The positional coding system in the early eye rudiment of Xenopus laevis, and its modification after grafting operations

During vertebrate development, fibres from the retinal ganglion cells form a set of connections across the midbrain optic tectum which give a simple, characteristically orientated representation of visual space. The initial formation of this retinotectal projection, or ‘map’, which is largely if not completely independent of visual experience, has been the subject of much investigation in amphibians (Gaze, Jacobson & Székely, 1963; Gaze & Keating, 1972; Chung & Cooke, 1975; Scholes, 1979) and theoretical modelling (Prestige & Willshaw, 1975; Hope, Hammond & Gaze, 1976; Willshaw & von der Malsburg, 1979). A widely held view at the present time is that for the development of such a set of connections only the retinal cell population requires a set of positional labels of any degree of precision. Construction of the projection may then be of a fibreinteractive or fibre-instructive nature, not utilizing precise prearranged tectal target markers but making use instead of differential recognition cues within the substance of the brain, along central pathways of the optic tract and/or at the tectum.

Recent studies have indicated that the amphibian eye rudiment as a whole is autonomously programmed, both for construction of the orientated set of connections and for development of the morphological pattern around the eye’s circumference (optic fissure, pigment gradients etc.), at the time of its évagination from the neural tube (Sharma & Hollyfield, 1974, 1980; Gaze, Feldman, Cooke & Chung, 1979). No modification of this programme by surrounding non-neural tissues seems to occur after such stages, since cutting the optic stalk and rotating the rudiment (a vesicle of a few hundred neuroepithelial cells) results in morphogenesis of an eye appropriately rotated within the head, and mediating at the tectum a correspondingly rotated map of visual space. Thus in terms of cellular descent, ganglion cells connecting with each tectal location after this operation are those that would normally have connected there, despite their abnormal positioning within the head and in relation to visual space. Comparable autonomy of behaviour is also found in various surgically reconstructed eyes, comprising abnormal combinations of half-rudiments made at relatively more advanced eyecup stages (Straznicky & Gaze, 1980; but see Discussion). Such observations show that development of the projection involves more than local axonal interactions and a reliance on the normal geometry and timing of axon outgrowth etc. To some degree, fibres from retina originating in different parts of the eyecup must bear distinctive markers which must interact with the guidance cues encountered within the central pathways. More direct evidence that this is so is now available (Fawcett & Gaze, 1982 and in preparation).

How is the required pattern of positional values (Wolpert, 1971) or codes in ocular tissue established? In how detailed a manner is tissue position in the eye marked out by such a coding system at early stages, and how is an appropriate array of codes preserved, and perhaps expanded, as the population of ganglion cells is expanded at the ciliary margin during the extended period over which the projection develops? In an approach to these questions we have constructed in the clawed frog Xenopus laevis a particular class of compound eyes; that is, eyes developing from surgical construction of an abnormal array of sectors from the rudiment. The particular recombination employed involves replacement of a small part (less than a quarter) of the normal tissue complement with a sector that duplicates another small part situated at the opposite point in the circumference. The major component, part of the host eye, remains normally orientated, while the grafted duplicate component, having come from a donor eye on the same side of the head, is rotated through some 180° as well as translocated.

