DOLLANDER, A. (1950). Etude des ph6nomenes de regulation consecutifs a la separation des premiers blastomeres de l'oeuf de Triton. Archs Biol. 81, 1-110. ELINSON, R. P. (1984). Cytoplasmic phases in the first cell cycle of the activated frog egg. Devi Biol. 100, 440-451. FANKHAUSER, G. (1948). The organization of the amphibian egg during fertilisation and cleavage. Ann. NY. Acad. Sci. 49, 684-712. FORMAN, D. & SLACK, J. M. W. (1980). Determination and cellular commitment in the embryonic amphibian mesoderm. Nature 286, 482-484. FREEMAN, G. (1983). Experimental studies on embryogenesis in Hydrozoans (Trachylina and Siphonophora) with direct development. Biol. Bull. mar. Lab., Woods Hole 165, 591-618. GERHART, J. (1980). Mechanisms regulating pattern formation in the amphibian egg and early embryo. In Biological Regulation and Development, (ed. Goldberger), pp. 133-316. New York: Plenum Press. GERHART, J., UBBELS, G., BLACK, S., HARA, K. & KIRSCHNER, M. (1981). A re-investigation of the role of the grey crescent in axis formation in Xenopus laevis. Nature 292, 511-516. GIERER, A. & MEINHARDT, H. (1972). A theory of biological pattern formation. Kybernetik 12, 30-39. GIERER, A. (1981). Generation of biological patterns and form: some physical mathematical and logical aspects. Prog. Biophys. molec. Biol. 37, 1-47. GIMLICH, R. L. & COOKE, J. (1983). Cell lineage and the induction of second nervous systems in amphibian development. Nature 306, 471-473. GIMLICH, R. L. & GERHART, J. (1984). Early cellular interactions promote embryonic axis formation in Xenopus laevis. Devi Biol. 104, 117-130. HAMILTON, L. (1969). The formation of somites in Xenopus laevis. J. Embryol. exp. Morph. 22, 253-260. HEAYSMAN, J., WYLIE, C. C, HAUSEN, P. & SMITH, J. C. (1984). Fates and states of determination of single vegetal pole blastomeres of Xenopus laevis. Cell 37,185-194. HOLTFRETER, J. & HAMBURGER, V. (1955). Embryogenesis and progressive determination in amphibians. In Analysis of Development, (ed. Willier, Weiss & Hamburger), pp. 230-297. Philadelphia: Saunders. JACOBSON, M. (1984). Cell lineage analysis of neural induction; origin of cells forming the induced nervous system. Devi Biol. 102,122-129. KAGEURA, H. & YAMANA, K. (1983). Pattern regulation in isolated halves and blastomeres of early Xenopus laevis. J. Embryol. exp. Morph. 74, 221-234. KELLER, R. E. (1976). Dye mapping of the gastrula and neurula of Xenopus laevis. II. Prospective areas and morphogenetic movements of the deep layers. Devi Biol. 51, 118-137. KIRSCHNER, M., GERHART, J., HARA, K. & UBBELS, G. A. (1980). Initiation of the cell cycle and establishment of bilateral symmetry in Xenopus eggs. In The Cell Surface pp. 187-215. New York and London: Academic Press Inc. LEWIS, J. & WOLPERT, L. (1976). The principle of non-equivalence in development. J. theor. Biol. 62, 478-490. NAKAMURA, O. & KISHIYAMA, K. (1971). Prospective fates of blastomeres at the 32-cell stage of Xenopus laevis embryos. Proc. Japan. Acad. 47, 407-412. 112 J. COOKE AND J. A. WEBBER NAKAMURA, O., TAKASAKI, H. & MIZOHATA, T. (1970). Acquisition of selfdifferentiation capacity of the dorsal marginal zone. Proc. Japan. Acad. 46. NEFF, A. W., WAKAHARA, M., JURAND, A. & MALACINSKI, G. M. (1984). Experimental analysis of cytoplasmic rearrangements which follow fertilization and accompany symmetrization of inverted Xenopus eggs. /. Embryol. exp. Morph. 80, 197-224. NIEUWKOOP, P. D. (1973). The 'organisation center' of the amphibian embryo: Its origin, spatial organisation and morphogenetic action. Advances in Morphogenesis 10, 2-39. NIEUWKOOP, P. D. & FABER, J. (1967). Normal Table of Xenopus Laevis (Daudin). 2nd. edn. Amsterdam: Elsevier/North Holland. NIEUWKOOP, P. D. & FLORSCHUTZ, P. A. (1950). Quelques characteres sp€ciaux de la gastrulation et de la neurulation de l'oeuf de Xenopus laevis (Daudin) et the quelques autres anoures. I. Etude descriptive. Archs Biol. Liege 61, 113-150. NIEUWKOOP, P. D. & SUTASURYA, L. A. (1976). Embryological evidence for a possible polyphyletic origin of the recent amphibians. /. Embryol. exp. Morph. 35, 159-167. SANDER, K. (1975). Pattern specification in the insect embryo. In Cell Patterning. CIBA Symposium 29 (new series) pp. 241-263. SCHARF, S. R. & GERHART, J. (1980). Determination of the dorsal-ventral axis in eggs of Xenopus laevis: Complete rescue of UV-impaired eggs by oblique orientation before first cleavage. Devi Biol. 79, 181-198. SCHARF, S. R. & GERHART, J. (1983). Axis determination in eggs of Xenopus laevis: A critical period before first cleavage, identified by the common effects of cold, pressure and UV- irradiation. Devi Biol. 99, 75-87. SLACK, J. M. W. (1982). Regeneration and the second anatomy of animals. In Developmental Order: Its Origin and Regulation, (ed. S. Subtelny), pp. 423-436. New York: Alan R. Liss. SLACK, J. M. W. (1983). From Egg to Embryo: Determinative Events in Early Development. (BSDB 13), Cambridge UK: Cambridge University Press. SMITH, J. C. & SLACK, J. M. W. (1983). Dorsalisation and neural induction; properties of the organizer in Xenopus laevis. J. Embryol. exp. Morph. 78, 299-317. SPEMANN, H. (1902). Entwicklungsphysiologische Studien am Triton-Ei, II Wilhelm Roux Arch. EntwMech. Org. 15, 448-557. SPEMANN, H. (1938). Embryonic Development and Induction. Yale University Press. Reprinted Hafner, New York, 1967. SUDARWATI, SRI & NIEUWKOOP, P. D.(1971). Mesoderm formation in the anuran Xenopus laevis (Daudin). Wilhelm Roux Arch. EntwMech. Org. 166, 189-204. SULSTON, J. E., SCHIRENBERG, E., WHITE, J. G. & THOMSON, J. N. (1983). The embryonic cell lineage of Caenorhabditis elegans. Devi Biol. 100, 64-119. WOLPERT, L. (1971). Positional information and pattern formation. In Current Topics in Devi Biol. 6, 183-223.

(Accepted 20 February 1985)