Diverse subtypes of KCs form the Drosophila MB neuropil and their axons extend in the pedunculus and in the medial and vertical lobes 33 48 . The whole Drosophila MB structure can be revealed with anti-DC0 (PKA-C1) antibodies that label all parts of the KCs with different intensity (Figure 1A1, A2): the cell bodies (K), the dendrites in the calyx (ca), and the axons in the pedunculus (ped) and lobes. In the γ lobe, KCs extend only one axonal branch, forming a medial lobe and the spur (sp), whereas the axons of other KCs divide at the level of the spur to give rise to vertical and medial axonal branches, thus forming the α'/β' and α/β divisions of the lobes. The KC axons are organized in concentric strata in the pedunculus and α/β lobes. This is schematically depicted in Figure 1A3, the first-born cells, that is, the γ neurons, being the most external, and the last-born cells, the α/βc neurons, being embedded within the surrounding processes of α/β neurons. Like the other KCs, these last-born α/βc neurons originate from four neuroblasts, which provide four identical axonal bundles in the pedunculus and throughout the MBs (Figure 1A3).

The α/βc KCs present a high level of Glu-like immunoreactivity in their newborn axons immediately after adult eclosion (Figure 1B, C). Figure 1B (insert) shows that Glu-like immunoreactivity in the pedunculus is restricted to four thin axonal bundles of α/βc KCs embedded within the α/β axons. In contrast, no Glu-like immunoreactivity can be detected in the older neurons in the core of the α/β lobes in 10-day-old adult flies (Figure 1D1, D2). At this age, we observed Glu-like immunoreactivity scattered in the α and γ lobes and in the spur. This pattern is likely to represent the distribution of Glu-like immunoreactivity in the processes of MB extrinsic neurons rather than in KCs.

The accumulation of Glu in the young α/βc KCs suggests that this amino acid could be transiently used as a neurotransmitter by these cells. The vesicular Glu transporter DVGluT is involved in Glu synaptic vesicle storage, prior to neurotransmitter release, and can be used as a marker of glutamatergic neurons in Drosophila 29 30 . Therefore, we performed anti-DVGluT immunostaining immediately after adult eclosion, at a time when the Glu level is high in the MB core neurons. The DVGluT antibodies strongly labeled the protocerebrum and antennal lobe neuropils. In contrast, in the MB neuropil, neither the core neurons nor any intrinsic KCs were found to be immunopositive for DVGluT (Figure 1E-J). The DVGluT immunoreactivity observed in the MBs, particularly in the spur of the pedunculus and in the γ and α lobes (Figure 1E-J), most likely corresponds to synapses from extrinsic glutamatergic neurons. This suggests that Glu in the young α/βc KCs either is not stored in a vesicular pool or is stored in vesicles by another transporter not yet identified.

The previous results suggest that the last-born MB core neurons mature within just a few days after eclosion from an early Glu-expressing state to a differentiated cell in which Glu no longer accumulates. We asked whether other MB markers could be used to differentiate between these two maturation states. The enhancer trap lines 17d-, c739-, and 201Y-GAL4 mimic the expression of MB markers and are commonly used as MB-specific drivers. We crossed each of these lines to UAS-mCD8::GFP flies to characterize their expression pattern in young adults. When flies were collected immediately after eclosion, we found every time that the Glu-expressing α/βc neurons consist of two populations: inner α/βc neurons, which corresponds to the younger cells that do not express green fluorescent protein (GFP); and outer α/βc neurons, which do express GFP (Figure 2A-C and inserts). In contrast, in 10-day-old flies, we observed that these three drivers express mCD8-GFP in all the α/βc KCs: 17d-GAL4 in the α/βc neurons only (Figure 2D), c739-GAL4 in the α/β and α/βc cells (Figure 2E), and 201Y-GAL4 in the α/βc and γ lobe neurons (Figure 2F). In each of these lines, GFP immunostaining arranges in a specific way revealed by the pedunculus sections shown in the Figure 2 inserts: 17d-GAL4 expresses GFP in four inner bundles corresponding to the α/βc cells, c739-GAL4 in the whole area of the pedunculus that contains the α/β axons, and 201Y-GAL4 in the four α/βc bundles and a peripheral surrounding area that contains the γ lobe axons. Thus, at least three different MB markers are not expressed in the immature last-born α/βc KCs at a time when these cells show high levels of Glu immunoreactivity, while these markers are strongly expressed in these same cells in 10-day-old flies.

