|
  ReviewDrosophila as a genetic and cellular model for studies on axonal growthNatalia Sánchez-Soriano1 1The Wellcome Trust Centre for Cell-Matrix Research, Faculty of Life Sciences, The University of Manchester, Manchester, UK 2MRC Centre for Developmental Neurobiology, Guy's Campus, King's College, London, UK 3Department of Anatomy and Cell Biology, University of Melbourne, Victoria, Australia
Neural Development 2007, 2:9doi:10.1186/1749-8104-2-9 The electronic version of this article is the complete one and can be found online at: http://www.neuraldevelopment.com/content/2/1/9
©
2007 Sánchez-Soriano et al; licensee BioMed Central Ltd. AbstractOne of the most fascinating processes during nervous system development is the establishment of stereotypic neuronal networks. An essential step in this process is the outgrowth and precise navigation (pathfinding) of axons and dendrites towards their synaptic partner cells. This phenomenon was first described more than a century ago and, over the past decades, increasing insights have been gained into the cellular and molecular mechanisms regulating neuronal growth and navigation. Progress in this area has been greatly assisted by the use of simple and genetically tractable invertebrate model systems, such as the fruit fly Drosophila melanogaster. This review is dedicated to Drosophila as a genetic and cellular model to study axonal growth and demonstrates how it can and has been used for this research. We describe the various cellular systems of Drosophila used for such studies, insights into axonal growth cones and their cytoskeletal dynamics, and summarise identified molecular signalling pathways required for growth cone navigation, with particular focus on pathfinding decisions in the ventral nerve cord of Drosophila embryos. These Drosophila-specific aspects are viewed in the general context of our current knowledge about neuronal growth. BackgroundThe function of a nervous system depends on the proper arrangement of its cellular elements, that is, neurons and glia cells. Amongst these, neurons bear axonal processes that establish synaptic contacts with other cells (neurons, muscles or gland cells) that can be a significant distance away. The transfer of information between these cells is the key feature of nervous system function and is usually mediated by action potentials that propagate along axons and are passed on to other cells at synapses. The wiring of such a system has to be precise and reproducible from individual to individual, as was first highlighted by Ramón y Cajal for the nervous systems of humans, other vertebrates, and also invertebrates [1]. Such precision is achieved during development through the guided growth of axons along specific paths, a process clearly governed by genetic mechanisms [2,3]. Essential work contributing to our current understanding of axonal growth has been carried out in vertebrates and invertebrates alike, in many instances demonstrating the conservation of principal mechanisms across the animal kingdom. One strategy towards improving molecular insights into axonal growth is the use of genetically tractable invertebrate model organisms, such as the worm Caenorhabditis elegans or the fruit fly Drosophila melanogaster. The strength of these species lies in their amenability to genetic manipulation (see below) and the fact that their nervous systems are composed of relatively low numbers of cellular elements. Whereas specific neuronal connections in vertebrates are usually formed by larger groups of neurons that develop and act in parallel, these connections are mostly represented by unique, individually recognisable neurons in invertebrates. Studies capitalising on such identifiable neurons, for example in insects, have helped to unravel principles of neuronal circuit formation. For example, the initial observation of guidepost cells as stepping stones for axonal growth came from studies of the grasshopper limb bud [4]. The concept of pioneer guidance, which proposes that axonal tracts are established by single pioneer neurons that are subsequently used as guidance cues for follower neurons, was discovered in vertebrates [5]. However, work on insects has helped to refine these concepts ('selective fasciculation' and 'labelled pathways' hypotheses) [6-8] and to contribute molecular players [9]. Once it was revealed that Drosophila embryos were experimentally accessible and largely homologous to grasshoppers at the single cell level [10], insights that had to that date mostly been gained on larger insects could be transferred into a genetically tractable model system. Such work in Drosophila was further fuelled by the development of new technologies, strategies and molecular tools, such as specific anatomical antibody probes [11,12] and the development of ever more sophisticated genetic screening strategies [13]. Furthermore, such research has profited enormously from the increasingly well organised research environment of Drosophila, including the systematic generation and provision at a large scale of new genetic tools and mutant fly stocks, and the improvement of access to research-relevant information [14-18]. Undoubtedly, the use of Drosophila as a model system for the study of axonal growth has been a prolific endeavour, providing a plethora of novel and/or refined insights into relevant molecular mechanisms, many of which have been shown to be conserved in higher organisms. This review will provide an overview of the major cellular model systems established for the fly, our current insights into Drosophila growth cones (the key structure executing axonal growth), the housekeeping machinery regulating the cytoskeletal dynamics required for axonal growth, and the signalling events involved in axon guidance, focussing on cellular models in the Drosophila embryo. Models for axonal growth in DrosophilaOne of the major strengths of Drosophila lies in its use as a model where axonal growth can be studied in situ. A widely used strategy for the unbiased discovery of neuronal growth mechanisms in Drosophila is the search for mutations that cause morphological aberrations of axonal tracts or neuronal connections in situ. Subsequently, the genes associated with such mutations can be identified, and the molecular nature, function and type of interactions of their products can be studied. To facilitate such genetic screens and functional studies of gene action, descriptions of axonal pathways and neuronal connections in wild-type animals have been provided for a number of different neural systems in Drosophila. A selection of these cellular models is illustrated in Figure 1, and some of their principal features will be discussed below.
First, neurons of all classes, that is, sensory neurons, interneurons and motorneurons, have been used for studies of axonal growth in Drosophila (red, green and blue in Figure 1, respectively). The principal structures of vertebrate and invertebrate neurons have been proposed to be homologous [19], despite there being certain organisational differences (Figure 2). These differences may be associated with deviations in some aspects of axonal pathfinding behaviours (details in Figure 2). For example, in the vertebrate trunk, sensory neurons are (pseudo-)unipolar and located in the dorsal root ganglia derived from a migratory stream of neural crest cells [20]. In contrast, cell bodies of sensory neurons in Drosophila are usually bi- or multipolar and are born and located in the periphery close to the sense organs they innervate or represent. Therefore, sensory axons in Drosophila grow unidirectionally towards the central nervous system (CNS), whereas sensory neurites in vertebrates bifurcate and grow bidirectionally both towards the CNS and into the periphery. A clear exception to this rule are the ocellar photoreceptors of Drosophila (located on the dorsal surface of the head; Figure 1c,d), which do not form axons themselves, but are connected to the brain via interneurons [21]. Somata of inter- and motorneurons are commonly multipolar in vertebrates and located in the synaptic region (grey matter; Figure 2b). In contrast, cell bodies of comparable neurons in Drosophila are unipolar and localised outside the neuropile within the cortex of the neuromeres. Finally, in vertebrates, axons ascending and descending to/from the brain are located in defined tracts in the white matter (1–3 in Figure 2b) where they become heavily myelinated, whereas comparable axons in Drosophila are located in the synaptic neuropile and usually lack glial ensheathment. Hence, mechanisms placing ascending/descending axons in the Drosophila neuropile might be distinct from those placing them in the white matter in vertebrates. Alternatively, if the same guidance cues are utilised, their spatial expression must differ considerably between vertebrates and arthropods.
