Introduction

The question of whether the last common ancestor of bilaterians had an anatomically complex central nervous system (CNS) is controversial. Evidence from the new molecular-based phylogeny implicates the absence of intermediate taxa at the basis of protostome–deuterostome lineage separation (Figure 1) (Adoutte et al, 2000). One important consequence is that traits homologous in arthropods and vertebrates must have been present in the common ‘urbilaterian’ ancestor from which protostomes and deuterostomes diverged. Several attempts to reconstruct the last common bilaterian ancestor and determine the origin of the CNS of organisms as different as insects and vertebrates have been made in the past. Based on differences in embryonic topography and morphogenesis of the nervous system, bilaterian animals have been subdivided into different groups thought to be characterized by the evolutionary independent origin of their nervous systems (eg Brusca and Brusca, 1990). Contrasting with this notion of independent origins is a large amount of molecular genetic data generated in several vertebrate and invertebrate model systems which indicate that key developmental processes, such as proliferation, regionalization, and specification of the embryonic nervous system, are controlled by homologous genes in vertebrates and insects (reviewed in Arendt and Nübler-Jung, 1999; Reichert and Simeone, 1999; Sprecher and Reichert, 2003). Indeed, evidence from recent experiments in Drosophila melanogaster (D. melanogaster) and mouse indicates that basic genetic mechanisms involved in embryonic brain development are conserved and suggest a common evolutionary origin of the protostome and deuterostome brain. Here we review the basic regulatory mechanisms of brain development in D. melanogaster and mouse from a comparative developmental genetic perspective. Recent expression data and functional experiments on key developmental control genes, such as the dorsoventral patterning genes, the cephalic gap genes otd/Otx and ems/Emx, or the Hox and Pax2/5/8 genes, are reconsidered in the light of a possible common origin of the bilaterian brain.

Figure 1
figure 1

Phylogenetic relationship of mouse and D. melanogaster. Simplified version of the new molecular-based phylogeny showing a selection of bilaterian phyla with the Cnidaria as outgroup. Bilaterian phyla are grouped into major cladistic classifications indicated at the right side (modified after Adoutte et al, 2000). Vertebrates and arthropods are evidenced in bold. The phylogenetic tree indicates that homologous features of mouse and D. melanogaster already existed in the common ancestor of all bilaterian animals.

Full size image

Overview of embryogenesis of the brain in insects and vertebrates

The insect brain is composed of an anterior supraesophageal ganglion and a posterior subesophageal ganglion. The supraesophageal ganglion is subdivided into the protocerebrum, the deutocerebrum, and the tritocerebrum, whereas the subesophageal ganglion is subdivided into the mandibular, maxillary, and labial neuromeres (Therianos et al, 1995; Younossi-Hartenstein et al, 1996; Campos-Ortega and Hartenstein, 1997; Reichert and Boyan, 1997). The anterior brain anlage of D. melanogaster derives from the procephalic neurogenic region, which is specified to become neuroectoderm through genetic interactions during gastrulation (Jürgens and Hartenstein, 1993). The posterior embryonic brain derives from the rostral-most ventral neurogenic region and is specified in a manner similar to that of the ventral nerve cord (Doe and Skeath, 1996). Within the cephalic neuroectoderm, single progenitor cells called neuroblasts delaminate and start to proliferate, giving rise to the developing brain of D. melanogaster.

In vertebrates, inductive interactions between germlayers during gastrulation cause an early regionalization of the developing neural tube. This leads to a rostrocaudal subdivision of the anterior neural tube into the rostral forebrain (prosencephalon or telencephalon/diencephalon) and midbrain (mesencephalon) regions and into the caudal hindbrain regions (rhombencephalon or metencephalon/myelencephalon). The developing hindbrain reveals a clear metameric organization based on seven or eight rhombomeres with pairwise compartment-like organization (Lumsden and Krumlauf, 1996). The segmental organization of the embryonic prosencephalon is still debated; however, a number of studies suggest that this region, like the hindbrain, is subdivided into six neuromeres known as prosomeres (Rubenstein et al, 1994, 1998).

Conserved dorsoventral patterning mechanisms indicate a CNS axis inversion in protostome and deuterostome evolution

One of the major arguments during the last two centuries against the common origin of the nervous systems of protostomes and deuterostomes has been the morphologically opposite position of the nerve cords in arthropods (ventral) and vertebrates (dorsal). This striking morphological discrepancy has led to the concept of two taxonomic groups, whose CNS evolved independently from a primitive common ancestor. Invertebrates exhibiting a ventrally located nerve cord such as arthropods, annelids, and mollusks were grouped into the gastroneuralia, whereas the notoneuralia include urochordates, cephalochordates, and vertebrates that are characterized by a dorsal nerve cord (Hatschek, 1891; Brusca and Brusca, 1990). This general notion was first challenged by Geoffroy St Hilaire in the early 19th century who argued, based on morphological considerations, that the ventral side of arthropods corresponds to the dorsal side of vertebrates. Molecular genetic evidence from recent developmental studies in D. melanogaster and different vertebrate model organisms have strengthened the view, that the dorsoventral bauplan of protostomes, such as arthropods, represents an inversion of the bauplan of deuterostomes, such as vertebrates. From an evolutionary point of view, this is thought to be the consequence of the inversion of dorsoventral body axis in one of the two animal groups (Arendt and Nübler-Jung, 1994; De Robertis and Sasai, 1996). One of the implications of the dorsoventral inversion theory is that the last common ancestor of protostomes and deuterostomes might already have had a centralized nervous system that was inherited to both descendant lines.

