Corresponding author: M. Stern, Department of Biochemistry and Cell Biology MS-140, Rice University, 6100 Main Street, Houston, TX 77005, USA. E-mail: email@example.com
The lipid kinase PI3K plays key roles in cellular responses to activation of receptor tyrosine kinases or G protein coupled receptors such as the metabotropic glutamate receptor (mGluR). Activation of the PI3K catalytic subunit p110 occurs when the PI3K regulatory subunit p85 binds to phosphotyrosine residues present in upstream activating proteins. In addition, Ras is uniquely capable of activating PI3K in a p85-independent manner by binding to p110 at amino acids distinct from those recognized by p85. Because Ras, like p85, is activated by phosphotyrosines in upstream activators, it can be difficult to determine if particular PI3K-dependent processes require p85 or Ras. Here, we ask if PI3K requires Ras activity for either of two different PI3K-regulated processes within Drosophila larval motor neurons. To address this question, we determined the effects on each process of transgenes and chromosomal mutations that decrease Ras activity, or mutations that eliminate the ability of PI3K to respond to activated Ras. We found that PI3K requires Ras activity to decrease motor neuron excitability, an effect mediated by ligand activation of the single Drosophila mGluR DmGluRA. In contrast, the ability of PI3K to increase nerve terminal growth is Ras-independent. These results suggest that distinct regulatory mechanisms underlie the effects of PI3K on distinct phenotypic outputs.
The lipid kinase PI3K, which phosphorylates the phospholipid PIP2 to form PIP3, is activated by extracellular ligands such as growth factors and neurotransmitters, acting through receptor tyrosine kinases (RTKs) and G protein coupled receptors, respectively (Pignataro & Ascoli 1990; Zhu et al. 2001). The most prominent outcome of PI3K activity is phosphorylation and activation of the downstream kinase Akt (Cohen et al. 1997). In the nervous system, activated PI3K/Akt promotes neuronal survival (Brunet et al. 2001), growth and maturation of synaptic dendrites (Jaworski et al. 2005; Kumar et al. 2005), nerve terminal growth (Howlett et al. 2008; Martín-Peña et al. 2006) and several aspects of synaptic plasticity, including long term synaptic depression mediated by the metabotropic glutamate receptor [mGluR-mediated long term depression (LTD)] (Hou & Klann 2004; Klann & Dever 2004). PI3K is activated by RTKs when the p85 regulatory subunit of PI3K binds to the phosphotyrosines that accompany RTK activation. This p85-driven membrane localization is generally sufficient for PI3K activation (Funamoto et al. 2002; Klippel et al. 1996).
An alternative, p85-independent route to PI3K activation is provided by activated Ras (Rodriguez-Viciana et al. 1994). Active Ras binds directly to the PI3K catalytic subunit p110 at amino acids distinct from those required for p85 binding (Orme et al. 2006; Pacold et al. 2000). Because many stimuli that activate PI3K activate both p85 and Ras, it is often unclear if a particular stimulus activates PI3K via Ras-dependent, p85-independent or Ras-independent, p85-dependent mechanisms (Mazzoni et al. 1999). Addressing this question is necessary for a more complete understanding of the mechanisms by which PI3K activity is regulated. Furthermore, several neurological disorders including autism and Neurofibromatosis are associated with increased PI3K activity (Cuscó et al. 2009; Dasgupta et al. 2005; Jeon et al. 2011; Johannessen et al. 2005; Serajee et al. 2003), but it is not known to what extent, if any, Ras is required for this increase. Therefore elucidating the Ras dependence or Ras independence of PI3K activation in neuronal processes will have medical significance.
