Dendritic diversification through transcription factor-mediated suppression of alternative morphologies

The morphologies of neurons are highly diverse, fitting with distinct functions and connectivity. Transcriptional regulators drive neuronal diversification, as demonstrated by studies in several systems, including rodents, worms and flies (Allan and Thor, 2015). Given the enormous cellular diversity in the nervous system, understanding how multiple transcriptional regulators interact to drive subclass and individual neuronal features is an important goal. Studies of the specification of motor neurons in fly and vertebrate systems have revealed multiple strategies for generating specific axonal trajectories including combinatorial codes, hierarchical genetic cascades, and spatial and temporal patterning (Landgraf and Thor, 2006; Dasen and Jessell, 2009; Enriquez et al., 2015). Drosophila multidendritic sensory neurons have provided a powerful model for identifying transcriptional programs that underlie the morphological diversification of dendritic arbors; nevertheless, regulators have been studied largely in isolation and the strategies for diversification are poorly understood (Corty et al., 2009; Jan and Jan, 2010). Here, we examine transcriptional strategies for the diversification of multiple morphologically and functionally distinct types of somatosensory neurons.

Drosophila sensory neurons are segregated into morphological classes distinguished by sensory dendrite branching pattern, end organ innervation, and axonal targeting in the central nervous system (Bodmer and Jan, 1987; Merritt and Whitington, 1995; Grueber et al., 2002, 2007). A subset of md neurons, the dendritic arborization (da) neurons, extend sensory dendrites across the body wall to cover large territories. The da neurons have been further divided into classes I-IV based on dendritic morphology (Grueber et al., 2002) (Fig. 1A). Dendritic morphology correlates with sensory function, with class I, dbd and dmd1 neurons functioning as proprioceptors (Hughes and Thomas, 2007), class III neurons mediating responses to gentle touch (Yan et al., 2013), and class IV neurons functioning as multimodal nociceptors (Hwang et al., 2007; Xiang et al., 2010). Class II neurons are touch sensitive, but a specific functional role has not been resolved (Tsubouchi et al., 2012).

Multiple transcription factors are known to play a role in sensory neuron dendrite morphogenesis (Moore et al., 2002; Grueber et al., 2003; Li et al., 2004; Sugimura et al., 2004; Kim et al., 2006; Hattori et al., 2007; Jinushi-Nakao et al., 2007; Crozatier and Vincent, 2008; Ye et al., 2011; Ferreira et al., 2014). The transcriptional regulator Cut is a key factor in the establishment of diverse da neuron morphologies (Grueber et al., 2003; Jinushi-Nakao et al., 2007; Iyer et al., 2013). Cut promotes the development of class II, III and IV da neurons with larger territories and complex branching patterns. By contrast, class I, dbd and dmd1 neurons with smaller fields and less complex morphologies lack detectable Cut expression and do not require Cut for dendritic morphogenesis. Cut is expressed at different levels in different da neuron classes. These levels help to specify the development of distinct neuronal morphologies (Grueber et al., 2003). Cut expression is promoted by the Longitudinals lacking (Lola) transcription factor in class IV neurons (Ferreira et al., 2014), but it is still not known how distinct Cut levels are established in different cell types and, more generally, how class-specific expression patterns of transcriptional regulators are generated to diversify sensory neuron morphology.

We addressed these questions by examining links between Cut and the transcription factors Pdm1/2, Scalloped and Vestigial in sensory neuron subset diversification. Our results identify a transcriptional pathway for diversification of sensory neuron morphology via suppression of distinct alternative morphological states.

The homeodomain transcription factor Cut is expressed in class II (low expression), III (high expression) and IV (intermediate expression) da neurons. Prior studies of Cut in da neuron dendritic morphogenesis identified two very distinct phenotypes in cut null mutant MARCM clones. Depending on the specific neuron examined, cut clones either showed simplified dendritic terminal branching or a more severe dendritic growth defect (Grueber et al., 2003). We examined the basis for these qualitatively distinct phenotypes. Neurons that showed severe growth defects included the class II and class III neurons ddaA, ddaB, ddaF and ldaB in the dorsal and lateral clusters (Fig. 1B-F, Fig.2G-K, Fig. S1A,B). Mutant neurons either extended a single primary dendrite that ended in a compact, dense arborization a short distance from the cell body, or adopted a bipolar dendrite morphology (Fig. 1B-D). Stunted dendrites often targeted to a nearby chordotonal organ or nerve (Fig. 1F,F′, Fig. S1B′).

