40 years of homeodomain transcription factors in the Drosophila nervous system

After the discovery of the homeobox (summarised by Scott, 2024), it quickly became clear that homeodomain transcription factors (HDTFs) were evolutionarily conserved (reviewed by Wanninger, 2024). Genetic studies have since revealed that HDTFs played key roles in most – if not all – phases of Drosophila central nervous system (CNS) development (Fig. 1) and HDTFs are often detected in subsets of neurons (Feng et al., 2021).

The first homeodomain proteins shown to be expressed in the CNS and required for neuronal fate specification were Fushi tarazu (Ftz) (Doe et al., 1988a) and Even-skipped (Eve) (Doe et al., 1988b). Both were discovered in the pioneering screen of Nusslein-Volhard and Wieschaus, where they exhibited ‘pair-rule’ mutant phenotypes (Nüsslein-Volhard and Wieschaus, 1980) and, subsequently, cloning revealed that both had seven stripe ‘pair-rule’ expression in early embryos (Macdonald et al., 1986; Weiner et al., 1984). Unexpectedly, however, they also had a second phase of expression in neuronal subsets (Doe et al., 1988a; Frasch et al., 1987). To determine the function of ftz in neurons, a transgene containing the pair-rule enhancers but lacking the neuronal enhancer was used to remove Ftz specifically in the CNS and showed it was required in RP2 motor neurons for the expression of Eve and proper axon pathfinding (Doe et al., 1988a). Similarly, eve encodes a HDTF expressed in a pair-rule pattern followed by segmental expression in a sparse population of neurons, including RP2. An eve temperature-sensitive allele was used to inactivate Eve specifically during neurogenesis and, again, there was abnormal axon pathfinding in RP2, as well as other neurons (Doe et al., 1988b). Thus, both Ftz and Eve were shown to be expressed and function in the developing CNS. Further work over the years has shown that Eve plays an important role in specifying dorsal-projecting motor neurons (see below), whereas there has been little follow-up work on Ftz in neurogenesis.

In the late 1980s and early 1990s, several more HDTFs were characterized for their expression and function during neurogenesis. The cut gene was identified in a screen for defective embryonic sensory neuron development; cut mutants showed a transformation of external sensory neurons to chordotonal sensory neurons (Bodmer et al., 1987). One year later, the cloning of the cut locus revealed it encoded a HDTF (Blochlinger et al., 1988). Two other early characterized HDTFs were Ventral veinless (Vvl; formerly called Cf1a) and Rough. Vvl was shown to be required for the proper development of a subset of dopaminergic neurons in the larval CNS (Johnson and Hirsh, 1990), whereas the rough gene was identified based on its rough eye phenotype and shown to be expressed in photoreceptor (R)2 and R5. The misexpression of Rough throughout the developing eye resulted in changes in photoreceptor neuron identity (Basler et al., 1990; Kimmel et al., 1990). In the decades following these pioneering studies, an increase in the number of molecular markers and neuronal morphology tracing tools allowed further precision in phenotypic characterization, but all approaches validate the early studies: HDTFs are often detected in subsets of neurons, where they play important roles in specifying neuronal identity (Box 1). In this Spotlight, we introduce how the Drosophila central nervous system develops before discussing some of the key HDTFs that function at different stages of the process in the embryo and larva, but we note that recent work has shown roles for HDTFs in the adult nervous system (Allen et al., 2020; Baek et al., 2013; Baek and Mann, 2009; Issa et al., 2019; Raouf Issa et al., 2022; Xu et al., 2024a,b).

