Abstract
In insects, many complex behaviors, including olfactory memory, are controlled by a paired brain structure, the so-called mushroom bodies (MB). In Drosophila, the development, neuroanatomy, and function of intrinsic neurons of the MB, the Kenyon cells, have been well characterized. Until now, several potential neurotransmitters or neuromodulators of Kenyon cells have been anatomically identified. However, whether these neuroactive substances of the Kenyon cells are functional has not been clarified yet. Here we show that a neuropeptide precursor gene encoding four types of short neuropeptide F (sNPF) is required in the Kenyon cells for appetitive olfactory memory. We found that activation of Kenyon cells by expressing a thermosensitive cation channel (dTrpA1) leads to a decrease in sNPF immunoreactivity in the MB lobes. Targeted expression of RNA interference against the sNPF precursor in Kenyon cells results in a highly significant knockdown of sNPF levels. This knockdown of sNPF in the Kenyon cells impairs sugar-rewarded olfactory memory. This impairment is not due to a defect in the reflexive sugar preference or odor response. Consistently, knockdown of sNPF receptors outside the MB causes deficits in appetitive memory. Altogether, these results suggest that sNPF is a functional neuromodulator released by Kenyon cells.
Introduction
In insects a wide range of behaviors including olfactory learning and memory rely on a bilaterally symmetric brain structure, the mushroom bodies (MB) (Heisenberg, 2003; Davis, 2011). These prominent neuropils consist of intrinsic neurons, the Kenyon cells. Their dendrites form a globular microglomerular complex called calyx, which receives the main olfactory input from the antennal lobes via olfactory projection neurons (Stocker et al., 1990). The bundled axons of the Kenyon cells form characteristic vertical and medial substructures, the lobes (Strausfeld et al., 2003). Despite the importance of the MB, a functional neurotransmitter or neuromodulator produced by the Kenyon cells has not been conclusively identified in any insect. An immunohistochemical screen for candidate neuroactive substances identified the precursor for short neuropeptides F (sNPF) and its processed peptides in a majority of the Kenyon cells, both in larval and adult Drosophila (Johard et al., 2008; Nässel et al., 2008). In most cases, neuropeptides act as neuromodulators (Nässel and Winther, 2010), but rarely as neurotransmitters in the strict sense. Here we quantified relative changes of sNPF levels after-induced activation of Kenyon cells and investigated the role of sNPF and its receptor in the Kenyon cells for olfactory memory. Our findings strongly suggest that sNPF in Kenyon cells of the MB plays an important role in olfactory learning and/or retrieval, probably as an output signal from these neurons.
Materials and Methods
Fly strains.
The following transgenic strains of Drosophila melanogaster were used for crosses: OK107-GAL4 (IV) (Connolly et al., 1996), MB247-GAL4 (III) (Zars et al., 2000), UAS-dicer2 (II) (Dietzl et al., 2007), UAS-sNPF-RNAi (X;II) (Lee et al., 2004), UAS-sNPF-RNAi (X), MB247-GAL80 (II) (Krashes et al., 2007), elav-GAL4 (X) (Lin and Goodman, 1994), UAS-sNPF (II;III) (Lee et al., 2004), UAS-sNPFR1-RNAi (III) (Hong et al., 2012), UAS-TrpA1 (II) (Hamada et al., 2008), NP3061-GAL4 (III) (Aso et al., 2009), and white (w1118). UAS-dicer2 was used together with MB247-GAL4 in all knockdown experiments and respective control experiments. All transgenes are inserted in w1118 with a wild-type CantonS background.
Immunohistochemistry.
Adult Drosophila heads were prepared for whole mount or cryosections as previously described (Nässel et al., 2008) with the following modifications. Thickness of optical and cryostat sections was 0.2–0.5 and 25 μm, respectively. To visualize synaptic neuropil regions, mouse monoclonal antibodies for either nc82 (Hofbauer et al., 2009) or Synapsin (Klagges et al., 1996) were used (both 1:10; Developmental Studies Hybridoma Bank, Iowa City, IA). For Figure 1, A and C, dissected brains were fixed for 30 min at room temperature in 4% formaldehyde in PBS with 0.1% Triton X-100 (PBS-Tx), washed with 0.3% PBS-Tx, and stained using antiserum to a sequence of the Drosophila sNPF (rabbit, 1:4000) and anti-Synapsin antiserum (mouse, 1:200). For detection of primary antiserum, Alexa 488-tagged goat anti-rabbit (1:1000; Invitrogen) and Cy3-tagged goat anti-mouse antiserum (1:250; Dianova) were used. Preparations were mounted in Vectashield (Vector Laboratories) and imaged with Olympus FV-1000 or Zeiss LSM 510 confocal microscopes using 20× (NA: 0.85, oil-immersion or 0.8, respectively) or 40× (NA: 1.0, oil immersion) objectives at an optical section thickness of 1.0 μm.
