Nuclear Receptor Unfulfilled Regulates Axonal Guidance and Cell Identity of Drosophila Mushroom Body Neurons

Nuclear receptors (NRs) comprise a family of ligand-regulated transcription factors that control diverse critical biological processes including various aspects of brain development. Eighteen NR genes exist in the Drosophila genome. To explore their roles in brain development, we knocked down individual NRs through the development of the mushroom bodies (MBs) by targeted RNAi. Besides recapitulating the known MB phenotypes for three NRs, we found that unfulfilled (unf), an ortholog of human photoreceptor specific nuclear receptor (PNR), regulates axonal morphogenesis and neuronal subtype identity. The adult MBs develop through remodeling of γ neurons plus de-novo elaboration of both α′/β′ and α/β neurons. Notably, unf is largely dispensable for the initial elaboration of γ neurons, but plays an essential role in their re-extension of axons after pruning during early metamorphosis. The subsequently derived MB neuron types also require unf for extension of axons beyond the terminus of the pruned bundle. Tracing single axons revealed misrouting rather than simple truncation. Further, silencing unf in single-cell clones elicited misguidance of axons in otherwise unperturbed MBs. Such axon guidance defects may occur as MB neurons partially lose their subtype identity, as evidenced by suppression of various MB subtype markers in unf knockdown MBs. In sum, unf governs axonal morphogenesis of multiple MB neuron types, possibly through regulating neuronal subtype identity.


Introduction
The brain consists of neurons that are wired in specific patterns, and establishing a complex brain involves multiple tightly regulated developmental processes. In Drosophila, it starts with birth of neuroblasts (Nbs) with specific fates that are largely acquired through spatial patterning [1]. The Nbs then proliferate to produce multiple neuron types often in an invariant sequence [2,3]. Post-mitotic neurons subsequently undergo extensive morphogenesis and some neurons remodel to form the circuitry [4,5]. Although these basic processes are known, the detailed mechanisms that govern each step of brain development remain only partially resolved. Identifying more genes required for the various aspects of brain development is essential for elucidating further how the complex brain develops.
Nuclear receptors (NRs) are ligand-regulated transcription factors that play important roles in key metabolic and developmental pathways, including lipid and glucose homeostasis, aging, and cell fate determination [6,7,8,9,10,11,12,13,14,15,16]. NRs have also been shown to govern diverse aspects of neural development, such as the maintenance of neuronal precursors [17,18,19], neuronal cell death [20], axon guidance [21,22,23] and neuronal remodeling [24,25]. Furthermore, mutations in NRs have been implicated in several neurodegenerative diseases [11,15,26,27,28,29,30,31,32]. Determining the functions of NRs will provide ample opportunities for better understanding brain development and neuron degeneration. NR genes are highly conserved across the animal kingdom. The 48 NR genes identified in mammals can be categorized into six subfamilies according to their protein structure similarity [33]. The neural functions of most NRs have not been explored at all. Compared to the 48 NR genes in mammals, the Drosophila genome contains only 18 NR genes, though all six NR subfamilies are represented [33]. The smaller number of NR genes makes it easy to survey NR functions in brain development. Four fly NRs have been shown to regulate various aspects of neural development. Seven-up (svp), which encodes a homolog of human COUP-TF orphan NR, represses the expression of a temporal identity gene hunchback (hb) in embryonic neuroblasts after the first mitosis to ensure the subsequently produced neurons acquire proper cell fates [8]. Moreover, in the late larval stage, a burst expression of svp is required for neuroblasts to exist the cell cycle or undergo apoptosis [34]. Tailless (tll), the Drosophila homolog of human Tlx orphan NR, is robustly expressed in the larval brain in certain neuroblasts and ganglion mother cells (GMCs) to promote cell cycling and prevent apoptosis [17]. Ecdysone receptor (EcR) and its heterodimeric partner ultraspiracle (usp) work together to regulate axonal and dendritic remodeling during metamorphosis [24,25,35,36,37]. EcR have also been shown to regulate programmed neuronal cell death [20,38] and secondary arbors of adult-specific neurons [39]. However, the functions of the remaining 14 NRs in neural development have not been studied.
