A flexible genetic toolkit for arthropod neurogenesis

Abstract

Arthropods show considerable variations in early neurogenesis. This includes the pattern of specification, division and movement of neural precursors and progenitors. In all metazoans with nervous systems, including arthropods, conserved genes regulate neurogenesis, which raises the question of how the various morphological mechanisms have emerged and how the same genetic toolkit might generate different morphological outcomes. Here I address this question by comparing neurogenesis across arthropods and show how variations in the regulation and function of the neural genes might explain this phenomenon and how they might have facilitated the evolution of the diverse morphological mechanisms of neurogenesis.

1. Introduction

During neurogenesis neurons and glia are formed and their number, types, morphologies, final positions and connections shape the architecture of the brain. Consequently, variations in the morphology and function of metazoan nervous systems must originate from evolutionary changes of the developmental processes that generate the neural cells and regulate their differentiation.

In order to understand the causes and sequence of evolutionary variations, Volker Hartenstein and I, in a recent review [1], compared early neurogenesis across the animal kingdom by subdividing the developmental processes of early neurogenesis into distinct morphological modules. Neurogenesis starts with patterning cells with neurogenic potential (module A) and further early steps include the patterning (module B), proliferation (module C) and movement (module D) of neural progenitors. Variants of these modules can be detected across the animal kingdom and allow conclusions to be drawn on the mechanisms of neurogenesis in the urbilaterian [1]. For example, the module ‘patterning cells with neurogenic potential’ (module A) is present as three different variants. In animals like cnidarians that have a nerve net, the whole ectoderm has neurogenic potential (‘generalized neurogenic ectoderm’, module A1), whereas arthropods and vertebrates have restricted ectodermal domains that generate the central nervous system (CNS) (‘neuroectoderm’, module A2). In some clades, the neurogenic potential is restricted to subectodermal cells (module A3) such as the neoblasts in flatworms (Platyhelminthes) [1].

Interestingly, the genetic toolkit of early neurogenesis is highly conserved in all animal phyla despite significant variations in the morphological modules. In this review, I will address the question of how the various morphological modules might have emerged and how the same genetic toolkit might generate the different modules of neurogenesis, using arthropods as model organisms. Arthropods are particularly well suited to analyse this question because of the large size of the phylum, great variations in shapes and behaviours, and the availability of comparative data on neurogenesis for all groups [1,2]. Furthermore, large-scale molecular phylogenies have greatly improved the resolution of arthropod relationships and thus facilitate evolutionary investigations of arthropod neurogenesis (figure 1) [3]. It is now widely accepted that the paraphyletic crustaceans group with the insects forming the pancrustacean clade (also named Tetraconata clade). The pancrustaceans are a sister group of the myriapods and are united with the latter in the Mandibulata group [3–6]. The chelicerates represent a basal branch and are a sister group of the Mandibulata. Insects, crustaceans, myriapods and chelicerates together compose the euarthropods. The onychophorans (velvet worms) are the sister group of the euarthropods and together with the latter form the phylum Arthropoda (figure 1). The following discussion of the neurogenesis modules is based on this phylogenetic framework (but see [7–10] for discussion).

Figure 1.

Arthropod phylogeny and the evolution of the neurogenesis module. The arthropod relationships shown in the phylogenetic tree are the framework for the discussions in this review. A2–D4 represent parts of the module variants recently described by [1]. Module A: patterning cells with neurogenic potential. In contrast with animals with a generalized ectoderm (module A1, not shown), all arthropods have a restricted area with neurogenic potential (neuroectoderm) that gives rise to the central nervous system (module A2). Module B: patterning of neural progenitors. Neural progenitors/precursor groups show a similar, invariant pattern in all euarthropods (module B2), while the neural progenitors of onychophorans appear in random positions (module B1). The ancestral pattern of euarthropod neurogenesis is the selection of neural precursor groups. Modules C and D: proliferation and movement of neural progenitors, respectively. The neural precursors of chelicerates and myriapods (mainly) do not divide (module C1) and directly differentiate into neurons and glia. The neural precursor groups are internalized by ingression (module D3) or invagination (module D4). Both mechanisms are seen in chelicerates (including pycnogonids) and myriapods. In addition, pycnogonids have asymmetrically dividing progenitors (module C3) that remain in the epithelium (module D1). By contrast, the neural progenitors of onychophorans divide symmetrically (module C2) after their delamination (module D2). Insect and crustacean neuroblasts divide asymmetrically (module C3), however insect neuroblasts delaminate (module D2), while crustacean neuroblasts remain in the epithelium (module D1). The scheme and description refer to the processes in the ventral neuroectoderm.