The Drosophila central nervous system (CNS) develops in a multi-step manner (reviewed by Skeath and Thor, 2003) (Fig. 1). Initially, specification of the neuroectoderm occurs in two ventral ectoderm domains, which produce the ventral nerve cord (VNC), and two anterior dorsal domains, which generate the brain. The neuroectoderm is patterned by anterior-posterior (A-P) and dorso-ventral (D-V) cues to generate molecularly and developmentally distinct neuroblasts (NBs), which are the stem cells of the CNS (see below), within each CNS half segment (hemisegment) (Fig. 1A). The A-P patterning also ensures that these NBs are distinct in their behavior between each CNS segment. NBs commence neurogenesis by asymmetric cell division (see below), renewing themselves and budding daughter cells with a more limited developmental potential in one of three ways: directly differentiating as neurons (type 0 lineage), producing a ganglion mother cell (GMC) that undergoes a terminal division to make two sibling neurons (type I lineage) or producing intermediate neural progenitors (INPs) that divide multiple times (type II lineage) (reviewed by Doe, 2017) (Fig. 1B). NBs also undergo temporal patterning, providing intra-lineage diversity (see below; Fig. 1C). The number of divisions of each NB and the daughter cell division modes employed are both linked to NB identity and temporal progression, and result in NB-specific lineage sizes from 2-40 cells (Fig. 1D). Last, in most type I lineages, each GMC undergoes a molecularly asymmetric division in which one daughter inherits the Notch inhibitor Numb (Notch^OFF sibling) and a sibling lacking Numb (Notch^ON sibling) (reviewed by Yaghmaeian Salamani and Thor, 2020). An important unresolved issue is whether this mechanism is also used in type II lineages. Within each hemisegment, each neuron appears to have a distinct identity, including the identity of programmed cell death (PCD), which occurs in a stereotyped number of neurons, suggesting that PCD is one variant of cell fate (Miguel-Aliaga and Thor, 2009) (Fig. 1E). The A-P patterning cues that ensure segment-specific NB specialization also result in segment-specific cell fates, such as motor neurons and glia (Fig. 1F). Thus, these three mechanisms – spatial patterning, temporal patterning and Notch signaling – result in the final number of ∼400 distinct neurons and glia in each hemisegment, with a further diversification along the CNS A-P axis. The peripheral nervous system (PNS) is generated along the lateral body wall with sensory organ precursors (SOPs) undergoing specific lineages to generate the diverse pool of sensory neurons (Jan and Jan, 1990).

HDTFs are expressed in the CNS from the earliest stages. Within the neuroectoderm of the VNC, an orthogonal array of HDTFs patterns the neuroectoderm to generate unique NB identities. The first NBs to form align in four rows (1, 3, 5 and 7) and three columns (ventral, intermediate and dorsal)(Skeath et al., 1995). Rows 5 and 7 express the HDTFs Gooseberry (Gsb) and Engrailed (En), respectively; the two anterior rows currently lack a defining HDTF (Skeath et al., 1995; Zhang et al., 1994). Misexpression of Gsb transforms row 3 into row 5 but has no effect on the anterior row NB identity (Skeath et al., 1995; Zhang et al., 1994). En has not been rigorously tested for a role in specifying row 7 identity (Bhat and Schedl, 1997). HDTFs are also essential for specifying columnar identity. The HDTFs Ventral nervous system defective (Vnd), Intermediate neuroblasts defective (Ind) and Muscle-specific homeobox (Msh or Drop in FlyBase) are expressed in the ventral, intermediate and dorsal columns, respectively (McDonald et al., 1998; Weiss et al., 1998), similar to their mammalian orthologs (Weiss et al., 1998). Loss of Vnd results in expansion of Ind into the ventral neuroectoderm; similarly, loss of Ind results in expansion of Msh into the intermediate domain (McDonald et al., 1998; Weiss et al., 1998). The orthogonal array of unique NB identities present within each VNC hemisegment is modified along the A-P axis, chiefly by the Hox homeotic genes, which act at multiple steps to alter the NB array, control NB and daughter cell proliferation, as well as cell fate and PCD (Miguel-Aliaga and Thor, 2009; Technau et al., 2014; Yaghmaeian Salmani and Thor, 2020). Although our understanding of spatial patterning is incomplete, HDTFs clearly play a key role in the process.

Similar to the patterning of the neuroectoderm that gives rise to the VNC, some HDTFs are expressed in spatial domains of the dorsal brain anlage; these include Orthopedia (Otp), Empty spiracles (Ems), Homeobrain (Hbn), Retinal homeobox protein (Rx), Nkx6 (HGTX in FlyBase) and Brain-specific-homeobox (Bsh) (reviewed by Urbach and Technau, 2003). These genes, and others, are required to pattern the dorsal neuroectoderm and thus specify distinct brain NB identities.