The Drosophila mushroom body (MB), the olfactory learning/ memory center, has been shown as an excellent model to study gene functions in neural development (e.g. [24,40,41,42]), and thus could also be good for studying NR functions. The adult MB consists of three major types of neurons-c, a9/b9 and a/b [43]. These neuron types are specified according to their date of birth by temporal identity factors, such as chinmo [44]. The c neurons are born before the mid-3rd instar larval stage [around 4 days after larval hatching (ALH)]; their axons initially form bifurcated branches in the larval brain. Later, these axons are pruned through fragmentation and glia engulfment at early pupal stage, followed by the re-extension of adult-specific axons that only elaborate horizontally toward the midline of the adult brain [43,45]. The a9/b9 neurons are born between the mid-3rd instar and puparium formation, and a/b neurons are born at the pupal stage. Both a/b and a9/b9 axons form bifurcated branches without remodeling during development, and thus may only be functional in the adult brain [43]. As to roles of NRs in MB development, EcR, USP and TLL have been shown playing important roles in regulating MB remodeling and neurogenesis, respectively (see below) [17,24].
In this paper, we systematically silenced each of the 18 Drosophila NR genes in MBs using miRNA-based RNA interference [46,47]. In addition to EcR, usp, and tll, we isolated unfulfilled (unf), the fly NR2E3 subfamily ortholog of C. elegans fax-1 and human photoreceptor-specific nuclear receptor (PNR) [48,49]. unf has a novel function in MB development, and unf knockdown caused a severe MB lobe extension defect. Further analysis revealed that unf was required for proper axonal guidance of all three major types of the MB neurons. Without unf, the MB axons were misguided and wandered around at the end of peduncle. Interestingly, unf was mainly required for adult-specific axonogenesis. In later-born adult-specific a9/b9 and a/b neurons, unf played a critical role in initial neurite extension; in contrast, for early-born c neurons, unf was largely dispensable to the establishment of larval projections, but absolutely required for their re-elaboration of axons during early metamorphosis.
We also found that the expression of Trio, a guanine nucleotide exchange factor important for the MB axonal guidance [50], as well as several subtype-specific markers were significantly reduced in the unf knockdown adult MB. This result implies that unf is required for MB neurons to acquire correct gene expression profile (i.e. cell identity) and that improper expression of multiple guidance molecules resulting from the loss of neuron identities might underlie the axonal defects induced by unf knockdown.

Silencing Individual NRs by miRNA Unveils a Role of Nuclear Receptor unf in MB Development
To study NR functions in fly brain development, we generated UAS-miRNA transgenic fly lines against each of the 18 nuclear receptor (NR) genes so far identified in the fly genome. We drove the miRNA transgenes using GAL4-OK107 [51] to knock down individual NRs through development of the MBs. We detected MB abnormalities when the miRNAs against ecdysone receptor (EcR), ultraspiracle (usp), tailless (tll), ftz transcription factor 1 (ftz-f1) or unfulfilled (unf) were induced (Table S1).
EcR and its heterodimeric partner usp have been shown to regulate the axonal pruning of MB c neurons [24], and tll was recently demonstrated to promote the efficient proliferation of the MB neuroblasts and ganglion mother cells [17]. Our miRNA screening results were consistent with the previous studies. Knocking down EcR or usp blocked pruning of MB c neurons, which maintained larval-type projections in the adult brain ( Fig. 1A-C). In contrast, silencing tll affected MB proliferation, resulting in a tiny MB purely consisting of early-type MB neurons (Fig. 1D,E). This recapitulates the tll mutant MB phenotype [17]. The consistency between the miRNA results and published data suggests that our miRNA approach worked efficiently in knocking down the endogenous nuclear receptors.
In addition to the nuclear receptors known to be required for MB development, we found that silencing ftz-f1 and unf caused abnormal MB morphologies. In the MB, distinct sets of MB axon bundles show different levels of FasII expression [52]; but silencing ftz-f1 elicited ectopic FasII-positive bundles in about 30% of MBs (Table S1; data not shown). Ectopic FasII bundles could result from axon guidance defects, misregulation of cell fate, or incomplete c axon pruning. Since the phenotype was subtle and the penetrance was low, we did not pursue analysis of ftz-f1 in this study. In contrast, silencing unf caused a much stronger MB phenotype (Table S1; Fig. 1F); the five MBcharacteristic axon lobes were missing in all unf knockdown MBs (Fig. 1F). While most axons might extend across the brain through the peduncle, they failed to form the lobes that normally project on the anterior surface of the brain hemisphere. Instead, the stereotyped MB lobes were replaced with a ball-like structure around the end of peduncle (Fig. 1F). The MBs were otherwise grossly normal, consisting of many cell bodies residing over the dendrite-based calyx that sometimes appears slightly larger than normal (data not shown).