Two processes within Drosophila larval motor neurons are regulated by PI3K: synaptic bouton number and motor neuron excitability in response to activation of the single metabotropic glutamate receptor DmGluRA (Howlett et al. 2008; Martín-Peña et al. 2006). To determine if PI3K activation requires Ras for these processes, we tested effects of chromosomal Ras mutations, transgenes predicted to decrease Ras activity, and a PI3K variant (termed ‘gen-PI3KRBD’) in which Ras binding, but not p85 binding, was eliminated by mutation. We found that the effect of PI3K on nerve terminal growth was Ras-independent. In contrast, PI3K required Ras activity to mediate effects of DmGluRA on motor neuron excitability. Consistent with this observation, Ras was also required for glutamate-induced increases in phosphorylated Akt, a major product of PI3K activity within motor nerve terminals. Thus, PI3K requires distinct activators depending on the specific physiological process affected.
For all experiments, Drosophila larvae were reared on standard cornmeal/agar media at 22°C. The D42 Gal4 driver, which expresses in motor neurons (Parkes et al. 1998), was provided by Tom Schwarz (Harvard Medical School, Boston, MA, USA). The UAS-PI3KDN, UAS-PI3K-CAAX, the PI3K Ras binding variant gen-PI3KRBD and the gen-PI3K+ control transgenes were provided by Sally Leevers (London Research Institute, London, UK), the PI3K null allele PI3KA was provided by Ernst Hafen (ETH, Zurich, Switzerland), UAS-PTEN was provided by Bruce Edgar (Fred Hutchinson Cancer Research Center, Seattle, WA, USA), UAS-RasN17 was provided by Denise Montell (Johns Hopkins Medical School, Baltimore, MD, USA), Rase2F was provided by Gerald Rubin (University of California, Berkeley CA, USA), Ras12a was provided by Celeste Berg (University of Washington, Seattle, WA, USA), and the Ras-RNAi transgene was provided by the National Institute of Genetics (Mishima, Japan). All other fly stocks were provided by the Drosophila stock center (Bloomington, IN, USA).
Larvae were grown to the wandering third instar stage in uncrowded half-pint bottles at 22°C and were collected only from bottles that had been producing wandering larvae for fewer than 3 days. For phosphorylated Akt (p-Akt) measurements made prior to glutamate application, larvae were dissected in Schneider's Drosophila media (Gibco, Grand Island, NY, USA) and fixed in Schneider's Drosophila media containing 4% paraformaldehyde. For p-Akt measurements following glutamate application, larvae were dissected as above and then incubated for 1 min in Schneider's Drosophila media containing 100 μm glutamate, followed by fixation. Fixed larval tissues were incubated with a rabbit anti-Drosophila p-Akt (Ser505) primary antibody (1:500 dilution, Cell Signaling Technologies, Beverly, MA, USA), followed by Rhodamine Red conjugated goat anti-rabbit IgG (1:500 dilution, Jackson ImmunoResearch), and Cy-2 conjugated antibodies against horseradish peroxidase (1:200 dilution, Jackson ImmunoResearch, West Grove, PA, USA). Immunolabelled larval tissues in standard phosphate buffered saline (PBS, 0.128 m NaCl, 2.0 mm KCl, 4.0 mm MgCl2, 0.34 m sucrose, 5.0 mm HEPES, pH 7.1 and 0.15 mm CaCl2) containing 50% glycerol were mounted onto slides. Neuromuscular junctions (nmjs) from muscles 6 and 7 in segments A3 through A6 were imaged on a Zeiss LSM 510 confocal microscope system (Zeiss, Oberkochen, Germany) with a × 20 objective. Optical sections were 10 μm thick, which encompassed the entire nmj. Optical parameters, including pinhole, gain, contrast and brightness, were held constant for each experimental set. Nmjs, marked by anti-HRP, were traced using the Image-J (NIH, Bethesda, MD, USA) freehand selection tool and the selection was transferred to the anti-p-Akt image where pixel intensity value was quantified within the traced region. Background obtained from the non-neuronal area of muscles 6 and 7 from the same abdominal segment was averaged and subtracted from the mean p-Akt pixel intensity. Mean p-Akt pixel intensity was normalized to wildtype (WT).
