The unfulfilled gene is required for the development of mushroom body neuropil in Drosophila

Background The mushroom bodies (MBs) of Drosophila are required for complex behaviors and consist of three types of neurons, γ, α'/β' and α/β. Previously, roles for transcription factors in MB neuronal differentiation have only been described for a subset of MB neurons. We are investigating the roles of unfulfilled (unf; HR51, CG16801) in MB development. unf encodes a nuclear receptor that is orthologous to the nuclear receptors fasciculation of axons defective 1 (FAX-1) of the nematode and photoreceptor specific nuclear receptor (PNR) of mammals. Based on our previous observations that unf transcripts accumulate in MB neurons at all developmental stages and the presence of axon pathfinding defects in fax-1 mutants, we hypothesized that unf regulates MB axon growth and pathfinding. Results We show that unf mutants exhibit a range of highly penetrant axon stalling phenotypes affecting all neurons of the larval and adult MBs. Phenotypic analysis of unfX1 mutants revealed that α'/β' and α/β neurons initially project axons but stall prior to the formation of medial or dorsal MB lobes. unfZ0001 mutants form medial lobes, although these axons fail to branch, which results in a failure to form the α or α' dorsal lobes. In either mutant background, γ neurons fail to develop larval-specific dorsal projections. These mutant γ neurons undergo normal pruning, but fail to re-extend axons medially during pupal development. unfRNAi animals displayed phenotypes similar to those seen in unfZ0001 mutants. Unique asymmetrical phenotypes were observed in unfX1/unfZ0001 compound heterozygotes. Expression of UAS-unf transgenes in MB neurons rescues the larval and adult unf mutant phenotypes. Conclusions These data support the hypothesis that unf plays a common role in the development of all types of MB neurons. Our data indicate that unf is necessary for MB axon extension and branching and that the formation of dorsal collaterals is more sensitive to the loss of unf function than medial projections. The asymmetrical phenotypes observed in compound heterozygotes support the hypothesis that the earliest MB axons may serve as pioneers for the later-born MB neurons, providing evidence for pioneer MB axon guidance in post-embryonic development.


Background
The mushroom bodies (MBs) of Drosophila melanogaster, which are required for olfactory learning and other complex behaviors [1,2], are ideal for studying the transcriptional regulation of interneuronal development because they form discrete axonal projections that are well-characterized [3][4][5] and easily visualized [4,[6][7][8][9]. Four neuroblasts in each brain hemisphere sequentially generate three types of Kenyon cells, the γ, α'/β', and α/ β MB neurons that begin dividing during embryogenesis and continue to divide through development [10,11]. Each neuron projects dendrites that contribute to a large dendritic field in the calyx, and an axon that travels anteroventrally, forming a tightly bundled peduncle before branching medially to form the γ, β', and β lobes, and dorsally to form the α' and α lobes ( Figure 1A). The earliest born γ neurons initially extend axons both medially and dorsally during late embryonic and early larval stages. These larval-specific γ axons are then pruned back to the peduncle by 18 hours after puparium formation (APF) and re-extend medially during pupal remodeling; the late larval-born α'/β' and pupalborn α/β neurons do not remodel their axonal projection patterns during metamorphosis [3][4][5].
