Presynaptic PI3K activity triggers the formation of glutamate receptors at neuromuscular terminals of Drosophila

Synapses represent the major structural and functional contact points between neurons. These structures couple the neurotransmitter release machinery (in the presynaptic cell) with the neurotransmitter receptor complexes (in the postsynaptic cell). The Drosophila larval neuromuscular junction (NMJ) is suitable to dissect the molecular mechanisms controlling synaptic morphology and plasticity due to its accessibility to perform electrophysiology, inmunohistochemistry, live imaging or electron microscopy (Collins and DiAntonio, 2007; Owald and Sigrist, 2009). Similar to vertebrate synapses in the central nervous system (CNS), the Drosophila NMJ is glutamatergic. The projection pattern of each single motor neuron innervating the larval body wall muscles is fully characterised (Hoang and Chiba, 2001), as well as the formation and maturation of the synapse (Fouquet et al., 2009; Owald et al., 2010; Rasse et al., 2005; Schmid et al., 2008). The NMJ postsynaptic sides express non-NMDA receptors, homologous to vertebrate non-NMDA ionotropic receptors (DiAntonio et al., 1999; Schuster et al., 1991), which mediate fast excitatory transmission (Dingledine et al., 1999). Ionotropic glutamate receptors (iGluRs) concentrate in the postsynaptic density (PSD), a dense specialisation attached to the postsynaptic membrane, opposite to presynaptic neurotransmitter release sites, also known as active zones (Ziff, 1997). These receptors exhibit five subunits, GluRIIA to GluRIIE (DiAntonio, 2006; DiAntonio et al., 1999; Featherstone et al., 2005; Marrus et al., 2004; Petersen et al., 1997; Qin, G. et al., 2005), although a functional receptor can be formed with either GluRIIA or GluRIIB (DiAntonio et al., 1999).

Out of the extensive number of studies on the larval NMJ model, especially relevant here are works focusing on the interaction between the presynaptic neurotransmitter release sites and their opposed glutamate receptor clusters at the postsynaptic site. Thus, while GluRIIA levels are higher than those of GluRIIB at immature PSDs, GluRIIB is prevalent at mature PSDs (Schmid et al., 2008). Conversely, in the presynaptic compartment, synapse stability and function depends on the proper assembly of the active zone (AZ), the cytoplasmic matrix at the active zone (CAZ) and the large proteinaceous scaffold (Owald and Sigrist, 2009; Siksou et al., 2007). Genetic analyses have shown that the gene Shaggy/GSK3 (homologue of mammalian glycogen synthase kinase 3 beta, GSK3B) regulates negatively the growth of the NMJ only when expressed in motor neurons, but not in muscles (Franco et al., 2004). In the same context, the overexpression of phosphoinositide 3-kinase (PI3K) in larval motor neurons increases the number of functional synapses in a cell-autonomous manner (Martín-Peña et al., 2006). Recently, the synaptogenic activity of PI3K has been demonstrated in mammals, being able to induce behavioural changes in the memory performance of rats (Cuesto et al., 2011).

Here, we address the dynamics of synapse growth and GluRII subunit composition under conditions of presynaptic- and postsynaptic-induced changes in PI3K expression. The data demonstrate an increase in the number of PSDs only when PI3K is overexpressed presynaptically.

In Drosophila, PI3K overexpression leads to an increase in the number of functional synapses in larval motor neurons (Martín-Peña et al., 2006). Due to this synaptogenic effect, we first investigated whether increasing PI3K levels also exerts a possible influence on postsynaptic receptors. Under electron microscopy, the glutamate receptors form an electron dense compartment, the postsynaptic density (PSD), a key structure for the stabilisation and dynamic regulation of neurotransmitter receptor populations (Renner et al., 2008). To perform our analysis we focused on the well characterised RP3 motor neuron innervating muscles 6/7 of abdominal segment 3 (A3) from the third instar larvae. Synapse identification and counting were achieved by double immunostaining with anti-HRP antibody, to visualise neuronal membranes (Jan and Jan, 1982), and with the monoclonal antibody nc82 to recognise Brp (Bruchpilot), a synaptic protein (Rasse et al., 2005; Wagh et al., 2006). To label and quantify postsynaptic densities (PSDs), we applied antibodies against the non-essential glutamate receptor subunits A, B (GluRIIA and GluRIIB) and also against two of the fundamental subunits C and D (GluRIIC and GluRIID) (DiAntonio, 2006).