We then precisely compared the temporal patterns of Glu and GFP expression in the α/βc KCs of MB driver lines expressing mCD8-GFP by looking at the pedunculus in frontally cut agarose sections (Figure 3). In all three GAL4 lines (17d-, c739-, and 201Y-GAL4) early after eclosion, the α/βc neurons could be divided into three subtypes: the most inner α/βc neurons that express Glu but not GFP; more peripheral α/βc neurons that co-express glutamate and GFP; neurons in the α/βc outer area that express only GFP. Within 1 hour after eclosion, only the outer parts of the α/βc region were simultaneously positive for both GFP and Glu in each driver line (Figure 3A-C, panels a1-a3). Because the border of the core regions co-localizes with GFP, this suggests that the maturing α/βc neurons begin to express GFP at the time when Glu is still present. Twenty-four hours after eclosion, GFP starts to be expressed in most α/βc neurons. This is when Glu-like immunoreactivity was already becoming dramatically reduced (Figure 3A-C, panels b1-b3). Ten days after eclosion, there is no detectable Glu immunoreactivity in the α/βc KCs, which all express GFP at a high intensity (Figure 3A-C, panels c1-c3). Therefore, maturation of the last-born KCs appears to be a fast process in Drosophila that is concluded in a few hours only after adult eclosion.

We observed that Glu is already present in the α/βc axons of pharate adults, removed from the pupal cases a few hours before eclosion (Figure 4A, B). At this stage, Glu immunoreactivity is also detectable in the γ region of the MB pedunculus. Representative pictures are shown in Figure 4A1-A3, section close to the lobes, and Figure 4B1-B3, section close to the calyx. The MB drivers 201Y-GAL4 and 17d-GAL4 do not express GFP in the Glu-expressing α/βc neurons of pupal flies, except for a few neurons at the border of the core. In 201Y-GAL4, GFP is also present in the γ area in part of the pedunculus. The level of Glu immunoreactivity in the γ lobe varied significantly amongst flies sampled before eclosion, probably depending on the exact developmental progress of each pupal animal. This signal probably originates from extrinsic glutametergic neurons because we found that only cell bodies of newborn KCs express high Glu-immunoreactivity (see Figure 4C, I and text below).

In order to explore if Glu is similarly present in developing KCs during earlier phases of MB development, we performed Glu immunostaining on brain agarose sections in late third instar larva and at various stages of pupal development. At these stages, we generally found four clusters of Glu-immunoreactive KC bodies that send four Glu-positive bundles in the pedunculus and lobes (Figure 4C, I). Strikingly, these Glu-positive bundles are always located in the core areas of the pedunculus (arrowheads in Figure 4C, E, G, I), indicating that they correspond to newborn cells. A Glu signal was also detected in the γ lobe area in late pupae (Figure 4G) that most likely originates from extrinsic neurons as the only Glu-positive KC bodies are those of newborn cells. Therefore, although more work is needed to draw a precise picture of Glu expression in MB during metamorphosis, it seems to be a rule that developing KCs transiently express high levels of Glu during the formation and/or maturation of their neural circuits in the Drosophila MB.

Because we suspected that the Glu accumulated in the last-born immature KCs could be released and play a role in KC maturation, we looked for expression of further glutamatergic markers in the vicinity of the α/βc in newly eclosed flies. In mammalian nervous systems, subtypes of glial cells are equipped with specific proteins to ensure Glu recapture, metabolism and recycling. These include high affnity Glu transporters 63 64 and the enzyme glutamine synthetase (GS), which converts the captured Glu into glutamine 65 66 67 . Such glial cells can extend cytoplasmic processes close to Glu neurotransmitter release sites. Similarly, in the Drosophila central nervous system, the cell surface Glu transporter dEAAT1 is expressed by peripheral glia of the cell body cortex and addressed to glial processes that invade the neuropil proper 31 , and the fly glutamine synthetase orthologue GS2 is expressed in glial cell subsets 68 69 70 (T Rival and S Birman, unpublished).