To measure bouton number, larvae were grown, selected and dissected as described above, fixed in standard PBS containing 4% paraformaldehyde and labelled with Cy-2 conjugated antibodies against horseradish peroxidase (1:200 dilution). The neurons innervating muscles 6 and 7 from abdominal segments A3 and A4 were utilized for this analysis. Images were obtained as described above. The Image-J cell counter function was used to quantify bouton numbers. The two axons innervating muscles 6 and 7 generate distinct bouton types, the larger (3–6 µm) termed 1b and the smaller (2–4 µm) termed 1s (Atwood et al. 1993). The number of 1b and 1s boutons was counted manually using the Image-J cell counter application.
Larvae were grown and selected as described above, dissected Jan's buffer (128 mm NaCl, 2.0 mm KCl, 4.0 mm MgCl2, 34 mm sucrose, 4.8 mm HEPES, pH 7.1 and CaCl2 concentration as specified in the text). Peripheral nerves were cut immediately posterior to their exit from the ventral ganglion, and were stimulated with a suction electrode at a 5-V stimulus intensity. Muscle recordings were taken from muscle 6 in abdominal sections 3–5. Stimulus duration, approximately 0.05 milliseconds, was adjusted to 1.5 times threshold which reproducibly stimulates both axons innervating muscle cell 6. Intracellular recording electrodes for muscle potentials were pulled with a Flaming/Brown micropipette puller to a tip resistance of 10–40 MΩ and filled with 3 m KCl.
For all experiments, wandering third instar larvae of either sex were selected. The arithmetic mean was used for all of the data except for the electrophysiological experiments as shown in Fig. 3. For these experiments the geometric means were used because the data show a positive skew (Bland & Altman 1996). Student's t-test was used for the p-Akt analysis as shown in Fig. 2, and for examining the effects of the RafDN transgene as shown in Fig. 5. All other data were analysed with a one-way anova. Fisher's LSD was used as the post hoc test. Genotype was variable for all experiments except for the p-Akt analysis shown in Fig. 2b, for which glutamate application was variable. Values with a P < 0.05 were judged as significantly different. Statistical analysis was performed using Microsoft Excel.
Ras is required for glutamate-mediated activation of PI3K
Glutamate application to filleted third instar Drosophila larvae activates PI3K in motor nerve terminals, as assayed by monitoring levels of phosphorylated Akt (p-Akt), which are increased by PI3K (Colombani et al. 2005; Dionne et al. 2006; Howlett et al. 2008; Palomero et al. 2007). Glutamate activates PI3K via the single metabotropic glutamate receptor DmGluRA: the DmGluRA112b null mutation, or presynaptic RNAi knockdown of DmGluRA, each block this glutamate-evoked PI3K activation (Howlett et al. 2008).
PI3K is activated by either the p85 regulatory subunit or Ras (Fig. 1). To determine if Ras is required for DmGluRA-mediated PI3K activation, we inhibited Ras activity in three ways. First, we used the D42 motor neuron Gal4 driver (Brand & Perrimon 1993; Parkes et al. 1998) to express the dominant-negative RasN17 variant within motor neurons (Lee et al. 1996). This RasN17 mutant sequesters the guanine nucleotide exchange factor, an essential Ras activator, thus preventing activation of the WT Ras (Powers et al. 1989). We term larvae of this genotype ‘D42>RasN17’ to indicate that the D42 Gal4 element drives expression of the UAS-driven RasN17 transgene. Second, we used the Gal4 system to express a Ras-RNAi construct in motor neurons (termed D42>Ras-RNAi). Third, we constructed larvae heterozygous for two chromosomal Ras mutations, Rase2F and Ras12a, which in combination decrease Ras activity sufficiently to confer phenotypes, but retain sufficient activity to maintain viability (Zhong 1995). We found that whereas glutamate application to WT larvae induced a significant increase in p-Akt levels (Fig. 2; t(63) = 3.17, P = 0.002), any of the three methods employed for decreasing Ras activity blocked the ability of glutamate to increase p-Akt levels (Fig. 2; D42>RasN17, t(46) = 1.76, P = 0.092; D42>Ras-RNAi, t(56) = 0.569, P = 0.571; Rase2F/Ras12a, t(58) = 0.436, P = 0.665). These observations suggest that Ras is a critical intermediate in the PI3K activation mediated by glutamate-liganded DmGluRA.