Since the three different classes of MB neurons are born sequentially, generate a single dendritic field, project axons that fasciculate prior to branching medially and/or dorsally to form type-specific lobes, it is interesting to consider whether any differentiative events of the γ, α'/β', and α/β neurons are regulated by a common set of genes or whether they utilize independent transcriptional networks. Existing data on the role of transcription factors in MB differentiation provide little insight into this question. The genes eyeless [12][13][14], tramtrak [15], mushroom body miniature [16,17], chinmo [18], polyhomeotic [19], and tailless [20] regulate proliferation, specification, and viability of MB neurons, events that precede differentiation. dachshund (dac), ecdysone receptor B1 (EcR-B1), ultraspiracle (usp), and dSmad2 act in subtype-specific pathways [12,13,[21][22][23], consistent with the hypothesis that the differentiation of the γ, α'/β', and α/β neurons utilize independent transcriptional pathways. dac mutants display axonal branching and pathfinding defects in subsets of α'/β' and α/β MB neurons [12,21]. EcR-B1 and its heterodimeric partner, Ultraspiracle (USP), both members of the nuclear receptor superfamily, form an ecdysone-regulated transcription factor that is required for the pruning of MB γ neurons at the outset of metamorphosis [22]. dSmad2 regulates the transcription of EcR-B1 in MB γ neurons during neuronal remodeling [24]. Thus, whether any Figure 1 unf is required for mushroom body lobe formation. In the adult brain, the mushroom body (MB) is a paired neuropil structure that comprises five axonal lobes, γ, α'/β', and α/β. Each neuron projects dendrites that contribute to a large dendritic field (calyx), and an axon that travels anteroventrally. MB axons fasciculate with other MB axons forming a peduncle (Ped) before projecting axons medially and dorsally. α' and α axons project dorsally, whereas γ, β', and β axons project medially, forming five distinctive lobes. To visualize the MB lobes, OK107-GAL4 was used to drive expression of the UAS-mCD8GFP transgene in all MB neurons (Kenyon cells) and their axons. Lobes were distinguished by using anti-Fasciclin II (anti-Fas II) to label α and β lobes and anti-Trio to label α', β', and γ lobes. (A) In adult UAS-mCD8GFP;;OK107-GAL4 control animals labeled with anti-Fas II (red), all MB lobes have formed. (B) In unf X1 mutants, MB axons have formed a peduncle (arrowhead), but have spread out and stalled prior to lobe formation (arrow). (C) In unf Z0001 mutants, γ, β', and β axons projected medially, but were disorganized. No dorsal lobes were formed (star). (D) This +/Df2426;UAS-unf RNAi ;OK107-GAL4 adult displayed dramatically reduced dorsal lobes in one brain hemisphere (arrow). (E, F) In unf X1 and unf Z0001 rescue animals, in which a wild-type unf trangene was expressed in all MB neurons in an otherwise mutant background, all MB lobes were present. It is interesting to note that in rescued flies, MB lobes may have fewer axons, and that some medially projecting axons have extended past the midline (E, arrow). Eb, ellipsoid body; Meb, median bundle; Ped, peduncle. Scale bars = 10 μm. differentiative events of the γ, α'/β', and α/β neurons are regulated by a common set of genes has not been previously reported.
In this study we show that the gene unfulfilled is required for the development of all three types of MB neurons, supporting the hypothesis that some differentiative events of the three types of MB neurons are regulated by a common set of genes. The unfulfilled gene (unf; HR51, CG16801) encodes the Drosophila NR2E3 member of the nuclear receptor superfamily [25]. UNF, like all classical nuclear receptors, contains an aminoterminal transactivational domain, a DNA-binding domain, a hinge region, and a carboxy-terminal ligandbinding domain [26]. unf is an ortholog of the Caenorhabditis elegans gene fasciculation of axons defective (fax-1) and the human gene photoreceptor specific nuclear receptor (PNR) [27]. Both fax-1 and PNR mutations disrupt developmental events in a limited number of neurons and result in behavioral or sensory deficits. fax-1 mutants are uncoordinated and display axon pathfinding and neurotransmitter defects [28][29][30]. The observed axon pathfinding defects are inferred to be due to the misregulation of fax-1 target genes. PNR impacts neuronal identity of vertebrate photoreceptors, functions as a dimer, and acts as a dual function transcriptional regulator, able to act as a transcriptional activator and a transcriptional repressor [31][32][33][34][35][36][37].
Based on our previous observations that robust levels of unf transcripts accumulate in MB neurons at all developmental stages [25] and the axon pathfinding defects of fax-1 mutants [28][29][30], we hypothesized that unf regulates MB axon growth and pathfinding. Phenotypic analysis of unf mutants revealed that MB axons stall prior to the formation of the lobes with the exception of the larval-specific γ neurons, which project axons medially, but fail to project dorsally. These axons are pruned appropriately but fail to re-extend during pupal stages. Expression of an unf transgene in the MBs in a mutant background rescued the unf mutant phenotypes, demonstrating that MB defects of unf mutants are due to loss of unf function in the MB neurons. These data demonstrate that unf is required for the proper formation of γ, α'/β', and α/β lobes, consistent with the hypothesis that at least some differentiative events of the γ, α'/β', and α/β neurons are regulated by a common set of genes.
