Ancestral regulatory mechanisms specify conserved midbrain circuitry in arthropods and vertebrates

We focused on phylotypic (35) stage 11 to 14 embryos (Fig. 1 A and B) to characterize morphological and molecular signatures of the developing Drosophila DTB. In addition to the adjoining otd and unpg expression and function reported earlier (34), we found specific domains of expression of the Pax2/5/8 homologs shaven(sv)/dPax2 and Pox neuro (Poxn), as well as engrailed (en), wingless (wg/dWnt), muscle specific homeobox (msh/dMsx), ventral nervous system defective (vnd/dNkx2), and empty spiracles (ems/dEmx) (Fig. 1C and SI Appendix, Fig. S1A). For axial patterning, we examined the expression and function of the key genes otd + wg (antero-posterior) and msh + vnd (dorsal-ventral), which revealed essential roles in DTB formation (SI Appendix, Fig. S1B).

A cardinal feature of the vertebrate MHB is its organizer activity, mainly mediated by the FGF8 effector molecule (30–33, 36). Previous studies failed to identify a DTB-related function of the FGF8 homolog branchless and its receptor breathless in embryonic brain development of Drosophila (34). We now show that a second set of FGF8-like orthologs, thisbe/FGF8-like1 and pyramus/FGF8-like2 (37, 38), and the FGF8 receptor heartless (htl) are expressed at the DTB (SI Appendix, Fig. S2). Consistent with earlier reports (37, 38), phylogenetic comparison of annotated protein sequences reveals that Pyramus and Thisbe, like Branchless, are homologs of human and mouse FGF8, FGF17, and FGF18, as well as of the ascidian Ciona intestinalis FGF8/17/18, the Annelid Capitella teleta FGF, and the Cnidarian Nematostella vectensis FGF8a (SI Appendix, Fig. S3). Similarly, phylogenetic comparison of annotated protein sequences reveals that the Drosophila FGF receptors Breathless and Heartless are homologs of human and mouse FGF receptors FGFR1-FGFR4, as well as of the ascidian C. intestinalis FGFR, the annelid C. teleta FGFR, and the cnidarian N. vectensis FGFR (SI Appendix, Fig. S4). A functional role for FGF8 signaling at the Drosophila DTB was revealed by altered engrailed expression patterns and morphological defects in embryos homozygous for a deficiency, Df(2R)BSC25, uncovering both FGF8-like1 and FGF8-like2 genomic loci, and of an htl-null allele, htl^AB42−/− (Fig. 2A). As indicated in the Insets of Fig. 2 A, a–c, these defects impact the prefrontal commissure (indicated by arrowheads) and the size and integrity of longitudinal connectives (indicated by the white brackets), as well as axonal projections, also indicated by the white arrow in each panel at the bifurcation between prefrontal commissure and longitudinal connective. These observations were further substantiated by progressive changes and loss of the DTB expression patterns of unpg and ems in htl-null mutant embryos, between embryonic stages 12 to 16 (Fig. 2 B, d–f; compare to panels a–c). These data identify a role of FGF8-like signaling in the maintenance of the embryonic DTB region in Drosophila. Given the expression pattern of FGF8-like homologs in the brain and procephalic endoderm (SI Appendix, Fig. S2), it remains to be shown whether cell autonomous or cell nonautonomous FGF8 signaling is required for DTB formation and maintenance. Ectopic expression of the FGF8-like homolog thisbe in ems-specific brain regions did not cause any detectable changes in morphology or molecular signatures of the DTB region (SI Appendix, Fig. S5). Despite conserved regulatory interactions between otd/Otx and unpg/Gbx (34), these data indicate the absence of FGF8-mediated organizer activity in the embryonic DTB. We conclude that direct or indirect FGF8-like signaling is required for the maintenance of the embryonic DTB but, contrary to what is seen in vertebrate MHB development, appears not to organize the DTB region in Drosophila.

To identify the adult brain structures and associated functional modalities that derive from the embryonic DTB, we utilized genetic tracing of neural lineages employing the tracer line UAS-mCD8::GFP, tub-FRT-CD2-FRT-Gal4, UAS-FLP/CyO GMR Dfd YFP (39, 40). We first deployed en-Gal4 as driver, which, combined with the tracer line, causes permanent GFP labeling of progeny of engrailed-expressing cells, even when those progeny themselves no longer express engrailed. This is due to the fact that the genetic tracer line carries a flippase (FLP) recombinase-related flip-out cassette, in which a strong and ubiquitous tubulin enhancer is separated from a Gal4 transcriptional activator by CD2, flanked by a pair of FRT sites, thereby preventing Gal4 expression. As soon as the genetic tracer line is crossed with an independent Gal4 driver, in this case en-Gal4, it initiates UAS-FLP expression as part of the genetic tracer, which in turn induces recombination at the CD2-flanking FRT sites. As a result, the CD2 element is excised, resulting in a fusion of the tubulin enhancer to drive Gal4, which in turn leads to expression of UAS-mCD8::GFP in that cell and all of its progeny. Since tubulin is constitutively active, the genetic tracing thus permanently marks targeted cells and their progeny, thereby revealing lineage-relationship even if some of the traced cells no longer express engrailed.