Axon morphology and position in the neuropil are also distinguishing features of sensory cell types. Cut⁺ touch-sensing and nociceptive neurons (classes II, III and IV) terminate in the ventral neuropil, whereas Cut⁻ proprioceptive neurons (class I and dbd) terminate more dorsally (Merritt and Whitington, 1995; Schrader and Merritt, 2000; Grueber et al., 2007). We assessed the positions of afferent terminals relative to Fasciclin 2 (Fas2)-labeled ventromedial (VM) or dorsomedial (DM) fascicles. In contrast to the targeting of wild-type neurons to the VM fascicle, the terminals of a subset of cut mutant MARCM clones were repositioned to the DM fascicle (n=20; Fig. 1G-J). Notably, axon mistargeting to the DM fascicle was specific to the neurons (ddaA, ddaB, ddaF and ldaB) that also showed severe dendritic growth phenotypes upon loss of Cut (Fig. S1C,D). These data suggest that Cut is required to repress morphological characteristics of proprioceptive neurons in a subset of dorsal cluster touch-sensing neurons.

Cut has been shown to repress Pdm1 [Nubbin (Nub) – FlyBase] in embryonic sensory neurons (Brewster et al., 2001), prompting us to study this factor and the closely related transcription factor Pdm2 (collectively Pdm1/2) in more detail. Pdm1/2 are normally co-expressed in dbd and dmd1 sensory neurons, both of which are Cut⁻ and predicted to function as proprioceptors (Billin et al., 1991; Dick et al., 1991; Brewster et al., 2001; Hughes and Thomas, 2007). dbd resides along a connective tissue strand and extends dendrites in a simple bipolar fashion to span each segment (Schrader and Merritt, 2007). We found that dmd1 resides on the body wall and extends a compact dendrite bundle away from the body wall to the intersegmental nerve (ISN) (Fig. 2A,B). Thus, in contrast to da neurons, Pdm1/2⁺ neurons extend simple dendrites that are not associated with an epidermal substrate. dbd and dmd1 axons both terminate near the DM fascicle, consistent with a proprioceptive function, but arrive via distinct paths, with dbd axons traveling through the dorsal neuropil (Schrader and Merritt, 2000) and dmd1 axons taking a ventral-to-dorsal route (Fig. 2C,D). Thus, the normal dendritic and axonal projections of Pdm1/2-expressing dbd and dmd1 resembled those of the transformed cut clones described above.

Using antibodies against Pdm1 and Pdm2, we confirmed that Pdm1/2 expression was expanded in cut mutant embryos to include extra da neurons, which we assigned by position as ddaA, ddaB and ddaF (Fig. 2E,F). To determine whether Cut represses Pdm1/2 expression cell-autonomously we assessed Pdm1/2 expression in cut MARCM clones. We found strong expression of Pdm1/2 in cut^c145 clones of ddaA, ddaB and ddaF (Fig. 2G,H,L, Fig. S1E) but not in class IV or class I neurons (Fig. S1F-H). Thus, ectopic Pdm1/2 expression was observed only in neurons that showed dendritic or axonal transformations. The strong correlation is exemplified by phenotypes observed for the lateral class III ldaB neuron. For this cell, the dendritic transformation phenotype was observed in only a subset of MARCM clones (transformation n=12/15), and ectopic expression of Pdm1/2 was observed only in transformed clones (Fig. 2I-L). Ectopic expression of Cut was sufficient to suppress Pdm1/2 expression (Brewster et al., 2001) and promote ectopic dendritic branching in dmd1 and dbd neurons (Fig. 2M-P). Together, these data point to a role for Cut in suppressing Pdm1/2 expression in dorsal/lateral class II and III neurons.

