The sequence of appearance of different structures during regulation was ascertained for three wing disc fragments. It was found that structures were added sequentially from the cut edge. Regeneration in one fragment was very fast and successful, and it is suggested that this is difficult to explain in terms of wound healing being responsible for the decision to regenerate rather than duplicate. In another fragment, regeneration appeared to proceed from one part of the cut edge only, this being the region through which the anteroposterior compartment boundary runs. Compartment boundaries have been implicated in both proximodistal and circumferential regeneration, and it is suggested that regeneration is controlled at these sites. The decision to regenerate rather than duplicate may be made on the basis of which compartment boundaries are exposed at the cut edge, and their preferred direction of regeneration. The implications of this interpretation for the control of growth are discussed.

A prerequisite for pattern formation in imaginai discs is that the correct number of cells, in the correct configuration, should be present at the time of differentiation. Since this is accomplished mainly by cell division, growth is an essential part of the pattern-forming process. Regeneration re-establishes this essential configuration, and therefore can be expected to provide information about the growth processes responsible.

If the wing disc is cut into two fragments, one regenerates during culture while the other duplicates (Bryant, 1975; Karlsson, 1981). This rule is in general very well obeyed, but there are two notable exceptions, in both of which fragments duplicate structures expected from the fate map but also regenerate others. In one, the regenerated structures are proximal (Karlsson & Smith, 1981), and in the other distal (Karlsson, 1980). The ability to regenerate distal structures while duplicating was found to be related to compartments; only fragments containing the ventral end of the anteroposterior (AP) compartment boundary were able to regenerate wing blade (Karlsson, 1980). After dissociation, fragments containing the dorsal end are also able to produce wing blade (Wilcox & Smith, 1980). Thus it appears that the AP boundary is important in controlling proximodistal regeneration; this also seems to be the case in the leg disc (Schubiger & Schubiger, 1978), and implies that growth in this axis is also controlled at these places.

There are other observations which support this idea. Simpson (1976) has found that clones of slow-growing mutants disappear from the middle of compartments but not from the edges. This suggests that growth rules are different in the two places, a conclusion supported by the finding of Lawrence & Morata (1976) that small engrailed clones are capable of causing a shape distortion in the wing blade if they touch the dorsoventral compartment boundary but not otherwise. These authors suggested that this was because the compartment boundary was somehow instrumental in controlling growth, and that the reason internal clones caused no shape distortion was that they came under the control of the wild-type boundary.

Recently, further evidence for the involvement of compartment boundaries in regeneration has been reported (Karlsson, 1981). It was found that of two complementary wing disc fragments, the physically larger is not always the one which regenerates; in terms of the polar coordinate model (French, Bryant & Bryant, 1976), the smaller fragment has over half the circumferential positional values. Testing of a large number of fragment pairs led to the conclusion that the values are very unevenly spaced indeed, half of them being tightly clustered round the ends of the AP boundary. Since it is this boundary which affects proximodistal regeneration, this seemed to suggest a connexion between the ability to regenerate distally and the ability to regenerate rather than duplicate circumferentially, and that both phenomena were somehow connected to compartments.

It seemed possible that more information could be gained about this relationship by studying the sequence in which different structures were produced during regulation. For instance, it could be that the different parts of the cut edge contribute unequally to the regenerate, and it could also be that distal structures are produced in different ways from fragments containing different ends of the AP boundary. Two fragments were therefore chosen for detailed study, both having 8 of the 12 circumferential positional values and each containing a different end of the AP boundary. A third fragment was also studied; this was one which duplicated circumferentially and regenerated distally.

Host and donor flies were of ebony-11 genotype, and were raised at 25°C on standard corn meal/agar/syrup medium seeded with live yeast. Wing discs were removed from late third-instar larvae in insect Ringer. Fragments were cut with tungsten needles and implanted into the body cavities of well-fed 1-to 3-day-old fertilized adult females, where they remained for varying periods. They were then removed and re-implanted into the body cavities of late third-instar larvae, using the method of Ephrussi & Beadle (Ursprung, 1967). When the hosts emerged as adults, the metamorphosed implants were removed, mounted in Gurr’s Hydramount, and scored for the structures shown in Fig. 1. Implants cultured for 0 days (Tables 1–3) were implanted directly into late third-instar larvae.

