1.1 Diversifying Networks in Development and Evolution
Every behaviour requires machinery able to perform movement. In insects, the
most fundamental unit of this machinery consists of two cells, one muscle and
one neuron controlling its contraction. This unit is repeated and the
repetitions are arranged into patterns and embedded into a locomotor network.
Such networks primarily consist of the following elements: structures in the
exoskeleton that provide friction on the substrate; muscles that are anchored to
the exoskeleton through tendon cells; motorneurons located in the central
nervous system that extend axons to these muscles; interneurons that are
presynaptic to the motorneurons and include central pattern generating circuits;
and sensory neurons that transmit proprioceptive feedback and information on
The way in which these elements are arranged and differentiated reflects the
function that they serve. Selection pressure acts most immediately on the
outer-most elements of the network, the morphology of the exoskeleton and the
muscle pattern, since this is where the organism meets its environment.
Functional and structural adaptations in the periphery have to be matched with
respective neurons, and so the neuromuscular networks of insects have evolved
manifold variations. This reflects the diverse habitats in which different
species of insects live and the specialised behaviours they have to perform in
these environments in order to survive.
For example, the drone fly Eristalis sp. has an aquatic larva. Like all fly
larvae, it possesses a pair of posterior spiracles. In this species, they extend
into a siphon longer than the rest of the body, which is used in a snorkel-like
manner to enable the animal to forage whilst under water (see figure 1.1a). The
larvae of many Nematoceran flies such as the moth fly Clogmia albipunctata are
aquatic, too. The posterior spiracles of this species are much smaller, yet the
cuticular fans at the posterior end of either the spiracle or the telson are
dramatically enlarged. Thereby, this larva uses the adhesive force of water to
attach itself to the surface with the head end down. The fans can be retracted
(personal observation; see figure 1.1b). I would like to emphasise that these
two species are both Diptera, sharing all the principal elements of
neuromuscular networks outlined above. These elements vary only in number or
morphology. Both species share a common ancestor not long back in their
phylogeny, and their neuronal machinery was developed and diversified from a
common groundplan (THOMAS et al., 1984).
One can now ask a central question: What are the mechanisms that lead to such
diversifications? This is an important issue from both a developmental and an
evolutionary perspective. To explore these mechanisms, I originally designed
this project as a comparative study involving larvae of a range of species. In
particular, I mapped the embryonic and larval muscles of the Phorid fly
Megaselia scalaris and compared them to those in Drosophila melanogaster (see
figures 1.2 to 1.4 and table 1.1). While working with Megaselia, I encountered
difficulties with the maintenance of the cultures and other technical issues.
Therefore, I abandoned this species and, going back to my fundamental questions,
I chose a new system to address them.
In order to study the mechanisms that lead to a diversification of a
neuromuscular network, I required two such networks that were variations of a
common ground plan. Originally, I had intended to relate the neuromuscular
networks of two different species with a shared ancestral state as a ground
plan. In the Drosophila abdomen, a ground plan can be found in the array of
neuronal stem-cells that is identical in all segments early in development.
These stem cells give rise to cells that are then specified and interconnected,
thereby forming individual elements in the neuromuscular network. The mature
networks in the various segments finally have diverged from each other: they
share some features, but are different in others (see figure 1.5).
Such diversification occurs in two different species that increasingly diverge
in the course of evolution; but also in iterative features of a single animal,
such as different metameres of the Drosophila larva that have specialised to
serve specific functions. Here one can observe exactly the same principles of
diversification: a set of neuronal stem-cells that is identical in all thoracic
and abdominal segments gives rise to a network of neurons and glia, of which
some are subsequently altered in a segment-specific manner. In the case of
motorneurons, they are matched with target muscles of which some are similarly
altered in certain segments.
Much is known of the development, physiology and morphology of the Drosophila
neuromuscular network (see below). This knowledge and the availability of tools
to manipulate and study this system made it more attractive than my original
approach exploring neuromuscular diversification in different species. Most
studies on the motor system’s development were done in the abdominal segments
A2-A7. These segments are identical, but different from the gnathal and thoracic
segments as well as the most posterior segments A8 and A9 (BATE, 1993).