2. Patterning regions with neurogenic potential

In arthropods, the procephalic and the ventral neuroectoderm generate the tripartite brain and the ventral ganglia, respectively (module A2; figure 1) [11–13]. The size of the neuroectoderm differs from a few cell rows in crustaceans and insects to wide areas containing many small cells in myriapods and chelicerates (figure 2) [14–19]. Data from insects and a representative of a basal branch of the euarthropods, the spider Parasteatoda tepidariorum (chelicerate), show that dorsoventral (DV) cell fates, including the regions with neurogenic potential, are determined by morphogen gradients involving Dpp/Sog [20–24]. The determination and patterning of the ventral neurogenic region by Dpp/Sog might, therefore, be an ancestral feature of arthropod neurogenesis. This is further supported by data from other Ecdysozoa, Lophotrochozoa and Deuterostomia, which indicate that the BMP(Dpp)/chordin(Sog) morphogen system is conserved across bilaterians. However, there seem to be differences in the degree of the contribution of the dpp/sog homologues in patterning the neurogenic region in arthropods. In Drosophila melanogaster sog null mutants, the DV extension is slightly reduced [22], while in the flour beetle Tribolium castaneum and in the spider (chelicerate) P. tepidariorum, sog RNAi results in an (almost) complete loss of the ventral neuroectoderm [20,23]. Data are not available for myriapods and data from crustaceans are inconclusive because the only sog gene that has been identified shows a restricted expression in the mandibular ventral midline cells [25]. The fact that the regulation of DPP signalling has considerably diverged even within insects [23,24] suggests that evolutionary variations in the regulation of the DV patterning genes and their targets might correlate with the different extensions of the neurogenic regions in arthropods (figure 2). This assumption is further supported by comparative studies of the DV extension of the ventral midline in Diptera and Hymenoptera [26]. While in D. melanogaster and in the mosquito Anopheles gambiae the ventral midline is 1–2 cells wide along the DV axis, it extends over 5–6 cells in the honeybee Apis mellifera. This difference is because of variations in the regulation of single-minded—a transcription factor which is essential for the development of the midline cells—by DV patterning genes [26].

Figure 2.

Patterning of the neurogenic region in euarthropods. The dashed lines indicate the ventral midline of the neuroectoderm. (a) Flat preparation of a Glomeris marginata embryo (millipede) showing the RNA expression pattern of the SoxB1 gene in all neuroectodermal cells. (b,c) Comparison of the msh RNA expression pattern in flat preparations of three ventral neuromeres of G. marginata and Tribolium castaneum (beetle). Note the difference in the width of the neuroectoderm in these species. The boxes frame the neuromere of the second leg segment in G. marginata and the second thoracic neuromere of T. castaneum, respectively. (d,e) Schematic of the arrangement of the neural precursor groups in hemi-neuromeres of G. marginata and the spider Cupiennius salei. The precursor groups expressing msh are shown in brown. (f,g) Schematic drawings comparing the expression of the dorsoventral patterning genes vnd (blue), ind (red) and msh (brown) in the neuroblasts of Drosophila melanogaster and T. castaneum. The neuroblasts 6-2 and 7-2 express both vnd and ind in D. melanogaster, while neuroblasts 4-2 and 5-3 express ind and msh in T. castaneum. l1–l3, leg segment 1 to 3; pc, procephalic neuroectoderm; t1–t3, thoracic segments 1 to 3; vne, ventral neuroectoderm.