Thermo-activation of Kenyon cells.
Crosses for thermo-activation experiments were raised at 18°C. F1 progeny (18 to 13 d old) was put for 10 min in empty prewarmed vials in a 33°C water bath and afterward transferred into new vials at room temperature. After 5 min flies were cold anesthetized and immediately dissected. All genotypes were treated in parallel and under the same experimental conditions.
Quantification of immunofluorescence.
To quantify sNPF levels after thermo-activation of Kenyon cells, we measured sNPF immunofluorescence in single optical sections at the anterior end of the MB peduncle using the histogram tool of ImageJ, (National Institutes of Health, Bethesda, MD; http://rsb.info.nih.gov/ij) (Knapek et al., 2011). Fluorescence was quantified in a manually set region of interest (ROI) of the α/β-neurons that were activated by UAS-TrpA1 expression driven by NP3061-GAL4. To compensate for differences in immunofluorescence between different specimens the fluorescence of the thermo-activated areas (α/β-neurons) was normalized to the fluorescence of γ-neurons in the peduncle that are not part of the expression pattern of NP3061-GAL4. Fluorescence levels of sNPF after sNPF knockdown with MB247-GAL4 were measured in several manually set adjacent ROI that cover most of the targeted α- and γ-lobe spur and normalized to a group of neuronal cell bodies adjacent to the MB (in the same specimen) that are not included in the MB247-GAL4 and thus not targeted by RNA interferences (RNAi). All specimens were imaged under identical conditions.
Behavioral experiments.
Crosses were raised at 25°C. F1 progeny was collected and transferred to fresh food vials for at least 24 h and subsequently starved for 20–24 h in moistened vials. Mixed populations of 2- to 6-d-old males and females were trained and tested for 2 h appetitive olfactory memory as previously described (Tempel et al., 1983; Thum et al., 2007). Odors (4-Methylcyclohexanol and 3-Octanol; Fluka, diluted 1:80 and 1:100 in paraffin oil) were presented in odor cups with a diameter of 14 mm. A learning index was calculated as the mean preference of two separate groups of reciprocally trained flies. In half of the experiments the first presented odor was rewarded and in the other half, the second presented odor was rewarded (Knapek et al., 2010). Controls of odor responses were measured as previously described (Knapek et al., 2011). A negative value indicates avoidance from the odor, a performance index (PI) of zero indicates no response. Response to 2 m sucrose was measured in a previously described setup (Schnaitmann et al., 2010). Distribution of the flies was recorded for 240 s at one frame per second. A PI was calculated by subtracting the number of flies on the control side from the number of flies on the sucrose side, divided by the total number of flies. To avoid any potential side preference effects, the same number of tests was made with the sucrose on the left or on the right side of the arena.
Statistics.
Data were evaluated using the software Prism (GraphPad) as previously described (Kremer et al., 2010) using Kolmogorov–Smirnov and Bartlett's test, followed by one-way ANOVA or Kruskal–Wallis test. If significant at p < 0.05, Bonferroni- or Dunn-corrected pairwise comparisons were applied; *p < 0.05, **p < 0.01, and ***p < 0.001.
Results
Thermo-activation of Kenyon cells reduces sNPF-immunoreactive signals
The expression of sNPF transcript and sNPF peptide in Kenyon cells of MB was demonstrated by in situ hybridization, immunocytochemistry and an sNPF-Gal4 line (Johard et al., 2008; Nässel et al., 2008). Expression levels of the peptide differ in the MB, being strongest in the α/β-lobes, less in the γ-lobe, even weaker in the core of α/β-lobes, and not detectable in the α′/β′-lobes (Johard et al., 2008); Figs. 1A,B, 2A). Within the lobe regions sNPF peptide is distributed in a punctate fashion in the lobes and the γ-lobe spur (Fig. 1B,B′). These puncta probably represent peptide release sites and are frequently, but not consistently colocalized with an active zone protein, Bruchpilot (Fig. 1B,B′), suggesting that sNPF is commonly present near presynaptic sites throughout the lobes. Generally neuropeptides are stored in vesicles located in the perisynaptic area rather than directly in the nearby active zone (Salio et al., 2006), thus a perfect match between the two markers is not expected.