Recent developmental genetic evidence supports the dorsoventral inversion theory at two different levels of neuroectoderm specification (Figure 2). At the level of regionalization of the dorsoventral axis with respect to the presumptive neurogenic region, the early embryos of vertebrates and insects are both patterned by two opposed gradients of homologous morphogens. In accordance with the dorsoventral inversion hypothesis, the transforming growth factor β (TGFβ) family member encoded by the decapentaplegic (dpp) gene is expressed dorsally in the insect D. melanogaster, whereas its vertebrate orthologue bone morphogenetic protein 4 (BMP4) is localized at the ventral side in vertebrates. These factors are antagonized by the secreted products of the homologous genes short gastrulation (sog) in D. melanogaster and Chordin in vertebrates (Holley et al, 1995; De Robertis and Sasai, 1996; Holley and Ferguson, 1997). The site of action where sog/Chordin expression inhibits dpp/BMP4 signaling corresponds in fly and mouse to the region of the dorsoventral axis that gives rise to the neuroectoderm in the early embryos. Thus, in insects and vertebrates the antineural function of dpp/BMP4 and the antagonizing neurogenic potential of sog/Chordin seem to be conserved, whereas their expression gradients are inverted with respect to the dorsoventral body axis.

Figure 2
figure 2

Schematic representation of the molecular genetic patterning of the dorsoventral axis in vertebrates and arthropods. Only half of the body wall is represented for vertebrates and arthropods in the schematic dorsoventral sections with dorsal to the top for both animal groups. The secreted products of the homologous genes dpp/Bmp4 form a dorsoventrally inverted gradient in vertebrates with respect to D. melanogaster. They are antagonized by sog/Chordin, another homologous gene pair, from the region of the embryo that will adopt a neurogenic potential. This region is further patterned by a set of homeobox genes into medial (vnd/Nkx2), intermediate (ind/Gsh) and lateral (msh/Msx) neurogenic domains.

Full size image

A second level of dorsoventral patterning of the neuroectoderm has been found to be conserved in evolution as well. A set of homologous genes are involved in the formation of dorsoventral regions of the developing CNS in insects and vertebrates. Again, their relative expression domains are inverted in the sense of a dorsoventral axis inversion between protostomes and deuterostomes (Chan and Jan, 1999; Cornell and Ohlen, 2000). In D. melanogaster proneural clusters and early delaminating neuroblasts in the ventral neurectoderm are arranged in three longitudinal columns (medial, intermediate, and lateral) on either side of the midline cells (reviewed in Skeath and Thor, 2003). Similarly, in vertebrates, such as frog (Chitnis et al, 1995) and zebrafish (Haddon et al, 1998), proneural clusters that give rise to primary neurons are arranged in three columns on each side of the neural plate (medial, intermediate, and lateral). In D. melanogaster, the homeobox genes ventral nerve cord defective (vnd), intermediate neuroblasts defective (ind) and muscle-specific homeobox (msh) are essential for the formation and specification of neuroblasts in the medial, intermediate, and lateral longitudinal columns (Chu et al, 1998; McDonald et al, 1998; Weiss et al, 1998). In the neural plate of vertebrates, the expression of the homologous genes of the Nkx2 (vnd), Gsh (ind), and Msx (msh) families defines the medial, intermediate, and lateral neurogenic columns and are involved in their specification (reviewed in Arendt and Nübler-Jung, 1999).

The functional conservation and the similar relative expression patterns of these dorsoventral patterning genes in vertebrates and insects strongly suggest a common origin of the CNS of protostomes and deuterostomes. Accordingly, a reasonable explanation for the opposed positions of the CNS in these two animal groups is the dorsoventral axis inversion between protostomes and deuterostomes. This is further supported by independent molecular evidence from gene expression data in the developing heart of chordates and arthropods. In D. melanogaster, the homeobox gene tinman is expressed in the dorsal vessel, an insect equivalent of the vertebrate heart (Bodmer, 1993). Csx (also called Nkx2.5) is a murine orthologue of tinman, and is expressed in the ventrally located heart primordium of the mouse embryo (Tanaka et al, 1999).

Alternative scenarios for the evolution of centralized nervous systems in protostomes and deuterostomes have been proposed where centralization occurred independently, after the split of the two taxonomic groups, without a dorsoventral inversion (reviewed in Gerhart, 2000; Holland, 2003; Lacalli, 2003). Recently, two hypotheses gained support from molecular genetic studies on hemichordates, a basal deuterostome phylum (Figure 1). In the auricularia hypothesis (Garstang, 1928), the evolutionary origins of the chordate nervous system may be found in the ciliary bands of a deuterostome dipleurula-like larval ancestor. Bilateral ciliary rows and the associated nerves moved dorsally, fused at the midline, and sank inside to form a dorsal cord. A number of genes that are involved in chordate CNS development, including SoxB3, Nkx2.1 and Otx are expressed in ciliary bands of larval hemichordates and/or echinoderms (Tagawa et al, 2001; Taguchi et al, 2002; Takacs et al, 2002). So far, however, the ciliary band derivatives have not been shown to give rise to cells of the adult nervous system after metamorphosis. Furthermore, the auricularia hypothesis does not take into account the molecular genetic similarities between the CNS of protostomes and chordates.

A recent comparative study on an enteropneust hemichordate has shown that the anteroposterior expression pattern of a large number of genes, which are involved in axial patterning of the vertebrate and arthropod CNS, is conserved in the diffuse nervous system of the enteropneust worm. The body-encircling basiepithelial nerve net of the directly developing hemichordate, Saccoglossus kowalevskii, expresses complex regulatory gene networks in a circumferential way (Lowe et al, 2003). Based on the findings in S. kowalevskii, Lowe et al proposed that the nervous system of the common ancestor of hemichordates, chordates and protostomes was organized in a diffuse, body-encircling, basiepithelial way. A diffuse nervous system is also found in the potentially relevant outgroups to the bilaterian animals, the cnidarians and ctenophores, and could therefore be an ancient condition of Bilateria. According to this view, independent centralization events in chordates and protostomes without dorsoventral inversion could have resulted in anteroposteriorly oriented CNSs with similar gene expression domains. The asymmetric expression along the dorsoventral axis of three genes (rx, nkx2.1, and hox4) described by Lowe et al indicates the presence of a dorsoventral patterning program. The antineural mechanism involving dpp/BMP4 that has been shown to limit the nervous systems of D. melanogaster and vertebrates to one side of the body is obviously not acting on the nervous system of the enteropneust hemichordate. Therefore, this antineural mechanism would have emerged independently in protostomes and chordates, assuming a noncommon origin of the CNS from a diffusely organized nervous system. This, however, would represent an example of parallel or even convergent evolution and thus not be parsimonious (Gould, 2002). Alternatively, an antineural mechanism along the dorsoventral body axis could have been present in the common ancestor of protostomes and deuterostomes, assuming that the diffuse nervous system of S. kowalevskii represents the secondary loss of a CNS and the antineural signaling system. Taken together, the conserved patterning mechanisms giving rise to a neurogenic region and an opposed antineural region along the dorsoventral axis in arthropods and vertebrates support a common origin of the CNS of protostomes and deuterostomes, including a dorsoventral inversion between the two animal groups.

