An urbilaterian origin of the tripartite brain: developmental genetic insights from Drosophila

Classical phylogenetic, neuroanatomical and embryological studies have led to the view of an independent evolutionary origin of protostome and deuterostome brains. Accordingly, bilaterians have been divided into two major groups with different CNS morphologies: the gastroneuralia, which are characterized by a ventral nerve cord, include protostomes such as arthropods,annelids and molluscs; and the notoneuralia, which are characterized by a dorsal nerve cord, include all chordates(Siewing, 1985; Brusca and Brusca, 1990; Nielsen, 2001). By contrast,more recent studies examining the expression patterns and functions of orthologous genes in embryonic nervous systems have revealed similarities in the embryonic brains of protostomes, such as annelids and arthropods, and deuterostomes like tunicates and mammals. These results suggest that lophotrochozoan and ecdysozoan invertebrate brains and chordate brains of vertebrates, acraniates and urochordates are all characterized by a rostral region specified by genes of the otd (oc –FlyBase)/Otx family and a caudal region specified by genes of the Hox family(reviewed by Bruce and Shankland,1998; Arendt and Nübler-Jung, 1999; Hirth and Reichert, 1999; Holland and Holland, 1999; Reichert and Simeone, 1999; Wada and Satoh, 2001; Reichert,2002).

In ascidian and vertebrate chordates, a Pax2/5/8 expression domain is located between the anterior Otx and the posterior Hox expression regions of the embryonic brain (reviewed by Holland and Holland, 1999; Wada and Satoh, 2001). In vertebrate brain development, this Pax2/5/8 expression domain is positioned at the interface of the brain-specific Otx2 and Gbx2 expression domains; it is an early marker for the isthmic organizer at the midbrain-hindbrain boundary (MHB), which controls the development of the midbrain and the anterior hindbrain (reviewed by Liu and Joyner, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). The central role of this MHB region in brain development together with the conserved expression patterns of Pax2/5/8 genes in this region have led to the proposal that a fundamental characteristic of the ancestral chordate brain was its tripartite organization characterized by Otx, Pax2/5/8 and Hox gene expressing regions(Wada et al., 1998). To investigate whether protostomes also possess a tripartite organization of the brain, and to gain insight into the evolution of the bilaterian brain, we carried out a comparative analysis of expression and function in the ecdysozoan Drosophila melanogaster of the orthologues that pattern the vertebrate MHB region.

Our study reveals striking similarities in the expression and function of genes that pattern the embryonic brains of Drosophila and vertebrates. We find that a Pax2/5/8 expressing domain is located between an anterior otd-expressing region and a posterior Hox-expressing region in the embryonic brain of Drosophila. Moreover,in Drosophila, as in vertebrates, we find that this Pax2/5/8expressing domain is positioned at the interface between the otd/Otx2expression domain and a posteriorly abutting unplugged/Gbx2expression domain. Finally, we demonstrate that inactivation of otd/Otx2 or of unplugged/Gbx2 results in comparable effects on mispositioning or loss of brain-specific expression domains of orthologous genes in both embryonic brain types. These developmental genetic similarities indicate that the tripartite ground plan, which characterizes the developing chordate brain, is also at the basis of the developing insect brain,conferring on all bilaterians a deep similarity in brain development. This suggests that a corresponding tripartite organization already existed in the brain of the last common urbilaterian ancestor of insects and chordates.

Wild-type flies are of the strain Oregon-R. All mutant stocks are from the Bloomington Stock Center unless indicated otherwise. The following reporter lacZ lines and mutant alleles were used: P{lacZ}unpg^f85 (an unpg-lacZ reporter gene that expresses cytoplasmic β-galactosidase in the same pattern as endogenous unpg) (Chiang et al.,1995), P{lacZ}Pax2^Δ122 (a Pax2-lacZ reporter gene that expresses β-galactosidase in the same pattern as endogenous Pax2; E.F., M. Daube, J. Kronhamn, A. Rasmuson-Lestander and M.N., unpublished), otd^JA101(Hirth et al., 1995), P{3′lacZ}unpg^r37 (unpg null allele with a unpg-lacZ reporter gene that expresses nuclearβ-galactosidase in the same pattern as endogenous unpg)(Chiang et al., 1995).