We confirmed here that larvae overexpressing PI3K in presynaptic motor neurons (D42-Gal4/UAS-PI3K) form more synapses than controls (D42-Gal4) by the criterion of nc82-positive staining (Martín-Peña et al., 2006), and also by counting Brp-positive spots detected with a second antibody (anti-Brp) (Fig. 1A; Table 1). Both markers identify different amino acid residues of the BRP protein (Fouquet et al., 2009). In addition, we evaluated the effects at the postsynaptic density by immunostaining against the four different GluRII subunits (A–D) and counting immunopositive punctae as a direct estimation of PSD number (Fig. 1A–D). We found that overexpressing PI3K in motor neurons leads to an increase of 32±7% in the IIA subunit, 40±10% in IIB, 33±6% for IIC and 43±4% for IID subunits (Fig. 1F; Table 1). In a second series of experiments, we questioned whether the downregulation of PI3K levels, by using the presynaptic overexpression of a PI3K dominant negative construct (Leevers et al., 1996), alters synapse and PSDs numbers. The data show a significant decrease in synapse number in D42-Gal4/UAS-PI3K^DN larvae: 400±25 synapses (n = 9) when compared with control D42-Gal4 larvae: 548±38 synapses (n = 9) (supplementary material Fig. S1C). We also detected a reduction in the number of PSDs visualised by the GluRIID antibody in PI3K^DN larvae: 477±40 (n = 10) versus D42-Gal4 larvae: 618±48 (n = 9) (supplementary material Fig. S1D). Together, these data demonstrate that PI3K overexpressed presynaptically elicits changes in the number of postsynaptic glutamate receptor fields, in accordance with the observed increase in synapse number.

We then asked whether the increase in the number of postsynaptic receptors would be accompanied by a presynaptic PI3K-dependent effect. To this end, we quantified by western blot the changes in expression level of GluRIIA. These data are relevant because GluRIIA levels are higher compared with the other subunits at the developing, immature PSDs (Schmid et al., 2008). In these experiments we processed fillets from third instar larvae expressing the pan-neural elav-Gal4 driver as a control (Lin and Goodman, 1994), elav-Gal4/UAS-PI3K and elav-Gal4/UAS-PI3K^DN as experimental genotypes. As an additional control we also expressed a UAS-GluRIIA^RNAi construct in the elav domain (data not shown). Using an antibody specific for GluRIIA, we found that PI3K upregulation leads to an increase of 42±20% in the GluRIIA subunit levels, compared to control larvae (n = 5 independent western blots for each genotype, three larvae per lane) (Fig. 2A,B). However, the downregulation of PI3K does not seem to affect GluRIIA levels (n = 5, three larvae per lane) (Fig. 2A,B). We have also performed western blot quantifications from elav-Gal4, elav-Gal4/UAS-PI3K and elav-Gal4/UAS-PI3K^DN third instar larval brains (supplementary Fig. S2). The data show a moderated increase of 24±6% in GluRIIA levels in elav-Gal4/UAS-PI3K (supplementary material Fig. S2). Furthermore, we also detected a 40% decrease in GluRIIA in larval brains expressing an RNAi construct against GluRIIA (data not shown). All these data indicate that the presynaptic PI3K overexpression, but not its downregulation, is able to modify the levels of GluRIIA in the postsynaptic compartment.

