Ommatidium Assembly and Formation of the Retina-Lamina Projection in Interspecific Chimeras of Cockroach

/. Embvycl. exp. Morph. Vol. 60, pp. 345-358, 1980 345 Printed in Great Britain © Company of Biologists Limited 1980

Ommatidium assembly and formation of the retina-lamina projection in interspecific chimeras of cockroach

By MARK S. NOWEL1 From the Department of Zoology, University of Leicester

SUMMARY By grafting operations, interspecific eye chimeras of the cockroaches Gromphadorhina portentosa andLeucophaea maderae were produced. Mechanisms involved in the development of both the compound eye and the retina-lamina projection have been studied. Most cell types composing the eyes of these cockroaches are cytologically distinguishable in the chimera; also, retinula axons forming the retina-lamina projection in the two species are of vastly different lengths. At the border between host and graft eye tissue, individual ommatidia are formed containing cells of both types. In particular, it is shov/n that the four cone cells can be found in any of the possible combinations of the two cell types. This shows that the cone cells within one ommatidium are not necessarily related by lineage. These results are in agreement with the hypothesis that cells within an ommatidium are determined by position rather than by a lineage mechanism. Furthermore, formation of mosaic ommatidia suggests that mechanisms governing eye formation are similar in these two species. The formation of the projection from donor retina to host lamina shows that axon elongation is not rigidly programmed, but that the axons grow until they reach a suitable target at which point connexions are made.

INTRODUCTION By generating interspecific chimeras it is possible to investigate the question of whether pattern-forming mechanisms in different organisms are similar. In this study two species of cockroach, Gromphadorhina portentosa and Leucophaea maderae, have been used to create chimeric compound eyes. This experimental material enables us to investigate the specific question of cell lineage in development, because the eye cells in the two species are cytologically identifiable. It also enables us to ask the general question of whether or not ommatidia in different genera are formed by the same mechanisms. Finally, the compound eye as a component of the nervous system has enabled us to examine questions concerning nerve growth in a chimera. Concerning the question of cell lineage, it has been possible to confirm the observations of previous workers that there is no causal relationship between lineage and determination in the formation of ommatidia (Yagi & Koyama, 1 Author's address: Department of Zoology, University of Leicester, Leicester LEI 7RH, U.K. 346 M. S. NOWEL 1963; Hanson, Ready & Benzer, 1972; Ready, 1973; Benzer, 1973; Shelton & Lawrence, 1974; Green & Lawrence, 1975; Ready, Hanson & Benzer, 1976; Nardi, 1977; Shelton, Anderson & Eley, 1977; Lawrence & Green, 1979). In particular, cone-cell lineages have been examined. At the borders of graft and host tissues, ommatidia are found containing cells of both genotypes. The two types of cells behave autonomously and mosaic ommatidia have normal numbers of cells even though some are of one genotype and some are of the other. Thus, whichever pattern-forming mechanism is responsible for ommatidium assembly, it is the same in these different genera. Concerning nerve growth, the mode of establishment of connexions between the eye and optic lobe has been examined. In the two species, retinula axons growing into the optic lobe have to grow different distances to form their connexions. In an interspecific chimera, L. maderae retinal tissue is three times farther away from its target, the lamina of its G. portentosa host, than it would be from its own lamina in the unoperated situation. Nevertheless, retinula axons from the foreign retina can form connexions in the lamina of the under- lying optic lobe. This provides valuable information on factors governing nerve growth and target discrimination.