The homeotic genes pattern the posterior brain in insects and vertebrates

The homeobox or Hox genes code for transcription factors with a characteristic helix–turn–helix DNA-binding motif called the homeodomain. Homeotic genes involved in specifying anteroposterior segment identity in the ectoderm were first discovered in D. melanogaster. Subsequently, similar clustered homeotic genes were found in a wide range of species where they have been shown to have an essential role in anteroposterior body axis patterning (Ferrier and Holland, 2001; Schilling and Knight, 2001; Carpenter, 2002; Hughes and Kaufman, 2002; Vervoort, 2002). In D. melanogaster, the Hox genes are arranged in two gene clusters known as the Antennapedia (ANT-C) and Bithorax (BX-C) complex. The ANT-C contains the five more anteriorly expressed Hox genes: labial (lab), proboscipedia (pb), Deformed (Dfd), Sex combs reduced (Scr) and Antennapedia (Antp). The BX-C contains the three posteriorly expressed genes: Ultrabithorax (Ubx), abdominal-A (abd-A), and Abdominal-B (Abd-B). Interestingly, the relative position of the genes within the clusters show a correlation with their spatial and temporal expression pattern in the body; genes located towards the 3′ end of the cluster are expressed more anteriorly and earlier in development than genes closer to the 5′ end. This correlation has been termed spatial and temporal colinearity (Mann, 1997). Furthermore, there appears to be a conserved functional hierarchy among the members of the homeotic gene clusters in that more posterior acting genes are functionally dominant over more anterior expressed genes, a fact that has been called ‘phenotypic suppression’ (Duboule and Morata, 1994). Mammalian Hox genes are aligned into 13 paralogous groups which are organized in four chromosomal clusters called Hox A – Hox D. The four clusters contain 9–11 Hox genes and only the Hox-B cluster comprises orthologues of all D. melanogaster homeotic complex genes. Similarly, as in D. melanogaster, the principle of spatial and temporal colinearity among the paralogous groups is also observed for vertebrate Hox genes, and more posterior acting genes impose their developmental specificities upon anterior acting genes what has been termed ‘posterior prevalence’ (Duboule and Morata, 1994; Mann, 1997).

Hox genes are expressed in the developing CNS of insects and vertebrates in a remarkably similar anteroposterior order (Figure 3a). In D. melanogaster genes of the Hox clusters are expressed in discrete domains in the developing brain and the ventral nerve cord and their anterior expression boundaries often coincide with neuromere compartment boundaries. In contrast to the embryonic epidermal structures of D. melanogaster, the anteroposterior arrangement of the homeotic genes in the fly CNS does not strictly fulfill the criterium of spatial colinearity (Kaufman et al, 1990; Hirth et al, 1998). The expression domains of the two 3′ most Hox genes of the ANT-C are inverted in that the anterior expression boundary of lab is posterior to that of pb. Interestingly, with respect to the relative spatial order of homeotic gene expression, the CNS of D. melanogaster is more similar to the CNS of the mouse than to the epidermis of the fly itself. In vertebrates, Hox genes are expressed in the hindbrain and spinal cord of the developing CNS. Expression precedes rhombomere formation and becomes progressively restricted to specific domains during embryogenesis. The most anterior Hox gene expression in the mouse brain is at the boundary between rhombomeres 2 and 3. This is followed posteriorly by a set of Hox gene expression domains, which generally coincide at their anteriormost domains with rhombomere boundaries. As in the D. melanogaster CNS, the mouse orthologues of the lab and pb genes, Hoxb-1 and Hoxb-2, show an inversion concerning the spatial colinearity rule of Hox cluster genes (Figure 3b). This inversion is more likely to have emerged in the CNS of a common ancestor of protostomes and deuterostomes, than independently after the divergence of the two groups.

Figure 3
figure 3

Simplified schematic comparison of Hox gene expression domains and mutant phenotypes in the CNS of D. melanogaster and mouse. (a) Expression domains of the homeotic genes of the Antennapedia and Bithorax complexes in the CNS of D. melanogaster: lab (labial), pb (proboscipedia), Dfd (Deformed), Scr (Sex combs reduced), Antp (Antennapedia), Ubx (Ultrabithorax), abdominal-A (abd-A) and Abdominal-B (Abd-B). In lab null mutant embryos (lab−/−), cells of the posterior part of the tritocerebrum (b3) are correctly located in the mutant domain, but fail to assume their correct neuronal cell fate (indicated by dashed lines). (b) Expression of the homeotic genes Hoxb-1, Hoxb-2, Hoxb-3, Hoxb-4, Hoxb-5, Hoxb-6, Hoxb-7, Hoxb-8 and Hoxb-9 in the embryonic CNS of mouse. Double mutant embryos of Hoxa-1 and Hoxb-1 (Hoxa-1−/− Hoxb-1−/−) result in a reduzed size of rhombomere 4 (4) and additionally a loss of expression of rhombomere 4-specific markers (indicated by dashed lines). The synergistic action of Hoxa-1 and Hoxb-1 in the specification of rhombomere 4 is comparable to the action of their single orthologue lab in the posterior tritocerebrum of D. melanogaster. Abbreviations: b1, protocerebrum; b2, deutocerebrum; b3, tritocerebrum; s1, mandibular neuromere; s2, maxillary neuromere; s3, labial neuromere; T, telencephalon; D, diencephalon; M, mesencephalon; 1–8, rhombomeres 1–8; wt, wild type (modified after Hirth and Reichert, 1999).