Second, Drosophila is a holometabolous insect that exists as a maggot during larval life but is substantially reorganised into the adult fly during metamorphosis at the intermediate pupal stage. This process is not dissimilar to metamorphic events known from lower vertebrates, such as amphibians. One phase of de novo axonal growth takes place in the embryo, and a vast number of these cellular elements are maintained into the adult stage, albeit being remodelled during the pupal period [22]. However, to adapt to the far more complex behavioural repertoires of the adult, a second phase of de novo neurogenesis and axonal growth is initiated during larval life and completed during the process of metamorphosis. There is abundant evidence that common cellular and molecular mechanisms for axon guidance apply at both stages, allowing us, in many instances, to study the same molecules with the multiple techniques and cellular systems available for both stages of axonal growth in Drosophila [3]. Cellular models from both developmental phases (Figure 1a,b versus 1c,d) have been used to screen for genes involved in neuronal growth regulation. At embryonic/larval stages, genetic screens have mostly been based on analyses of animals bearing mutations that were induced by chemical treatment, radiation or via transposable elements. Morphological read-outs used in these screens have included motornerves, projections in the CNS, or sensory projections in the trunk or head (for example, Bolwig's organ; 1–4 in Figure 1) [23-32]. Some screens have been carried out on adult flies using as read-outs the giant fibre system (13 and 14 in Figure 1) or general brain morphology [33,34]. However, since many mutations cause embryonic or larval lethality, the utility of adult animals for genetic screens is limited. To overcome this problem, mosaic strategies have been used in which only specific tissues express the mutant phenotype and can be maintained by an otherwise normal body [13]. Such mosaic screens have been successfully carried out using adult photoreceptor axons as read-outs (8 in Figure 1) [35,36]. A further refined mosaic strategy is the MARCM technique, in which homozygous mutant neurons can be visualised at the single cell level, surrounded by heterozygous, non-mutant cells [37]. This technique has been used to study and screen for growth aberrations of mushroom body, olfactory and photoreceptor neurons (8, 10 and 12 in Figure 1) [38-40]. Based on these various screens in embryonic, larval or adult individuals, many mutations that result in defective axon growth have been identified [3]. The gene functions underlying a number of these mutations will be discussed below (see 'Regulators of actin dynamics in Drosophila growth cones' below). Studies of neurons in situ have the essential advantage that they address the relevance of genetic mechanisms for developmental processes in a natural context. However, the ability to utilise precise and well-controlled pharmacological, physiological or imaging methods in such in situ systems is limited. To address these shortcomings, cultures of primary neurons, obtained from embryonic or larval tissues, have been established in Drosophila. These primary culture systems have been used successfully to demonstrate and/or analyse various phenomena, such as promotion of axonal growth by the neural cell adhesion molecule (N-CAM) homologue Fasciclin 2 or the extracellular matrix molecule laminin [41-43], induction of axonal fasciculation through neuronal activity [44], dependence of axonal growth on endocytosis [45,46], or cAMP-dependent neurotransmitter release from growing neurons [47]. All of these phenomena are likely to be of relevance to neuronal growth in situ [9,21,48], and the established culture systems provide valuable additional means for their study. Taken together, the Drosophila system provides a wide range of cellular models to be used for studies of axonal growth. In addition, a number of models exist for other forms of neuronal growth, such as dendritic arborisation or the branching and plasticity of synaptic terminals [49-52]. However, these systems will not be considered here. Growth cone dynamics in DrosophilaPrincipal structure and function of growth conesThe elongation of an axon is led at its tip by a highly dynamic structure, the growth cone, first described and named by S Ramón y Cajal ("cono de crecimiento") and then confirmed in live studies by RG Harrison [53-55]. Growth cones navigate along stereotypical paths, steadily elongating the axon by adding new structural components such as microtubules and membrane to its tip [56]. Just like the leading front of migrating cells, growth cones display actomyosin-containing filopodia and lamellipodia, focal contacts, a dynamic population of microtubules, and cell polarity markers such as Par-3 and -6 [57-62]. In contrast to migrating cells, where stress fibres pull the entire cell forward, neuronal cell bodies generally remain behind when growth cones advance, remaining connected by the steadily elongating axons, the core of which contains bundled, stable microtubules representing the highways for axonal transport. It has been shown for growth cones of many species, that their guidance requires actin dynamics in the growth cone periphery [55,59,60,63]. The peripheral filamentous actin cytoskeleton executes a continuous, myosin-driven retrograde flow that involves ATP-dependent addition of globular actin to actin filaments at their barbed ends (pointing towards the growth cone's periphery), a gradual change of actin-bound ATP to ADP, and hydrolysis or severing at their pointed ends (pointing towards the growth cone's centre) [64-66]. This process of actin assembly and disassembly and its organisation into lamellipodial networks and filopodial bundles is regulated by a complex molecular machinery [65,67-69], which we detail below in the context of Drosophila components (see 'Regulators of actin dynamics in Drosophila growth cones'). Whereas growth cone guidance crucially depends on actin dynamics, growth cone advance depends on microtubules. This is clearly demonstrated by the persistence of axonal growth in the presence of actin polymerisation blockers, in vertebrates as well as in Drosophila [55,70-72], whereas pharmacological destabilisation of microtubules causes growth cone retraction [55,73]. Furthermore, extending axons can even be induced to turn in the absence of filamentous actin if they are exposed to an electrical field [74]. Microtubules in the axon shaft and central zone of growth cones are stable and bundled, whereas single unstable microtubules elongate into and retract from the peripheral actin-rich zone in a highly dynamic fashion. These microtubules can be trapped or stabilised through signals in the growth cone periphery, thus determining the direction in which microtubules of the axon shaft will extend [55,58-60,63]. Microtubule growth and shrinkage is, per se, an autocatalytic process [75-77]. However, a number of cellular factors regulate microtubule dynamics, as detailed below in the context of Drosophila components (see 'Regulators of microtubule dynamics in Drosophila growth cones'). Growth cones in DrosophilaInitial work on Drosophila growth cones dates back