A monoclonal antibody raised against sheep brain GS recognizes only one band on western blots of Drosophila brain proteins, which migrates close to the predicted size of GS2 (Figure 5A). In situ immunolabeling revealed that the fly GS is present in dEAAT1-expressing glial cells that surround the whole MB (Figure 5B). These glial cells also extend elaborate velate process inside the MB neuropil (Figure 5B, C). We observed in our preparations that GS-positive glial processes primarily surround the α/β neurons and enwrap their axons in the pedunculus and lobes, separating them from the other (γ, α'/β') divisions of the lobes. These glial processes originate from cells that lie on the surface over the MBs as well as between the lobes, thereby enclosing and isolating the MB from the surrounding protocerebral neuropils. Co-immunolabeling against GS and DVGluT showed that some GS-positive glial processes lie close to glutamatergic synapses, particularly in the α lobes (Figure 5C). However, a large network of GS-positive glial processes can be observed throughout the entire MB neuropil, including volumes such as the β lobe where DVGluT-positive processes are obviously absent (Figure 5C).

We found that GS expression is highly dynamic in the MB region during the first days of Drosophila adult life. The network of GS-positive glial profiles in the α/β lobe area becomes progressively dense and elaborated with time. Immediately after eclosion, GS is expressed by glial cells that surround the whole MB and send processes into it. These glial processes branch into the α/β region of the pedunculus and lobes and envelop the α/β axons. Surprisingly, no glial branching occurs in the α/βc areas; these are merely surrounded by the GS-positive processes in newly eclosed flies (Figure 5D, E).

Twenty-four hours after eclosion, glial cell processes also begin to invade the core neuropil of the pedunculus and α/β lobes. Processes also distribute equally throughout the lobes, with the highest density of processes at the lobe's periphery. The GS-expressing glial cells form a thick layer that surrounds the α/β lobes and generate a fine mesh-like network everywhere inside these lobes, including the α/βc area (Figure 5F, G). In 10-day-old flies, glial cells express high levels of GS immunoreactivity and surround all the MB lobes, forming much more elaborate and extensive distributions that in newly eclosed flies (Figure 5H, I).

Next, we searched for the presence and distribution of Glu receptors potentially involved in learning and memory in the Drosophila MB. The NMDA receptors (NMDARs) are subtypes of iGluRs composed of an essential NR1 subunit and variable NR2 subunits. We applied antibodies against the Drosophila dNR1 protein on whole-mount or frontal agarose sections of the fly brain. The immunostaining on whole mounts revealed a brightly labeled cluster of neurosecretory neurons, in agreement with a previous report 36 . Additionally, we found dNR1 staining in the MB lobes that was only detectable on agarose sections. In newly eclosed flies, significant dNR1 immunoreactivity was observed in the MB pedunculus and lobes, with a higher level of staining in the γ axons, specifically on the surface neurons that make up the spur region of the MB neuropil. In contrast, the α/βc neurons were not stained with these antibodies (Figure 6A, B).

The dNR1 immunostaining appeared somewhat different in flies examined 10 days after eclosion (Figure 6C-E). The lateral horn of the protocerebrum exhibited a high level of dNR1 immunoreactivity. Some KCs express dNR1 but the MB calyx showed only sparse staining, suggesting that the NMDAR subunit is mainly addressed to cell bodies and axons of KCs invading any of the MB lobes except those formed by the α/βc neurons (Figure 6C-E). We observed that the staining in the MB lobes and pedunculus was dramatically reduced in 10-day-old flies compared to the staining of the same areas in flies collected within 1 hour after eclosion. The spur area of the pedunculus had the highest level of dNR1 immunoreactivity (Figure 6E). Interestingly, the spur region is also the area of the MB that receives the highest density of DVGluT-immunoreactive profiles (Figure 1E-J). The dNR1-positive processes might also be associated with extrinsic neurons of the MBs.

DmGluRA is the only G-protein-coupled mGluR in Drosophila. This receptor has been shown to localize in the presynaptic site at the neuromuscular junction 24 and it is also expressed in the brain - for example, in clock neurons, where it regulates circadian locomotor behavior 39 . The GAL4 enhancer trap technique is a widely used method for analyzing tissue-specific gene expression patterns in Drosophila. To generate the insertion of a GAL4-containing P element in the regulatory region of the DmGluRA gene, localized on the fourth chromosome, we used a targeted transposition strategy as described by Sepp and Auld 71 . This technique induces the precise replacement of one P element for another. Starting from the line 39C42, in which an outmoded P element lacking any expression reporter is inserted 5.94 kb upstream of the translation initiation codon of DmGluRA 24 , we generated a new strain in which this P element was replaced with a P(Gal4) enhancer trap element, here called DmGluRA-GAL4. We used this driver line to monitor the expression of the DmGluRA receptor in the MBs, with mCD8::GFP as a reporter gene. Strikingly, we found that the last-born KCs located in the inner part of the α/βc neurons express GFP in their cell bodies, dendrites, and axons in the pedunculus and lobes (Figure 7A). According to their position, these cells undoubtedly correspond to Glu-expressing immature α/βc neurons. Although expression of the GAL4 reporter may, in part, differ from the mGluR pattern, such a precise localization suggests that the new born KCs express the DmGluRA receptor as well.