We next examined the role of Pdm1/2 in specifying dbd and dmd1 morphology. Mutation of either pdm1 or pdm2 individually caused no morphological defects and thus each is separately dispensable for dendritogenesis (Fig. S2A,B), consistent with similar redundant function in CNS neuroblasts (Yeo et al., 1995; Grosskortenhaus et al., 2006). We characterized a lethal allele of pdm1, nub^R5, generated by imprecise P-element excision (Terriente et al., 2008), and found that homozygous embryos and MARCM clones lacked detectable expression of both Pdm1 and Pdm2, indicating that this deletion is likely to be a Pdm1/2 protein null (Fig. S2C). MARCM analysis revealed a dendritic overgrowth phenotype in nub^R5 dmd1 and dbd clones. Most dmd1 clones projected dendrites across the epidermis rather than the ISN (Fig. 3A-D). Most dbd clones retained their longitudinal dendrites, but sprouted ectopic branches that grew along the epidermis (Fig. 3E-H). We did not observe Cut immunoreactivity in dmd1 or dbd nub^R5 clones, indicating that Pdm1/2 does not reciprocally repress Cut (Fig. S2D,E). ddaC nub^R5 clones (class IV, normally Pdm1/2⁻) showed reduced growth (n=8; Fig. S2F,G), although we noted no obvious defects in class III arborization patterns (n=3). Thus, one or more genes within the nub^R5 deficiency might promote complex dendritic branching in class IV da neurons.

To narrow the region of the nub^R5 deficiency responsible for the overgrowth phenotypes we analyzed dbd morphology in nub^R5/Df(2L)ED773 animals. This combination of deficiencies eliminates Pdm1 and Pdm2 expression and putatively disrupts only two other genes in the region, RNA and export factor binding protein 2 (Ref2; coding region deleted) and bunched (bun; loss of all upstream sequence and part of its 5′ sequence). We visualized dbd morphology using live imaging in surviving early first instar larvae labeled with clh8-Gal4, UAS-mCD8::GFP and observed abnormal morphologies with excessive dendritic growth and branching (n=6; Fig. 3I-L, Fig. S2H), consistent with the overgrowth phenotypes seen in dbd nub^R5 MARCM clones.

In contrast to the overgrowth and targeting defects observed in dendrites, axons in nub^R5 dmd1 and dbd MARCM clones had grossly normal morphologies and retained their terminations near the DM fascicle (Fig. S2I-K). These loss-of-function studies, together with misexpression experiments presented below, point to a role for Pdm1/2 in restricting dendritic growth, but not in directing axonal targeting, of dmd1 and dbd neurons.

To test whether Pdm1/2 expression is sufficient to restrict dendritic growth we drove expression of each gene with Gal4-109(2)80 and resolved dendritic morphologies using the FLP-out method (Struhl and Basler, 1993). Expression of either gene in class II, III or IV neurons dramatically reduced dendritic growth and branching (Fig. 4, Fig. S3). Misexpression did not cause targeting of dendrites to the ISN or to connective tissue, as seen in wild-type Pdm1/2⁺ neurons. Thus, although Pdm1/2 can strongly suppress growth and branching, Pdm1/2 misexpression is not able to shift dendrite substrate preference or terminal targeting. Growth and targeting might be under separate transcriptional control in Pdm1/2-expressing neurons, or the timing of Pdm1/2 expression might be critical for dendritic targeting.

The above data demonstrate a role for Cut in directing dorsal cluster class II and III da neurons toward a complex dendritic morphology via suppression of a default Pdm1/2-dependent program of proprioceptor morphogenesis. Diversification of class II and class III sensory terminal dendrites depends at least in part on differential Cut expression levels, with high levels in class III neurons and low levels in class II neurons (Grueber et al., 2003). We explored the basis of differential Cut expression as a route to further dendritic diversification. From a collection of GFP trap lines (Morin et al., 2001; Buszczak et al., 2007; Quinones-Coello et al., 2007) we identified prominent expression of the TEAD/TEF-1 transcription factor Scalloped (Sd) (Campbell et al., 1992) in the class II neuron ddaB and the class III neuron ddaF in third instar larvae (Fig. 5A). Very low GFP expression could occasionally be observed in the class I neuron ddaE and in the class III neuron ddaA. We used an antibody against Sd that had previously revealed its expression in sensory neurons of the embryonic PNS (Guss et al., 2013), and confirmed Sd expression in ddaB and ddaF neurons in third instar larvae consistent with the GFP trap (Fig. 5B). Expression of Sd in a subset of Cut⁺ da neurons and the interaction between Sd and Cut in wing development (Morcillo et al., 1996) led us to examine a possible role in dendritic diversification.