more than 20 years when growth cones were first described incidentally in a study on the role of the activity regulator Maleless/Nap in primary neurons cultured from the larval nervous systems [78]. In vivo observations carried out on growth cones of primary neurons in culture demonstrated classifiable shape differences (longer filopodia of ventral ganglion-derived neurons versus those derived from brain), and also established strategies to measure their dynamics [79]. These strategies were later used to demonstrate that genetic or pharmacological alterations of cAMP levels influence growth cone dynamics [80]. Analyses of dye-filled motorneurons in Drosophila embryos demonstrated that their growth cones make characteristic and reproducible directional decisions when advancing in the CNS and periphery [81]. In agreement with this finding, ultrastructural studies in the developing embryonic Drosophila CNS showed that filopodia of pioneering growth cones establish close contacts with characteristic sets of cells in their immediate environment [82]. In vivo studies of the identified RP2 motorneuron in the same context revealed that these growth cones grow at rates comparable to other animal model systems and carry out a set of characteristic shape changes along their reproducible paths [83]. Follow-up studies using this system showed that genetic removal of the Roundabout receptor (see 'Mediolateral patterning of longitudinal fascicles' and 'Growth cone guidance at the Drosophila CNS midline' sections below) in these growth cones led to an increase in the length and persistence of filopodia [84]. In vivo observations on growth cones of a different set of motorneurons (ISNb) demonstrated that their filopodial dynamics and guidance are regulated through GTPases of the Rho-family and their downstream effectors [85,86]. Further studies on embryonic motorneurons have shown that growth cones of pioneer neurons display a significantly more complex shape than their follower axons [9], a finding that is consistent with descriptions of the growth cones of pioneer/follower neurons in vertebrates [87,88]. In the Drosophila system, this behaviour is regulated by the N-CAM homologue Fasciclin 2 [9]. Motorneuronal growth cones have also been studied in their transition phase into a characteristically shaped neuromuscular junction. Such analyses have revealed how growth cones, which initially explore a wider muscle field, transform into precisely targeted neuromuscular terminals [89,90]. Ultrastructural analyses of such growth cones have shown how filopodia of identified motorneurons establish intimate contacts only with their specific target muscles, and that this process is regulated through the presentation of specific cell recognition molecules [91]. Consequently, ablation of such target muscles changes the behaviour of growth cones in their target area [92]. Drosophila growth cones display the characteristic hallmarks of those in other systems, such as highly dynamic filopodia, enrichment of microtubules at their core and the presence of filamentous actin in their periphery (Figure 3). Yet, to date, subcellular studies of Drosophila growth cones are fairly sparse. Most of the insights into axonal growth mechanisms in Drosophila (as reviewed below) have been derived from analyses of the axonal projection defects found in the mature embryonic or adult nervous system in mutant animals or under experimentally manipulated conditions. There are only few descriptions of the behaviours of growth cones during the axon elongation phase in these animals.
Regulators of actin dynamics in Drosophila growth conesThe morphological studies of Drosophila growth cones described above ('Growth cones in Drosophila' section) suggest that they possess the same cytoskeletal machinery as found in other model organisms ('Principal structure and function of growth cones' section). In this section we summarise and speculate on the (potential) factors regulating cytoskeletal dynamics in Drosophila growth cones. An overview of these components is given in Figure 3 and Table 1. Table 1. Direct regulators of the cytoskeleton with (potential) functions at growth cones Several factors have been reported to be able to seed new actin filaments, a process called 'nucleation'. One of these factors is the Arp2/3 (actin-related protein) complex, which is composed of seven subunits. The genes encoding all its subunits have been identified in Drosophila and some of them have been functionally assessed [93,94]. Loss of function of the genes Suppressor of profilin 2 (ArpC1) and/or Arp66B (Arp3) causes severe axonal projection defects when analysed at late embryonic stages (Figure 4, xxxix) [95]. Another independent nucleator of actin filaments, Spire, has been described for Drosophila [96] but seems not to be localised in the nervous system [14]. As opposed to Arp2/3-complex or Spire, members of the Formin family of proteins are nucleators and, in addition, effective barbed end-binding elongators of actin filaments [97]. Of the six Drosophila Formins [98], DAAM (dishevelled associated activator of morphogenesis) is expressed pan-neurally while two others, CG32030 and CG32138, are specifically expressed within midline glia (the AA142 transposon insertion, a classic midline glia marker, is inserted within CG32030; Richard Tuxworth and GT, unpublished observations). The other Formins, cappuccino, diaphanous and formin3, do not appear to be expressed at detectable levels in developing neurons [98]. No neural phenotypes have yet been associated with mutations in either of these genes, but ectopic expression of Formin 3 in the CNS can induce strong pathfinding phenotypes [98]. Thus, based on our current knowledge, the Arp2/3 complex and DAAM are the potential actin nucleators in Drosophila growth cones. Of these, Arp2/3 is unlikely to regulate subsequent actin filament elongation, since it stays with the pointed ends of elongating actin filaments, which move away from the very cell periphery after nucleation has occurred [68,97,99]. Instead other factors have been proposed to facilitate actin filament elongation. Likely candidates in Drosophila are DAAM (see above) or Enabled, the Drosophila member of the Enabled/VASP (vasodilator-stimulated phosphoprotein) family of proteins, which is a direct target of the cytoplasmic Abelson tyrosine kinase in the context of neuronal growth [100,101]. Further factors regulating actin assembly are molecules that bind monomeric actin, such as Capulet, Ciboulot and Profilin/Chickadee, all of which are essential for neuronal growth regulation in Drosophila [101-103]. A further actin monomer-binding molecule, Twinfilin, is likely to be expressed in the nervous system, but no data exist concerning its cellular function [104]. Profilin/Chickadee has been shown to interact with Calmodulin [105], suggesting [Ca2+]i may influence its activity. Consistent with this observation, growth cones of Drosophila, like those of other species, show measurable changes in local calcium concentration [106-108]. Barbed-end capping proteins are typical inhibitors of actin polymerisation. For Drosophila, the capping proteins Cpa and Cpb have been reported, but their role in the nervous system has yet to be addressed [109].