Decrease of sNPF upon activation of MB neurons. A, Double labeling of sNPF (magenta) and a ubiquitous synaptic marker Synapsin (Syn, blue) shows strong expression of the neuropeptide in the MB of a wild-type brain. Partial projection includes the anterior part of the brain. Scale bar, 50 μm. B, Double labeling of sNPF (magenta) and a synaptic marker Bruchpilot (BRP, green; nc82 antibody) to show relations between peptide and synapse distribution. γ-Lobe spur and β-lobe show punctate sNPF and Bruchpilot labeling, and the adjacent β′-lobe displays no peptide labeling. B′, Higher magnification of sNPF/Bruchpilot double labeling in the spur region of another optical section. sNPF distribution partly matches that of synapse distribution (e.g., three punctuates at arrow). Note that part of Bruchpilot puncta originate from extrinsic neurons. Scale bars: 10 μm. C, A representative slice for sNPF quantification is chosen at the anterior end of the peduncle of the GAL4 control (inset in A), showing distinct labeling of the α/β (arrow)- and γ-neurons (arrowhead). The outlined regions are used for quantification. Scale bar, 10 μm. D, Intensity ratio of sNPF signals in the α/β-neurons (labeled by NP3061-GAL4) and the γ-neurons of the Kenyon cells (control region, not in the NP3061-GAL4 expression pattern) after 10 min thermo-activation by UAS-TrpA1. The sNPF signal is significantly reduced in the α/β-neurons in the experimental group (**p < 0.01, n = 19–23, bars and error bars indicate medians and quartiles, respectively). E, Total intensity of sNPF signals in the γ-neurons of the Kenyon cells (control region) after 10 min at 33°C. The groups do not differ significantly (p > 0.05, same data as shown in D). Bars and error bars indicate mean and SEM, respectively.
Knockdown of sNPF in MB, but not overexpression, leads to impaired olfactory appetitive 2 h memory. A–E, Reduced sNPF-immunofluorescence is observed in the MB lobes in MB247/sNPF-RNAi (B) compared with control flies (MB247/+; A). Representative images are shown. We measured immunofluorescence in the α-lobe (α) and in the γ-lobe spur (s). Note that strong immunofluorescence of non-Kenyon cells (asterisks) is not affected in MB247/sNPF-RNAi. Magnification of the spur of a control brain (C) and a knockdown brain (D). A and B are projections of multiple MB confocal sections; C and D are single optical sections. Scale bars: A, B, 20 μm; C, D, 10 μm. E, Effect of the knockdown on the relative immunofluorescence in five specimens per genotype. Bars and error bars indicate mean and SEM, respectively. ***p < 0.001 in α-lobe, **p < 0.01 in γ-lobe spur. F, sNPF knockdown in the MB with UAS-sNPF-RNAi expressed either in all Kenyon cells (OK107) or the α/β- and γ-neurons (MB247) leads to a significantly impaired appetitive olfactory memory tested at 2 h after training (n = 10–38, **p < 0.01). Bars and error bars indicate mean and SEM, respectively. G, Expression of the GAL4 repressor GAL80 under control of the MB247 promoter in the Kenyon cells rescues the memory impairment of sNPF knockdown with OK107-GAL4 (n = 10–14, *p < 0.5). For this specific experiment we expressed a single copy of sNPF-RNAi (on X), which decreased memory performance to a similar level as in the previous experiments where two copies of RNAi were used. Bars and error bars indicate mean and SEM, respectively. H, In contrast to the knockdown, overexpression of sNPF in the MB with neither OK107-GAL4 nor MB247-GAL4 changes 2 h appetitive olfactory memory (n = 8–13). Bars and error bars indicate median and quartile range, respectively. I, Normal dynamics and maintenance of sugar responses of flies with OK107-GAL4 driving sNPF-RNAi compared with control flies. Preference of 100 means that all flies are on the sugar side. n = 8–10, means and SEM are presented. J, Normal response of the first 120 s of sugar presentation of OK107 or MB247 driving sNPF-RNAi (for OK107 same data as in G). Medians with interquartile range are shown.