Table 1

Structures differentiated by implants of fragment (a)

Structures differentiated by implants of fragment (a)
Structures differentiated by implants of fragment (a)
Fig. 1

Fate map of the wing disc, the fragments tested, and the sequence of regeneration. Left: fate map of the wing disc (after Bryant, 1975). Presumptive wing blade is stippled, and compartment boundaries are shown in broken lines. The dorsoventral boundary runs along the wing margin. The position of the anteroposterior boundary was ascertained by Brower, Lawrence & Wilcox, 1981, using clones of a succinic dehydrogenase mutant. Positional values 1–12/0 from Karlsson (1981). Right: fragments (a), (b) and (c). The numbers outside the fragments denote the first day on which a structure was present at a frequency of 30% or more. Along the wing margin, in fragments (a) and (b), the first number refers to the day on which triple row appeared at a frequency of 30%, and the next to the day on which the average number of triple row bristles reached 10. The numbers inside fragment (c) denote the first day on which a structure was duplicated more often than not. Along the wing margin, the upper figure refers to the first day on which a structure was present at a frequency of 30% and the lower figure to the first day on which it was more often duplicated than not.

Abbreviations:

Scu, scutellum; HP, humeral plate; UP, unnamed plate; AS 1, 2, 3, first, second and third axillary sclerites; T, D, Rows, triple and double rows of chaetes; P Row, posterior row of hairs; PDR, proximal dorsal radius; A Lobe, alar lobe; A Cord, axillary cord; YC, yellow club; PVR, proximal ventral radius; PS, pleural sclerite; PWP, pleural wing process.

Fig. 1

Fate map of the wing disc, the fragments tested, and the sequence of regeneration. Left: fate map of the wing disc (after Bryant, 1975). Presumptive wing blade is stippled, and compartment boundaries are shown in broken lines. The dorsoventral boundary runs along the wing margin. The position of the anteroposterior boundary was ascertained by Brower, Lawrence & Wilcox, 1981, using clones of a succinic dehydrogenase mutant. Positional values 1–12/0 from Karlsson (1981). Right: fragments (a), (b) and (c). The numbers outside the fragments denote the first day on which a structure was present at a frequency of 30% or more. Along the wing margin, in fragments (a) and (b), the first number refers to the day on which triple row appeared at a frequency of 30%, and the next to the day on which the average number of triple row bristles reached 10. The numbers inside fragment (c) denote the first day on which a structure was duplicated more often than not. Along the wing margin, the upper figure refers to the first day on which a structure was present at a frequency of 30% and the lower figure to the first day on which it was more often duplicated than not.

Abbreviations:

Scu, scutellum; HP, humeral plate; UP, unnamed plate; AS 1, 2, 3, first, second and third axillary sclerites; T, D, Rows, triple and double rows of chaetes; P Row, posterior row of hairs; PDR, proximal dorsal radius; A Lobe, alar lobe; A Cord, axillary cord; YC, yellow club; PVR, proximal ventral radius; PS, pleural sclerite; PWP, pleural wing process.

Terminology

For convenience, the four ends of the two compartment boundaries are referred to in the text as D-AP, V-AP (dorsal and ventral ends of the anteroposterior boundary) and A-DV, P-DV (anterior and posterior ends of the dorsoventral boundary).

Figure 1 shows a fate map of the wing disc, the fragments tested, and the day on which each structure first appeared (or was duplicated) at high frequency. Tables 1–3 show the frequencies with which the different structures were present.

The data show that structures were produced sequentially from the cut edge in a very orderly way. No structure far from the cut edge appeared until intervening structures were present. The amount of wing blade rose steadily with increasing culture periods, and the same was true for the numbers of wing margin (triple row) bristles which were counted in fragments (a) and (b). If a given structure appeared for the first time on one day, it was almost always present at maximum frequency the next. However, there were considerable differences in the behaviour of the two regenerating fragments.

Fragment (a)

The most noticeable difference between this fragment and fragment (b) was that wing blade and wing margin structures (triple row, double row and posterior row) appeared considerably sooner in fragment (a). Some wing blade was present in all implants injected directly into larvae (0 days of culture); that this was the result of regeneration is indicated by its complete absence from fragment (b) after 0 days, since the two fragments were cut at exactly the same distance from the edge of the disc.