The animals were allowed to develop until just after metamorphosis, and the behaviour of the retinal tissue derived from the graft was then analysed by electrophysiological mapping of the retinotectal fibre projection (i.e. visuotectal mapping). Thus our evidence for the existence of retinal positional differences comes from the way that populations of retinal ganglion cell arbors are distributed across the tectum. A recent interactive theory of the mapconstruction process would claim that the array of positional labels on a minor population of fibres cannot entirely be deduced from its mapping behaviour in conjunction with a major population on one tectum (Fawcett & Willshaw, 1982). We shall touch on this reservation in discussing certain details of the results. We are confident, however, that the recorded projections from the operated eyes reveal the widespread presence of distinctive positional codes, around the circumference of the rudiment, at stages preceding the formation of neurones, and that these codes are autonomously maintained within tissue after surgical translocation. Thus, if we characterize each graft by the position of the centre of its circumferential site of origin in the donor rudiment, we find that grafts, despite their new and abnormal positions in host eyes, project to the general tectal region appropriate for their origin. There is, however, a marked tendency for the graft-derived sector as a whole to project onto the tectum with a ‘handedness’ the reverse of that to be expected from its origin on the same side of the head as the host eye; that is, it tends to make a minor projection related by mirror symmetry rather than by point symmetry to the main projection. We explain this more fully in describing and discussing the results, and we believe that these results may help us to relate the behaviour of the eye rudiment to that of other systems of biological pattern formation that have been studied in detail. The experiments reported here are complementary to those of Willshaw, Fawcett & Gaze (1983) in which sectors were grafted between right and left eyes, so that the graft sector could be translocated in ocular position without rotation. However, the presence of a detailed marking system within ocular tissue at vesicle or preneural stages has not been described before, to our knowledge.

Maintenance of Xenopus, including production of fertile eggs and the rearing of larvae between operations and metamorphic stages, was as described elsewhere (Gaze et al. 1919).

Sectors of less than 90° were extirpated, and replaced in a host at as nearly as possible the 180°-opposed orientation and position as shown in Fig. 1. Sectors were moved with tungsten needles and hair loops, and accompanied by overlying head ectoderm to aid healing, into a site made by extirpating a sector from the host. Embryos were anaesthetized with 1:10000 MS222 (Sandoz), to avoid the movements that can occur from stage 25 onwards. They were situated in shallow depressions in a wax bed, and light coverglass bridge was used for some 20 min to retain grafts. Operating solution was 66 % Niu-Twitty saline, but with pH reduced to 7·3–7·4 with 0·1 N-HCI. One hour after operation, when epidermal continuity appeared normal, embryos were transferred for 48 h to 20% Niu-Twitty before commencement of rearing to mapping stages.

Metamorphic tadpoles and juveniles were maintained under 1:15 000 MS 222 (Sandoz) anaesthesia in oxygenated 100 % Niu-Twitty saline, in a perspex sphere permitting access to nearly all the dorsal and ventral visual field. The recording microelectrode (metal-filled micropipette, tip diameter 15 μm approximately, impedance 100 Ω at 1000 KΩ) was moved serially to the sets of tectal positions shown in each diagram, to explore the area invaded by optic terminals in each animal. Responses were evoked by movement of a 10° black disc against the arc of an Aimark projection perimeter with the eye centred at the fixation point. Recording was from pre-synaptic optic fibre terminals, and was characteristically multi-unit with receptor fields of 15–30° of arc (see Figs 2, 3, 4, 6 and 7).

Heads of animals after recording were fixed in Heidenhain’s Susa, sectioned coronally and stained by Holmes’ silver technique for examination of the anatomy of the eye and optic tract.

The type of compound rudiment made in this study had a sector subtending somewhat less than 90° in face view, transplanted from a donor eye vesicle into an excised site at the diametrically opposed position in the vesicle on the same side of the host embryo (see Fig. 1A). Most operations were performed at embryonic stages between 24 and 28 (Nieuwkoop & Faber, 1956), but five maps were made from compounds constructed at stage 32/3. Normal relationships were preserved between donor and host rudiments with respect to presumptive retinal centre versus periphery, and with respect to inner, flattened epithelial and outer, pseudostratified neuroepithelial components of the vesicle. The sector was thus necessarily rotated some 180° as well as being repositioned in the eye. Transfers were attempted between all possible pairs of circumferential locations centred 180° apart.

