5.1 Introduction

As I have outlined in Chapter 2, there are three principal ways in which elements of a diversifying network could behave: New elesments could differentiate; elements could be preserved without significant change; or elements could be removed from the network. The reduction of elements could be achieved in two ways. Either cells forming certain elements would not differentiate at all; or they do differentiate to a certain stage and then get eliminated at a later stage in development.

As far as the muscle pattern is concerned, I have outlined in Chapter 2 that the absence of muscles in the dorsal domain of A8 is a very distinctive feature. This applies in particular to dorsal longitudinal muscles. Apart from the minute and barely characterised MO1, there is only the large muscle DO1 in A8 in a region that contains eight muscles in the reference segment. Taking the muscle pattern as the most accessible part of the neuromuscular machinery, one can ask whether the observed reduction of elements is echoed in the CNS on the level of the respective motorneurons.

The removal of neurons leading to a diversification of the neuromuscular architecture is a phenomenon well known in vertebrates. Between 50% and 80% of all embryonic motorneurons undergo apoptosis after extending their axons as a result of a failure in finding their appropriate target (OPPENHEIM, 1991). Experiments investigating similar mechanisms in the grasshopper Schistocerca gregaria could not show a target-dependence in motorneuron survival: If muscles are surgically removed from grasshopper embryos, their respective motorneurons still extend their axons into the domain of their target and continue growing even though they fail to find it (BALL et al., 1985; WHITINGTON et al., 1982). Previously, segmental differences in the number of neurons and glia have been linked to changes in the lineage of specific neuroblasts (TRUMAN & BALL, 1998; PROKOP & TECHNAU, 1994; SCHMID et al., 1999; UDOLPH et al., 1993; PROKOP et al., 1998; BERGER et al., 2004).

These changes include both alterations in proliferation and apoptosis. One study showed that apoptosis occurs in postembryonic neuroblasts, so that the number of neurons in the adult is locally reduced. A pulse of expression in the Hox-gene abdominal-A specifies the time at which cell death occurred. Thereby, the authors showed a mechanism that linked abdominal-A with the determination of the final number of the progeny that a single neuroblast generates (BELLO et al., 2003).

Recently, apoptosis has been shown to occur segment-specifically in post-mitotic neurons of Drosophila: In some segments, the pioneer neurons dMP2 and MP1 undergo cell-death at stage 17, well after their axons have been extended and innervate the hindgut. This cell-death occurs specifically in segments anterior of A6. It was linked to the differential expression of the Hox-gene Abdominal-B, which prevents death in these cells in A6, A7 and A8 by repressing the apoptotic genes reaper and grim (MIGUEL-ALIAGA & THOR, 2004; see also Chapter 6).

Based on my muscle map and the studies described above, I asked two questions: Is the absence of muscles in the periphery correlated with the loss of their respective motorneurons? If so, what mechanism is responsible for this loss – do the relevant neurons fail to differentiate or are they apoptotically “ablated” at post-mitotic stages?

5.2 No Indication of a Differentiating DA1 Provides a Model System

Muscles DA1 and DA2 receive their innervation from the motorneurons aCC and RP2 respectively. As I have described in detail in Chapter 3, these muscles appear to be absent from segment A8. Their motorneurons are contained in the RN2-gal4 pattern.

The aCC and pCC are generated in one segment in early stage 11 and then migrate just beyond the segment border into the next anterior metamere by late stage 11 (BROADUS et al., 1995). However, aCC still innervates muscles DA1 intersegmentally in the segment it originated from. RP2 remains in the segment of its origin and innervates DA2 in this segment, along with other dorsal longitudinal muscles (LANDGRAF et al., 1997; LANDGRAF et al., 2003; Matthias Landgraf, personal communication). With the absence of DA1 and DA2 in A8, aCC in A7 and RP2 in A8 would thus be supernumerary neurons. The RN2 line therefore provided me with a model system with which to address my questions (see the diagram of figure 5.1). Beforehand, however, I looked for further arguments in support of my claim that the dorsal acute muscles are not present in A8.

As described in Chapter 4, the expression pattern of the transcription factor evenskipped includes dorsally projecting motorneurons, some interneurons, pericardial cells and a ring of cells outlining the anal pad in the late embryo. It is also expressed in DA1 in all segments from T2 to A7 and in muscle GS4 in A9 (see also Chapter 4). Using evenskipped as a marker for the differentiating DA1, I traced the development of this muscle back to approximately stage 12, when individual muscle founder cells (MFCs) appear darkly stained. One can clearly see one evenskipped positive MFC differentiate per hemisegment from A1 to A7 in addition to one or two pericardial cells (see figure 5.2).

