Cell division during intercalary regeneration in the cockroach leg

In many developing animals cellular interactions have been shown to be fundamental in determining the fate of individual cells and the extent to which they divide. In the single-layered insect epidermis, for example, grafting and other surgical operations which juxtapose cells which are not normally neighbours lead to a modification in the pattern of cuticular structures formed and an increase in growth of the tissue. This has been demonstrated in the epidermis of the leg (e.g. Bohn, 1970; French, 1978), and abdominal segment (e.g. Nübler-Jung, 1977; Wright & Lawrence, 1981), and in the imaginal discs of Drosophila (e.g. Haynie & Bryant, 1976; Abbott, Karpen & Schubiger, 1981).

In the cockroach leg the single layer of surface epidermal cells forms precise patterns of structures (e.g. bristles, spines, pigmented regions) in the overlying cuticle. If different proximal-distal levels of a leg segment are grafted together, growth occurs in the region of the junction and the intervening levels are formed by intercalary regeneration (Bohn, 1970; Bullière, 1971). Similarly, if cells from different circumferential positions on a segment are grafted together, the intervening circumferential positions are produced by intercalary regeneration (French, 1978). These results (and others) lead to the formulation of the Polar Coordinate Model (French, Bryant & Bryant, 1976; Bryant, French & Bryant, 1981). This proposes that the limb epidermis has a two-dimensional map of positional values arranged along its longitudinal and circumferential axes, and that pattern regulation occurs by epimorphosis (Morgan, 1901). Interaction between cells with different values provokes division of those cells and the new daughter cells adopt intermediate values, producing an intercalary regenerate.

The Polar Coordinate Model (PCM) was derived almost entirely from observation of the final pattern of cuticular structures formed after various experimental perturbations. To evaluate the model, and more generally to explore the relationship between pattern formation and cell division, it is vital to also observe cell behaviour during the processes of wound healing and regeneration. Some such studies have been made on imaginal disc fragments (e.g. Reinhardt, Hodgkin & Bryant, 1977; Reinhardt & Bryant, 1981; Dale & Bownes, 1980), and on cockroach limbs undergoing distal regeneration (Truby, 1983). In the experiments reported here we have grafted together epidermal cells from different circumferential positions on the cockroach femur and have looked directly at healing and cell division during subsequent intercalary regeneration. We demonstrate that intercalary regeneration involves local cell divisions which start around the time of healing, well before the normal cell divisions associated with growth during the moult cycle.

Larvae of the cockroach, Blabera craniifer were kept at 28·5°C (±1°C) in constant darkness, and were provided with laboratory rat pellets and moist cotton wool. Grafting operations were performed on CO₂-anaesthetized 5th instar animals, 2– 3 days after moulting. Using fine forceps and a razor-blade knife, a longitudinal strip of cuticle plus epidermis was removed from the femur of the donor mesothoracic leg (after the leg had been removed from the animal) and grafted into a site prepared on the metathoracic femur of the same animal. In this way the same confrontation of graft and host cells (from different positions or from the same circumferential position) was created all along each longitudinal edge of the graft. Grafts were secured by dried haemolymph and the operated cockroaches were then kept in small groups in plastic sandwich boxes.

Some animals were kept until after the 1st or 2nd postoperative moult when their operated legs were removed, fixed in 70% ethanol and examined for cuticular features. Other animals were selected for histological examination and were taken at progressively later times after the operation, anaesthetized, injected with colchicine (10 μl of a 0·25 % solution made up in distilled water) and left for 11 h. The grafted femurs were then removed, fixed for 2 h in alcoholic Bouin, stored for several weeks in 70 % isopropyl alcohol to soften the cuticle, embedded in paraffin wax and sectioned at 10μm in the transverse plane. Sections were stained in haemalum and eosin and, in the region of the centre of the graft, 20 sections taken at five-section (i.e. 50 μm) intervals were examined in detail and the positions of any arrested mitoses were plotted on standard diagrams of the femur circumference. The thick cuticle was sometimes difficult to section and in cases where a chosen section was incomplete or damaged, an adjacent section was scored instead.

The metathoracic (host) or mesothoracic (donor) femur of Blabera is an anteroposteriorly flattened cylinder of cuticle bearing clearly recognizable rows of bristles and bands of pigmentation at different circumferential positions (mainly on the medial and lateral faces), as shown in Fig. 1. The femur also contains apodemes which are cuticular invaginations from the femur/tibia joint and from the distal tip of the tarsus. The surface cuticle and internal apodemes are secreted every moult by an underlying single-layered epidermis. The femur contains large muscles which attach to the apodeme epidermis and to particular regions of lateral, midanterior, and midposterior surface epidermis (see Fig. 1).