Fig. 1B shows an exploded diagrammatic view of eye vesicle anatomy at the average stage (stage 26) of operations, while Fig. 1C shows appearance at the time of mapping of chimaeric compound eyes, where a wild-type donor sector existed in a periodic albino host (Hoperskaia & Golubeva, 1980). Such host embryos have been used to reveal the wedge-shaped territory in the retinal pigment epithelium (RPE) derived from the graft. The wild-type RPE wedges illustrated represent extremes of a range found, the angles subtended by graft-derived territories becoming much more variable than those subtended by original implants 24 h after operation. No marker system was used for the neural retinal component (NR) normally derived from the pseudostratified lateral wall of the eyecup. The expected outcome of successful operations, however, would be graft-derived sectors of RPE and NR in close spatial relation to one another, extending essentially from the fundus of the eye to the ciliary margin as wedges. This would follow from a coherent cell group having colonized and maintained a sector of the eyecup margin where new retinal tissue of both types is produced in a growth zone (Conway, Feiock & Hunt, 1980).

The genetic marker was not used throughout the work, but observation of very young albino larvae in which a wild-type sector had been implanted suggests an overall success rate of some 60%, in this series, in finally populating a wedge of growing retina from a graft made at eye vesicle stages. No compound projections were ever observed when genetic evidence showed that ectopic tissue had not survived embryonic life in this way, and conversely, a normal entire projection of the visual field in the genetically proven presence of a graft was an unusual result (see Table 1). A marked wedge in the RPE layer of the eye does not define unequivocally the position of graft-derived NR as being beneath it, because évagination of some ocular tissue destined to be ventronasal occurs after mid-twenties stages and may be followed by sliding movements between RPE and NR layers before these fuse to give a retinal structure (Holt, 1980). Locations around the Xenopus eye are also thought to be programmed for specific rates of enlargement during larval life (Beach & Jacobson, 1979), and this could influence the success of grafts from different origins in establishing sectors of ciliary margin and thus founding retinal wedges. The foregoing factors may explain our failure to obtain evidence for many successful dorsal or dorsotemporal to ventral or ventronasal tissue transfers in the sample of animals mapped physiologically, despite having made representative numbers of the appropriate grafts.

Most sectors of genetically marked RPE have simple coherent outlines even at the cellular level in histological sections (Fig. 1D), but a certain incidence of ‘outlier’ patches or splitting and rejoining of pigmented territories near the ciliary margin was observed, as by others (Conway et al. 1980). A great majority of these compound eyes showed normal continuity of retinal structure between graft- and host-derived regions over most sections. In general, retinal fibres gained access to brain and tectum via a grossly normal nerve and tract, in contrast with the abnormal pathways in many cases where whole early vesicles were rotated, as previously described (Gaze et al. 1979).

Visuotectal projections through the operated eyes were recorded in 67 animals at stages ranging from just-before to just-after metamorphosis. To provide a comparison for the abnormal maps to be shown, Fig. 2 illustrates a normal visuotectal map. This projection, although normal and complete, came from an eye carrying a graft and represents a minor class of results discussed later. Figs 3, 4, 6 and 7 give samples of the major classes of reduplicated projection obtained.

Of the animals recorded, 33 were not considered further, since they did not permit useful conclusions to be reached, for various reasons. In 18 of these animals the graft and host were both wild type and the map was normal; no conclusions about the continued existence or behaviour of the graft could Rows of numbers and filled circles on the tectum represent electrode positions. Similarly numbered rows in the chart of the visual field indicate corresponding field positions. The chart extends for 100° outwards from the fixation point in the centre.

The map, though normally organized, was obtained as a relatively unusual result from an experimental animal (see Table 1). therefore be made (see section on eye development). We realize that to ignore these results may lead to an overestimate of the degree of autonomy in developing eye tissue since we cannot rule out the possibility that the tissue positional codes in grafts may have accommodated to the surroundings in these cases. The results from genetically marked grafts suggest that this outcome is less common than autonomy, however. In two other animals there was a gross deformity of the eye and in a further 13 animals the recordings were inadequate for technical reasons or poor connectivity.