Peptide levels can be quantified by immunohistochemistry, allowing to indirectly quantifying peptide release (Han et al., 1999; Park et al., 2000; Rao et al., 2001). We used antibody staining to measure sNPF signals after induced activation in a defined subset of Kenyon cells. To specifically activate the α/β-lobe neurons we expressed the thermosensitive cation channel dTRPA1 (Hamada et al., 2008) under the control of NP3061-GAL4 (Aso et al., 2009) and shifted temperature to 33°C for 10 min. To compensate for variability in immunofluorescence among different specimens, the fluorescence of the thermo-activated areas (α/β-neurons) was normalized to the fluorescence of a control region (γ-neurons) in a same confocal section of the anterior peduncle (Fig. 1C). This close internal control should cancel out any potential general effects of the temperature shift on neurons. We found that thermo-activation of Kenyon cells significantly reduced sNPF levels compared with control flies (Fig. 1C,D; differences in rank sum: 17.74 and 34.36, p < 0.01 and p < 0.001 for comparisons of TrpA1/NP3061 vs NP3061/+ and vs TrpA1/+, respectively. n = 19–23). The selected ROIs are very distinct (being located at the anterior part of the peduncle) and can be reproducibly identified among different samples. Nevertheless, we repeated the quantification by another uninformed experimenter, which yielded a similar result (data not shown). Levels of sNPF in the γ-neurons (internal control region) are not significantly different among the three tested groups (Fig. 1E; ANOVA, F(2,58) = 2.88, p = 0.064). Given rapid diffusion and degradation of released peptides in the extracellular space, these data suggest activity-dependent release of sNPF from Kenyon cells.
sNPF in the MB is required for appetitive olfactory memory
To manipulate the amount of sNPF specifically in Kenyon cells, we used targeted RNAi for the sNPF precursor using the GAL4/UAS system. Driving sNPF-RNAi (two copies, X and II) with MB247-GAL4 in the α/β- and γ-lobes of the MB significantly reduced sNPF signal by ∼70% in the targeted lobes (Fig. 2A–E; t(8) = 6.67, p < 0.001 in α-lobes, t(8) = 4.72, p = 0.002 in γ-lobe spur, n = 5 for each genotype). Overexpression of sNPF with UAS-sNPF driven by OK107-GAL4 did not lead to a detectable increase of sNPF levels in the MB (data not shown).
We next addressed the function of sNPF in the MB for appetitive olfactory memory. Knockdown in all lobe systems using OK107-GAL4 significantly impaired 2 h memory (Fig. 2F; t(19) = 3.43 and t(47) = 3.70 for comparisons of OK107/sNPF-RNAi vs OK107/+ and vs sNPF-RNAi/+, respectively, p < 0.01, n ≥ 10). This line strongly drives transgenic expression in most, if not all, Kenyon cells, as well as in few additional cells (Aso et al., 2009). To attribute the memory impairment to the MB, we tested an additional GAL4 line with overlapping expression in the α/β- and γ-neurons (MB247-GAL4) where sNPF has shown to be highly enriched (Johard et al., 2008). Similar to the sNPF knockdown by OK107-GAL4, the MB247-GAL4 significantly impaired appetitive memory (Fig. 2F; t(48) = 4.72, p < 0.001 and t(65) = 3.72, p < 0.01 for comparisons of MB247/sNPF-RNAi vs MB247/+ and vs sNPF-RNAi/+, respectively, n ≥ 21). We suppressed the sNPF knockdown with OK107-GAL4 only in the MB by expressing the GAL4-repressor GAL80 under control of the MB247 promoter (MB247-GAL80). GAL4 repression in the MB significantly rescued memory to control levels (Fig. 2G; p < 0.05, t(22) = 2.97 and t(24) = 2.76 for comparisons of sNPF-RNAi/OK107 vs MB-GAL80/OK107 and vs sNPF-RNAi;MB-GAL80/OK107, respectively, n ≥ 10).
A receptor for sNPF is required for appetitive olfactory 2 h memory. A, Pan-neuronal knockdown of the sNPF receptor (UAS-sNPFR-RNAi driven by elav-GAL4) leads to significantly impaired appetitive 2 h memory (**p < 0.01), whereas specific knockdown in the MB with neither OK107-GAL4 nor MB247-GAL4 significantly affects the memory (p > 0.05, n = 10–27). Bars and error bars indicate mean and SEM, respectively. B, Memory impairment after pan-neuronal knockdown of sNPFR is not significantly enhanced by expression of the GAL4 repressor GAL80 in the MB (p > 0.05, n = 10–12). Bars and error bars indicate mean and SEM, respectively.