It seems to be the case that the time of appearance of structures close to a particular part of the cut edge is indicative of whether it contributes to the regenerate (Abbott, Karpen & Schubiger, 1981). Thus the very early appearance of wing blade in fragment (a) indicates that the ventral half of this fragment, which lies next to the wing blade, contributes from the start of regeneration. In fact, since no wing blade can be produced without the ventral end of the AP boundary (V-AP), it is probably this point, at the extreme ventral end of the fragment, at which regeneration starts. This can only be inferred since all parts of the wing blade appear the same in implants.

Whether the dorsal half of the fragment contributes is less clear. The structures of the dorsal hinge appeared somewhat later than the wing blade and could possibly have been produced from it; however, if this were so, the first hinge structure to appear should be the proximal dorsal radius (see Fig. 1), and in fact this appeared one day later than the remaining ones. Thus it seems likely that some of the dorsal edge does contribute. The extreme dorsal end possibly does not, since scutellum, which lies close to it, never appeared at high frequency.

The above observations suggest that a wave of regeneration spreads along the cut edge from ventral to dorsal.

Regeneration was very successful in this fragment. Few implants failed to regenerate, and only six had duplicated structures, a single one in each case (two cases each of pleural sclerite, humeral plate and third axillary sclerite). Regeneration could be completed in as little as 3 days; the only structure to continue increasing in frequency after this time was third axillary sclerite. Axillary cord, alar lobe and scutellum never appeared at high frequency; Bryant (1975) has cultured a similar fragment for 7 days, after which these structures were still present at low frequency, so it seems likely that in many cases they never appear.

Fragment (b)

Wing blade appeared in this fragment considerably later than in fragment (a). The first structures to appear, on day 2, were those of the dorsal hinge, which lie close to the edge in the dorsal half of the fragment. It seems very likely that these must be produced before wing blade can appear, as virtually all (18/22) implants having wing blade after 2 and 3 days also had one or more of these structures. Thus wing blade is probably produced at first entirely from this reg-on The ventral half which hes close to triple row, appears not to contr.bute at least until day 4, which is the first day on which this structure appeared in any quantity. It seems quite likely that the very ends of the fragment do not contribute at all, since structures lying very close to the edge at these sites did not appear until day 5, and yellow club, at the extreme ventral end, was even duplicated in many cases.

A high frequency of duplication is the other major difference between this fragment and fragment (a); a total of 26 implants had duplicated structures compared with 6 in fragment (a). These were mostly yellow club (18 cases of yellow club, 7 each of humeral plate and notum, 2 each of scutellum and tegula, 1 each of costa, unnamed plate, and first and second axillary sclerites). Seven of the 26 had regenerated structures as well.

Thus it seems that the region of the dorsal hinge, through which the AP boundary runs, produces most and possibly all of the regenerate, while the rest of the cut edge is either doing nothing or duplicating. Indeed, many implants did nothing whatsoever and differentiated so poorly that they could not be scored and were not included in the data. This is always true of a few implants, but seldom to the extent seen in this fragment.

Fragment (c)

Most implants of this fragment regenerated distally and duplicated circumferentially; about 20% regenerated completely with little or no evidence of duplication. Wing blade was produced very early as in fragment (a). Wing blade lies closer to the edge of the disc in the posterior than in the anterior, and approximately 1 unit of wing blade (on a scale of 1–5) was expected from the fate map (Karlsson, 1980). In fact, 2·5 units were present in implants injected directly into larvae. It appears likely that the whole of the distal-facing cut edge contributes to the production of wing blade, since the amounts of triple row and posterior row both increased on days 1 and 2, and double row, which lies at the distal wing tip, only appeared in implants having virtually complete triple and posterior rows. It was not possible to count the numbers of triple row bristles in implants of this fragment because this row was often duplicated and the bristles lay on top of one another.

Duplication appeared to spread inwards from the ends of the fragment; third axillary sclerite, at one end, was always duplicated after only 1 day of culture. Tegula, at the other end, was according to Table 3 not duplicated to any extent until day 4, but this is probably misleading since only part of the tegula was present and duplication in this part can be very difficult to detect. Humeral plate, which lies next to the tegula, was duplicated more often than not on day 3, and this was only true for those structures furthest from the ends after day 4. Thus duplication spreads from the posterior end and probably from the anterior end as well. Wing margin structures (triple and posterior rows) were only duplicated to any great extent by day 4, indicating that duplication of proximal structures is followed or accompanied by duplication of the newly regenerated wing blade.

