To further our understanding of dendrite wiring specificity in Drosophila olfactory PNs, we performed a MARCM-based forward genetic screen using piggyBac insertional mutants 22 . MARCM allows visualization and genetic manipulation of single cell or neuroblast clones in an otherwise heterozygous background, permitting the study of essential genes in mosaic animals 23 . In this screen, we used GH146-GAL4 to label a single PN born soon after larval hatching 1 , which in wild-type (WT) animals always projects its dendrites to the dorsolateral glomerulus DL1 in the posterior of the antennal lobe (Figure 2A). The DL1 PN also exhibits a stereotyped axon projection, forming an L-shaped pattern in the lateral horn, with additional branches in the mushroom body calyx (Figure 2B). We identified a mutant, called LL05955, in which DL1 PNs mistargeted to the dorsolateral glomerulus DA4l in the anterior of the antennal lobe (Figure 2C). This phenotype is strikingly specific, with 100% penetrance (Table 1). Arborization of mutant axons, however, was not obviously altered (Figure 2D). We identified the piggyBac insertion site using inverse PCR 22 and Splinkerette PCR 24 . LL05955 is inserted into the coding sequence of Bap55 (Figure 1C), encoding a homolog of human BAF53a and BAF53b. Precise excision of the piggyBac insertion reverted the dendrite mistargeting phenotype, confirming that disruption of the Bap55 gene causes the dendrite mistargeting (Table 1).

In addition to causing DL1 mistargeting, Bap55 mutants also display neuroblast clone phenotypes. In WT, GH146-GAL4 can label three distinct types of PN neuroblast clones generated in newly hatched larvae. Two of these clones, the anterodorsal neuroblast clone (Figure 2E) and the lateral neuroblast clone (Figure 2F), possess cell bodies that lie dorsal or lateral to the antennal lobe, respectively. PNs from these two lineages project their dendrites to stereotyped and nonoverlapping subsets of glomeruli in the antennal lobe. The third type of clone, the ventral neuroblast clone, has cell bodies that lie ventral to the antennal lobe and dendrites that target throughout the antennal lobe due to the inclusion of at least one PN that elaborates its dendrites to all glomeruli (Figure 2G) 1 2 .

In Bap55 ^-/- PNs, anterodorsal neuroblast clones display a mild reduction in cell number, and their dendrites are abnormally clustered in the anterior dorsal region of the antennal lobe, including the DA4l glomerulus (Figure 2H). Lateral neuroblast clones display a severe reduction in cell number, and the remaining dendrites are unable to target to many glomeruli throughout the antennal lobe (Figure 2I). Ventral neuroblast clones display a mild reduction in cell number and a reduced dendrite mass throughout the antennal lobe (Figure 2J). During development, the lateral neuroblast first gives rise to local interneurons before switching to produce PNs 25 ; in mutants affecting cell proliferation, this property of the lateral neuroblast displays as a severe reduction in GH146-labeled PNs. The severely reduced cell number in Bap55 mutants suggests that Bap55 is essential for neuroblast proliferation or neuronal survival. In the anterodorsal and ventral neuroblasts, PN numbers are not drastically changed; thus, the phenotypes indicate that Bap55 is important for dendrite targeting in multiple classes of PNs.

In WT, Mz19-GAL4 labels a subset of the GH146-GAL4 labeling pattern. It labels a small number of PNs derived from two neuroblasts, which can be clearly identified in WT clones generated in newly hatched larvae. Anterodorsal neuroblast clones target their dendrites to the VA1d glomerulus (Figure 3A), as well as the DC3 glomerulus residing immediately posterior to VA1d (not easily visible in confocal stacks). Lateral neuroblast clones target all dendrites to the DA1 glomerulus (Figure 3B). Unlike GH146-GAL4, WT Mz19-GAL4 never labels ventral neuroblast clones because it is not normally expressed in those cells (Figure 3C).

In Bap55 mutant PN clones, however, Mz19-GAL4 labels additional PNs in anterodorsal, lateral, and ventral clones (Figures 3D-F) compared to their WT counterparts (Figures 3A-C). This phenotype suggests that some Mz19-negative PNs now express Mz19-GAL4. In anterodorsal clones, Mz19-GAL4 labels additional cells, although not as many as are labeled by GH146-GAL4. The PNs also mistarget their dendrites to the anterior antennal lobe (Figure 3D), similar to mutant GH146-GAL4 anterodorsal neuroblast clones (Figure 2H). WT lateral neuroblast clones normally contain GH146-positive PNs and GH146-negative local interneurons 25 . In Bap55 ^-/- lateral neuroblast clones, Mz19-GAL4 predominantly labels local interneurons that send their processes to many glomeruli throughout the antennal lobe (Figure 3E) and do not send axon projections to higher brain centers. Lateral clones also show ectopic PN labeling with a lower frequency (data not shown). The Bap55 mutant appears to cause derepression of Mz19-GAL4, resulting in labeled local interneurons. Ventral neuroblast clones are never labeled in WT Mz19-GAL4 (Figure 3C), yet are labeled in Bap55 mutants (Figure 3F). This further indicates a derepression of the Mz19-GAL4 labeling pattern.