Full size image

In D. melanogaster, mutational inactivation of either of the Hox genes lab or Dfd results in severe axonal patterning defects in the embryonic brain (Hirth et al, 1998). In lab null mutants, axonal projection defects occur in the posterior tritocerebrum where lab is expressed in the wild-type brain. In the mutant, longitudinal pathways connecting supraesophageal and subesophageal ganglia as well as the projections in the tritocerebral commissure are absent or reduced. Interestingly, the brain defects are not due to a deletion in the tritocerebral neuromere; neuronal progenitors are present and give rise to progeny in the mutant domain. These postmitotic cells, however, do not form axonal and dendritic extensions and are not contacted by axons from other parts of the brain. The lab mutant cells do not acquire a neuronal fate, as revealed by the absence of neuronal markers, but rather remain in an undifferentiated state (Figure 3a). Comparable defects are seen in the D. melanogaster Dfd mutant in the corresponding mandibular domain, where the wild-type expression of the gene is located. Thus, the appropriate expression of the homeotic genes lab and Dfd is essential for the establishment of regionalized neuronal identity in the brain of D. melanogaster.

In mouse, the lab orthologues Hoxa-1 and Hoxb-1 are expressed in overlapping domains, with a sharp anterior boundary coinciding with the presumptive rhombomere 3/4 border. Single loss-of-function mutations of particular Hox paralogues show marked phenotypic differences suggesting synergy or functional compensation mechanisms (Maconochie et al, 1996; Rijli et al, 1998; Carpenter, 2002). Functional inactivation of Hoxa-1 causes segmentation aberrations leading to a reduced size of rhombomeres 4 and 5, defects of motor neuron axonal projections, and malformations of the trigeminal and facial/vestibuloacoustic nerve, but the normal identity of rhombomere 4 is not altered (Gavalas et al, 1998; Rijli et al, 1998; Studer et al, 1998). In contrast, loss of Hoxb-1 function has no effect on the size of rhombomere 4, but results in a loss of identity of the segment and a partial transformation into a rhombomere 2 identity (Goddard et al, 1996; Studer et al, 1996). The Hoxa-1, Hoxb-1 double loss-of-function mutant results in a territory of unknown identity and reduced size between rhombomeres 3 and 5, suggesting a synergistic action of the two genes in rhombomere 4 specification (Figure 3b) (Gavalas et al, 1998; Studer et al, 1998). Thus, Hoxa-1 and Hoxb-1 have very similar roles in the specification of neuronal identity during embryonic brain development as their orthologue lab in D. melanogaster. The similar functions in addition to similar expression domains of the homologous Hox genes in the developing hindbrains and posterior brains of fly and mouse support the idea of a common origin of the CNS.

The ems/Emx genes are involved in anterior brain development of D. melanogaster and mouse

The D. melanogaster ems gene belongs to the cephalic gap genes together with tailless (tll), orthodenticle (otd), buttonhead (btd), and sloppy paired (slp). At the early blastoderm stage of embryogenesis, the cephalic gap genes are broadly expressed in overlapping anterior stripes where their expression is initially regulated by maternal effect genes (Dalton et al, 1989; Finkelstein and Perrimon, 1990; Walldorf and Gehring, 1992). The functional inactivation of any of these genes results in gap-like phenotypes where structures of several head segments are missing (Cohen and Jürgens, 1991; Grossniklaus et al, 1994). The cephalic gap genes tll, otd, ems, and btd have been shown to be essential in early brain development. By the time of neuroblast delamination in the anterior brain, their expression domains become restricted to specific subsets of neural progenitors (Younossi-Hartenstein et al, 1997; Urbach and Technau, 2003). Mutational inactivation of a given cephalic gap gene results in the deletion of a specific brain area, indicating the requirement of these genes in early specification of the anterior brain primordium (Hirth et al, 1995; Younossi-Hartenstein et al, 1997).

The expression domain of the homeodomain transcription factor ems in the procephalic neuroectoderm and in the subsequently formed early embryonic brain of D. melanogaster comprises two stripes in the anterior parts of the deutocerebral (b2) und tritocerebral (b3) neuromeres (Figure 4a). A reiterated segmental expression pattern is also seen in the ventral nerve cord at later embryonic stages (not shown in Figure 4a). Loss-of-function of the ems gene results in a gap-like phenotype in the brain due to the absence of cells in the deutocerebral and anterior tritocerebral neuromeres (Hirth et al, 1995; Hartmann et al, 2000). In the ems mutant domain the expression of the proneural gene lethal of scute (l'sc) is lost and neuroblasts fail to form (Younossi-Hartenstein et al, 1997). This phenotype can be rescued by ubiquitous overexpression of ems, which results in proper brain development (Hartmann et al, 2000). Thus, ems function is required for the specification and formation of the anterior embryonic brain in D. melanogaster.

Figure 4
figure 4

Schematic representation of expression patterns and loss-of-function mutant phenotypes of ems in D. melanogaster and Emx2 and Pax6 in mouse. (a) In insects, the ems gene is expressed in the anterior part of the deutocerebrum and the anterior part of the tritocerebrum. Mutational inactivation of ems (ems−/−) results in the absence of the deutocerebrum and anterior part of the tritocerebrum. (b) In the developing mouse neocortex Emx2 is expressed in a gradient, with high caudomedial and low rostrolateral expression levels. In Emx2 null mutants (Emx2−/−), the anterior motor (black) and sensory (dark gray) cortical areas are expanded caudally, whereas the posterior visual cortical areas (white) are reduced in size. Pax6 is expressed in a gradient opposite to that of Emx2 expression in the developing neocortex. An opposite expansion of the cortical areas with respect to Emx2 mutants is observed in the Pax6 mutant Small eye (Sey/Sey), which indicates the interaction of Emx2 and Pax6 in regionalizing the neocortex (abbreviations, see Figure 3; modified after Hartmann et al (2000) and Bishop et al (2002)).