To examine whether Sd regulates sensory neuron morphology, we generated homozygous mutant sd MARCM clones. We did not observe any obvious phenotypes for class I or IV neurons. We found no clones with a typical ddaB morphology (0/90). By contrast, among wild-type clones ddaB made up 12% (14/119) of the total. These data suggested that Sd could be required for either the production, survival or proper differentiation of ddaB. Using the md-neuron marker E7-2-36-lacZ, we found no difference in the number of md neurons in the dorsal cluster in sd^ΔB embryos [8.0±0.0 cells (±s.d.), n=53 clusters] compared with the sd^+/− control (8.0±0.24 cells, n=6 clusters). These data argue that Sd is not required for the survival or initial generation of ddaB, and raised the possibility of a role in morphogenesis.

We investigated whether sd mutant ddaB clones show morphological transformations. Examination of sd^ΔB clones revealed mutant neurons in the typical location of ddaB that had developed a highly branched morphology with the short fine terminal dendrite ‘spikes’ that are a hallmark of class III neurons. We inferred the identity of these clones by verifying the presence of ddaA as well as the absence of the ddaB-characteristic dendritic arbor using HRP labeling (Fig. S4A,B). Quantification of the dendrites of these clones revealed they were statistically indistinguishable from wild-type class III ddaA clones in each of the parameters examined (Fig. 5C-G, Fig. S4E). Relative to ddaB wild-type clones, sd clones showed dramatically increased terminal branching (Fig. 5F,G). Similar dendritic terminal phenotypes were observed with the sd^ΔC allele (n=3). A modest, but significant increase in total dendritic branch points was also observed for ddaF, which also expresses Sd, without a change in main arbor length (Fig. 5F,G). We did not observe transformations of other class II neurons to a class III-like morphology (ldaA, n=8; vdaA, n=4; Fig. S4C,D).

We conclude that Sd is required in ddaB to repress elaboration of a class III-like morphology and in ddaF to limit terminal branching. Our data also suggest that different genetic programs specify class II neuron morphology in different regions of the body wall.

One hallmark of class II versus class III identity is the expression level of Cut. Class II neurons express low levels of Cut, whereas class III neurons express high levels that promote extensive terminal branching. To examine whether sd class II ddaB clones with altered morphology also show altered Cut expression we quantified the levels of Cut immunoreactivity in sd^−/− ddaB clones relative to adjacent sd^+/− class III ddaA neurons. As controls, we quantified Cut levels in wild-type ddaB clones relative to their neighboring ddaA neurons, as well as the levels of Cut in ddaB and ddaA neurons in sd^+/− larvae. In both controls, levels of Cut immunoreactivity in ddaB neurons were, on average, about half those of ddaA neurons (mean Cut^ddaB:Cut^ddaA ratio=0.41; Fig. 6A,I). By contrast, sd mutant ddaB clones showed high levels of Cut that were indistinguishable from those of ddaA (mean Cut^ddaB:Cut^ddaA ratio=0.98; Fig. 6B,I). These results suggest that Sd is required in ddaB to limit the Cut expression level.

Either of two transcriptional intermediary factors, Vestigial (Vg) or Yorkie (Yki), interact with Sd in different tissues (Halder et al., 1998; Paumard-Rigal et al., 1998; Simmonds et al., 1998; Goulev et al., 2008; Wu et al., 2008; Zhang et al., 2008). yki MARCM clones showed normal ddaB dendritic morphology (n=6; Fig. S4F). By contrast, vg ddaB clones showed extra terminal branching and short fine terminal processes similar to class III neurons and sd mutant ddaB clones (n=5; Fig. 6C,D,F-G). As with sd mutants, Cut immunoreactivity in vg ddaB clones was significantly increased (Fig. 6E,H,J). In addition, we found that a Vg::mCitrine fusion protein is expressed in ddaB and ddaF and that anti-Vg immunolabeling overlapped with Sd::GFP expression in those neurons (Fig. S4G,H). These results suggest that Sd and Vg function together in ddaB to repress class III characteristics, including a high Cut expression level and extensive terminal branching.

To examine whether Sd is sufficient to suppress branching in high Cut-expressing class III neurons we performed ectopic expression experiments. We examined the consequences of ectopic sd expression under the control of the tubulin (αTub84B) promoter (tub>HA-sd). tub>HA-sd rescues lethality and wing phenotypes of sd mutants and thus can functionally substitute for sd. Furthermore, in an sd⁺ background tub>HA-sd does not cause lethality or generate the overexpression phenotypes caused by UAS-based overexpression, such as wing scalloping (M. Zecca, personal communication).