For in situ hybridization experiments, digoxigenin-labelled sense and antisense RNA probes of the spa/Pax2 cDNA clone cpx1(Fu and Noll, 1997) were generated in vitro with a DIG-labelling kit (Roche diagnostics) and hybridized to Drosophila whole-mount embryos by following standard procedures(Tautz and Pfeifle, 1989). Whole-mount immunocytochemistry was performed as previously described(Hirth et al., 1998). Rabbit anti-PAX2 antiserum (Fu and Noll,1997) and monoclonal anti-POXN antibodies(Hassanzadeh et al., 1998)were used as primary antibodies at 1:50 and 1:20 dilutions, respectively. To generate an anti-Otd antiserum, a 835 bp fragment from an otd cDNA(position 1708-2542) (Finkelstein et al.,1990) was amplified by PCR and cloned into the pCR2.1-TOPO vector(Invitrogen), from which it was removed by digestion at flanking EcoRI sites and cloned in frame into the pGEX-2T expression vector(Smith and Johnson, 1988). The glutathione-S-transferase/Otd fusion protein was purified according to Smith and Johnson (Smith and Johnson,1988), except that induction occurred at 18°C overnight. Immunization of rabbits was carried out by Pab Productions (Hebertshausen). For whole-mount immunocytochemistry, anti-Otd antiserum was used as primary antiserum at a 1:100 dilution.

Embryos were staged according to Campos-Ortega and Hartenstein(Campos-Ortega and Hartenstein,1997).

For laser confocal microscopy, a Leica TCS SP microscope was used. Optical sections ranged from 0.4 to 2 μm, recorded in line average mode with picture size of 512×512 pixels. Captured images from optical sections were arranged and processed by the use of IMARIS (Bitplane). Figures were arranged and labelled by the use of Adobe Photoshop.

In the embryonic CNS of vertebrates, the Pax2, Pax5 and Pax8 genes are expressed in specific domains that overlap in the presumptive MHB region (reviewed by Liu and Joyner, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). Drosophila has two Pax2/5/8orthologues, Pox neuro (Poxn) and Pax2(Noll, 1993; Fu and Noll, 1997).

The embryonic brain of Drosophila can be subdivided into the protocerebrum (PC or b1), deutocerebrum (DC or b2) and tritocerebrum (TC or b3) of the supra-oesophageal ganglion and the mandibular (S1), maxillary (S2)and labial (S3) neuromeres of the sub-oesophageal ganglion. Expression of engrailed (en) delimits these subdivisions by marking their most posterior neurones (reviewed by Hartmann and Reichert, 1998)(Fig. 1). Because of morphogenetic processes, such as the beginning of head involution, the neuraxis of the embryonic brain curves dorsoposteriorly within the embryo. Accordingly, henceforth anteroposterior coordinates refer to the neuraxis rather than the embryonic body axis.

We first determined whether Pax2 is expressed in specific domains of the Drosophila brain, by analysing its expression pattern using in situ hybridization, immunolabelling and lacZ reporter gene expression. Pax2 transcripts initially appear during gastrulation and at stage 9/10 are observed in a segmentally reiterated pattern of the developing procephalic and ventral neuroectoderm, with its anteriormost expression domain located at the future deutocerebral-tritocerebral boundary(Fig. 2A, arrow). Expression of Pax2 transcripts in the developing brain begins at stage 10/11 and is most prominent in a longitudinal stripe at the medial part of the protocerebrum and in a transversal stripe at the posterior border of the deutocerebrum (Fig. 2B-D,arrowheads). Immunolabelling with a Pax2-specific polyclonal antibody(Fu and Noll, 1997) revealed that Pax2 protein distribution resembles that of Pax2 transcripts, as does a Pax2-lacZ reporter gene expressing β-galactosidase(Fig. 3A-C). In addition to its expression in the developing anterior brain, Pax2 expression is also seen in six to eight cells located at the lateral margin of each hemisegment throughout the more posterior CNS regions of the sub-oesophageal ganglion and ventral nerve cord (Fig. 3C).