In all arthropods the neurogenic potential differs along the anterior–posterior (AP) axis. The (anterior) procephalic neuroectoderm not only generates a higher and more diverse number of neurons than the ventral neuroectoderm but also forms the complex brain centres such as the mushroom bodies that integrate sensory information and regulate behaviour [27]. In vertebrates, the Wnt/β-catenin pathway patterns the CNS along the AP axis and in some species determines the overall AP cell fates (reviewed in [28]). An involvement of this signalling pathway in patterning the AP axis has also been demonstrated in lophotrochozoans, in particular in regenerating planarians, where the reduction of Wnt signalling results in radial hypercephalized individuals [29].

In arthropods, Wnt signalling seems to regulate posterior elongation and segmentation but is neither required for the formation of anterior (head) segments nor the overall AP patterning of the CNS [28,30]; rather, the function of Wnt in establishing AP fates in the arthropod CNS seems to be confined to neuromeres, the metameric units of the CNS [31–34].

The mechanisms establishing different AP fates in the anterior part of the embryo show considerable variations within and between arthropod groups. The AP morphogen Bicoid, for example, is not present outside of dipterans and Hunchback, another head morphogen, does not regulate the expression of AP patterning genes (e.g. Hox genes) in the spider P. tepidariorum, in contrast with the case in D. melanogaster [35,36]. If the overall AP patterning mechanisms are connected with the regionalization of the CNS, these differences might contribute to existing variations of the arthropod CNS. In D. melanogaster, AP axis formation is in fact directly linked to the AP regionalization of the CNS. AP morphogens (e.g. Bicoid, Hunchback) regulate the genes that pattern the CNS along the AP axis (e.g. orthodenticle, ems, Hox genes) (e.g. [37,38]). However, there are no data available on how overall AP patterning is linked to CNS regionalization in arthropods other than insects.

Thus, while the overall function of BMP in DV axis formation and patterning of the neurogenic region seems to be conserved across the animal kingdom, the function of Wnt in AP patterning of the CNS has been restricted to the metameric units in arthropods. Furthermore, the upstream mechanisms that set up the initial differences in CNS AP fates have diverged in arthropods.

However, many of the genes downstream of the primary AP/DV subdivision are conserved across the animal kingdom, such as the transcriptional regulators of the SoxB family, which maintain the proliferative state of the neuroepithelium and at the same time inhibit neural differentiation (figure 2) [39–42].

Furthermore, many genes that translate the positional information into neural/lineage subtype identity are conserved in metazoans. Downstream of the primary AP and DV patterning mechanisms, additional signalling pathways as well as homeodomain proteins establish neuronal subtype identity [43]. For example, the function of Hox genes in determining segmental identity is conserved across the animal kingdom [43]. However, although these genes seem to be an integral downstream part of the molecular processes that pattern the neurogenic region along the AP axis of the CNS in all arthropods (e.g. [44–47]), there are considerable functional variations even within individual arthropod groups. For example, the expression of the Hox gene labial in the head segment that gives rise to the tritocerebrum is conserved in all arthropods [48]; however, studies in insects show divergent functions of Labial in brain development. While in D. melanogaster labial mutants the tritocerebrum and additional brain structures are affected, labial RNAi does not show a phenotype in the milkweed bug Oncopeltus fasciatus and the functional defects are restricted to the tritocerebrum in the flour beetle T. castaneum [49–51].

The genes patterning the neurogenic region along the DV axis are also highly conserved across the animal kingdom [52]. Yet again, detailed comparative analyses of the expression of the DV patterning genes muscle segment homeobox (msh), intermediate neuroblasts defective (ind) and ventral nervous system (vnd) defective in arthropods revealed considerable variations in the expression of these genes relative to the neural progenitors/precursors in the respective domains (figure 2) [53,54]. Furthermore, evolutionary changes have occurred in the neuronal subtype specific genes that are regulated by the DV patterning genes. For example, msh regulates the expression of the motor- and interneuronal marker islet in the spider Cupiennius salei, whereas this is not the case in D. melanogaster [53].