A strip between circumferential positions 8 and 4 (see labelling of femur circumference in Fig. 1) was removed from the left mesothoracic femur and grafted to a site between positions 9 and 8 on the left metathoracic femur, as shown in Fig. 2. On the same animal, a control graft was performed on the right metathoracic femur: a strip between positions 9 and 8 was removed from the leg and then replaced in the same site. Both grafts involved freeing some muscle attachments in the region of position 9.

A series of 29 operated animals moulted after 18– 24 days (27/29 moulting in 20 ± 1 days); 13 were fixed and the rest left to the 2nd postoperative moult.

After the 1st moult the experimental (left) femur had a considerably increased overall circumference in the grafted region. The bristle rows and well-defined pigmentation bands of the original host femur and the grafted medial face were unmodified (except for the growth normally occurring between 5th and 6th instars) but along both graft-host junctions intercalary regeneration had produced new tissue. At the junction between graft position 8 and host position 9 (the 8–9 junction) this tissue had no recognizable cuticular features. At the junction between graft position 4 and host position 8 (the 4–8 junction) the new tissue had either no recognizable features, occasional short bristles and some light pigmentation (Fig. 2Aiii), or had two irregular rows of short bristles separated by a poorly defined lightly pigmented band.

After the second moult, all parts of the femur circumference appeared to have grown fairly evenly, the 8– 9 intercalary regenerate still bore no recognizable features, and the 4– 8 intercalary regenerate was in all cases a clear medial face (i.e. positions 5,6 and 7 - Fig. 2Aiv).

After the 1st moult, the control (right) femur had the circumference normal for the 6th instar, and the control graft had healed in, provoking no extra growth or modification of cuticular structures (Fig. 2Biii).

On each alternate day following the grafting operations (Days 1,3,5,7,9, 11,13,15,17), some operated animals were treated with colchicine, and their experimental and control legs fixed, sectioned and examined for the appearance of the epidermis and the occurrence and distribution of mitotic cells.

The length of the normal moult cycle varies between animals within the same instar. Therefore experimental and control operations were performed on the same animals so that the numbers of mitotic cells could be compared statistically as pairs by analysis of variance. We compared the total numbers of mitotic cells, the numbers of mitotic cells at the graft-host borders, and the numbers of mitotic cells elsewhere, in the experimental and control legs for each day. The numbers of mitotic cells in each class on each day are shown in Fig. 7.

The graft-host junction was indicated by the obvious break in the leg cuticle. On either side of this break was a region where epidermis had been damaged or removed, to a greater or lesser extent, during the grafting operation, and which was filled initially by clotted haemolymph and haemocytes. Subsequently epidermal cells adjacent to this region detached from their overlying cuticle and migrated to form a continuous sheet under the wound region. In experimental legs these epidermal cells were considered to be ‘the graft-host border’ but in control legs, where the degree of wounding was less, an equivalent band of 20– 30 cells was considered to be the ‘graft-host border’.

The first event after grafting is closure of the wound by clotted haemolymph and accumulation of one or more layers of haemocytes beneath the clot. These are distinguishable from epidermal cells by their small darkly staining nuclei and flattened shape. Sections through experimental grafts showed that at neither the 8– 9 nor the 4– 8 borders were the host and graft epidermis in contact (Fig. 3A,B). The control grafts fitted better than the experimental grafts (Fig. 3D) so that the host and graft epidermis were nearly in contact (Fig. 3E). At all host-graft junctions haemocytes had accumulated under the cut edges of epidermis, so that a continuous cell layer was present under the wound (Fig. 3B,E).

Very few mitotic cells were observed at this stage (Fig. 7).

By Day 3 there had been some migration of epidermal cells through the plug of haemocytes so that in experimental legs host and graft epidermis were in contact or in close proximity (Fig. 3C) and in control legs they were usually in contact (Fig. 3F). The epidermal cells remain as a coherent sheet during migration.

There were few dividing cells in regions away from the graft-host borders on the experimental and control legs, and the difference was not significant. There were more divisions at the graft-host borders and there were significantly (P = 0·05) more at those of experimental legs than those of control legs (Fig. 7).