Unlike GH146-GAL4, WT Mz19-GAL4 never labels single cell clones when clone induction is performed in newly hatched larvae (Figure 3G, H). This is because Mz19-GAL4 is not expressed in the DL1 PN, the only GH146-positive cell generated during this heat shock time of clone generation. However, in Bap55 mutant PN clones, Mz19-GAL4 ectopically labels single cell anterodorsal PN clones targeting to the DA4l glomerulus (Figure 3I), which show an L-shaped pattern in the lateral horn with branches in the mushroom body calyx (Figure 3J), similar to GH146-GAL4 labeling (Figures 2C, D). The simplest interpretation is that this compound phenotype reflects first a derepression of Mz19-GAL4 in the DL1 PN, and second a DL1 to DA4l mistargeting phenotype in Bap55 mutants.

To test whether Bap55 functions in the neuroblast or postmitotically in PNs, we used GH146-GAL4, which expresses only in postmitotic PNs 7 , to express UAS-Bap55 in a Bap55 ^-/- single cell clone. We show that the dendrite mistargeting phenotype is rescued to the WT DL1 glomerulus (Figure 4A, Table 1) and conclude that Bap55 functions postmitotically to regulate PN dendrite targeting. The axon phenotype remains the stereotypical L-shaped pattern (Figure 4B).

However, in 2 out of 21 cases, expression of UAS-Bap55 in a Bap55 ^-/- single cell clone resulted in a de novo phenotype. The PN dendrites targeted to neither the DA4l nor the DL1 glomeruli, but to the DM6 glomerulus in the anterior medial region of the antennal lobe (Figure 4C, Table 1). In addition, the axon showed a mistargeting phenotype, extending ventrally to the lateral horn (Figure 4D). The two cases showed correlated DM6 dendrite and ventral axon mistargeting; the remaining 19 out of 21 cases showed full DL1 rescue and an L-shaped axon pattern. Expression of UAS-Bap55 in a WT single cell clone, however, did not cause any phenotype (n = 16; Table 1).

The Drosophila Bap55 protein is 70% similar and 54% identical to human BAF53a and 71% similar and 55% identical to human BAF53b. BAF53a and b are 91% similar and 84% identical to each other. Using GH146-GAL4 to express human BAF53a or b in a Bap55 ^-/- single cell clone, we found that the human homologs can effectively rescue the Bap55 ^-/- dendrite mistargeting phenotype (Figures 4E-H). Interestingly, both also cause the de novo DM6 dendrite and ventral axon mistargeting phenotypes in 6 out of 19 cases for BAF53a and 2 out of 32 cases for BAF53b. These phenotypes are quantified in Figure 4I and summarized in Table 1. Thus, human BAF53a and b can largely replace the function of Drosophila Bap55 in PNs.

To address whether Bap55 functions as a part of the BRM complex in PN dendrite targeting, we tested two additional BRM complex mutants for their PN dendrite phenotypes. We first tested Brahma (brm), the catalytic ATPase subunit of the BRM complex, which is required for the activation of many homeotic genes in Drosophila 26 (Figure 1D). Null mutations have been shown to decrease cell viability and cause peripheral nervous system defects 27 . RNA interference knockdown of brm in embryonic class I da neurons revealed dendrite misrouting phenotypes, although not identical to the Bap55 phenotype 15 . The human homologs of brm, BRM and BRG1 (Brahma-related gene-1), both have DNA-dependent ATPase activity. Inactivation of BRM in mice does not yield obvious neural phenotypes, but inactivation of BRG1 in neural progenitors results in reduced proliferation. BRG1 is likely to be required for various aspects of neural development, including proper neural tube development 28 .

In PNs, brm mutants displayed anterodorsal single cell clone mistargeting to non-stereotyped glomeruli throughout the antennal lobe, with each clone differing from the next (Figure 5A; n = 16). This is in contrast to the highly stereotyped DA4l mistargeting of Bap55 mutants. brm ^-/- neuroblast clones also displayed phenotypes where dendrites make small, meandering projections throughout the antennal lobe, which does not resemble the Bap55 ^-/- phenotype (Figures 5B-D). They additionally exhibit a perturbed cell morphology phenotype, which is markedly more severe than the Bap55 mutant phenotype.

