The regeneration of supernumerary cockroach antennae

During each larval instar the single layer of epidermal cells of hemimetabolous insects separates from the cuticle, grows and secretes a new cuticle bearing a precise pattern of structures such as bristles and spines. Many grafting experiments have been performed on the legs of larval cockroaches (Bohn, 1965, 1970, 1972; Bullière, 1970, 1971; French, 1976, 1978) and other insects (Bart, 1971a; French, 1981 ; Shaw & Bryant, 1975; Balazuc, 1948; Bodenstein, 1937), to analyse the role of cellular interactions in controlling growth and the formation of pattern.

Many other appendages of insects are known to regenerate after amputation but there have been few grafting experiments to see whether their cells will interact to form intercalary or supernumerary regenerates. The results of grafting the forceps of earwigs (Furukawa, 1940) and, especially, the cerci of crickets (Palka & Schubiger, 1975) suggests that the abdominal appendages and the legs respond in the same way to confrontation of normally non-adjacent cells. This paper reports several grafting experiments on the basal scape segment of the larval cockroach antenna. Since the results are comparable with those of similar grafts performed on the leg, we argue that the two developing appendages have a similar spatial organization.

Fourth to sixth instar larvae of the cockroach, Blabera craniifer, were taken from mass cultures (see French, 1978) 2–4 days after moulting, and anaesthetized with CO₂. Amputations and grafts were performed at precise levels in the antenna, using fine forceps, spring scissors and knives made from fragments of razor blade. Operated animals were kept in groups of 20–30 and were killed after the first (rarely), second or third post-operative moult. The heads were fixed in alcohol and examined.

The cockroach antenna (see Fig. 1) consists of three proximal segments - the scape, the pedicel and the third segment - and a long segmented flagellum. It articulates within the antennal socket which is positioned on the head capsule anterior and medial to the compound eye, and it contains a major articulation between the scape and the pedicel. During larval life growth occurs in the proximal segments, particularly the third segment, which forms about 15 new flagellum segments, in distal-to-proximal sequence, during an instar. During the next instar each of these primary flagellum segments splits into two secondary segments (Haas, 1955).

The transverse axes of the antenna are labelled anterior/posterior (A/P) and medial/lateral (M/L); this corresponds only roughly to their orientation on the head, but corresponds to the axes of the leg. The scape is slightly convex on the medial side, but the only circumferential marker structures reliably formed on regenerated antennae are the articulatory structures of the socket and the distinct A and P hinges between scape and pedicel (see Fig. 1 B, C).

Antennae were amputated at the proximal/distal levels used in the grafting operations. Amputation in the proximal scape (42 animals), the mid scape (25 animals) or the distal scape (42 animals) was always followed by distal regeneration. The small antenna present after the first moult was fairly normal in structure, but had a reduced number of flagellum segments and sometimes had abnormalities in the segmentation of the flagellum. In subsequent instars the regenerate developed a completely normal structure, including the distinct hinges between scape and pedicel.

Amputation in the antennal socket was also followed by regeneration (32 animals), while Urvoy (1963) found that removal of the socket plus the surrounding region of head capsule prevents regeneration. The antenna does not regenerate from an amputation surface in the flagellum, but increases the rate of formation of new segments from the third segment (Haas, 1955).

Grafts were performed between the right antennae of different animals of the same instar (grafts 1 and 4) or between left and right antennae of the same animal (grafts 2 and 3). The donor antenna was cut in the proximal third of the scape (see Fig. 1) and the graft was telescoped with the appropriate orientation into the host antenna amputated in the distal third of the scape. This choice of levels ensured a good fit and the graft was secured by dried haemolymph.

All 85 successful control grafts healed (often with a slight bulge) with no change in orientation of the graft and no formation of supernumerary structures (Fig. 2).