Genetically marked ectopic grafts existed in 18 of the 34 remaining animals that were analysed in this study. Where the grafts were albino tissue in a wildtype host (two cases) the final tissue distribution could only be assessed approximately, due to the imperfect cell autonomy of the pigmentation (Hoperskaia & Golubeva, 1980).

Table 1 displays the results of these experiments, which fall into the categories described below.

In this, the largest single class of maps (16/34), a subset of tectal positions received input from two widely separated visual field positions each, so that the final projection consisted of an essentially normal representation, spread over the tectum, of the entire visual field with the exclusion of a variable-sized sector, together with a cluster of field positions from within that missing sector which mapped as a duplicate projection on one region of tectum (Figs 3 and 4). This tectal region was usually that appropriate for the diametrically opposite part of the visual field, which indeed also mapped there, but no obvious organization within the minor projection could be discerned. In genetically marked cases the duplicate map was usually centred on the visual field watched by a wedge of retina underlying graft-derived RPE, and such maps (criterion: three or more abnormal duplicate positions) have been seen in conjunction with wedges subtending down to 30° of arc. In eight animals the grafts had been wild type into wild type and the persistence and functioning of the grafts was deduced from the existence of duplicate maps as described above. Most of these compound projections came from 20’s stage operations, but one was from an animal operated on at the eyecup stage 32+.

The grafting operation both translocates and rotates the graft. Thus in assessing the organization within the duplicate projections of these compound eyes, we consider whether duplicate pairs of field positions activating successive tectal sites tend to march with the same ‘handedness’ around the field, so as to preserve the diametrical opposition between the members of each pair, or with opposite ‘handedness’, so that the two members of each pair make parallel progress in the visual field (Fig. 5). The former mode we call point-symmetrical compound maps, of which only one probable example was found. It is the type of map that would be expected from the geometry of the operation performed, given total cell autonomy. The latter mode comprises mirror-symmetrical maps (Figs 6 and 7) of which ten were seen. Mirror-symmetrical duplicate projections were found especially when the graft-derived wedge had increased its share of the eye compared with expectation, but this could reflect the fact that the organization within very small duplicate projections can scarcely be assayed by our methods. Three of the mirror-symmetrical reduplications came from genetically-marked grafts and the other seven came from wild-type grafts, as did the only point-symmetrical duplicate found.

Four projections were found in which an apparently normal map contained visual field positions which must have been seen by retina underlying genetically marked RPE (Fig. 2).

Fig. 8 gives the estimated central positions of origin of all rudiment sectors leading to the development of accessory projections of the three types described above, as well as those genetically marked cases giving normal maps.

In one case the duplicate projection corresponded only partially, and in one case not at all, with the field watched by retina overlain by genetically marked pigment epithelium. In one further case a narrow sector of marked tissue appeared to split the host retina partially into separate domains, without itself connecting to the tectum.

The high frequency of minor duplicate projections reveals that grafts of the order of a hundred cells, from most locations in the eye vesicle, can preserve and utilize distinctive position values during development at a site opposite their original one in the rudiment. We must assume that sets of positional labels developed on optic axons of retina founded by the grafts duplicate those being developed on retina of homologous origin in its normal position in the mature eye. Many of the smallest autonomously mapping wedges derive from grafts made at the earliest stages used. These results therefore advance the time of stable diversification of positional markers within amphibian eye tissue into the vesicle, preneural, stages, whereas previously published work has dealt with recombinations in stage-30+ eyecup tissue (Willshaw, Fawcett & Gaze, 1983; Conway et al. 1980).

A few large, marked RPE wedges are associated with normal, single maps of the whole retina onto the tectum. From our understanding of retinal growth at the ciliary margin we would expect that the NR in such animals also contains a wedge of graft-derived tissue, so that these cases presumably reflect the possibility that a graft can accommodate to the position value of its surroundings (in classical experimental embryological terms, the phenomenon of ‘assimilation’).

