4.1 Introduction

The characterisation of muscles and their respective motorneurons in the previous chapters led me directly to the task of relating these pairs of cells from A8/9 to pairs in the reference segment. This is ideally done through the identification of muscles and neurons in the reference segment that are equivalent to those in A8/9. Once this is achieved, one can ask whether the identity of muscles and neurons can be altered, how and to what extent; and whether the match between a muscle and its neuron is fixed in both segments or is open to re-arrangements.

In order to avoid engaging in a discussion about “homologues” with its many functional, phylogenetic and developmental implications (WAGNER, 1989), I am using the word “equivalent” to describe the relationship between muscles and neurons among the compared segments. I will base my claim for equivalence between two muscles or neurons on shared features, which I will describe in every individual case.

In muscles, these features include position, size, orientation, insertion sites in the epidermis, patterns of gene expression and innervation. In neurons, these features include position, dendritic and axonal projections, patterns of gene expression and targets. These features can be subject to dynamic changes in a developmental context and therefore, ought to be compared only at equivalent developmental stages. As in any comparative study, determination of equivalence gains weight the more features are shared among the compared segments. Of similar interest are features they do not share – if equivalence can be derived from other properties of the cells, such instances can point towards means of diversification.

4.2 Relating Muscles: Marker Expression Profiles

In muscles, I regard morphological information (position, size, orientation and insertion sites into the epidermis) as relatively unreliable indications of equivalence; the large changes in A8/9 morphology could easily account for passive changes in any of these features and thereby, disguise equivalence. I have discussed such a case in the apparent shift of dorsal muscles to the ventral side of the body in A9 (see Chapter 2). However, there are four muscles in A8 with almost completely or completely conserved anatomical features: VL1 to VL4 in A8 are very similar to those in the reference segment. The conserved morphology of their respective neurons (see Chapter 3) adds further support for claiming equivalence: Their dendritic projections look very similar to those of equivalent neurons in more anterior segments. This notion also implies that “posteriorness” alone does not appear to be sufficient for changes in the dendritic morphology. This supports the view that the grossly enlarged dendritic projections of aCC in A8 and RP2 in A9 (see Chapter 3) are indeed related to a change in the function of these neurons and their respective target muscles.

For all other muscles, more evidence is required. The identity of muscles can be reliably determined by their patterns of gene expression (reviewed in BAYLIES et al., 1998). This includes in particular the transcription factors evenskipped (FRASCH et al., 1987), Krüppel (GAUL et al., 1987; RUIZ-GOMEZ et al., 1997), ladybird (JAGLA et al., 1998), vestigial (WILLIAMS et al., 1991) and S59 (DOHRMANN et al., 1990; CARMENA et al., 1995); and the cell adhesion molecule connectin (NOSE et al., 1992). As an ongoing endeavour during my research, I progressively mapped the expression of these markers in the differentiated muscles of A8 and A9 at around 14 hrs AEL.

The findings of these experiments are summarised in the maps of figure 4.1 and in table 4.1. Images of the original specimens for the most relevant findings (profiles of LT1, DT1 and DO1 in A8; and TO1, TO2 and GS4 in A9) are given in figures 4.2 to 4.6. In the following, I will discuss the implications of these profiles concerning a possible equivalence.

Firstly, it is worth noting that the expression of vestigial in the VLs of A8 fits with their role as equivalents of other VLs in the reference segment. Furthermore, I could not detect the expression of any other markers, which again matches with more anterior VLs. The only exception for this is Krüppel, which I was unable to detect in any of the VLs; however, Krüppel expression in the muscles is dynamic and the signal in VA1 of A8 might have obscured the signal in the overlapping VLs.

Secondly, it has been reported (JAGLA et al., 1998) that there is no expression of ladybird posterior of the SBM between A7 and A8. This is interesting, since the position and orientation of muscle TT1 suggests that it could be the equivalent to the SBM. It runs transversally along the segment border between A8 and A9; however, there is no indication of ladybird expression in TT1, as one can see in the images published by (JAGLA et al., 1998).

Thirdly, and much more interestingly, evenskipped is expressed in muscle GS4. Originally, I had failed to detect signal in this muscle, which is easily missed due to the strong eve expression in cells surrounding the anal pad. The only eve expressing muscle in the reference segment is DA1, which receives its innervation intersegmentally from the motorneuron aCC (see Chapter 3). After I had found no sign of a DA1 in A8 and subsequently discovered the death of aCC in A7 (see Chapter 5), I noticed that aCC in A8 survives. Therefore, I asked whether there might be an evenskipped expressing muscle in A9. Staining in flat preparations revealed that muscle GS4 actually does express evenskipped. It is expressed in the most dorsal muscle from T1 to A7, suggesting that it might be a feature of the dorsal-most muscle. I take this finding as further evidence that the very ventral muscles of GS4, TO1 and TO2 are actually in the dorsal domain of A9 (as discussed in Chapter 2 with respect to the arrangement of nerve A9 and the targets of ventrally versus dorsally projecting motorneurons). Later I tested if the surviving aCC in A8 innervates muscle GS4 (see chapter 4.5).

There are other features of GS4 that it does not share with DA1: I could not find vestigial signal, whereas there is strong vestigial expression in TO1 and TO2. Nor could I detect any Krüppel signal. Dissections of RN2-Gal4:mCD8GFP and CQ-Gal4:mCD8GFP embryos showed that GS4 received innervation from at least one motorneuron of the RN2 pattern and at least one U-neuron (see figure 3.8 and figure 3.7a, Chapter 3). It is worth noting that both RN2 and CQ neurons express evenskipped. I regard GS4 as a dorsal muscle with no direct equivalent, which is innervated by at least two motorneurons, of which at least the one within the RN2 pattern can be easily compared with its equivalent from the reference segment. This made the identification of GS4’s neurons a priority (see 4.5 Relating Neurons: aCC and RP2).