Table 2

Structures differentiated by implants of fragment (b)

Structures differentiated by implants of fragment (b)
Structures differentiated by implants of fragment (b)
Table 3

Structures differentiated by implants of fragment (c)

Structures differentiated by implants of fragment (c)
Structures differentiated by implants of fragment (c)

The results show that pattern regulation in the fragments studied proceeds sequentially and in an orderly way from the cut edge, and suggest that the regenerate can be produced from one part of the edge only. Both these results have been obtained by Abbott and co-workers (1981) in the leg disc; these authors have shown by means of clonal analysis that a part of the cut edge which appears, according to the sequence of structures produced, not to contribute, does indeed not do so.

It is not wholly expected that regeneration should proceed sequentially from the cut edge, since it is thought to be intercalary, and therefore to resemble an averaging process. Averaging should produce those structures furthest from the cut edge first, since these would have positional values intermediate between those at either end. Thus this result suggests that once the initial decision to regenerate or duplicate is made, the rest of the regenerative process proceeds autonomously until all cells have their correct nearest neighbours again.

Since therefore wound healing does not appear to be necessary in the later stages of regulation, one might ask whether it is necessary in the initial stages when the decision to regenerate or duplicate is made. If it is, the behaviour of fragment (a) is not easy to explain. At one end of this fragment, three positional values are crammed into a very small space; without this small piece the fragment would duplicate (Karlsson, 1981). This means that the rest of the fragment has to be aware of the presence of this small piece. If wound healing is to be responsible for the decision to regenerate rather than duplicate, it must somehow ensure that this small piece is in contact with the whole of the rest of the cut edge. Implants of this fragment inspected after 1 day of culture do appear to have curled up and therefore probably healed, but in a rather random fashion, some like the letter S, and others like P, C or O (Karlsson, unpublished observations). Nevertheless, this fragment is extremely successful at regenerating and only very rarely produces duplicated structures. One suspects that other mechanisms besides wound healing might be responsible for providing the signal which leads to the decision to regenerate or duplicate. Since no wing blade can be produced in this type of fragment without the ventral end of the AP boundary (V-AP), the initial signal must come from this region; the sequence of appearance of the different structures suggests that cell division spreads from this part of the edge. Thus the signal might in this case be quite simply an advancing front of cell division.

The spacing of positional values suggests that the AP boundary plays a crucial role in the decision to regenerate or duplicate. There also appears to be some value clustering in the region of the A-DV; fragments which contain more than a certain minimal amount of this boundary, plus either end of the AP boundary, are able to regenerate (Karlsson, 1981). From this it seems possible that the decision is made on the basis of which boundaries are contained in the fragment, certain pairs being able to co-operate in stimulating regeneration.

A possible clue to the role of the AP boundary in regeneration comes from its behaviour in isolation. While the ventral end has a strong tendency to regenerate distally (Karlsson, 1980), the dorsal end appears to have the reverse tendency. Thus many duplicating fragments which lack presumptive notum, through which the AP boundary runs, regenerate notum at high frequency. Notai structures are added in sequence from distal to proximal, and regeneration stops once most of the notum is present (Karlsson & Smith, 1981). That the boundary is somehow responsible seems very likely because fragments lacking any part of it are unable to regenerate notum, and the structures produced lie along the boundary. Thus the two cases of simultaneous regeneration and duplication seem to indicate that new proximodistal positional values can be added at compartment boundaries, and that the two ends of the AP boundary preferentially add values in opposite directions.

These observations lead to the suggestion that the involvement of compartments in regeneration means that new proximodistal values must be laid down at the boundaries before they can be added elsewhere. It might then be fruitful to consider whether the regeneration/duplication decision could be entirely the result of this process. A fragment containing two boundaries which can cooperate in regenerating distally to the extent of producing half the wing blade, say, might be able to continue regenerating until the entire disc was present. Exactly the same process would be happening in the complementary fragment, but here of course it would mean duplication.

Thus the very successful regeneration of fragment (a) in the present experiments might be the result of the strong tendency for distal regeneration of the V-AP, together with the ability of the A-DV to sustain this process. The much less successful regeneration of fragment (b) might result from the tendency of the D-AP to regenerate proximally, which must be overcome before distal regeneration can occur.