We quantified the expression changes of another postsynaptic protein, Discs large (Dlg), a protein essential for the architecture of excitatory synapses in mammals (Montgomery et al., 2004; Parnas et al., 2001). In Drosophila, Dlg expression is restricted to postsynaptic areas, and it is implicated in the clustering of receptors that contain GluRIIB, but not GluRIIA (Chen and Featherstone, 2005; Thomas et al., 2010). Here, we observed possible changes in Dlg expression with an antibody recognising the two Dlg isoforms: DlgA and DlgS97 (Mendoza-Topaz et al., 2008) (Fig. 3A). To this end, we analysed western blots from elav-Gal4, elav-Gal4/UAS-PI3K and elav-Gal4/UAS-PI3K^DN genotypes. The data indicate that the upregulation of PI3K causes an increase of 83±31% in the DlgA isoform levels (Fig. 3B) and 72±29% in DlgS97 isoform (Fig. 3C) (n = 5 in each genotype, three third instar larval fillets per lane). In turn, the reduction of PI3K causes no differences in any of the two Dlg isoforms (Fig. 3B,C). Both sets of data are reproduced when Dlg proteins levels were quantified in elav-Gal4, elav-Gal4/UAS-PI3K and elav-Gal4/UAS-PI3K^DN third instar larval brains (supplementary material Fig. S3). Thus, we observed increases of 14±7% in DlgA and 15±7% of DlgS97 isoforms in elav-Gal4/UAS-PI3K brains and no level changes elav-Gal4/UAS-PI3K^DN when compared with the control. In parallel, larval brains expressing the UAS-Dlg^RNAi construct elicited a 20% reduction for both Dlg proteins (data not shown) in agreement with previous observations (Mendoza-Topaz et al., 2008). Taken together, these results indicate that PI3K from the presynaptic side modify the amount of Dlg protein isoforms in the developing NMJ.

We next tested the possible involvement of PI3K in the PSD dynamics for GluRs. To this end, we performed fluorescence recovery after photobleaching (FRAP) experiments in the larval NMJ (Rasse et al., 2005). FRAP demonstrated previously that GluRIIA subunit entry was directly related to PSD formation and growth (Owald and Sigrist, 2009; Rasse et al., 2005; Schmid et al., 2008). GluR dynamics during formation and maturation of PSDs have been imaged in larvae carrying GluRIIA constructs tagged with green fluorescent protein (GluRIIA^GFP) and red fluorescent protein (GluRIIA^RFP) markers inserted in the intracellular C-terminal of GluRIIA to allow endogenous expression (Rasse et al., 2005). Thus, the recovery of GluRIIA^RFP fluorescence after its specific bleaching allows the quantification of glutamate receptor incorporation into the PSDs of the growing larvae, whereas GluRIIA^GFP signal is monitored as an internal control of PSD formation (‘steady state channel’). We generated larvae in which PI3K is selectively expressed in motor neurons, using the OK6-Gal4 line (Aberle et al., 2002). This line is expressed in all motor neurons, but we focused in muscle 27 and 26 due to their easier accessibility. These muscles are located in the ventral part of larvae musculature close to the cuticle, allowing a better observation of the fluorescence in vivo (Hoang and Chiba, 2001; Rasse et al., 2005).

Individual NMJs were imaged at three different time-points used: Pre-bleach (0 hours), Bleach (photobleaching of GluRIIA^RFP) and Post-bleach (14 hours after GluRIIA^RFP bleaching) for each genotype (Fig. 4A–G). 14 hours later, we quantified the number of new PSDs formed after photobleaching in muscle 27. Our data indicates no differences between control larvae (OK6-Gal4, 32±3%, n = 10) and those overexpressing PI3K (OK6-Gal4/UAS-PI3K, 36±5%, n = 12) (Fig. 4H). Next, we estimated the percentage of recovery by measuring differences in the GluRIIA^RFP signal intensity in the bleached area compared with the unbleached area 14 hours after photobleaching the NMJ (Rasse et al., 2005; Schmid et al., 2008). Again, there were no statistically significant differences in GluRIIA^RFP signal recovery between control (52±12%) and experimental (39±6%) groups (Fig. 4I). Collectively, our data show that PI3K upregulation does not alter the dynamics of GluR incorporation during PSDs formation. Instead, the system seems to trigger the formation of additional synapses, with GluR incorporation at individual PSDs operating normally. Consistent with the larger number of PSDs, the overall amounts of GluRIIA per terminal were also increased.