On Day 5 the host and graft epidermis had formed a continuous sheet (Fig. 4A,B,E,F) and had produced a continuous layer of endocuticle (Fig. 4B,F).

There were still relatively few dividing cells at regions other than the graft-host borders’ and these did not differ significantly between experimental and control legs. There were however consistently more mitotic cells at the experimental graft-host borders than at the control graft-host borders (Fig. 7), although this was not statistically significant at the 5 % level.

The epidermis along the anterior face had begun to thicken slightly and the chromatin within the epidermal cells to condense. Some degenerating cells were observed, either as pycnotic nuclei or as clusters of darkly stained droplets, but these were always confined to the graft-host border.

Experimental legs showed significantly (P = 0·05) more dividing cells than control legs especially at the graft-host borders (Figs 4C,G and 7), but also elsewhere around the circumference.

The appearance of the epidermis was much the same as on Day 7.

The numbers of mitotic cells had increased greatly throughout the legs (Figs 4D,H and 7). At the graft-host borders experimental legs showed a significantly (P = 0·05) higher number of mitotic cells than did control legs, but elsewhere there was no significant difference.

At this stage all of the epidermis around the circumference of the leg was thickened, including that of the posterior face.

Mitotic cells were observed in large numbers all over both control and experimental legs (Figs 5A,B and 7). At the graft-host border experimental legs showed significantly (P = 0·05) more mitotic cells than did controls. At other positions, however, control legs had significantly more mitoses.

Apolysis, the withdrawal of the epidermis from the overlying cuticle, had started along the host medial face in a few animals. There were numerous cell divisions throughout the epidermis of experimental and control legs (Fig. 5C,D) and the difference in numbers was not significant, either at the graft-host borders or elsewhere (Fig. 7).

The epidermis had completely withdrawn from the cuticle and expanded to form extensive folds (Fig. 6A,B) over which the cells had begun to secrete a new cuticle. The folding was more extensive in the region of the experimental graft than in the region of the control graft (compare Fig. 6A,B). No dividing cells were observed at this stage.

The overall appearance of the epidermis was similar to that of Day 15 except that the newly secreted cuticle was considerably thicker (Fig. 6C,D). No mitotic cells were observed.

These results suggest that events after grafting fall into the following four phases:

which takes place during Days 1– 3 and results in the production of a continuous epidermis. By Day 3 mitotic cells are observed at the graft-host borders and are more numerous in experimental legs.

which takes place on Days 5,7 and 9, during which there are low levels of mitosis at the control graft-host borders, and much higher levels at experimental graft-host borders, with mitosis initially low elsewhere but becoming frequent by Day 9.

which occur throughout the epidermis on Days 11 and 13 with mitosis still considerably higher at the graft-host border in the experimental legs on Day 11 but not significantly different on Day 13.

which takes place during Days 15– 17 when no mitoses are observed. Thus this experiment has shown a considerable difference in behaviour at the graft-host borders, with cell divisions much higher at the experimental borders from the end of healing through the intercalation and into the profiferation phase. The experimental borders have a positional disparity between the graft and host epidermis, which the control borders do not have. However, the experimental borders also differ from control borders in that, because of poorer fit, more cell migration is necessary to attain epidermal confluence in the healing phase and this may be accompanied by cell divisions in the surrounding area of low cell density (Wigglesworth, 1937). Hence it could be that the observed difference in the number of dividing cells in our experiments resulted from the differing extent of repair in control and experimental legs rather than the positional disparity. However, the enhanced cell division at experimental borders lasted long after a new cuticle was secreted (c.f. Wigglesworth, 1937), so that it is likely to indicate intercalation rather than healing. However, to confirm this we performed another set of grafting operations where the degree of mechanical disruption of the epidermis at control and experimental junctions was well matched.

(C) Graft II: right medial face grafted to left anterior face

A strip between circumferential positions 4 and 8 was removed from the right mesothoracic femur and grafted to a site between positions 9 and 8 on the left metathoracic femur, as shown in Fig. 8. As in Graft I, muscle attachments in the region of position 9 had to be broken in preparing the host site.

A series of 25 operated animals moulted in 18– 28 days (20/25 moulting in 20– 25 days): 14 were fixed and the rest left to the 2nd postoperative moult.