Genetic tracing of engrailed-expressing lineages of the embryonic neuroectodermal DTB (SI Appendix, Fig. S6 A–C) identified neurons and projections of the antennal mechanosensory and motor center (AMMC) and select antennal glomeruli in the adult brain (Fig. 1 D and E and SI Appendix, Fig. S6 D–F). We also determined the fate of Poxn-expressing cells, which in the embryonic brain are located next to DTB-specific engrailed lineages (Fig. 3 A–H). In the adult brain, Poxn-expressing cells are present in a caudal volume of the deutocerebrum called the Wedge receiving auditory and gravity sensing afferents (41), where its interneurons, as do En-positive cells, express the neurotransmitter GABA (Fig. 3 I–P). Genetic tracing revealed that the Poxn-expressing AMMC/Wedge neurons derive from DTB lineages. Furthermore, their region-specific projection patterns resemble the mechanosensory pathway architecture of local interneurons and projection neurons (Fig. 3 Q–T) previously identified for the AMMC and the Wedge (41–44).

The Drosophila AMMC and Wedge neuropils comprise neurons that mediate auditory, vestibular, mechanosensory, and somatosensory information processing in pathways with similarities to the mammalian auditory and vestibular pathways (41–44). In vertebrates, auditory, vestibular, somatosensory, and motor information are processed by neural populations of the tectum and cerebellum, adult brain structures derived from the MHB region (36, 45). These fate-mapping studies in mice did not identify specific neural circuits within the tectum and cerebellum, which limits a structural comparison until more refined studies at cell and circuit resolution in vertebrates become available.

Despite these limitations, we reflected on whether functional comparisons might be possible. The tectum and cerebellum receive auditory and vestibular, as well as motor, information and, among other functions, are important for balance, body posture, sensorimotor integration, and motor coordination (1, 4, 46). To test whether DTB-derived circuits in Drosophila might exert similar functions, the GAL4-UAS system was used to express Tetanus-Toxin-Light-Chain (TNT) and inhibit synaptic transmission (47) in subsets of AMMC neurons (42). Flies were tested for their startle-induced negative geotaxis (SING) response, which, after being shaken to the bottom in a test tube, quantifies their ability to right themselves and climb upwards (48). Except R19E09 > TNT, all of the tested genotypes showed significantly impaired SING behavior (Fig. 4A and SI Appendix, Table S1). This includes R79D08, R45D07, and R30A07 that target AMMC neurons and coexpress engrailed, or encompass Poxn and engrailed expression domains (Fig. 4 C–I). Of note, several of the tested genotypes showed difficulties with balance and to right themselves, as exemplified for R52F05 > TNT compared to control (Movies S1 and S2).

To further analyze AMMC-mediated motor coordination, we employed video-assisted motion tracking and recorded freely moving flies (Fig. 4B). During 135-min recordings, activity bouts and movement trajectories were analyzed to quantify locomotion parameters: frequency of episodic movements, how often they were initiated, and their length and average velocity, as well as the duration and frequency of pauses. Response to sensory stimulation triggered by repeated mechanical shock was also recorded. UAS-TNT expression by R79D08 (CG9650/dZic-B)-Gal4, which targets zone B (41–43) of the AMMC (Fig. 4C, arrows), significantly impaired overall activity and duration, with fewer actions initiated, shorter episodes of activity and their intervals, and reduced velocity and distances traveled (Fig. 4D and SI Appendix, Fig. S7A). UAS-TNT expression by R88B12 (en/inv)-Gal4 targeting zone A (41–43) of the AMMC (Fig. 4E, arrows) significantly impaired average and prestimulus speed, with shorter bouts of activity, together resulting in less distance traveled (Fig. 4F and SI Appendix, Fig. S7B). Comparable motor phenotypes were seen with R30A07 (NetA)-Gal4, which targets AMMC neurons that coexpress engrailed (Fig. 4G, arrowheads, Fig. 4H, and SI Appendix, Fig. S7C), but not with R45D07 (hth)-Gal4 targeting parts of the AMMC-specific giant fiber system (Fig. 4 I and J and SI Appendix, Fig. S7D). Together, with SING data, our behavioral observations establish essential functions of the AMMC for sensorimotor integration (41, 42, 44), balance, righting reflex, and motor coordination in Drosophila, behavioral manifestations similar to MHB region-derived circuits in vertebrates.