Finally, the expression profiles of LT1, DT1 and DO1 in A8 allow some interesting conclusions. I could not detect any signal of Krüppel in DO1, whereas LT1 and DT1 show strong staining (see figure 4.5b). DO1 in the reference segment expresses Krüppel, but not vestigial. DO1 in A8 expresses vestigial, only. Generally, the marker expression profiles of DO1 are more like those of the smaller DA muscles of the reference segment (such as DA2), rather than the morphologically more similar DO muscles. From dissections of embryos of CQ-Gal4:CD8-GFP (LANDGRAF et al, 2003), I know that DO1 of A8 is innervated by a U neuron (see NMJs in figure 3.7, Chapter 3). I did not find this neuron through flipout experiments. Backfill experiments in combination with staining for evenskipped could resolve which of the U neurons innervates this muscle. In summary, I did not find a conclusive basis for determining equivalence between DO1 in A8 and any particular muscle of the reference segment.

The expression profiles of LT1 and DT1 in A8 are identical: they consist of connectin, Krüppel and S59. This matches with DT1 from the reference segment but contradicts my original hypothesis that LT1 might be an equivalent to an LT muscle from the reference segment. Initially, this hypothesis had been supported by the similarities in the basic branching pattern of the motorneuron for LT1 in A8 with the one of LT3 in the reference segment (see figure 4.7): The cell bodies of both neurons are located very laterally in the VNC and both neurons have large dendrites that are divided into three branches. However, the LT3 neuron is located in the posterior domain of the segment, whereas the cell body of the LT1 neuron lies more anterior. Therefore, I consider these two neurons not to be equivalent.

I try to avoid irrelevant details, but an evaluation of how cells relate to each other among different segments requires a lot of attention to a variety of features. The matching expression profiles between LT1 and DT1 in A8 and DT1 in the reference segment brought a new hypothesis into favour: That the two muscles in A8 are a “duplicated” variation of what is DT1 in the anterior segments.

One can speculate that a functional muscle could be quite easily generated by a single mutation that affected the segregation lineage in the mesoderm. Progenitor cell divisions that give rise to muscle founder cells are characteristically asymmetric. Each of these divisions gives rise to two different classes of cells with each sibling either maintaining or losing the expression of genes that were originally expressed by the parent progenitor (RUIZ-GOMEZ & BATE, 1997). This outcome of the progenitor division depends on an asymmetric distribution of the cytoplasmic protein Numb between the two daughter cells (RUIZ-GOMEZ & BATE, 1997; UEMURA et al., 1989; SPANA et al., 1995; RHYU et al., 1994). The cell receiving Numb maintains progenitor gene expression; its sibling that lacks Numb loses this expression.

Previous experiments (RUIZ-GOMEZ & BATE, 1997; CARMENA et al., 1998) demonstrated that during wildtype development, two ventral S59 expressing muscle progenitors divide sequentially to give rise to four cells, the founders for muscles VA1, VA2 and VA3; as well as the ventral adult precursor VAP. Numb is distributed asymmetrically in both progenitors. The generation of the four individually distinct cells requires the segregation of Numb to one cell in each pair of siblings. Therefore, after the division of the most dorsal progenitor, Numb is segregated to the VA2 founder and excluded from VA1. Similarly the division of the second progenitor generates a VA3 founder containing Numb and the VAP that lacks Numb.

In numb mutant embryos, both siblings of a division adopt the fate of the cell that does not receive Numb in the wild type. Therefore, two duplicated VA1s and two VAPs are generated. If numb is expressed ectopically in the mesoderm, the opposite fates are acquired by the two sibling cells of each pair: Two VA2s and two VA3s differentiate.

If a mutation seeded the formation of two DT1s in A8, the independent phenomena of muscle attachment and innervation could contribute to the maintenance of both of these muscles in the course of evolution. If this had been the case with LT1 and DT1, this would provide an example for a novel mechanism of muscular diversification.

Alternatively, one could argue that the matching expression profiles have functional reasons; this is supported by the notion that the two muscles are perfectly aligned and roughly in the same area (latero-dorsal). Therefore, it is likely that they function as a unit. In any case, I decided to focus on the “novel” muscle LT1 with its unusual motorneuron. Noting that this muscle expresses S59, I decided to use this marker to trace the differentiation of LT1 over the course of its development.

4.3 Relating Muscles: Development of S59 Positive Muscles

The development of the S59 expressing cells in the somatic mesoderm of the reference segment has been characterised in great detail (DOHRMANN et al., 1990; CARMENA et al., 1995). S59 is first expressed in a pattern of mesodermal cells towards the end of stage 11. These are muscle progenitors that divide and give rise to muscle founder cells or adult muscle precursors. The muscle founder cells form three clusters: I (2 cells) in the ventral, anterior region of the segment; II (4 cells) in the ventral, posterior region; and III (2 cells), in the dorso-lateral region of the segment (see figures 4.9 to 4.11).

By stage 12, the dorsal muscle progenitor of cluster III divides and gives rise to two muscle founder cells. One of these seeds the formation of muscle DT1, the other one probably seeds muscle DO3 (CARMENA et al., 1995). By stage 13, the muscle precursor of DT1 strongly expresses S59. The two progenitors of the ventral, anterior cluster II give rise to four cells: Muscle founder cells for VA1, VA2 and VA3; and one adult ventral muscle precursor. Of these cells, only the precursor for VA2 maintains expression of S59 beyond stage 13. In all other muscles, S59 expression ceases.

The single progenitor of ventral posterior cluster I gives rise to the muscle founder cells of LO1 and VT1. The founder of LO1 expresses S59 very strongly and migrates dorsally and to the anterior, crosses the segment border and seeds precursor formation. However, by stage 13, the expression of S59 in the LO1 precursor is very faint. The muscle founder cell for VT1 migrates slightly more dorsally and seeds the formation of a precursor. It is crucial to understand that VT1 forms in the ventral, anterior domain of the segment, in which its founder cell was born, whereas its sibling migrates to one segment further anteriorly.