unf Z0001 /Df2426 hemizygous (unf Z0001 ,UAS-GFP/ Df2426;;OK107-GAL4; Figure 1C) adults formed only medial lobes and displayed γ axons that splayed as they approached the midline compared to the compact bulblike organization of γ lobes in wild-type animals ( Figure  1C; Table 1, row 10; Additional files 4 and 5). α'/β' and α/β neurons of unf Z0001 /Df2426 hemizygotes projected axons medially, but rarely projected dorsal collateral axons (data not shown). Dorsal lobes were observed more frequently in unf Z0001 /unf Z0001 homozygotes (unf Z0001 /unf Z0001 ,UAS-GFP/Df2426;;OK107-GAL4; Table 1, row 11). The observation that dorsal lobes were observed more frequently in unf Z0001 /unf Z0001 homozygotes than in unf Z0001 /Df2426 hemizygotes supports the hypothesis that the unf Z0001 allele is a hypomorph [25]. All MB lobes were observed in control genotypes (w 1118 , unf Z0001 /+, Df2426/CyOGFP, UAS-GFP;;OK107-GAL4 controls); however, missing dorsal lobes were observed in 1 of 10 unf Z0001 ,UAS-GFP/CyOGFP;;OK107-GAL4 control siblings ( To independently test whether unf plays a role in MB neuron development, a UAS-unf RNAi line was crossed to the OK107-GAL4 line to generate animals in which unf levels were reduced in the MBs via RNA interference (RNAi). Adult brains were double-labeled with anti-Fas II and anti-Trio to visualize the five MB lobes. When OK107-GAL4 was used to drive UAS-unf RNAi in a wildtype background, normal MBs were observed (data not shown). However, when OK107-GAL4 was used to drive UAS-unf RNAi in unf + /Df2426 hemizygotes (unf + /Df2426; UAS-unf RNAi ;OK107-GAL4) 50% of brains displayed dramatically reduced dorsal lobes or were missing dorsal lobes bilaterally or unilaterally ( Figure 1D; Table 1, row 13). These RNAi data are consistent with the analyses of unf X1 and unf Z0001 mutants, supporting the hypothesis that unf is necessary for axon extension and branching in all MB neurons and that dorsal collaterals are more sensitive to loss of unf function than medial projections.

unf expression in the mushroom bodies rescues lobe formation
We tested whether expression of unf in the MBs was sufficient to rescue the phenotypes observed in unf mutants by driving expression of a UAS-unf transgene with the OK107-GAL4 transgene. Interestingly, all UASunf rsqF ;OK107-GAL4 flies developed to late pupal stages, but failed to eclose. The failure to eclose is probably due to OK107-GAL4-driven expression of unf in regions other than the MBs causing a disruption that prevents further development. We therefore assessed the MBs of rescued animals at late pupal stages, 72 to 96 hours APF. At this late developmental time in wild-type pupae, the MBs are indistinguishable from those of adult MBs [4]. Medial and dorsal MB lobes were observed in all unfX1/Df2426 rescued (unf X1 ,UAS-GFP/ Df2426;UAS-unf rsqF /+;OK107-GAL4; Figure 1E) and unf Z0001 /Df2426 (unf Z0001 ,UAS-GFP/Df2426;UAS-unf rsqF / +;OK107-GAL4; Figure 1F) rescued pupae. In contrast, MB lobes were not observed in unf X1 /Df2426 control siblings (unf X1 ,UAS-GFP/Df2426;TM3/+;OK107-GAL4) that lacked the UAS-unf rsqF transgene (Table 1, compare  rows 15 and 18). Similarly, only medial lobes were observed in unf Z0001 /Df2426 control siblings (unf Z0001 / Df2426,UAS-GFP;TM3/+;OK107-GAL4) ( Table 1, compare rows 16 and 23) that lacked the UAS-unf rsqF transgene. All MB lobes were observed in all other control pupae (Table 1, rows 19 to 22). MB lobes appeared thin and less robust in some unf X1 and unf Z0001 rescued animals, suggesting that the rescue was imperfect. Nonetheless, MB axons contributing to each of the five MB lobes could be identified in all rescued animals. An independent rescue line, UAS-unf rsqC , was tested with OK107-GAL4 to express unf in the MBs of unf X1 / Df2426 hemizygotes. All MB lobes were observed in six of seven rescued unf X1 /Df2426 pupae with the UASunf rsqC transgene (unf X1 ,UAS-GFP/Df2426;UAS-unf rsqC / +;OK107-GAL4). In the seventh pupa, medial lobes were observed