The alterations in A8 and A9 with respect to more anterior segments of the
abdomen have served as a model system to address my questions: What are the
mechanisms that shape the cellular elements of the neuromuscular network over
the course of its development in a segment-specific manner? And what are the
cues that determine which particular mechanism is to be induced in individual
I will finish this introduction with a review on how the core of the
neuromuscular network is formed in Drosophila: The development of bodywall
muscles; the development of the Central Nervous System (CNS); and the
establishment of connections between these and other elements of the network.
1.2 Early Mesoderm Development
The mesoderm is derived from the most ventral cells of the blastoderm, which
express twist and snail (sna, which encodes a zinc-finger protein), and are set
aside by the action of maternal dorsal (CHASAN & ANDERSON, 1992), an axial
patterning gene. At gastrulation, these cells invaginate along the ventral
furrow, divide twice and spread dorsally. There they form a monolayer, which is
in close contact with the overlying ectoderm (LEPTIN & GRUENEWALD, 1990).
Derivatives of these will include the somatic muscles, the visceral muscles, the
heart and the fat body (BATE, 1993). Initially, all cells of the internal
mesodermal layer appear uniform, expressing the transcription factor twist (THISSE
et al., 1987), its targets DMEF-2 (LILLY et al., 1995; TAYLOR et al., 1995) and
tinman (AZPIAZU & FRASCH, 1993), as well as snail, which largely represses the
expression of genes characteristic for more lateral sectors of the blastoderm
(BATE et al., 1999).
As the mesodermal cells migrate dorsally, they respond to decapentaplegic (dpp)
(STAEHLING-HAMPTON et al., 1994). By then, dpp is expressed in a dorsal band of
ectodermal cells and acts on underlying mesodermal cells to maintain the
expression of tin and repressing ventrally expressed genes like pox meso. In
this way, dpp signalling divides the mesoderm into dorsal and ventral sectors,
while the segmentation genes evenskipped and sloppy paired divide it along the
anterior-posterior axis (AZPIAZU et al., 1996; RIECHMANN et al., 1997; reviewed
in BAYLIES et al., 1998).
Beyond this, the differences in twist expression between the eve (low twist) and
the slp (high twist) domain are crucial for the assignment of cells to somatic
myogenesis and first become apparent after gastrulation, when the mesoderm is
subdivided through the expression of pair-rule genes. twist inhibits the
development of derivatives such as visceral mesoderm progenitors and promotes
the formation of somatic muscles in their place (BAYLIES & BATE, 1996).
1.3 Muscle formation
As in many organisms, muscles in Drosophila melanogaster consist of syncitial
fibres, also called myotubes. Position, size, orientation, insertion sites in
the epidermis, patterns of gene expression and innervation make every muscle
individually identifiable. The process of muscle formation from
mid-embryogenesis can be divided into two phases:
Phase 1: Fate determination, from stage 11. Within the domain of twist
expressing cells are clusters of cells that express the transcription factor
lethal-of-scute (CARMENA et al., 1995). Via a notch-mediated lateral inhibition,
one cell within each of these clusters becomes a progenitor myoblast, while the
remaining cells will become naïve, fusion-competent myoblasts. Each of the
progenitor myoblasts divides to give rise either to one muscle founder cell (MFC)
and one adult muscle precursor or two muscle founder cells.
It is important to emphasise that the ~30 MFCs per abdominal hemisegment carry
all the information required to produce a specific muscle, whereas the fusion
competent cells are not able to fuse with each other. Evidence for this is
provided by flies that are deficient for myoblast city (mbc), which is essential
for myoblast fusion. In the absence of fusion, the muscle founder cells
differentiate into small, fully functional but mononucleated muscles at the
appropriate location. The muscle marker profile remains unaffected as the
wild-type complement of genes is expressed (RUSHTON et al., 1995).
Individual MFCs achieve their specific identities partly by the expression of
transcription factors such as S59, apterous, Krüppel, ladybird, even-skipped,
muscle specific homeobox and nautilus (reviewed in DWORAK & SINK, 2002).