3. Patterning of neural precursors/progenitors

Neural precursors/progenitors can appear in either random (module B1) or fixed (module B2) positions within regions of neurogenic potential (figure 3) [1]. Furthermore, single (modules D1 and D2) or groups of contiguous precursors can be selected (modules D3 and D4). In the arthropod neuroectoderm, different types of founder cells of the nervous system are generated over a period of time ranging from hours (e.g. in the insect D. melanogaster) to months (e.g. in the onychophoran Euperpatoides kanangrensis) [55]. In chelicerates and myriapods groups of contiguous mainly postmitotic neural precursors are selected in the neuroectoderm, while in onychophorans, crustaceans and insects single neural progenitors are selected that divide either symmetrically or in a stem-cell like mode (see below) (figure 3) [7,14,15,17,18,55,56]. In onychophorans, the sister group of euarthropods (chelicerates, myriapods, crustaceans, insects), neural progenitors appear in random positions in the neuroectoderm and their number per hemi-neuromere is variable (60–100 progenitors; figure 3). By contrast, in all euarthropods neural progenitors/precursors are generated in fixed positions and show species-specific invariable numbers within the same neuromeres (figures 2 and 3).

Figure 3.

Pattern of neural precursor/progenitor specification and division. Small sections of the neuroectoderm (white squares: neuroectodermal cells) are shown for each group. (a) Onychophorans: ASH shows a low homogeneous expression in the neuroectoderm (light grey) and is upregulated in the neural progenitors (dark grey) after their delamination. Chelicerates: ASH is upregulated in large domains and becomes restricted to groups of cells. A similar pattern is seen in myriapods (not shown). Insects: ASH is expressed in small proneural clusters. Expression becomes restricted to single neuroblasts. Crustaceans (branchiopods): ASH is upregulated after neuroblast specification. Over time more neuroblasts express ASH and the gene is expressed at different levels indicated by lighter and darker grey. (b) Division pattern and movement of neural precursors/progenitors. Onychophorans: single neural progenitors delaminate and divide symmetrically to produce intermediate neural precursors, which divide again. The expression of snail and prospero has not been studied in onychophorans. Chelicerates: most neural precursors are postmitotic. Neural precursor groups express snail (orange). Neural precursors co-express prospero (red) before detaching from the group. Insects: neuroblasts express snail and prospero and divide asymmetrically to produce GMCs, which divide again to produce neurons and glia. Crustaceans show the same division pattern but snail is expressed long before prospero. Snail-positive neuroblasts can divide symmetrically.

The morphological differences can be explained by variations in the regulation of two classes of genes, the achaete-scute homologues (ASH) and the Notch signalling pathway. These genes are used for the generation and regulation of neural cells throughout the animal kingdom; however, they show variations in their spatial and temporal expression (reviewed by [1]). In insects, chelicerates and myriapods the invariant spatial–temporal expression of ASH genes sets up the areas where neural progenitors/precursor groups form [17,18,57,58]. The positions of the neural progenitors/precursors are refined by a transcriptional feedback loop between the Notch signalling pathway and the ASH genes, which results in the selection of regular arranged, spaced neural precursor groups/progenitors [59,60]. In branchiopod crustaceans, the ASH genes are not expressed before neural progenitors (neuroblasts) are formed and therefore are not involved in positioning them (figure 3) [15]. However, an invariant spatial pattern of ASH expression appears with a time delay. Neuroblasts are also arranged in an invariant pattern in another crustacean group, the malacostracans, but molecular expression data are not available [14].

By contrast, in the onychophoran E. kanangrensis, ASH is expressed at very low, homogeneous levels in the neuroectoderm [55]. High levels of expression can be detected and are maintained in the neural progenitors after their delamination, suggesting that Ek ASH is required for neural progenitor maturation (figure 3).

