Molecular Biology Unit, Tata Institute of Fundamental Research, Homi Bhabha Road, Navy Nagar, Colaba,
Bombay 400 005, India

†Corresponding author

We studied the metamorphosis of the central nervous system (CNS) and neighbouring muscles of Drosophila melanogaster (Diptera: Drosophilidae) during pupation by transmission electron microscopy (TEM). The age of white pupa was assumed to be 0 h and the process of metamorphosis was monitored, onward between 6 and 96 h at 25°C. The profiles in the neuropil showed degeneration at 6 h and its extent increased by 12 h. The presence of glycogen in some of these profiles indicated their larval character. Between 12–18 h, the neuronal profiles became separated from one another, the intervening space was filled with extracellular fluid, and some of the larval synapses degenerated. Synaptic vesicles started reappearing around 18 h and synapses were detectable by 24 h. Neuronal processes compactly filled the neuropil by 65 h and the maturation of synapses continued until 86 h. The degeneration of profiles in the neuropil was found to be bimodal, peaking at 12 and 42 h, and that of cortical cells was unimodal with a peak at 42 h. The number of neuronal profiles increased with the development time, indicating that more branching of neuronal profiles occurs in neuropils as the metamorphosis progresses. Average number of synapses per unit area (or volume) is minimum at 18 h and maximum at 72 h, when the average number of synapse per axon profile is 0·54. Because 2 axon profiles share one synapse, a value close to 0·5 for monad synapses shows that, on an average, each axon profile at least makes one synapse at this stage of development. Subsequently, there is more than 75% of reduction in the number of synapses during 73 and 78 h.

In muscles, vacuoles suggesting histolysis appeared by 6 h. Their ultrastructure became deranged between
12–18 h and myoblasts were found to be present since 8 h. Except for a few muscles in the thorax, such as larval oblique muscles and pharyngeal muscles, most of the muscles in the head and thorax lost all the ultrastructural details and histolyzed by 18 h. Around 38 h, imaginal muscles were detectable, and well-developed muscles were found by 55 h. However, myofibrils continued to be added laterally to the preformed muscles even at 96 h.

Electron-dense mitochondria (EDMITs) were found in the neuropil, cortex and muscles of pupa, along with mitochondria of characteristic shape and normal appearance. These EDMITs often occurred in large clusters of more than 100, at times near the surface of the tissue. A few of these were enclosed in vacuoles and were darker than the rest of the EDMITs and normal looking mitochondria. Histochemistry with diaminobenzidine showed the presence of cytochrome c and marker enzyme cytochrome oxidase, both in EDMITs and normal mitochondria. EDMITs were not found to be present in any tissue of the adult Drosophila.

Compared with hemimetabolous insects in which the major features of the building plan of the adult central nervous system (CNS) are laid down during embryogenesis (Malzacher 1968; Edwards 1969; Bate 1976; Doe and Goodman 1985), a holometabolous insect, such as Drosophila, has 2 prominent phases of neurogenesis –embryonic (Poulson 1950; Hartenstein and Campos-Ortega 1984; Campos-Ortega and Hartenstein 1985; Canal and Ferrus 1986; Hartenstein et al 1987; Jacob and Goodman 1989a, b), and postembryonic (Bodenstein 1950; Lawrence 1973). Further, in postembryonic period, 2 distinct phases of development of CNS exist (i) larval (Sonnenblick 1950; White and Kankel 1978; Thomas et al 1984; Truman and Bate 1988; Hartenstein 1988; Singh et al 1989; Prokop and Technau 1991), and (ii) post-larval, where metamorphosis of the CNS occurs during pupation leading to early adult stage of the fly (Hertweck 1931; Robertson 1936; Power 1943, 1948, 1952; Zalokar 1947; Bodenstein 1950; Kankel et al 1980; Technau and Heisenberg 1982; Truman 1990). Extensive studies have been done on the development of eyes (Lees and Waddington 1942; Power 1943; Meinertzhagen 1973; Kankel and Hall 1976; Ready et al 1976; Campos-Ortega and Hofbauer 1977; Kankel et al 1980; Zipursky et al 1984; Tomlinson and Ready 1987), and the peripheral nervous system of Drosophila (Bate 1978; Murray et al 1984; Hartenstein 1988; Singh et al 1989; Ghysen and Dambly-Chaudière 1990; Giangrande and Palka 1990) using light and electron microscopy. Except for the studies of White and Kankel (1978), Technau and Heisenberg (1982), and Truman (1990) the metamorphosis of the CNS in Drosophila during pupation has been rather neglected. Some prominent reasons for the lack of attention to the development of CNS during pupation could be that, right from the beginning, the puparium has a non-transparent exocuticle that is impervious to most of the aqueous fixative solutions. The high tergal pressure inside the puparium, is an added disadvantage, as the inner contents are likely to ooze out through any hole or slit made in the cuticle, for the facilitation of the entry of fixative. Because of these reasons, direct observation of the pupa by light microscope or the preservation of tissue structures for electron microscopy without recourse to injection of fixative through glass microcapillary electrode, becomes difficult.