Twenty-four hours after eclosion, GFP immunostaining is still bright in labeled α/βc cell bodies and axons and individual axonal branches are clearly visible at the end of the α and β lobes (Figure 7B). All around the labeled axons in the lobes, one can notice GFP-positive dots, which may represent sections of smaller axonal branches derived from the main axons (Figure 7B). Ten days later, GFP expression is dramatically reduced in the MB α/βc neurons (Figure 7C). Remarkably, only two bundles out of initially four are still GFP-positive, as can be seen in the calyx and pedunculus presented in Figure 7C. GFP immunoreactivity associated with dendritic fields in the calyx and axons in the lobes is dramatically reduced. In addition, any bright axonal branches and processes have vanished, compared to the staining observed in newly eclosed flies. Such a dramatic change suggests that DmGluRA expression in α/βc neurons is transient and does not persist for long after eclosion. The immunoreactivity still detectable at 10 days might be the remnant of earlier expression and accumulation of the stable GFP protein when the DmGluRA promoter was still active.

Ubiquitin is a small protein that binds covalently to specific proteins and either marks them for degradation by the proteasome or modifies their activity 72 . Substrate protein ubiquitination plays important roles in neuronal differentiation 73 74 and synaptic plasticity 75 76 . We found selectively high levels of ubiquitin immunoreactivity in the α/βc neurons of just eclosed flies, as shown in Figure 8A1, A2. The spur region of the γ lobe was also labeled but with lower intensity than the core neurons. In 10-day-old flies, ubiquitin immunoreactivity was found to be homogenous all over the MB lobes, with higher intensity in the spur region (Figure 8B). This result further indicates that the α/βc neurons of newly eclosed flies are in a transient condition of active metabolism related to their maturation.

Although the last-born α/βc KCs show a high level of Glu immunoreactivity a few hours prior and after adult eclosion, Glu immunostaining is dramatically reduced in these cells 24 hours after eclosion and is entirely absent a few days later. Disappearance of this signal could result from the release or intracellular metabolism of this amino acid. Similarly, it was observed in cockroach MBs that newborn KCs loose Glu immunoreactivity when they become mature and establish contacts with extrinsic neurons 77 81 . Here we also present the first evidence that Glu transiently accumulates at a high level in developing newborn KCs of Drosophila in late larva and during pupal stages. Therefore, transient Glu expression could correlate with KC growth and maturation not only in the α/βc around eclosion time but also in other lobes during earlier stages of MB development.

Three subtypes of vesicular Glu transporters (VGluTs) have been identified in the mammalian nervous system with similar Glu transport functionality 20 . Two of these (VGluT1 and VGluT2) present complementary distribution in central glutamatergic neurons 82 83 . The third isoform, VGluT3, appears to be primarily expressed in neurons that release another transmitter (serotonin, dopamine, acetylcholine or GABA), where it may be required for efficient synaptic transmission 84 85 . In the present study, neither the α/βc neurons nor any other intrinsic MB KCs were found to express the Drosophila vesicular transporter DVGluT. This may indicate that the Glu that is accumulated in the inner α/βc neurons is not stored in synaptic vesicles. However, we cannot exclude the possibility that these cells express another vesicular Glu transporter not yet identified in Drosophila. We did observe DVGluT immunoreactivity in the MBs, particularly in the γ lobe and spur region and in the α lobes (Figure 1E-J), but the punctuate labeling and localization suggest that this distribution corresponds to glutamatergic synapses belonging to extrinsic neurons.