In A8, the situation is different. By late stage 11, a cluster of four S59-expressing cells can be seen in the posterior, ventral part of the segment and a cluster of two cells in the anterior, ventral part. These cells are most likely the equivalent of cluster II and cluster I respectively. However, the pattern of migration and the differentiation into muscle precursors proceeds differently compared to the one shown for cluster II of the reference segment. It is likely that muscle VA1 in A8 is derived from this cluster; it maintains the expression of S59 beyond stage 13. There is also a cluster of two cells in the dorsal domain of A8, approximate equivalent to cluster III in the reference segment. A cell from this cluster seeds the formation of DT1 in A8. Two more muscles express S59, one in A8, one in A9, but their identity is unclear and they are labelled with question marks in the figures. From its position and orientation, the muscle in A8 might be muscle LL1 or VL1. In A9, I found S59 expression exclusively in the neuroectoderm.

Looking at the anterior cluster I in A8, one can see that the progenitor divides, as previously described at late stage 11. This is followed by the migration of a very darkly stained muscle founder cell across the segment boundary and into the lateral domain of A7, where it seeds the formation of LO1. Its sibling, which seeds the formation of VT1 in the reference segment, remains in A8 and migrates dorsally. There it seeds the formation of the original objective for these observations: muscle LT1.

One can now ask whether VT1 in the reference segment and LT1 in A8 are therefore equivalents. In this instance it is obvious why I refrain from using the term “homologue” to describe the relationship between these muscles: some features are shared, others are not. Both VT1 and LT1 are derived from equivalent lineages and muscle founder cells; both muscles are transverse. They are different in their position (ventral versus lateral); size (approximately 5 myoblasts contribute to VT1 as against approximately 10 that contribute to LT1; n>10); and to some extent gene marker expression (LT1 expresses Krüppel and connectin, whereas VT1 does not appear to).

This covers lineage, orientation, position, size and patterns of gene expression. From a developmental perspective, I regard VT1 and LT1 clearly as equivalents at least up to the stage of myoblast recruitment. It is not fruitful to argue about homologies, yet most interesting to explore how two elements of a network starting as identical units diverge over the course of their specification to finally serve two different functions. One of the features that I have outlined in the introduction to this chapter has so far been neglected: the innervation. In terms of different functions, the next logical step was to investigate the neurons that control these two muscles.

4.4 Relating Neurons: VT1, LT1 and the Role of the Transverse Nerve

The transverse nerve (TN) contains the axons of only two motorneurons per hemisegment: A posterior one that branches together with its equivalent in the other hemisegment; and an anterior one. The posterior TN neuron innervates the alary muscle dorsal of the SBM and is located at the same level as pCC. The anterior TN neuron innervates muscle VT1 and is on the same level as RP2 (see figure 4.17 A; MACLEOD et al., 2003). These motorneurons project their axons to the midline, exit the VNC dorsally and grow along the posterior segment border (GORCZYCA et al., 1994; CHIANG et al., 1994; MACLEOD et al., 2003; LANDGRAF et al., 1997). The external VT1 is the only somatic muscle innervated from the next anterior segment through the transverse nerve (TN); however, the nerve extends beyond this muscle to the lateral domain of the segment (see figure 3.15, Chapter 3; figure 4.18). On the posterior edge of the SBM, two sensory neurons join the transverse nerve: the lateral bipolar dendrite neuron (LBD) and the tracheal dendrite neuron (TDN) (GORCZYCA et al., 1994; see figure 4.18).

Muscle VT1 is specific to abdominal segments A2 to A7 (see figure 2.4 in Chapter 2). Its equivalent muscle LT1 in A8 has a prominent neuromuscular junction and receives its innervation from a motorneuron with very elaborate dendritic branches (see Chapter 3), which does not extend its axon through the TN. Through flip-out experiments, I never found a cell resembling a VT1 neuron in A7. However, as I have shown in Chapter 2, there is clearly a transverse nerve in exiting A7 and running along the A7/8 border. Therefore, I wondered whether, despite my findings with the flip-out approach, an anterior VT1 neuron existed in A7, for example a cell, which, like the A7 aCC, was eliminated by apoptosis.

In 13.5 hrs AEL embryos, one can see a transverse nerve exiting the VNC from T1 to A7; at around 14 hrs AEL, the TNs of T1 to T3 progressively disintegrate and finally do not project into the muscle field at around 15 hrs AEL (data not shown). As the VNC contracts progressively during this time, the mesodermal cells on the dorsal surface that are part of the TN delaminate and the nerves get “pulled” to the posterior.

To investigate whether there is a VT1-like neuron in A7 and in the thoracic segments, I performed retrograde labelling experiments by directly applying a DiI crystal to the transverse nerve of A8, of the thoracic segments and of the reference segment (see Chapter 8 for Material and Methods). Such a treatment often leads to the labelling of the posterior TN neuron in both hemisegments, because the posterior neuron is branched and therefore, both cells have axons exiting on each side.

To avoid labelling a cell that was assigned to die, I first carried out these experiments as late as I was technically able (around 15 to 15.5 hrs AEL). This is well after aCC in A7 and RP2 in A8 have started to disintegrate. In most specimens, they have completely disappeared by this stage. By using embryos of the RN2-Gal4 line crossed to UAS-GFP, I had reliable landmarks to discriminate between the anterior TN neuron (aligned with RP2) and the posterior TN neuron (aligned with pCC).

In A7 (from where the TN of A8 is derived), I labelled one motorneuron with a cell body at the same anterior-posterior level as pCC and dendritic projections at the level of the anterior commissure (see figures 4.15 and 4.17; n > 10). This identified it as the “posterior TN neuron” innervating the alary muscle as previously described (MACLEOD et al., 2003).

In the reference segment (anterior of A7), I labelled two motorneurons (see figures 4.15 and 4.17; n > 10). One motorneuron with a cell body at the same anterior-posterior level as pCC and dendritic projections to the anterior. This identified it as the “posterior TN neuron” innervating the alary muscle as previously described (LANDGRAF et al., 1997; LANDGRAF et al., 2003; MACLEOD et al., 2003). The other motorneuron had a cell body at the same level as RP2. This identified it as the “anterior TN neuron” innervating muscle VT1 as previously described (MACLEOD et al., 2003).