In this figure one can also see evenskipped-positive cells differentiating in the mesoderm of A9, although they are labelled less strongly. They later migrate ventrally and become difficult to discriminate from the cells outlining the anal pad. It is likely that they will seed the formation of muscle GS4. In A8, there is no indication of a differentiating DA1 muscle founder cell. This supports my claim that DA1 is missing in this segment. With this in mind one can consider the motorneuron aCC in A7 to be a supernumerary neuron.

5.3 Cell-Death of Differentiated Motorneurons: aCC in A7 and RP2 in A8

In a 13 hrs AEL embryo of the RN2 line, the pattern of aCC, pCC and RP2 is similar in all abdominal segments from A1 to A8 (see figure 5.3). In A9, only RP2 is present along with two to three neurons posterior of RP2 that extend axons and innervate the hindgut through the hindgut nerve (see also Chapter 4). Note that this pattern is consistent with the anterior migration of aCC and pCC as well as the position of A9 as the most posterior segment.

Cells in all segments extend axons that grow to the dorsal domain of the muscle field. This includes aCC in A7 and RP2 in A8, highlighting the ISNa in A8 up to the level of DO1. There they defasciculate and clearly show growth cones (see fig. 3.8, Chapter 3). At around 14 hrs AEL, the cells show first signs of disintegration: the axons start to fragment and the cell bodies shrink until they show a clearly pyknotic appearance or have disappeared at around 15 hrs AEL (see figure 5.4).

This is around the time when aCCs and RP2s in other segments have established contact with their target muscles. Preparations of first instar larvae show aCC, pCC and RP2 with their elaborate dendritic projections in all segments. Only aCC in A7 and RP2 in A8 are missing from the pattern. I concluded from these experiments that these supernumerary neurons are ablated from the neuromuscular network after they have differentiated and went on to show that this ablation was achieved through apoptosis.

5.4 Cas-3 Signal as direct Evidence for Apoptosis

Caspases are a group of cysteine proteases that are required for apoptosis (HENGARTNER, 2000). Caspase-3 (Cas-3) protein is a reliable marker for apoptotic cells. Antibody staining against Cas-3 clearly co-localises with the pyknotic cell bodies of aCC in A7 and RP2 in A8 at 15.5 hrs AEL (see fig. 5.5).

This provides direct evidence that the ablation of these cells is achieved apoptotically. I then asked whether aCC in A7 and RP2 in A8 would survive in an apoptosis mutant. This was of interest not only because of the neuronal apoptosis, but also to investigate whether the missing muscles DA1 and DA2 would differentiate in such a mutant.

5.5 Survival of aCC in A7 and RP2 in A8 in a deficiency that removes grim, reaper and hid

In Drosophila, apoptosis depends on the RHG-motif genes reaper (rpr) (WHITE et al., 1994), grim (CHEN et al., 1996) and head involution defective (hid) (GRETHER et al., 1995) (reviewed in BERGMANN et al., 2003). Embryos that are homozygous for the chromosomal deletion Df(3L)H99 lack these three genes; they are considered to be deficient for apoptosis and have not been found to show any sign of programmed cell death (WHITE et al., 1994). Using once again antibody staining against Evenskipped, I could highlight not only the neuronal pattern of dorsally projecting motorneurons (including aCC and RP2) together with some interneurons, but also the forming muscle DA1 in the dorsal musclefield.

The formation of DA1 is not disrupted in Df(3L)H99 and identical to the wildtype pattern (see fig. 5.6). Muscle DA1 is clearly visible in all abdominal segments from A1 to A7. There is no sign of an evenskipped signal in the dorsal domain of A8 other than in a single pericardial cell. This finding shows that apoptosis is not required for the absence of DA1 in A8. It seems therefore that this muscle fails to differentiate in this segment.

On the other hand, the evenskipped pattern in the ventral nerve cord is clearly different from the one seen in the wildtype: At 14.5 hours, the wildtype aCC in A7 and RP2 in A8 have either clearly started to disintegrate or disappeared – together with U neurons and lateral cells of the eve pattern in A7, A8 and A9. In homozygous Df(3L)H99 embryos, however, these neurons do not undergo cell death (see figures 5.7 and 5.8). This finding provides further evidence that the ablation of supernumerary neurons is achieved through apoptosis.

5.6 Discussion

The involvement of apoptosis that I have described in the diversification of this neuromuscular network provides a potentially rewarding system to which one could address a variety of questions which I outline below. It is the first known case of apoptosis in post-mitotic neurons in the Drosophila CNS that can be linked to a specific function – the innervation of their respective target muscles.

