|Home | About | Journals | Submit | Contact Us | Français|
Compared to the mechanisms of axon guidance, relatively little is known about the transcriptional control of dendrite guidance. The Drosophila olfactory system with its stereotyped organization provides an excellent model to study the transcriptional control of dendrite wiring specificity. Each projection neuron (PN) targets its dendrites to a specific glomerulus in the antennal lobe and its axon stereotypically to higher brain centers. Using a forward genetic screen, we identified a mutation in Rpd3 that disrupts PN targeting specificity. Rpd3 encodes a class I histone deacetylase (HDAC) homologous to mammalian HDAC1 and HDAC2. Rpd3−/− PN dendrites that normally target to a dorsolateral glomerulus mistarget to medial glomeruli in the antennal lobe, and axons exhibit a severe overbranching phenotype. These phenotypes can be rescued by postmitotic expression of Rpd3 but not HDAC3, the only other class I HDAC in Drosophila. Furthermore, disruption of the atypical homeodomain transcription factor Prospero (Pros) yields similar phenotypes, which can be rescued by Pros expression in postmitotic neurons. Strikingly, overexpression of Pros can suppress Rpd3−/− phenotypes. Our study suggests a specific function for the general chromatin remodeling factor Rpd3 in regulating dendrite targeting in neurons, largely through the postmitotic action of the Pros transcription factor.
The Drosophila olfactory system with its stereotyped organization provides an excellent model to study wiring specificity. Each olfactory projection neuron (PN) targets its dendrites to one of ~50 glomeruli in the antennal lobe to make synaptic connections with a specific class of olfactory receptor neurons, and sends its axon stereotypically to higher brain centers (Jefferis et al., 2001; Marin et al., 2002; Wong et al., 2002). Previous studies have identified several transcription factors that act cell-autonomously to regulate PN dendrite targeting specificity (Komiyama et al., 2003; Zhu et al., 2006; Komiyama and Luo, 2007; Spletter et al., 2007). We report here a genetic screen for additional factors that regulate PN dendrite wiring specificity, and identify Rpd3 (homolog of mammalian HDAC1/2) and the transcription factor Prospero (Pros) as additional regulators of PN dendrite wiring specificity.
HDACs are best studied for their role in deacetylating lysines on histones, resulting in a more compact chromatin structure correlated with gene repression (Thiagalingam et al., 2003). In the nervous system, HDACs have been most intensely studied in neurodegeneration. For instance, inhibition of HDACs can improve memory in mouse models of neurodegenerative diseases (Fischer et al., 2007) and can reduce polyglutamine-dependent neurodegeneration in Drosophila (Steffan et al., 2001). There are three major classes of HDACs, grouped based on their homology to yeast histone deacetylases: class I HDACs bear high homology to the yeast RPD3 protein, class II to Hda1, and class III to Sir2 (Thiagalingam et al., 2003). In Drosophila, Rpd3 and HDAC3 are the only two class I HDACs (De Rubertis et al., 1996; Johnson et al., 1998). They are present in the Drosophila nucleus and function in chromatin remodeling (Foglietti et al., 2006). In addition, RNAi knockdown of Rpd3 in embryos causes increased dendrite arborization in class I da neurons (Parrish et al., 2006). The role of HDACs in wiring specificity has not been characterized.
Prospero is an evolutionarily conserved atypical homeodomain transcription factor originally identified in Drosophila for its role in regulating homeobox gene expression in the developing central nervous system (Doe et al., 1991). Pros has been shown to regulate axon pathfinding (Doe et al., 1991), axon outgrowth (Vaessin et al., 1991) and dendrite morphogenesis (Gao et al., 1999). During asymmetric cell division of embryonic neuroblasts (Figure 1A), pros is only transcribed in neuroblasts (Broadus et al., 1998), and Pros protein is asymmetrically segregated into the nucleus of ganglion mother cells (Knoblich et al., 1995; Spana and Doe, 1995). Pros is thought to act in ganglion mother cells to repress neuroblast genes required for self renewal while activating genes for terminal differentiation (Choksi et al., 2006). Although several studies show that Pros is downregulated in embryonic postmitotic neurons (Doe et al., 1991; Vaessin et al., 1991; Choksi et al., 2006), recent studies show that Pros is present in some larval neurons (Guenin et al., 2007; Guenin et al., 2010). Our study reveals a function for Rpd3 in regulating dendrite targeting in neurons, largely through the postmitotic action of the Pros transcription factor.