To confirm that Ras inhibition prevented p-Akt increases by preventing PI3K activation, rather than via a distinct Ras effector pathway, we tested the prediction that a PI3K variant that cannot be activated by Ras would fail to be activated by glutamate application. This PI3K variant, termed ‘PI3KRBD’, carries a four amino acid substitution in the PI3K Ras binding domain that prevents activation by Ras but permits activation by p85 (Orme et al. 2006; Pacold et al. 2000). This variant was introduced into a genomic clone, which placed this allele under transcriptional control of the PI3K endogenous promoter, and then introduced into Drosophila as a transgene (Orme et al. 2006). A control transgene, encoding PI3K+, was constructed and introduced into Drosophila in parallel. Then the PI3KA deletion, which removes about 300 amino acids (including the kinase domain) from the C terminus of PI3K (Halfar et al. 2001; Weinkove et al. 1999), was introduced into each fly line above by a genetic cross. Thus, the transgene-based PI3KRBD or PI3K+ (called gen-PI3KRBD and gen-PI3K+ to distinguish these transgenes from the endogenous PI3K) were the sole source of active PI3K in these flies. gen-PI3K+ behaves equivalently to endogenous PI3K+ in all assays applied (Orme et al. 2006). Finally, PI3KA, gen-PI3K+ and PI3KA, gen-PI3KRBD flies were each crossed to flies bearing PI3KA, and experiments were performed on the resulting larvae.
We found that flies expressing gen-PI3K+ displayed a robust glutamate-evoked p-Akt increase (Fig. 2, t(113) = 4.79, P < 0.0005). In contrast, glutamate-evoked p-Akt increases were blocked in larvae expressing the gen-PI3KRBD allele (t(77) = 6.31, P = 0.55). These experiments confirm that Ras is a critical intermediate in the activation of PI3K by glutamate.
We also noted that basal p-Akt levels (p-Akt levels prior to glutamate application) were decreased in larvae carrying gen-PI3K+ in the PI3KA background (Fig. 2). The decrease in basal p-Akt in gen-PI3K+ might be a consequence of decreased PI3K gene dosage in these larvae. Alternatively, this decrease could reflect properties of the PI3KA allele: PI3KA is predicted to encode a 668 amino acid protein truncated at the C terminus, which might confer partial dominant-negative properties. To distinguish between these possibilities, we compared basal p-Akt levels in WT larvae, in PI3KA/PI3K+ larvae, and in PI3KA, gen-PI3K+/PI3K+ larvae. We found that larvae expressing PI3KA exhibited significantly less basal p-Akt than WT regardless of whether these larvae simultaneously expressed one dose (PI3KA/PI3K+) or two doses (PI3KA, gen-PI3K+/PI3K+) of PI3K+ (Fig. 2c, WT vs. PI3KA/PI3K+, t(56) = 2.25, P = 0.029; WT vs. PI3KA gen-PI3K+/PI3K+, t(56) = 3.69, P < 0.001.). We conclude that expression of PI3KA decreases basal p-Akt levels, most likely because PI3KA encodes a PI3K fragment with dominant-negative activity.
Basal p-Akt levels were even further decreased in gen-PI3KRBD larvae (Fig. 2), suggesting that PI3K activity in motor neurons is chronically decreased when activation by Ras is prevented. A similar decrease in basal p-Akt levels and a block in glutamate-evoked p-Akt increases were reported when either of two additional molecules, the Calcium/Calmodulin-dependent kinase II (CaMKII) and the non-receptor tyrosine kinase focal adhesion kinase (DFak), were inhibited (Lin et al. 2011).