While actin filaments elongate at their barbed ends, they are severed and/or depolymerised at their pointed ends, and a number of factors have been assigned to this process [110]. One of these factors is ADF/cofilin (Twinstar in Drosophila), the activity of which is directly regulated by LIM kinase and the Slingshot phosphatase in vertebrate growth cones and the developing Drosophila nervous system alike [111-113]. Another group of potent filamentous actin severing molecules comprises members of the gelsolin family [114], at least two of which have been identified in Drosophila, flightless I and quail [115,116]. However, nothing is known about their potential requirement for axonal growth. Distal elongation and proximal shortening of actin filaments, generally referred to as 'treadmilling', would potentially lead to a constant outward movement of actin filaments. However, such movement is antagonised by myosin motor proteins, which produce a steady retrograde/reaward flow of the actin network. This action of myosin II plays an important regulatory role downstream of guidance-related signalling events [58,117,118]. Accordingly, two subunits of non-muscle myosin, the heavy chain (Zipper) and the regulatory light chain (Spaghetti-squash), and the direct regulators myosin light chain kinase (MLCK) and myosin light chain phosphatase have been implicated in neuronal growth in Drosophila [119-121]. Accumulating evidence suggests that the set of actin regulators involved in lamellipodia formation (producing branching networks of filamentous actin) is not congruent to those in filopodia (bundled linear actin filaments) [69]. The typical parallel bundling of actin filaments in filopodia can be initiated by the actin polymerisation factor Enabled/VASP in collaboration with Formins [122,123], and the principle involvement of both these factors in axonal growth of Drosophila has been reported (see above). Another suggested mode of filopodium formation is through bundling of filamentous actin via the powerful cross-linker Fascin [124,125]. Singed, the Drosophila orthologue of Fascin, has been shown to influence neuronal growth [126]. Other factors cross-linking actin filaments, such as α Actinin, Fimbrin or Filamin (cheerio in Drosophila) are able to substitute Fascin in the context of Listeria propulsion [127]. In Drosophila, α Actinin is expressed in the developing CNS [128], and cheerio has been associated with associative learning [129], but none of these factors has been studied in the context of neuronal growth so far [14]. The complex machinery mediating actin dynamics in growth cones is essentially orchestrated by Rho-family GTPases [130,131], six of which have been reported for Drosophila. Of these, Rac1, Rac2 and Mtl have an overlapping requirement during the regulation of axon extension, branching and guidance (Figure 4, xxxvi and xxxvii) [132,133]. Cdc42 is similarly involved in growth cone guidance, but its functions appear to involve downstream effectors different from those of Rac1 (for example, Pak) [85,86,134]. RhoA has been shown to inhibit neuronal growth in Drosophila [135], whereas little is known about the other two Rho-like GTPases, RhoL and RhoBTB [136]. Many of the factors shown to act up- or downstream of Rho-GTPases in vertebrates have been identified in Drosophila and their functions in neuronal growth were revealed by their mutant phenotypes. Thus, GTPases are active in a GTP-bound state, promoted by RhoGEFs (guanine nucleotide exchange factors), and inactive in a GDP-bound state, catalysed by RhoGAPs (GTPase activating proteins) [131]. The Drosophila genome contains 20 predicted RhoGAPs and 22 RhoGEFs [137-139]. At least three RhoGEFs, Still life, Gef64C and Trio, have been associated with neuronal growth regulation of the axons of motorneurons, photoreceptors, and neurons in the ventral nerve cord (Figure 4, xxxviii) [138,140-142]. Three of the 20 RhoGAPs, RhoGAP-71E1, RhoGAP-50C14 and RhoGAP-16B12/p190, show neuronal growth phenotypes when assayed in the mushroom body neuropile (MB in Figure 1d) [139], and RhoGAP93B/CrGAP/Vilse is involved in midline crossing of axons (see 'Growth cone guidance at the Drosophila CNS midline' below) [143,144]. An involvement in neuronal growth regulation has also been demonstrated in various cellular contexts for typical effectors of Rho-like GTPases, such as p21-activated kinase (PAK) and its closely associated partner, the SH2/SH3 adaptor protein Nck (Dreadlocks = Dock in Drosophila), the RhoA effector Rho-kinase (Drok or Rock in Drosophila), and the Cdc42 effector WASP, a direct activator of the Arp2/3 complex [95,141,145,146]. WASP's paralogue SCAR in itself is not an immediate effector of GTPases, but the WAVE/SCAR complex (containing SCAR, CYFIP, Kette, Abi and HSPC300) can be targeted by Rac through its direct interaction with CYFIP [147,148]. The WAVE/SCAR complex activates the Arp2/3 complex and has a clear impact on neuronal growth (Figure 4, xl) [95,149]. Taken together, most if not all classes of actin regulatory factors have been identified in Drosophila and many of them have been associated to varying degrees of certainty with neuronal growth. Regulators of microtubule dynamics in Drosophila growth conesThe second cytoskeletal component essential for growth cone advance is the microtubule; factors regulating their dynamics are summarised in Figure 3 and Table 1. One important class of proteins regulating the elongation and shortening of microtubules comprises the plus-end-tracking molecules (+TIPs) [150,151]. A number of these have been identified in Drosophila. CLIP-190 (orthologue of vertebrate CLIP-170) is strongly expressed in the nervous system [152] and has been shown to interact with the +TIP protein Chb (Chromosome bows/Orbit/MAST, an orthologue of vertebrate CLASP, CLIP-associated protein) [153]. Further +TIP molecules described for Drosophila are APC1 and APC2 (adenomatous polyposis coli), both of which are expressed in the brain (though mainly in neuronal somata) [154], and three possible homologues of EB1 (Endbinding protein 1; Eb1, CG18190, and CG32371), the potential role of which has yet to be addressed in the nervous system [155]. A potential regulator of these molecules is the Par3/6-complex, which localises to growth cones in both mammals and Drosophila, and has been shown to organise the cytoskeleton in Drosophila motorneurons [19,57,156,157]. Another molecule required for neuronal growth in Drosophila is the lissencephaly-associated factor Lis1, a molecule shown to be capable of binding microtubules to reduce microtubule catastrophe events [158,159]. Characterisation of the lis1 mutant growth phenotype in Drosophila suggests Lis1 may function through an interaction with the motor protein Dynein [158,160]. Finally, an important class of microtubule-binding proteins, also reported for Drosophila, comprises the microtubule associated proteins (MAPs), which stabilise and facilitate transport along microtubules [151]. MAP2 seems to be absent from the Drosophila genome, but one Drosophila Tau-like protein that localises within axons has been described, although no functional data have been reported to date [161,162]. A MAP1B-like molecule called Futsch has been shown to regulate axonal growth in Drosophila, presumably through stabilisation and loop formation of microtubules [163]. Therefore, as for the actin cytoskeleton, many regulators of microtubule dynamics in the Drosophila nervous