After the 1st moult the original graft and host tissues had grown as normal and their cuticular structures remained unchanged. At the control 8– 8 junction, the graft had healed in with no production of extra tissue, but at the experimental 4– 9 junction there was a large intercalary regenerate with either no recognizable structures, scattered short bristles and some light pigmentation (Fig. 8iii), or two irregular rows of short bristles separated by a lightly pigmented region.

After the 2nd postoperative moult there was still no extra tissue at the 8– 8 junction and, in all cases, the large intercalary regenerate at the 4– – 9 junction included a clear medial face (positions 5,6 and 7, Fig. 8iv).

On Days 3,5 and 12 some animals were treated with colchicine and their operated legs fixed, sectioned and examined. These days were chosen to give a representative picture of the stages of wound healing, intercalary regeneration, and general growth as observed for graft I.

At one junction, graft epidermis from position 8 was confronted with host epidermis from position 8 (the 8– 8 junction, Fig. 8) and at the other, graft epidermis from position 4 was confronted with host epidermis from position 9 (the 4– 9 junction, Fig. 8) providing control (non-intercalating) and experimental (intercalating) regions within the same leg. We compared the numbers of mitotic cells at the 8– 8 and 4– 9 graft-host borders (the graft-host border defined as for graft I) for each of the 3 days using the same statistical methods as before. These results are shown in Fig. 10.

The general appearance of the epidermis was the same as for Graft I on the corresponding day and a description will not be repeated here. As anticipated, the 8– 8 and 4– 9 junctions looked very similar in configuration (Fig. 9A,B) and presumably required a similar amount of wound healing and cell migration to produce, in all cases, a continuous sheet of epidermis with its overlying cuticle by Day 5.

On Day 3 there were consistently more dividing cells at the 4– 9 border than at the 8– 8 border (Fig. 10), although the difference was not statistically significant. On Day 5 there were significantly more mitotic cells at the 4– 9 border than at the 8– 8 border which now showed fewer than on Day 3. On Day 12 there were again significantly more dividing cells at the 4– 9 border than at the 8– 8 border. The latter again showed fewer dividing cells than on Day 3.

Since the amount of wound healing is similar at the two borders, the far greater number of dividing cells found at the 4–9 border on Days 5 and 12 is caused by the positional discrepancy created in this region by the grafting operation. This result therefore strongly supports the conclusion reached from Graft I that cell division occurring at graft-host borders on and after Day 5 reflects intercalary regeneration.

Grafts which disturb normal neighbourhood relationships within the insect leg epidermis result in intercalary regeneration whereby growth is stimulated and new pattern elements are formed. In the present experiments a strip graft is moved around the circumference, creating circumferential positional disparities at both (Graft I) or one (Graft II) graft-host junctions. The total limb circumference increases and the new structures formed at each graft-host junction correspond, by the shortest route, to the missing part of the circumference (French, 1978, 1980).

Examination of the cuticular structures formed at the first and subsequent postoperative moults suggests that the extra growth occurs locally at the junction and that the new structures form within the new tissue. Hence the graft and host retain their differentiated structures (e.g. bristle rows some 10 cell diameters away from the cut edges at the time of grafting) but are separated by an intercalary regenerate which initially bears few cuticular structures (e.g. Figs 2Aiii, 8iii) but which subsequently develops them (Figs 2Aiv, 8iv).

The present histological study was undertaken to show directly that a positional discontinuity does provoke cell divisions in addition to those resulting from wounding inevitably accompanying grafting and those normally occurring in the moult cycle, and to investigate when these divisions occur and whether they are indeed localized at the site of the discontinuity.

During each instar the epidermis grows by proliferative cell divisions and new bristles, hairs and chemosensilla are added to the increased body surface by differentiative cell divisions. Proliferative divisions generally occur during the middle third of the moult cycle (Lawrence, 1966b; Bullière, 1972; Kunkel, 1975;

Truby, 1983), while differentiative divisions occur either before proliferative divisions, in the case of chemosensilla or bristles (Lawrence, 1966b; Kunkel, 1975), or after, in the case of hairs (Lawrence, 1966b). In all cases, cell divisions have ceased at the time of cuticle secretion.