MATERIALS AND METHODS Cultures of G. portentosa and L. maderae were maintained under conditions of constant temperature (24 °C) and an alternating cycle of 12 h light/12 h dark, and fed on a diet of rat pellets and water. Grafts of eye margin and adjacent vertex epidermis were exchanged between young, newly moulted G. portentosa and L. maderae nymphs. The animals were anaesthetized by cooling in ice for 10-20 min and restrained with strips of Plasticine on a bed of moulded Plasticine. Operations were carried out under a dissecting microscope. Excisions of the integument were made using a razor blade fragment (Gillette francais) supported in a pin vice, and grafts were transferred from donor to host site using tungsten needles. After positioning the graft in the host site which had been kept moist with insect saline (Hoyle, 1953), a small droplet of insect wax (Krogh & Weis-Fogh, 1951) was used to seal the operation site. Operated animals were allowed to grow to the imago. To map areas of neural projection from graft-derived ommatidia in the chimeric eye into the lamina neuropile of the optic lobe, small localized lesions were made within the graft-derived eye tissue of the adults. After immobilizing the anaesthetized cockroach as for a grafting operation, a silver earthing wire was inserted into the head through a hole cut in the cuticle at a posterior medial point. A tungsten microelectrode was connected to a function generator set to deliver 2 /iA at a frequency of 1 MHz. After removal of narrow strips of the overlying cornea, the microelectrode was inserted into the exposed ommatidia to be electrolytically destroyed, and left in each point of insertion for 10 sec. The Cockroach compound-eye development 347 wounds were covered with insect wax. Sixteen hours later the animal was killed and the eye and optic lobe were fixed in a mixture of glutaraldehyde and paraformaldehyde (Karnovsky, 1965) in a 0-1 M-phosphate buffer (Hayat, 1970) at pH 7-4 for 2-4 h. The tissue was post-fixed in phosphate-buffered 1 % osmium tetroxide for 2-12 h, then dehydrated in an acetone series, cleared in propylene oxide, and embedded in Spurr's resin following a long period of infiltration. The optic lobe was serially sectioned in 1 ^m horizontal sections with a Huxley Ultramicrotome and glass knives, mounted in order on subbed slides, stained with toluidine blue, and examined with a Zeiss compound microscope for degenerating retinula axon terminals which appear dark blue following such treatment (Geisert & Altner, 1974). Chimeric eyes were sectioned perpendicular to the ommatidial long axis at the graft/host border. Serial semithin (1 ju,m) sections were collected in order on subbed slides and stained with toluidine blue. Ultrathin (80-100 nm) sections cut on a Dupont diamond knife were collected on collodion films (Pease, 1964), placed on to slot grids, stained with uranyl acetate and Reynolds' (1963) lead citrate, and examined with an AE1-802 Electron Microscope or a Siemens 102 Elmiskop. For wax histology, quarter heads were fixed for at least 3 h in alcoholic Bouin's fluid (Dubosq-Brasil) (Pantin, 1969) and left in 70 % isopropyl alcohol for periods of several days to several weeks to wash out the fixative and to soften the cuticle. After dehydration and embedding in paraffin wax, 10 /*m thick horizontal sections were cut on a Cambridge rocking microtome, dried on to albuminized slides and stained using a modification of the Mallory- Azan technique (Schiimperli, 1977). Experimental animals were photographed using a Zeiss Tessovar Photo- macrographic Zoom system. Sectioned material was photographed on a Zeiss Photomicroscope II. RESULTS Identification of cell genotypes in G. portentosa and L. maderae Sections through ommatidia in adult G. portentosa and L. maderae at the level of the crystalline cones are shown in Figs. 1 and 2. Cone cells in G. portentosa have a denser granularity and are considerably larger than those of L. maderae: crystalline cones (composed of four Semper's or cone cells) in G. portentosa measure approximately 35-40 pm in diameter at their bases with a base to apex length of 60 [im, while those of L. maderae measure 25 /im in diameter with a base to apex length of 35-40 /*m. In the chimeric situation (Figs. 5-11) the relative sizes of not only the cone cells (Figs. 5-8) but also the pigment granules which provide distinguishing features between both the retinula and primary pigment cells of the two species (Figs. 7, 10) are easily seen. In the primary pigment cells, pigment grains of G. portentosa are large 348 M. S. NOWEL

FIGURES 1 AND 2 Semithin sections through the compound eyes of G. portentosa (Fig. 1) and L. maderae (Fig. 2) cut perpendicular to the ommatidial long axis at the level of the crystalline cones (cc). The Semper's cells comprising the crystalline cones may be distinguished on the basis of their relative sizes: cones of G. portentosa are composed of four large cells, while those of L. maderae are made up of four smaller cells. Each crystalline cone is surrounded by two primary pigment cells (pp) and numerous secondary pigment cells (sp). (The basis of the differential staining of cone cells within the same cone, often observed, is not understood.) Bars represent 10/*m. Cockroach compound-eye development 349 (1-3-1-7/mi) while those of L. maderae are slightly smaller (1-0-1-2/mi). Pigmentation of the retinula cells is similarly distinguishable: larger (0-7- 1-0 /mi) pigment grains in G. portentosa cells, smaller (0-35-0-5 /mi) grains in those of L. maderae. Secondary pigment cells are not easily distinguishable.