We next tested Snr1, a highly conserved subunit of the BRM complex (Figure 1E). It is required to restrict BRM complex activity during the development of wing vein and intervein cells 29 and functions as a growth regulator 30 31 . Its human homolog, SNF5, is strongly correlated with many cancers 32 , yet little is known about its specific role in neurons.

In PNs, Snr1 mutants displayed similar phenotypes to brm mutants. The single cell clones displayed mistargeting to non-stereotyped glomeruli throughout the antennal lobe, with each clone demonstrating a unique phenotype (Figure 5E; n = 31). The neuroblast clones exhibited small meandering dendrites throughout the antennal lobe (Figures 5F-H), which also showed extremely perturbed cell morphology. These results, especially the non-sterotyped single cell clone phenotypes, indicate that the BRM complex functions differently from Bap55 in controlling PN dendrite targeting.

We further analyzed brm and Snr1 mutants with Mz19-GAL4 to determine whether their phenotypes resembled the Bap55 mutant phenotype of derepression. We were unable to generate any labeled clones, indicating one of three possibilities: increased cell death, severe cell proliferation defects, or repression of the Mz19-GAL4 labeling pattern. In any of the three cases, the phenotype does not resemble the Bap55 ^-/- mutant phenotype of abnormal activation of Mz19-GAL4 in single cell or neuroblast clones, indicating that the BRM complex functions differently from Bap55 in PNs.

In the same screen in which we identified the Bap55 mutation, we also independently identified LL05537, a mutation in dom that resulted in a qualitatively similar phenotype to Bap55 mutants. dom ^-/- DL1 PNs split their dendrites between the posterior glomerulus DL1 and the anterior glomerulus DA4l (Figure 6A). Their axons exhibit a WT L-shaped pattern in the lateral horn (Figure 6B).

The LL05537 allele contains a piggyBac insertion in an intron just upstream of the translation start of dom (Figure 1F). Because the piggyBac insertion contains splice acceptor sites and stop codons in all three coding frames 22 , this allele likely disrupts all isoforms of dom. Similarly to Bap55, we identified the piggyBac insertion site using inverse PCR 22 and Splinkerette PCR 24 . Precise excision of the piggyBac insertion reverted the dendrite targeting phenotype, confirming that disruption of the dom gene causes the dendrite mistargeting (Table 1). In addition, a BAC transgene that contains the entire dom transcriptional unit rescued the dom ^-/- mutant phenotypes (Figure 6E).

Dom is the catalytic DNA-dependent ATPase of the TIP60 complex and has been shown to contribute to a repressive chromatin structure and silencing of homeotic genes. Dom is a member of the SWI/SNF family and its ATPase domain is highly similar to the Drosophila Brahma and human BRG1 ATPase domains 20 . The human homolog of Dom is p400, which is important for regulating nucleosome stability during repair of double-stranded DNA breaks 33 and an important regulator of embryonic stem cell identity 34 .

To determine whether Bap55 and Dom genetically interact, we expressed UAS-Bap55 in a dom ^-/- PN. This manipulation did not significantly alter the dendrite phenotype (Figures 6C, D, quantified in Figure 6E and Table 1; P > 0.05 using two-way ANOVA with a Bonferroni posttest comparison across all columns). The axon branching pattern also was not altered.

We also examined another component of the TIP60 complex, E(Pc) (Figure 1G). In Drosophila, E(Pc) is a suppressor of position-effect variegation 21 35 and heterozygous mutations in E(Pc) result in an increase in homologous recombination over nonhomologous end joining at double-stranded DNA breaks 36 . Following ionizing radiation, heterozygous animals also exhibit higher genome stability and lower incidence of apoptosis 36 . Yet little is known about its role in neurons.

In our study, we find that E(Pc) ^-/- DL1 PN dendrites also mistarget to the anterior glomerulus DA4l (Figure 7A) and exhibit the stereotyped L-shaped axon pattern in the lateral horn (Figure 7B). A BAC transgene that contains the entire E(Pc) transcription unit rescued the E(Pc) mutant phenotypes (Figure 7E). To determine whether Bap55 and E(Pc) genetically interact, we expressed UAS-Bap55 in an E(Pc) ^-/- DL1 PN. This manipulation caused the dendrites to split between the DA4l and DM6 glomeruli (Figure 7C), and resulted in axons targeting ventrally to the lateral horn (Figure 7D, E, Table 1).

Neuroblast clones mutant for dom also exhibit dendrite mistargeting phenotypes to inappropriate glomeruli throughout the antennal lobe. Anterodorsal and lateral neuroblast clones show a very mild reduction in cell number and their dendrites do not target to the full set of proper glomeruli (Figure 8A, B). Ventral neuroblast clones, when compared to WT, exhibit incomplete targeting throughout the antennal lobe (Figure 8C).