We again visualized individual neurons using the FLP-out technique. Class III ddaF neurons in a tub>HA-sd background showed highly simplified dendrites characterized by the loss of fine terminal branches (Fig. 7A,B,F,I), while the main arbor length of these cells was unchanged (Fig. S4I). The class III neuron ddaA also showed a significant decrease in branch points per total dendrite length and reduced branch number, whereas ddaB showed no dendritic defects (Fig. 7D,E,G,H, Fig. S4J-K). Thus, forced Sd expression can reduce dendritic branching of class III neurons. These data also suggest that either the level or the timing of endogenous Sd expression is not sufficient to strongly suppress terminal branching in ddaF. UAS-cut expression provided significant rescue of ddaF branch points per length in a tub>HA-sd background but did not restore arbors to wild-type morphology and did not restore total branch number in this cell (Fig. 7C,F,I). Altogether, the partial restoration of dendritic complexity in tub>HA-sd by UAS-cut, together with our loss-of-function data, suggest that Sd regulates dendritic morphology in part by specifying low levels of Cut expression. These data also raise the possibility that Sd can interfere with dendritic branching programs that operate downstream of, or in parallel to, Cut.

We have studied how three distinct somatosensory neuron morphologies are established in Drosophila. Together, our results reveal a repressive strategy for the diversification of dendritic arbor morphology involving the transcription factors Cut, Pdm1/2, Sd and Vg (Fig. 8).

Cut/Cux transcription factors have conserved roles in dendritic morphogenesis, including control of dendritic elaboration of da neurons (Grueber et al., 2003; Jinushi-Nakao et al., 2007), dendritic targeting of Drosophila olfactory projection neurons (Komiyama and Luo, 2007), and control of dendritic morphogenesis of upper layer cortical neurons in vertebrates (Cubelos et al., 2010). Previous studies showing that Cut levels impact class-specific complexity of da neuron dendritic arbors (Grueber et al., 2003) raised several questions, including how Cut activity establishes divergent sensory neuron morphologies, how different levels of Cut are specified, and whether Cut controls axonal targeting as well as dendritic branching in da neurons. Our data begin to address these questions by characterizing regulatory interactions with Pdm1/2, Sd and Vg, and by identifying a role for Cut in da neuron axon targeting (Fig. 8).

Loss-of-function and gain-of-function experiments indicate that Pdm1 and Pdm2 inhibit dendritic growth. Cut represses the action of Pdm1/2 and, by extension, derepresses a program for extensive dendritic elaboration (Fig. 8). Complementing level-dependent functions for Cut, this function of Cut is likely to be level independent as both low (class II) and high (class III) levels appear sufficient to repress Pdm1/2 expression. We suggest that a minimum level of Cut is required for Pdm1/2 repression in dorsal cluster class II and III neurons, allowing for the establishment of an expansive epidermal territory. Higher levels of Cut can promote increased higher-order branching and further diversify sensory neuron morphology. POU domain transcription factors are important regulators of neuronal morphogenesis in both vertebrate and invertebrate sensory systems (Komiyama et al., 2003; Badea et al., 2009) and it would be interesting to determine whether repression of alternative or default states is a more general mechanism by which transcriptional programs involving POU proteins generate diversity in neuronal morphology.

In the fly somatosensory system, sensory neuron subtype is correlated with the morphology of dendrites in the periphery and the positioning of sensory afferents centrally. While the mechanisms that differentiate dendrite versus axon development are beginning to be uncovered (Ye et al., 2007, 2011; Satoh et al., 2008; Zheng et al., 2008), the mechanisms that coordinate md neuron dendrite and axon morphology have not been established. Our results suggest that Cut fulfills such a role in a subset of somatosensory neurons. Loss of Cut led to the development of proprioceptive dendritic features and to a shift in axon targeting to regions where proprioceptive neurons terminate. Still unknown are the mechanisms by which Cut regulates this switch, as Pdm1/2 appear not to affect axon terminal positioning of da neurons. Prior studies have shown that axon positioning along the dorsoventral axis is controlled in part by Semaphorin-Plexin signaling (Zlatic et al., 2009), and it will be important to examine whether Cut directly or indirectly regulates this pathway to control axonal targeting.