Full size image

The two mouse orthologues, Emx1 and Emx2, of the D. melanogaster cephalic gene ems, show largely overlapping expression domains in the developing brain. Whereas Emx1 mutant mice are postnatal viable and show rather subtle phenotypes restricted to the forebrain, Emx2 mutant mice die immediately after birth (Pellegrini et al, 1996; Qiu et al, 1996; Yoshida et al, 1997). Emx2 expression is seen in the germinative neuroepithelium of the presumptive cerebral cortex in the developing forebrain at around embryonic day 9.5. During early corticogenesis, Emx2 is restricted to the germinative layer in the ventricular zone, where it is expressed in proliferating neuroblasts. Subsequently, Emx2 expression is also found in Cajal-Retzius cells and the most marginal cortical plate neurons in the marginal zone (Gulisano et al, 1996; Pellegrini et al, 1996; Mallamaci et al, 2000). The anteriormost expression of Emx2 in the brain is found in the olfactory epithelium, whereas posteriorly the expression domain extends into the roof plate of the diencephalon. Emx2 is expressed throughout the developing neocortex in a graded manner with high levels at caudomedial and low levels at rostrolateral regions (Figure 4b). An opposed gradient is built up by the Pax6 gene that has been shown to interact with Emx2 in the regionalization of the neocortex. Mutational inactivation of Emx2 results in an expansion of the rostrolateral motor and somatosensory areas at the expense of the caudomedial neocortical areas, such as the visual area. An opposite shift in regional identity is seen in the Pax6 loss-of-function mutant, while in the Emx2 and Pax6 double mutant the cerebral cortex seems to acquire the identity of basal ganglia (Bishop et al, 2002; O'Leary and Nakagawa, 2002; Muzio et al, 2002). Interestingly, two orthologues of Pax6, eyeless (ey) and twin of eyeless (toy) are expressed in the anterior brain of the D. melanogaster embryo (Kammermeier et al, 2001). This raises the question whether they interact with ems in the regional specification of the embryonic fly brain.

The expressions of the D. melanogaster ems and the mouse Emx genes in the developing embryonic brain are similar, as is their ability to confer regional identity to the cells of a specific domain in the brain. Further evidence for the functional equivalence of the ems and Emx2 gene products comes from a cross-phylum rescue experiment carried out in D. melanogaster embryos. Ubiquitous overexpression of a mouse Emx2 transgene in an ems null mutant background rescues the brain phenotype of the mutant fly embryos (Hartmann et al, 2000).

Functional conservation of otd/Otx genes in embryonic brain development of D. melanogaster and mouse

The D. melanogaster cephalic gap gene otd encodes a transcription factor with a bicoid-like homeodomain and is required for head development and segmental patterning in the fly embryo. The first otd transcripts appear in the anterior region of the early blastoderm stage embryo, where they are expressed in a broad circumferential stripe. During gastrulation this anterior expression domain becomes more and more restricted to the procephalic neuroectoderm, and otd is expressed in most delaminating neuroblasts, of the presumptive protocerebrum and anterior deutocerebrum. This corresponds largely to the domain where otd is expressed at later embryonic stages in the brain (Hirth et al, 1995; Younossi-Hartenstein et al, 1997; Urbach and Technau, 2003). Interestingly, otd expression is not detected in the anteriormost part of the brain (Figure 5a). A second otd expressing domain is found at the ventral midline of the fly embryo in mesectodermal cells that will give rise to neurons and glia of the ventral nerve cord (not shown in Figure 5a). Mutational inactivation of otd results in a striking phenotype of the fly embryo in which the entire anterior part of the brain is lacking (Hirth et al, 1995). Mutant analysis has shown that most protocerebral and part of the adjacent deutocerebral neuroblasts are absent in the otd mutant, a fact that correlates with loss in the expression of the l'sc gene, which is thought to be required for neuroectodermal cells to adopt a neuroblast fate (Younossi-Hartenstein et al, 1997). In addition to the gap phenotype in the anterior brain, otd loss-of-function flies exhibit impairments in the development of visual structures as well as midline defects in the ventral nerve cord (Finkelstein et al, 1990). Ubiquitous overexpression of otd in a null mutant background at stages preceding neuroblast formation is able to restore anterior brain structures and ventral nerve cord defects. In a wild-type background, ubiquitous overexpression of otd results in the generation of ectopic neuronal structures, such as enlarged ganglia. Interestingly, some of the ectopic cells express the protocerebrum-specific gene brain-specific homeobox (bsh) indicating that otd expression might induce a partial protocerebral identity in these neuronal structures (Leuzinger et al, 1998).

Figure 5
figure 5

Summary scheme of expression domains, null mutant phenotypes and cross-phylum rescue experiments of the otd/Otx2 genes in D. melanogaster and mouse. Genotypic indications on the left of the corresponding rows are indicated in (a) and (b). The photographs show frontal views of D. melanogaster embryonic brains (anti-HRP immunostaining; scale bar: 10 μm) in (a) and lateral views of whole mount mouse embryos (embryonic day 10.5; scale bar: 250 μm) in (b). The column on the right-hand side shows schematic representations of expression domains and phenotypes in the brain of the corresponding animal and genotype. (a) In the fly the otd gene is expressed throughout most of the protocerebrum and the anterior part of the deutocerebrum. In the frontal view of the embryonic D. melanogaster wild-type brain, the preoral commissure interconnecting the two anterior brain hemispheres is indicated by an arrow (the frontal connective is marked with an asterisk). In otd mutant embryos (otd−/−), the protocerebrum including the preoral commissure and the anterior deutocerebrum are absent (indicated by triangles in the picture and by dashed lines in the scheme). Overexpression of human Otx2 gene in otd mutant embryos (otd−/−; hs-Otx2) results in a rescue of the anterior brain including the preoral commissure. (b) In mouse the Otx2 gene is expressed in the anterior part of the embryonic brain including the presumptive telencephalon (except for anteriormost part), diencephalon, and mesencephalon. In the lateral view of the mouse embryo the major brain regions are labelled as forebrain (fb), midbrain (mb), and hindbrain (hb). In Otx2 null mutants in which the D. melanogaster otd replaces the Otx2 gene (Otx2−/−; otd2/otd2) the entire forebrain and midbrain (as well as rhombomeres 1 and 2) are absent. In Otx2 mutants, in which the otd coding sequence has been fused to the 3′ and 5′ UTRs of Otx2 gene (Otx2−/−; otd2FL/otd2FL), the hybrid transcript is translated in the anterior neuroectoderm of the mouse embryo and the rostral forebrain is restored (modified after Leuzinger et al, 1998; Acampora et al, 2001).