bilaterally, while dorsal lobes were observed only in one hemisphere (Table 1, row 17). MB lobes were not detected in any unf X1 /Df2426 control siblings that lacked the UAS-unf rsqC transgene (unf X1 ,UAS-GFP/ Df2426;TM3/+;OK107-GAL4; Table 1, row 24). To confirm that the rescue depended upon the expression of an unf open reading frame, we tested the ability of a UAS-lacZ transgene to rescue unf X1 /Df2426 hemizygotes. As expected, MB lobes were not observed in unf X1 /Df2426 pupae containing the UAS-lacZ transgene (unf X1 ,UAS-GFP/Df2426;UAS-lacZ/+;OK107-GAL4; Table 1, row 25). These data demonstrate that the axonal defects observed in unf X1 mutants are due to the lack of UNF function. unf MB05909 : GAL4 insertion unf X1 : splicing defect unf Z0001 : G120R missense DBD Figure 2 Summary of unf alleles. The unf X1 allele disrupts the 5' donor splice site of intron 2, whereas the unf Z0001 allele has a missense mutation due to a guanine to adenine transition at base 312 of exon 2, resulting in a glycine to arginine substitution (G120R) [25]. The unf MB05909 line contains a GAL4 insertion in intron 1 of the unf gene (FlyBase). DBD, DNA-binding domain. Table 1 Mushroom body phenotypes in unf mutants and rescue animals.
unf X1 /unf Z0001 compound heterozygotes exhibit a range of mushroom body phenotypes Phenotypic analysis of unf X1 /unf Z0001 compound heterozygotes (unf X1 ,UAS-GFP/unf Z0001 ;;OK107-GAL4) revealed a range of aberrant MB phenotypes. Thirteen percent of unf X1 /unf Z0001 compound heterozygotes lacked all MB lobes ( Figure 3A), similar to the unf X1 mutant phenotype. Forty-seven percent of the unf X1 / unf Z0001 compound heterozygotes developed only medial lobes and were missing dorsal lobes ( Figure 3B, C), similar to the unf Z0001 mutant phenotype. These unf Xl / unf Z0001 compound heterozygotes occasionally displayed a thin fascicle of dorsally projecting α' axons ( Figure 3B, C). Interestingly, 40% of unf X1 /unf Z0001 compound heterozygotes exhibited asymmetrical phenotypes in which a dorsal and/or medial lobe were present in one hemisphere but missing in the other (Table 1, row 14). In some cases, medial axons misprojected or extended past the midline ( Figure 3B, C, E). γ neurons were also variably affected in unf X1 /unf Z0001 compound heterozygotes and often appeared defasciculated, and stalled at various points along their medial trajectory ( Figure 3C, D, F; Additional files 6, 7 and 8). The novel phenotypes of unf X1 /unf Z0001 compound heterozygotes that are different from either unf X1 /Df2426 or unf Z0001 /Df2426 hemizygotes demonstrate that the unf X1 and unf Z0001 alleles interact.
unf is required for larval-specific g dorsal collaterals and the re-extension of g axons during metamorphosis The axon stalling phenotypes of unf X1 /Df2426 hemizygotes suggested that unf is required in all MB neurons for axons to extend in any direction beyond the heel region of the MB. This region is the branching point for dorsal collateral projections from medially projecting axons. The MB axons of adult unf X1 /Df2426 hemizygotes may have all stalled during the initial phase of their outgrowth, either during larval or pupal development. Alternatively, it is possible that MB neurons may initially project axons medially and dorsally, but these axons may not be maintained into the adult. To determine whether unf is required for the initial projection patterns of all MB neurons, the MBs of experimental and control animals were analyzed at various larval and pupal stages.
Expression of the UAS-unf rsqF transgene in the MBs in unf X1 /Df2426 hemizygotes (unf X1 ,UAS-GFP/Df2426; UAS-unf rsqF /+;OK107-GAL4) rescued the dorsal collaterals of the larval γ neurons in five of six third instar larvae ( Figure 4C; Table 1, row 15) and supported the re-extension of γ medial axon projections in pupae (Figure 1E, 1F). These data confirm that the unf function in MBs is necessary for the formation of dorsal collaterals in γ neurons during larval development.