system have been identified. Interestingly, the proteins described here are regulated in part through the same factors described previously in relation to actin dynamics. For example, the activity of the +TIP protein Chb/CLASP is regulated by the tyrosine kinase Abelson during axon guidance at the CNS midline [153], which is also a regulator of the actin filament elongating factor Enabled/VASP [164]. Furthermore, GTPases of the Rho-family, which are major regulators of the actin cytoskeleton (see 'Regulators of actin dynamics in Drosophila growth cones' above), also influence a number of microtubule regulating molecules [150,156]. Regulators of actin-microtubule cross-talkCross-talk between microtubules and actin is an essential aspect of regulation of growth cone behaviour (see 'Principal structure and function of growth cones' above), and this process has long been known to be facilitated by microtubule-associated proteins [165]. Such a function may, in part, be carried out by MAPs, as suggested by the observation that MAP2 can bind both microtubules and actin [166]. In Drosophila, two molecules with dual actin- and microtubule-binding capabilities have been reported. Short stop/Kakapo is a member of the Spectraplakin family of cytoskeletal linker molecules [167], and its cytoskeletal binding activity is required for axon extension [168]. Another actin-microtubule linker in Drosophila is the highly conserved DPod-1, which has been suggested to be required for growth cone guidance rather than axon extension [169]. In conclusion, many (potential) components of the machinery regulating the cytoskeletal dynamics of growth cones have been identified and characterised to differing degrees in Drosophila. It is essential to extend the analysis of these factors both at the subcellular level and in combination with genetic strategies, using the various growth cone models available in Drosophila to identify precisely how guidance cues are transduced to direct cytoskeletal re-organisation. Signalling mechanisms involved in axonal pathfinding in DrosophilaNature of axon guidance cuesCytoskeletal regulation is a house-keeping function required in all cells and, thus, a large common set of molecules serving this purpose would be expected to be present in all neurons. In contrast, the signals involved in growth cone guidance and/or the molecular mechanisms that transduce those signals are likely to differ between different types of neurons, as each neuron makes a specific set of pathfinding decisions to reach its synaptic target. It follows that the neuronal cytoskeletal machinery should be responsive to a broad spectrum of axon guidance cues and signalling pathways. Factors that have been shown to guide growing axons include simple structural barriers such as impenetrable tissues or gaps due to wounding [170], electric fields (galvanotropism) [171,172], light (phototropism) [173], or mechanical forces (stereotropism) that can be either applied externally [174,175] or generated via the growth cone's own pulling forces antagonised by adhesive interactions with the environment [176]. However, the best understood form of axon guidance is through external chemical/molecular cues and their receptive machinery within growth cones. Many ligand/receptor pathways involved in axon guidance have been identified [177-179]. Guidance molecules can be either diffusible, or attached to the extracellular matrix or cell surfaces. They are present in the environment of growth cones in an evenly distributed, graded or spatially restricted manner. Depending on its receptive machinery, a growth cone may either fail to respond to certain guidance molecules, or interpret them as growth-permissive, attractive or repulsive signals [2]. It is now a universally accepted idea that the same guidance molecules can act both as repellents and attractants on different axons, as was shown for ligands such as Unc-6/Netrin or Semaphorins in invertebrates and vertebrates alike [180-183]. To add to this complexity, the behaviour of an individual growth cone can be modified over time, for example, through alterations in its electrical activity patterns [184,185]. In the following, we will focus on a subset of neuronal model systems studied in Drosophila to illustrate the complexity of the molecular signalling pathways discovered and analysed in this model organism. Pathfinding in the trunk of the Drosophila embryoAxons of motorneurons and sensory neuronsA variety of neuronal model systems have been used for studies of axonal growth in Drosophila (see 'Models for axonal growth in Drosophila' above; Figure 1). Of these, the most intensely studied are the systems in the trunk of the Drosophila embryo, in particular, motoraxons in the periphery and axons of motor-, inter-, and sensory neurons within the CNS (neurons 1–3 in Figure 1). From developmental and morphological analyses of these neurons, it has become clear that the identity of individual neurons is specified mainly by their stereotypic lineages [186-188]. Each neuron then makes a number of stereotypic pathfinding decisions as they advance to their particular target and the mature morphology of many of these axons have been characterised (Figure 5c,d) [189-193].
For example, motorneuronal growth cones make a series of choices as they grow into the periphery. First, they have to take the appropriate stepwise decisions that guarantee that their axons exit the CNS and join one of six possible principal nerve branches leading to their appropriate target muscle fields. Second, they defasciculate from other axons in their nerve at precise points to approach their appropriate target muscles. Third, they establish adhesive contacts with their appropriate target muscles at precise positions. Fourth, the motorneuronal growth cones eventually undergo drastic morphogenetic changes and differentiate into neuromuscular junctions. In this cellular system, we currently have insights into: the specification of motorneuron and muscle identities (which in turn determine motor axonal pathfinding behaviour); relevant cells and tissues used as guideposts by motor neurons; mechanisms of fasciculation, defasciculation and pioneer guidance; and the combination of permissive, attractive or repulsive molecular guidance cues governing motor axonal navigation. The systems utilised for studying motor axon guidance and the mechanisms they have revealed are discussed in detail elsewhere [49,194]. In contrast to the motoraxonal system, the sensory system of the trunk has been somewhat neglected, in spite of its obvious advantages, which include the following. First, each of the approximately 40 sensory neurons per hemisegment has been individually identified [195]. Second, sensory cell bodies are relatively accessible and can be visualised at the single cell level in the living embryo [192]. Third, the normal pattern of axon growth from several of these neurons, and the environment through which they navigate in the periphery and where they terminate in the CNS, have been characterised in detail [192,196-201], providing an ideal background for studies of the genetic control of sensory axon growth. While at least two morphological screens have been performed to uncover such genes [25,30], our understanding of the molecular basis for sensory axon guidance is still limited. Nonetheless, several factors that guide motor axons in the periphery and interneuronal axons in the CNS have been shown to