Especially because sNPF was reported to regulate feeding behavior (Lee et al., 2004), we tested whether flies with sNPF knockdown in the MB respond normally to the to-be-associated stimuli used in the learning paradigm. First, we examined the response to the sugar stimulus (Fig. 2I,J). None of the pairwise comparisons of sNPF knockdown in the MB with the respective controls was significantly impaired in overall sucrose preference (Fig. 2I,J; Kruskal–Wallis test 13.42, p = 0.0094; p > 0.05 for all selected comparisons, differences in rank sums: 8.65 and 7.78 for sNPF-RNAi/OK107 vs OK107/+ and vs sNPF-RNAi/+, 3.00 and −14.13 for sNPF-RNAi/MB247 vs MB247/+ and vs sNPF-RNAi/+, respectively, n ≥ 8). The dynamics of sugar detection and maintenance of the choice were also indistinguishable (Fig. 2I). Furthermore, the response of the knockdown flies to the used odorants (4-methylcyclohexanol and 3-octanol) was not significantly impaired compared with the corresponding genetic controls (Table 1).
Next, we tested the effect of sNPF overexpression on memory. Consistent with no significant increase of sNPF immunoreactivity upon overexpression (data not shown), we observed neither impairment nor enhancement of appetitive memory in OK107/UAS-sNPF or MB247/UAS-sNPF (Fig. 2H; Kruskal–Wallis test 1.78, p > 0.05, n = ≥ 8).
Kenyon cells may form reciprocal synapses in the trajectory of their parallel axons (Christiansen et al., 2011). To examine whether sNPF expressed in the MB targets nearby Kenyon cells, we knocked down the expression of a receptor for sNPF (sNPFR). Before local knock down in Kenyon cells, we verified the potency of the transgenic RNAi for sNPFR using a pan-neuronal driver (elav-GAL4). We found significant impairment in appetitive memory after pan-neuronal knockdown (Fig. 3A; t(18) = 3.24, p < 0.01 and t(35) = 4.54, p < 0.001 for comparisons of elav/sNPFR-RNAi vs elav/+ and vs sNPFR-RNAi/+, respectively, n ≥ 10). Naive sugar and odor responses of these flies were not significantly impaired in our assay (Table 1), although a role of sNPF in feeding- and odor-driven food search has been reported earlier (Lee et al., 2004; 2008; Root et al., 2011). In contrast to the significant memory impairment after pan-neuronal knockdown, MB-specific knockdown with OK107 or MB247 did not significantly affect appetitive memory (Fig. 3A; p > 0.05, t(26) = 0.89, t(39) = 2.04 for comparisons of MB247/sNPFR-RNAi vs MB247/+ and vs sNPFR-RNAi/+, t(29) = 2.26, t(41) = 3.14 (p > 0.05) for OK107/sNPFR-RNAi vs OK107/+ and vs sNPFR-RNAi/+, respectively, n ≥ 12). To further support the role of sNPF outside the MB, we expressed sNPFR-RNAi using elav-GAL4, but at the same time inhibited GAL4 activity in the MB by expression of MB-GAL80. Consistent with the previous data (Fig. 3A), this manipulation did not significantly improve the impaired memory performance of pan-neuronal knockdown of sNPFR (Fig. 3B; t(19) = 0.34, p > 0.05, n ≥ 10). In line with these behavioral data, we found no detectable reporter expression in Kenyon cells in sNPFR-GAL4 (data not shown) (Kahsai et al., 2012). These results together imply that sNPF released by Kenyon cells may be received by extrinsic neurons of the MB to mediate olfactory memory.
Discussion
The MBs are a prominent brain structure in insects required for olfactory learning and other complex behaviors (Heisenberg, 2003; Kahsai and Zars, 2011). Although several neurotransmitters and neuromodulators have been associated with extrinsic neurons of the MB (Liu et al., 2007, 2012; Johard et al., 2008; Tanaka et al., 2008; Liu and Davis, 2009; Aso et al., 2010; Waddell, 2010; Blenau and Thamm, 2011; Lee et al., 2011; Sejourn é et al., 2011; Aso et al., 2012), a functional neurotransmitter or neuromodulator that mediates any of the MB-dependent behaviors in intrinsic MB neurons (Kenyon cells) has not been conclusively identified yet. However, a previous study suggested a potential role of peptidergic signaling in the MB for olfactory learning and memory (Iliadi et al., 2008). We show that a precursor for sNPF plays an important role as a neuroactive substance for olfactory memory in Kenyon cells of the MB (Fig. 2F,G).