201Y-GAL4 and c739-GAL4 driven GFP expression in the mushroom bodies is unf-dependent
To independently confirm that the axon stalling phenotypes of unf X1 /Df2426 hemizygotes were not an artifact associated with the OK107-GAL4 transgene (Figure 1), we examined UAS-GFP expression in the MBs of unf X1 / Df2426 hemizygotes using three other GAL4 drivers known to express in all or subsets of MB neurons, 201Y-GAL4, c739-GAL4, and c747-GAL4 [7,40]. In each case, adult brains were counterstained with fluorescently labeled phalloidin to visualize actin-rich structures, including the extensive dendritic arbors of MB neurons that fill the calyces. The c747-GAL4 line showed GFP expression in MB neurons in control and unf X1 /Df2426 MB neurons ( Figure 5A, 5B). The unf X1 /Df2426 hemizygotes (unf X1 ,UAS-GFP/Df2426,c747-GAL4) displayed stalled axons and failed to develop any MB lobes, confirming the unf X1 mutant phenotypes described using the OK107-GAL4 transgene to drive the reporter transgene in MB neurons. Surprisingly, unf X1 /Df2426 hemizygotes carrying the 201Y-GAL4 transgene failed to express GFP in MB neurons (compare Figure 5C and Figure 5D), and GFP expression in unf X1 /Df2426 In this unf X1 /unf Z0001 heterozygote only medial lobes are present, similar to the unf Z0001 mutant phenotype; a thin fascicle of α' axons is present in the right hemisphere (star). (C) In this unf X1 /unf Z0001 heterozygote, left hemisphere β' and β axons extend medially beyond the midline (arrow), whereas γ axons appear to stall; right hemisphere β and β' axons misproject ventrally (arrowhead) and γ axons are highly disorganized; only a few α' dorsal axon projections are present in either hemisphere (stars). (D) In this unf X1 /unf Z0001 heterozygote α and α' dorsal axon projections are present in the left hemisphere, whereas only a thin fascicle of α' axons is present in the right hemisphere (star); β axons are present in the right hemisphere and appear to stall (arrow), whereas they are completely absent in the left hemisphere. (E) In this unf X1 /unf Z0001 heterozygote β', β, and γ axons project medially and cross the midline (arrow), but dorsal axons are missing in the left hemisphere; in the right hemisphere γ axons project medially and α and α' axons project dorsally but appear to stall (arrowhead). (F) In this unf X1 /unf Z0001 heterozygote, α and α' dorsal axon projections are present in the left hemisphere, whereas only α' dorsal axons are present in the right hemisphere; medial axon projections are disorganized and stall before reaching the midline in both right and left hemispheres (arrows). Eb, ellipsoid body; Meb, median bundle; Ped, peduncle. Scale bars = 25 μm.
hemizygotes carrying the c739-GAL4 transgene was greatly diminished (compare Figure 5E and Figure 5F). These observations indicate that 201Y-GAL4 and c739-GAL4 expression is unf-dependent. The 201Y-GAL4 transgene is an insertion in the TAK1-associated binding protein 2 (Tab2) gene (FlyBase). Inverse PCR revealed that the c739-GAL4 transgene is inserted in the second intron of hormone receptor-like in 39 (HR39; data not shown).