The ability of PI3K to decrease motor neuron excitability requires Ras
In addition to blocking glutamate-evoked increases in p-Akt, the DmGluRA112b null mutation also increases motor neuron excitability (Bogdanik et al. 2004). This increased excitability results from an inability to activate PI3K: transgene-induced PI3K inhibition in motor neurons, accomplished by expressing the dominant-negative PI3KDN (Leevers et al. 1996), increases motor neuron excitability similarly to DmGluRA112b, whereas transgene-induced PI3K activation, accomplished by expressing the constitutively active PI3K-CAAX, decreases excitability (Howlett et al. 2008). Furthermore, PI3K activation is completely epistatic to DmGluRA112b for the control of neuronal excitability: excitability in motor neurons expressing PI3K-CAAX is unaffected by the DmGluRA genotype (Howlett et al. 2008).
If Ras is required for the DmGluRA-dependent activation of PI3K, as the experiments shown in Fig. 2 suggest, then preventing Ras from activating PI3K would similarly be predicted to increase motor neuron excitability. Thus we tested the prediction that Ras inhibition, or the presence of gen-PI3KRBD, would increase neuronal excitability in a manner similar to DmGluRA112b or PI3KDN.
To measure neuronal excitability, we used the same readout as was employed to measure neuronal excitability in DmGluRA112b and in PI3KDN: the rate of onset of a phenomenon termed ‘long term facilitation’ (LTF) (Jan & Jan 1978). LTF is a form of synaptic plasticity induced when a larval motor neuron is subjected to a train of repetitive nerve stimulations at a low external [Ca2+]. Because Ca2+ is essential for neurotransmitter release to occur, nerve stimulation at this low external [Ca2+] initially induces little neurotransmitter release, and thus low amplitude depolarizations (termed excitatory junctional potentials, or EJPs) in the target muscle. However, as repetitive stimulation continues, at a certain point during the stimulus train a threshold is reached, and subsequent stimulations elicit EJPs greatly increased in amplitude and duration. These increased EJPs occur as a consequence of increased and asynchronous neurotransmitter release, which in turn results from prolonged depolarization of the nerve terminal (Jan & Jan 1978) associated with the occurrence of supernumerary action potentials within the nerve (Stern et al. 1990; Stern et al. 1995; Howlett et al. 2008).
The number of stimulations required to reach this LTF threshold (LTF onset rate) is decreased by genetic conditions that increase neuronal excitability either by increasing sodium currents or decreasing potassium currents (Jan & Jan 1978; Mallart et al. 1991; Parker et al. 2011; Poulain et al. 1994; Schweers et al. 2002; Stern & Ganetzky 1989; Stern et al. 1990; Stern et al. 1995). Both the DmGluRA112b null mutation and PI3KDN expression greatly decrease the number of nerve stimulations required to reach LTF (Bogdanik et al. 2004; Howlett et al. 2008), which indicates increased excitability in these genotypes.
The observation that Ras is required for DmGluRA-dependent PI3K activation predicts that Ras inhibition would increase the LTF onset rate in a manner similar to DmGluRA112b and PI3KDN expression. To test this prediction, we measured LTF onset rate in larvae expressing the dominant-negative RasN17 in motor neurons, and in the Rase2F/Ras12a heterozygous combination. We found that as predicted, both genotypes significantly increased LTF onset rate (Fig. 3; One-way anova and Fisher's LSD gave the following differences at 3 Hz, 5 Hz, 7 Hz and 10 Hz, respectively: For D42>+ vs. D42>RasN17, F1,22 = 6.34, P = 0.02; F1,23 = 11.9, P = 0.002; F1,23 = 14.8, P = 0.0008; F1,23 = 8.99, P = 0.006; vs. Rase2F/Ras12A, F1,16 = 4.11, P = 0.06, F1,16 = 10.7, P = 0.005, F1,16 = 10, P = 0.006, F1,16 = 7.18, P = 0.016).