Myoblasts on the other hand express the Gli superfamily transcription factor
myoblasts incompetent (RUIZ-GOMEZ et al., 2002).
There is a second class of muscle precursors in the embryo that will initiate
the formation of adult muscles at the end of larval life. These cells are known
as adult muscle precursors (APs).
Phase 2: Myoblast fusion begins during stage 12 and continues for less than
seven hours until stage 16. This starts when the muscle founder cells seed the
fusion of 4 to 25 myoblasts into one multinucleated myotube. Myoblasts or fusion
competent cells express sticks and stones (sns) on the cell surface. Other genes
essential for fusion include the previously mentioned myoblast city (mbc;
RUSHTON et al., 1995), blown fuse (blow, DOBERSTEIN et al., 1997), rolling stone
(rost; PAULULAT et al., 1995) and dumbfounded (RUIZ-GOMEZ et al., 2000). The
process of fusion can be divided into cell-cell recognition which includes the
development of filopodia on myoblasts; adhesion; alignment; and eventually
membrane fusion (MENON & CHIA, 2001). The number of myoblasts that fuse with the
MFC determines the size of the resulting muscle.
An early sign of muscle development is the formation of small syncitia,
including 2 to 3 nuclei that appear at stage 12 at specific positions within the
ventral part of the mesoderm, close to the ectoderm (BATE, 1990). The cells now
lose twist expression, whereas adult muscle precursors maintain twist expression
throughout embryogenesis and larval life, when they divide to form pools of
adult myoblasts from which adult muscles will be formed. During germband
shortening, this process is continued and new syncitia form in other parts of
the somatic mesoderm; by the end of germband retraction, every developing muscle
is represented by a syncitium, also called ‘muscle precursor’ (reviewed in
DWORAK & SINK, 2002).
These give rise to 30 myotubes per abdominal hemisegment in A2 to A7 (A1 has two
muscles absent and one additional muscle, segments posterior to A7 vary as
well), of which each represents a contractile unit and is the equivalent to a
primary or secondary myofibre in a vertebrate muscle (BAYLIES et al., 1998). The
muscle pattern is segmentally repeated, and this adds up to a network of
approximately 600 individual muscles in the larva (BATE et al., 1999). Muscles
are divided into external (with exceptions transverse) and internal (with
exceptions longitudinal) ones. The muscle pattern in Drosophila and its
development has been analysed and mapped in greater detail than in any other
1.4 Muscle Attachment
The developing myotubes are attached to the epidermal body wall by a set of
ectodermal attachment cells, called tendon cells. They are part of the larval
exoskeleton and derived from subsets of epidermal cells that acquire the
competence to differentiate into tendon cells through the expression of the
transcription factor stripe (FROMMER et al., 1996; LEE et al., 1995). The
precise match between myotubes and tendon cells is determined by a continuous
dialogue between these two cells. Tendon cells direct the migration of myotubes;
myotubes on the other hand influence the maintenance of the tendon cells in an
Several genes that are essential for the formation of these
attachment sites have been described: delilah, which codes the protein b1
tubulin (ARMAND et al., 1994); groovin (VOLK & VIGAYRAGHAVAN, 1994); and the
transcription factor stripe (sr; LEE et al., 1995). The required signalling
depends widely on the neuregulin-like ligand Vein and its receptor Egfr, an
epidermal growth factor (VOLK, 1999). Myotubes develop filopodia at the tips and
expand into a posterior-anterior orientation underneath the ectoderm, where they
make contact and get attached to the tendon precursor cells (BATE, 1990).
Roundabout (robo) receptors and their ligand Slit act as both repellent and
attractant: Robo-expressing mesodermal cells initially migrate away from Slit at
the midline; a few hours later the same cells change their behaviour and require
Slit at muscle attachment sites to establish contact (KRAMER et al., 2001). The
attachment ends the myotube extension and with increased expression of delilah,
b1 tubulin marks the final differentiation of tendon precursors into elongated
tendon cells. These cells eventually have microtubules and microfilaments that
are oriented in the apical-basal cell axis and are connected to myotubes and a
thick layer of extracellular matrix by hemi-adherens-type junctions (HAJs)
(SOUSTELLE et al., 2004). The fate of muscle-unbound tendon precursors is
unclear. The initial determination of tendon precursors is myogenesis
independent (reviewed by VOLK, 1999).