Different levels of transcriptional regulators can affect neuronal morphogenesis (Grueber et al., 2003; Chen et al., 2006), but how specific levels are established is not understood. We found that Sd and Vg establish low Cut levels and repress a more highly branched class III morphology in ddaB (Fig. 8). Sd and Vg are also expressed in ddaF, which develops a class III-like branching pattern, and our loss-of-function experiments suggest that Sd has a modest effect on branching in this cell as well. Our results also raise the possibility (not schematized in Fig. 8) that Sd might also act via a parallel pathway to set an upper limit on the branch-promoting activity of Cut, perhaps by blocking pathways that are required downstream of Cut for dendrogenic activity. In C. elegans sensory neuron specification, the AHR-1 transcription factor operates by controlling both a transcription factor and a subset of its downstream targets (Smith et al., 2013). Dual-level regulatory strategies might add robustness and range to transcription factor encoding of specific morphological features or identity. The role of class III terminal branches in mediating responsiveness to gentle touch (Tsubouchi et al., 2012; Bagley et al., 2014) raises the possibility that Sd regulation of morphology could fine-tune mechanoreceptor sensitivity.

Sd and Cut interact during wing development (Morcillo et al., 1996), and during adult development Vg suppresses Cut levels in a subset of muscle precursors (Sudarsan et al., 2001). Sd and Vg therefore operate in several contexts to regulate Cut levels and promote cell diversification, and our results identify roles in differentiating neurons in the control of dendrite morphology. Similarly, in C. elegans, the sd homolog egl-44 suppresses late stage ectopic branching in multidendritic PVD neurons (Smith et al., 2010). Given these roles in dendritic branching in invertebrate neurons, and the broad expression of Tead1 (the mouse homolog of Sd) in the mouse brain (Lein et al., 2007), it will be interesting to examine whether TEAD transcription factors also regulate dendritic branching in vertebrate neurons.

Our findings provide an example of how linked repressive interactions among transcriptional regulators can diversify dendritic morphology. In Drosophila sensory neurons, it appears that a broad competence to produce dbd/dmd1-type morphologies in dorsal regions of the larva is restricted by the action of Cut (Fig. 8). This switch generates at least two types of sensory neurons: Cut⁺/Pdm⁻ (which perform tactile functions) and Cut⁻/Pdm⁺ (which function as proprioceptors) (Fig. 8). Further morphological diversification of Cut⁺/Pdm⁻ neurons is achieved by modulation of Cut expression levels (Fig. 8). The partial, rather than complete, repression of Cut levels in ddaB becomes important in this context, since it allows for the maintenance of Pdm1/2 repression, as well as the development of a third distinct Cut-dependent morphology (Fig. 8). High-level Cut expression might represent a default state for the dorsal cluster class II neuron that is refined by Sd and Vg (Fig. 8).

Repression of alternative identities and morphologies by transcriptional regulators is an important driver of neuronal diversification during the development of sensory and motor systems (Jung et al., 2010; Li et al., 2013; Philippidou and Dasen, 2013; Gordon and Hobert, 2015). Transcription factors may suppress alternative programs of differentiation and promote neuronal diversity by direct repression of other transcription factors, repression of their downstream target genes, or by competitive binding of common co-factors to block transcription factor function (Smith et al., 2013; Borromeo et al., 2014; Gordon and Hobert, 2015). Differential expression levels of the same transcription factor can also drive distinct morphology, identity and connectivity to further increase diversity in the nervous system (Grueber et al., 2003; Chen et al., 2006; Dasen et al., 2008; Smith et al., 2013). It has been proposed that such transformations from one cell type, or cell characteristic, to another by mutations in terminal selector genes might provide a basis for the evolutionary diversification of cell types in the nervous system (Arlotta and Hobert, 2015). One speculative scenario is that, in the Drosophila PNS, the suppression of default programs in subsets of equivalent neurons by the action of transcriptional regulators provided a mechanism for the evolutionary diversification of somatosensory cell types without disruption of crucial existing modalities.