To determine the expression of the second Drosophila Pax2/5/8orthologue, we characterized Poxn expression using immunolabelling and lacZ reporter genes. POXN protein is first detected in the developing brain at the end of germband extension (stage 10/11) in two stripes of the procephalic neuroectoderm, which subsequently become restricted to the posterior protocerebrum and the posterior deutocerebrum(Fig. 3D,E). Poxnexpression in more posterior regions of the CNS also occurs in segmentally reiterated patterns (Fig. 3F)(Bopp et al., 1989). A comparison between Pax2 and Poxn expression domains reveals,that Pax2 and Poxn are never co-expressed in the same cells of the CNS. Moreover, expression of Pax2 and Poxn does not occur at a comparable anteroposterior position along the neuraxis(Fig. 3I), with one exception. This exception is in the posterior deutocerebrum where adjacent Pax2and Poxn expression domains define a transversal domain immediately anterior to the tritocerebral brain neuromere (arrows in Fig. 3G,H). This transversal domain of adjacent Pax2 and Poxn expression is distinguishable from 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 (compare Fig. 3H with Fig. 1D). We refer to this transversal domain of adjacent Pax2/5/8 orthologue expression as the deutocerebral-tritocerebral boundary (DTB) region.

It is important to note that the DTB is located anterior to the expression domain of the Drosophila Hox1 orthologue labial(lab), which is expressed in the posterior tritocerebrum(Hirth et al., 1998). Moreover, the DTB is located posterior to the expression domain of the Drosophila Otx orthologue otd in the protocerebrum and anterior deutocerebrum (Hirth et al.,1995). Thus, in Drosophila as in vertebrates, a Pax2/Poxn (Pax2/5/8) expression domain is located between the anterior otd/Otx2 and the posterior Hox-expressing regions. This raises the question of whether the DTB in the embryonic Drosophilabrain might have developmental genetic features similar to those observed for the MHB in vertebrate brain development.

In the embryonic vertebrate brain, Otx2 is expressed anterior to and abutting Gbx2. The future MHB as well as the overlapping domains of Pax2, Pax5 and Pax8 expression are positioned at this Otx2-Gbx2 interface (Liu and Joyner, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). To investigate if comparable expression patterns are found in the embryonic fly brain, we determined the brain-specific expression of the Drosophila Gbx2 orthologue unplugged (unpg) in relation to that of otd, using immunolabelling and an unpg-lacZ reporter gene that expressesβ-galactosidase like endogenous unpg(Chiang et al., 1995). The otd gene is expressed in the protocerebrum and anterior deutocerebrum of the embryonic brain (Fig. 4A,B) (Hirth et al.,1995), as well as in midline cells in more posterior regions of the CNS (Fig. 4C)(Finkelstein et al., 1990; Wieschaus et al., 1992). Expression of unpg-lacZ in the embryonic CNS is first detected at stage 8 in neuroectodermal and mesectodermal cells at the ventral midline, with an anterior limit of expression at the cephalic furrow(Chiang et al., 1995). Subsequently, the unpg expression domains in the CNS widen and have their most anterior border in the posterior deutocerebrum(Fig. 4D-F). Double immunolabelling of OTD and β-galactosidase revealed that the posterior border of the brain-specific otd expression domain coincides with the anteriormost border of the unpg expression domains along the anteroposterior neuraxis (Fig. 4G,H, arrows). There is no overlap of otd and unpg expression in the brain or in more posterior regions of the CNS(Fig. 4G-I).