We have felt the need to follow the development of the CNS and related structures such as glia (Fredieu and Mahowald 1989) and the neighbouring muscles (Sink and Whitington 1991), by transmission electron microscopy (TEM). TEM by virtue of the higher magnification compared with the light microscopy, was in our view better suited for monitoring the process of morphological development of the CNS and the neighbouring muscles during pupation in Drosophila. Although, at higher magnifications, the field of view becomes smaller in TEM this was suitably overcome by the light microscopic examination of 1 m m thick sections for a general view of the morphological development.

At the start of metamorphosis the larval CNS has both larval neurons and arrested imaginal neurons. From the patterns of neurogenesis the thoracic neuropil is likely to be enriched for imaginal neurons, whereas the posterior abdominal neuromere would consist mainly of larval processes (Truman 1990). It would be interesting, if it was possible to discriminate between these two types of processes. However, these processes could not be distinguished from each other by TEM.

The age of the white pupa was assumed to be 0 h. The development process was monitored from 6 to 96 h after white pupa formation (APF) at 4 to 8 (usually 6) hourly intervals at 25°C using mainly TEM. These observations were supplemented with light microscopic examinations.

Clusters of EDMITs, considered to be degenerating mitochondria, were discovered in the nervous tissue and muscles during pupation. These EDMITs were absent in the adult Drosophila tissues.

2. Materials and methods

2.1 Development of pupa at controlled temperature

Five to 6 day-old D. melanogaster Canton special (CS) flies were allowed to lay eggs at 25°C on Drosophila food medium containing yeast. The first 2 lots of eggs during first 2 h were discarded. Subsequent batches of eggs were harvested at hourly intervals on fresh food surface containing yeast. Eggs and the emerged larvae were allowed to grow at 25°C and the time of white pupa formation was regarded as 0 h for the sample. The white pupae were isolated with a soft hair-brush and grown for different lengths of time at 25°C in vials containing a piece of wet filter paper. The temperature during all these operations and especially during development was maintained at 25°C ± 1°C.

2.2 Transmission electron microscopy of
Drosophila pupae

At the end of the desired time, individual pupa was submerged under a drop of fixative (Karnovsky 1965). After nicking its anterior and posterior ends, the pupa was transferred to a vial containing approximately 5 ml of fixative and the fixation was done for 6 h at room temperature (RT). The exocuticle of the pupa was removed after 2 h of fixation in samples older than 18 h, whereas in younger samples the exocuticle was removed after full length 6 h of fixation. After several washings with 0·1 M sodium phosphate buffer of pH 7·4 during an hour, the specimens were postfixed with Dalton’s chrome-osmium tetroxide fixative (Dalton 1955) on ice for 2 h followed by 1 h fixation at RT. By this time, the entire specimen turned dark brown. The specimens were washed with sodium phosphate buffer, dehydrated with increasing concentrations of ethyl alcohol, followed by propylene oxide, infiltrated with Durcupan ACM (Fluka), oriented and embedded in the same resin mixture. Polymerization of the resin mixture was done at 60°C overnight. Silvery ultrathin sections were cut on a ‘LKB Ultrotome III’, mounted on ‘Formvar’ coated grids, and stained with a saturated aqueous solution of uranyl acetate at 60°C for 10 min, washed with water and stained with lead citrate at RT for 5 min (Reynolds 1963). Subsequent examination of sections was done with a ‘Jeol JEM 100 S’ electron microscope. Approximately 2800 photographs were taken for these studies. Comparative light microscopic observations were made with a ‘Zeiss Universal RS III’ or a ‘Leitz Orthoplan’ photomicroscope with plan-apo objectives.