Two mechanisms might contribute to the variations in patterning neural precursors/progenitors in the neuroectoderm in arthropods. Firstly, the variations could be due to changes in the prepatterning mechanisms regulating ASH expression. Given the homogeneous low ASH expression in onychophorans, molecular changes in the achaete-scute homologues must have occurred in the lineage leading to the euarthropods that facilitated the spatio-temporal regulation of these genes allowing for the upregulation of transcripts in invariant positions. However, data on ASH regulation in the embryonic CNS are fragmentary in insects and not available at all in other arthropods. Skeath et al. [61] report that the D. melanogaster ASH genes achaete, scute and lethal of scute are regulated by the products of the pair-rule and DV axis patterning genes (e.g. dpp) in the ventral neuroectoderm and their expression patterns are maintained by the segment polarity proteins. In the D. melanogaster procephalic neuroectoderm, lethal of scute, which is the main proneural gene in the brain, is regulated by the head gap genes tailless (tll), orthodenticle (otd), empty spiracles (ems) and buttonhead (bhd) [62]. It can be speculated that evolutionary changes in the spatial–temporal expression of ASH genes are associated with variations in the expression and interactions of the segmentation genes, which have considerably diverged in arthropods, in particular regarding the gap and pair-rule genes (e.g. [63–69]).

However, segmentation genes might not be used at all for positioning neural precursors/progenitors in some groups. For example, in the neuroectoderm of the branchiopod crustacean Daphnia magna, the default neural cell fate is suppressed in the central area of the hemi-neuromeres by the activation of Notch signalling so that neural progenitors are arranged in a ring-like structure (figure 3). This pattern seems to be representative for branchiopods; however, in malacostracan crustaceans the neuroblast arrangement is not disrupted by a non-neural central area, suggesting that the mechanisms of neuroblast positioning have diverged in crustaceans [14].

Secondly, variations in the regulation of Notch activity could underlie the variation in the pattern of neural precursors/progenitors in arthropods.

The core mechanism of the Notch signalling pathway is the activation of the Notch receptor by the ligand Delta, which results in the intramembrane proteolysis of the receptor and the release of an intracellular fragment that acts as a transcriptional regulator [70]. The interactions between the transmembrane proteins Notch and Delta are complex, and different regulatory mechanisms are used in different developmental processes such as ligand–receptor trans- and cis-activation, receptor cis-inhibition by the ligand or mutual receptor–ligand cis-inhibition [70,71].

Biological and mathematical models show that the selection pattern of precursors can be modulated by regulating Notch signalling in different ways [71,72]. The selection of regular arranged, spaced neural precursor groups/progenitors can be explained by the ‘lateral inhibition model’ where Notch signalling and the ASH genes are linked in a transcriptional feedback mechanism [70]. In insects, the ASH genes are expressed in small clusters in the ventral neuroectoderm. One of the cells adopts the neural fate and expresses higher levels of ASH, which in turn leads to an increase of Delta expression [59]. Delta trans-activates the Notch receptor in the neighbouring cells. The products of the Notch target genes, the hairy/Enhancer of Split (HES) genes, downregulate ASH and inhibit the neighbouring cells from adopting the neural fate. In chelicerates and myriapods, lateral inhibition operates between the neural precursors and the neuroectoermal cells surrounding the selected precursor group [18,60]. However, the neural precursor group itself seems to be insensitive to Notch signalling. This is reflected in the high levels of Delta expressed by the precursor groups suggesting that the Notch receptor is cis-inactivated [60,71].

In onychophorans, the homogeneous, low expression of ASH and the strong expression of Notch and Delta in the ventral neuroectoderm suggest that Notch signalling keeps ASH expression at low levels in all cells (figure 3) [55]. Random changes in the sensitivity to Notch signalling, possibly by cis-inactivation of the Notch receptor by the ligand Delta, could allow for individual cells to adopt the neural fate and delaminate. This mechanism could explain the seemingly random positions and great range in progenitor numbers per hemi-neuromere.