Our findings thus far establish correspondences between Drosophila DTB and vertebrate MHB at multiple levels, including adult brain circuits and the behaviors they regulate. We hypothesized that this will be reflected in commonalities among character identity networks of DTB and MHB that are mediated by homologous gene regulatory networks (27, 28). To test this hypothesis, we screened the Janelia Gal4 collection (49) for cis-regulatory elements (CREs) mediating the spatiotemporal expression of developmental genes controlling DTB formation in Drosophila. We identified CREs for msh, vnd, ems, and for thisbe/FGF8-like1 (SI Appendix, Fig. S8 A–E), genes that are essential for the formation and/or maintenance of the embryonic DTB (Fig. 2 and SI Appendix, Fig. S1) (14). In addition, we identified CREs for Wnt10, Sex combs reduced/Hox5; the Drosophila homologs of zinc finger of the cerebellum (ZIC), odd-paired (opa/dZic-A) and CG9650/dZic-B; of Purkinje cell protein 4 (PCP4), igloo (igl/dPCP4); of Ptf1a, Fer1/dPtf1a, and of Lim1 (SI Appendix, Fig. S8 F–J). All mammalian homologs of these genes have been implicated in vertebrate MHB formation and the specification of midbrain-cerebellar circuitry (30–34, 50) (SI Appendix, Table S2).

We also identified CREs for dachshund (dac) and the Pax2 homolog shaven (sv/dPax2). dac-specific CRE R65A11-Gal4 targeted UAS-mCD8::GFP expression to the DTB region, in derived lineages of the embryonic brain and to the AMMC in a pattern encompassed by DTB-specific engrailed and Poxn expression domains (Fig. 5 A–C). The regulatory element VT51937 (51) located in an intron of sv/dPax2 targets Gal4 expression in a segment-specific pattern similar to endogenous sv/dPax2, including DTB expression domains (SI Appendix, Fig. S9 A–F). VT51937-Gal4–mediated genetic tracing also identified cells and projections in the AMMC (SI Appendix, Fig. S9G). Together, these data identify CREs of the DTB-AMMC character identity network in Drosophila that mediate the spatiotemporal expression patterns of genes that are homologous to genes involved in the formation and specification of the vertebrate MHB and derived midbrain-cerebellar circuitry (30–34, 50).

Given the correspondences between the Drosophila DTB and vertebrate MHB gene regulatory and character identity networks, we asked whether the CREs that control the expression of these genes might be conserved. To identify conserved cross-phylum CREs, we utilized DTB-AMMC–specific regulatory sequences and applied bioinformatics tools (52) including VISTA (53), MLAGAN (54), and EMBOSS MATCHER (55) to screen for corresponding CREs in mouse and human genomes (56). Stringent selection criteria (57) were applied to identify CRE sequences 1) that are linked to homologous genes in the different species; 2) with a minimum of 62% sequence identity over at least 55 base pairs (bp) and at least 1e⁻¹ confidence level as the BLAST e-value; 3) that are not unannotated protein sequences; and 4) that are not repetitive elements. Additionally, BLAT (58) was applied for searching vertebrate genomes using the same selection criteria.

We first analyzed the DTB-AMMC–specific CRE of sv/dPax2 (= VT51937 sequence) and identified noncoding CREs for mouse Pax2 and human PAX2 with extensive sequence similarities (SI Appendix, Fig. S9H and Supplementary Dataset S1), and comparable intragenic location (SI Appendix, Fig. S9I). To validate the significance of these sequences, we carried out BLAST/BLAT searches of the sv/PAX2 conserved sequence against the Drosophila melanogaster, Mus musculus, and human genome sequences. These data revealed that, in D. melanogaster, only the sv/dPax2-related CRE of 255 bp with 1.6E−142 Blast e-value matches the cutoff criteria of >62% sequence identity over at least 55 bp with minimum 1^e-1 confidence level as the BLAST e-value whereas all other identified sequences were of 18 bp or shorter length (SI Appendix, Supplementary Dataset S1). Of note, BLAT searches of the sv/PAX2 conserved sequence against the mouse and human genome identified CREs only related to mouse Pax2 and human PAX2 and no other genomic sequence (SI Appendix, Supplementary Dataset S1). In addition, using the JASPAR algorithm (59), we identified potential transcription factor binding sites that match stretches of the MLAGAN-aligned sv/PAX2 conserved CRE sequence (SI Appendix, Supplementary Dataset S1).