MARCM-based mosaic screen and SNP mapping leading to the identification of Rpd312-37 were analogous to procedures previously described (Chihara et al., 2007). After narrowing the region to ~67 kb, all open reading frames and splice acceptor and donor sites were sequenced.
pros38 was isolated from a MARCM-based P-element forward genetic screen. pros38 possesses a p[mCD8-GFP, y+] transposon inserted early in the open reading frame of prospero, at amino acid 188. The insertion site was confirmed by inverse PCR.
MARCM was performed as described (Wu and Luo, 2006a). Fly brains of both genders were dissected, fixed, and stained as described (Wu and Luo, 2006b). Antibody conditions: rabbit anti-Rpd3 1:1000 (gift from J.T. Kadonaga), rabbit anti-acetylated lysine 1:1000 (Cell Signaling Technology #9441), mouse anti-Prospero 1:4 (Developmental Studies Hybridoma Bank #MR1A, C.Q. Doe), rat anti-mCD8α 1:100 (Invitrogen Caltag #RM2200), and mouse anti-nc82 1:40 (Developmental Studies Hybridoma Bank #nc82, E. Buchner).
DL targeting (Figures (Figures1I,1I, ,5D,5D, ,6C)6C) includes full innervation of the DL1 glomerulus, as well as minor dendrite spillover into nearby glomeruli in the dorsolateral quadrant of the antennal lobe.
Confocal stacks of axons were blindly rank ordered by two independent observers with consistent results. Axons were ranked by degree of axon overbranching, with the lowest score denoting the least branching (closest to wild-type) (Figures (Figures2D,2D, ,5H,5H, ,6F6F).
Confocal images were taken with all pixel values in the linear range. Background fluorescence was subtracted from the average pixel intensities in the DL1 PN nucleus and the highest Pros expressing nucleus in the section. Ratio of DL1 to highest intensity was calculated as (DL1-background)/(highest Pros-background) (Figure 6K).
To generate UAS-Rpd3-V5, we obtained Rpd3-V5 cDNA from K.T. Min (NIH, USA) (Cho et al., 2005). The Rpd3-V5 fragment was amplified using the following primers (5′-3′): GGGGTACCCCAAAATGCAGTCTCACAGCAAAAAGCGCG and GACTAGTCTACGTAGAATCGAGACCGAGGAGAGGGTTAGG. The first primer adds a KpnI site and a Kozak sequence, and the second primer adds a stop codon and an SpeI site to the amplified Rpd3-V5 fragment. The PCR product was subcloned into the pUAST vector (KpnI and XbaI). Germline transformation was performed using standard P-element transformation; an insertion on the 2nd chromosome was used in all experiments.
To generate UAS-HDAC3-3xFLAG, a full length cDNA (LD23745, Berkeley Drosophila Genome Project Gold cDNA, Drosophila Genomics Resource Center, Bloomington, IN) was amplified using the following primers (5′-3′): CACCCAAAATGACGGACCGTAGGGTGTC and CTACTTGTCATCGTCATCCTTGTAATCGATGTCATGATCTTTATAATCACCGTCATGG TCTTTGTAGTCACTTTCTGCCGAATCGGGCTTGTCTTG. The first primer amplifies from the 5′ end and adds a CACC overhang for the TOPO reaction and a Kozak sequence. The second primer adds a C-terminal 3xFLAG tag. The PCR product was subcloned into pENTR-D/TOPO (Invitrogen) and recombined into the destination vector pUAST-Gateway-attB (described below) using LR Clonase II (Invitrogen).
Another UAS-HDAC3 was constructed similarly with a C-terminal V5 tag an alternative second primer (5′-3′): CTACGTAGAATCGAGACCGAGGAGAGGGTTAGGGATAGGCTTACCACTTTCTGCCGAATCGGGCTTGTCTTG. This construct yielded similar results to the 3xFLAG tagged construct (data not shown).
Both UAS-HDAC3 constructs were integrated into the P24 landing site (Markstein et al., 2008) on the 2nd chromosome.
The pUAST-Gateway-attB destination vector was created by PCR amplifying the Gateway cassette from pBPGUw (Pfeiffer et al., 2008) using the following primers (5′-3′): CACCTCGAGGTATCACGAGGCCCTTTC and CACTCTAGACTCGGCCGGCCGTTTATCAC. The first primer amplifies from the 5′ end and adds an XhoI site. The second primer adds an XbaI site to the 3′ end. The PCR product was cut and ligated into pUAST-attB (XhoI and XbaI) (Bischof et al., 2007).