To confirm that Ras inhibition increases excitability by preventing PI3K activation, rather than via a distinct effector pathway, we compared LTF onset rate in larvae carrying the control gen-PI3K+ with larvae carrying gen-PI3KRBD. As shown in Fig. 3, LTF onset rate was significantly greater in gen-PI3KRBD than in gen-PI3K+ at the two higher stimulus frequencies (Fig. 3; PI3KA gen-PI3K+/PI3KA vs. PI3KA gen-PI3KRBD/PI3KA, F1,11 = 5.6, P = 0.037 and F1,11 = 17.1, P = 0.002 at 7 Hz and 10 Hz, respectively) which indicates increased motor neuron excitability in gen-PI3KRBD. The observation that an increased LTF onset rate was not observed at the lower stimulus frequencies is consistent with previous reports suggesting that LTF onset rates are not as affected at low stimulus frequencies in hyperexcitable genotypes (Howlett et al. 2008; Lin et al. 2011). We conclude that the ability of PI3K to decrease neuronal excitability requires Ras. This conclusion is consistent with the previous experiments suggesting that ligand-activated DmGluRA requires PI3K activity to decrease motor neuron excitability (Howlett et al. 2008), and that the ability of ligand activation of DmGluRA to activate PI3K requires Ras (Fig. 2).
Ras activity is not required for PI3K-mediated increases in bouton number
Bouton number at the larval nmj is regulated by PI3K activity: transgene-induced PI3K pathway inhibition, accomplished by expressing either the dominant-negative PI3KDN or PTEN, which encodes the phosphatase that opposes the effects of PI3K, decreases bouton number about twofold. In contrast, transgene-induced PI3K activation increases bouton number almost threefold (Howlett et al. 2008; Martín-Peña et al. 2006). The PI3K activators important for bouton number were not identified. However, the observation that DmGluRA112b decreases bouton number only modestly (Bogdanik et al. 2004) suggests that PI3K activators distinct from glutamate and DmGluRA might be necessary for nerve terminal growth.
It was previously reported that Ras regulates larval bouton number (Koh et al. 2002). We confirmed these findings: inhibiting Ras activity in motor neurons by expressing either the dominant-negative RasN17 or Ras-RNAi or by introducing the Rase2F/Ras12a alleles, significantly decreased bouton number (Fig. 4b, D42>+ vs. D42>RasN17, F1,49 = 6.47, P = 0.014; vs. D42>Ras-RNAi, F1,29 = 14.4, P < 0.001; WT vs. Rase2f/Ras12A, F1,45 = 31.4, P < 0.001). The larval muscles that were analysed are innervated by two motor axons that form boutons of different sizes, termed 1b and 1s (Atwood et al. 1993), and certain mutations can differentially affect 1b vs. 1s numbers (Romero-Pozuelo et al. 2007; Bhogal et al. 2011). We similarly found that altered Ras and PI3K differentially affect 1b and 1s numbers. In particular, PI3K pathway inhibition conferred by PTEN overexpression decreased each bouton type by about twofold (Fig. 4c, F1,23 = 44.3, P < 0.001; Fig. 4d, F1,23 = 11.36, P = 0.003), whereas decreased Ras activity also decreased 1b number by about twofold (Fig. 4c), but did not significantly affect 1s number (Fig. 4d). Thus, 1s bouton number is less sensitive than 1b number to decreased Ras activity.
Koh et al. (2002) reported that Ras activation was not sufficient to activate PI3K in motor nerve terminals. To determine if Ras activity is necessary for the PI3K-dependent increase bouton number, we compared bouton number in larvae carrying the control gen-PI3K+ with larvae carrying gen-PI3KRBD. We found WT bouton numbers in both of these genotypes (Fig. 5a,b). This observation suggests that the ability of PI3K to increase bouton number is Ras-independent. In contrast, motor neuron expression of the dominant-negative RafDN transgene decreased bouton number to a value similar to those conferred by Ras inhibition (Fig. 5a,b; D42>+ vs. D42>Raf DN, t(24) = 2.7, P = 0.012). In addition, Raf inhibition, like Ras inhibition, significantly decreases 1b bouton number (to 33 ± 7.6 vs. 51.5 ± 3.2, F1,16 = 10.6, P = 0.005) but not 1s bouton number (F1,16 = 0.59, P = 0.45;data not shown). We conclude that Ras regulates bouton number via Raf, as was reported previously (Koh et al. 2002), and that Ras activity is not required for PI3K to increase bouton number.