1.5 Early CNS Development
In the early blastoderm, a ventral-lateral region is defined as the
neuroectoderm. Patterning genes along the anterior-posterior (AP) and
dorsal-ventral axis (DV) subdivide the neuroectoderm. Sequential action of the
maternal AP co-ordinate, gap and pair-rule genes define individual AP stripes of
segment polarity gene expression in each metamere (AKAM, 1987a). At the same
time, a dorsal-ventral gradient of the nuclear factor NF-kB, dpp and epidermal
growth factor receptor (EGFR) signalling pathways define the dorsal-ventral
borders of the neuroectoderm. These borders also underlie the dorsal-ventral
subdivision of the neuroectoderm into longitudinal stripes that express of the
columnar genes (vnd, ind and msh; VON OHLEN & DOE, 2000; reviewed in SKEATH &
THOR, 2003). The columnar genes determine gene expression along the
dorsal-ventral axis and thereby lead to the acquisition of distinct fates of
neuroblasts from different dorsal-ventral columns (reviewed in SKEATH & THOR,
Following this subdivision into equivalent groups of cells, all cells in a group
acquire neural potential through the expression of the proneural genes achaete
(ac), scute (sc) and lethal of scute (l’sc) (GARCIA-BELLIDO & SANTAMARIA, 1978;
reviewed in SKEATH & THOR 2003), collectively known as ac/sc genes.
They compete for neural fate, which is finally determined through the Notch
pathway: ac/sc expression can activate Delta, the ligand of Notch; Notch can
repress the expression of the ac/sc genes. In the initial equivalence group, all
cells signal though the Notch pathway: Notch activation results in the
downregulation of ac/sc expression and consequently a decrease in Delta
expression. This reduces the targeted cell’s ability to activate Notch
signalling in its neighbouring cells. Among cells of an equivalence group, one
is singled out which has higher levels of ac/sc or Delta activity or lower
levels of Notch. This cell has a higher ability to activate Notch signalling in
neighbouring cells. Thereby, initially subtle differences in ac/sc or Notch
pathway activity are amplified and lead to the differentiation of one single
neuroblast per equivalence group (HEITZLER et al., 1996; HEITZLER & SIMPSON,
1991; reviewed in SKEATH & THOR, 2003). Cells acquiring neuroblast fate enlarge
and delaminate in five sequential waves to the interior of the embryo. The
remaining cells of the equivalence groups remain undifferentiated or acquire
epidermal fate (HEITZLER et al., 1996).
Approximately 100 neuroblasts start cell lineages that contribute to brain
development, 30 neuroblasts per Ventral Nerve Cord (VNC) hemisegment delaminate
from more posterior parts of the neuroectoderm. According to its spatiotemporal
position and its expression patterns, each neuroblast is individually
identifiable (reviewed in SKEATH & THOR, 2003).
Neuroblasts divide asymmetrically and thereby give rise to one ganglion mother
cell (GMC) by an asymmetric cell division cycle. This GMC then divides and gives
rise to two neurons or one neuron and one glial cell (KRAUT et al., 1996). Each
set of segmental neuroblasts gives rise to approximately 350 neurons, 30 glial
cells and 10 neurosecretory cells, whereby each neuroblast lineage can be
identified and traced in spatiotemporal terms (BOSSING & TECHNAU, 1994; DOE,
1992; SCHMIDT et al., 1997; BOSSING et al., 1996; PROKOP & TECHNAU, 1991). The
homeotic gene network has been shown to be involved with segmental differences
in the lineages of homologous neuroblasts, which result for example in
approximately 500 neurons and glial cells generated by thoracic neuroblasts
compared with approximately 380 neurons and glia that are derived from abdominal
neuroblasts (PROKOP et al., 1998; BELLO et al., 2003).