UAS-1xFLAG-pros-L DNA was obtained from F. Matsuzaki. We amplified approximately 4.5kb of the 5′ side of pros-L using the following primers (5′-3′): GAAGATCTTCATGAGTAGCGATTACAAGGATGATG and CTCCCGCAGAGTCGATTCGACCACG. The first primer adds a BglII site to the 5′ end. The amplified fragment was digested with BglII and KpnI. The fragment was then subcloned into a pUAST vector (BglII and KpnI), which already contained the N-terminal 3xFLAG fragment between EcoRI and BglII sites. Then we amplified approximately 2.5kb of the 3′ side of pros-L, digested the fragment with KpnI and XbaI, and ligated with the plasmid above to obtain UAS-3xFLAG-pros-L.
UAS-3xFLAG-pros-L was used for transfection of S2 cell culture (see below).
For all genetic experiments, flies containing an insertion of UAS-pros-L on the X chromosome were kindly provided by C.Q. Doe.
Fly embryos were collected on grape plates for 14 hours, washed, and homogenized using a dounce homogenizer in RIPA buffer (150mM NaCl, 0.5% deoxycholic acid, 0.1% SDS, 50mM Tris pH 8.0, protease inhibitor cocktail (1:100, Sigma-Aldrich #P8340), 150nM trichostatin A (Sigma-Aldrich #T1952), 5mM sodium butyrate (Sigma-Aldrich #B5887), 10mM niacinamide (Sigma-Aldrich #N0636), and with or without 1% Triton-X). Immunoprecipitation was performed using anti-Pros antibody bound to Protein G beads.
The final precipitate was denatured in SDS sample buffer and protein was run on a standard SDS-PAGE gel, transferred, and blotted using anti-Rpd3 or anti-acetylated lysine antibody.
S2 cells were raised in Shields and Sang M3 Insect Medium (Sigma-Aldrich #S8398) supplemented with potassium bicarbonate, penicillin, streptomycin, and 10% fetal bovine serum. Cells were split at 24 hours prior and immediately before co-transfection of tubP-GAL4 and UAS-3xFLAG-pros-L using Effectene Transfection Reagent (Qiagen #301425). S2 cells endogenously express Rpd3.
48 hours after transfection, HDAC inhibitors were added for 4 hours at the following final concentrations: trichostatin A 150nM and sodium butyrate 5mM. Cells were lysed in the following immunoprecipitation buffer: 150mM NaCl, 1% Triton-X, 2mM EDTA, 50mM Tris pH 8.0, 10% glycerol, protease inhibitor cocktail 1:100, 150nM trichostatin A, and 5mM sodium butyrate. Immunoprecipitation was performed using anti-Pros antibody bound to Protein G beads.
The final precipitate was denatured in SDS sample buffer, run on a standard SDS-PAGE gel, transferred, and blotted using anti-Pros, anti-Rpd3, or anti-acetylated lysine antibody. Alternatively, the final precipitate was denatured in SDS sample buffer, run on a standard SDS-PAGE gel, stained using GelCode Blue Stain Reagent (Thermo Scientific #24590), and the Pros band cut from the gel for mass spectrometry. Mass spectrometry was performed by NextGen Sciences (Ann Arbor, MI) to determine the acetylation state of Pros.
To identify genes that are essential for dendrite wiring specificity in Drosophila olfactory projection neurons (PNs), we performed a MARCM-based forward genetic screen using ethyl methanesulfonate as mutagen (Chihara et al., 2007). MARCM allows visualization and genetic manipulation of single cell or neuroblast clones in an otherwise heterozygous animal, allowing the study of essential genes in mosaic animals (Figure 1A) (Lee and Luo, 1999). We used GH146-GAL4 to label a single PN born at newly hatched larva (Jefferis et al., 2001), which in wild-type animals always projects its dendrites to the dorsolateral glomerulus DL1 in the antennal lobe (Figure 1C). We identified a mutant, called 12-37, in which DL1 PNs mistargeted towards dorsomedial or ventromedial regions of the antennal lobe (Figure 1D). SNP and deletion mapping identified the causal gene to be Rpd3, encoding a homolog of mammalian HDAC1 and HDAC2. The mutant 12-37 causes a G136R missense mutation in the catalytic domain; the glycine at this position is conserved from yeast to humans (Figure 1B).