Can the Glu transiently stored in the newborn MB neurons be released into the extracellular space? In the absence of DVGluT or another similar transporter, this could involve a non-vesicular release of Glu. Non-conventional release of Glu from immature neurons has been previously demonstrated in the developing rat hippocampus 21 where Glu release exerts a paracrine action that seems to particularly affect the migration of neighboring maturing neurons 22 23 . To address this question indirectly, we looked for the presence in the MB of other proteins known to be involved in the recycling and degradation of Glu at glutamergic synapses.

An important role of glial cells is to capture Glu released from the synapse with specific transporters and then convert Glu to glutamine with GS. The only Drosophila high-affinity Glu transporter, dEAAT1, is expressed in subtypes of glial cells and is associated with Glu-release sites 28 31 86 . GS2 is similarly expressed in glial cells in the Drosophila nervous system 70 (T Rival and S Birman, unpublished). We show here for the first time that glial cells expressing dEAAT1 and GS surround the Drosophila MB lobe neuropiles, closely enwrapping the α/β lobes, thus isolating them from other lobes, and sending a mesh-like system of extensions inside these lobes. Enwrapping and invading of the MB β lobes by glia was also observed to occur in cockroach MBs, where glial cells are implicated in the removal of degenerating transient KC processes that occur during their establishment of mature connections with extrinsic cell dendrites 81 . The high levels of glial dEAAT1 and GS within the Drosophila MB lobes suggest that this neuropil is tightly cordoned off from other parts of the brain and regulates the extracellular Glu level between the axons.

Our data show that GS expression is highly dynamic in the MB during the first day of adult life, suggesting that glial cells play a role in establishing the MB's functional network. During the first hour after eclosion, the meshwork of glial processes expressing Glu signaling-associated molecules (GS and dEAAT1) is not present in the inner α/βc region, but within 24 hours this area becomes covered by glial extensions (Figure 9). These glial elements are possibly guided towards the α/βc area by the gradient of Glu released by the last born KCs. Glia could be involved in reducing Glu concentration in this area and play a role in axonal guidance and final maturation of KCs. Evidence that Glu transporters are required for coordinated brain development has been previously reported for mice: the absence of two glial Glu transporters resulted in excess of extracellular Glu and abnormal formation of the neocortex 87 .

Assuming Glu is released by the newborn MB neurons, it has to interact with specific receptors. Therefore, we searched for the expression of Glu receptors in MB neuropiles of young adult Drosophila, particularly those receptors that are likely to regulate neuronal growth and maturation through second-messenger pathways. Once activated by simultaneous Glu binding and membrane depolarization, the NMDAR channel allows calcium influx into the postsynaptic cell, where this ion triggers a cascade of biochemical events resulting in synaptic maturation and plasticity 88 . We used available antibodies against the constitutive dNR1 subunit of the Drosophila NMDAR 36 . Immediately after eclosion, many processes in the MB neuropil were found to be dNR1-positive, with the exception of the α/βc neurons. The Glu released from either these α/βc neurons, or the surrounding glial cells, or extrinsic MB glutamatergic neurons may activate these NMDAR receptors. Thus, a widespread localization of NMDAR characterizes the MB immediately after eclosion, at the beginning of adult life when the MB is expected to receive the least inputs from sensory interneurons. Subsequently, with increasing sensory data being received and relayed to projection neurons, there is a dramatic and concomitant restructuring of NMDAR signaling: the majority of MB neurons no longer express these receptors. It is only those neurons that receive constant glutamatergic signaling that still address the dNR1 subunit in the vicinity of glutamatergic synapses expressing DVGluT. This occurs in particular within the spur region of the MB and the lateral horn. Such developmentally related regulation of NMDAR expression in the MBs of young adult flies may relate to adaptations of synaptic activity in response to sensory experience 88 .

mGluRs are neuromodulatory G-protein-coupled receptors that are involved in many aspects of brain physiology, including neuronal development, synaptic plasticity, and neurological diseases 8 89 . Whereas eight distinct mGluRs are present in the mammalian genome, a single functional mGluR is expressed in Drosophila, DmGluRA. The fly mGluR is structurally and pharmacologically closer to the mammalian group II mGluRs, which are mainly presynaptic receptors negatively coupled to adenylate cyclase. Our attempts to locate DmGluRA with the commercially available monoclonal antibody 7G11 90 were not successful because the antibody produced by the hybridoma clone recently lost its binding specificity. To monitor DmGluRA distribution, we used a new GAL4 line that carries an enhancer trap insertion close to the mGluR start site of transcription, keeping in mind that expression of this GAL4 reporter may, in part, differ from the mGluR pattern. Strikingly, the DmGluRA-GAL4 line was found to express GFP selectively in the Glu-accumulating α/βc KCs of newly eclosed adult flies. This is in contrast to commonly used MB GAL4 driver lines (17d-, c739- and 201Y-GAL4) that do not express GFP in these neurons immediately after eclosion. Ten days later, the GFP staining in the DmGluRA-GAL4 line appeared strongly reduced in the α/βc; in contrast, the MB drivers now expressed GFP in these neurons.