Interspecific mosaic ommatidia Of approximately 400 grafting operations, approximately 20 resulted in interspecific chimeras. Following such operations, L. maderae eye tissue appears to grow at a slightly faster rate and G. portentosa at a slightly slower rate than does their host eye tissue. This gives the compound eye of each sort of chimera a characteristic appearance (Figs. 3, 4). Three chimeric eyes containing G. portentosa and L. maderae ommatidia were sectioned and examined for the presence of ommatjdia composed of cells of both sources, especially those whose crystalline cones were mosaics of cells of the donor and host. Such ommatidia were common and readily identifiable at the border between graft- and host-derived eye tissue (Fig. 5). Several mosaic cones of the three possible combinations of constituent cells (one G. portentosa cell with three L. maderae cells; one L. maderae with three G. portentosa cells; or two cells from each source) were found (Figs. 6-8). Mosaic ommatidia appear at random intervals along the graft/host border, and occasionally two are found alongside each other. It is concluded that there is no fixed clonal relationship between cells of a single crystalline cone. Mosaic ommatidia in which each of the two primary pigment cells is from a different source (one exhibiting donor-specific pigmentation and the other exhibiting host-specific pigmentation) are found (Fig. 7). Ommatidia in which the constituent retinula cells come from different sources are also seen (Figs. 5, 9). In Fig. 9 it can be seen that a single retinula cell has G. portentosa-spQcific characteristics while the other seven retinula cells (as well as all four Semper's cells and both primary pigment cells) are from L. maderae. These observations confirm the conclusions of previous investigators that retinula cells and primary pigment cells within a single ommatidium are not determined by sequence of determinative cell divisions, and argue in favour of lineage independence of

FIGURES 3-5 (see page 350). Fig. 3. Interspecific chimeric eye of a G. portentosa (G) host with a graft of L. maderae (L.) eye margin and head epidermis. Bar represents 0-25 mm. Fig. 4. Interspecific chimeric eye of a L. maderae (L) host with a graft of G. portentosa (G) eye margin and head epidermis. Bar represents 0-25 mm. Fig. 5. Semithin section through the graft/host border of a chimera similar to that shown in Fig. 3. G. portentosa (G) host ommatidia have very large cones, while L. maderae (L) graft ommatidia have small cones. Note two ommatidia (arrows) with mosaic crystalline cones (shown again in Figs. 6 and. 7); * indicates an ommatidium (shown again in Fig. 9) with a mosaic sensory retinula. Bar represents 10 /im. 23 EMB 60 350 M. S. NOWEL

& M

For legend see page 349. FIGURES 6-8 Three electron micrographs showing mosaic cones in interspecific chimeras. Such mosaic cones show that neither the ommatidium nor the crystalJine cone is produced from a single ommatidial or cone mother cell. Bars represen t5 /*m. Fig. 6. The crystalline cone is composed of one G. portentosa cell (G) and three L. maderae cells (L). Fig. 7. The crystalline cone is composed of two G. portentosa cells (G) and two L. maderae cells (L). In addition, each of the two primary pigment cells comes from a different source (G and L). Fig. 8. The crystalline cone is composed of three G. portentosa cells (G) and one L. maderae cell (L). 352 M. S. NOWEL

r/7

1CL Cockroach compound-eye development 353 any particular cell component of an ommatidium from other cells within that structure.

Projection of graft-derived retinula axons to host lamina in interspecific chimera In the adult cockroach, the projection of axons from the basement membrane of the retina to the outer optic anlage is approximately 0-5-0-6 mm in G. portentosa, and 0-2 mm in L. maderae (Figs. 12, 13). These differences are apparent during larval development at the time that the connexions are origin- ally established. At the larval stages during which the operations were per- formed, the separation between retina and outer optic anlage was approximately 0-2 mm in G. portentosa and 0-1 mm in L. maderae. Following interspecific transplant operations, if G. portentosa retinula axons are to make connexions in the lamina neuropile of their L. maderae hosts, they must be able to stop elongating after growing only one half to one third of their normal length. For retinula axons of L. maderae to make connexions in the lamina of their G. portentosa hosts, they must double or treble the distance they have to grow before reaching the target outer optic anlage. Of three G. portentosa eyes (with a graft of L. maderae tissue) examined, one showed degenerating retinula terminals in the host lamina neuropile following microcautery of a set of graft-derived ommatidia (Fig. 14). Of seven L. maderae eyes containing grafts of G. portentosa tissue which were examined in the same way, three showed degenerating retinula axon terminals in the host lamina neuropile (Fig. 15). DISCUSSION Ommatidium assembly Bernard (1937) had proposed that all cells forming a particular ommatidium were the division products of a single ommatidium mother cell. The presence of mosaic ommatidia disproves this mechanism's involvement in compound-eye development.