Full size image

The two mouse orthologues, Otx1 and Otx2, of the otd gene are expressed in nested domains of the developing brain. Otx1 expression is first observed at approximately 8 days post coitum (dpc) in the neuroepithelium of the presumptive telencephalon, diencephalon, and mesencephalon (Simeone et al, 1992). During corticogenesis, Otx1 expression is maintained in the ventricular zone of the cortical anlage, but decreases as upper layer neurons are generated. By this time, postmigratory neurons of layers 5 and 6 progressively start to express Otx1, whereas later differentiated neurons of upper layers 1–4 remain devoid of Otx1 expression (Frantz et al, 1994). Otx1 is also expressed at early embryonic stages in precursor structures of sense organs, such as the olfactory epithelium and the inner ear (Simeone et al, 1993). Otx1 null mice are viable but suffer from spontaneous epileptic seizures and exhibit a smaller brain size, mainly due to a reduced thickness of the telencephalic cortex. In addition, the development of the vestibuloacoustic sense organs is impaired, as the lateral semicircular duct of the inner ear is lost (Acampora et al, 1996).

The earliest expression of Otx2 is found in the epiblast and in the visceral endoderm (VE) prior to the onset of gastrulation. During gastrulation, Otx2 expression is observed in the epiblast and anterior neuroectoderm as well as in the underlaying anterior visceral endoderm (AVE) and the node-derived axial mesendoderm (ame). The AVE and ame are believed to generate Otx2-mediated instructive signals that are required in the early specification and patterning of the overlaying anterior neuroectoderm (reviewed in Simeone, 1998; Acampora and Simeone, 1999). Otx2 expression in the anterior neuroectoderm is maintained during brain regionalization and extends from the telencephalon to the posterior border of the mesencephalon, anterior of the midbrain-hindbrain boundary (MHB) (Figure 5b). Interestingly, the domain of Otx2 expression does not include the most anterior brain region, which is similar to the expression pattern of otd in the embryonic fly brain (Simeone et al, 1992).

Otx2 null mice die early in embryogenesis and lack the rostral neuroectoderm fated to become forebrain, midbrain, and rostral hindbrain as a result of an impairment in early specification of the anterior neuroectoderm by the VE (Acampora et al, 1995; Rhinn et al, 1998). This has been demonstrated in chimeric mouse embryos containing Otx2−/− epiblast and wild-type VE, where the early induction of the anterior neural plate was transiently rescued, but subsequent brain development remained impaired. No rescue was obtained in chimeras containing a wild-type epiblast and a Otx2−/− VE (Rhinn et al, 1998).

Cross-phylum rescue experiments between fly otd and mammalian Otx1 and Otx2 genes were carried out in order to assess the functional equivalence or diverged properties of the gene homologues. Ubiquitous overexpression of either human Otx1 or human Otx2 in an otd mutant fly in both cases restored the anterior brain structures absent in the otd null mutant (Figure 5a) (Leuzinger et al, 1998).

Similar cross-phylum experiments were carried out in mouse with otd replacing the vertebrate Otx orthologues. In an Otx1 null mutant background, otd is able to fully rescue epilepsy and corticogenesis abnormalities restoring wild-type brain size. However, the lateral semicircular duct of the inner ear is never restored (Acampora et al, 1998a). A similar rescue potential is also observed in homozygous mutant mouse embryos, where Otx1 was replaced with human Otx2 (Acampora et al, 1999; Morsli et al, 1999). Thus, the ability to specify the lateral semicircular duct of the inner ear may be an Otx1-specific property (Acampora and Simeone, 1999). Gene replacement experiments where different portions of the Otx2 locus were exchanged with the cDNA of the fly otd or human Otx1 genes revealed a crucial role of regulatory control mechanisms in Otx2-specific action during anterior neuroectoderm specification. Two different replacement strategies were utilized. A first mouse model (otd2/otd2) was generated in which an Otx2 region including 5′ and 3′ untranslated regions (UTRs) was replaced with the fly otd cDNA, whereas in a second mutant (otd2FL/otd2FL) the otd coding sequence was directly fused to the intact 5′ and 3′ UTRs of Otx2. In the otd2/otd2 mouse model, otd is able to take over the early function of the Otx2 gene in the AVE, leading to a transient restoration of the anterior neural plate absent in Otx2 mutants. However, otd2/otd2 mutants fail to maintain the anterior identities of the neuroectoderm, giving rise to a headless phenotype (Figure 5b). Mutant analysis revealed that D. melanogaster otd transcripts were present in both AVE and presumptive anterior neuroectoderm, whereas translation only occurred in the AVE. Additional evidence from similar experiments where Otx2 including UTRs was replaced with human Otx1 favored the view that post-transcriptional control was involved in the cell type-specific translation of Otx2 mRNA in the epiblast and anterior neuroectoderm (Acampora et al, 1998b; Boyl et al, 2001). This was confirmed in the second mouse model otd2FL/otd2FL, where translation of the hybrid transcript consisting of the fly otd fused to the 5′ and 3′ UTRs of Otx2 occurred in the epiblast and anterior neuroectoderm. Moreover, the correct translation of otd in the epiblast and anterior neuroectoderm restored the maintenance of anterior brain pattening in Otx2 null mutants including the normal positioning of the MHB (Figure 5b) (Acampora et al, 2001). This was also shown by a similar hybrid mouse model where human Otx1 was fused to the 5′ and 3′ UTRs in the mouse Otx2 locus (Acampora et al, 2003). Taken together, Otx1, Otx2, and otd genes show a high degree of functional equivalence in the regions of the developing organism where they are normally expressed. This supports the idea that otd/Otx functions were originally established in a common ancestor of fly and mouse and conserved throughout evolution. On the other hand, their regulatory control mechanisms appear to have been modified during evolution, thus, generating the specific properties of the genes.