PI3K affects distinct physiological processes via distinct regulatory pathways
It was previously shown that activity of the lipid kinase PI3K affects two properties within Drosophila larval motor neurons: first, nerve terminal growth and second, neuronal excitability in response to activation of the single Drosophila metabotropic glutamate receptor DmGluRA. Here, we present evidence suggesting that PI3K affects these two properties in response to distinct upstream regulators. The ability of DmGluRA to activate PI3K and regulate neuronal excitability is Ras-dependent, whereas the ability of PI3K to regulate synaptic bouton number is Ras-independent. Thus, the DmGluRA112b null mutation strongly increases motor neuron excitability but affects arborization in only a modest way (Bogdanik et al. 2004), which suggests that the PI3K activity generated by DmGluRA activation is necessary to regulate excitability but not synaptic growth. In contrast, the PI3KRBD variant, which cannot be activated by Ras, increases neuronal excitability but elicits WT bouton number, suggesting that the PI3K that regulates bouton formation is not sufficient (or available) to inhibit neuronal excitability.
Two possible mechanisms might underlie these observations. First, given that PI3K regulates neuronal excitability and bouton number via distinct effector pathways (via Foxo inhibition and Tor/S6 Kinase activation, respectively; Howlett et al. 2008), it may be that the PI3K activity sufficient to activate the Tor/S6 Kinase pathway is not sufficient to inhibit Foxo. In this view, Ras is required for the PI3K-dependent regulation of excitability to provide the additional PI3K activity required for Foxo inhibition; this additional PI3K activity is not required for PI3K to activate Tor/S6 Kinase to a level sufficient to increase bouton number. Alternatively, neuronal excitability and bouton number might be regulated by functionally discrete, non-interacting, pools of PI3K; one pool, which regulates excitability, is activated by DmGluRA and requires Ras, whereas the second pool, which regulates nerve terminal growth, responds to unknown activators and is Ras-independent (Fig. 6).
Our data cannot distinguish between these possibilities. However, the possibility that excitability and bouton number are regulated by two discrete pools of PI3K is supported by the previous demonstration that Ras is not only unnecessary for the PI3K activity that regulates bouton number, but is also not sufficient for this activation (Koh et al. 2002). In particular, Koh et al. tested effects of the RasV12C40 double mutant on bouton number. RasV12C40 is rendered constitutively active by the V12 substitution, whereas the C40 substitution prevents Raf activation but permits PI3K activation (Halfar et al. 2001). Despite retaining ability to activate PI3K, RasV12C40 failed to increase bouton number (Koh et al. 2002). This observation does not appear to be consistent with the possibility described above in which less PI3K activity is required to increase nerve terminal growth than to inhibit excitability. Rather, this observation suggests that nerve terminal growth is affected by a pool of PI3K that is impervious to Ras. It has similarly been observed in the mammalian hippocampus that PI3K activated within specific dendritic spines is not free to diffuse to other spines within the same cell (Man et al. 2003); thus PI3K activity is capable of regulating physiological processes within specific subcompartments of a neuron.
Pathways regulating the PI3K-dependent promotion of arborization
The signals activating the ability of PI3K to increase synaptic bouton number are not completely known. However, the CaMKII is both necessary and sufficient to promote arborization, at least in part via PI3K (Beumer et al. 2002; Griffith et al. 1994; Koh et al. 1999; Lin et al. 2011; Park et al. 2002). For example, bouton number is decreased when CaMKII is inhibited by expression of an inhibitory peptide, and this decrease is completely suppressed by co-expression of the constitutively active PI3K-CAAX (Lin et al. 2011). This observation suggests that CaMKII inhibition decreases nerve terminal growth by preventing PI3K activation. This observation further suggests that synaptic activity and the resultant increase in intracellular [Ca2+] plays a critical role in the PI3K activation that promotes bouton formation.