FIGURES 9-11 Fig. 9. Electron micrograph of a mosaic ommatidium lying on the graft/host border of an interspecific chimera, taken at the level of the sensory retinula. There is the normal compliment of eight retinula cells (numbered arbitrarily). The larger pigment grains of the host G. portentosa retinula cell (6) clearly distinguishes it from the seven graft-derived L. maderae cells (1-5, 7, 8). rh rhabdom. Fig. 10. Electron micrograph through an interspecific mosaic ommatidium at the level of the sensory retinula. The retinula cell of G. portentosa (G) origin has larger pigment grains than those of L. maderae (L). Note the desmosome (arrow) joining the adjacent cells of different specific origin at the periphery of the rhabdom (rh). cp, Cone-cell process. Fig. 11. Electron micrograph through an interspecific mosaic cone. Cone cells of L. maderae (L) are smaller with a denser granularity than those of G.portentosa (G). pp, primary pigment cell. Bars represent 2/*m. 354 M. S. NOWEL

^V 'r »VH! Cockroach compound-eye development 355 Lawrence & Green (1979) have generated clones of red-pigmented cells in white Drosophila eyes. Extremely small clones (even clones composed of only two scorable cells) can contain both pigment and retinula cells. It is clear that these different cell types composing the ommatidium are not derived by separate cell lineages and determinative divisions. As a result of the present experiments, which include the cone cells as well as the pigmented retinula and primary pigment cells, it may now be said that any ommatidial cell may be unrelated by lineage to any other cell component of that ommatidium. This reaffirms the hypothesis that spatial considerations, or the position of cells within the developing ommatidium, are determinative in the generation of particular cell types (Shelton & Lawrence, 1974; Lawrence & Green, 1979). How cell position within a developing pre-ommatidial cell cluster is assessed is a matter for speculation. Certainly, information exchange is possible at the time of cluster formation when gap junctions are widely distributed within the undifferentiated cells of the eye margin of the locust (Eley & Shelton, 1976). Another suggestion is that particular retinula cells are determined with respect to their positional relationship with the five cone-cell processes in the locust (Wilson, Garrard & McGinness, 1978). The ability of a group of cells coming from two different genera to interact and form a perfect ommatidium (four cone cells, two primary pigment cells, eight retinula cells, plus secondary pigment cells) argues in favour of the similarity or equivalence of the ommatidium-forming mechanisms in these two different insects.

FIGURES 12-15 Fig. 12. Horizontal section through the compound eye (ce) and optic lobe of an adult L. maderae. Note the relatively short (approximately 0-2 mm) retina-lamina projection of retinula axons (ra). In, Lamina neuropile; mn, medulla neuropile. Bar represents 01 mm. Fig. 13. Horizontal section through the compound eye (ce) and optic lobe of an adult G. portentosa shown at the same magnification as Fig. 12. Note the relatively long (approximately 0-5-0-6 mm) retina-lamina projection of retinula axons (ra). In, Lamina neuropile. Bar represents 01 mm. Fig. 14. Lamina neuropile (In) of G. portentosa adult in which there was a graft of L. maderae tissue. Following the introduction of a small lesion in the grafted eye tissue, the appearance of degenerating terminals (arrow) in the lamina indicates that axons project to the lamina from the (wounded) graft ommatidia. Bar represents 10/un. Fig. 15. Lamina neuropile (In) of L. maderae adult in which there was a graft of G. portentosa tissue. Following the introduction of a small lesion in the grafted eye tissue, the appearance of degenerating terminals (arrow) in the lamina indicates that axons project to the lamina from the (wounded) graft ommatidia. Bar represents 10/mi. 356 M. S. NOWEL