Evidence for a common tripartite ground-plan of the bilaterian brain

A detailed comparison of gene expression patterns and developmental neuroanatomy in vertebrates and urochordates (ascidians) has uncovered a common tripartite ground-plan along the anteroposterior axis for the embryonic CNS. In all cases studied, a rostral brain region expressing Otx family genes (corresponding to the vertebrate forebrain and midbrain) is followed by a central region expressing Pax2/5/8 genes (delimiting the MHB of vertebrates), and subsequently a Hox gene expressing caudal region (hindbrain and spinal cord of vertebrates) (Wada et al, 1998; Holland and Holland, 1999; Wada and Satoh, 2001). Recently, a similar tripartite pattern of gene expression has been reported for arthropods (see below) and hemichordates, suggesting an evolutionarily more ancient origin of the tripartite organization of brains than chordates (Figure 6a–e) (Hirth et al, 2003; Lowe et al, 2003; CJ Lowe, personal communication). (Interestingly, no Pax2/5/8 expression can be detected between the anterior Otx and posterior Hox expression domains in the neural tube of the cephalochordate Amphioxus; the most parsimonious explanation for this is the secondary loss of the tripartite pattern in the Amphioxus CNS (Kozmik et al, 1999; Takahashi and Holland, 2004).)

Figure 6
figure 6

Tripartite ground-plan of the bilaterian nervous system based on expression patterns of orthologous genes in Drosophila, mouse, Amphioxus, ascidian, and hemichordate. The expression of otd/Otx2, unpg/Gbx2, Pax2/5/8, and Hox1 gene orthlogues in the developing nervous systems of (a) stage 13/14 D. melanogaster embryo (Hirth et al, 2003, b) embryonic day 10 mouse embryo (Wurst and Bally-Cuif, 2001, c) 10 somite stage Amphioxus embryo (Wada and Satoh, 2001, d) neurula ascidian (Wada et al, 1998) and (e) 1 gill slit stage hemichordate embryo (Lowe et al, 2003). In all cases an otd/Otx2-expressing region is located anterior to a Hox-expressing region in the posterior nervous system. In D. melanogaster and mouse, a Pax2/5/8-expressing domain is positioned at the interface between the anterior otd/Otx2 and the posteriorly abutting unpg/Gbx2 expression domains. In D. melanogaster, the Pax2/5/8 orthologues Pax2 and Poxn also show a segmentally reiterated expression pattern (see text for details). Up to now, no unpg/Gbx2 orthologues have been isolated in Amphioxus and Ascidians. The expression domains of the hemichordate otd/Otx2 and unpg/Gbx2 orthologues show no sharp boundary, but overlap in an intermediate region of the basiepithelial nerve net. Nevertheless, the expression of the hemichordate Pax2/5/8 orthologue is consistent with its relative location in chordates (CJ Lowe, personal communication). No Pax2/5/8 expression can be found between the otd/Otx2 domain and the Hox1 domain in Amphioxus, which is thought to be due to a secondary reduction.

Full size image

In vertebrate brain development, the Pax2/5/8 domain at the MHB is an early marker for the isthmic organizer (IsO), which controls both the growth and the ordered rostrocaudal specification of mesencephalic and metencephalic territories (reviewed by Liu and Joyner, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). The IsO was first identified through transplantation experiments, in which MHB tissue grafts were transplanted to more rostral or caudal neural locations, resulting in the induction of mesencephalic–metencephalic fate in the host tissue surrounding the graft (Martinez et al, 1991; Marin and Puelles, 1994). This organizer-like activity on the surrounding neural tissue is thought to be mediated by fibroblast growth factor 8 (Fgf8) and Wnt1 proteins which are secreted from the MHB tissue. During late gastrulation and early neural plate stages of the vertebrate embryo, the two homeodomain transcription factors Otx2 and Gbx2 are expressed in a complementary, mutually exclusive fashion anterior and posterior to the MHB. Whereas Otx2 null mutant mice lack the brain rostral to rhombomere 3 (see above), mice of the genotype Otx1−/− Otx2+/− show a rostral extension of metencephalic tissue and the absence of the mesencephalon and caudal diencephalon. Furthermore, the expressions of MHB-specific markers, such as Fgf8, Gbx2, and Wnt1, align in a domain that is shifted rostrally to the corresponding position of prosomere 2 (Acampora et al, 1997). Conversely, a caudal shift of MHB markers can be observed in Gbx2 null mutants, where isthmic nuclei, cerebellum, and rhombomeres 1–3 of the hindbrain are absent (Figure 7a) (Wassarman et al, 1997; Millet et al, 1999). Together with evidence from misexpression experiments, these results suggest that an antagonistic interaction between Gbx2 and Otx2 during early embryonic stages is responsible for the correct positioning of the MHB at their common interface.

Figure 7
figure 7

Antagonistic interactions of the otd/Otx and unpg/Gbx2 genes in the positioning of their common interface. (a) Expression domains of Otx2 and Gbx2 in the developing mouse CNS corresponding to the six-somite stage in the Gbx2 null mutant (Gbx2−/−), wild type (wt), and Otx1−/− Otx2+/− (Otx1−/− Otx2+/−) genetic background. In the wild-type mouse embryo, Otx2 is expressed with a sharp limit at the posterior mesencephalon and Gbx2 expression abuts the Otx2 expression domain, creating a common interface. In mice homozygous mutant for Otx1 and heterozygous mutant for Otx2 (Otx1−/− Otx2+/−), the common interface is shifted anteriorly into the forebrain (dark gray arrow). A posterior expansion of the Otx2 expression into the hindbrain is observed in Gbx2 null mutant (Gbx2−/−) brains. (b) Expression domains of otd and unpg in the developing CNS of D. melanogaster in the unpg null mutant (unpg−/−), wild type (wt), and otd null mutant (otd−/−) genetic background. The expression domains of otd and unpg in the wild-type D. melanogaster CNS form a sharp common boundary in the posterior deutocerebrum. In the otd null mutant embryo (otd−/−), the protocerebrum and the anterior deutocerebrum are absent (dashed lines). In addition, the unpg expression is shifted anteriorly (dark gray arrow). In the brain of the unpg null mutant embryo (unpg−/−), the otd-expressing domain expands posteriorly (light gray arrow). Abbreviations: P, protocerebrum; D, deutocerebrum; T, tritocerebrum; VNC, ventral nerve cord; F, forebrain; M, midbrain; H, hindbrain; SC, spinal cord (modified after Hirth et al, 2003; Joyner et al, 2000).