Further analysis of dom mutants by labeling with Mz19-GAL4 revealed the same derepression as in Bap55 mutants (Figure 3). dom mutant Mz19-GAL4 PN clones also label anterodorsal, lateral, and ventral neuroblast clones (Figures 8D-F) with phenotypes similar to GH146-GAL4 labeled neuroblast clones (Figures 8A-C). In anterodorsal and lateral neuroblast clones, Mz19-GAL4 labels a large number of PNs that target to many glomeruli throughout the antennal lobe, although the cell number is smaller than GH146-GAL4 labeling (Figures 8D, E). Ventral neuroblast clones are never labeled in WT Mz19-GAL4 (Figure 3C), yet are labeled in dom mutants (Figure 8F). Mz19-GAL4 also labels single cell clones that split their dendrites between the DA4l and DL1 glomeruli (Figure 8G) and form the stereotypical L-shaped axon pattern in the lateral horn (Figure 8H). As in Bap55 mutants, this compound phenotype likely results from ectopic labeling of a DL1 PN, which further mistargets to DA4l.

The E(Pc) phenotypes in GH146 and Mz19-GAL4 labeled neuroblast clones (Figure 8I-N), as well as Mz19-GAL4 labeled single cell clones (Figure 8O, P) displayed similar phenotypes to dom as described above (Figure 8A-H). The phenotypic similarities in single cell clone dendrite mistargeting and derepression of a PN-GAL4 in mutations that disrupt Bap55, dom and E(Pc) strongly suggest that these three proteins act together in the TIP60 complex to regulate PN development.

A MARCM-based mosaic screen was performed using piggyBac insertional mutants previously described and currently available at the Kyoto Drosophila Genetic Resource Center 22 . The insertion sites of piggyBac mutants in Bap55 (LL05955) and domino (LL05537) were confirmed by inverse PCR 22 , precise excision, and Splinkerette PCR 24 .

The brm ² , Snr1 ⁰¹³¹⁹ , and E(Pc) ¹ alleles were obtained from the Bloomington Drosophila Stock Center and Snr1 ^6C was kindly provided by JA Simon 40 . We recombined the alleles onto FRT-containing chromosomes using standard techniques.

The GH146-GAL4 transgene has been previously described 41 . An insertion on the fourth chromosome was used in all MARCM experiments on 2R (Figures 2, 4, 6, 7, 8A-C, I-K).

MARCM was performed as described and flies were raised at 25°C 42 . Fly brains of both genders were dissected, fixed, and stained as described 43 . Antibody conditions: rat anti-mCD8α 1:100 (Invitrogen Caltag #RM2200 or #MCD0800, Carlsbad, CA, USA), mouse anti-nc82 1:40 (Developmental Studies Hybridoma Bank #nc82, contributed by E Buchner, Iowa City, IA, USA).

To generate UAS-Bap55, a full length cDNA (LD29458, Berkeley Drosophila Genome Project Gold cDNA, Drosophila Genomics Resource Center, Bloomington, IN, USA) was amplified using the following primers (5'-3'): CACCCAAAATGAGTGGCGGCACCATGCTATATG and TTACGGACACTTCCGCTCGACTTGG. The first primer amplifies from the 5' end and adds a CACC overhang for the TOPO reaction and a Kozak sequence. The second primer amplifies to the stop codon. The PCR product was subcloned into pENTR-D/TOPO (Invitrogen) and recombined into the destination vector pUAST-Gateway-attB 8 using LR Clonase II (Invitrogen).

The final construct was integrated into the 86Fb landing site 44 on the third chromosome.

To generate UAS-BAF53a and UAS-BAF53b, full length cDNAs (kind gifts of GR Crabtree 17 ) were amplified using the following primers (5'-3'): for UAS-BAF53a, CACCCAAAATGAGCGGCGGCGTGTACGGCG and TCAAGGGCACTTTCTTTCTACACACTG; for UAS-BAF53b, CACCCAAAATGAGCGGGGGCGTCTACGGCG and TCAGGGGCACTTCCGCTCCACGCACTG. The first primer amplifies from the 5' end and adds a CACC overhang for the TOPO reaction and a Kozak sequence. The second primer amplifies to the stop codon. The PCR product was subcloned into pENTR-D/TOPO (Invitrogen) and recombined into the destination vector pUAST-Gateway-attB 8 using LR Clonase II (Invitrogen).

The final construct was integrated into the 86Fb landing site 44 on the third chromosome.

To generate BAC transgenic flies containing dom, BAC #CH321-01P07 in the attB-P[acman]-CmR vector from the BACPAC Resources Center (Oakland, CA, USA) was verified and sent to BestGene, Inc. (Chino Hills, CA, USA) for integration into the 86Fb landing site 44 on the third chromosome.

BAC transgenic flies containing E(Pc) were generated in the same manner, using BAC #CH322-140G22.