Axon projection Cotman & Banker (1974) describe synapse formation as a two-step process: axon elongation to approach the general vicinity of its set of target cells followed by the establishment of synaptic contacts with a limited number of cells within this region. The signal to the growing nerve fibres to stop elongating and form connexions poses an interesting problem of developmental mechanisms: are fibre tips directed to stop at a particular site and there form connexions, or do they grow until they make connexions and then stop elongating? Swisher & Hibbard (1967) have removed the target sites of Mauthner's fibres in Xenopus embryos by removing the tails of two individuals and grafting them together with their heads pointing in opposite directions. The Mauthner's fibres pass caudally down the spinal cord of one embryo and, not encountering their sites of termination (in the excised tail), continue into the spinal cord of the grafted embryo rostrally into the second brain. Altman's (1972, 1973) studies on the rat cerebellar cortex show that if the normal migration of target neurons is retarded, projecting fibres are capable of further elongation. The fibres bypass other (inappropriate) neurons without making extensive synaptic contacts with these cells, and terminate on their appropriate target cells. Having made these proper connexions, both the elongation of the projecting axons and the migration of the target cells cease. In the present studies, it is clear that the developmental programme for the advancing retinula axons of the cockroach does not include instructions to grow a particular distance and then form connexions. Rather, the variations introduced by experimental manipulations of the projection distance indicate that fibres grow until they reach cells which are recognized as prospective target sites, possibly bypassing inappropriate (i.e. already-differentiated) lamina ganglion cells to reach the target, the outer optic anlage. Apparently these ganglion cells are deemed unsuitable despite their location in a more typical projection distance for the presynaptic axons. In Daphnia, growing retinula tips bypass differentiated ganglion cells to reach the proliferation zone of the lamina (LoPresti, Macagno & Levinthal, 1973). Once suitable target cells are contacted, synaptogenesis begins. Several things remain unknown: (a) How adjustable is the axon's capacity for elongation ? What restraints are imposed on the extent of axon elongation ? (b) What is the nature of the signal to stop fibre elongation ? Is it the presence of the target cell itself, or is it some property associated with the outer optic anlage ?

CONCLUSIONS In hemimetabolous insects, new retinal cells are generated by cell proliferation within a budding zone of the eye margin (Nowel & Shelton, 1980). The results of the present experiments, namely (a) the generation of inter-specific chimeric Cockroach compound-eye development 357 retinae; (b) the formation of mosaic ommatidia; (c) the formation of mosaic crystalline cones; (d) the formation of mosaic pigment cell sleeves, and (e) the formation of mosaic sensory retinulae, including an example in which only one retinula cell differs in origin from the other seven as well as from all the other scored cells in the ommatidium indicate that new ommatidia - and also parts of new ommatidia - are not determined by a cell-lineage mechanism but by some sort of interactive mechanism within the cell clusters forming at the margin of the compound eye. This leads to determination and differentiation of specific cell types making up the ommatidium. With regard to axon projection to its target, the present experiments allow certain conclusions to be drawn: (a) Axons from receptor cells of one species of cockroach can grow into the optic lobe of a host animal of a different species; they can apparently form connexions with second-order cells and otherwise be integrated into the host neuropile. Under normal circumstances, interspecific connexions between these receptor and second-order cells would never be made, and species-specific cell surface properties must exist. Nevertheless, successful establishment of connexions can occur between cells of different specific origins. Presumably the affinities of retinula terminals for lamina ganglion cells override any interspecific differences. (b) It is significant that the graft retinula cells allow apparently normal development of the lamina (Figs. 14, 15). This implies that the stimulus for the development of lamina ganglion cells and neuropile are similar in different species. (c) Retinula neurons of G. portentosa are able to curtail axon extension prematurely to form connexions with second-order cells, apparently upon reaching a particular signal which is met sooner in this experimental situation than in a normal situation. {d) Retinula neurons of L. maderae are able to grow further than their normal programme of growth would have allowed. In the experimental situation (in which they are growing in a G. portentosa host), axons apparently receive a particular signal to cease elongation and begin synaptogenesis later than in a normal situation.

My sincere thanks go to Dr Peter M. J. Shelton for supervising this work and for his encouragement. We are grateful to the Science Research Council for its grant to P. M. J. Shelton.