Because the GAL4 reporter method reveals whole neurons, we could not determine where the receptor is addressed restrictively in cell bodies, dendrites or axons. A previous study performed with an active lot of 7G11 antibody indicated that DmGluRA is present in nearly all neuropiles of the mature adult fly brain, including the MB calyces, but not in the MB lobes 91 . However, this article did not report on the localization of DmGluRA in newly eclosed Drosophila. Further work is required to precisely locate the subcellular localization of DmGluRA in the newborn α/βc neurons, either with a new antibody or a DmGluRA-GFP fusion gene. The source of Glu binding to this mGluR receptor may be the neighboring glial cells or newborn KCs themselves, or both. Through activation of these receptors, Glu is likely to have a transient paracrine action on the α/βc neurons during the first day after eclosion that could be required for dendrite growth or synaptic maturation 92 .

Although the α/βc KCs represent a minor part of the α/β lobe neurons, the maturation of these cells appears to be essential for proper MB functioning. Selective expression of the rutabaga (rut)-encoded adenylate cyclase in the α/βc neurons with 17d-GAL4 was shown to partially restore olfactory learning and memory in 2- to 5-day-old rut mutant flies 49 . In contrast, no rescue of the rut defect was observed with c739-GAL4, which expresses in more peripheral α/β neurons at this stage. Therefore, the network involved in olfactory learning and memory apparently requires the α/βc neurons and is already functional in 2- to 5-day-old flies. Furthermore, treatment with mGluR antagonists restored courtship behavior, memory deficits and MB structural defects in DFmr1 mutants, a Drosophila model of FXS. These positive effects are even stronger when the pharmacological treatment is applied both during larval development and after eclosion 59 . This suggests that these behavioral defects relate to an abnormally high level or prolonged duration of DmGluRA expression in the α/βc neurons of DFmr1 mutants 93 . Further study should determine the distribution of Glu and DmGluRA during MB development in Drosophila FXS models.

The ubiquitin-proteasome system is one of the major conserved cellular pathways controlling protein turnover in eukaryotic cells. Substrate protein ubiquitination plays important roles in neuronal differentiation 73 74 , axonal targeting 94 , synapse formation and plasticity 75 76 95 . In addition to strong Glu immunolabeling in the inner α/βc KCs, we also observed a high level of anti-ubiquitin immunoreactivity in these neurons immediately after eclosion. Such a high staining level was no longer detected in 10-day-old flies. In contrast, the spur region of the MB showed a constant high ubiquitin immunoreactivity that did not change with the age of the animal. This could suggest that synaptic plasticity is particularly active in this MB area.

Similarly, labeling of the cockroach MB β lobe with anti-ubiquitin showed, at specific stages in each developmental instar, as well as at an early adult stage, consistent staining of newly generated KC axons. Anti-ubiquitin also labeled the extending transiently Glu-immunoreactive collateral processes from developing KCs in the ingrowth zone, the hemimetabolous homologue of Drosophila's core KCs 81 . These authors showed that ubiquitin expression precedes degeneration of these collaterals and their subsequent removal by scavenging glial cells. Glu receptors can be endocytosed by an ubiquitin-dependent mechanism 75 . The down-regulation of Glu and its receptor protein, possibly mediated by ubiquitin, thus appear to be important steps in the maturation and differentiation of the α/βc KCs.

Fly stocks and crosses were maintained at 25°C on standard corn meal-yeast-agar medium supplemented with methyl-4-hydroxy-benzoate as a mold protector. The following strains were used: wild-type Oregon R; UAS-mCD8::GFP to express cell surface membrane-associated GFP 55 ; the MB drivers 17d-GAL4 96 , c739-GAL4 and 201Y-GAL4 97 ; dEAAT1-GAL4, which targets Glu transporter-expressing glial cells 31 ; and DmGluRA-GAL4, a GAL4 enhancer trap inserted upstream and close to the DmGluRA transcription start site (Y Grau, unpublished). This line was generated by a P-replacement method (see below).