We confirmed that the mistargeting phenotype was caused by the mutation in Rpd3 using the following two criteria. First, MARCM single cell clones of two previously existing Rpd3 mutants (Figure 1B) (Mottus et al., 2000) gave similar mistargeting phenotypes (Figures 1E, F). Second, MARCM expression of UAS-Rpd3 (see Materials and Methods) in DL1 single cell clones using GH146-GAL4 significantly rescued the phenotype for all three Rpd3 alleles (Figures 1G, H, data for Rpd3def24 not shown), especially at higher temperatures where GAL4 activity increases (Figure 1I). Because GH146-GAL4 expresses UAS-Rpd3 only in postmitotic PNs (Spletter et al., 2007), we conclude that Rpd3 plays an essential role in postmitotic PNs to regulate dendrite targeting.
Rpd3 is ubiquitously expressed and located in the nucleus of all cells surrounding the developing antennal lobe at 24 hours after puparium formation (24hAPF; Figure 1J), when wiring specificity in the antennal lobe is being established. Rpd3 immunostaining is absent in single cell Rpd312-37 MARCM clones (Figure 1K), suggesting that the G136R mutation destabilizes the Rpd3 protein. Rpd3 protein expression is restored by MARCM-mediated activation of UAS-Rpd3 (Figure 1L). Neuroblast clones homozygous mutant for Rpd3 resulted in a 50% reduction in the number of PNs (WT cell bodies 33.8 ± 2.0; Rpd3−/− 14.9 ± 1.2, mean ± standard deviation, n=8 each), indicating that Rpd3 is essential for the proliferation or survival of neuroblasts. Postmitotic expression of Rpd3 does not rescue the reduced cell number phenotype. This is consistent with a previously described role for Rpd3 in cell proliferation (Zhu et al., 2008).
To further study the role of Rpd3 in postmitotic neuronal development, we examined the terminal arborization patterns of DL1 PN axons in the lateral horn. Wild-type DL1 PNs have a stereotypical L-shaped axon projection in the lateral horn, with one dorsal collateral and a major branch extending laterally (Figure 2A) (Marin et al., 2002). Rpd312-37 (hereafter referred to as Rpd3−/−) DL1 PNs exhibit an axon overbranching phenotype in the lateral horn (Figure 2B), regardless of where dendrites mistarget. This is evidence against the possibility that loss of Rpd3 simply causes a PN fate switch. This axon overbranching phenotype can be rescued by postmitotic expression of UAS-Rpd3 (Figure 2C and D). Thus, Rpd3 functions cell autonomously in PNs to limit the arborization of axon terminal branches.
To test whether the only other class I HDAC in Drosophila is also involved in PN development, we generated MARCM clones mutant for HDAC3N, an early stop and presumably null mutation (Zhu et al., 2008). HDAC3−/− neuroblast clones exhibited a 20% reduction in cell number (26.4 ± 1.8, n=8), suggesting that HDAC3 is also required for PN neuroblast proliferation or survival. This is consistent with previous findings that HDAC3 is involved in suppression of apoptosis (Zhu et al., 2008). However, HDAC3−/− DL1 PNs exhibited normal dendrite targeting (Figure 3A) and axon arborization in the lateral horn (Figure 3B), indicating that HDAC3 is not required in postmitotic PNs for dendrite targeting or axon arborization.
To determine whether Rpd3 and HDAC3 have distinct roles, we tested the rescue of Rpd3−/− PN phenotypes by increasing HDAC3 expression. MARCM-mediated expression of UAS-HDAC3 was confirmed by nuclear staining of HDAC3-3xFLAG as revealed by anti-FLAG staining (Figure 3E, see Materials and Methods), but HDAC3 expression did not rescue the dendrite targeting phenotype (Figure 3C, compared to 1D1) or the axon arborization phenotype (Figure 3D, compared to to2B)2B) of Rpd3−/− DL1 PNs. Similar results were obtained for PN neuroblast clones (Figures 3F-J). Thus, HDAC3 cannot replace Rpd3 function in PN dendrite targeting and axon arborization.