The pattern of neural precursors/progenitors is much more complex in the developing brain. In chelicerates and myriapods, the same types of founder cells (neural precursor groups) are visible in the procephalic and ventral neuroectoderm [73]. In both neurogenic areas, the ASH genes are expressed in large domains from which several neural precursor groups arise (although the underlying morphological processes are more complex in the procephalic neuroectoderm, see below). By contrast, in insects, ASH genes are expressed in large domains (rather than small clusters) in some areas of the procephalic neuroectoderm, from which several neuroblasts delaminate [74]. Furthermore, in the dorsal midline of the procephalic neuroectoderm of D. melanogaster, groups of neural precursors form within a large lethal of scute domain [74]. Yet again, different patterns and regulations of ASH and Notch signalling can be seen in different ontogenetic stages (e.g. larvae) and in the sensory nervous system (e.g. [75,76]). The mechanisms seen in the developing insect brain show similarities to the ancestral patterns of neurogenesis described in onychophorans and chelicerates/myriapods.

These data have led to the hypothesis that the neural precursor groups of chelicerates and myriapods have evolved from areas of massive, random neural progenitor delamination due to changes in the regulation of the ASH genes and in the regulation of Notch signalling [55]. There is strong evidence that neural precursor groups were present in the last common ancestor of euarthropods and thus represent the ancestral pattern of euarthropod neurogenesis [55]. Additional changes in the regulation of the participating genes have led to the evolution of neuroblasts from the neural precursor groups in the last common ancestor of insects and crustaceans. Thus, changes in the prepatterning mechanisms and the versatility of the Notch signalling pathway might explain the various patterns of neural precursors/progenitors in arthropods.

4. Proliferation and movement of neural precursors

Arthropods show also variations in the developmental modules following neural determination. Selected cells can either differentiate directly into neural cells (neurons/glia) (module C1), as is the case for many of the neural precursors in chelicerates and myriapods, or they can undergo symmetric (e.g. onychophorans) or asymmetric divisions (insects, crustaceans) (modules C2 and C3; figure 1). They can also show combinations of these variants (e.g. sea spiders; figure 1). The associated movements either of the neural precursors/progenitors themselves and/or of their progeny lead to the formation of internal cell layers, basal to the neuroectoderm (figure 3). The movements show variations that are partially linked to the differences in the specification patterns of neural precursors/progenitors. Insect neuroblasts start to divide asymmetrically to produce ganglion mother cells (GMCs) after segregating individually from the neuroectoderm (delamination; modules C3 and D2), while crustacean neuroblasts remain in the neuroectoderm and generate GMCs at the basal side by directed asymmetric mitosis (module C3 and D1; figures 1 and 3) [14–16,77]. Onychophoran neural progenitors delaminate before they divide symmetrically to produce intermediate neural precursors (modules C2 and D2; figures 1 and 3) [7,55]. By contrast, there are no cell divisions within the neural precursor groups of chelicerates and myriapods (module C1) and only few neural precursors divide after detaching from the precursor groups (ingression; module D3) or the internalization of the whole group (invagination; module D4; figures 1 and 3).

However, in myriapods single and groups of mitotic cells seem to prefigure the areas where neural precursor groups are selected, suggesting a link between the formation of neural precursors and cell proliferation [18]. However, the division pattern has not been analysed in detail and it is not known if the neural precursors within a group are related by lineage. Interestingly, sea spiders, which are most probably the sister group of all other chelicerates (euchelicerates), show a combination of asymmetrically dividing neural progenitors and neural precursor groups [78]. However, in this case, neural precursor groups are generated before the neural progenitors appear.

Despite these differences, committed neural progenitors/precursors express a set of conserved factors that were shown to be involved in asymmetric division and neural cell fate determination in D. melanogaster [79]. This raises the question of how these genes can support the various morphological outcomes in all euarthropod groups (data are not available for onychophorans).

In the following, I will discuss a few of the factors for which comparative data exist in euarthropods. In D. melanogaster, Asense activates many genes in neuroblasts, among others those that are required for asymmetric division [80]. The neuroblasts immediately delaminate after their formation and start to divide (figure 3). The neural differentiation factor Prospero is asymmetrically distributed into the GMCs where it enters the nucleus, acts antagonistically to Asense and promotes cell-cycle exit [81–83]. The Snail family members are required for the mitotic activity of the neuroblast and the spindle rotation, which occurs prior to the production of GMCs [79]. The asymmetric distribution of Prospero to GMCs is affected in mutants carrying a deletion of three Snail family genes [79]. Furthermore, the Snail family members might be directly involved in triggering the delamination of neuroblasts by downregulating E-cadherin [84,85].

