Spatial and temporal patterns of neurogenesis in the embryo of the locust (Schistocerca gregaria)

During neurogenesis in insects, the neurons in each segmental ganglion are produced by a segmentally repeated and stereotyped set of neuronal precursor cells (neuroblasts NBs) and midline precursor cells (Bate, 1976; Bate and Grünewald, 1981; Doe and Goodman 1985a). Each NB divides asymmetrically and repeatedly to produce chains of smaller ganglion mother cells (GMCs) which divide once more to produce a pair of sibling neurones (Bate, 1976). The identity of these neurons is understood to be specified in three stages: first, the identity of each NB is uniquely specified by its position in the neural ectoderm to produce a particular and invariant family of neurons; second, the identity of a pair of neurons is specified by the rank order of their parent GMC in the lineage produced by the NB and, third, an interaction between sibling neurons decides which of two fates each will adopt (Doe and Goodman, 1985a; 1985b; Doe et al. 1985; Kuwada and Goodman, 1985). This model is based almost entirely on observations of the earliest stages of neurogenesis, from the enlargement of the NBs in the neural ectoderm to the production of just the first three or four GMCs. Some NBs however, produce as many as 50 GMCs and the model has been extrapolated to cover the entire period of embryonic neurogenesis. Whether this model still applies at the later stages is not known, and it is certainly possible, that at these later stages, neuron identity may not be so rigidly specified. Studies of the later divisions of the NBs have not been undertaken, because of the practical difficulties involved in the identification of individual NBs and their progeny. Beyond the earliest divisions of the NBs, the number of NBs that can be distinguished decreases and the increasing number of neurons makes following the progeny of individual NBs impossible.

Here we describe methods that have allowed NBs to be visualised throughout embryonic development and provide a detailed description of the entire process of embryonic neurogenesis from the first appearance of the NBs to the last division of the last NB. More importantly, we have been able to trace the complete life-history of each of the original complement of NBs in the mesothoracic neuromere. With this information, we can now individually identify all of the NBs present in the mesothoracic ganglion at any particular stage of development. The consequences of this are that we have made the later stages of neurogenesis accessible to a level of investigation not previously possible.

Embryos of Schistocerca gregaria (Forskal) from a laboratory colony were used for all the work described here. The eggs were laid in moist sand and stored in sand at a constant 30°C. Eggs for experiments were removed from the sand and cultured on moist filter paper in Petri dishes, at 30°C. Embryos were staged according to the timetable of Bentley et al. (1979).

Toluidine blue (Raymond Lamb) was used to stain living nervous systems according to the recipe of Altman and Bell (1973). Ventral nerve cords were incubated in the stain for 15–30 min at room temperature, then fixed and destained in Bodian’s fixative, dehydrated, cleared in xylene and mounted in Canada balsam.

DNA synthesis was monitored using the incorporation of the substituted nucleotide, 5-bromodeoxyuridine (BUdR), revealed immunocytochemically with a monoclonal antibody against BUdR (Gratzner, 1982). This method has already been used to study neurogenesis in the CNS of larvae of Drosophila melanogaster (Truman and Bate, 1988). BUdR (Sigma) was introduced into embryos by injecting a small quantity (<1 pl) of lOmM-BUdR in embryo saline directly into the embryo in ovo, with a glass micropipette. To assist injection and to permit accurate staging of the embryos, the eggs were dechorionated in bleach for 30–60 s and washed in clean saline. Various concentrations of BUdR were tried; concentrations less than 10 mM gave poor and variable labelling and concentrations higher than 10 mM were not used. With lOmM-BUdR, labelling was strong and embryos developed normally and hatched. Therefore this concentration was used in all experiments presented here.

Following injection, embryos were incubated on moist filter paper at 30°C until required. In general, two protocols were used. In the first, embryos were injected and incubated for just 24 h and then processed. In the second, embryos were left after injection to hatch and processed on the first day as 1st instar larvae.