Animals were reared using standard methods. We used pdm1-Gal4 [nub-Gal4 (Calleja et al., 2000)], clh8-Gal4 (Hughes and Thomas, 2007), GMR11F05-Gal4 (Pfeiffer et al., 2008; Li et al., 2014), 109(2)80-Gal4 (Gao et al., 1999) and clh201-Gal4 (Hughes and Thomas, 2007). tubP>HA-sd and FRT42D vg^83b27R were provided by Drs M. Zecca and G. Struhl (Columbia University), vg::mCitrine by Dr G. Struhl (unpublished), and sd^ΔB and sd^ΔC alleles were from Dr J. Jiang (UT Southwestern) (Zhang et al., 2008). FRT42D yki^B5 was from Dr Duojia Pan (Johns Hopkins). cut^c145 and cut^db3 are null alleles (Jack, 1985; Blochlinger et al., 1988). cut^db3 was combined with the md neuron marker E7-2-36 lacZ for studies of embryonic stages. The nub^R5 deletion is the result of P-element excision (Terriente et al., 2008) and was provided by Dr F. Diaz-Benjumea (Universidad Autónoma de Madrid, Spain). PCR mapping revealed that the nub^R5 deletion removes ∼224 kb of genomic material spanning 11 protein-coding regions, which is about half the number of genes deleted in Df(2L)ED773, which also deletes both pdm1 and pdm2 (Ryder et al., 2004, 2007; Grosskortenhaus et al., 2006). Other stocks were Df(2L)ED773 (Bloomington Stock Center), pdm2^E46 (Bloomington Stock Center), Sd::GFP (CA07575; Buszczak et al., 2007), full-length UAS-cut (Ludlow et al., 1996; Grueber et al., 2003), UAS-pdm1 (Neumann and Cohen, 1998), UAS-pdm2 (Grosskortenhaus et al., 2006), UAS-mCD8::GFP and UAS-mCD8::Cherry (Bloomington Stock Center). Animals were analyzed as third instar larvae, except when noted (i.e. embryos or first instar).

For MARCM experiments (Lee and Luo, 1999) the following fly stocks were used: (1) cut^c145 FRT19A/FM7, (2) sd^ΔB FRT19A, (3) FRT19A;;ry, (4) tub-Gal80, hs-FLP, FRT19A; 109(2)80-Gal4, UAS-mcD8::GFP, (5) ywhs-FLP¹²²; nub^R5 FRT40A/CyO, (6) nub¹ FRT40A/CyO, (7) FRT40A/CyO, (8) hs-FLP, elav-Gal4, UAS-mCD8::GFP; tub-Gal80 FRT 40A, (9) FRT42D yki^B5, (10) FRT42D vg^83b27R, (11) w; FRT 42D, (12) hs-FLP, elav-Gal4, UAS-mCD8::GFP; FRT 42D tub-Gal80, (13) tub-Gal80 FRT 40A/CyO; GMR11F05-Gal4, UAS-mCD8::GFP, (14) ywhs-FLP¹²²; FRT 40A/CyO. MARCM clones were generated as previously described (Matthews et al., 2007). For overexpression experiments, we crossed hs-FLP; 109(2)80-Gal4; UAS>rCD2>mCD8::GFP to either w¹¹¹⁸, w¹¹¹⁸; UAS-pdm1/CyO, Dfd-YFP, w¹¹¹⁸; UAS-pdm2/CyO, Dfd-YFP or tub>HA-sd. FLP-out clones were generated using a 30 min heat shock at 38°C to induce mosaic expression of rCD2 and mCD8::GFP.

Immunohistochemistry was performed according to published methods (Grueber et al., 2002). Antibodies used were rabbit anti-Vg (1:20; Halder et al., 1998; courtesy of S. Carroll, University of Wisconsin), guinea pig anti-Sd (1:500; Guss et al., 2013), guinea pig anti-Vg (1:10; Zecca and Struhl, 2010; courtesy of G. Struhl), mouse anti-Cut (1:20; DSHB #2B10), goat anti-HRP (1:200; Jackson ImmunoResearch #123-005-021), chicken anti-GFP (1:1000; Abcam ab13970), rabbit anti-β-gal (1:200; Cappell #55976), rat anti-ELAV (1:10, DSHB #7E8A10, mouse anti-Fas2 (1:10; DSHB #1D4); rabbit anti-Pdm1 (1:1000; Yeo, 1995), and rat anti-Pdm2 (1:10; Grosskortenhaus et al., 2006). Secondary antibodies (Jackson ImmunoResearch) were used at 1:200. Following antibody labeling, animals were dehydrated in an ethanol series, cleared in xylenes, and mounted in DPX (Electron Microscopy Sciences) as described (Grueber et al., 2002). For the experiments depicted in Fig. S2I-K, animals were mounted in Vectashield (Vector Laboratories).