play a role in sensory axon guidance, such as Delta and its receptor Notch [197], the actin-microtubule linker molecule Short stop/Kakapo [25,202], Slit and its cognate receptors Robo and Robo2 [203], and the secreted Semaphorin, Sema2a and its receptors Plexin B and Plexin A [204]. Principal axonal pathfinding decisions in the CNSMotor-, sensory and interneurons all extend axons within the CNS. To study their growth decisions in the embryonic CNS, a number of visualisation strategies have been used, comprising global neuropile markers (Figure 4, i) [23], molecular markers for reproducible axon fascicles, such as Fasciclin 2 (Figure 4, ix) [205,206], transplantation-based analyses of neural lineages (Figure 5c,d) [207], labelling of identified neurons through dye injections or via genetically targeted markers (Figure 4, xxi-xxxi) [81,208], or ultrastructural analyses [8]. The level of precision provided by each of these techniques varies significantly, which can make it difficult to compare studies obtained via different strategies. However, when combining the various approaches, a very precise view of neuronal growth processes in the embryonic Drosophila ventral nerve cord has been obtained, as outlined below. One obvious feature of axonal growth behaviour in the CNS is that projections of motor-, inter- and sensory neurons aggregate into a cell body-free region, the neuropile, in which synaptic contacts will be established (Figure 2a). This behaviour of neurites to segregate away from cell bodies and constitute the neuropile seems to be driven by axo-axonal affinity, as illustrated by axons of larger groups of neurons in primary cell culture, which always associate into small neuropile-like arrangements [44]. The neuropile stereotypically forms in the dorsal plane of the ventral nerve cord (Figures 1b and 2a), where many of the pioneer motor- or interneurons that initiate formation of the neuropile are located. In contrast, motor- and interneurons in ventral or lateral positions of the nerve cord, as well as sensory neurons entering through the segmental nerves, have to grow through the cortex towards the neuropile. On their way towards the neuropile, axons of neurons derived from the same neuroblast lineage tend to fasciculate [189,190], a property that is also true for the larval nerve cord [209]. Within the (prospective) neuropile, growth cones are faced with a number of decisions (Figures 1b and 4c,d). They may first simply terminate shortly after entering the ipsilateral neuropile (some interneurons); second, grow towards a peripheral nerve root to exit the neuropile (ipsilaterally projecting motorneurons); third, turn anteriorly or posteriorly to project along longitudinal fascicles in the connectives (see 'The formation of longitudinal fascicles in the embryonic ventral nerve cord' below; almost all sensory axons and some interneurons); or fourth, cross the midline (see 'Growth cone guidance at the Drosophila CNS midline'; most interneurons and some motorneurons). Those axons that cross the midline are faced with the first and third choices again on the contralateral side. Those axons that do not leave the neuropile terminate their growth in stereotypic positions, presumably at points of future synapse formation [210]. Motoraxonal growth towards the CNS exit pointThe third and fourth growth decisions outlined above (choice of longitudinal fascicles and midline crossing) have been studied extensively during the past two decades, and insights gained from these studies will be reviewed in detail in the sections 'The formation of longitudinal fascicles in the embryonic ventral nerve cord' and 'Growth cone guidance at the Drosophila CNS midline' below. In contrast, virtually nothing is known about the mechanisms underlying axonal growth termination in the neuropile. In the context of the navigational capability of motorneuronal growth cones to exit the CNS, the involvement of only two signalling pathways has been suggested so far: the diffusible ligands Netrin A+B and their receptors Unc5 and Frazzled, and the receptor tyrosine kinase Eph and its ligand Ephrin. Targeted expression of Unc5 or of double stranded Eph or Ephrin RNA in interneurons causes their aberrant exit from the CNS (Figure 4, xxx) [211,212]. However, later analyses using unc5 and Eph loss-of-function mutations failed to detect equivalent axonal phenotypes, calling the original findings into debate [213,214]. At the cellular level, detailed analyses of the contralaterally projecting motorneurons RP1 and RP3 have revealed that their growth cones depend on the presence of the somata of their contralateral homologues, which are required as stepping stones to continue normal pathfinding towards the neuropile exit points. In contrast, ablation studies have revealed that intimate contacts established by these motorneurons with a number of glial guide post cells are not required [81,82,215]. If RP3 motorneurons are prevented from crossing the midline experimentally (axotomy) or genetically (commissureless mutant background; see below), their growth cones will carry out the normal growth regime, however, in mirror image fashion on the ipsilateral side, demonstrating that midline crossing is not an essential prerequisite for their capability to exit the CNS [216,217]. Obviously, the RP3 cell body (ipsi- or contralaterally) lies in a strategic position from which the pathfinding cues towards the exit point can be reached. However, under normal conditions, early growth cones of these motorneurons seem first to be attracted exclusively towards the midline, which dominates over all other pathfinding cues. The formation of longitudinal fascicles in the embryonic ventral nerve cordThe formation of longitudinal fasciclesDetailed analyses of the developing ventral nerve cord have revealed the time course of initial antero-posterior axonal growth within the longitudinal connectives [8,82,218,219]. These investigations demonstrated that a small number of early ipsilateral interneurons, namely MP1, dMP2, vMP2 and pCC, pioneer a scaffold of longitudinal fascicles that are subsequently joined by other ipsi- and contralateral axons projecting in anterior or posterior directions (Figure 5a,b). Cell-specific ablations of MP1, dMP2, vMP2 and pCC in various combinations have demonstrated that these pioneer axons are indeed essential for the establishment of connectives as a whole (Figure 4, xxiv) [219,220]. During their initial advance, growth cones of MP1, dMP2, vMP2 and pCC establish intimate contacts with longitudinal/interface glia cells, which line up in the area of the future connectives prior to the axonal growth phase (Figure 5b) [8,206,218,221]. These intimate axono-glial contacts reflect an essential role of this glial population as guidepost cells for the four pioneer neurons, since ablation of these glia cells causes severe disruption of the pioneer neurons' pathfinding and subsequent establishment of connectives (Figure 4, xxxii) [222,223]. Additional guidance cues for early pioneer neuron growth are the secreted factors Netrin A and Netrin B. Both these proteins are released from midline glia cells and become bound in restricted dorso-lateral areas of the ventral nerve cord by neurons in these areas that express Frazzled (a member of the DCC immunoglobulin subfamily; Deleted in colorectal cancer) on their dorsal surfaces (Figure 5b) [224]. Frazzled is able to bind Netrins in these lateral areas, since