The outcome of positional interactions involving small numbers of cells could be capricious, because dependent upon variables such as rate of healing, or cell viability in the hours after operation. In well-studied pattern-forming systems such as the chick limb or the amphibian gastrula, healed grafts between particular positions that normally result in new or abnormal pattern formation, fail to do so in a significant proportion of cases. A positive record of position values retained in tissue from any one location is therefore compelling evidence for their normal presence there (see Fig. 8). The occasional duplicate projection centred on visual field wholly separate from that watched by an RPE wedge is presumably the result of the early morphogenetic independence of inner and outer eyecup layers, alluded to earlier.

There is wide distribution, across the rudiment, of the capacity whereby graft-derived axons ultimately project to a tectal sector corresponding to their original destinations. This suggests the existence of a considerably diverse or fine-grained set of codes on pre-eyecup-stage neuroblasts, marking position round the rudiment’s circumference, and thus the relative placement on the tectum of the terminals of the descendant ganglion cells. There is no good evidence for any special ‘origins’ or reference points for graded cues within the eye that are in line with the conventional anatomical axes of the body as a whole. Thus reference to Fig. 8 shows that sectors from most parts of the circumference maintain distinctive values, and a survey of the areas of tectum covered by the accessory projections resulting from these, reveals that they form a continuous, overlapping series of tectal sectors, rather than examples of defined anterior, posterior, medial or lateral tectal occupancy. The set of codes would appear to be continuous and at least moderately numerous around the eye rudiment, and could be thought of as a cycle of states distributed around the circumference. Compound eyes of this type, assaying the autonomous mapping behaviour of small retinal portions ectopically situated in a normal surround, are now in use in conjunction with fibre-tracing techniques, to investigate the actual mechanism of projection assembly by axons (Fawcett & Gaze, 1982 and in preparation).

Thus far, we have described all grafted sectors giving rise to duplicate projections as ‘autonomous’, whatever the organization or apparent lack of it within those projections. The original operation translocates and rotates by 180 ° an eye sector of a given handedness with respect to its surroundings (i.e. within the eye of one side of the head). True autonomy of a sector (that is, construction of a spatial array of ganglion cell positional labels that is unchanged in cell-lineage terms from the contribution it would normally have made) would therefore be represented only by a point-symmetrical compound map. The results in fact obtained demonstrate a strong tendency towards mirror-symmetrical relationships between the projections from host- and graft-centred retinal sectors, and thus a departure from true autonomy.

This is illustrated by comparing the maps of Figs 6 and 7 and the right-hand diagram of Fig. 5, with the original scheme of operation in Fig. 1A and the lefthand diagram of Fig. 5. The former maps and diagram represent mirror symmetry, in that for a linear series of recording positions on tectum that is occupied by both projections, the paired series of visual field positions tend to march in parallel (i.e. to progress in opposite directions in terms of circumference), although situated in opposing general regions of the field. In a good point-symmetrical compound projection, such paired series of visual field positions ‘chase’ one another, progressing in the same direction around the circumference though situated in opposite regions. Conventional maps like those of Figs 2, 3, 4, 6 and 7 result from the rectilinear nature of micromanipulator advancers, and our axial way of lining animals up in the mapping apparatus. They obscure the fact that retina and tectum are more nearly disc shaped, and that position within each is more naturally defined by angular coordinates.

Mirror-symmetrical compound projections involve a change of the ‘handedness’ with which progression round the graft-derived retina is reflected in positions of tectal terminals, even though a memory of the overall graft origin has been retained. Eleven compound projections were recorded in which the minor component showed a fair degree of internal order and of these, ten appeared mirror symmetrical and only one might be interpretable as an example of point symmetry. Most mirror-symmetrical compound maps were from rudiments made in mid-twenties-staged embryos, but two came from operations performed on eyecups at stage 32+.