Histone deacetylases are best studied for their roles in deacetylating lysine residues on histones, which are the most abundant substrates for HDACs (Thiagalingam et al., 2003). To examine the effect of Rpd3 on lysine acetylation in PNs, we compared the levels of acetylated lysine immunoreactivity in nuclei of PNs, which are labeled by mCD8-GFP via MARCM, with nuclei of neighboring cells. In wild-type controls, we found anti-acetylated lysine immunoreactivity at a low level in all nuclei (Figure 4A). In contrast, Rpd3−/− PN nuclei exhibit a much higher level of acetylated lysine compared to neighboring nuclei (Figure 4B). This finding indicates that Rpd3 plays a predominant role in deacetylation of lysines in PNs. The elevated level of acetylated lysine in Rpd3−/− PNs can be reduced to wild-type low levels upon postmitotic expression of Rpd3 (Figure 4C). Interestingly, HDAC3−/− PN cell bodies exhibited levels of acetylated lysine indistinguishable from control cells (Figure 4D), suggesting a minor role of HDAC3 in PN histone deacetylation. Expression of HDAC3 cannot detectably compensate for loss of Rpd3, as exhibited by the remaining high level of acetylated lysine in the nucleus (Figure 4E). Similar results were obtained for PN neuroblast clones (Figures 4F-J). Together these data indicate that Rpd3 is the major histone deacetylase in PNs, and that its function cannot be compensated for by the only other class I HDAC in the fly.
In a parallel screen for PN mistargeting phenotypes, we found that a P-element insertion early in the open reading frame of prospero yielded a homozygous clonal phenotype similar to Rpd3 mutants. Comparing with wild-type DL1 PNs, pros−/− DL1 PNs also project to medial glomeruli (Figures 5A, B, D). Interestingly, pros−/− DL1 PNs also exhibit an axon overbranching phenotype, similar to Rpd3−/− PNs (Figures 5E, F, H). The similarity of the phenotypes of Rpd3−/− and pros−/− PNs suggests that the two proteins might function in the same pathway in PNs.
Three lines of evidence indicate that Pros expression and function in the PN lineage differs from its function in the embryonic lineage previously described (see Introduction). First, Pros is clearly present in postmitotic PNs. Immunostaining for Pros at 24hAPF shows that wild-type postmitotic PNs express varying amounts of Pros protein (Figure 5L); this variable expression persists throughout development and adulthood (data not shown). Wild-type DL1 PNs express an intermediate level of Pros protein (Figure 5I). pros−/− DL1 or neuroblast clone PNs have an undetectable level of Pros protein in the cell body (Figure 5J and M), confirming both antibody specificity and the nature of the loss-of-function mutation. Pros immunoreactivity can be restored by MARCM expression of UAS-pros in postmitotic PNs (Figures 5K and N).
Second, the pros−/− single cell clone phenotype indicates that pros mRNA must be transcribed in postmitotic neurons and/or ganglion mother cells; it cannot be transcribed only in neuroblasts. There are two scenarios in which a single cell MARCM clone can be produced. The mitotic recombination could occur right before the ganglion mother cell divides to give rise to two postmitotic cells, and therefore only the postmitotic DL1 PN is pros−/− (Figure 1A, left). Alternatively, because the sibling of the DL1 PN dies during development (Lin et al., 2010; Potter et al., 2010), the mitotic recombination could also occur in the neuroblast; in this case, the ganglion mother cell giving rise to the DL1 PN is pros−/− (Figure 1A, center). In either case, the neuroblast remains pros+/−. Therefore pros must be transcribed either in the postmitotic PN or the ganglion mother cell to account for a single cell phenotype.
Third, postmitotic UAS-pros expression can rescue the mistargeting defect of pros−/− DL1 dendrites (Figures 5C and D) as well as overbranching of DL1 axons (Figures 5G and H), indicating that Pros functions in postmitotic neurons to regulate PN dendrite targeting and axon branching.
Given the similarity of their phenotypes, we sought to test the relationship between Rpd3 and Pros using genetic interactions. We expressed UAS-pros in Rpd3−/− PNs using MARCM, and found that expression of Pros suppresses the dendrite mistargeting phenotype of Rpd3 mutants (Figures 6A and C). In contrast, expressing UAS-Rpd3 did not suppress the pros−/− PN phenotype (Figure 6B). UAS-pros also partially but significantly suppresses the Rpd3−/− axon overbranching phenotype at higher temperatures where GAL4 activity increases (Figures 6D and F). Conversely, UAS-Rpd3 cannot suppress the pros−/− axon overbranching phenotype (Figure 6E). Overexpression of Pros in wild-type DL1 PNs does not cause any dendrite targeting defects, and Pros expression cannot suppress several other medially mistargeting mutants (data not shown). These data suggest that Pros acts downstream of Rpd3 to regulate PN dendrite targeting and to limit PN axon terminal arborization.