To generate the DmGluRA-GAL4 driver, we mobilized the original P element 39C42, w ⁺, localized on the fourth chromosome, 5.94 kb upstream of the translation initiation codon of DmGluRA 24 . Among the white revertant lines analyzed by PCR using genomics primers spanning 39C42, one deletion event that removed most of the P element, including the miniwhite gene, but not the P element inverted repeats, was recovered; this is called the P-XVII3:w ^- line. This homozygous viable line was used as a target to isolate replacement events of the residual P element by P(GAL4, w ⁺). Df(4)O2 has been described in Bogdanik et al. 24 . Df(4)G(spa ^- ) is lethal over Df(4)O2 and was obtained from the Bloomington stock center. These and other used mutations are described in FlyBase. The donor strain carrying a P(GAL4, w ⁺) located on the X chromosome was provided by Dr J-M Dura (CNRS, Montpellier, France).

The P-XVII3 enhancer trap targeted transposition screen consisted of the cross of wP(Gal4, w ⁺); SbD2-3/+; P-XVII3:w ^- /Df(4)O2 males to yw; Df(4)O2/Ci ^D spa females. w ⁺ ; Sb+; Ci ^D males were collected and mated to yw; Df(4)O2/Ci ^D spa as single-pair mating. Then, single w ⁺ ; Ci ^D males were crossed to yw; Df(4)G/Ci ^D spa females. Tubes where all Ci ^D spa+ progeny was w ⁺ indicated a mobilization of the P(GAL4, w ⁺) transposon to the fourth chromosome. These lines were established as stable stocks and further analyzed by PCR to determine whether P(GAL4, w ⁺) had indeed replaced the P-XVII3:w ^- element. We determined whether the donor P element had been replaced in the appropriate location by molecular verification of targeted transpositions events. PCR analysis was performed on genomic DNA with primers specific for the DmGluRA gene surrounding the P-insertion site as described in Bogdanik et al. 24 as well as primers internal to the GAL4 enhancer trap construct (Plac1 and Pry2; described by the Berkeley Drosophila Genome Project). All PCR reactions were done as recommended by the manufacturer (Takara, Kyoto). To determine the precise location of the P(GAL4, w ⁺) insertion in the obtained lines, small regions spanning the junction between the gene and inserted P element ends were amplified using PCR and sequenced. We thus found that one line, called R39-GAL4, was the result of a targeted transposition event that successfully converted the P-XVII3 insertion to a P-GAL4 located 5.6 kb upstream of the DmGluRA translation start.

Proteins from Drosophila head and dissected brain homogenates were separated on 7.5% SDS-PAGE and transferred onto 0.45 μm nitrocellulose membranes (Bio-Rad Laboratories, Hercules, CA, USA) in 25 mM Tris, 192 mM glycine, 15% methanol at 0.6 A for 2 h at 4°C. Membranes were blocked for 1 hour in phosphate-buffered saline (PBS) buffer containing 10% low-fat powdered milk at room temperature and incubated with mouse monoclonal anti-GS antibodies (Chemicon, Millipore MAB302, Billerica, MA, USA) at 1:2,000 in PBS plus 5% milk overnight in the cold room. Following four 5-minute washes in PBS, membranes were incubated with anti-mouse horse radish peroxidase-conjugated IgG (Jackson Immunoresearch Laboratories, West Grove, PA, USA) at 1:7,000 in PBS plus 5% milk for 2 h. Membranes were washed four times in PBS and developed using chemiluminescence as described by the manufacturer (SuperSignal West Pico Chemiluminescent Substrate, Pierce Thermo Scientific, Rockford, IL, USA). Several exposures were captured on Kodak BioMax MR films.