As Miguel-Aliaga and Thor have pointed out before (MIGUEL-ALIAGA & THOR, 2004), apoptosis in invertebrate embryos in contrast to vertebrate embryos typically occurs in cells that have just arisen (referring to LUNDELL et al., 2003; SULSTON & HORVITZ, 1977; SULSTON et al., 1983). They propose the midline precursor cells dMP2 and MP1 as a model system to study the developmental regulation of apoptosis and the mechanisms that control programmed cell death and associated events. These include in particular the role of macrophages and glia in engulfing and removing apoptotic neurons from the CNS (SONNENFELD & JACOBS, 1995; SEARS et al., 2003; BAEHRECKE, 2002), as well as the molecular mechanisms that are involved in initiation and execution of programmed cell death. The three apoptotic genes have been demonstrated to work in an additive manner (ZHOU et al., 1997), but the cues that underlie their activation remain to be resolved.

There are several reasons why the death of aCC and RP2 might provide a rewarding system to study these aspects of developmental apoptosis in Drosophila. The cells of the RN2 pattern are very dorsally located, easily accessible and extremely well characterised. Cell death occurs relatively late and not only in posterior segments, but also in the suboesophageal ganglion S3 and in RN2 cells anterior of this segment (Li Feng, personal communication). Furthermore, the sibling of aCC, the interneuron pCC, does not undergo apoptosis in any segment. Its axon runs along the intermediate FasII fascicle, but nothing is known about its function. This provides us with a system of two sibling cells that face very different fates – further studies could identify cues that contribute to these cells’ specification and in the case of aCC in A7, the programmed cell death. There are a variety of evenskipped-related GAL4 lines available that allow precise manipulations of the cells that are contained in the RN2 pattern.

For my own project, the episode of cell death provided me with a clear example for an elimination of network elements in which the reduction of muscles in the periphery and of their partners in the CNS goes hand in hand. It also showed that the mechanism that leads to the loss of DA1 in A8 must be independent of apoptosis. This is consistent with the fact that no instance of apoptosis has been found in the developing Drosophila mesoderm (BATE, 1993). I base my claim that DA1 does not differentiate primarily on the fact that I could not detect Evenskipped in any muscle or MFC of A8. Therefore, one can hypothesise that the muscle progenitor that is equivalent by lineage to the muscle progenitor of the MFC for DA1 in the reference segment, differentiates differently in A8. It could either give rise to an adult muscle precursor rather than an MFC, or to an MFC with a different identity that does not express eve, or it could be that the progenitor is simply not formed in this segment of the mesoderm.

To investigate this issue, the most obvious experiment to do would be to trace the lineage of muscle progenitors and MFCs in the dorsal domain of A8. Labelling against twist, l’sc and eve at different developmental stages could help to resolve the question of the fate in A8 of the lineal equivalent to DA1 of the reference segment. Investigating the cues that underwrite this difference might be the next step.

The hypothesis that the cell-death of aCC and RP2 could be mediated by their failure to find target muscles is tempting, especially if one looks at the prominent growth cones that aCC and RP2 develop to explore the dorsal domain of A7. However, the muscle ablation experiments in Schistocerca (BALL et al., 1985;wrong reference see above) and other studies of cell-autonomous cell death in the developing Drosophila (MIGUEL-ALIAGA & THOR, 2004; LOHMANN et al., 2003) point in another direction. A genetic ablation of the mesoderm showed that motorneurons differentiate and form dendrites independently of the absence of target muscles (LANDGRAF et al., 2003).

Another interesting aspect of my findings is the stage at which the observed apoptosis occurs. Why do aCC and RP2 differentiate normally, extend an axon, and then die? One could think that the underlying cues that encode positional information in these cells and that tell them to die could equally be used for changes in the neuroblast lineages. There is no obvious advantage in differentiating a whole pattern of cells from which certain elements are then ablated. However, it could indicate that aCC and RP2 have essential roles aside from the obvious one of motorneurons controlling the contractions of specific muscles. In the case of aCC, the survival of its sibling pCC gives an example for this model: Both cells arise from the lineage of neuroblast 1-1 (BROADUS et al., 1995). Interneuron pCC has never been observed to undergo apoptosis in any of the abdominal or even more anterior segments; one can assume that it is required for whatever its function is in all segments. This prevents the system from ablating or altering the lineage to an extent that would lead to the loss of both aCC and pCC.