In the previous experiments, PI3K activity was chronically up- or downregulated after embryogenesis. It could be argued that the observed PI3K-dependent changes in synapse and PSDs number could result from synaptic rearrangements produced along developmental stages. To investigate this possibility, we performed experiments allowing temporal control of the driver activation by using a temperature-sensitive form of the Gal4 inhibitor, Gal80^ts (McGuire et al., 2003). Thus, when we raised embryo and larvae at 18°C, Gal80^ts repressed Gal4, preventing PI3K overexpression. At early third instar larvae, temperature was shifted to 30°C and Gal4 repression was abolished allowing selective PI3K overexpression in larval motor neurons. The system needs at least 6 hours to have full Gal4 protein expression detected by a CD8-GFP reporter (data not shown). In these experiments, we considered three data and time points: before PI3K expression (0h), 24 hours after temperature shift to 30°C (24h) and 40 hours at 30°C (40h).

Synapse numbers were counted after double immunostaining with nc82 and HRP antibodies in larval muscles 6–7 in control (Fig. 5A) and PI3K larvae (Fig. 5B). Data indicate that 24 hours after the ‘switch-on’ protocol, both control (n = 8) and PI3K overexpressing (n = 9) larvae showed increased synapse number (83% and 99%, respectively). Although this increment tended to be higher in PI3K-overexpressing larvae compared with controls, we did not observe statistical differences between both groups (P = 0.6954), probably due to an insufficient PI3K activation (Fig. 5A). However, after 40 hours at 30°C, PI3K larvae showed a statistically significant increase in synapse number compared with control larvae. The data values were 560±27 (n = 10) and 694±43 (n = 12) synapses in control and PI3K larvae, respectively (Fig. 5A). Also, we repeated the same experiment in D42-Gal4/UAS PI3K^DN larvae quantifying the number of synapses. We found a decrease in the number of synapses in D42-Gal4/UAS PI3K^DN: 437±40 synapses (n = 10) compared with D42-Gal4: 542±35 synapses (n = 11) (supplementary material Fig. S4). Taken together, these results demonstrate that the presence of high levels of PI3K during third instar larvae is sufficient to trigger NMJ synaptogenesis.

As shown above, the PI3K-dependent increase in synapse number correlates with increments in the number of glutamate receptor subunits and, hence, PSDs (Fig. 1). To further investigate this finding, we expressed PI3K under the control of D42-Gal4 only in third instar larvae using the Gal80^ts system, and quantified the number of PSDs. As in the previous experiment, we considered three time points: 0 hours before PI3K expression, 24 hours after temperature shift to 30°C (24h) and 40 hours at 30°C (40h). PSDs were labelled by co-immunostaining with anti-GluRIIA and anti-HRP (Fig. 6A,B). The data show that 24 hours after ‘switching-on’ the system, the overexpression of PI3K (606±55, n = 10) increased the number of PSDs with respect to the control D42-Gal4 larvae (458±38, n = 9) (Fig. 6C). This augmentation occurs even prior to the increase in the number of synapses (by the criterion of nc82 antigen visualisation). This is in agreement with data showing that GluRIIA receptors cluster previously to the incorporation of Brp into the growing synapses (Fouquet et al., 2009, Owald et al., 2010). This increase in PSD number is maintained at 40 hours after PI3K expression (D42-Gal4/UAS-PI3K larvae, 817±61, n = 11; D42-Gal4 larvae: 637±35, n = 8) (Fig. 6C). Conversely, we observed a significant decrease in the number of PSDs at 40 hours after PI3K^DN overexpression in D42-Gal4/UAS-PI3K^DN (444±29, n = 5) versus D42-Gal4 (619±42, n = 5) larvae (supplementary material Fig. S4). The consistency of all these data sets demonstrates that presynaptic PI3K is necessary for both the formation of new synapses and nascent PSDs in the growing NMJ.