These findings indicate that the otd-unpg interface is positioned at the anterior border of the DTB. This was confirmed by additional immunolabelling studies examining unpg-lacZ, otd, Poxn and en expression in the protocerebral/deutocerebral region of the embryonic brain. Thus, double immunolabelling of OTD and EN confirms that the posterior border of otd expression extends beyond the protocerebral en-b1 stripe into the anterior deutocerebral domain(Fig. 5A). Labelling OTD and POXN confirms that the Poxn expression domain of the DTB is posterior to this deutocerebral otd expression boundary(Fig. 5B). Labelling EN andβ-galactosidase, which is indicative of unpg expression,confirms that the anteriormost unpg expression domain overlaps with the en-b2 stripe (Fig. 5C). Finally, labelling β-galactosidase and POXN confirms that this anteriormost unpg expression domain overlaps with the Poxnexpression domain of the DTB (Fig. 5D). Therefore, in terms of overall gene expression patterns, we find that a transversal domain of adjacent Pax2/Poxn expression defines the DTB region of the embryonic Drosophila brain. Furthermore, this region is located between an anterior otdexpression domain and a posterior Hox expression domain. Moreover, it is located abutting and posterior to the interface of otd and unpg expression along the anteroposterior neuraxis. A schematic summary of these expression patterns in the embryonic fly brain is illustrated in Fig. 8A.

In mammalian brain development, homozygous Otx2-null mutant embryos lack the rostral brain, including the MHB-specific Pax2/5/8expression domain (Acampora et al.,1995; Matsuo et al.,1995; Ang et al.,1996), whereas Gbx2 null mutants misexpress Otx2and Hoxb1 in the brain (Wassarman et al., 1997). Moreover, Otx2 and Gbx2negatively regulate each other at the interface of their expression domains(Li and Joyner, 2001; Liu and Joyner, 2001; Martinez-Barbera et al., 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001; Ye et al., 2001). To test if similar regulatory interactions occur in the embryonic brain of Drosophila, we analysed the expression of the corresponding orthologues in otd and unpg mutant embryos.

In otd-null mutant embryos, the protocerebrum is absent because protocerebral neuroblasts are not specified(Hirth et al., 1995; Younossi-Hartenstein et al.,1997). Analysis of unpg, en and Poxn expression in otd-null mutant embryos revealed that the anteriormost border of unpg expression shifts anteriorly into the anterior deutocerebrum(Fig. 6A,B), while Poxn fails to be expressed in the deutocerebrum(Fig. 6C). In contrast to inactivation of otd, inactivation of unpg does not result in a loss of cells in the mutant domain of the embryonic brain, as is evident from the expression of an unpg-lacZ reporter construct in unpg-null mutant embryos (Chiang et al., 1995). Analysis of otd expression in unpg-null mutants shows that the posterior limit of brain-specific otd expression shifts posteriorly into the posterior deutocerebrum,thus extending into the DTB (Fig. 7A,B). This was confirmed by additional immunolabelling studies examining otd, Poxn and en expression in the protocerebral/deutocerebral region of the embryonic brain in unpg-null mutants. Double immunolabelling of OTD and EN in unpg-null mutants confirms that the posterior border of brain-specific otd expression extends posteriorly to the deutocerebral en-b2 stripe into the posterior deutocerebrum(Fig. 7C, compare with Fig. 5A). In addition, double immunolabelling of OTD and POXN in unpg-null mutants confirms that the posterior border of brain-specific otd expression extends posteriorly into the Poxn expression domain of the DTB(Fig. 7D, compare with Fig. 5B). Moreover, analysis of lab expression in unpg-null mutants shows that brain-specific lab expression shifts anteriorly into the anterior tritocerebrum (Fig. 7E,F). Thus, in both Drosophila and mammals, mutational inactivation of otd/Otx2 and unpg/Gbx2 result in the loss or misplacement of the brain-specific expression domains of orthologous Pax and Hox genes. Moreover, otd and unpg appear to negatively regulate each other at the interface of their expression domains.

In vertebrate brain development, the Pax2 gene, and subsequently the Pax5 and Pax8 genes, are among the first genes expressed at the Otx2/Gbx2 interface, followed by the overlapping expression of En1 and Fgf8 genes (Ye et al., 2001). Inactivation of Pax2, Pax5, En1 or Fgf8 results in the loss of the midbrain and cerebellum because of a failure to maintain development of this brain region (reviewed by Joyner, 1996; Wassef and Joyner, 1997; Acampora and Simeone, 1999; Martinez et al., 1999; Liu and Joyner, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif, 2001). In Drosophila, no obvious brain phenotypes were seen after mutational inactivation of Pax2, Poxn, en/inv or the Drosophila Fgf homologue branchless (bnl) (data not shown). The absence of brain phenotypes in these mutants contrasts with those observed in the vertebrate brain following mutational inactivation of the orthologous Pax2, Pax5, En1 and Fgf8 genes(Liu and Joyner, 2001; Rhinn and Brand, 2001; Wurst and Bally-Cuif,2001).