Although crustacean neuroblasts do not delaminate, they show the same division pattern as insect neuroblasts, and it is therefore conceivable to assume that the function of Snail and Asense in asymmetric cell division is conserved in both groups. Furthermore, prospero is asymmetrically distributed into the GMCs in crustaceans and its expression coincides with the onset of asymmetric division of the neuroblasts, suggesting that it acts as a neural cell fate determinant similar to the case in insects [15,86]. However, Snail must have an additional earlier function in crustacean neurogenesis because it is the earliest gene to be expressed in neuroblasts and there is a considerable time delay between neuroblast formation and the start of asymmetric divisions [15,86]. The fact that snail is expressed before ASH suggests that the former rather than the latter gene is required for neural cell fate determination. This potential function of Snail is supported by data from D. melanogaster showing that Snail can activate neuroblast fate genes such as grainyhead, numb, prospero and seven-up [80]. Furthermore, Snail could be involved in the symmetric divisions of crustacean neuroblasts, which occur before the start of GMC production (figure 3) [15]. The function of Snail in regulating cell-cycle genes and the proliferative divisions of stem cells has been demonstrated both in D. melanogaster and in vertebrates [80,87–89].

The delayed progression of crustacean neuroblasts to asymmetric divisions compared with insects can again be explained by changes in the use of the participating genes [86]. Notch signalling is used not only for binary cell fate decisions in the hemi-neuromeres between the central presumptive epidermal cells and the surrounding neuroblasts, but also for keeping neuroblasts in the transition phase which is characterized by the expression of neural genes but the absence of asymmetric divisions. Neuroblasts individually leave the transition phase showing strong expression of ASH, asense and prospero and start producing GMCs [86]. Both ASH and prospero expression are derepressed when Notch activity is lowered, leading to premature production of GMCs. Thus in this context, ASH seems to be required for the maturation of the neuroblasts, rather than the specification of the neural fate as is the case in insects.

Furthermore, in embryos with reduced Notch activity the epithelial morphology is disturbed, indicating that Notch signalling might have an additional function in keeping neuroblasts within the neuroectoderm [86]. This function of Notch has been described both in vertebrates and in arthropods. For example, inactivation of the Notch effector genes Hes 1, 3 and 5 in mice results in reduction of tight and adherens junctions at the apical end feed of neural stem cells and loss of these cells from the spinal cord neuroepithelium [90]. In the spider C. salei (chelicerate) neural precursor groups disintegrate prematurely in the CNS and peripheral nervous system (PNS), showing that this function of Notch can also support the formation and maintenance of neural precursor groups [17,91].

In spiders (chelicerates), most of the neural precursors expressing snail and prospero are postmitotic and the few precursors that proliferate divide symmetrically, thus excluding a function of these panneural factors in asymmetric division [92]. However, Prospero seems to act as neural differentiation factor and might promote cell-cycle exit similar to the case in insects and crustaceans because it is expressed in the nuclei of neurons and their precursors in the CNS and PNS [92,93]. Snail, on the other hand, must have a different function. Interestingly, it has been shown in vertebrates and recently also in D. melanogaster that members of the Snail family are sufficient to induce cell-shape changes leading to the delamination of neural precursors from the neuroepithelium [94–96]. It is, therefore, conceivable that Snail is required for the morphological changes that occur during invagination/ingression of the neural precursors both in chelicerates and myriapods, where snail is expressed in the same pattern [35].

Asense is neither present in chelicerates nor in myriapods; however, the spider C. salei has two ASH genes, which arose by independent duplication in the spider lineage [97]. While CsASH1 is initially expressed in large proneural domains, CsASH2 is exclusively expressed in neural precursors and is able to rescue asense-specific defects in D. melanogaster mutants [97]. This suggests that there is yet another mechanism that could account for the morphological differences of neurogenesis in arthropods—the convergent evolution of genes promoting the neural programme.