Previous electrophysiological recordings obtained on motor neurons overexpressing PI3K indicated an increase in frequency and amplitude of miniature endplate potentials (MEPPs, spontaneous release), whereas the evoked release produced only an increase in EPSP amplitude (Martín-Peña et al., 2006). Here, we aimed to directly assess whether the synaptogenic effect of PI3K can be assigned to the pre- or postsynaptic side. To this end, we generated larvae in which PI3K was selectively overexpressed in the postsynaptic compartment by using a myosin heavy chain Gal4 (Mhc-Gal4) driver (Schuster et al., 1996). We found that PI3K expressed postsynaptically did not elicit significant differences in the number of synapses compared to their sibling controls (Fig. 7C). Interestingly, the general morphology of the NMJ of Mhc-Gal4 larvae (Fig. 7A) is remarkably similar to Mhc-Gal4 PI3K-overexpressing larvae (Fig. 7B). This structural feature is radically different when it is compared with the increased branching observed when PI3K was driven presynaptically (e.g. D42-Gal4/UAS-PI3K larvae) (Martín-Peña et al., 2006). Furthermore, we analysed whether postsynaptic overexpression of PI3K affects PSD number using anti-GluRIIA and anti-HRP in both control (Mhc-Gal4) and PI3K larvae (Mhc-Gal4/UAS-PI3K) (Fig. 7D,E). No significant differences in the number of PSDs were found in PI3K larvae (747±48, n = 8) with respect to their sibling controls (691±52, n = 10) (Fig. 7F). Taken together, all these data indicate that the synaptogenic effect of PI3K inducing supernumerary, functional synapses, including PSDs increases, is attributable to its overexpression in the presynaptic compartment only.

As previously shown, the persistent high levels of PI3K activity are necessary not only for synapse formation, but also for its subsequent maintenance and functionality (Cuesto et al., 2011; Martín-Peña et al., 2006). Unfortunately, however, the lack of a valuable PI3K antibody in Drosophila precludes an unequivocal detection of the protein in the pre- or postsynaptic compartments. Electrophysiological recordings from larval motor neurons overexpressing PI3K indicated an increase in evoked EPSP size and increments in both MEPP frequency and MEPP amplitude (Martín-Peña et al., 2006). These features could be accounted for by either a presynaptic or a postsynaptic role of PI3K. Additionally, rat hippocampal cultured neurons stimulated with PTD4-PI3KAc, a PI3K-activating transduction peptide, showed an increase in basal mEPSC frequency without any augmentation in mEPSC amplitude (Cuesto et al., 2011), consistent with an increase in functional synapses without changes in the postsynaptic receptor properties. Moreover, a postsynaptic role of PI3K has been demonstrated on the maintenance of the physiological PIP3 levels at postsynaptic densities, most likely by controlling AMPA receptor turnover (Arendt et al., 2010). Thus, to date, both immunochemical and electrophysiological findings failed to fully assign the synaptic compartment from which PI3K is causing these phenotypes.

The data reported here demonstrate that postsynaptic GluRIIs respond to PI3K changes induced in the presynaptic side of the developing NMJ. It has been previously reported that immature PSDs typically were dominated by GluRIIA, being subsequently balanced by GluRIIB incorporation during PSD and synapse maturation (Schmid et al., 2008). This balance modifies the electrical properties of the glutamate receptor and seems to be directly correlated to increases of the presynaptic levels of Brp, crucial for mature presynaptic glutamate release (Schmid et al., 2008) and also for structural integrity of the active zone (Wagh et al., 2006). Thus, the integration of both pre- and postsynaptic elements controls efficiently the number of synapses and, hence, PSDs per NMJs (Reiff et al., 2002; Schmid et al., 2008; Sigrist et al., 2000; Sigrist et al., 2002). We have found here that presynaptic, but not postsynaptic, PI3K overexpression yields to an augmentation of all NMJ synaptic markers examined (Figs1–f02,3). By contrast, presynaptic PI3K downregulation does alter neither GluRIIA (Fig. 2) nor the two Dlg isoforms (Fig. 3) tested. However, the reduction of PI3K activity using PI3K^DN gives rise to a decrease in both synapse and PSD (quantified by GluRIID immunostaining) numbers (supplementary material Figs S1,S4). This apparent discrepancy is probably due to the fact that the mechanism of attenuating PI3K activity with a dominant negative (PI3K^DN) is not fully efficient to generate a detectable reduction in all GluRII subunits or Dlg protein levels.

