Remodelling neuronal connections by synaptic activity requires membrane trafficking. We present evidence for a signalling pathway by which synaptic activity and its consequent Ca2+ influx activate the small GTPase Ral and thereby recruit exocyst proteins to postsynaptic zones. In accord with the ability of the exocyst to direct delivery of post‐Golgi vesicles, constitutively active Ral expressed in Drosophila muscle causes the exocyst to be concentrated in the region surrounding synaptic boutons and consequently enlarges the membrane folds of the postsynaptic plasma membrane (the subsynaptic reticulum, SSR). SSR growth requires Ral and the exocyst component Sec5 and Ral‐induced enlargement of these membrane folds does not occur in sec5−/− muscles. Chronic changes in synaptic activity influence the plastic growth of this membrane in a manner consistent with activity‐dependent activation of Ral. Thus, Ral regulation of the exocyst represents a control point for postsynaptic plasticity. This pathway may also function in mammals as expression of activated RalA in hippocampal neurons increases dendritic spine density in an exocyst‐dependent manner and increases Sec5 in spines.
Cell morphology determines many functional aspects of a neuronal network and achieving the correct morphology requires precisely regulated protein and membrane traffic to specific domains. The primary shape of a neuron is established during axon and dendrite outgrowth and synapse formation, but is subjected to subsequent modifications by physiological events. In response to changes in synaptic activity, neurons can alter both pre‐ and postsynaptic elements of the synapse, including the number, size, and shape of dendritic spines (Kennedy and Ehlers, 2006; Alvarez and Sabatini, 2007; Bourne and Harris, 2007; Hanus and Ehlers, 2008; Newpher and Ehlers, 2009; Kasai et al, 2010). Because of the importance of these morphological events, membrane trafficking is emerging as a key aspect of neuronal development and plasticity. Membrane addition is critical for permitting neurite outgrowth and branching (Steiner et al, 2002; Lalli and Hall, 2005) and both membrane addition and membrane internalization are required for the growth and retraction of spines (Holtmaat and Svoboda, 2009; Newpher and Ehlers, 2009; Kelly et al, 2011). Exocytic trafficking from recycling compartments contributes to dendritic spine growth in response to activity (Park et al, 2006; Kennedy et al, 2010), but the main source of membrane responsible for this growth and the signals that control membrane addition remain elusive. Moreover, the localization of these events to precise regions of the neuronal surface is necessary to establish or modify appropriately synaptic connectivity.
The exocyst is a protein complex that can govern the polarized cell‐surface delivery of membrane and membrane proteins (Munson and Novick, 2006; Wu et al, 2008; He and Guo, 2009; Jin et al, 2011). The exocyst comprises eight proteins conserved from yeast to man and, although not required for the exocytosis of synaptic vesicles (Murthy et al, 2003), the exocyst may be important in other aspects of synapse growth and plasticity through its involvement in and regulation of the tethering, docking, and fusion of post‐Golgi vesicles with the plasma membrane. It is required for neurite outgrowth and the addition of neuronal membrane proteins, including the insertion of glutamate receptors, and the maturation of photoreceptors (Brymora et al, 2001; Vega and Hsu, 2001; Murthy et al, 2003; Sans et al, 2003; Beronja et al, 2005; Liebl et al, 2005; Mehta et al, 2005; Gerges et al, 2006). The distribution of the exocyst within a cell can be highly dynamic, consistent with its role in directing membrane fusion to specific domains (Boyd et al, 2004; Beronja et al, 2005; Mehta et al, 2005; Zhang et al, 2008; Murthy et al, 2010).
Recent studies have examined the manner in which the exocyst is assembled and localized and have identified regulatory interactions of exocyst proteins with several small GTPases (Lipschutz and Mostov, 2002; Wu et al, 2008), including the two mammalian Ral isoforms, RalA and RalB. Like the exocyst, Ral is expressed in the nervous system (Ngsee et al, 1991; Huber et al, 1994; Peng et al, 2004; Han et al, 2009). RalA binds directly to the exocyst members Sec5 and Exo84 and this interaction is thought to promote complex assembly, vesicle exocytosis, and membrane addition (Moskalenko et al, 2002, 2003; Sugihara et al, 2002; Fukai et al, 2003; Mott et al, 2003; Wang et al, 2004; Jin et al, 2005; Hase et al, 2009). Ral can be activated either by Ras indirectly via a Ral‐GEF (Guanine nucleotide Exchange Factor) or by Ca2+/calmodulin binding (Hofer et al, 1994, 1998; Kikuchi et al, 1994; Wolthuis et al, 1998; Wang and Roufogalis, 1999; Wolthuis and Bos, 1999; Harvey et al, 2008), and it is inactivated by PKC phosphorylation of the effector Sec5 (Chen et al, 2011). Both Ca2+ and Ras can regulate synaptic development and plasticity (Tada and Sheng, 2006; Alvarez and Sabatini, 2007; Harvey et al, 2008). We therefore took advantage of the evolutionary conservation of the exocyst and Ral pathways in Drosophila, to ask whether a Ral/exocyst pathway also plays a role in modifying synapses.