Nervous systems labelled with BUdR were either fixed in situ (25-40% stage embryos) or dissected free prior to fixation (40% to hatchling stage). Fixation was for 30 min in Camoy’s fixative. Ganglia were then rehydrated and stored in phosphate-buffered saline (PBS) with 0.3% Triton X-100 (PBS-TX). Prior to immunocytochemistry the preparations were treated with 4N-HC1 in PBS-TX (1 to 1) for 30 min. After thorough washing in at least six changes of PBS-TX, the tissue was incubated in 10 % goat serum for at least 1 h at 4 °C, followed by incubation in a 1:200 dilution of the anti-BUdR antiserum (Becton-Dickinson) in PBS-TX for 48 h at 4 °C with agitation. After the primary antiserum, the tissue was washed for 8h with at least 8 changes of PBS-TX with agitation. The tissue was then exposed to a rabbit anti-mouse IgG antibody conjugated to peroxidase (1:250 in PBS-TX) for a further 48 h at 4°C with agitation. The peroxidase was visualised according to the method of Watson and Burrows (1981). After the reaction, the tissue was dehydrated, cleared in xylene and mounted in Canada balsam. All the materials were examined using x25, x40 and x63 oil immersion objectives on a Zeiss microscope fitted with Nomarski optics.

A study of the later stages of neurogenesis in the locust requires that the NBs and their mitotic activity be monitored throughout the full course of embryonic development in every segment. At the early stages of development (25–40% stage), it is possible to identify all the NBs and observe their divisions in living embryos using Nomarski interference contrast optics. During the later stages of development, the growth of the nervous system caused by the addition of more neurons and their enlargement as they differentiate means that the original NBs are displaced and become relatively small making the resolution of individual NBs difficult. In the absence of a NB specific stain, we cannot directly observe neurogenesis beyond the earliest divisions.

Toluidine blue staining is a relatively selective stain for NBs and Fig. 1 shows a series of TB stains of the mesothoracic ganglion at different stages of embryogenesis. At the earliest stage (about 35 %), the NBs are visible as large, ventrally located cells with large nuclei and darkly staining cytoplasm (Fig. 1A). In addition to the NBs, TB staining shows the early neuronal progeny of the NBs and the intervening sheath cells. At this stage, the NBs are quite distinct with an almost complete array visible in the mesothorax. By 50% of development, the number of neurons is greatly increased but NBs are still evident although relatively smaller and spatially dispersed (Fig. IB). It is also apparent by this stage that there are fewer NBs suggesting that some of them have completed their lineages and have already degenerated. By 70 %, the number of NBs still visible is smaller (approx. 30), and these remaining NBs can be smaller than the surrounding neurons (which have now enlarged) and recognisable only by their shape and the associated chain of progeny (Fig. 1C). It is still possible to distinguish NBs until about the 85% stage, but beyond this stage it is no longer possible to identify them in any of the ganglia of the ventral nerve cord.

Using TB we have been able to distinguish NBs in the nervous system throughout most of embryonic development. It is possible, however, that the TB does not reveal all the NBs in each ganglion, particularly in the later stages of development when the apparent number of NBs is low and they are small in size relative to their rapidly enlarging progeny. We therefore used the incorporation of BUdR to reveal the presence of NBs by their mitotic activity. For these experiments, embryos were staged, injected with BUdR and allowed to continue their development for a further 24 h before the preparations were fixed and processed immunocytochemically with an antibody raised against BUdR. In such preparations, clusters of small cells are labelled (Fig. 2A), with labelling restricted to the nuclei of the cells, each cluster contains a single large NB, several (usually 3) GMCs and a variable number of neurons (Fig. 2B). All the cells in a labelled cluster are the progeny of the NB and each cluster therefore represents an active NB. Every NB that can be recognised is associated with such a cluster of labelled cells. In positively stained preparations, we never saw a NB that was not associated with a cluster of labelled cells, although occasionally we saw clusters of labelled neuronal cells that were not associated with a NB. Since the timing and location of these clusters coincide with the disappearance of NBs in TB-stained ganglia, it is likely that these cells represent the final progeny of the NBs concerned.

Labelling with BUdR at various developmental stages can reveal the presence of NBs. Fig. 3 presents examples of the metathoracic ganglion after injection of BUdR at different stages, 30% (Fig. 3A), 55% (Fig. 3B) and 70% (Fig. 3C). In all cases, the NBs are revealed by their incorporation of the BUdR. Thus the incorporation of BUdR is a graphic way of revealing NBs in the developing nervous system. Furthermore, these preparations highlight the developmental changes in the number and distribution of NBs in this ganglion.