Notably, here we have also found changes in the number of PSDs along larval development and NMJ maturation when PI3K overexpression is selectively restricted to third instar larvae (Fig. 6). Our data also indicate a timing of around 24 hours to generate increases in PSD number. However, we found no differences in synapse number at 24 hours, although there is a tendency towards increment in PI3K versus control larvae (Fig. 5). It is reasonable to assume that PI3K needs time to be transcribed, translated and accumulated in the NMJ to generate new synapses. Thus, the temperature shift, which allows PI3K expression only in the last 18 hours, is insufficient for the generation of fully mature synapses, detected by nc82 antibody, but enough to detect immature PSDs by GluRIIA immunostaining. Previous data have shown a synaptic half-life of around 24 h in larval NMJ (Rasse et al., 2005), but also in fully differentiated brain neurons (Martín-Peña et al., 2006). In mammals, dendritic spines are functional within a day after induction of long-term plasticity (LTP) in hippocampal slices (De Roo et al., 2008).

The dynamics of GluRIIA incorporation into PSDs is not affected by PI3K (Fig. 4). This feature could be explained by a PI3K role triggering the formation of additional new nascent synapses by increasing the number of sites where active zones will be formed, leaving unaffected the time of GluR incorporation at individual PSD. In turn, the number of PSDs could be incremented due to the recruitment of glutamate receptors from pools dispersed over the whole muscle cell membrane (Rasse et al., 2005) that could be able to respond to the PI3K-dependent increase of suitable synaptic sites. Previous findings obtained in excitatory synapses in vertebrates indicate that the PIP3 pathway is linked to AMPAR insertion in the membrane (Qin, Y. et al., 2005</citref>). In hippocampal neurons, PI3K localises and directly binds to the cytosolic C-terminus of the AMPAR (Man et al., 2003) and this PI3K-AMPAR association plays a significant role in sustaining synaptic transmission (Arendt et al., 2010).

Different mechanisms could for the PI3K effects on postsynaptic proteins revealed in our study: first, a higher expression of glutamate receptor subunits should be achieved by increased synthesis or by reduced local receptor protein degradation. Indeed, the regulation of both protein synthesis and turnover are key factors for synaptic terminal development and function at the Drosophila NMJ (van Roessel et al., 2004; Zalfa et al., 2006). Second, receptor subunits enter to the newly forming clusters from a diffuse pool of receptors, not by splitting from previously formed ones (Rasse et al., 2005). This indicates a crucial role for protein trafficking, diffusion and clustering mechanisms in the developing NMJ. In this context, SYD-2/Liprin-alpha has been implicated in both pre- and postsynaptic assembly by interacting with a multitude of synaptic proteins and by regulating synaptic cargo transport (Spangler and Hoogenraad, 2007) guiding transport of active zone components. Liprin family proteins steer transport in axons and dendrites (e.g. of AMPA receptors) to support synaptic specialisations and play a key role in AZ assembly function (Shin et al., 2003; Wagner et al., 2009; Wyszynski et al., 2002). Moreover, the presynaptic AZ-localised RhoGAP DSyd-1 acts in a trans-synaptic manner, by targeting DLiprin-α to maturing AZs, and also defining the amount and composition of glutamate receptors (GluRs) accumulating at PSDs (Owald et al., 2010).