it is one of two identified Netrin receptors [225]. This distribution of presented Netrins is used as a navigational cue by outgrowing pioneer neurons, as was demonstrated for dMP2 [224]. Thus, Frazzled acts in a non-autonomous fashion in the context of longitudinal pioneer growth. Whether it is also the receptor used by the pioneering growth cones has not been explored [224]. In addition to responding to external cues such as glial surfaces and Netrins, the axons of the four pioneer neurons interact with each other in a stereotypical way, involving sequential steps of fasciculation, defasciculation and refasciculation (details in Figure 5b) [219,223]. Two factors seem essential for this complex fasciculation behaviour. First, defasciculation of axonal tracts depends crucially on longitudinal glial cells [223], potentially through physical intercalation of glial protrusions. Second, CAMs are expressed on longitudinal axons in a dynamic fashion and are believed to be the immediate factors regulating selective fasciculation processes. These CAMs comprise N-Cadherin [226], Neurotactin (bearing a catalytically inactive cholinesterase-like domain) [227,228], the GPI-anchored molecules Fasciclin 1 (composed of four Fasciclin 1 domains) [229], Connectin (containing leucin-rich repeats) [230,231], and the three immunoglobulin family members Neuroglian (orthologue of vertebrate L1), Fasciclin 2 (orthologue of vertebrate N-CAM) and Fasciclin 3 [232-235]. Of these, N-Cadherin is strongly expressed throughout the neuropile and its loss causes aberrations of longitudinal fascicles (Figure 4, xxxiv) [226]. Fasciclin 2 is initially expressed only in the pCC pioneer neuron (and its sibling aCC) [234], but its expression gradually spreads to further neurons, constituting a final number of at least 11 longitudinal Fasciclin 2-positive axon tracts in the mature embryonic ventral nerve cord (Figure 4, ix) [236]. In the absence of Fasciclin 2, the selective fasciculation behaviour of the pioneer neurons is considerably impaired, although this does not have other obvious developmental consequences [8,206]. An explanation for this mild phenotype seems to be redundancy of factors involved in axonal pathfinding, as best exemplified by Neurotactin. Neurotactin acts as a homophilic CAM in cell culture assays, but this interaction requires its heterophilic binding to the secreted immunoglobulin protein Amalgam [237-240]. Both ligand and receptor are widely expressed in the developing nervous system, but their absence causes no serious detectable phenotypes. However, neurotactin mutant phenotypes emerge in combination with mutations in other neuronal growth-related genes [241]. Thus, embryos carrying loss-of-function mutations of both neurotactin and neuroglian display fasciculation defects of dMP2 and MP1 axons and disrupted connectives at high frequency (Figure 4, xxxv) [241]. Further strong neurotactin mutant phenotypes were revealed in combination with mutations of genes encoding the receptor tyrosine kinase Derailed, the cytoplasmic tyrosine kinase Abelson, or the CAM-like molecule Kekkon, but not with a number of other tested factors (Ptp69D, Ptp99A, Fasciclins 1–3, Pollux and Neuromusculin) [241,242]. As illustrated by neurotactin's genetic interactions, the range of molecules involved in longitudinal guidance of central axons in Drosophila clearly goes beyond adhesion factors, and this is supported by a number of further mutant phenotypes. For example, lack of the transmembrane protein Semaphorin-1A and its receptor PlexinA causes disruption of a lateral Fasciclin 2 tract (Figure 4, xxxiii), whereas lack of PlexinB (the Semaphorin-2A receptor) causes defasciculation of a median Fasciclin 2 tract [243-245]. Selective manipulations of [Ca2+]i-dependent Calmodulin function in the four pioneer neurons affect both their growth ability and fasciculation behaviour [246]. Embryos carrying mutations in the genes wnt5 or glaikit (encoding a phospholipase D superfamily protein) show severe disruption of connectives and strong defasciculation phenotypes [247,248]. Also, embryos lacking Kuzbanian, a secreted metalloprotease of the ADAM family, or the two receptor-linked protein tyrosine phosphatases Ptp10D and Ptp69D, show severe aberrations of the neuropile. However, in these latter two cases the aberrations are not caused through aberrant growth or fasciculation behaviours of the pioneer neurons, but rather due to the misguidance of neurons growing at subsequent stages [249,250]. The expression of these various neuronal receptors, adhesion molecules and their associated molecules involved in the neuronal response to environmental cues must be precisely regulated by pioneer neuron-intrinsic gene regulatory programmes. One candidate regulator is the neuronal transcription factor Longitudinals lacking (Lola). In the absence of Lola, longitudinal glia cells form normally, but pCC, vMP2 and MP1 growth cones stall dramatically and connectives fail to form [23,251]. Lola encodes a variety of isoforms that are expressed within subsets of neurons and these are likely to regulate the expression of many guidance molecules [252]. As discussed above, all these genetic pathways implicated in longitudinal fascicle formation eventually have to act on the cytoskeletal machinery of the growing neurons (Table 1 and Figure 3). In agreement with this assumption, genetic manipulations of various cytoskeletal regulators cause severe disturbances of the longitudinal tracts, as was demonstrated for Chickadee/Profilin, Abelson, members of the the Arp2/3 or SCAR/WAVE complexes, Futsch/MAP-1B, Rho-family GTPases, such as Cdc42, combinations of Rac1, Rac2 and Mtl, or the Rac activator Trio (Figure 4, xxxvi to xl) [95,101,132,134,163]. In conclusion, a clear temporal pattern of the cellular events and requirements in longitudinal pathway formation has been described at high resolution. A good number of genes, comprising adhesion, signalling or cytoskeletal regulatory factors, have been demonstrated to mediate these precise events. These reveal a significant role for interactions mediated by cell adhesive molecules to direct the neuronal fasciculation decisions necessary during the formation of longitudinal fascicles. These decisions are regulated by additional contact mediated signals and effected by the activation of cytoskeletal regulators. Mediolateral patterning of longitudinal fasciclesIndividual longitudinally projecting axons are positioned at particular locations within the connectives. This is mediated both by selective fasciculation behaviour and the use of protein gradients that establish positional information within the connectives to dictate where individual axons project. As explained above, longitudinal pioneer axons are initially grouped into one fascicle, but they subsequently separate into distinct fascicles in characteristic dorso-ventral and medio-lateral positions (Figure 5b) [219,223]. This may suggest a gradual increase in distinct positional values in the neuropile of the ventral nerve cord. In agreement with this hypothesis, the expression of the three paralogous Roundabout receptors (Robo, Robo3, Leak/Robo2) is largely overlapping at early stages of axonal growth. However, at later stages, these three receptors take on clearly distinguishable expression domains, that is, Leak/Robo2 becomes restricted to the lateral area, Robo3 spans from lateral to medial, and Robo covers the whole width of the