The results obtained using the incorporation of BUdR are entirely consistent with those observed with the TB staining. TB staining of early nervous systems (<70 %) reveal the same pattern of NBs as found using BUdR. BUdR incorporation is however, a more reliable and repeatable method and at later stages of development revealed NBs not seen using the TB.

The earliest signs of mitotic activity in NBs can be seen at 25 %, when NBs can be identified in every segment. At this stage, however, not all the NBs are undergoing DNA synthesis and cell division. There is an anterior-to-posterior gradient of neurogenic activity. Thus as shown in Fig. 4, it is clear that, although most of the NBs of abdominal segments 1, 2 and 3 are actively dividing, in the more posterior segments there is less evidence of division. This is reflected in two ways: first, there are fewer NBs showing signs of BUdR incorporation and second, the labelled lineages produced by the active NBs during a pulse appear to be smaller. Eventually. all the NBs start dividing but there is a slight anterior-to-posterior delay and, as development proceeds, progressively more posterior cells commence DNA synthesis. Thus, by 30%, all NBs in every segment can be seen to be synthesising DNA and dividing. At this time, a clear pattern of NBs can be recognised in every segment (Figs 5 and 6).

At the 45-50 % stage, actively dividing NBs can be seen in every segmental ganglion. The NBs no longer form the tightly packed array typical of the earlier stages but are becoming spatially separated (Figs 5 and 6). It is also apparent that many NBs have produced their entire lineages and have already degenerated. It is assumed that the disappearance of a NB indicates that it has degenerated. From our observations we have seen that many NBs show clear signs of cell death (Bate, 1976) prior to their disappearance but, since we cannot provide proof of the degeneration of all NBs, it is possible that some NBs just cease DNA synthesis and reduce in size. The losses are most obvious in the abdominal segments where over 50 % of the NBs have already disappeared (Table 1). In the thoracic ganglia, they are fewer, with only 14 of the original complement of NBs missing (Table 1).

The pattern of NB disappearance in each ganglion is highly invariant and suggests that each NB has its own specific timetable of development, that the active life of each NB is uniquely determined and that it degenerates at approximately the same stage of development in every individual. Thus it is possible to trace the fates of individual NBs and specify the point of degeneration. For instance, NBs 6–3 and 7–3 are the first 2 NBs to degenerate in the mesothorax and, by the 45% stage, both are missing. It is possible to recognise the pattern of NB division and degeneration in all ganglia and so compare the patterns of NB death in different ganglia. Thus, Fig. 5 shows a camera-lucida drawing of the patterns of NB activity in the three thoracic ganglia of a 50 % embryo, a comparable pattern of NBs can be seen in all three ganglia, with the exception of the anterior median NB, which is found only in the prothorax. The same can be done for the abdominal ganglia (Fig. 6); in this case, all of the unfused abdominal ganglia (Al to A7) show an almost identical pattern of persistent NBs at the 50% stage. It is, however, not possible to demonstrate a complete homology between the abdominal and thoracic patterns, although there are examples of individual homologies, thus NBs 6–3 and 7–3 are the first NBs to degenerate in the abdominal as well as the thoracic ganglia.

Any comparison with the terminal abdominal ganglion (TAG) is difficult, as the TAG is unique in representing the fusion of at least 3 abdominal segments. Even so, a segmentally repeated pattern of NBs can be recognised in the anteriormost of its constituent segments at the earliest stages of development (Fig. 6), but, by the 50% stage, it is not possible to recognise with confidence any homologies with the more anterior ganglia.

In each thoracic neuromere, about half of the original NBs have now degenerated and there are about 30 NBs actively dividing (Fig. 5 and Table 1). As at the previous stages, an identical pattern of extant NBs can be seen in each of the three thoracic ganglia.