In the thoracic segments, I applied the DiI crystal to the already disintegrated transverse nerve. By this stage, the TN did not project into the muscle field, although the nerve endings were still in place on the dorsal surface of the VNC. The dye revealed a motorneuron with a cell body at the same level as RP2 and with dendritic projections at the same level as the anterior commissure. Based on its position, I regard this neuron as being equivalent to the anterior TN neuron of the reference segment, where it innervates muscle VT1 (n = 3). I did not label any posterior TN neurons, but this could be due to the small number of specimens. The absence of alary muscles in the thoracic segment also implies that the respective neuron might not differentiate in the first place.

My interest now focused on the apparent “absence” of the anterior TN neuron in A7. In order to see whether it differentiated and extended an axon earlier in development, I repeated the experiment using younger embryos (13.5 hrs AEL or younger; preceding the death of aCC in A7). In most specimens, I succeeded in labelling only the posterior TN motorneuron (n = 8). Thus, I can conclude that there is probably no motorneuron extending its axon through the TN into A8 besides the posterior one that innervates the alary muscle (see discussion below).

I then decided to investigate the neuromuscular junctions formed by axons of the TN in the reference segment and in A8 at 15.5 hrs AEL using DiI. In the reference segment, the motorneuron axons project into the muscle field along the segment border. At this stage, the most distal part of the nerve extends to the level of VL1 and the SBM (see figure 4.17; n > 20).

In A8, the nerve projects into the muscle field along the segment border, too. The most distal part extends to the level of VL3, which is approximately where one would expect a muscle VT1. In most specimens examined (7 of 9), the nerve bifurcated. Since I was not able to find a NMJ in any segment at this stage, I decided to investigate specimens later in development.

I repeated my observations with glue preparations of first instar larvae (for methods see Chapter 8.9). I was not able to preserve the cuticle of segment A8 or the TN for this segment. However, labelling the TN in the reference segment revealed the VT1 neuron (n = 4). I could trace the axon to the periphery and identify a neuromuscular junction on VT1. In two specimens, the dye also revealed the posterior TN motorneuron (alary muscle) in the same hemisegment with its cell body posterior of the VT1 neuron.

In order to investigate the patterns of innervation of the two TN motorneurons in greater detail and at later stages, I examined the projections of motorneurons in third instar larvae of the OK371-Gal4 line. The arrangement of muscles with respect to features of the cuticle such as denticles is fixed. Therefore, denticles can serve as landmarks to determine the position of neuromuscular junctions from the OK371 pattern. Using the G203 2C12144 line with GFP:titin (expresses GFP in the muscles), I determined the position of VT1 with respect to the denticles (see fig. 4.18a). Then I localised this position in a OK371:CD8GFP third instar larva (expresses GFP in motorneurons and allows one to trace axons) and found NMJs on VT1 derived from the TN (see figures 4.18b and 4.18c). I could not find any NMJ ventral of the SBM derived from the TN in A8 (see figures 3.15 and 3.16 in Chapter 3).

In the reference segment, the TN nerve continues to grow beyond the region of VT1. It extends to the region of the LBD neuron without forming a detectable NMJ. Interestingly, the TN in A8 projects slightly more dorsally: it ends dorsal of the NMJ of the LT1 neuron (see figure 3.15, Chapter 3). From these observations, I draw the following conclusions for larval TNs: There are two motorneurons per hemisegment in the TN of A1 to A7, of which the anterior one innervates VT1, the posterior one the alary muscle (as described in MACLEOD et al., 2003). There is one motorneuron per hemisegment in the TN of A8, which is the posterior one that innervates alary muscle. The very dorsal extension of the TN of A8 suggests that the alary muscle of this segment is probably located more dorsally than in the reference segment.

To compare the motorneuron innervating VT1 in the reference segment and the LT1 neuron in A8 genetically labelled with mCD8:GFP, I repeated the flipout experiments described in Chapter 3. This time, I specifically looked for the motorneuron of VT1. I found a total of three specimens in which I could identify this cell based on two features: Its axon exits the VNC through the TN; and its cell body is anterior of its dendrite. Figure 4.16B shows a fourth specimen provided by Annemarie North. It looks identical to the neurons I found; however, the quality of the specimen is significantly better in terms of its general preservation and the lack of background from labelled cells in nearby segments. In comparison with the previously described LT1 neuron, neither cell body position nor dendrite morphology or the axonal projection bears similarities. An equivalence of the two neurons is therefore extremely unlikely.

The fate of the anterior TN neuron of A8 is unclear. There are several possibilities: The neuron could never differentiate in this segment; it could fail to extend an axon through the TN; it could die even before the TN differentiates; or it could differentiate into a completely different motorneuron that extends its axon through a different nerve. With the aid of NB markers or labelling techniques, it seems conceivable that the tracing of the lineage could finally reveal the answer. For my work, however, I reject the hypothesis that there is a skeletal muscle in A8 that receives its innervation from the TN, based on the (negative) evidence presented above.

I showed in Chapter 3 that LT1 in A8 is innervated by a neuron that does not project through the TN. Tracing the nerves of OK371:CD8GFP larvae showed that VT1 in the reference segment and LT1 in A8 are most likely innervated by only one excitatory motorneuron each (see figures 3.15 and 3.16 in Chapter 3). In the light of the developmental equivalence of these two muscles determined earlier in this chapter, these findings have interesting implications for the understanding of neuron/muscle relationships (see 4.6 Discussion).