Stable point-symmetrical compound projections can in fact result from certain other operations, which also recombine eyecup sectors of the same original ‘handedness’ (i.e. from one side of the head - Straznicky & Gaze, 1980), so that there are no dynamics in the process of assembling projections that lead inherently towards the development of mirror symmetry in eyes of compound structure. The outcome that we observe in the present experiments may result from the confrontation of a minor fragment of the circumference with a major one, in contrast with the previous work that had confronted large fragments, approximately equal in size and in their diversity of positional codes.

Conway et al. (1980) report a departure from complete autonomy in minor translocated eyecup sectors, similar to that found here. Their results are not, however, presented so as to allow an appraisal of the diversity of positions within the eyecup which can support development of projections on appropriate tectal sectors after grafting. The work is also discussed in terms of a ‘Cartesian’ scheme involving combinatorial coding of tissues according to values along two axial dimensions, presumably organized from special boundary regions in the system. Our own results (see Fig. 8), as already stated, offer no evidence for any such organization, nor for any specially anteroposterior or mediolateral organization of tectal cues used in assembly of projections.

In attempting to understand the development of the patterning system in the early eye, it would seem best to treat this as some more or less fine-grained set of codes distributed around the circumferential dimension, perhaps with a centre-to-periphery coordinate of coding laid out according to cell birthdate in the sequence of peripheral growth, and not addressed by the present work. Properties of interaction among the circumferential codes must be such that when a duplicate edition of any subset of them is made to replace an opposite subset in the cyclic series (see the explanatory Fig. 9; and also Fig. 1A), then by the time retina has started to connect with the tectum the duplicate positional codes are often being developed in mirror-symmetrical sequence.

This behaviour may be understandable if we consider the eye in relation to other work on pattern regulation, especially in structures which are also growing. A recent general description of interactions in a variety of such systems still seems valid (French, Bryant & Bryant, 1976) despite modification or redundancy that has been proposed for its detailed rules (Bryant, French & Bryant, 1981; Lewis, 1981). It is assumed that at an early stage cells are equipped with codes, of unknown molecular nature, in relation to their geometrical positions within a structure that is to form pattern. During subsequent growth they display their codes to their neighbours, allowing some form of interaction that causes daughter cells to adopt coding values that lie in some sense ‘between’ those of their neighbours and precursors in an expanding set of positional values. Thus the increasing multiplicity of positional codes required to direct the final cell differentiation in an expanding tissue is automatically provided in an appropriate arrangement. A special intercalation process can heal coding discontinuities caused by natural or surgical traumas which bring together cells which have derived their codes from different, normally non-adjacent positions. In the case of the eyecup and subsequent retina, any increased multiplicity of codings required for fibres from later-maturing and more populous rings of retina would be accomplished in normal development by intercalation or expansion of the cyclic set of values during growth at the margin.

Because of the cyclic nature of such a set of markers the healing of the discontinuities, originally caused by operations such as those described here, would necessarily pose ambiguities in the ‘routes’ taken by the process of intercalation to replace values at the zones of junction in the growing tissue. Such intercalation could have two outcomes in terms of the new circumferential values supplied, and work on other systems has revealed that the ‘shorter’ route across a discontinuity is characteristically taken (French, Bryant & Bryant, 1976). More recently, French (1978,1980) has suggested that experimental confrontation of certain exactly opposite members of a series of codes results in ‘paralysis’ whereby no intercalation occurs (there being no ‘shorter’ route), but simply autonomous development of part patterns by graft and host components. Fig. 9 A and B depict two possible outcomes of the operation where a minor eyecup fragment finds itself 180° rotated and repositioned, to replace the diametrically opposite tissue in a rudiment of the same ‘handedness’.