GMCs acquire unique fates, too. The transcription factors Hunchback (Hb),
Krüppel (Kr), Pdm, Castor (Cas) and Grainyhead (Gh) are expressed in a temporal
cascade, starting with Hb in the neuroblasts as they delaminate. GMCs and
neurons produced by Hb expressing NBs retain this expression. Later Hb is
downregulated in the NBs and Kr is activated. This is followed by Pdm, Cas and
Gh, generating a layered pattern of gene expression in the neurons and glia
produced by each NB (KAMBADUR et al., 1998; ISSHIKI et al., 2001; NOVOTNY et
al., 2002; BRODY & OLDENWALD, 2000). Each temporal factor activates the next
gene and represses the ‘next plus one’ gene (reviewed in SKEATH & THOR, 2003).
1.6 Neuron and Glial Cell Specification
The postmitotic cells of the VNC are divided in three groups: motorneurons,
interneurons and glia. Each hemisegment of the VNC contains approximately 30
glia, which are divided into the two subclasses of midline glia and lateral
glia. Furthermore, each VNC hemisegment contains about 40 motorneurons, divided
into subclasses according to their projection pattern (see 1.7). Genes involved
in motor neuron specification include evenskipped and Hb9, islet and lim3.
Mutations in these genes lead to axon pathfinding phenotypes. Little is known
about the specification of interneurons; however, it is clear that they have
individual fates, too (LANDGRAF & THOR, 2006; SKEATH & THOR, 2003; THOR &
1.7 Motorneuron Outgrowth, Target Finding and Innervation
At stage 12, axons of motorneurons leave the CNS, grouped into three nerves: the
segmental nerve (SN), the intersegmental nerve (ISN) and the transverse nerve
(TN, along the segment border with only two motorneurons; see Chapter 4). This
structure represents a groundplan that appears to be stereotypical for all
arthropods (THOMAS et al., 1984). The ISN has two side-branches, ISNb and ISNd;
the SN is sub-divided into SNa and SNc (THOR & THOMAS, 2002; see fig. 2.14).
According to their axonal projections, motorneurons can be subdivided into
distinct classes, most generally into ventrally (vMN) and dorsally (dMN)
projecting motorneurons (LANDGRAF & THOR, 2006).
Ventrally projecting motorneurons are specified through the expression of Nkx6
(BROIHIER et al., 2004; CHEESMAN et al., 2004) and Hb9 (BROIHIER & SKEATH,
2002), whilst dorsally projecting motorneurons are specified through evenskipped
expression (LANDGRAF et al., 1999). All three of these genes code for
homeodomain proteins. Further subdivisions in each of these two groups can be
drawn from the expression of Lim3, drifter and Islet (ventral) (CERTEL & THOR,
2004; THOR et al., 1999; THOR & THOMAS, 1997); as well as grain and zfh1
(dorsal) (reviewed in LANDGRAF & THOR, 2006).
The outgrowth of the nerves is regulated by mechanisms of axon guidance, leading
to a progressive defasciculation of the nerves as soon as they approach the
muscles in the periphery around stage 15. Defasciculation is regulated by
signals from the muscle founder cell, whereby a single MFC is able to trigger
the branching of axon bundles off the main nerve tracks (LANDGRAF et al., 1999).
The establishment of neuromuscular junctions (NMJs) requires target recognition,
synapse formation and stabilisation. At hatching, there are approximately 36
motorneurons innervating 30 muscles per abdominal hemisegment from A1 to A7
(LANDGRAF et al., 1997; SINK & WHITINGTON, 1991).
In contrast to neurons of vertebrates, insect neurons form dendritic arbours not
on the cell body, but at a distance from it. Motorneuron dendritic fields are
arranged into domains, which represent centrally the arrangement of body wall
muscles in the periphery (LANDGRAF et al., 2003). Unlike the segmental muscles,
this myotopic map is parasegmentally organised. It forms independently of the
presence of muscles, glial differentiation or competitive interactions with
adjacent motorneurons (LANDGRAF et al., 2003). The positions of the cell bodies
do not correlate with their innervation pattern and this myotopic map and the
determining cues represent the first layer of organisation in the motor system.
As this layer proves to be very robust, one can speculate that its underlying
cues are laid down early in development as the embryo subdivides into
parasegmental units (THOMAS et al., 1984). The genes and processes involved in
developing dendritic arbours are far from being fully understood yet (reviewed
by GAO & BOGERT, 2003).