2.3 Histochemistry of cytochrome c and cytochrome oxidase in normal and electron-dense mitochondria

Cytochrome oxidase is one of the several marker enzymes among those present in normal mitochondria (Lehninger 1990). Seligman et al (1968), Reith and Schüler (1972), and Crammer and Moore (1973) clearly demonstrated that diaminobenzidine (DAB) can donate electrons to the oxidized form of cytochrome c in mitochondria. The reduced form is reoxidized by cytochrome oxidase in the presence of oxygen. Thus, the DAB localizes cytochrome oxidase and cytochrome c both inside the mitochondria.

Sixty-five hour-old pupae were cut lengthwise into 2 approximately equal parts. The cut end of right half was immersed for half an hour in a 30% solution of horseradish peroxidase (Sigma type VI) dissolved in water. Subsequently, the tissue was fixed with Karnovsky’s fixative for 4 h at RT. The tissue was washed several times with 0·2 M sodium cacodylate buffer of pH 7·2. The DAB reaction was carried out on the tissue in 0·05 M TRIS buffer of pH 7·6 according to Nässel (1983) followed by post fixation with 0·5% osmium tetroxide in 0·1 M cacodylate buffer for 1·5 h on ice. Further processing of the tissue and its TEM was done as described above.

3. Results

TEM observations on the brain, ventral or thoracic ganglia and nearby muscles were recorded.

3.1 Metamorphosis of neuropils

In neuropils of brain and thoracic ganglia, signs of degeneration such as electron-dense profiles, pseudo-myelin figures and abnormal vacuoles, are predominantly present at 6 h and subsequent times until 50 h (figures 1A–F, 3A, B). The presence of glycogen until approximately 12 h in some of the profiles in the neuropil, indicates their larval character (Singh et al 1991), (figures 1A, B, arrowhead). In the neuropil, glial profiles by being more electron-dense than the neuronal profiles, are easily distinguishable from each other by TEM. Neuronal profiles start separating from each other by 12 h and the intervening space so created is filled by extracellular fluid and vacuoles (figures 1C–E; 3B–D, arrow). Vacuoles are significantly present in the neuropil until 55 h (figures 1F, 2A, 3E marked v). The neuropil becomes compactly filled with neuronal profiles by 72 h (figure 2C). The larval synapses decrease in number during 12 to 18 h (figures 1A–C; 3A–C; 4C).

Neuronal growth cones and synaptic vesicles start appearing by 18 h (figure 1C, black star). Fasciculation of neuron profiles (figures 1D; 3D, marked f) and the formation of imaginal synapses were first detectable by 24 h (figures 1D; 3D, arrowhead). Morphological maturation of synapses is attained over a long period of time between 24 and 92 h (figures 2A–F; 3D–F). Around 65 h, synapses appear as large contiguous electron-dense regions (figures 2B, D; 3F, arrowhead) where only a few synaptic vesicles were visible. As the development progresses, the electron-dense material along the synaptic region condenses into discrete synaptic patches (figure 2C, E, arrow) and the synaptic vesicles were prominently visible. By 82 h, synapses appear fully mature with electron-dense synaptic thickenings and synaptic vesicles (figure 2F). At least 3 morphological types of synaptic vesicles were detectable: (i) electron-lucent vesicles (25–60 nm in diam), (ii) electron-dense vesicles (50–80 nm in diam) and (iii) dense core vesicles (80 nm in diam). In addition, neurosecretory granules were found in some neuron profiles. Neurosecretory profiles were found to be present throughout the period of metamorphosis (figures 1E; 3A–C, marked s).

Estimates of degenerating neuron profiles in the neuropil and degenerating cells in the cortex were made by scoring them on electron micrographs covering 50,000 sq. m m area of ultrathin sections of tissue-samples at various times of pupal development. Results so obtained are shown in figure 4A. Increase in the number of neuronal profiles per unit area of ultrathin sections of brain-neuropil were similarly estimated from electron micrographs of known enlargement (figure 4B). It was observed that the neuronal profile counts continuously increase until 54 h of development, then tend to stabilize by 85 h (figure 4B). The number of synapses per unit area of ultrathin sections during the course of development were also estimated from electron micrographs and are given in figure 4C. The average number of synapse per neuronal profile in the brain-neuropil was estimated at a given time by dividing the number of synapses given in figure 4C with number of axon profiles read from the curve in figure 4B. This was considered necessary to overcome the scattering of observed points in figure 4B. Results so obtained are plotted in figure 4D.