The case of larval motor neurons studied here has a precedent on the optic ganglia in which presynaptic photoreceptors determine the number of synapses established with the lamina neurons (Canal et al., 1994). Here, the new data provide a molecular mechanism and highlight the instructive role of PI3K in the regulation of synapse number and postsynaptic proteins in the NMJ. Also, these data have paved the way to understand the trans-synaptic signals needed for the formation, maturation and dynamic regulation of the synapse.

Fly line D42-Gal4 was provided by Harold L. Atwood (University of Toronto, Canada) (Parkes et al., 1998); OK6-Gal4 was from by Cahir J. O'Kane (University of Cambridge, United Kingdom) (Aberle et al., 2002); elav-Gal4 was from Bloomington Stock Center (Lin and Goodman, 1994) and Mhc-Gal4 line was a gift from Jean-François. Ferveur (Burgundy University, France) (Schuster et al., 1996). UAS-PI3K strain was provided by Juan Botas (Baylor College of Medicine, Houston, TX) (Leevers et al., 1996) and the UAS-PI3K^DN line, carrying a dominant-negative PI3K mutation, was a gift from Sally Leevers (Cancer Research Center, London) (Leevers et al., 1996). Both UAS-GluRIIA^RNAi and UAS-Dlg^RNAi lines were obtained from the Vienna Stock Center (Vienna, Austria; http://stockcenter.vdrc.at/control/main). Line Gal80^ts was a gift from Ronald Davis (The Scripps Research Institute, FL) (McGuire et al., 2003). The GluRIIA^RFP and GluRIIA^GFP constructs were provided by Stephan J. Sigrist (Institute of Biology of the FU Berlin and of the NeuroCure cluster of Excellence, Germany) (Rasse et al., 2005; Schmid et al., 2008). All crosses were reared at 25°C unless otherwise indicated.

Third instar larvae were dissected and fixed in 4% paraformaldehyde in PBT (PBS + 0.05% Triton X-100), except for anti-GluRIIA, for which larvae were fixed with 100% methanol at −20°C. Larvae were then washed briefly in 0.05% PBT, preincubated in blocking solution [95% PBT + 5% normal goat serum (NGS, Sigma)] for 30 minutes and incubated overnight at 4°C in blocking solution with the following primary antibodies: monoclonal antibody nc82 (1:10, DSHB, IA), anti-Brp (1:250, from Stephan J. Sigrist) to recognise Bruchpilot (Brp), an active zone protein at synapses. Whereas nc82 antibody recognises the C-terminal of Brp, the anti-Brp antibody is directed against a N-terminal peptide (Fouquet et al., 2009), anti-Dlg (1:250, DSHB) to label Discs Large, a postsynaptic protein homologous to the vertebrate DLG-MAGUK SAP97 protein (Thomas et al., 2010), anti-GluRIIA (1:100, DSHB), anti-GluRIIB (1:2000, from David Featherstone, UIC Biological Science, Chicago, IL), anti-GluRIIC (1:1000, from Aaron DiAntonio, Developmental Biology Department, University of Washington, WA), anti-GluRIID (1:500, from Stephan J. Sigrist) and anti-HRP (1:200, Jackson ImmunoResearch Laboratories), which binds to the membrane of NMJ boutons. The following secondary antibodies were applied for 3 hours at room temperature: Alexa 488 (goat anti-mouse, 1:500, Molecular Probes), Alexa 568 (goat anti-rabbit, 1:500, Molecular Probes), Cy3 (goat anti-mouse, 1:500, Invitrogen) and Cy5-HRP (goat, 1:250, Jackson ImmunoResearch Laboratories). Larvae were finally mounted in Vectashield (Vector Labs). Images from muscles 6-7 (segment A3) were acquired with a Leica Confocal Microscope TCS SP5 II (Mannheim, Germany). Serial optical sections at 1024×512 (100×50 µm) or 1024×1024 pixels (100×100 µm) were obtained at 0.5 µm with the 63× objective. At least eight larvae per genotype were analysed.