sNPF is expressed in α/β- and γ-lobes of the MB (Fig. 1A,B), but is undetectable in α′/β′-lobe neurons. This suggests that different Kenyon cells may use different neuroactive substances in addition to sNPF. Furthermore, the same Kenyon cells may express potential cotransmitters together with sNPF. However, functional evidence of such neuroactive substances in the MB has not been provided yet. In other insects neuropeptides have been detected in subpopulations of Kenyon cells: tachykinin-related peptide, FMRFamide, and gastrin cholecystokinin in honeybees (Schürmann and Erber, 1990; Strausfeld et al., 2000; Takeuchi et al., 2004) and possibly orcokinin in locust (Hofer et al., 2005). Amino acids like aspartate and taurine have been detected by immunocytochemistry in portions of the Kenyon cell populations in Drosophila, and two other insect species (Schäfer et al., 1988; Sinakevitch et al., 2001, 2008; Strausfeld et al., 2003;). Parts of the MB are shown to be glutamatergic (Bicker et al., 1988; Strausfeld et al., 2003; Sinakevitch et al., 2010). While NMDA receptors in Kenyon cells are involved in olfactory memory (Wu et al., 2007) there is so far no conclusive evidence whether glutamate is released from the Kenyon cells or whether it represents a metabolic intermediate. Further functional evidence of these peptides and amino acids awaits further experiments. In addition, vesicular transporters for glutamate (DVGluT) as well as for amines (DVMAT) and GABA (DVGAT), seem to be absent in Kenyon cells (Chang et al., 2006; Daniels et al., 2008; Fei et al., 2010; Sinakevitch et al., 2010). Interestingly, a recent study showed that a putative vesicular transporter encoded by the portabella (prt) gene mediates courtship behavior and olfactory memory in Drosophila MB (Brooks et al., 2011). However, the target molecule of the Portabella transporter has not been identified until now.
We identified sNPF as a functional neuromodulator in Kenyon cells and the target cells are likely MB extrinsic neurons, rather than Kenyon cells themselves. These findings incite to screen for the target neurons by manipulating the sNPF receptor. This would help to identify circuits that are critical for olfactory learning. However, expression levels of sNPFR in adult Drosophila brains appear to be too low for detection with immunocytochemistry, whereas in the larvae immunolabeling was seen in unidentified neurons with processes invading the MB lobes (data not shown). Another approach to find target cells of sNPF release is knockdown of receptors for sNPF using RNAi and test for memory impairment. We verified this approach by pan-neuronal knockdown of sNPF receptors, which caused olfactory memory impairment (Fig. 3). As a first step in finding specific target neurons, we addressed whether sNPF-dependent olfactory memory involves cross talk between Kenyon cells. To this end, we knocked down the receptor specifically in the MB Kenyon cells (Fig. 3). This resulted only in a slight, but not significant, decrease in memory performance and thus is not supportive for reciprocal signaling between Kenyon cells or autocrine signaling of sNPF on these cells. Thus the neuronal target of sNPF from Kenyon cells still remains unclear for the moment, but altogether our data imply that sNPF released by Kenyon cells may be received by extrinsic neurons of the MB. Mapping neurons that require sNPF receptors for olfactory memory could be a promising future approach for finding extrinsic circuits to higher brain centers in Drosophila.
Footnotes
This work was supported by the Swedish Research Council (621-2007-6500, D.R.N.), the Royal Swedish Academy of Science, the Carl Tryggers Foundation (Å.M.E.W.), the Magnus Bergvalls Foundation (Å.M.E.W.), the Bernstein Focus Learning from Bundesministerium für Bildung und Forschung (H.T.), the Emmy-Noether Program from Deutsche Forschungsgemeinschaft (H.T.), and the Max Planck-Gesellschaft (H.T.). UAS-sNPF-RNAi and UAS-sNPF were generously provided by Dr. K. Yu (Daejeon, Korea). We are also grateful to Mikael A. Carlsson, Anja B. Friedrich, and our lab members for technical assistance, discussion, and critical reading of this manuscript.
- Correspondence should be addressed to Stephan Knapek, Max-Planck-Institut für Neurobiologie, Am Klopferspitz 18, D-82152 Martinsried, Germany. knapek{at}neuro.mpg.de