Discussion
The unf X1 and unf Z0001 alleles interact showing that both alleles are at least partially functional unf mutants exhibit a range of highly penetrant axon stalling phenotypes affecting all neurons (γ, α'/β' and α/ β) of the larval and adult MBs. Similar phenotypes have been observed in unf microRNA knockdown animals [41]. unf X1 /Df2426 hemizygotes and unf X1 /unf X1 homozygotes fail to project larval-specific γ dorsal collaterals, fail to re-extend γ axons medially during metamorphosis, and fail to project any medial and dorsal axons of α'/β' and α/β neurons. The γ, α'/β' and α/β axons of unf Z0001 /Df2426 hemizygotes only project medially, whereas MBs were normal in some unf Z0001 /unf Z0001 homozygotes. These data together with previous observations [25] would seem to support the hypothesis that the unf X1 allele is an amorph, a null allele, and that the unf Z0001 allele is a hypomorph, a partial loss of function allele. However, while the unf Z0001 allele behaves as a hypomorph with respect to sterility, it displays dominant properties with respect to wing expansion [25]. Interestingly, the G56R allele of PNR, which displays dominant properties [42], is structurally equivalent to the unf Z0001 allele (G120R) [25]. The observation that the unf X1 / unf Z0001 compound heterozygotes display unique phenotypes was unexpected and demonstrates that these alleles interact, compelling us to conclude that the unf X1 allele is not a null allele. These data strongly suggest that the unf X1 allele encodes a unique isoform of the UNF protein, UNF X1 , which is predicted to contain the 110 residue amino-terminal domain and the complete first zinc finger of the DNA-binding domain [25]. These data do not allow us to infer the functional nature of the unf X1 allele or the mechanism of this genetic interaction.
Asymmetrical phenotypes suggest a role for unfulfilled in pioneer axon guidance The phenotypic variation and asymmetry observed in the MBs of unf X1 /unf Z0001 compound heterozygotes supports the hypothesis that pioneer axons are established early during MB development and that the pathfinding of these pioneers is unf-dependent. The observation that the β lobe axons in one hemisphere project medially while the β lobe axons in the other hemisphere project ventromedially ( Figure 3C, 45°a ngle down, arrowhead) demonstrates that the projection of the β lobes in these unf X1 /unf Z0001 compound heterozygotes is independent of their genotype. In an independent, yet genetically identical animal, unf X1 / unf Z0001 α/β MB neurons assume a different fate and fail to project any medially projecting β axons ( Figure  3F). Similar observations can be made for all unf X1 / unf Z0001 MB lobe axons when these and other samples are examined (Figure 3). The fact that the axons of the later-born α/β neurons consistently stall or misproject whenever the axons of the earlier-born α'/β' neurons stall or misproject suggests that the α'/β' axons may be acting as pioneers for the α/β axons. These data support a model of MB lobe formation in which unf is required for MB pioneer axons to navigate to their targets, and that later-born MB neurons project axons that fasciculate along these established axons. We propose that the variable and asymmetric phenotypes observed in unf X1 /unf Z0001 compound heterozygotes are due to inappropriate targeting of pioneer axons of the MB or the stalling of pioneer axons prematurely as a result of insufficient unf function in α'/β' pioneer axons. Thus, the asymmetric β lobe projection in Figure 3C may be due to asymmetric projections of pioneer axons, while the lack of β lobes in Figure 3F may be due to the stalling of these pioneer axons in the peduncle. These data are supported by an analysis of non-autonomous effects of Dscam mutant clones, which suggests that the α'/β' axons may be acting as pioneers for the α/β axons at least some of the time [43]. Our observations that larval γ dorsal axons, and the α' and α dorsal lobes all fail to develop in unf Z0001 /Df2426 mutants not only supports the hypothesis that the α'/β' axons act as pioneers for the α/β axons, but suggests that the larval γ axons act as pioneers for the α'/β' axons. Discerning the roles and mechanisms of unf in MB pioneers during postembryonic MB development requires further investigation.
unf plays a common role in the early development of all mushroom body neurons The data presented here demonstrate that unf plays a common role in the early development of all three subtypes of MB neurons by regulating axon extension and branching. While we cannot rule out the possibility that single axons, which normally project dorsally, may be misguided and project medially, our analysis is consistent with the hypothesis that unf mutant γ, α'β', and α/ β neurons fail to project dorsal axon branches. Our observations that unf mutant MB neurons express subtype-specific epitopes such as Fas II and Trio suggest that unf does not impact MB neuronal subtype identity. Interestingly, Lin et al. [41] disagree and conclude that unf does regulate MB neuronal subtype identity based on a series of unf RNAi knockdown experiments. We argue that until the transcriptional codes that distinguish MB neuronal subtypes are defined, one cannot conclusively determine whether the identity of these neurons has been impacted [44][45][46][47]. While we cannot yet place unf at any specific position in a transcriptional hierarchy that regulates MB development, our data suggest that unf acts earlier than dac, EcR-B1, and usp, since these genes regulate differentiation in specific subsets of MB neurons while unf regulates the differentiation of all three MB subtypes.