In most of the abdominal segments, however, only the final traces of neurogenesis can be seen with only 5 or 6 NBs remaining (Fig. 6 and Table 1) and by 75 % even these NBs have disappeared. Nonetheless at 70 %, it is still possible to recognise an identical pattern of persistent NBs in each of the segments A2 to A7. On the other hand, it is also apparent that the NB activity of segment Al differs from its more posterior counterparts in that there are as many as 10 NBs still dividing in the ganglion (Table 1). The TAG also shows an extended sequence of NB activity, where almost 30 NBs persist and continue to divide (Table 1 and Fig. 6).

By 85%, signs of neurogenesis have ceased in all ganglia apart from the 3 thoracic ganglia and the TAG. The number of active NBs in these ganglia is however small, 4 or 5 in the thoracic ganglia and 6 in the TAG (Table 1). As before, an identical pattern of NB activity can be seen in all three thoracic segments (Fig. 5), and no homologies can be recognised in the TAG. During the next 24–36 h, all persisting NBs will complete their lineages and degenerate, so that by 95 % all neurogenic activity has ceased in the ventral nerve cord.

Using BUdR, Truman and Bate (1988) showed that, in the terminal ganglia of Drosophila larvae, there was a sexual dimorphism in the time course of neurogenesis. This was such that, in males, neurogenesis was extended by more than 30h. With this in mind, we made particular observations for evidence of such a dimorphism in the TAG of locusts. Although it was not possible to determine the sex of embryos, we were unable, however, to detect any differences in neurogenesis that could be attributed to the sex of the animal.

As shown above, there is a consistent pattern of NB disappearance in each ganglion suggesting that each NB has a unique life-history, completing its lineage and disappearing at a specific stage of development. With the level of detail provided by the incorporation of BUdR, we have been able to trace each of the NBs in the mesothoracic neuromere from the first divisions to disappearance. Following individual identified NBs from one stage to the next is relatively straight forward and is not complicated by migration, the NBs remaining relatively close to their original positions. The biggest movements are made by NBs 4–1, 4–2 and 4–3 which, although they start in a line perpendicular to the anterior-posterior axis, move such that by 70% 4–2 and 4–3 are displaced posteriorly (Fig. 7). The only other movements are such that the more laterally placed NBs (e.g. 6–4, 7–4, 2–5 and 3–5) are displaced dorsally to come to lie on the dorso-lateral edges of the ganglion. These NBs are however still identifiable by their position relative to the other NBs.

From this information we are now able to identify unequivocally and make a map of all the NBs present in the ganglion at all developmental stages (Figs 7 and 8). The full life-histories of all the individual NBs is shown graphically in Fig. 9.

A fixed stage is given for the disappearance of each NB, but there are however, slight variations in the exact time of disappearance and therefore the stage given for the disappearance of a NB is the latest stage at which that NB has been identified.

Neurogenesis was further defined by comparing the rates at which NBs produce labelled progeny in different ganglia. Embryos were staged, injected with BUdR and then cultured in ovo for the required time before fixation and processing. Following injection, labelled progeny could first be detected after 3h and consistently thereafter for the next 21 h. Thus in a 24 h pulse, the labelled cells are those that have gone through at least one S-phase during the previous 21 h. The outcome of this regime is the labelling of the NB, 2 or 3 GMCs and a variable number of neurons. The rate of proliferation was determined by counting the number of labelled nuclei (GMCs and neurons) associated with each NB. This rate does not give an exact figure for the rate of NB division since some of the labelled cells may be the progeny of a preexisting GMC and labelled as a consequence of the division of the GMC. Given the sample of cells used in this study, it is unlikely that this fact influences the comparison of the rates for different NBs, but it exaggerates the absolute rate of NB division. A more realistic estimate of the rate of NB division can be made if one considers the patterns of division made by NBs and their GMCs (Fig. 10). From this, we can see that it is possible for a lineage to contain four labelled cells (in addition to the NB) after only one NB division. Only one of these cells was produced by that NB division, of the other cells two are produced by a GMC that has divided and the other is a GMC that has entered S-phase. With the same logic, a lineage of 11 labelled cells (plus a NB) can be produced after only 4 divisions of the NB (Fig. 10).