Of course this does not explain why aCC extends an axon before it dies. However, it indicates that “developmental constraints” (developmental pathways leading to its formation that cannot be overridden) or a “burden”, (the sum of all subsequent events or structures that rely on a feature) (LAUBICHLER, 2000), demand its conservation until later stages. One can see this differentiation of seemingly supernumerary, useless motorneurons as a consequence of developmental constraints or burdens in a diversifying system. Further studies investigating this aspect will have to characterise other features in the CNS that depend on the differentiation of aCC and RP2.

Most importantly for my own studies, the death of aCC and RP2 gave me a very obvious feature in the process of neuronal diversification. I used it as a handle to move from my first question (what mechanisms contribute to the diversification of the network) to my second one: What are the underlying cues that tell cells in certain segments how they have to behave? For example, what signal triggers the activation of a developmental pathway leading to cell death in the aCC of A7, whilst its equivalent in A6 differentiates and innervates a muscle? The drastic and relatively accessible event of cell death in two superficial and well-characterised neurons opened a way to investigating the role of positional cues in this process of diversification.


Figure 5.1: The RN2 pattern as a model for element reduction

RN2-gal4:CD8GFP expression highlights the motorneurons RP2 and aCC, as well as aCCs sibling, the interneuron pCC. RP2 innervates muscle DA2 segmentally (for instance, RP2 in A7 innervates DA2 in A7), whereas aCC innervates muscle DA1 intersegmentally from the next anterior segment of the VNC (for instance, aCC in A6 innervates DA1 in A7).

Since both muscles DA1 and DA2 are not present in A8, aCC in A7 and RP2 in A8 appear to be redundant. Therefore, the RN2 pattern provides a system to investigate the fate of supernumerary motorneurons.

Figure 5.2: Differentiation of a muscle founder cell for DA1

Antibody staining for evenskipped reveals the differentiation of a muscle founder cell and pericardial cells in all thoracic and abdominal segments except A8. This suggests that DA1 does not have an equivalent muscle in segment A8 (n > 10).

Figure 5. 3: The RN2 pattern in an embryo and a larva

The expression pattern of RN2 in a 13.5 hrs AEL embryo (A) and a first instar larva (B). RP2, aCC and pCC differentiate normally in A7 and A8 in the early embryo.

The same pattern in the larva, however, reveals the absence of aCC in A7 and RP2 in A8. The single cell shown in this picture in A7 is most likely a pCC interneuron.

Figure 5.4: Death of supernumerary motorneurons

Disintegration of supernumerary yet differentiated motorneurons: The RN2-gal2 driven expression of CD8GFP reveals the disintegration of nerve A8 (arrow) and the cell bodies of aCC in A7 as well as RP2 in A8 at this 14.5 hrs AEL embryo.

Figure 5.5: The enzyme caspase-3 as a marker for apoptosis

In this preparation of a 14.5 hrs AEL embryo with CD8-GFP expression in the RN2 cells, the localisation of caspase-3 highlights aCC in A7 and RP2 in A8. Thereby, it provides direct evidence for the apoptosis of these neurons (n > 8).

The intensity of the signal increases towards the posterior and is particularly strong in A8 and even more so in A9. In this respect, it looks very similar to a staining against AbdB. This shows that cell death occurs most frequently at the posterior segments.

Figure 5.6: evenskipped expression in the dorsal muscle field

evenskipped is expressed in the forming DA1 muscle in wildtype and the H99 apoptosis deficiency mutant: Shown here in a dorsal view of 16 hrs AEL embryo. There is no apparent difference between the formation of DA1 muscles in wildtype and H99 embryos, showing that apoptosis is not involved with the absence of DA1 in A8 (n > 8).

Figure 5.7: Overview of eve expression in the ventral nerve cord

evenskipped expression in the ventral view: The images are overviews, see the red outline boxes in high magnification in figure 5.8 (n > 8).

Figure 5.8: evenskipped expression in the ventral nerve cord

In this overlay of pictures, you can see the pattern of RP2, aCC and RP2 along with some interneurons in a wildtype and an H99 apoptosis deficiency embryo. U-neurons and other interneurons that are known to express evenskipped are not in focus. The embryos were fixed at 15 hrs AEL (n > 8).

Note that in the wildtype animal, aCC in A7 and RP2 in A8 show a fading of the evenskipped signal as well as the disintegration of one aCC and the lack of any signal for one RP2. In the H99 apoptosis deficiency mutant, there is no sign of cell death in any of the neurons, providing evidence for the role of apoptosis shaping segmental differences in the RN2 pattern.

return to "thesis overview"