As shown in Fig. 3, the M/L axis of the grafted scape was reversed relative to that of the host and, in 46/51 successful cases, this orientation of the graft was retained. The remaining five mis-orientated grafts may have resulted from errors in grafting or from subsequent rotation, and they will be considered below, together with similar mis-orientated results from Graft 3. The 46 correctly orientated grafts produced two separate supernumerary antennae (40 cases) or one supernumerary double-antenna (6 cases).

Separate supernumeraries usually arose in M and L positions (33/40 cases) and their A/P axes were orientated like that of the host (Fig. 3B). In all but one case they originated from the level of the graft/host junction in the scape (often one was slightly more distal in origin than the other) and consisted of distal scape, pedicel and flagellum. Other positions of origin were AM and AL (3 cases), PM and PL (3 cases), and M and PL (1 case), and all these supernumeraries were also orientated like the host.

Supernumerary double-antennae usually arose from a P position (5/6 cases) on the graft/host junction. They were very large and were clearly double structures since, at the scape/pedicel articulation, they had two A hinges (in M and L positions relative to the host) and two P hinges (Fig. 3C). The remaining double-antenna arose from an A position and also had two hinges and two P hinges (in M and L positions), as shown in Fig. 3D.

As shown in Fig. 4, the A/P axis of the graft was reversed relative to the host and, in 67/77 successful cases, this orientation was retained. The remaining 10 mis-orientated grafts are considered separately below. The 67 A/P reversed grafts produced two separate supernumeraries (57 cases), one supernumerary (5 cases) or one supernumerary double-antenna (5 cases).

The two separate supernumeraries arose from the graft/host junction in A and P positions (53/57 cases) or in A and PM positions, and their A/P axes were orientated like that of the host (except for a few broken or rudimentary structures where the axis could not be identified). In the 5 cases where only one supernumerary was formed, this arose in a P position and had host orientation.

Supernumerary double-antennae usually arose from a M position (4/5 cases) and bore two A hinges (in distal MP and proximal MA positions) and two P hinges. The remaining double-antenna arose from a L position and also had two A hinges (in distal LP and proximal LA positions) and two P hinges (Fig. 4C).

The 15 mis-orientated cases from grafts 2 and 3 can be considered together. In 4 cases the A face of the graft was orientated medially, corresponding to a reversal of the MP/LA axis of the graft relative to the host. In all 4 cases separate supernumeraries of host orientation were formed in approximately MP and LA positions. In 11 cases the A face of the graft was orientated laterally, reversing the MA/LP axis of the graft. Eight of these grafts formed separate supernumeraries in approximately MA and LP positions while, in the remaining 3 cases, a double-antenna was formed in a MP position.

As shown in Fig. 5, both transverse axes of the graft were reversed relative to the host and, in all 29 successful cases, this orientation was changed by derotation of the graft. De-rotation in either clockwise or anti-clockwise direction brought the graft back into alignment with the host by the second post-operative moult. In 22/29 cases, graft and host levels healed together with the formation of no extra structures (Fig. 5B), but in 4 cases a partial supernumerary was formed, consisting of distal scape, articulatory membrane and a tiny lobe. The partial supernumeraries could arise from any circumferential position, could be identified in 2 cases as symmetrical structures (Fig. 5 C) and did not regenerate further in subsequent instars. In 2 cases graft and host levels independently regenerated distal structures and fused at the level of the pedical (Fig. 5D). The remaining grafted antenna formed supernumerary regenerates from A and P positions on the pedicel.

Grafts were performed between the leg tarsus and the antenna by telescoping the prothoracic tarsus (cut at proximal level) into the distal scape of the right antenna of the same animal. Grafts were done with control orientation (62 cases) 180° rotation (62 cases), or reversal of the M/L axis (62 cases) or the A/P axis (62 cases). At the first moult a few grafts were still attached by a thin neck of tissue to the host scape, but no graft survived to the second moult.