4.5 Relating Neurons: aCC and RP2

I have described the pattern of the RN2-Gal4 line in detail in Chapter 3. In brief, the motorneuron aCC is located with its sibling interneuron pCC, just anterior to the segment border of the metamere it originated in, into which it projects intersegmentally and innervates muscle DA1. For example, motorneuron aCC in A6 innervates muscle DA1 in A7. The second motorneuron in the pattern, RP2, segmentally innervates - among others - muscle DA2. For example, motorneuron RP2 in A6 innervates muscle DA2 in A6 (see figure 5.1 in Chapter 5). Muscles DA1 and DA2 do not have obvious equivalents in A8 and the “supernumerary” neurons aCC in A7 and RP2 in A8 differentiate and then die (see Chapter 5).

Posterior of the dying RP2 in A8 at approximately 14 hrs AEL, there is a pair of cells that I take to be aCC/pCC in the same segment; there is also an RP2 in A9 and two to three unidentified cells per hemisegment at the posterior-most margin of A9 (aCC, pCC and RP2 were identified in the RN2 pattern, see figure 3.3, Chapter 3). I had found that muscle GS4 (which expresses evenskipped, a marker for the most dorsal muscle) alongside with TO2 (which expresses vestigial, a marker for longitudinal muscles) receives innervation from a motorneuron of the RN2 pattern (see figure 3.8, Chapter 3). There is also a contribution of the RN2 pattern to the hindgut nerve, which I have resolved later in this chapter.

TO2 and GS4 have obviously different features from DA1 and DA2 in the reference segment, which are innervated by aCC and RP2 motorneurons. Note that GS4 is also innervated by a U-neuron of A8 (see Chapter 3, figure 3.8). Since aCC, RP2 and U1 motorneurons can be reliably identified as equivalents at 14 hrs AEL among different segments, the RN2 and CQ line provided me with the ideal model to explore morphological changes of motorneurons accompanying alterations in the muscle field.

I used flies of the RN2-flp, tub-frt-stop-frt-Gal4, UAS-mCD8GFP line (see PIGNONI & ZIPURSKI, 2001 for tubulin-CD2 stock; Annemarie North, unpublished) to label single RN2 neurons in third instar larvae. In this line, the gene for the flipase is placed under the control of a weak evenskipped promoter. I exposed embryos at a variety of stages to 29ºC for 24 hours and raised the larvae at 25ºC to third instar (see Chapter 8, Material and Methods). Flip-out events occurred only in a subset of cells in the RN2 pattern, leading to Gal4 expression under the control of the Tub-promoter. I dissected larvae in which single cells expressed GFP and then stained the CNS for evenskipped as a reference marker, as well as GFP.

In this way I identified aCC in A8 (n = 7; see figure 4.19) and RP2 in A9 (n = 6; see figure 4.20). The aCC in A8 has two symmetrical dendritic branches, one ipsilateral, one contralateral and both originating from the cell body. They extend far to the anterior, covering much of the neuropile in A8 and probably reaching into A7. The RP2 in A9 has a dendrite that projects to the anterior and then bifurcates into an ipsilateral and a contralateral branch. Just as in the aCC, these two branches are symmetrical and cover much of the neuropile in A8. In both these neurons, the dendritic projections are strikingly different from those of their equivalents in the reference segment (see figure 4.21 for direct comparison; and the discussion of this chapter for implications).

The two or three neurons posterior of RP2 in A9 extend axons to the hindgut (see figure 3.8, Chapter 3) but have no detectable dendrites (data not shown). Looking at my specimens of aCC in A8, I realised that a neuron I had identified through the flipout experiments of Chapter 3 and matched with TO2 was actually an aCC with strongly altered dendrites (see figure 3.12g in Chapter 3).

Since phalloidin injections failed to reveal the most posterior muscles of A9, I could not confidently match aCC in A8 and RP2 in A9 individually with their target muscles GS4 and TO2. To identify the targets confidently, I filled these neurons with the intracellular dye Lucifer Yellow in 14.5 hrs AEL embryos (see Chapter 8, Material and Methods). I traced the axonal projection of aCC in A8 to muscle TO2 (n = 7; see figure 4.22a) and RP2 in A9 to muscle GS4 (n = 5; see figure 4.22b).

4.6 Discussion

I began this chapter by outlining the idea that elements of the neuromuscular network can share particular features at different stages of their development. Depending on the degree of resemblance, one can consider certain iterative elements to be equivalent and investigate their role in the locomotor network, their partners in this network, and their developmental fate.

By tracing the lineage of the S59 positive muscle progenitors and their progeny, I could determine the equivalence of the muscle founder cells of VT1 in the reference segments and LT1 in A8. These two muscle founder cells fuse with specific numbers of naïve myoblasts to form two distinct types of muscle precursors. I investigated the cues underlying this diversification in Chapter 6. More detailed information about how the recruitment of certain numbers of myoblasts is regulated in a segment-specific manner is of great interest. It could help to understand the interaction of genes that are involved with myoblast recruitment and the underlying spatial cues on a molecular level.

I showed in this and the previous chapter that VT1 in the reference segment and LT1 in A8 are innervated by two different motorneurons. Not only do they have very distinct dendritic arbours, they also project to their target muscles through different nerves. This demonstrates that the match between a neuron and its target muscle is not fixed among all segments and raises the question of what features have to change in order to accomplish a switch in the innervation.

The current understanding how the match between a motorneuron axon and its specific target muscle is achieved is based on two models that are not mutually exclusive: The “relative balance” (WINBERG et al., 1998) and the “lock-and-key” model (HOANG & CHIBA, 1999) both predict the existence of molecular interactions between motorneurons and muscles, as well as repulsive and attractive cues that modulate those. Several studies have demonstrated that many genes affecting motor axon pathfinding are widely expressed by muscles and motorneurons: Alterations in the levels of cell surface proteins such as Fasciclin II in muscles can lead to a biased innervation of specific muscle fibres (DAVIS et al. 1997). This supports the “relative balance“ model, in which the final step in target recognition would be controlled by the overall balance of attractive and repulsive cues mediated by a combination of cell adhesion molecules (CAMs) and other guidance cues.