After a wound, cell divisions occur at other times. Following a cut, epidermal cells adjacent to the wound migrate across it to re-establish a continuous epidermis (Wigglesworth, 1937). Nearby regions of reduced cell density undergo a period of local cell division which ceases when epidermal continuity has been restored and the cells in the wound area have secreted a new cuticular covering (Wigglesworth, 1937). After amputation of the distal part of the leg, cell divisions also appear much earlier in the moult cycle than normal. Truby (1983) observed DNA synthesis and mitosis at 2 days (in a 16-day moult cycle), while Bullière (1972) detected DNA synthesis on Day 7 (in a 40-day moult cycle) but, presumably because he did not harvest mitoses with colchicine, he did not detect cell divisions until Day 18. Following amputation late in the instar, wound healing and some cell division occurs prior to the moult, but regeneration only starts after the moult and, again, cell divisions are found as early as Day 2 of the following moult cycle (Truby, 1985). However, unlike cell divisions occurring after a simple incision (which cease shortly after the wound has healed), the additional cell divisions following amputation continue into the proliferative stage of the moult cycle (Bullière, 1972; Truby, 1983). The additional cell divisions result not only from the extensive migration required for healing over the amputation site, but also from the production from stump tissue of a regenerated leg.

In the experiments described here we were able to perform experimental and control grafts and could therefore distinguish cell divisions involved in healing, intercalation, and growth. We observed the following pattern of cell divisions after grafting operations:

(a) In regions away from the graft-host borders, cell divisions in both experimental and control legs were negligible on Day 1, were present in low numbers on Days 3 and 5, increasing on Days 7 and 9 to give very high levels on Days 11 and 13. On Days 15 and 17 cuticle secretion had begun and mitosis had ceased.

The few early divisions (Days 3 and 5) could be differentiative divisions giving rise to new hairs and bristles. Most were too far away from the graft-host borders to represent the wound healing response, although some adjacent to the borders may be involved in this. The later divisions correspond to the proliferative divisions observed in the absence of colchicine in non-regenerating epidermis of Blabera craniifer from Day 22 to Day 34 in a cycle of 40 days (Bullière, 1972).

We therefore interpret mitosis away from the borders as corresponding to the normal events of the moult cycle.

(b) At the graft-host border, the number of cell divisions was significantly higher on Day 3 in the poorly fitting Graft I experimental grafts than in the control grafts. However, in Graft II where there was no difference in the degree of wounding at the control and experimental borders, there was no significant difference on Day 3 in cell division at the experimental and control graft-host borders. Cell divisions at the graft-host borders on Day 3 therefore represent a wound healing response.

(c) At the graft-host borders on Day 5, epidermal continuity had been restored and a new endocuticle had been secreted by cells under the wound. Any mitotic cells observed at, and after this stage are therefore probably associated with intercalary regeneration of new tissue. This view is confirmed by the very different levels of cell division at the poorly fitting Graft II experimental and control junctions. We interpret the low levels seen at control junctions on Day 5 and subsequently, as representing a small degree of intercalation provoked by inexactly fitting grafts, or as a residual wound response. Much higher levels of cell division were observed at experimental graft-host borders than at control graft-host borders on all days from 5 to 13 following Graft I and on Days 5 and 12 following Graft II. We interpret this local cell division as representing intercalary regeneration of tissue at the region of positional discrepancy.

Intercalary regeneration therefore does not result from an augmentation of cell division during the normal proliferative phase of the moult cycle, nor from an extension of this phase later into the cycle, but results from cell divisions beginning much earlier in the moult cycle and extending into the proliferative phase (Fig. 11).

We now have shown directly that intercalation is associated with division of cells at the ‘graft-host border’, defined as a band of 20– 30 cells disrupted by damage and subsequent healing. Because of the mechanical disturbance involved in the grafting operation, it is not possible at present to be more accurate about the localization of the initial mitoses of intercalation. As intercalation proceeds, a region of folded and thickened epidermis is formed, bulging a little away from the old cuticle, and again it is difficult to be precise but it seems that mitoses can occur anywhere within it, rather than just at a narrow front corresponding to the confrontation of host- and graft-derived cells. From the histological observations, it is not possible to say whether both host and graft contribute to the intercalary regenerate, but this is known to be the case in similar grafts between leg segments with different cuticular structures (French, 1980).

Intercalary regeneration following grafting in cockroach legs is very similar to the pattern regulation observed when fragments of Drosophila imaginal disc are cultured for some time in vivo before metamorphosis (Haynie & Bryant, 1976; French et al. 1976). In general, small fragments of the wing disc duplicate, forming a mirror-symmetrical partial pattern, while complementary large fragments regenerate, completing the pattern (Bryant, 1975). Many fragments have been shown to heal their cut edges together in such a way that shortest route intercalation would generate the observed pattern (Reinhardt et al. 1977; Reinhardt & Bryant, 1981), and other fragments which show variable patterns of regeneration show corresponding variability in their mode of healing (Dale & Bownes, 1985).