neuropile (Figure 4, ii to iv) [253,254]. This provides a combinatorial code based on the number of these three Robo receptor types expressed by each particular axon, thus dictating their spatial arrangements within the connectives (Figure 5e). Experimental changes of the Robo receptor composition/dosage expressed by single neurons are sufficient to shift the position of their projections in the medio-lateral plane (Figure 4, xxix) [205,254-256]. During normal development, the state of Robo receptor expression of individual neurons appears to be determined by their cell-autonomous programmes. Thus, the expression or non-expression of the fate-determining transcription factor Atonal in chordotonal versus multidendritic sensory neurons differentially regulates their distinct medio-lateral fasciculation behaviours by instructing the expression state of the Robo3 receptor in a cell type-specific manner [255]. Similarly, the Lola transcription factor has a role to regulate the expression of Robo within central neurons [257]. The ligand to all three Robo receptors is the secreted factor Slit, which is released from glia cells in the midline of the ventral nerve cord where it is present at its maximal concentration (Figure 5e) [258]. It is also present, although at much lower levels, within the connectives, and this localisation is partly mediated by the heparan sulfate proteoglycan Syndecan (Figure 4, vi) [259]. Axons expressing a higher dosage or greater number of Robo receptor types are more sensitive to the Slit repellent and are positioned further from the midline (Figure 5e) [253,254]. In addition, Robo function inactivates N-Cadherin-mediated adhesion [260]. Thus, higher Robo activity weakens the adhesive interactions driving fasciculation. This might allow growth cones with high Robo expression to progress further laterally and aid in regulating their mediolateral positioning. Consistent with this hypothesis, N-cadherin mutant embryos display aberrations of longitudinal fascicle patterns (Figure 4, xxxiv) and, at the identified cell level (apterous-expressing neurons), growth cones alter their medio-lateral pathfinding and longitudinal fasciculation behaviour in stereotypical ways (Figure 4, xxxi) [226]. It remains to be seen whether additional adhesion molecules are regulated similarly through Robo-dependent pathways. Interestingly, axons shifted in the medio-lateral plane through targeted manipulations of their Robo genes do not project randomly but seem to partition in an organised way with a different group of axons, as suggested by their fasciculated appearance (Figure 4, xxix) [205,254,255]. A potential molecular explanation for this phenomenon is that the types and patterns of adhesion molecules is repeated within different mediolateral positions, allowing shifted axons to find an appropriate fascicle even in an inappropriate Robo zone. In conclusion, both short-range interactions as well as long range organiser activity contribute to the spatial arrangement of longitudinal fascicles along the medio-lateral axis of connectives. Growth cone guidance at the Drosophila CNS midlineThe majority of axons cross the midline within either of two commissures per segment, called anterior and posterior commissures (Figures 1b, 4c,d and 5i). The formation of these commissures begins at embryonic stage 12 and involves dynamic but reproducible interactions between pioneering growth cones, neuronal cell bodies and migrating glia cells [28], and a dynamic up- and downregulation of cell adhesion molecules, such as Fasciclin 1, during the period of axonal invasion of these commissures [8,261]. Each neuron makes a stereotypic choice as to whether to cross the midline and, for those that cross, whether to grow through the anterior or posterior commissure (Figure 5c,d). Within each commissure axons are arranged in fascicles, as illustrated by the expression pattern of molecules such as the adhesion molecule Connectin, which is restricted to specific sub-fascicles within commissures [230,231]. The Wnt5/Derailed signalling pathway (Figure 5f) plays an important role to regulate growth cone choice of anterior versus posterior commissure. Axons of the anterior commissure express the atypical receptor tyrosine kinase Derailed, which mediates their repulsion from the posterior commissure (Figure 4, xix) [208,262]. A ligand of Derailed that has been shown to mediate this commissure-specific repulsion is Wnt5, which is enriched in the area of the posterior commissure (Figure 4, xx) [248,263]. Absence of Derailed or Wnt5 causes random misrouting of axons from the anterior towards the posterior commissure (Figure 4, xviii), whereas Derailed gain-of-function repels axons from the posterior commissure (Figure 4, xxii and xxvi) [208,248,263]. Obvious candidate factors organising the intrasegmental distribution of guidance cues along the antero-posterior axis are the segment polarity genes. Consistent with this assumption, embryos mutant for the segment polarity gene gooseberry have a severely reduced posterior commissure (Figure 4, xvii) [23,264]. With respect to the axonal choice of whether to cross the midline or stay ipsilateral, two major regulatory pathways, the Netrin/Frazzled and Slit/Robo systems, have been identified. Cells at the midline of the ventral nerve cord express the secreted factors Netrin A and B, in the absence of which commissure formation is clearly reduced (Figure 4, vii and xv) [265,266]. Although the Netrin molecules can influence the growth of axons some distance from the midline, this does not appear to be necessary for midline crossing in Drosophila, as membrane tethered versions of Netrin that necessarily act over a short-range are sufficient to attract axons across the midline [267]. In contrast, long-range signalling activity of Netrin is necessary for commissural axons to traverse the floor plate of the vertebrate spinal cord [268]. Netrin A and B act as attractive ligands to CNS axons expressing the Frazzled receptor to bring them across the midline (Figure 4, viii and xvi) [225]. However, not all neurons are attracted by Netrins, and neurons expressing the alternative receptor Unc5 are repelled by them (Figure 4, xxx) [212]. First insights are being gained into the spatial regulation of Frazzled expression. Thus, in neurons of the posterior commissure, Frazzled (together with Trio and Enabled) is directly upregulated by the segmentation gene Engrailed [269]. Another major factor guiding axons at the CNS midline and opposing the effects of Netrins is the secreted repellent factor Slit, which is also expressed by the glia cells at the midline of the ventral nerve cord (Figure 4, vi) [258,270]. In the absence of slit function, all axons of the ventral nerve cord collapse towards the midline (Figure 4, xiv) [23]. Slit activity to control midline crossing is mediated by the two immunoglobulin domain containing receptors, Robo and Leak/Robo2 (see 'Mediolateral patterning of longitudinal fascicles' above). Absence of these two proteins phenocopies the slit mutant phenotype (Figure 4, xi) [205,271,272]. Thus, Slit is required to prevent axons from inappropriately reaching the midline; it acts to maintain all ipsilateral growth cones away from the midlin |
|
|
|
 |   | Log on / register BioMed Central home | Journals A-Z | Feedback | Support | My details |
on Google Scholar
author email
corresponding author email
Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.