The “lock-and-key” model, on the other hand, proposes a complementary code of molecular cues expressed by the motorneuron and its target muscle that ensures a precise match during the process of target recognition. Support for this model comes from NetrinB, which is expressed in only three muscles: VL3, VL4 and DA2. Motorneuron RP3 innervates specifically muscles VL3 and VL4, but fails to do so in 35% of the cases in NetrinB mutants (MITCHELL et al., 1996).

In this context it is worth noting that LT1 in A8 expresses connectin, which has also been shown to play a role in target finding (NOSE et al., 1994), whereas VT1 in the reference segment does not. Similarly, Krüppel is expressed in LT1, but not in VT1. If an evolutionary change in the expression profile of muscle marker proteins had occurred at the level of muscle founder cells, this would have determined a different identity, and could have been accompanied by a new neuron finding this novel target. This could be a neuron that is supernumerary in anterior segments where it dies at later stages in development.

If the lineage of the LT1 motorneuron in A8 was known, one could investigate whether such supernumerary neurons differentiate in the reference segment. A network that arises not only from a range of muscles that one particular motorneuron has to choose from, but also from a range of motorneurons that can be attracted by a particular muscle would allow dynamic re-matching through mutations leading to altered identities of either motorneurons or muscles.

With respect to the VT1 innervation, one can ask why the VT1 neuron in the thoracic segments survives until about 15.5 hrs AEL and maybe even well into larval life. The labelling was done on disintegrated nerves that did not reach into the muscle field anymore and at a time when aCC in A7 and RP2 in A8 have already disintegrated. Yet there is no obvious difference in this neuron’s dendrite or cell body and its equivalents in abdominal segments, where it forms a functional NMJ. In the thorax, this neuron could act as a pioneer and die at much later stages.

With respect to alterations in motor neuron identity, I approached the neuromuscular network from the opposite angle. Instead of looking for the neurons innervating muscles that are derived from equivalent founder cells, I investigated the development of motorneurons aCC and RP2 that innervate different muscles in different segments. Until approximately 14 hrs AEL, the aCC and RP2 motorneurons of all abdominal segments look identical. Within the next hour, however, the supernumerary aCC in A7 and RP2 in A8 disintegrate and die (see Chapter 5). The neurons that started as aCC in A8 and RP2 in A9, however, differentiate into two very distinct neurons that make contact with target muscles in A9. However, these muscles are different from DA1 and DA2, which are normally innervated by aCC and RP2. In larval stages, aCC in A8 and RP2 in A9 have bilateral dendrites that extend far to the anterior.

These distinct features of aCC and RP2 dendrites could be explained through passive alterations due to the over-all changes in this segment’s morphology. Alternatively, one could hypothesise that the death of aCC in A7 and RP2 in A8 leaves a “gap” in the neuropile into which neurons of more posterior segments can project dendrites. This would mean that the presence of a cell in one segment would create a repulsive signal at least for its equivalents from other segments. Previous experiments showed that neighbouring neurons do not extend axons into the domain of a dead neuron of a different class (Matthias Landgraf, personal communication). However, this applies to neurons of a different class and not to equivalent neurons from other segments that are more likely to respond to the same attractive and repulsive cues in the neuropile. In this context, it is worth noting that branches of dendrites do occasionally expand beyond the segment in which they normally branch. Thereby they span up to three segments even with their equivalents developing normally and occupying this domain in the neuropile (Matthias Landgraf, personal communication).

To test this hypothesis, either the flipout system or laser ablation could be used to selectively eliminate neurons from the RN2 pattern early in development and investigate whether other (equivalent) cells take over their dendritic domains. An alternative explanation for the altered dendrite morphology of aCC in A8 and RP2 in A9 is their “different” function with respect to equivalents from more anterior segments. Because they innervate two muscles with a different position, size and orientation than their equivalents in the reference segment, one might expect that these pairs of neurons and muscles play a different role in locomotion. This is likely to require specific synaptic input and an appropriate dendritic morphology to accommodate this.

A different issue that applies to both muscles and neurons is the one of identity. Are muscles VT1 in the reference segment and LT1 in A8 homologues? Is aCC in A8 still an aCC? These examples show how important it is to “study things in the process of development from the beginning” (ARISTOTELES, 360 BC). The cells described start as identical entities that differentiate into morphologically and functionally very distinct elements of the neuromuscular network depending on their environment (A8/9 versus the reference segment). The question as to whether the resulting structures are still homologues or not is a rather semantic issue and will not be discussed in further detail. It is more important to reveal the molecular and cellular basis of the developmental pathways that lead to the diversification of originally identical elements of the networks. This leads directly to the basis of segmental identity, which I investigated in Chapter 6.

For such investigations, the RN2-Gal4 line has provided an exceptionally rewarding system. Within the three segments of A7, A8 and A9, I could describe three different muscle patterns. This is accompanied on the level of central neurons by all three principal ways of diversification, as I have defined them in Chapter 2: The RN2 motorneurons for A7 differentiate the same way as they do in more anterior segments (conservation of elements); the RN2 motorneurons for A8 die postmitotically (removal of elements); and the RN2 motorneurons for A9 develop new features (differentiation of novel elements). The small number of cells contained in the RN2 line, their accessibility for mechanical manipulation and the detailed knowledge about the specification of dorsally projecting motorneurons makes this system particularly attractive to study the mechanisms of diversification.

Investigating the development of U1 neurons in A6, A7 and A8 could also lead to interesting findings. U1 in A6 innervates muscle DO1 in A7; U1 in A7 most likely the much bigger muscle DO1 in A8; and U1 in A8 probably muscle GS4. This system, in which one finds a conserved, a slightly different and a drastically different target could also provide insights into the process of neuromuscular diversification.


Figure 4.1: Expression of muscle markers

The expression profile of marker proteins, such as transcription factors or cell surface proteins, gives a unique identity to groups of muscles or even individual muscles in the reference segment. In A8/9, they can provide valuable cues for evaluating the relationships among muscles of different segments.