The insulin growth factors, acting through the single Drosophila Insulin Receptor, provide one set of potential activators in addition to synaptic activity and intracellular Ca2+. Both insulin-like and Insulin Receptor-like immunoreactivity are present at Drosophila larval nmj (Budnik et al. 1990a,1990b) and both insulin and insulin growth factors are capable of activating PI3K in a Ras-independent manner (Backer et al. 1992).
PI3K-independent regulation of synaptic bouton number
Two pathways in addition to the PI3K pathway regulate bouton number. Activation of the Ras pathway increases bouton number via Raf/Erk: active Erk phosphorylates and consequently down-regulates the cell adhesion molecule FasII, the Drosophila orthologue of NCAM (Koh et al. 2002). This down-regulation appears to be necessary for bouton formation to proceed. CaMKII similarly promotes arborization by down-regulating FasII and its binding partner, DLG (Koh et al. 1999; Beumer et al. 2002, see Fig. 6). As described above, PI3K-CAAX remained able to increase bouton number even when CaMKII activity was blocked (Lin et al. 2011), suggesting that CaMKII pathway activity is no longer necessary to promote nerve terminal growth when PI3K is activated. It is not known if ectopic activation of PI3K can similarly abrogate the effects of Raf/Erk inhibition on nerve terminal growth.
A role for Ras in mammalian metabotropic glutamate receptor-mediated LTD?
In several subregions of the mammalian brain, ligand activation of mGluRs activates PI3K; this PI3K activation is necessary for a type of synaptic plasticity termed ‘mGluR-mediated LTD’ (Hou & Klann 2004; Ronesi & Huber 2008). Hypersensitivity to induction of mGluR-mediated LTD is implicated in the neurological deficits of fragile X and autism (Kelleher & Bear 2008), which is consistent with the observation that both fragile X and autism are associated with increased PI3K activity (Butler et al. 2005; Cuscó et al. 2009; Gross et al. 2010; Jeon et al. 2011; Kwon et al. 2006; Mills et al. 2007; Neves-Pereira et al. 2009; Serajee et al. 2003). Therefore there is interest in determining the mechanisms by which mGluR activation activates PI3K. The observation that mGluR-mediated LTD both activates and requires the activation of the MAP kinase Erk (Gallagher et al. 2004), which, like PI3K, is activated by Ras, supports the possible involvement of Ras in mGluR-mediated LTD. However, this possibility has not been thoroughly investigated.
The possibility that Ras participates in mGluR-mediated LTD has important implications for the genetic disease Neurofibromatosis (Nf1). Similar to individuals with fragile X, individuals with Nf1 exhibit a high incidence of autism (Kelleher & Bear 2008) as well as other cognitive deficits. Because Nf1 encodes a Ras-GTPase activator protein, which negatively regulates Ras, the loss of Nf1 in individuals with Neurofibromatosis causes Ras hyperactivation (Xu et al. 1990). If Ras mediates the effects of mGluR activation on LTD, then this Ras hyperactivation would cause hypersensitivity to mGluR-mediated LTD, similar to what is observed in fragile X, and therefore provide a mechanism underlying the increased frequency of autism in Nf1. Thus, our observation that Ras mediates the mGluR-dependent activation of PI3K in Drosophila motor nerve terminals might have important implications for cognitive deficits in Neurofibromatosis.
We are grateful to James McNew and Daniel S. Wagner for comments on the manuscript, Natalia Molinas for assistance in experiments, Sally Leevers, Gerald Rubin, Celeste Berg, Denise Montell, the Drosophila stock center at Bloomington, IN, and the National Institute of Genetics (Mishima, Japan) for providing fly stocks. Funded by grant W81XWH-09-1-0106 from the Department of Defense Neurofibromatosis Research Program and grant IOS-0820660 from the National Science Foundation (to M.S.).