The unf miRNA-a used in the initial screening targets two sites on the unf coding sequence, one at 289-310 nt and the other at 1130-1151 nt ( Fig. 2A). To ascribe the above phenotype to loss of unf, we first learned that the induction of this miRNA by GAL4-OK107 did effectively deplete unf protein in MB neurons (see below). We then tried to rule out off-target effects by generating unf miRNA-b to target an independent site at 1001-1022 nt ( Fig. 2A). GAL4-OK107-dependent induction of unf miRNA-b also led to a similar MB lobe extension defect, though the phenotype was weaker with some axons fully extended (Fig. 2B). To confirm that these phenotypes are indeed due to unf knockdown, we induced unf miRNA-b in flies heterozygous for a deficiency [Df(2R)ED2426] that covers the unf locus, and found that the MB axonal lobe defect became as severe as that caused by unf miRNA-a (Fig. 2C,D). Notably, when the unf miRNA-a was induced in the deficiency heterozygous background, the lobe phenotype was not enhanced (data not shown), suggesting that the abnormality caused by unf miRNA-a was a very strong unf loss-of-function phenotype. Taken together, these results show that unf is essential for the extension of the MB axonal lobes.

unf Is Mainly Required for the Formation of Adult-Specific MB Lobes
To determine the pathological mechanisms underlying the lobe defects, we followed unf knockdown MBs through development. We found that most unf knockdown MBs (88%, n = 75) were grossly normal at mid-3rd instar larval stage when the MB primarily consists of c neurons (Fig. 3A,B), although 12% of unf knockdown MBs had thinner dorsal lobes (n = 75; inset in Fig. 3B). This is in great contrast to the 100% lobe extension defect of unf knockdown adult MBs (e.g. Fig. 1F). This result was not due to the insufficient RNAi knockdown in the early larval stage, because GAL4-OK107 induced unf miRNA can knockdown UNF protein completely within 6 hr ALH (See below). The observations suggest that unf is mainly involved in the formation of the adult MBs.
The adult MBs develop through remodeling of c neurons plus de-novo elaboration of both a9/b9 and a/b neurons. These dynamic changes in the MB structure can be closely examined during early pupal development. In wild-type MBs, the bifurcated larval c lobes were completely pruned around 18 hr APF, and at the same time the a9/b9 lobes that were initially wrapped by the larval c lobes can be clearly seen (Fig. 3C,D). Around 24 hr APF when the adult-specific c lobes started to extend, the nascent a/b axonal bundles derived from the pupal-born a/b neurons can be detected with anti-FasII mAb 1D4 (Fig. 3E). By 48 hr APF, the adult-specific c lobe is fully extended and the a/b bundles become much thicker (Fig. 3F).
Examination of unf knockdown MBs through early pupal development revealed multiple abnormalities. First, mutant MBs at 6 hr APF displayed no obvious morphological defect (Fig. 3G). By contrast, at 18 hr APF when the larval c lobes were largely pruned, there did not exist a9/b9 lobes (Fig. 3H). Second, no axon has extended beyond the peduncle end at 24 hr APF, indicating defects in both c axon re-extension and a/b bundle formation (Fig. 3I). Third, by 48 hr APF, instead of seeing all five MB lobes, we detected a ball-like structure bulging around the terminus of the peduncle (Fig. 3J). It appears that MB axons were lost at the peduncle end. They elaborated locally, but failed to project into discrete domains based on their subtype identity. These observations reveal that 1) unf is largely dispensable for the initial larvalspecific axonal morphogenesis of the c neurons, but absolutely essential for the extension of the adult-specific c axons during remodeling; 2) unf is necessary for the de-novo formation of a9/b9 and a/b lobes.
unf Acts in Mature c Neurons to Govern Axon Re-Extension, while Supporting Initial Axonal Morphogenesis in Later Types of MB Neurons Given its pleiotropic functions in MB morphogenesis, we sought to determine the role of unf in c neuron remodeling independent of its effects on other MB neuronal morphogenetic processes. We selectively knocked down unf in mature larval c neurons using GAL4-201Y [53], resulting in missing adult c lobes despite presence of other MB lobes (Fig. 4A,B). Phenotypic analysis through development revealed normal pruning of the larvalspecific c lobes (Fig. 4C-F) but no formation of the adult c lobe (Fig. 4B). This result suggests that unf acts in mature c neurons to promote axon re-extension during MB remodeling. In addition, lacking the c lobe did not affect formation of MB a/b lobes (Fig. 4B).