To summarize, these explanations show that the emergence of the various neurogenesis modules might have been facilitated by the diverse functions of the participating genes as well as their flexibility of acting in different morphological contexts. Furthermore, the modularity of early neurogenesis has allowed for combining various modules so that neurogenesis could be adapted to different requirements not only in the individual arthropod groups/species but also in different areas of neurogenesis in the same organism.

5. Conservation of structure despite changes of gene expression

So far, I have discussed how the expression of conserved genes in neural precursors/progenitors can lead to different morphological outcomes. However, in arthropod neurogenesis the reverse mechanism also exists: the expression of conserved genes in different patterns resulting in the same morphological outcome. This phenomenon might contribute to internal buffering mechanisms ensuring that evolutionary modifications do not result in destabilization and functional breakdown of neuronal structures. In the following, I will give two examples relating to the axonal scaffold of euarthropods.

In all euarthropods, axonal tracts are established in the developing ventral nerve chord in a highly conserved pattern [53]. Two longitudinal tracts connect the brain and the ventral ganglia. In each ventral neuromere, two transverse tracts, the anterior and posterior commissures, connect the longitudinal fascicles. During later development this arrangement can be obscured by the fusion of ganglia. The conserved axonal scaffold seems to be related to the equally highly conserved arrangement of the neural precursor groups/progenitors in euarthropods. In all four groups, they are arranged in seven rows and three to six columns in each hemi-neuromere (figure 2) [53]. Spatial and temporal cues within the neuroectoderm lead to the expression of a unique set of genes in the neural precursors/progenitors, which in turn determines the subtype identity of the progeny [43,53,54,98–100]. The neural precursors/progenitors produce early sets of pioneer neurons that establish the axonal scaffold [101]. One set of pioneer neurons is characterized by the expression of the motor and interneuronal marker even-skipped [54,102,103]. The position of eve-positive neurons is highly conserved in insects and crustaceans. However, a recent comparison of the neuroblast expression patterns in D. melanogaster and in the flour beetle T. castaneum revealed that despite their conserved positions the expression profiles of the neuroblasts have diverged considerably, including the profiles of those generating the eve-positive pioneer neurons [54,99]. Interestingly, these evolutionary changes include the expression of patterning genes which are highly conserved throughout the animal kingdom, such as the DV patterning genes msh, ind and vnd (figure 2). These data show that the early neuronal programme, which generates the pioneer neurons, tolerates considerable molecular variations, thus ensuring the formation of the conserved axonal scaffold.

In addition to the pioneer neurons, a sophisticated system of axonal guidance molecules at the ventral CNS midline plays an important role in generating the axonal scaffold in euarthropods [104–106]. For example, Netrin, a conserved chemoattractant, attracts commissural axons to the ventral midline [107]. In insects and crustaceans, specialized midline cells consisting of neurons and glia are the source of Netrin expression [106,107]. In spiders, however, neuronal midline cells are absent and the midline consists of epithelial cells [108]. Furthermore, during embryogenesis, spiders undergo a process called inversion whereby the germband splits into two halves at the ventral side along the AP axis. The two halves move apart and during this process a ventral midline epithelium appears which extends between the germband halves. After dorsal closure the germband halves come together again ventrally and the midline epithelium disappears [108]. Interestingly, commissural axons cross the space between the germband halves during inversion and Netrin, which is expressed in transverse stripes in the midline epithelium, is required for attracting the axons and guiding them across the midline [105]. This example shows an evolutionary change of gene expression to accommodate the morphological peculiarity of spider embryogenesis so that an important axonal guidance molecule can perform its conserved function in establishing the commissural tracts.

6. Conclusion

Molecular genetic changes that positively influenced the flexible use of the gene product, for example by introducing additional regulatory mechanisms and expanding the molecular interactions of the protein, might have been the starting point of the evolution of the diverse morphological mechanisms of neurogenesis. The conservation of the genetic tools of neurogenesis across the animal kingdom suggests that this might be a general principle of morphological evolution.

Competing interests

I declare we have no competing interests.

Funding

I received no funding for this study.