Methodology and experiments were performed as previously published (Füger et al., 2007; Rasse et al., 2005). Briefly, we used a UAS line in which the GluRIIA subunit has been tagged with both RFP (GluRIIA^mRFP, red fluorescent protein) and GFP (GluRIIA^GFP, green fluorescent protein) markers in control (OK6-Gal4) and experimental larvae (OK6-Gal4/UAS-PI3K). Larvae were anaesthetised with Desfluorane (Baxter) during image acquisition to avoid body wall movements. Serial optical sections (512×512 pixels) were taken at 0.5 µm with a 63× objective (50×50 µm) from NMJs of third instar larvae (muscles 26–27, segment A3) using a Leica Confocal Microscope TCS SP5 II (Mannheim, Germany). We considered three different points of analysis: before bleaching, 0 hours after bleaching and 14 hours after bleaching. Only the red channel (excited with a 561 nm diode-pumped solid-state laser) was bleached leaving the green channel (excited using the 488 nm ArKr laser line) unaffected as an internal reference. Following anaesthesia, larvae were placed in a plate containing agar, sugar and apple juice. 14 hours later, we took the last picture and observed the recovery of the GluRIIA^mRFP signal. This recovery was measured as a difference (%) between the intensity value in the bleached area and the intensity value in the surrounding unbleached regions, allowing the quantification of the GluRII entry in the NMJ. Data are based on a minimum of seven larvae per genotype.

Control (D42-Gal4) and experimental (D42-Gal4/UAS-PI3K and D42-Gal4/UAS-PI3K^DN) specimens were maintained after egg hatching at 18°C to keep Gal80^ts repressor active, blocking D42-Gal4 expression (McGuire et al., 2003). Third instar larvae were then shifted to 30°C, allowing Gal4 expression. After 24 and 40 hours of development, larvae were dissected and processed for immunostaining. On average, ten larvae per genotype were analysed.

Fillets or brains of third instar larvae were collected, frozen at −80°C, and lysed in lysis buffer: 50 mM Tris-HCl, 0.3 M NaCl, Complete Protease Inhibitor Cocktail (Roche, Switzerland), 1% Triton X-100 and phosphatase inhibitor cocktails 1 and 2 (Sigma-Aldrich). Membranes were then incubated with primary antibodies: anti-Dlg, 1:1250, and anti-GluRIIA, 1:1000 (Concentrate form, DSHB, Iowa), with 3% non-fat dry milk (Bio-Rad) in 0.05% PBST (PBS + Tween 20, Sigma-Aldrich) overnight at 4°C. They were then washed with double distilled water twice and incubated with a secondary peroxidase-conjugated antibody goat anti-mouse IgG for 1.5 hours at room temperature. Membranes were finally washed with PBST (2×10 minutes) and revealed with SuperSignal (Thermo Scientific). Western blot bands were scanned and quantified using a GS800 densitometer and Quantity One software (Bio-Rad). The graphs in Figs 2,3 and Supplementary material Figs S2,S3 represent the average of five independent western blots experiments.

ImageJ software (version 1.44, http://rsb.info.nih.gov/ij/) was used to determine the number of synapses and PSDs in immunostaining and FRAP experiments. PSDs (visualised with GluRIIs antibodies) were isolated and counted using the Analyze Particle tool (Rasse et al., 2005) and individual synapses (nc82 positive puncta) were quantified by using the Point Picker plug-in. A minimum of 10–12 larvae from each genotype was analysed.

All data are shown as mean ± s.e.m. Statistical significance was calculated using a Student's two-tailed t-test. The Kolmogorov-Smirnov test for normality was always performed before application of the Student's t-test for statistical significance. ANOVA test was used to compare absolute number of PSDs of all GluRII subunits in Fig. 1 and also to compare protein levels (western blots) in Figs 2,3. The software GraphPad Instat 3 was used throughout. Significant differences between compared groups was noted by *P<0.05, **P<0.005 and ***P<0.0005.

We thank colleagues who kindly provided fly strains. We also thank the Bloomington and Vienna Stock Centers. We appreciate the critical comments of Alberto Ferrús, laboratory members and colleagues of the Cajal Institute and the technical help from the Sigrist lab to S.J.A.