3.2 Changes in the rind/cortex

Electron-dense cells showing signs of degeneration were found in significant numbers in the cortex right from 6 h

through 60 h (figures 4A; 5A–C). It was not possible to distinguish between, the degenerating neuronal or glial cells of the cortex by TEM. The number of degenerating cells present with progressing time is given in figure 4A.

3.3 Changes in muscles

Muscles of larval origin in pupa develop vacuoles by 6 h (figures 6A, B; 7A, B) and they start showing definite signs of degeneration (figures 6A, B; 7B). These muscles lose most of their ultrastructural details between 12–18 h. By 18 h, no trace of organized larval muscle is detectable by TEM in the head, while in the thorax, except for a few known muscles such as larval oblique muscles (Fernandes et al 1991) (figure 7C), most of the muscles histolyze. Myoblasts are detectable from 8 h onward (figure 6D, 38 h sample) and progressive development of the muscle precursor structure proceeds in them (figure 6C). Gradually, these myoblasts form characteristic muscle structure, and by 80–96 h, the muscles are almost fully formed (figures 6E, F; 7E, F). However, a few myofibrils continue to be added laterally to the preformed muscles even during
86–96 h (figure 7D, 86 h sample, short thick arrow).

3.4 Electron-dense mitochondria

In addition to the normal-looking mitochondria of characteristic shape, electron-dense mitochondria (EDMITs)

are detectable in neuropil (figure 1F, asterisk), cortex (figures 5C, D; 8B) and muscles during pupal development (figures 6F, 7A; 8A, C, asterisk). These EDMITs are at times found in isolation together with normal mitochondria (figures 5C; 6F) and are often found in large clusters, with more than 100 mitochondria (figures 5D, 7A, 8B, C). Usually, these clusters are near the surface of the cortex or muscle, but they also occur in the interior of the tissue (figure 5C). A few of them, referred to as electron-opaque mitochondria, are found enclosed in a vacuole and appear darker than the rest of EDMITs and normal-looking mitochondria (figures 5D, 7A, 8A arrow). These EDMITs were not found in the tissues of adult Drosophila.

Among several marker molecules known to be present in normal mitochondria (Lehninger 1990), cytochrome c and cytochrome oxidase were selected for histochemical characterization of EDMITs. Horseradish peroxidase-DAB reaction shows darkly stained small particles (10–30 nm in diam) inside the normal-looking mitochondria (figure 8C, arrowhead). Similar darkly stained particles occur also in EDMITs (figure 8C, arrowheads), indicating the presence of cytochrome oxidase and cytochrome c in EDMITs.

For a general view of the progress of morphological development of the CNS of Drosophila,photomicrographs of 1 m m-thick sections of the brain region are shown in figure 9. The time course of major events during the development of the imaginal CNS of Drosophila during pupation at 25°C is given in figure 10.

4. Discussion

Prior to undergoing metamorphosis, mature non-feeding larvae attain maximum carbohydrate reserves, which are mainly present as glycogen in the fat body and trehalose in the haemolymph (Chippendale 1978). The formation of imaginal CNS from the larval CNS takes place inside the confined space of puparium without intake of any nutrients, except air from the external surroundings. This necessitates that the system has sufficient energy reserves and be able to generate utilizable energy for the metamorphosis to be complete. During this period the biochemical decomposition of glycogen and trehalose supply glucose, as energy source for the formation of pupal and adult tissues (Chippendale 1978). At the initial stage of metamorphosis (6–12 h, figure 1A, B), many profiles within the neuropil contain glycogen deposits (Singh et al 1991), (figure 5A–C). Since such glycogen deposits are not found in the neuropils of adult Drosophila, it is inferred that the glycogen present in the neuropils of pupa is utilized during metamorphosis as an energy source, in addition to other sources of energy such as: glycogen present in fat bodies, trehalose present in haemolymph and products of processes such as, cell degeneration. Though the energy derived from glycogen in the neuropil may be small compared to other mentioned processes, yet being located right within the neuropil, it is likely to have a significant role in the development of the neuropil and the processes within it.