Cell division is necessary for regeneration of disc fragments and, by wholemount autoradiography after thymidine pulse labelling (Dale & Bownes, 1980) and mitotic counts of sectioned discs (O’Brochta & Bryant, submitted) it has been shown that the dividing cells are localized in the region of wound healing. The estimated width of the zone of cell division is 4– 13 cells (Dale & Bownes, 1980) or 15 cells (O’Brochta & Bryant, submitted) although, using rather different techniques of long-term labelling with thymidine, Adler (1984) suggests that an increase in cell division spreads much further back from the wound surface (see also Adler, 1981).

In Drosophila, partial duplication and triplication of legs can arise after heat treatment during larval development of temperature-sensitive cell-lethal mutants. These structures have been interpreted as resulting from localized cell death, healing, and intercalary regeneration within the developing disc (Girton, 1981). From clonal analysis, Girton & Russell (1980) estimated that, in a duplicated pattern, the extra leg was derived from only 10–20 cells within the damaged disc, again suggesting that intercalary regeneration involves only very local cell division at a site of positional discontinuity. Abbott, Karpen & Schubiger (1981) used clonal analysis similarly to demonstrate that the regenerate or duplicate produced by an imaginal leg disc fragment derives from cells originating at or near the edge (in this case, only one edge) of the fragment, but since they used the Minute technique they could not estimate the number of cells involved.

An epimorphic model (such as the PCM) involves stable cellular positional values changing only during cell division provoked by a discontinuity between adjacent cells, and it therefore predicts that regeneration will involve only local cell division at a graft-host junction or at an amputation surface. Clearly, circumferential intercalary regeneration in grafted cockroach legs or in imaginal disc fragments involves localized cell division, and is thus broadly compatible with epimorphosis rather than, at the other extreme, a morphallactic model based on widespread respecification of cells. There are several possible forms of epimorphosis (Fig. 12A–C) and all predict that the first cell divisions will be precisely at the confrontation of disparate cells and that all cells of the intercalary regenerate will have divided. The forms differ in predictions concerning subsequent spread of cell division back from the junction, loss of extreme values at the junction, and the location of cell division within the developing intercalary regenerate (Fig. 12). In practice, however, with the complication of damage and healing, with the limited cuticular markers for cellular positional value, and with the difficulty of precisely locating divisions relative to the confrontation of grafted cells, and determining whether all cells of the regenerate have indeed divided, it is difficult to distinguish between some of these possibilities (e.g. Fig. 12B,C) or even to rule out a rather different model involving an initial morphallactic phase of resetting cellular positional values (Fig. 12D).

In a recent histological study of distal regeneration after autotomy of the cockroach leg at the femur-trochanter level, Truby (1983) found that the initial cell divisions occurred during wound healing, just proximal to the amputation site.

Cell division then spreads back into the stump to the level of the distal coxa (50–80 cell diameters in 2nd instar animals). After an autotomy late in the instar, wound healing and associated cell divisions occurred but regeneration only commenced after the next moult. Cell divisions then started at the distal tip of the stump and spread back, again to distal coxa level (Truby, 1985). Cell division continued at all levels within the blastema, visible segmentation occurred, and the regenerate continued to grow.

The PCM proposes that distal regeneration occurs because healing across the end of the stump provokes intercalation between cells with different values and, because of the presence of adjacent cells with unaltered values, the new cells take up more distal positional values (Bryant et al. 1981). Thus distal regeneration, like intercalation after a graft, should involve only local cell division at the site of discontinuity. Some versions of the PCM would allow cell division to spread back into stump tissue (as in Fig. 12C) but this effect would not be expected to extend very far back from the amputation site (Lewis, 1981) and most cell divisions would still occur at the tip of the developing blastema. The difference between Truby’s results and our own may indicate that distal regeneration is different from intercalation or that they are similar processes both involving a degree of cellular respecification before cell division (Truby, 1983). Because much more new tissue is formed after a proximal amputation than after a strip graft, respecification and subsequent cell division would be much more extensive in distal regeneration than in intercalation.

This work has been supported by the British Science Research Council, the European Molecular Biology Organization, and the Royal Society.