1.8 Objective of this Thesis
With the background of this information on how the neuromuscular machinery
develops in the segments A2 to A7, I can now briefly re-iterate the aim of my
project as such: At many points in the course of neuromuscular development,
small events can cause alterations of the described pathways. Such alterations
can give rise to developmental side-lines branching off a “ground plan” of
network formation. This can ultimately cause the establishment of a very
different neuromuscular network.
I do not consider segments A2 to A7 a ground plan in the sense of an ancestral
state. However, the stereotypical array of muscles and neurons in these segments
has made them the favoured system for studying neuromuscular development.
Therefore, it represents the best reference system for one possible outcome of
the developmental pathways described above. The very different network in the
most posterior segments A8 and A9 must be accommodated by alterations in these
There are two principal questions that underlie this thesis: What are the
mechanisms by which neuromuscular networks are diversified in the course of
their development? And what are the cues that underlie this diversification?
Before I could approach these questions, I first classified the muscles in A8/9
(Chapter 2) and some of their motorneurons (Chapter 3) to describe the
alterations; and then tried to relate these pairs of muscles and neurons to the
muscles and neurons of segments A2 to A7 (Chapter 4). In order to resolve some
of the mechanisms and cues underlying the alterations, I then investigated the
role of programmed cell death as a mean for diversifying neuronal networks
(Chapter 5); and the contribution of Hox genes on the spatio-temporal
orchestration of developmental pathways that eventually determine
segment-specific output networks (Chapter 6).
Figure 1.1: Characteristics of two Dipteran larvae
Figure 1.1a: The aquatic larva of the drone fly Eristalis sp., which uses
its posterior spiracles that are expanded to a siphon for respiration while the
animal is under water. (from: OLDROYD, 1964)
Figure 1.1b: The aquatic larva of the moth fly Clogmia albipunctata uses
elaborate cuticular fans at the posterior spiracles to attach itself to the
water surface. These fans can be retracted. (from: SUNDERMANN & LOHSE, 2004)
Figure 1.2: The embryonic muscle pattern of Megaselia scalaris
Figure 1.2a shows staining for myosin in 14 hrs AEL embryos of Megaselia
scalaris in a dorsal view (left; arrow indicates the dorsal midline; PS
indicates the posterior spiracles) and a ventro-lateral view (right).
Figure 1.2b shows diagrams of the embryonic muscle patterns of Megaselia
scalaris compared to Drosophila melanogaster. Two of the most significant
differences (presence of muscles with no obvious homologue in Drosophila) are
highlighted with arrows. Note the strong similarity between the two patterns.
Red muscles are external, yellow more internal, blue most internal. Turquoise
muscles in Megaselia are unclear; purple are the segment border muscles. This
colour code applies to all diagrams in this thesis.
Figure 1.3: The larval muscle pattern of Megaselia scalaris
Third instar larvae were dissected and stained for HRP to visualise nerves (see
arrow in fig. 1.3, left). The muscles were mapped and named according to the
nomenclature used for Drosophila (after BATE, 1993; n > 8).
Table 1.1: Nomenclature for Megaselia muscles
The names for every individual muscle are given and matched with its most likely
equivalent in Drosophila. This table and the nomenclature are based on position
(ventral, lateral, dorsal) and orientation (oblique, transverse, acute,
longitudinal) (after BATE, 1993).
Figure 1.4: Diagram of the main taxa of Diptera
This phylogenetic tree summarises the relationships of all higher level taxa of
Diptera and includes an estimated 125,000 species. The relative species
abundance of each individual taxon is indicated by the width of each triangle. (from:
STAUBER et al., 2002)
Figure 1.5: Diagram of diversification principles
In the central nervous system, the development of the neuromuscular network
starts with an array of neuronal stem cells (left) that is identical in all
segments. At the end of embryogenesis, however, differences have developed
between segments as adaptations to specific functions (right). These involve
some elements (neurons and muscles), but not all.
This developmental diversification of iterative features follows the same
principles as an evolutionary diversification in two diverging species.
return to "thesis