This figure gives a diagrammatic summary of the expression profiles of connectin (cell surface protein); evenskipped, Krüppel, ladybird, S59 and vestigial (transcription factors) in approximately 15 hrs AEL embryos. These results are derived from a large number of specimens, each set of experiments: n > 10. This applies also to the figures that show staining in the actual cells (fig. 4.2 to 4.6).

Table 4.1 Summary of the expression profiles of muscle markers

Summary of the expression profiles of connectin (cell surface protein); evenskipped, Krüppel, ladybird, S59 and vestigial (transcription factors) in approximately 15 hrs AEL embryos. In addition to the diagram in figure 4.1, this table contains references to positive, but not identified muscles.

Figure 4.2: Expression pattern of connectin

Expression of connectin in a flatprep (~15 hrs AEL). The yellow arrow points to DT1, the blue one to LT1. Slide provided by Matthias Landgraf.

Figure 4.3: Expression of vestigial in A8/9

Expression of vestigial in a flatprep (~15 hrs AEL). PS marks the posterior spiracles. Note that the TO2 muscles (white arrows) are torn as they usually do in flat preparations.

Figure 4.4: Expression of vestigial in A8/9

Expression of vestigial in a whole mount embryo (~15 hrs AEL). PS marks the posterior spiracles, the black arrow indicates the midline. The nuclei of DO1 are clearly visible (red dotted circles).

Figure 4.5: Expression of S59 and Krüppel

Expression of S59 (A) and Krüppel (B) in whole mount embryos (~16 hrs AEL). The yellow arrow points to DT1, the blue one to LT1. Note that the S59 specimen is shown in a dorso-lateral view, the Krüppel one in a dorsal view. PS = posterior spiracle.

Figure 4.6: Expression of evenskipped

Expression of evenskipped in a flatprep (~15 hrs AEL). Note the expression in DA1 of A6 and A7 and strong expression on the margin of the anal pad in A8. Arrows indicate eve expression in two parallel GS4 muscles of the two hemisegments.

Figure 4.7: Similarities in neuronal morphology

LT3 neuron for A7 (A) in a phalloidin background (red) and the LT1 neuron for A8 (B) in a background for evenskipped (red). The cell bodies do not lie in the same antero-posterior region of the neuropile, however, the dendritic projections of the two neurons share some similarity. Both are divided into three major branches (not visible in the two dimensional z-projections) that project over two or even three segments (LT1 for A8).

Figure 4.8: Lineage of S59 positive cells in the mesoderm

This diagram summarises the formation of S59 positive muscles in the segments A6, A7 and A9.

(A) S59 positive muscle progenitors give rise to muscle founder cells (MFCs) in three clusters (I, II and II). At stage 12, these clusters are the same in all abdominal segments from A1 to A7. In A8, a cluster of four cells similar in their position to cluster II expresses S59.

(B) The two muscle founder cells of cluster I migrate; MFC Ib migrates into the lateral domain of the segment it originates from, whereas MFC Ia migrates to the anterior and crosses the segment border.

(C) MFC Ia seeds the formation of muscle LO1 in the reference segment. MFC Ib seeds the formation of muscle VT1 in the reference segment, and LT1 in A8. Other muscles in the reference segment that are derived from S59 expressing founder cells include DT1 and VA2.

Muscle VA1 in A8 is likely to be derived from the cluster ~II of A8 and is probably an equivalent to muscle VA2 from the reference segment. Muscle DT1 in A8 strongly expresses S59 from around 13 hrs AEL. The observations that led to this diagram were made in a set of experiments with n > 10. This applies to the following fig. 4.9 to 4.14.

Figure 4.9: Differentiation of S59 positive cells

Over the course of germband retraction, S59 is detectable in three clusters of cells: A ventral anterior cluster I, a ventral posterior cluster II, and a dorsal cluster III. In A8, only clusters I and II can be detected at this stage. In A9, only neuroectodermal cells are S59 positive. The black arrow indicates the midline.

Figure 4.10: Migration of S59 positive Muscle Founder Cells

Towards the end of germband retraction, individual muscle founder cells within the clusters differentiate and start to migrate. For example, cluster I is clearly discernible due to its size and intense S59 staining (see orange arrow).

Figure 4.11: Migration of S59 positive MFCs

In this specimen, one can see an overview of the final position in which the MFCs of the three clusters of S59 positive cells will start to recruit myoblasts for fusion in the reference segment: the anterior, ventral cluster Ib will give rise to VT1 (note that in some anterior segments, myoblasts are visible); the darkly stained MFC of Ia that will give rise to LO1 is in the lateral domain; the ventral, posterior cluster II consists of four cells; and the dorsal cluster III that will give rise to DT1.

In A8, MFC Ib has migrated to the anterior, lateral domain of the segment where it will seed the formation of muscle LT1. In the ventral, posterior cluster II, four MFCs are clearly labelled.

Figure 4.12: Ia has migrated to the next anterior segment

Just before muscle founder cells start to recruit myoblasts for fusion, MFC Ia (orange arrow) that will give rise to muscle LO1 has crossed the segment border and is now in the next anterior segment. The sibling MFC Ib that will give rise to LT1 in A8 has migrated to the lateral domain (compare with the more ventrally located Ib of A7).

In this specimen, there is one cell associated with Ib in A8 with an uncertain identity (black arrow). It is likely to be a myoblast fusing with the aligned Ib founder cell.

Figure 4.13: S59 signal in differentiating muscles

S59 positive syncytia outline the forming muscles of DT1, LO1, VT1 and VA2 in the reference segment and LT1 as well as an unidentified muscle ventral of LT1 in A8. Note that at this stage, there is no indication of S59 signal in DT1 of A8; the signal can be detected later than in the DT1 of the reference segment.

Figure 4.14: S59 in fully differentiated muscles

Fully differentiated muscles with S59 signal reveal DT1, LO1 faintly), VT1 and VA2 in the reference segment. In A8, muscles DT1 and LT1 are stained as well as an unidentified muscle ventral of LT1 and one unidentified muscle in A9.