Grafts between the leg tibia and the antenna were performed by telescoping the prothoracic tibia (cut at proximal level) of a second-instar donor into the distal scape of the host animal. Of 80 grafts with control orientation, most were rejected by the first moult and only 4 were retained at the second moult. The graft remained more or less aligned with the host but was poorly healed. Grafts were done to reverse the A/P axis (160 cases) or the M/L axis (92 cases) of the graft, but all these grafts were rejected by the second moult.

The most obvious feature of the results of the antenna grafts is their similarity to the results of corresponding grafts done on the cockroach leg. While a control graft simply heals (graft 1), when a grafted scape is rotated 180° so that both transverse axes are reversed relative to the host (graft 4), it de-rotates, sometimes forming a symmetrical partial supernumerary at the graft/host junction. When one transverse axis of the grafted scape is reversed by a contralateral graft, supernumerary regenerates are reliably formed (grafts 2 and 3). Typically, a complete antenna regenerates from each of the points of maximum incongruity between graft and host, and its A/P axis (the only one which can be identified) is orientated like that of the host. Sometimes only one supernumerary structure is formed, midway between the points of maximum incongruity, but it is a double structure and its orientation suggests that it arises from secondary fusion of the two supernumeraries growing in the confined space beneath the old cuticle (French, 1976). This is illustrated for the M/L axis reversal in Fig. 6.

These results suggest that the spatial organization and the rules for cellular behaviour are similar in the insect leg and in the scape segment of the antenna. There is no evidence yet relating to the more distal antennal segments, but the different mode of growth and response to amputation suggest there may be different rules for cellular behaviour in the 3rd segment (or blastema) and the numerous segments of the flagellum.

Insects are metamerically segmented and are assumed to have evolved from annelid-like forms which were not divided into distinct head, thorax and abdomen, but had a sequence of body segments bearing similar jointed appendages (see Snodgrass, 1935; Manton, 1977). During evolution the appendages of different segments have been lost or adapted for locomotory, feeding or sensory functions. As pointed out by Wolpert (1971) and others, changes in morphology may occur because of changes in the stimuli (positional information, hormone concentrations, etc.) which developing cells receive from their surroundings, or changes in the ways in which cells respond to a set of stimuli.

Indirect evidence for the existence of similar sets of positional values in the insect antenna and leg comes from two situations where chimeric appendages are formed: heteromorphic regeneration of the stick insect antenna (Brecher, 1924; Urvoy, 1970) and homoeotic transformations in Drosophila (Postlethwaite & Schneiderman, 1971). Numerous grafts between pro- and metathoracic legs (Bohn, 1970,1972; Bullière, 1970; Bart, 1971 b;,French, 1980,1981) suggest that legs differ considerably in size and details of structure because of differences in cellular response, rather than positional information. A similar approach has reached the same conclusion concerning the different cuticular patterns formed on different abdominal segments (Stumpf, 1968).

The present axial grafts between the leg tarsus or tibia and the scape were an attempt to extend this approach. Healing of control grafts, derotation of rotated grafts and production of supernumeraries from A/P or M/L reversed grafts would have suggested that the same positional values run around leg and antenna. Unfortunately almost all grafts were rejected. It is a common finding in insect grafting experiments that epidermis from different body regions tends to reduce contact, isolating a graft from the supply of haemolymph and leading to graft death and rejection. The basis of this response is not understood: presumably it results from the surface properties of the different populations of cells (Nardi, 1977) but perhaps not those properties related to position within a segment or an appendage. If the cells will not heal together, they cannot interact to produce movement, growth or regeneration.

Cases of de-rotation (Bohn, 1965) and supernumerary formation (Urvoy, 1963) from .other rotated leg/antenna grafts suggest that the two appendages may be capable of interacting and may have common positional values. The same conclusion is also suggested by the results of grafts of the cricket cercus to the leg stump (Schubiger, personal communication; French, unpublished), where the graft is often rejected, but a rotated graft can de-rotate, and reversal of one transverse axis of the graft tends to produce supernumerary structures.

This work is supported by the Science Research Council.