REFERENCES ALTMAN, J. (1972). Postnatal development of the cerebellar cortex in the rat. III. Maturation of the components of the granular layer. /. comp. Neurol. 145, 465-514. ALTMAN, J. (1973). Experimental reorganization of the cerebellar cortex. III. Regeneration of the external germinal layer and granule cell ectopia. /. comp. Neurol. 149, 153-180. BENZER, S. (1973). Genetic dissection of behaviour. Scient. Am. 229 (12), 24-37. 358 M. S. NOWEL BERNARD, F. (1937). Recherches sur la morphogenese des yeux composes d'arthropodes. Bull. biol. Fr. Belg. Suppl. 23, 1-162. COTMAN, C. W. & BANKER, G. A. (1974). The making of a synapse. In Reviews of Neuro- science, vol. I (ed. S. Ehrenpreis & I. J. Kopin), pp. 1-62. New York: Raven Press. ELEY, S. & SHELTON, P. M. J. (1976). Cell junctions in the developing compound eye of the desert locust Schistocerca gregan'a. J. Embryol. exp. Morph. 36, 409-423. GEISERT, B. & ALTNER, H. (1974). Analysis of the sensory projection from the tarsal sensilla of the blow-fly (Phormia terraenovae Rob.-Desv., Diptera). Experimental degeneration of primary fibers. Cell Tiss. Res. 150, 249-259. GREEN, S. M. & LAWRENCE, P. A. (1975). Recruitment of epidermal cells by the developing eye of Oncopeltus (Hemiptera). Wilhelm Roux1 Archiv. devl Biol. Ill, 61-65. HANSON, T. E., READY, D. F. & BENZER, S. (1972). Use of mosaics in the analysis of pattern formation in the retina of Drosophila. Rep. Div. Biol. Cal. hist. Tech. 37, 40. HAYAT, M. A. (1970). Principles and Techniques of Electron Microscopy: Biological Application, vol. i, pp. 342-343. London: Van Nostrand Reinhold. HOYLE, G. (1953). Potassium ions and insect nerve muscle. J. exp. Biol. 30, 121-135. KARNOVSKY, M. J. (1965). A formaldehyde-glutaraldehyde fixative of high osmolality for use in electron microscopy. /. Cell Biol. 27, 137 A. KROGH, A. & WEIS-FOGH, T. (1951). The respiratory exchange of the desert locust {Schisto- cerca gregan'a) before, during and after flight. /. exp. Biol. 28, 344-357. LAWRENCE, P. A. & GREEN, S. M. (1979). Cell lineage in the developing retina of Drosophila. Devi Biol. 11, 142-152. LOPRESTI, V., MACAGNO, E. R. & LEVINTHAL, C. (1973). Structure and development of neuronal connections in isogenic organisms: cellular interactions in the development of the optic lamina of Daphnia. Proc. natn. Acad. Sci. U.S.A. 70, 433-437. NARDI, J. B. (1977). The construction of the insect compound eye: The involvement of cell displacement and cell surface properties in the positioning of cells. Devi Biol. 61, 287-298. NOWEL, M. S. & SHELTON, P. M. J. (1980). The eye margin and compound-eye development in the cockroach: evidence against recruitment. /. Embryol. exp. Morph. 60, 329-343. PANTTN, C. F. A. (1969). Notes on Microscopical Technique for Zoologists. Cambridge Univer- sity Press. PEASE, D. C. (1964). Histological Techniques for Electron Microscopy. London: Academic Press. READY, D. F. (1973). Pattern formation in the retina of Drosophila. Rep. Div. Biol. Cal. hist. Tech. p. 203. READY, D. F., HANSON, T. E. & BENZER, S. (1976). Development of the Drosophila retina, a neurocrystalline lattice. Devi Biol. 42, 211-221. REYNOLDS, E. W. (1963). The use of lead citrate at high pH as an electron-opaque stain in electron microscopy. /. Cell Biol. 17, 208-212. SCHUMPERLI, R. A. (1977). Adaptation of the Mallory-Azan staining method to insect nervous tissue. Stain Technol. 52, 55-56. SHELTON, P. M. J., ANDERSON, H. J. & ELEY, S. (1977). Cell lineage and cell determination in the developing compound eye of the cockroach, Periplaneta americana. J. Embryol. exp. Morph. 39, 235-252. SHELTON, P. M. J. & LAWRENCE, P. A. (1974). Structure and development of ommatidia in Oncopeltus fasciatus. J. Embryol. exp. Morph. 32, 337-353. SwrsHER, J. E. & HIBBARD, E. (1967). The course of Mauthner axons in Janus-headed Xenopus embryos. /. exp. Zool. 165, 433-439. WILSON, M., GARRARD, P. & MCGINNESS, S. (1978). The unit structure of the locust compound eye. Cell Tiss. Res. 195, 205-226. YAGI, N. & KOYAMA, N. (1963). The Compound Eye of Lepidoptera. Tokyo: Maruzen.

{Received 15 May 1980, revised 18 June 1980)