After spatial patterning to establish distinct NB identities, NBs undergo a shared temporal transcription factor (TTF) cascade to diversify cell types within each NB lineage. In the VNC, the cascade is Hunchback (Hb), Kruppel (Kr), Nubbin/Pdm2 (Nub), Castor (Cas) and Grainy head (Grh) (reviewed by Doe, 2017) (Fig. 1C). Only Nubbin/Pdm2 are HDTFs. Similarly, HDTFs are poorly represented in central brain and optic lobe TTF cascades; the central brain mechanism primarily uses gradients of RNA-binding proteins, with no HDTFs found so far, and the optic lobe TTF cascade [Homothorax (Hth), Klumpfuss (Klu), Eyeless (Ey), Sloppy paired (Slp), Dichaete (D) and Tailless (Tll)] contains only Hth and Ey HDTFs (Doe, 2017). Thus, spatial patterning genes are highly represented by HDTFs, whereas different DNA- and RNA-binding proteins predominate in temporal patterning.

NB asymmetric cell division was first observed by Wheeler (1893), but it took over one century, until 1995, for the first molecular asymmetry of a HDTF to be reported. Two labs independently showed that the Prospero (Pros) protein was asymmetrically localized into the smaller GMC (Hirata et al., 1995; Spana and Doe, 1995) (Fig. 1B). Pros is an atypical homeodomain protein with an extended C-terminal extension that forms an integrated domain when crystallized (Ryter et al., 2002). Although atypical, Pros orthologs (Prox1) are found widely in both invertebrates and vertebrates.

HDTFs are not only important for neuronal cell types, but are also essential for glial specification. The HDTF Reversed potential (Repo) was first discovered through its role in the eye, but soon it was recognized to be a glial-specific HDTF (Campbell et al., 1994; Xiong et al., 1994) (Fig. 1F). Loss of Repo transforms glia into neurons, and misexpression of Repo results in the converse neuron to glial transformation (Campbell et al., 1994; Xiong et al., 1994).

HDTFs are prominently involved in the specification of motor neuron (MN) identity; here, we focus on embryonic ventral nerve cord MNs that project to ∼30 muscles arranged from ventral to dorsal along the body wall (Landgraf et al., 1997) (Fig. 2). There are ∼36 bilateral MNs in each abdominal segment, each projecting to a characteristic pattern of one or two muscles (Landgraf et al., 1997; Zarin et al., 2019). Although no HDTF determines generic MN identity, the zinc-finger homeodomain protein 1 (Zfh1) comes the closest, being expressed in all MNs, and in a small number of interneurons and glia (Layden et al., 2006); loss of Zfh1 reduces the ability of MNs to exit the CNS, whereas ectopic Zfh1 in some interneurons promotes ectopic outgrowth from the CNS (Layden et al., 2006).

By contrast, HDTF combinatorial codes play a key role in generating MN diversity. Eve is expressed in seven bilateral MNs that all project to specific dorsal longitudinal muscles and is required for proper dorsal MN axon outgrowth (Doe et al., 1988b; Landgraf et al., 1999). Lim1 is also expressed in these Eve⁺ dorsal projecting neurons, but not in ventral projecting neurons (Lilly et al., 1999). A combination of Nkx6, Islet (Tup in FlyBase), Lim3, Drifter (Dfr) and Hb9 (Exex in FlyBase) is expressed in ventral projecting MNs, but not in Eve⁺ dorsal projecting neurons (Broihier et al., 2004; Broihier and Skeath, 2002; Certel and Thor, 2004; Lacin et al., 2014; Landgraf and Thor, 2006; Odden et al., 2002; Thor et al., 1999; Thor and Thomas, 1997). Further reinforcing this non-overlapping dorsal/ventral connectivity are the cross-repressive interactions between Eve and Hb9, and between Eve and Nkx6 (Broihier et al., 2004; Broihier and Skeath, 2002). The ventral projecting MNs are further diversified by the selective expression of Dfr only in the ISNb subclass (Certel and Thor, 2004). In addition to the two classes of MNs described above, a third population expressing the HDTF BarH1 does not express dorsal or ventral HDTFs and projects to transverse muscles (Garces et al., 2006; Heckscher et al., 2014).