In this study, we determined the proliferation rate for 4 uniquely identified NBs in the mesothoracic ganglion (2-1, 4–2, 4–3 and 7–1) at 6 stages of development. Data from 3 of these NBs are presented graphically in Fig. 11; from this it is clear that NBs 4–2 and 4–3 are producing progeny at the same rate, approx. 11.5 cells/24h. Taking into consideration the patterns of division shown in Fig. 9, this suggests a rate of NB division of between 4 and 5 divisions in 24 h. This rate declines simultaneously for both NBs as they come to the completion of their lineages. The third NB (NB 2–1) appears at the early stages to be dividing at a rate comparable to the others (11.25 cells/24h at 50%) again suggesting an actual NB division rate of 4 to 5 divisions in 24 h (Fig. 10), but this rate declines as it nears its demise (65 %), earlier than NBs 4-2 and 4-3. Two other identified NBs in the same ganglion were examined and show more or less the same maximal rates of cell division, declining only as they come to the ends of their lineages (Table 2). The data suggest that NBs in the same ganglion divide at approximately the same maximal rate.

Do all the NBs divide at the same rate or are there segmental differences? To answer this question, we looked at the rates of division in other ganglia. In both the prothoracic and metathoracic ganglia, NBs 2–1 4–2 and 4–3 all divide at rates comparable to their mesothoracic homologues (Table 2). In the abdominal ganglia, it was not possible to examine the division rates of individual identified NBs and so the data averaged for at least 5 different but unidentified NBs are presented. These data show that the abdominal NBs divide at a slower rate of 6.5 cells/24h, a rate equivalent to only 2 NB divisions in 24 h (Fig. 10), a rate much slower than their thoracic counterparts (11.5 cells/24h) (Table 2, Fig. 11). Since it was not possible to identify individual NBs, we cannot compare the rates of division of individual NBs. The NBs of the TAG show a rate of division similar to the rates of the other abdominal ganglia (Table 2).

It is possible that death of GMCs or neurons shortly after their last division could serve to underestimate cycling times of NBs (especially in the abdominal ganglia), but from our observations cell death was not seen as a significant factor amongst the labelled lineages examined and we feel that it does not lead to a serious underestimation of the rates of proliferation.

Although it is possible using BUdR incorporation to identify the NBs and their progeny in the CNS unequivocally, neurons are not the only cell types labelled by this procedure. There are at least 2 other classes of proliferating cells in the embryonic CNS. The first class of cells has large (30 μm), flat nuclei that lie superficially on the ganglia, connectives and nerve roots, forming part of the extraganglionic sheath (Fig. 12). It seems likely that these are the nuclei of perineurial cells (Wigglesworth, 1960) which differentiate to form part of the perineurial sheath. The second class has small irregularly shaped nuclei and appear to form a layer between the outer cortex of neuronal cell bodies and the neuropile. This class appears to extend into the connectives and nerve roots (Fig. 12). On the basis of this distribution, it is likely that these are the nuclei of glial cells associated with the neuropile and connectives (Wigglesworth, 1960). Although both types of cell are numerous and readily labelled, they pose no problem for the identification of neurons since the size, shape and distribution of their nuclei are quite distinct from those of the neurons.

The labelling of these two classes of glial cell with BUdR reveals an interesting facet of their development. From the data presented here, it is apparent that the two classes develop with completely different time courses (Fig. 13). The so-called neuropilar glial cells are first detected as labelled nuclei in embryos labelled at about the 35 % stage and labelled cells of this class can be seen consistently and in profusion after injections of BUdR at all developmental stages thereafter up to about 80 %. After 80 %, no more labelling is seen until after hatching. The perineurial cells on the other hand label during a much narrower developmental window. The earliest labelling is at 45 % when a few (<20) labelled nuclei are seen in a small number of preparations. Labelling reaches a maximum by 60% when all preparations have a heavily labelled perineurium. Labelling of this cell type ceases by 70% (Fig, 13).