If the above interpretation of regeneration and duplication is correct, it provides support for the idea that growth is controlled at compartment boundaries. New proximodistal values may be laid down at these boundaries before they are laid down elsewhere. This could be the reason clones do not grow across them; if new growth emanates from these lines in both directions, clones will be unable to invade this region. If the boundary consists of a double row of cells, which can throw off daughters proximodistally and to one side but not the other, clones will follow the lines but never cross them.

Nothing as yet is known about the way in which mitosis might be used to create complex shapes. One extreme hypothesis would be that each cell, at each cell division, is given specific instructions about whether and how to divide. For the wing disc, this would mean about 50000 separate instructions, and one might ask whether such a large amount of information is really necessary to produce the not overly complex shape of the mature wing disc. Indeed, if all cells were given precise instructions then clone shape would be determinate, which of course it is generally not. The very fact that it is determinate at compartment boundaries suggests that cells have precise growth instructions here and not elsewhere.

An alternative, then, to giving all cells specific growth instructions is to give them only to cells at strategic positions, and allow other cells to follow passively. If this is what is happening, it provides an explanation for Simpson’s (1976) finding that clones of slow-growing mutants disappear from the middle of compartments but not at the edges; if cells in the middle all have the same instruction, namely to divide when space is available, a slow-growing cell would never be permitted to divide because the space around it would always be taken up by its faster-growing neighbours.

If the above analysis is valid, it suggests that it might be fruitful to look for these lines in other organisms. It may be that this principle is a general one, used wherever mitosis is the primary means of creating complex shapes.

This work was supported by an M.R.C. studentship and a Beit Memorial Fellowship.

Abbott
,
L. C.
,
Karpen
,
G. H.
&
Schubiger
,
G.
(
1981
).
Compartmental restrictions and blastema formation during pattern regulation in Drosophila imaginai leg discs
.
Devi Biol. (In the Press
.)
Brower
,
D. L.
,
Lawrence
,
P. A.
&
Wilcox
,
M.
(
1981
).
Clonal analysis of the undifferentiated wing disc of Drosophila
.
Devi Biol. (In the Press
.)
Bryant
,
P. J.
(
1975
).
Pattern formation in the imaginai wing disc of Drosophila melanogaster. fate map, regeneration and duplication
.
J. exp. Zool
.
193
,
49
78
.
French
,
V.
,
Bryant
,
P. J.
&
Bryant
,
S. V.
(
1976
).
Pattern regulation in epimorphic fields
.
Science
193
,
969
981
.
Karlsson
,
J.
(
1980
).
Distal regeneration in proximal fragments of the wing disc of Drosophila
.
J. Embryol. exp. Morph
.
59
,
315
323
.
Karlsson
,
J.
(
1981
).
The distribution of regenerative potential in the wing disc of Drosophila
.
J. Embryol. exp. Morph
.
61
,
303
316
.
Karlsson
,
J.
&
Smith
,
R. J.
(
1981
).
Regeneration from duplicating fragments of the Drosophila wing disc
.
J. Embryol. exp. Morph. (In the Press
.)
Lawrence
,
P. A.
&
Morata
,
G.
(
1976
).
Compartments in the wing of Drosophila: a study of the engrailed gene
.
Devi Biol
.
50
,
321
337
.
Schubiger
,
G.
&
Schubiger
,
M.
(
1978
).
Distal transformation in Drosophila leg imaginai disc fragments
.
Devi Biol
.
67
,
286
295
.
Simpson
,
P.
(
1976
).
Analysis of the compartments of the wing of Drosophila melanogaster mosaic for a temperature-sensitive mutation that reduces mitotic rate
.
Devi Biol
.
54
,
100
115
.
Ursprung
,
H.
(
1967
).
In vivo culture of Drosophila imaginai discs
.
In Methods in Developmental Biology
(ed.
F.H.
Wilt
&
N. K.
Wessels
).
New York
:
Crowell
.
Wilcox
,
M.
&
Smith
,
R. J.
(
1980
).
Compartments and distal outgrowth in the Drosophila imaginai wing disc
.
Wilhelm Roux’ Arch, devl Biol
.
188
,
157
161
.