It has long been thought that histone deacetylases play a general role in chromatin remodeling and transcriptional control, and many studies have examined genome-wide patterns of histone modifications (Millar and Grunstein, 2006). Yet recent studies have also suggested that “general” chromatin remodeling factors can have very specific roles (Ho and Crabtree, 2010). In this study, we show a new function for Rpd3, a ubiquitously expressed protein that is the major histone deacetylase in Drosophila olfactory projection neurons. Rpd3 plays a specific role in controlling dendrite targeting and axon terminal branching, and this function cannot be replaced by the only other class I HDAC. Furthermore, we show that the majority of its function in regulating dendrite targeting and a portion of its function in regulating axon branching are likely carried out through the downstream transcription factor Prospero.
Although we cannot rule out the possibility that Rpd3 and Pros act in parallel pathways to regulate dendrite targeting and axon branching, several lines of evidence support the notion that Rpd3 acts via Pros to regulate these events. First, loss-of-function mutations in single cell clones of Rpd3 and pros give similar dendrite targeting and axon branching phenotypes. Second, overexpression of Pros can largely suppress the dendrite mistargeting phenotype of Rpd3−/− and can partially suppress the axon overbranching phenotype of Rpd3−/−. This suppression is specific, as overexpressing Pros does not cause defects in wild-type cells nor does it suppress mistargeting phenotypes due to a few other mutations. Conversely, overexpression of Rpd3 does not suppress pros mutant phenotypes. Because the suppression is more robust for dendrite mistargeting defects compared to axon overbranching defects, it is possible that Pros function accounts for more of Rpd3’s activity in regulating dendrite targeting than axon branching.
To address the mechanism by which Rpd3 and Pros function together in regulating PN development, we tested several models of their interactions. One model is that Rpd3 functions to upregulate the expression of Pros, which predicts that Rpd3−/− would lead to a decrease in Pros protein. However, anti-Pros immunoreactivity in Rpd3−/− DL1 PN clones was not decreased compared to wild-type DL1 PN clones (Figures 6G-K). A second model is that Rpd3 and Pros directly bind and work together to regulate the transcription of Pros target genes. Yet in wild-type embryos or in S2 cell culture, we could not detect a complex between Rpd3 and Pros via immunoprecipitation (Figure 6L). A third model is that Rpd3 directly deacetylates the Pros protein to affect its function, but we could not detect Pros acetylation in the presence of HDAC inhibitors by immunostaining or mass spectrometry (data not shown). Together, these data suggest that Rpd3 indirectly affects the function of Pros. This effect may be through posttranslational modification of Pros to modify its activity. For example, Pros has previously been shown to be phosphorylated (Srinivasan et al., 1998). If posttranslational modification increases Pros activity, then Rpd3−/− would result in reduced Pros activity. Overexpression of Pros in Rpd3−/− may compensate for the reduced activity of unmodified Pros, and therefore suppress the Rpd3−/− phenotype.
Future studies will determine how Rpd3 regulates Prospero, how these factors act together with other transcription factors (Komiyama and Luo, 2007), and what transcriptional target genes they regulate in order to orchestrate the developmental program for precise wiring of the olfactory circuit.
We thank J.T. Kadonaga and the Developmental Studies Hybridoma Bank for antibodies; C.Q. Doe, J.A. Simon, the Bloomington Drosophila Stock Center, and the Kyoto Drosophila Genetic Resource Center for fly stocks; K.T. Min and the Drosophila Genomics Resource Center for DNA; R.J. Watts, E.D. Hoopfer, and O. Schuldiner for collaboration on the EMS screen; D. Luginbuhl and E.J. Rao for technical assistance; S. Sekine and M. Miura for supporting T.C. to complete this work; D. Berdnik, Y.H. Chou, V.S. Dani, W. Hong, L. Liang, M.L. Spletter, L.B. Sweeney, B. Tasic, and X. Yu for comments on the manuscript. This work was supported by an NIH grant (R01-DC005982), a fellowship from the National Science Foundation (J.S.T.), the Grant-in-Aid for Scientific Research on Priority Areas “Molecular Brain Science” and “Dynamics of Extracellular Environments”, a grant from the Japan Society for the Promotion of Science, and the PRESTO, JST program (T.C.). L.L. is a Howard Hughes Medical Institute Investigator.