In recent years, neurogenesis in insects has become a model system for studying the mechanisms that underlie the specification of neuronal identity (Doe and Goodman, 1985a; Doe and Goodman, 1985b; Doe et al. 1985; Kuwada and Goodman, 1985). All of these studies have, however, placed their emphasis on the earliest stages of neurogenesis; the determination of NB identity and specification of the identities of the first 6 to 8 neurons produced by a NB. This restriction has been enforced primarily by the difficulties encountered in looking at and understanding the later divisions of the NBs. With this work we have described a technique that has overcome this restriction and has provided a means of visualising and identifying NBs at all stages of embryogenesis. Using this method we have been able to: (1) describe the full timecourse and pattern of embryonic neurogenesis in the ventral nerve cord, (2) describe segmental differences in the timing and pattern of neurogenesis and (3) trace the fate of all of the original complement of NBs in the mesothoracic neuro-mere and make a complete record of the stage at which each NB ends its lineage.

As a means of extending our knowledge of the later stages of neurogenesis, perhaps the most valuable product of this work has been the tracing of individual NBs through development. Until this was done the major limitation to tracing neural lineages in the locust embryo was the inability to confidently identify NBs beyond 50 %. Using the information that we now have at our disposal, this is no longer a problem. It is quite possible to look at the mesothoracic ganglion of living embryos under Nomarski optics and identify unequivocally each NB. Coupling this with the conventional techniques of intracellular recording and dye injection, we can now examine the physiological and morphological properties of neurons produced by the later divisions of identified NBs. Moreover it is possible to trace the complete lineage of identified NBs. This means that the whole period of neurogenesis is now open to a level of analysis once only possible at the earliest stages of development.

Although a detailed examination of NB fate was made only for the mesothorax, the marked similarity of the patterns of NB persistence in the pro and metathoracic ganglia make it highly probable that the fate of the individual NBs are the same in all three thoracic ganglia and therefore future work need not be restricted to the mesothorax.

At the onset of neurogenesis each of the presumptive segmental ganglia has a more or less identical complement of NBs (Bate, 1976; Doe and Goodman, 1985a), which are responsible for the production of virtually all the central neurons in each ganglion. The ganglia produced by these stereotyped arrays are, however, not identical and, therefore, each array of NBs is capable of producing not only the neurons of the simple abdominal ganglia but also the neurons of the larger and more complex thoracic ganglia. The mechanisms by which this is achieved are not clear, although cell death and variations in the lifespan of NBs are known to be important (Booker and Truman, 1987a, 1987b). Using BUdR and TB, we have been able to trace neurogenic activity in each segment from its onset to the final divisions of the last NB. Our data suggest that in the locust at least two factors can contribute to this difference in the length of the lineages produced by NBs in different ganglia: that is, segmental differences in the life time of individual NBs, and in their rates of division.

The time course of neurogenesis divides roughly into a thoracic pattern and an abdominal pattern. The first signs of division in the thoracic NBs occur at about 25 %, but not until about 28 % do all thoracic NBs appear to be actively dividing. Neurogenesis then continues unabated in all 3 thoracic ganglia until shortly before hatching (92%). The persistence of neurogenesis to this late stage is quite remarkable since it shows that neurons are being generated to within 20 h of the embryo hatching. Studies of the progeny of the median NB of the metathoracic ganglion show that the development of a neuron from axon outgrowth to the onset of electrical excitability can take up to 25 % of development (i.e. 3.5 days in Schistocerca) (Goodman and Spitzer, 1979). With such a time course, it is unlikely that the neurons produced by NBs at these late stages could have fully differentiated before hatching and it may be that they complete their differentiation during larval life. The abdominal NBs (A2–A7), by contrast, start division later and in an anterior-posterior sequence. Thus division in A2 begins at 28% as against 35 % in A7. Neurogenesis in these ganglia ceases at 70%, but once again in an anterior-posterior sequence. The NBs of Al and the TAG have unique patterns of division. In Al the pattern is an intermediate one, division commencing at the same time as in the thoracic NBs and continuing until 77%, (i.e. longer than in the other abdominal ganglia but shorter than in the thorax). The NBs of the TAG are among the last to start dividing but dividing NBs persist until well after they have disappeared in all other abdominal ganglia. Neurogenesis in the TAG ends at the same time as in the thorax.