Notably, the selective loss of the adult c lobe was evident even when unf was silenced in most, if not all, post-mitotic MB neurons via induction of unf miRNA-a using GAL4-MB247 (Fig. 4G,H). GAL4-MB247 drives UAS-transgene expression in MB neurons that have undergone extensive morphogenesis [44]. This again indicates the requirement for unf in mature c neurons, and further suggests that a9/b9 and a/b neurons need unf during their initial development. Consistent with these notions, silencing of unf in newborn neurons using asense-GAL4 [44] did not affect larval c axon projections (Fig. 5A,B) but drastically arrested the nascent a9/b9 (Fig. 5C,D) or a/b axon bundles (Fig. 5E,F) around the     peduncle. Taken together, unf acts in mature c neurons to govern axon re-extension while supporting axonal morphogenesis in newly derived a9/b9 and a/b neurons.

Direct Involvement of unf in Later MB Morphogenetic Processes
The serially-derived MB neurons, made concurrently by four MB progenitors, have undergone morphogenesis in sequence. It has been shown that the projections of specific MBs could be affected by the trajectories of others [50]. Notably, all MB axons stalled around the peduncle end when unf was depleted through development of the MBs, raising the possibility that some later morphogenetic defects may occur as a consequence of earlier axons' failings. The above observation that the a/b lobes could form normally in the absence of the adult c lobe (Fig. 4B,H) partially rules out such possibility. However, the a9/b9 lobes remained in those c-lobe-missing MBs (Fig. 4B,H) and might play an essential role in guiding the later derived a/b axons at the peduncle end, a critical choice point to all MB axons.
To demonstrate that unf is directly involved in the axonogenesis of a/b neurons, we sought to knock down unf after morphogenesis of c and a9/b9 neurons to selectively deplete unf from newborn a/b neurons. We controlled the timing of targeted RNAi using a temperature-sensitive GAL4 repressor, GAL80[ts] [54]. Organisms with the genotype of tubp-GAL80[ts],UAS-mCD8GFP/UAS-unf-miRNA-a;tubp-GAL80[ts]/+;GAL4-OK107/+ were initially cultured at a permissive temperature (18 degree Celsius) and shift to a restrictive temperature (29 degree Celsius) to induce the expression of unf miRNA-a at desired developmental stages. Besides labeling all five MB lobes by GAL4-OK107, we counterstained MB with anti-Trio antibody that strongly labels a/b lobes and weakly labels c lobe [52] to make the five MB lobes distinguishable. Induction of unf RNAi from mid-3 rd instar and prior to birth of most a9/b9 neurons disrupted all MB lobes (Fig. 6A,B). In contrast, shifting the temperature after adult eclosion left the gross MB morphology intact (Fig. 6C,D). Intriguingly, when the temperature was shifted around puparium formation after the production of a9/b9 neurons, the a9/b9 lobes were grossly normal but the FasII-positive a/b lobes were malformed (Fig. 6E,F). These results demonstrate that the unf knockdown phenotype in a/b neurons was not due to the defects in a9/b9 lobes. Thus, unf directly governs axonal morphogenesis in all three major types of MB neurons.
unf Regulates Axon Pathfinding of MB Neurons MB axons uniformly stalled around the peduncle end, a common choice point to the migrating axons, raising the possibility that unf may regulate axon pathfinding rather than simply promoting axon extension. To elucidate the mechanism(s) underlying the axonal lobe defect of the unf knockdown MBs, we followed single axons to determine their trajectories in the lobe defective MBs. Single MB neurons were labeled using flip-out strategy [55,56]. In wild-type MBs, the flip-out clones consistently projected their axons directly into the MB lobes (Fig. 7A). In contrast, the axonal processes of flipout clones in the unf knockdown MBs wandered around in the balllike truncated lobes (Fig. 7B). These wandering axons were apparently lost, and often made unusual back turns or loops ( Fig. 7C-F). These phenomena ascribe the lobe defect of the unf knockdown MB to misrouting of axons.