Significant degeneration of cortical cells and their profiles in the neuropils, is seen during the first half of metamorphosis (up to 50 h) (figures 1A–F, 3A, B). It is seen from figure 4A, where estimates of degenerating cells in the cortex and the profiles in the neuropil, are given that: (i) the cell bodies in the cortex show degeneration broadly peaking around 42 h and, (ii) at least 2 types of profiles exist in the neuropil, one that is maximally susceptible to degeneration at 12 h and the other whose degeneration peaks around 42 h. This bi-modal distribution can be understood in the light of the fact that the cortex in the CNS of Drosophila does not contain sensory neurons, which are on the periphery or away from CNS. The cortex consists of cell bodies of interneurons, motor neurons, glial, and neurosecretory cells. We suggest that the 42 h degeneration peaks in figure 4A correspond to some of these cells and their profiles in the neuropil, whereas the 12 h peak (figure 4A) corresponds to the degeneration of sensory profiles, whose cell bodies are not present in the cortex. We are aware that there could be other explanations. For example it may be possible to explain these results by assuming that there are 2 types of neuron profiles namely inhibitory and excitatory.

It is observed from figure 4B that the number of axon profiles in a given area of a section, or for that matter in a given volume, increases with the progress of development. These results are comparable with the increase in the number of Kenyon cell axons during pupation studied by Technau and Heisenberg (1982). Contrary to the expectation, the size of axon profiles does not increase appreciably; it is their number that increases. This suggests that branching of axon profiles increases during the development.

In addition to the neurosecretory granules, we found only 3 prominent types of synaptic vesicles in the CNS: electron-lucent, electron dense and dense core vesicles. We were not able to detect flattened vesicles in significant numbers. These synaptic vesicles are comparable to those described earlier by Shepherd (1979) for vertebrate nervous system: (i) electron-lucent, spherical (40–60 nm in diam) vesicles containing acetylcholine or amino acids, (ii) electron-lucent flattened vesicles (30–60 nm long) having g -amino butyric acid or glycine, (iii) electron-dense or dense-core (40–100 nm in diam) vesicles containing catecholamines and (iv) large electron-dense vesicles (100–160 nm in diam) known to be neurosecretory granules. All the above types of vesicles were prominently present in Drosophila neuron profiles, except electron-lucent flattened vesicles.

The number of synapses per unit area of sections initially decreases during the development and is minimum at 18 h (figure 4C). By this time, some of the synapses of larval origin, histolyze. Then there is a 2-step increase in the number of synapses per unit area until a peak value is reached at 72 h; this suggests that developmentally, at least 2 classes of synapses are formed during this period. Soon after, the number of synapses decrease sharply until 80 h (figure 4C). The observed decrease of 70–75% in the total number of synapses, within a span of 6 h out of 100 h of pupal development could be dismissed as fictitious, if only there was 300–400% increase in the volume of CNS, during this 6 h. Certainly, this does not seem to be the case. Drosophila, like Musca domestica (Fröhlich and Meinertzhagen 1983), makes synapses in excess, and finally only 25–30% of the maximum number of synapses formed are retained and the rest eliminated.

The average number of synapse per axon profile in the neuropil (figure 4D) has a close resemblance with the average number of synapses per unit area during the course of development (figure 4C). Because 2 neuron profiles share one monad synapse, a value of 0·5 or more for an average number of synapse per axon profile indicates that approximately every neuron profile is engaged in making a synapse at 72 h (figure 4D).

The metamorphosis of larval muscles to imaginal muscles goes through a more drastic change than the CNS. In CNS during metamorphosis, the neuronal profiles do not vanish completely; rather they become loose and the space so generated is filled with extracellular fluid, most probably produced by histolysis. In muscle metamorphosis, however, most of the muscles in head and all muscles, except dorsal thoracic muscles and some pharyngeal muscles (Bodenstein 1950; Fernandes et al 1991) in thorax, histolyze. New muscles are probably formed from the molecules generated by histolysis. What sort of changes these molecules undergo before being incorporated into imaginal muscles, is unknown. The control processes operating at any of these steps are worth investigating.

The presence of a large number of EDMITs is a positive indication that the imaginal structures, in need of readjustment of their energy requirement, discard a number of such organelles. For the following reasons, these EDMITs are likely to be degenerating mitochondria: (i) presence of a delimiting double-membrane in the EDMITs, (ii) lack of ultrastructural details such as cristae, and (iii) occurrence of some EDMITs inside vacuoles, indicating that they are going to be histolyzed or ejected out of the cell.

We thank Mrs Shubha Shanbhag and Dr Rajashekhar Patil for their helpful suggestions.

References

MS received 31 May 1999; accepted 23 June 1999