Although the temporal sequence of neurogenesis we describe suggests a continuous sequence of NB divisions, this does not mean that all of the NBs are active throughout the entire period. There is in fact a continual loss of NBs from all segments throughout neurogenesis. It is clear that the thoracic ganglia differ from their abdominal counterparts not only in the temporal pattern of neurogenesis but also in the rate at which NBs complete their lineages and degenerate. For example at the 50% stage, only 15 of the original 61 NBs in each thoracic ganglion have degenerated as against 40 in each of the abdominal ganglia (A2-A7). In this way, neurogenesis is not only extended in the thoracic ganglia, but more NBs persist to the later stages than in the abdominal ganglia. This is very similar to the pattern of neurogenesis seen in the larval nervous system of holometabolous insects such as Drosophila (Truman and Bate, 1988) and Manduca (Booker and Truman, 1987a). In Schistocerca, however, neurogenesis is completed in the embryo and does not extend into larval life.

The differences between the ganglia are not reflected solely in the timing of onset and termination of neurogenic activity but also in the spatial pattern of NB degeneration. Each NB has a specific lifespan, and divides according to a particular programme to produce its progeny and then degenerates (there is no evidence to suggest that this is an exact number of divisions). Thus cell death is a consistent and predictable event for any NB and there is a distinct sequence of NB deaths in each segment. Each of the 3 thoracic ganglia have identical spatial patterns of NBs at all developmental stages (with the exception of the anterior median NB in the prothorax) suggesting that homologous NBs in each segment undergo exactly the same patterns of division and degenerate at the same stage in each of the 3 ganglia. Like the thoracic ganglia, an identical pattern of NBs can be seen in the abdominal segments (A2-A7) at all developmental stages (the A9-A11 possess an anterior median NB) again suggesting that segmentally homologous NBs share the same pattern of mitotic activity and degenerate at equivalent times in development. It is not possible, however, to see any homologies with the sequence of NB death seen in the thoracic ganglia. The first abdominal segment and the TAG are different from the other ganglia. In Al, the rate of NB death resembles that seen in the other abdominal segments until about 55 % when several NBs survive beyond their homologues in posterior segments. The TAG is unique and cannot be compared with any other ganglia and no homologies can be recognised.

All of these intersegmental differences indicate that the complexity of, for example, the thoracic ganglia is in part generated by the extended divisions of more NBs. Similar extensions of neurogenesis in Al and the TAG are also indicative of the specialisation of these ganglia; e.g. Al becomes fused with the metathoracic neuro-mere and presumably contributes central neurons re sponsible for the control and processing of information related to locomotion (Robertson and Pearson, 1984) and ventilation (Hill-Venning, 1988). Similarly, the TAG is specialised in being responsible for the control of the genitalia and primary processing of information from the sensory structures associated with the genitalia (Thompson, 1986; K.J. Seymour pers. comm.). In contrast to Drosophila where in the males neurogenesis in the terminal ganglia is extended to later stages than in females (Truman and Bate, 1988), there is no sexual dimorphism apparent in the pattern of neurogenesis in the terminal ganglia of the locust.

These data tell us more about the development of the segmental specialisations. If one considers the rates of division of the NBs in the different segments, it can be seen that not only do the thoracic NBs persist longer and in larger numbers but also produce progeny at a more rapid rate (11.5 cells day^-1), which is equivalent to 4 or 5 NB divisions every 24 h, than their abdominal homologues (6.5 cells day^- ), which equals a NB division rate of just 2 divisions in 24h. Therefore, the complexity of the thoracic ganglia is doubly manifest. Again these results show strong parallels with similar data obtained for the larval NBs of Drosophila, which have the same higher rate of division for thoracic NBs (Truman and Bate, 1988). In Drosophila, the overall rates of division are much higher than in the locust but the ratio of thoracic to abdominal NB proliferation rates is almost identical for both insects. Interestingly, Al and TAG, which are of intermediate complexity in terms of having an extended period of neurogenesis, do not show a rate of proliferation significantly greater than their counterparts in the simpler abdominal ganglia and therefore their specialisations are due primarily to the extension of their mitotic activity.

We would like to thank Drs Alfonso Martinez Arias, Gilles Laurent and Helen Skaer for reading the manuscript. This work was funded by the SERC(UK).