Figure 4.15: Fills of the embryonic Transverse Nerve (TN)

DiI (red) fills of the transverse nerve (TN) in an approximately 15 hrs AEL embryo (post cell death of aCC in A7 and RP2 in A8). RN2 background provides landmarks (green).

(A) Overview of the specimen of which figures C and D were taken (see yellow outline).

(B) DiI fill of the transverse nerve in the reference segment. Nerve endings reach out to the muscle field. The neuron is aligned with RP2, which identifies it as the anterior TN neuron that innervates VT1.

(C) DiI fill of the transverse nerve in the thoracic segment T3 (initially innervates A1). By this stage, the nerve has already disintegrated and does not enter the muscle field anymore. Here, too, the neuron is aligned with RP2, which identifies it as the anterior TN neuron that innervates VT1.

(D) DiI fill of the transverse nerve in A7 (innervates A8). Two neurons were labelled in this specimen, as is often the case with the TN. Note that aCC in A7 and RP2 in A8 have already disappeared. The cell body is approximately aligned with pCC, therefore, it is more posterior than the labelled neurons in figures B and C. This identifies it as the posterior TN neuron, which innervates the alary muscle.

Figure 4.16: Fills of the larval Transverse Nerve (TN)

DiI (red) fills of the transverse nerve (TN) in first instar larva (glue preparations). RN2 background provides landmarks (green). Note the strong GFP expression in aCC of A8 (white arrow in the top image).

Note also that the segment boundaries in the larva do not follow straight lines – the guidelines given as label for the segments are based on the position of aCC and RP2.

(A) DiI fills of the transverse nerve in the reference segment. In the larva, two neurons per hemisegment can be discerned: one posterior (innervating the alary muscle, yellow arrow) and one anterior (innervating muscle VT1 in the next posterior segment, blue arrow). Note that in this specimen, two TNs (A5 and A6) were filled.

(B) Flip-out clone of the VT1 neuron in a 3rd instar larva. The dashed line indicates the midline. Specimen provided by Annemarie North.

(C) Neuromuscular junction of the anterior motorneuron of the transverse nerve on muscle VT1 – the approximate outline of this muscle is indicated by a dotted line. The muscle is difficult to see in a glue preparation, since it is covered by VL and VO muscles. NMJs are indicated by the arrows.

Figure 4.17: Projections of the embryonic Transverse Nerve (TN)

DiI (red) fills of the transverse nerve (TN) in an approximately 13.5 hrs AEL embryo (4.17 A) and a 15 hrs AEL embryo (4.17 B-D). RN2 background provides landmarks (green).

(A) Reference Segment: A fill of a young (13.5 hrs AEL) embryo in the TN of A7 (cell bodies in A6) reveals one anterior TN neuron (blue arrow) that innervates muscle VT1; as well as the posterior TN neurons that innervate the alary muscles (one in each hemisegment; yellow arrow).

(B) A8/9: The TN of A8 (cell bodies in A7) grows along the segment border. The specimen shows only the posterior TN neurons – this applies to both young (approximately 13.5 hrs AEL embryos; n=8) and older (post aCC death, 14.5 hrs AEL and older; n>10) specimens.

(C) The TN of A7, growing along the segment border A7/8 in a different specimen. Note the bifurcation.

(D) The TN of A4, growing along the segment border A4/5 as an example for the TN in the reference segment. The nerve does not bifurcate; no obvious sign of a neuromuscular junction can be seen.

Figure 4.18: Innervation of VT1 through the TN

(A) Muscle VT1 (outlined in blue) is highlighted under UV light in a third instar larva of a line that expresses GFP associated with the muscle protein titin. Its position with respect to the denticle bands is fixed.

(B) and (C) The same region in a third instar larva of a line that expresses mCD8-GFP in all motorneurons (OK371), thereby, highlighting their axons and neuromuscular junctions in the periphery. Brightfield in (B), UV light in (C).

The transverse nerve is clearly visible and forms a neuromuscular junction (C; yellow arrows). It continues growing dorsally, as expected since there are at least two neurons contributing to the TN in the larva (blue arrow).

The position of the neuromuscular junctions corresponds with the position of VT1, as determined through the denticle bands as landmarks. The VA and VO muscles are also visible in this image, shining through from underneath the focal plane.

Figure 4.19: Dendritic morphology of aCC in A8

In order to visualise individual cells of the RN2 pattern, virgins of a RN2-flp, tub-frt-stop-frt-Gal4, UAS-mCD8GFP were used. In this line, the gene for the flipase is placed under the control of a weak evenskipped promoter. If embryos are raised at 29ºC, flip-out events occur only in a subset of cells in the RN2 pattern, leading to Gal4 expression under the control of the Tub-promoter.

Motorneuron aCC can often be recognised even in the full pattern, due to the “gap” that the death of aCC in A7 and RP2 in A8 leaves. In the specimen shown, evenskipped staining (red) provided additional landmarks to identify aCC in A8. The dashed line indicates the midline. The specimen shown was a third instar larva.

Figure 4.20: Dendritic morphology of RP2 in A9

Protocol as above. In the figures shown above, evenskipped staining (red) provided information on landmarks to identify RP2 in A9. The dashed line indicates the midline. The specimen shown was a third instar larva.

Figure 4.21: Direct comparison of aCCs and RP2s

A direct comparison between RP2 in the reference segment (A) with RP2 in A9 (B); and between aCC in the reference segment (C) and aCC in A8 (D).

The dashed line indicates the midline. All specimens are third instar larvae.

Figure 4.22: Fills of aCC in A8 and RP2 in A9

Intracellular dye fills with Lucifer Yellow (LY) allow one to trace the axonal projections of individual neurons from the CNS to the target muscle (green). A background of RN2 (red) provides positional information as a landmark and proves the identity of the filled cell. The specimens are flat preparations of 14 to 14.5 hrs AEL embryos.

(A) Fill of aCC in A8: The axon projects through nerve A9 to muscle TO2 (n = 7).

(B) Fill of RP2 in A9: The axon projects through nerve A9 to muscle GS4 (n = 5).

return to "thesis overview"