Following this strategy, we used DTB-AMMC–specific CRE sequences for dachshund and engrailed/invected and identified human CREs conserved among vertebrates that direct MHB-specific expression in mouse for the dachshund homologs Dach1/DACH1 (Fig. 5 D–F) and for the engrailed/invected homologs Engrailed2/EN2 (SI Appendix, Fig. S10). Further bioinformatics analysis identified core CRE sequences associated with dac and DACH1, en/inv and EN2 and sv and PAX2 in Drosophilidae and vertebrate genomes (SI Appendix, Supplementary Datasets S1–S3). As was the case for the sv/PAX2 conserved CRE sequences, BLAST/BLAT searches against the D. melanogaster, M. musculus, and human genomes revealed for the D. melanogaster genome that only the en/inv-related CRE of 616 bp with 0.0 Blast e-value (SI Appendix, Supplementary Dataset S2) and dac-related CRE of 247 bp with 9.4E-138 Blast e-value (SI Appendix, Supplementary Dataset S3) match the cutoff criteria whereas all other identified sequences are of maximum 23 bp (for en/inv-related) or 30 bp (for dac-related) or shorter length. Also, BLAT searches of the invected/engrailed-EN2 and dachshund/DACH1 conserved sequences against the mouse and human genome only identified CREs for mouse En2 and human EN2, and mouse Dach1 and human DACH1 respectively, whereas no other genomic sequences can be identified (SI Appendix, Supplementary Datasets S2 and S3). Furthermore, and again using the JASPAR algorithm, we identified potential transcription factor binding sites that match stretches of the MLAGAN-aligned invected/engrailed-EN2 (SI Appendix, Supplementary Dataset S2) and dachshund/DACH1 (SI Appendix, Supplementary Dataset S3) conserved CRE sequences, respectively. Together, these data suggest ancestral noncoding regulatory sequences and their function predate the radiation of insect-specific DTB and vertebrate-specific MHB circuits and morphologies.

We have identified gene regulatory and character identity networks that underlie the formation of the deutocerebral–tritocerebral boundary in Drosophila. Mutant analyses reveal that otd + wg and msh + vnd, acting along the AP and DV body axes, respectively, are required for the formation of the embryonic DTB, and that FGF8-like signaling is necessary for its developmental maintenance. Genetic tracing experiments, together with the analysis of CREs for engrailed/invected, dachshund, and shaven/dPax2, as well as behavioral analysis after synaptic inactivation, show that embryonic DTB lineages give rise to neural circuits in the AMMC/Wedge complex of the adult brain that mediate balance and motor coordination in Drosophila. Together, these findings establish a ground pattern of DTB formation and derived circuit function in Drosophila that corresponds to the ground pattern and gene regulatory and character identity networks of the vertebrate MHB and derived midbrain–cerebellar circuits (SI Appendix, Fig. S11).

In contrast to a previous gene expression study pointing toward the protocerebral/deutocerebral boundary (60), our findings based on gene regulation and function identify the Drosophila DTB as the boundary between the rostral brain and its genetically distinct caudal nervous system. These data imply that caudal domains of the arthropod deutocerebrum and its circuits in Drosophila correspond to the vertebrate MHB and its derived principle proprioceptive circuits (SI Appendix, Fig. S11). It must be emphasized here that these circuits are not ascribed to the cerebellum, the anlage of which forms as an asegmental volume within Gbx2 and non-Hox expression domains of the developing MHB (30–33, 36, 61). Rather, our comparative analysis suggests that vestibular/balance-related circuits characterize the ancestral territory from which the DTB and MHB evolved, and that MHB-derived cerebellar circuitry is a vertebrate innovation. In support of this notion, FGF8 signaling in vertebrates acts as a secondary organizer in boundary development of the MHB and by promoting growth essential for the formation of the tectum and cerebellum (30–34, 36, 61). We did not observe FGF8-like organizer activity in flies but a role in the maintenance of the DTB boundary, suggesting an ancestral boundary-related function for FGF8 (61). A comparable phenotype has been described in ascidian larvae mutant for Ffg8/17/18 (62). Normally expressed in the visceral ganglion, knockdown of Fgf8/17/18 altered Otx, Engrailed, and Pax2/5/8 gene expression in the anterior adjacent neck region of Ciona, essentially leading to a posterior expansion of rostral central nervous system (CNS) identity (62). These data suggest that, similar to the situation in the embryonic brain of Drosophila, FGF8 signaling in the tadpole larvae of C. intestinalis delineates the boundary between the rostral brain and the caudal nervous system. Moreover, the observed absence of extended proliferative activity in Drosophila (ref. 34 and this paper) and Ciona (62) suggests that growth-related organizer activity of FGF8 is a vertebrate innovation whereas the boundary region is defined by expression patterns of genes homologous to those observed at the DTB/MHB (Fig. 6). Indeed, despite comparable expression domains (11, 12, 63, 64), no phenotypic cerebellum is found in ascidians, nor in hemichordates and cephalochordates, none of which can be assumed as proxies for ancestral vertebrates, but all of which may represent highly derived and evolutionarily simplified crown species. However, ancestral circuits mediating vestibular and motor (balance) coordination, which are specified by genes and regulatory networks homologous to those described in this paper, are found in lampreys and hagfish, the persisting ancient lineages of early vertebrates (2), and maybe also in the chordate brain of C. intestinalis tadpole larvae (65).