A key identifying feature of different MNs is their peripheral muscle targets; logically, mutations in and/or misexpression of the aforementioned HDTFs do indeed result in axon-targeting defects (Zarin et al., 2014). However, how these HDTFs control the extensive battery of axon pathfinding genes is still not well understood. One example of this connection pertains to the regulation of the Beaten path (Beat) family of immunoglobulin-containing cell-adhesion molecules by Islet and Lim3 (Certel and Thor, 2004). Similarly, another key MN subtype property is the distinct electrophysiological features of each MN subtype. Studies have identified direct transcriptional links between Eve, Hb9, Islet and Lim3, and ion channel and neurotransmitter receptors (Pym et al., 2006; Wolfram et al., 2014).

Interestingly, there appears to be motor neurons that do not express any of the known HDTFs (Heckscher et al., 2014), raising the possibility that additional HDTFs that specify MN identity remain to be discovered. Thus, a combinatorial code of HDTFs provides major input into MN specification, although other transcription factors make some contribution (Garces and Thor, 2006). There has been great progress on MN specification by HDTFs over the past two decades, but many unanswered questions remain. What is the precise combinatorial expression of the ventral MN HDTFs, and do they provide coverage of all ventral projecting MNs? What are the HDTF cell surface molecule targets that guide neuromuscular connectivity? Do the MN HDTFs persist throughout larval development and do they regulate functional features of mature neurons, such as ion channels and neurotransmitters?

Many interneurons, MNs and secretory neurons, in addition to their other subtype-defining features, are also defined by their highly selective expression of one of the more than two dozen neuropeptides encoded by the Drosophila genome (Miguel-Aliaga et al., 2004; Park et al., 2008). In the VNC, the Apterous (Ap) HDTF is selectively expressed in two neuropeptide cell types: the FMRFamide (FMRFa) and Nplp1 neurons, and controls both axon pathfinding and neuropeptide expression in these cells (Baumgardt et al., 2007; Benveniste et al., 1998; Lundgren et al., 1995) (Fig. 3). The role of Ap in selectively governing FMRFa versus Nplp1 cell identity is determined by other TFs and co-TFs selectively expressed in the two cell types, acting in combinatorial codes with Ap (Allan et al., 2005; 2003; Baumgardt et al., 2007; Gabilondo et al., 2016; Miguel-Aliaga et al., 2004; Stratmann et al., 2016). Molecular studies reveal that Ap acts directly upon the Nplp1 and FMRFa genes (Stratmann and Thor, 2017). In addition, genetics and genomics have identified several putative axon pathfinding genes regulated by Ap that may act to govern the precise axon projects of these two neuropeptide neuron cell types (Stratmann et al., 2019). Ap expression in FMRFa and Nplp1 neurons is triggered by several HDTFs acting upstream, i.e. Pdm (temporal), Antp (Hox, spatial), Hth (Hox co-factor) and Lbe (NB-identity, spatial), demonstrating how cascades of HDTFs act to dictate terminal cell identity (Baumgardt et al., 2009; Gabilondo et al., 2016; Karlsson et al., 2010).

Despite extensive studies over the past four decades, several issues regarding HDTFs and nervous system development require further studies. One key question moving forward is to address how HDTFs mechanistically act to control different developmental decisions, such as the cell fate of specific neurons; what is the full repertoire of downstream genes at each developmental stage? Moreover, how is the dual role of some HDTFs controlled (e.g. acting early in a patterning role and later in a terminal cell specification role)? Finally, how are early spatial and temporal patterning cues ‘carried forward’ to terminal cell identity? Addressing these questions will help to provide a more comprehensive picture of the role of HDTFs during Drosophila nervous system development and will likely continue to guide studies in other systems.

This article is part of the collection ‘40 years of the homeobox’. See related articles in this collection at https://journals.biologists.com/dev/collection/10249/40-years-of-the-homeobox.

We thank Peter Newstein for Fig. 2. We thank Peter Newstein and Derek Epiney for their comments on the manuscript.