We further determined if unf is cell-autonomously required in individual MB neurons for proper axon guidance, by knocking down unf in single MB neurons. We generated isolated single-cell clones of c neurons by MARCM (Mosaic Analysis with a Repressible Cell Marker) [57]. Such uniquely labeled cells were the only MB neurons that lack the GAL4 repressor, GAL80, and thus actively expressed unf miRNA-a. This allowed us to knock down unf in single MB neurons within otherwise unperturbed MBs. Notably, the lone unf knockdown neurons exhibited abnormal axon trajectories in the grossly normal MBs (Fig. 7G,H). The misguided axons were not stalled at the peduncle end, possibly because their surrounding wild-type axons may somehow steer the unf knockdown axons into the lobes. Additional evidence for the involvement of unf in axon guidance came from the observation that about 10% of the strongest unf knockdown MBs showed ectopic FasII-positive axon bundles extending through the calyx into abnormal targets rather than migrating along the peduncle (Fig. 7I,J). Taken together, these observations suggest that unf acts cell-autonomously to regulate individual axons' pathfinding (Fig. 7K,L).

unf Is Expressed in MB Neurons Continuously to Regulate Neuron Subtype Identity
In the MB c neurons, unf is required in a stage-specific manner. Genes required for remodeling of MBs may dynamically express in response to ecdysone signaling. To determine the expression of unf, we generated a rabbit polyclonal antibody against a short peptide (DVTNDNEEPHA) characteristic of unf. The antibody recognized abundant unf expression in the MB neuronal cell bodies (Fig. 8).
Depleting unf by targeted RNAi greatly suppressed the immunostaining signal (Fig. 8A-D), confirming the specificity of the antibody. Notably, unf is enriched in all MB neurons through different developmental stages (Fig. 8). No dynamic changes in its expression could be detected during MB remodeling, and highlevel expression continues in adult MB neurons (Fig. 8). This expression pattern provides no clue as to why unf is essential for c neuron remodeling but largely dispensable for their initial morphogenesis. However, its enrichment in the MBs strongly supports our observations that unf controls some MB-characteristic aspects of neural development.
Human PNR and C. elegans fax-1, the orthologs of fly unf, have been shown to regulate neuron identity [11,22,58,59]. To investigate if unf controls MB neuron identity and subsequently governs the subtype-specific axon projections, we examined the neuronal cell fate in unf knockdown MBs. We achieved this by knocking down unf using asenase-GAL4 in combination with different Pan-MB or MB subtype-specific markers, including several GAL4s. Given that asense-GAL4 expresses in MB precursors and developing young MB neurons [44], it allows us to knock down unf throughout MB development; and because ansense-GAL4 does not express in adult MB, it will not interfere with the expression of the other MB-specific GAL4s in the adult brain. None of the pan-MB markers, including dachshund (dac) [60,61], GAL4-OK107 [51] and GAL4-MB247 [62], was affected by unf knockdown (data not shown), suggesting that the MB lineage identity was not regulated by unf. However, all the subtype-specific markers we examined lost their expression in the unf knockdown adult MBs. These include the c and a9/b9-specific marker Trio in MB cell bodies (some Trio protein can be detected in MB lobes, which might be the residual product of earlier trio expression; see below) [50] (Fig. 9A,B), the c-specific maker GAL4-NP21 [63] (Fig. 9C,D), the a9/b9-specific marker GAL4-c305a [64] (Fig. 9E,F), the pioneer a/b-specific marker GAL4-c708 [44](data not shown) and the a/b-specific marker GAL4-c739 [53] (Fig. 9G,H). Thus, MB neurons require unf to acquire their subtype identity.
Notably, at the wandering larval stage, when the morphology of the unf knockdown c neurons was mostly normal, the expression of the c-specific genes such as trio or EcR-B1 [24,50] was unaffected (Fig. 9I-L). The loss of the c-specific markers during metamorphosis suggests that there is a second wave of cell fate specification and/or consolidation occurring at the larva-topupa transition for the c neurons to acquire their final fate; and unf is possibly required for this process. In sum, loss of neuron subtype identity may underlie misguidance of axons in unf knockdown MBs.

Discussion
Silencing individual NRs through development of the MBs by transgenic miRNAs has allowed us to identify unf as another NR (in addition to EcR, usp, and tll) that regulates MB development. Previously, unf was shown to be essential for the wing expansion and fertility of adult flies, and abundantly expressed in the developing MBs [49]. Here we studied the function of unf in further detail and learned that unf acts in all three major types of MB neurons to promote proper neuron subtype identity and axon guidance. Comparable axon stalling defects of adult MB neurons, as well as missing larval MB dorsal axonal branches were observed in unf mutant organisms (Bates et al., submitted), validating our study of unf's mechanism of action by targeted RNAi.