Comparative developmental studies in urochordates and vertebrates have led to the notion that the basic ground plan for the chordate brain consists of a forebrain/midbrain region characterized by Otx gene expression, a hindbrain region characterized by Hox gene expression, and an intervening boundary region characterized by expression of Pax2/5/8 genes(Wada et al., 1998; Holland and Holland, 1999; Wada and Satoh, 2001). Our data show that this tripartite ground plan is also found in the insect brain(Fig. 8). This suggests that a corresponding, evolutionarily conserved, tripartite organization also characterized the brain of the last common ancestor of insects and chordates.

A comparison of the brain-specific topology of gene expression patterns that define this tripartite organization in Drosophila and in mouse suggests that the vertebrate midbrain/hindbrain boundary (MHB) region corresponds to the insect deutocerebral-tritocerebral boundary (DTB) region. If this is the case, one might expect that other patterning genes that characterize the MHB region are also expressed at the insect DTB. Although this expectation is fulfilled for the segment-polarity genes en and wingless (wg) in Drosophila, these two genes are expressed at the borders of all CNS neuromeres, as well as at parasegmental boundaries in the epidermis; hence, their expression may not be indicative of brain-specific requirements.

In addition to remarkable similarities in orthologous gene expression between insects and chordates, our study also shows that several functional interactions among key developmental control genes involved in establishing the Pax2/5/8-expressing MHB region of the vertebrate brain are also conserved in insects. Thus, in the embryonic brains of both fly and mouse, the intermediate boundary regions, DTB and MHB, are positioned at the interface of otd/Otx2 and unpg/Gbx2 expression domains(Fig. 8A,B). These boundary regions are deleted in otd/Otx2-null mutants and mispositioned in unpg/Gbx2-null mutants. Moreover, otd/Otx2 and unpg/Gbx2 genes engage in crossregulatory interactions, and appear to act as mutual repressors at the interface of their brain-specific expression domains. However, not all of the functional interactions among genes involved in MHB formation in the mouse appear to be conserved at the Drosophila DTB. Thus, in the embryonic Drosophila brain, no patterning defects are observed in null mutants of Pax2, Poxn, en or bnl. It remains to be seen if these genes play a role in the postembryonic development of the Drosophila brain.

It is conceivable that the similarities of orthologous gene expression patterns and functional interactions in brain development evolved independently in insects and vertebrates. However, a more reasonable explanation is that an evolutionary conserved genetic program underlies brain development in all bilaterians. This would imply that the generation of structural diversity in the embryonic brain is based on positional information that has been invented only once during evolution and is provided by genes such as otd/Otx2, unpg/Gbx2, Pax2/5/8 and Hox, conferring on all bilaterians a common basic principle of brain development. If this is the case, comparable orthologous gene expression and function should also characterize embryonic brain development in other invertebrate lineages such as the lophotrochozoans. This prediction can now be tested in lophotrochozoan model systems such as Platynereis(Arendt et al., 2001), Helobdella (Kourakis et al.,1997) or Dugesia(Pineda et al., 2002).

Taken together, our results indicate that the tripartite ground plan that characterizes the developing chordate brain is also present in the developing insect brain. This implies that a corresponding tripartite organization already existed in the brain of the last common urbilaterian ancestor of insects and chordates. Therefore, we propose an urbilaterian origin of the tripartite brain.

We thank D. Arendt, B. Bello, A. Gould, R. Leemans and M. Müller for comments on the manuscript. We also thank Y. R. Stoll and B. Stoll for excellent technical assistance, and M. Affolter, K. Basler, P. Beachy, the Bloomington Stock Center, K. Furukubo-Tokunaga, A. Ghysen, C. Goodman, T. Kaufman, M. Krasnow,. N. McGinnis, W. McGinnis and K. Matthews for flies and antibodies. This work was supported by the Swiss National Science Foundation.