Circumferential position is marked out in a series of (presumably) cytochemically encoded values, 1–12/0, in the early rudiment, which may be diversified and made more fine grained only by intercalary insertion of intermediate codings, and not by extension, during subsequent growth of tissue. In case (A) cell survival is nearly complete at the cut surfaces, so that no positional values are lost. The values confronted may then be sufficiently near maximally distant in the set that no intercalation follows. In terms of retinotectal map development, the result would be autonomous point-symmetrical mapping of the two components onto one tectum. Our results offer only one doubtful example of this, but it can be the normal result when point-symmetrical combinations of large equal fragments are made, in operating salines which we believe to foster survival of the embryonic cells (Straznicky & Gaze, 1980). In case (B), considerable cell death at the margin of the fragments before healing results in there being a pronounced ‘shorter route’ for intercalation on either side of a ‘nucleus’ of misplaced tissue which itself may represent only a small sector. The result, after intercalation and growth, would be the presence of a large retinal domain mirror symmetrical with the host domain in terms of positional codes, and perhaps derived from host- and graft-derived tissue. Those of our supplementary projections which are large enough to offer evidence of their internal organization suggest that they are the result of a retinal domain that has changed its array of codes in this way. We propose that under the more physiologically challenging conditions involved in the transfer of small numbers of cells at early stages, cell death may be extensive even in the high strength salines, and it is this which leads to a situation provoking the intercalation of a mirror-imaged domain.

In the present experiments the orientation of the projection from the graft-centred wedge of retina was different from what would have been expected on the basis of completely autonomous development, and the difference is in accord with the prediction on the basis of ‘intercalation via the shortest route’, as described. In a recent paper (Willshaw, Fawcett & Gaze, 1983), pie-slice grafts between somewhat later eyecups, from opposite sides of the head, gave rise to compound projections also showing mirror-symmetrical organization. For the latter operation such mirror symmetry corresponds with expectation either in the case of autonomous development without significant cell death or after intercalation within the recombinant eye rudiment. This is because the recombined right and left eyecup sectors have, in any case, arrays of positional codes with opposite handedness. There is thus no incompatibility between the present results and those reported in the above-mentioned paper.

In the literature so far, the idea of intercalation has co-existed principally with the idea that new values are interposed in the act of cell division itself, in a way that is therefore tightly linked to growth in tissue (i.e. to epimorphosis - French et al. 1976). This is not a necessary part of the idea, however, and our results can offer no evidence as to whether, after our operations, the handedness of arrays of labels is altered only in the newly growing zone of junction, or across considerable tissue distance by signals acting like diffusion, i.e. by ‘morphallaxis’. Our series of results, and our methods for spatial correlation between visual field positions and the boundaries between host- and graft-derived tissue, are inadequate for precise assessment of graft vs. host participation in the respecified retinal domain. A variety of detailed relationships between growth and positional respecification are however possible (see also Conway et al. 1980).

The hypothesis that pattern respecification is the result of intercalation across gaps in a continuous set of values, wherever a ‘shorter’ route for such intercalation exists in recombinant eye rudiments, may go far to explain other reported phenomena. Thus even large transferred fragments frequently grow into respecified retinal domains of the sort this model would predict, when operating salines of low ionic strength are used which may provoke cell death (Hunt & Frank, 1975). In some circumstances, rounded-up half-eye fragments develop with maintenance of half-sets of positional codes (i.e. ‘paralysis’ of intercalation) as judged by their tectal projections and the behaviour of their fibres in the diencephalon. Alternatively, they may develop with mirror-symmetrical specificity or regenerate a whole, single structure (i.e. intercalation round the two possible ‘shorter’ routes, according to the views expressed here - see Berman & Hunt, 1975; Feldman & Gaze, 1975 with appendix by N. Macdonald; Straznicky, Gaze & Keating, 1980).

It is satisfying to find that the relatively detailed positional system required for the establishment of organized central connections from the eye, may be controlled during embryogenesis in a way consonant with findings from other pattern-forming systems.

We thank June Colville for expert histological assistance, and our colleagues David Will-shaw and James Fawcett for stimulating discussions and comments during the work and the preparation of the manuscript.