The involvement of unf in cell fate determination is evident by the loss of subtype-specific markers in the unf knockdown MB. Similar mechanisms have been shown in C. elegans and human. In human, the unf ortholog PNR is specifically expressed in rod photoreceptor cells to promote the rod-cell identity by repressing the expression of S-cone cell specific genes [11,58]. Mutations in PNR leads to enhanced S-cone syndrome (ESCS) which is an inherited disease causing hypersensitivity to short-wave light due to increased numbers of S-cone cells at the expense of rod photoreceptor cells [11,58]. However, in the present study, we did not observe any reciprocal cell number change in unf knockdown MBs. Thus, unlike PNR in human, unf in flies is not used to repress a default cell fate. Given that unf is required for the proper gene expression in all three major types of MB neurons, it may play a general role in assisting the temporal identity factors, such as chinmo, to diversify neuronal cell fates [3,44].
Loss of proper identity might underlie the axon guidance defect observed in the unf knockdown MB. The C. elegans ortholog of unf, fax-1, has been suggested to regulate cell identities of 18 neurons, including both motor-and interneurons; in fax-1 mutant, some neurotransmitters and synaptic proteins are not properly expressed in the neurons normally expressing fax-1 [22,59,65]. Notably, like unf in fly MB neurons, fax-1 is also required for axonal pathfinding for several C. elegans neurons, indicating a functional conservation among unf orthologs in different species [22,59,65].
unf has been hypothesized to act as a transcriptional repressor, because its ligand-binding domain (LDB) failed to activate gene expression [66]. The function of the human ortholog PNR to repress S-corn specific genes supports this hypothesis [58]. Contrary to the human PNR, in both flies and worms, loss of unf or fax-1 leads to the down-regulation of many neural genes [22,59]. However, since there is no evidence for gene activation by direct binding of unf or fax-1, it is possible that unf and fax-1 regulate these neural genes indirectly by repressing other repressors.
In the MBs, unf primarily governs axonal morphogenesis during later larval and pupal development when different MB neuron types need to make distinct projections. For instance, in early pupae, a/b neurons undergo de-novo axonogenesis to form the a/ b lobes while c axons regenerate to make up the adult-specific c lobe. Given the notion that unf promotes MB axonogenesis possibly through regulating neuron subtype identity, its selective involvement in late MB morphogenesis could simply reflect the importance of neuron subtype identity in ensuring diverse subtypespecific axonal morphogenesis. However, the larval c neurons of unf knockdown MBs show normal cell fate as evidenced by proper expression of EcR-B1. This argues for a stage-specific function of unf in MB development. Notably, despite its stage-specific requirement, unf is enriched in the MBs through different developmental stages and into the adult, raising the possibility that its dynamic activity is patterned through temporal control of ligand availability. Recently, an in vitro study suggested that UNF is a heme binding protein [48]. Given the known relationship between heme and lipid metabolism [6,12,67,68], heme levels can serve as an indicator for energy resource and developmental progress. Perhaps by detecting heme levels, unf may potentially coordinate the timing of the unf-medicated adult-specific cell fate determination and axonogenesis. However, in vivo evidence for the interaction between heme and unf remains lacking.
In conclusion, patterned NR activities govern various temporally regulated neural developmental processes of interest. UNF and its orthologs probably promote subtype neuronal differentiation in temporally controlled manners. Elucidating the regulation of NR activities and their control of neuron subtype identity should shed additional light on how diverse neuron types undergo differential morphogenesis and acquire different subtype-specific projections to construct the complex brain.

Generation of miRNA Fly Lines
The UAS-miRNA constructs were generated as described in the previous study [46]. The miRNA target sequences of 18 NR genes were shown in Table S1. The transgenic flies were generated by inserting UAS-miRNA constructs into the attp-16 site on the second chromosome using the integration system as previous described [69].

Induction of Flip-Out Clones and MARCM Analysis
To generate flip-out clones, the newly hatched larvae with proper genotype were applied 30-minute heat-shock in 32uC water bath. For MARCM studies, larvae with proper genotype were collected in 2 hours after hatching, and cultured at the density of 100 larvae per food vial at 25uC. MARCM clones were induced by applying 1 hour heat-shock in 38uC water bath at various developmental stages.