Based on the homology of UNF with PNR, we anticipate that UNF is likely to act as a dual function transcription factor with the ability to activate the transcription of some target genes, while repressing others [32,33,35,36]. Palanker et al. [48] have previously proposed that UNF may function as a transcriptional repressor. We propose that the axon stalling phenotype observed in unf mutants is due to the misregulation of target genes. Two potential target genes are Tab2 and HR39 based on the observation that the expression of the 201Y-GAL4, an enhancer trap insertion in Tab2, and c747-GAL4, an enhancer trap insertion in HR39, is unf-dependent. If the transcriptional regulation of these two enhancer trap transgenes reflects the transcriptional regulation of the genes in which they are inserted, then it follows that Tab2 and HR39 are expressed in the MBs and that this expression is unf-dependent, a hypothesis that remains to be tested.
Axon stalling and branching phenotypes in unf mutants suggest that genes encoding axon guidance cues are likely to be regulated by unf. Genetic screens have already identified a number of guidance genes that are expressed in the MBs. For example, genes encoding cell adhesion molecules like volado, an α-integrin, Notch, a transmembrane receptor and transcription factor [1], and semaphorinla [49] and plexinA [15], which encode a ligand-receptor pair that is largely involved in axonal repulsion, have been identified. Ephrin receptor tyrosine kinase [50] and Dscam [43,51], other wellknown guidance cues, are necessary for the proper guidance of MB axons. Misregulation of these or other guidance genes could disrupt the normal balance of attractive and repulsive cues resulting in inappropriate axon pathfinding and stalling.

Conclusions
These data support the hypothesis that unf plays a common role in the early development of all three subtypes of MB neurons, γ, α'/β', and α/β, by regulating axon extension and branching during the initial phases of larval and pupal outgrowth. Expression of a UAS-unf transgene in MB neurons of unf mutants rescues the unf mutant MB phenotypes, demonstrating that the MB defects are due to the lack of unf. The phenotypic variation and asymmetry observed in the MBs of unf X1 / unf Z0001 compound heterozygotes suggests a role for unf in the targeting of pioneer axons.

Drosophila stocks
All stocks were raised on standard cornmeal and sugar medium. The unf X1 and unf Z0001 stocks have been characterized previously: the unf X1 allele disrupts the 5' donor splice site of intron 2, whereas the unf Z0001 allele has a missense mutation due to a guanine to adenine transition at base 312 of exon 2 resulting in a glycine to arginine substitution (G120R) [25]. The Df(2R)ED2426 (Df2426) chromosome carries a deletion of 482,016 bp on the second chromosome that removes 57 genes or annotated genes in their entirety, including unf [52]. The Hr51 MB05909 (unf MB05909 ) line contains a GAL4 insertion in intron 1 of the unf gene (FlyBase). The Df (2R)ED2426, Hr51 MB05909 (unf MB05909 ), UAS-CD8::GFP;; OK107-GAL4, 201Y-GAL4, c747-GAL4, and c739-GAL4 lines were obtained from Bloomington Stock Center. The UAS-unf RNAi line was obtained from Vienna Drosophila RNAi Center (VDRC). All mutant or transgenic stocks were maintained over GFP-marked chromosomes to facilitate genotyping. Animals were reared at 25°C with the exception of UAS-unf RNAi crosses, which were reared at 29°C.

Immunohistochemistry and microscopy
Third instar larvae and pupae were staged as described [54,55]. The central nervous system of third instar larvae, pupae, and 0-to 2-day adults were collected, fixed in 4% paraformaldehyde, and processed using standard protocols [9]. mAb1D4 [56] (anti-Fas II; 1:10) and mAb9.4A [39]  Preparations were examined and imaged using an Olympus Fluoview FV-1000 laser scanning confocal system mounted on an Olympus IX-81 inverted microscope. Images were processed using Image J and Adobe Photoshop. Movies of z stacks were processed using QuickTime. During z stack collections for Additional files 1 to 3, the photomultiplier tube was manually adjusted for optimal brightness at different focal planes.

Statistics
The Fisher exact test was used to determine whether the frequency of defects in experimental animals was significantly different from the frequency of defects in control animals. Relevant genotypes were tested in pair-wise combinations. P-values less than 0.01 were considered significant.