The observed correspondences in circuit formation extend to behaviors they regulate. Synaptic inactivation of DTB-derived subcircuits of the AMMC/Wedge complex results in flies with impaired balance, defective action initiation and maintenance, and compromised sequences of motor actions. These AMMC circuits have been shown to mediate auditory and vestibular information processing and coordination (41, 42), suggesting that the DTB-derived AMMC/Wedge circuits integrate mechanosensory submodalities for motor homeostasis. These functions correspond to the activity of MHB-derived acoustic and vestibular receptor pathways in vertebrates (41, 44) which, when impaired in inherited disorders, affect both auditory and vestibular functions such as seen in ataxic patients (66). The observed correspondences therefore suggest that, similar to MHB-derived circuits, the DTB-derived AMMC/Wedge circuits in arthropods are required for sensorimotor integration, body posture, and motor coordination. These findings identify correspondences between ground patterns of the insect DTB and vertebrate MHB that extend beyond spatiotemporal expression patterns and functions of homologous genes, to neural circuits and behavior.

The present results identify CREs associated with highly conserved developmental control genes (67–70) regulating boundary formation between the rostral brain and the caudal nervous system in insects and vertebrates. Core elements of the identified CREs are highly conserved, as demonstrated by sequence identities across large phylogenetic distances and by the identification of potential transcription factor binding sites that are located within these conserved CREs. Although there is evidence that orthologous CREs can be completely divergent at the sequence level (62), our findings demonstrate that the identified conserved CREs are employed for the formation of circuits with comparable roles in neuronal processing. These data suggest the identified CREs are ancestral noncoding regulatory sequences with which insect-specific DTB and vertebrate-specific MHB circuits and morphologies evolved.

In conclusion, the corresponding ground patterns of insect DTB and vertebrate MHB suggest the early appearance in bilaterian evolution of a cephalic nervous system that evolved predictive motor homeostasis before the divergence of the protostome lineages and before the origin of deuterostomes. Based on the observed correspondences, we hypothesize that the retention across phyla of conserved regulatory mechanisms is necessary and sufficient for the formation and function of neural networks for adaptive behaviors common to all animals that possess a brain.

We thank H. Karten, C. Ragsdale, A. Kamikouchi, A. Delogu, M. Fanto, and T. J. Weidemann for discussions; G. M. Rubin and L. Pennacchio for CRE data; D. Diaper and M. Dyson for graphics; and U. Walldorf, M. Noll, S. Carroll, A. Mueller, M. Michaelson, M. Landgraf, A. Ghysen, and A. Gould, as well as the Developmental Studies Hybridoma Bank and the Vienna Drosophila Resource Center and Bloomington Stock Center, for providing antibodies and fly strains. This work was supported by a PhD fellowship from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior Foundation–Ministry of Education of Brazil (BEX 13162/13-6) (to J.C.B.); an Institute of Psychiatry, Psychology & Neuroscience-King’s College London Independent Researcher Award (to B.K.); a Research Foundation–Flanders (FWO) postdoctoral grant (to L.V.B.); Flemish FWO Grants G065408.N10 and G078914N (to P.C.); NSF Grant 1754798 (to N.J.S.); and United Kingdom (UK) Medical Research Council Grants G0701498 and MR/L010666/1, UK Biotechnology and Biological Sciences Research Council Grant BB/N001230/1, and Motor Neurone Disease Association Grants Hirth/Nov15/914-793, Hirth/Oct13/6202, Hirth/Mar12/6085, and Hirth/Oct07/6233 (to F.H.).