Full size image

Gene expression studies indicate that embryonic anteroposterior patterning of the D. melanogaster brain is strikingly similar to the tripartite ground-plan of the vertebrate brain. Expression of both D. melanogaster Pax2/5/8 orthologues, Pox neuro (Poxn) and Pax2, is present at the interface of otd and the Gbx2 orthologue unplugged (unpg), anterior to a Hox-expressing region (Noll, 1993; Fu and Noll, 1997; Hirth et al, 2003). The expression domains of Poxn and Pax2 span the whole embryonic CNS in a segmentally reiterated pattern, but the genes are never coexpressed in the same cells. Interestingly, the only anteroposterior position along the neuraxis where Pax2 and Poxn are expressed in adjacent domains is located immediately anterior to the deutocerebral–tritocerebral boundary (DTB). In addition, this transversal domain of adjacent Pax2 and Poxn expression differs from the segmentally reiterated expression in more posterior regions by the fact that it is the only position along the neuraxis where expression of both genes coincides with a neuromere boundary (Figure 6a) (Hirth et al, 2003). Mutational inactivation of otd results in the deletion of the anterior brain of the fly embryo (see above) as well as in the rostral extension of the unpg expression domain. In unpg loss-of-function mutants, the posterior limit of the anterior brain-specific otd expression shifts caudally (Figure 7b). Thus, in both D. melanogaster and mouse, mutational inactivation of otd/Otx2 and unpg/Gbx2 genes results in the loss or misplacement of an intermediate brain domain characterized by the otd/Otx2 and unpg/Gbx2 interface and by the expression of Pax2/5/8 genes. Moreover, otd/Otx2 and unpg/Gbx2 appear to negatively regulate each other at the interface of their expression domains in insects and vertebrates. (Interestingly, D. melanogaster otd is able to replace Otx gene function in the correct positioning of the MHB during mouse brain development as demonstrated in cross-phylum rescue experiments (see above) (Acampora et al, 2001).) Taken together, these results reveal remarkable similarities in gene expression and functional interactions involved in establishing the insect DTB and mouse MHB. However, not all functional interactions among genes involved in MHB formation in the mouse appear to be conserved at the DTB of D. melanogaster. Although expression of patterning genes that characterize the vertebrate MHB region, such as engrailed (en), Pax2, Poxn or the fly Fgf orthologue branchless (bnl) can be found at the DTB, no brain-patterning defects are observed in the corresponding null mutant embryos in the fly (Hirth et al, 2003). Moreover, even though D. melanogaster has a tripartite ground-plan for the developing brain and a boundary region genetically corresponding to the vertebrate MHB, evidence for organizer activity of the fly DTB has not been obtained.

In summary, current comparative data suggest that a tripartite ground-plan for the developing brain was already present in the common ancestor of bilateria. To date, organizer activity of the intermediate boundary region has only been demonstrated in vertebrates (Takahashi and Holland, 2004). As proposed by Wada and Satoh (2001), it may be useful to distinguish between the homology of two characteristics of the vertebrate MHB: homology as a developmental genetic region of the brain and homology as an organizer. In this sense, the D. melanogaster DTB can be considered as a region homologous to the vertebrate MHB.

Conclusions

Recent investigations on mechanisms controlling insect and vertebrate brain development have revealed an expanding number of homologous genes with similar expression patterns and comparable functions. The expression and interactions of homologous dorsoventral patterning genes show comparable relative patterning and orientation with respect to the presumptive neurogenic region. Genes of the otd/Otx and ems/Emx families are required for correct formation and specification of the developing anterior brain, and Hox genes are involved in patterning and specification of the developing posterior brain. Moreover, otd/Otx genes and unpg/Gbx2 genes position an intermediate domain between an anterior and a posterior brain region and thus contribute to the tripartite ground-plan of the insect and vertebrate brain.

Taken together, these results imply the evolutionary conservation of genetic programs underlying embryonic brain development in insects and vertebrates. Moreover, the similarities among orthologue groups are not restricted to the positions of the expression domains, but also include functional features. This supports the idea that the protostome and deuterostome brain is homologous in a developmental genetic sense, and thus the common urbilaterian ancestor already had a complex CNS (Arendt and Nübler-Jung, 1996; De Robertis and Sasai, 1996; Hirth and Reichert, 1999). The identification of downstream targets of conserved developmental control genes as well as the analysis of genetic mechanisms at more advanced stages of development should give a deeper insight into the degree of conservation of genetic programs between insects and vertebrates. Specific gene functions that are not shared between orthologous control genes in fly and mouse, such as the post-transcriptional control of the Otx2 gene in the mouse epiblast, appear primarily to involve modifications of regulatory control elements, but not the coding sequence of the gene. This suggests that genes involved in essential mechanisms of brain development could exert additional, novel functions by modification of their spatial or temporal regulatory control (Acampora and Simeone, 1999; Acampora et al, 2001).

A recent gene expression study on hemichordates has led to the view that the deuterostome ancestor might have been characterized by a diffuse basiepithelial nervous system and that a centralized brain could have evolved independently in the deuterostome and protostome lineages (Holland, 2003; Lacalli, 2003; Lowe et al, 2003). Homologies in embryonic brain development of vertebrates and insects would therefore derive from axis-patterning mechanisms or the correlating gene expression patterns, which were present in the circumferential nerve net of the last common ancestor. Other similarities that have not been inherited from a common ancestor characterized by a well-patterned nerve net would therefore be a product of convergent or parallel evolution (Gould, 2002). According to this view, the similar antineural function of dpp/BMP4 in insects and vertebrates represents an example of parallel evolution, assuming that the last common ancestor of protostomes and deuterostomes had a diffuse, body-encircling basiepithelial nervous system. An alternative explanation for the absence of a centralized nervous system in hemichordates is a secondary loss of a CNS together with the antineural activity of the dorsoventral signaling program and the retention of a peripherally located nerve net (Holland, 2003). The expression patterns of the hemichordate orthologues of dorsoventral patterning genes should nurture the discussion on the urbilaterian nervous system.