For in situ immunostaining, adult male flies of three different ages were collected: either immediately (within 1 hour), 24 hours, or 10 days after eclosion. For each condition, at least three independent experiments were conducted with three to five flies processed each time for brain dissection, agarose sectioning and immunolabeling. The following primary antibodies were used: rabbit Glu antisera and mouse monoclonal anti-Glu (Gemacbio, Saint Jean d'Illac, France; diluted 1:1,000); rabbit DVGluT antisera (gift of Aaron DiAntonio; 1:1,000) 31 33 ; mouse monoclonal anti-GS (Chemicon, Millipore MAB302; 1:500) 62 ; rabbit polyclonal anti-dNR1 (NMDAR subunit; gift of Ann-Shyn Chiang; 1:500) 36 ; rabbit polyclonal anti-ubiquitin (Dako Z0458, Glostrup, Denmark; 1:100) 73 ; chicken polyclonal anti-GFP (Abcam ab13970, Cambridge, MA, USA; 1:1,000); rabbit polyclonal anti-DC0 (alias PKA-C1, catalytic subunit of Drosophila protein kinase A; gift of Daniel Kalderon; 1:1,000) 98 . The secondary antibodies were used at 1:500 dilution: goat anti-rabbit, Cy5-conjugated (Fab') fragment of IgG (Jackson ImmunoResearch), goat anti-mouse IgG conjugated to Alexa Fluor 555 or goat anti-chicken Alexa Fluor 488 (Invitrogen Molecular Probes, Life Technologies, Carlsbad CA, USA). Preparations incubated in the absence of primary antiserum showed no labeling.

The protocol for Glu immunostaining together with controls of Glu specificity were described in detail previously 32 33 77 . For controls of immunostaining (Figure 1B, inserts), L-glutamic acid (Sigma Aldrich G-6904, Saint-Louis, MO, USA) was conjugated to bovine serum albumin (BSA) via glutaraldehyde Glu-G-BSA following a previously described protocol 99 . After adsorption of the Glu antiserum with Glu-G-BSA (10^-4 M with respect to the amino acid), sections of brains showed no labeling. Drosophila brains were removed in fixative containing 2.5% paraformaldehyde plus 1.5% glutaraldehyde (both from Electron Microscopy Sciences, Hatfield, PA, USA), 1% sodium metabisulfite (Na₂S₂O₅; SMB; Sigma) in 0.1 M sodium cacodylate buffer (pH 7.0), and postfixed in the same solution overnight at 4°C. The fixed whole brains were incubated for 15 minutes in 0.05 M Tris-HCl, pH 7.5 buffer supplemented with 0.5% SMB (Tris-SMB) and 0.5% sodium borohydride (NaBH4) to saturate double bonds. After a wash in Tris-SMB, brains were embedded in 8% agarose and cut in 60-μm serial sections with a vibratom (Leica). Sections were washed in Tris-SMB supplemented with 0.5% (v/v) Triton X-100 (Tris-SMB-Tx), then pre0incubated with 5% (v/v) normal swine serum (Dako, Glostrup, Denmark) in Tris-SMB-Tx for 1 hour. Brain sections were then simultaneously incubated with rabbit Glu antiserum (1:1,000) and chicken anti-GFP (1:1,000) in Tris-SMB-Tx overnight at room temperature. After a wash in Tris-Tx, sections were incubated overnight at room temperature with conjugated secondary antibodies diluted 1:250. After a final wash in Tris buffer, sections were embedded in 80% glycerol.

For marker proteins, Drosophila brains were dissected and fixed overnight in 4% paraformaldehyde in PBS made from tablets (Sigma P4417). The fixed brains were washed in PBS, embedded in 8% agarose and cut in serial 60-μm sections. Sections were washed four times in PBS containing 0.5% (v/v) Triton X-100 (PBS-Tx), then blocking solution was applied consisting of 10% (v/v) swine serum in PBS-Tx. After one night incubation with primary antibodies in PBS-Tx, the sections were thoroughly washed in PBS and then incubated with the secondary antibodies at 1:500. After a final wash in PBS, sections were embedded in 80% glycerol.

Images were collected on a Zeiss LSM 510 confocal microscope (Carl Zeiss, Oberkochen, Germany). Series of four to ten 0.5-μm optical sections (1,024 × 1,024 at 12-bit color depth) were scanned using 40 × 1.0 or 63 × 1.4 oil iris Plan-Apochromat objectives. The images were stored as TIFF files and edited in Adobe Photoshop CS2. The Glu and dNR1 antibodies gave only superficial staining in sections so the images were composed of two to five optical sections for these antigens. For DVGluT, GS and ubiquitin, images were made from up to ten optical sections from a stack. The contrast and intensity were adjusted accordingly.