Imaging was performed on a Zeiss 510 Meta confocal microscope with LSM software (Zeiss) using 40× Plan Neofluar and 63× objectives for all images except those in Fig. 3I-L and Fig. S2H-J, which were acquired using a 3i Everest spinning disk confocal microscope with Slidebook 6.0 (3i) imaging software and a Zeiss Plan-Apochromat 40× oil objective. Arbors were scanned as single frames or as partially overlapping frames that were assembled in Photoshop CS2 (Adobe Systems) or Neurolucida (MBF Bioscience). For quantitative analysis, neurons were traced as confocal stacks using Neurolucida and quantified using Neurolucida Explorer. Whole image adjustments of brightness and contrast were applied to aid in visualization of fine dendritic processes in figure images.

Quantification of Cut levels was performed similarly to prior studies (Grueber et al., 2003). Briefly, ddaA and ddaB Cut-labeled nuclei were identified in confocal stacks. Mean pixel intensity values for each nucleus were obtained from each successive confocal section in which the nucleus was visible using the Measure command in Fiji (Schindelin et al., 2012). The highest mean pixel value obtained for each analyzed nucleus was used to compute ddaB:ddaA ratios. Confocal settings were identical among the different imaging sessions for data collected for comparison between groups. Cut values were quantified from raw values for each image without adjustment of brightness, contrast, or levels. To aid visualization and comparison in figure panels, levels were adjusted uniformly across some images from (min-max) of 20-148 to 10-232 (Fig. 6A′) and 11-95 to 10-229 (Fig. 6B′). These changes were made following quantification.

Classification of nub^R5 phenotypes: dmd1 clones were classified as ‘normal’ if all dendrites project to the ISN; ‘mild’ if some dendrites project to the ISN with some dendrites that do not grow towards the target; or ‘severe’ if there was no ISN targeting and/or substantial growth of dendrites on the epidermis. dbd clones were classified as ‘normal’ if normal unbranched bipolar morphology is maintained; ‘mild’ if there is abnormal branching or mistargeting of at least one of the main dendrites; or ‘severe’ if there is an additional, branched dorsal dendrite and/or a lack of bipolar morphology.

Statistical analysis was performed in R (R Project for Statistical Computing) and GraphPad Prism. Data were tested for normality using a Shapiro-Wilk test. Nonparametric analyses (Kruskal–Wallis with Dunn's multiple comparisons test) were used for datasets that deviated from a normal distribution as noted (Fig. S4E, Fig. 6I, Fig. 7I). For all other datasets statistical significance was determined using a Student's t-test with Welch's correction for all comparisons between two groups and one-way ANOVA with Tukey's post-hoc test for all comparisons with three or more groups. n values refer to individual cells (i.e. MARCM or FLP-out clones) except where noted in the text. No statistical methods were used to predetermine sample size. The sample sizes used are consistent with those found in similar studies of da neuron morphogenesis. Animals were assigned to groups based on genotype, and researchers were not blind to conditions. Clones were excluded from analysis if arbors were damaged during processing. Datasets presented as box and whisker plots are as follows: quartiles Q1-Q3 (25%-75%, box), median (thick line), whiskers span minimum to maximum values. All individual data points are displayed as open circles. Multiplicity-adjusted P-values are given for all ANOVA/Tukey results. In figures, *P<0.05, **P<0.01 and ***P<0.001.

We thank Dr M. Zecca for generously sharing fly stocks prior to publication; Drs J. Bell, S. Carroll, W. Chia, S. Cohen, F. Diaz-Benjumea, C. Doe, K. Guss, J. Jiang, B. Ohlstein and G. Struhl for generous sharing of fly stocks and antibodies; Dr J. Dodd for comments on the manuscript; N. Ghani for input on branching analysis; and Dr O. Hobert and members of the W.B.G. lab for discussions. M.M.C. thanks Dr M. Freeman for support during preparation of the manuscript. The anti-Cut, anti-ELAV, and anti-Fas2 antibodies were obtained from the Developmental Studies Hybridoma Bank (DSHB), created for the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242, USA.