The Drosophila neuromuscular junction (NMJ) is a glutamatergic synapse that has proven to be a good model system to study synaptic development and plasticity (Collins and DiAntonio, 2007). Plastic changes to its morphology include a 10‐fold increase in synaptic boutons from the first‐ to the third‐larval instars and activity‐dependent changes in bouton number and shape. A prominent feature of this synapse, though one that has received less attention, is the elaborate set of postsynaptic folds of the plasma membrane called the subsynaptic reticulum (SSR). This structure develops during the second half of larval life and is also present at the NMJs of other arthropods (Jahromi and Atwood, 1974; Rheuben, 1985; Feeney et al, 1998). Glutamate receptors reside in the muscle membrane immediately across the synaptic cleft from presynaptic active zones. The folds of the SSR underlie these receptor fields and surround the presynaptic boutons. The SSR contains scaffolding proteins and cytoskeletal specializations (Rheuben et al, 1999; Ataman et al, 2006), but the significance of the SSR is poorly understood. The time course of its development suggests that it may be an adaptation that allows receptor‐activated currents to depolarize the low input resistance of these large muscle fibres. Its architecture, like that of mammalian dendritic spines, likely allows signal compartmentalization and the generation of postsynaptic micro‐domains that can sculpt synaptic responses.
We here report that the SSR is a plastic structure whose size depends on synaptic activity. Synaptically driven Ca2+ influx activates synaptically localized Ral and, by recruiting the exocyst complex to the postsynapse, causes the postsynaptic membranes to grow. Our initial studies in mammalian neurons further suggest conservation of Ral/exocyst‐induced postsynaptic growth as a form of plasticity.
Postsynaptic activated Ral recruits Sec5 to the synapse
To examine Ral regulation of the exocyst at the NMJ, we expressed previously characterized wild‐type, constitutively active, and inactive transgenes of Ral (RalWT, RalCA, and Ralinact) in the larval body wall by using a muscle‐specific driver, MHC‐Gal4. The RalCA and Ralinact transgenes contain mutations (G20V and S25N) that render Ral in the active GTP‐bound or inactive GDP‐bound state (Mirey et al, 2003). A newly developed antibody to Drosophila Ral demonstrated that both RalCA and Ralinact became highly concentrated at the NMJs of third‐instar larvae (Figure 1A and B). This antibody recognized a band corresponding to endogenous Ral and absent from ral mutant larvae on a western blot (Supplementary Figure S1A), but was not adequate to detect endogenous Ral by immunocytochemistry. Because Ral directly binds to Sec5 and Sec5 is a core component of the exocyst, we examined Sec5 localization at these synapses. As shown in Figure 1C, in wild‐type larvae, Sec5 immunoreactivity was observed in both the nerve and muscle and only modestly enriched around the NMJ. When RalCA was expressed in the muscles, we observed a significant recruitment of Sec5 towards the area around the synaptic boutons (Figure 1D). The phenomenon occurred at type I boutons throughout the larva but was analysed selectively at the well‐characterized synapses on muscles 6 and 7 in segments A2 and A3. In contrast to the effect of RalCA, Ralinact expression produced no detectable change in Sec5 distribution (Figure 1E). Expression of Ral wild‐type (RalWT) in the muscle showed a RalCA phenotype, though less pronounced, likely due to partial activation of RalWT (Figure 1H). RalCA was also expressed in motor neurons via the driver OK6‐Gal4, and Sec5 distribution at the NMJ was unchanged (Figure 1F). Nor did muscle expression of a constitutively active transgene of Rab11 (Rab11CA), a small GTPase that has been shown to interact with the exocyst complex protein Sec15 (Wu et al, 2005), change Sec5 distribution (Figure 1G). Significant enrichment of Sec5 to the synaptic region occurred only when RalCA or RalWT was expressed in the muscle (Figure 1H). When we examined other exocyst components, Sec15 immunoreactivity (Figure 1I–J) and the epitope‐tagged transgenes Sec3‐HA and Sec8‐HA (Supplementary Figure S1B–E) also became concentrated in the vicinity of the boutons in muscles expressing RalCA. Thus, activated Ral appears to recruit exocyst complexes and not just Sec5 to the NMJ. Because both RalCA and Ralinact (Figure 1A and B) localize to the NMJ, but only active Ral can recruit Sec5 to the NMJ (Figure 1D and E), it indicates that the synaptic localization of Ral is independent of its nucleotide state, but the recruitment of the exocyst occurs only in the GTP‐bound state.
Consequences of postsynaptic RalCAexpression on synaptic proteins
To understand the relevance of the Ral‐mediated Sec5 recruitment to the NMJ, we recorded from muscle 6 in third‐instar larvae, and examined the distribution of known components of that synapse. Only minor electrophysiological differences were observed between muscles expressing RalCA and controls: the amplitude of evoked excitatory postsynaptic potentials (EPSPs) was unchanged and the spontaneous miniature EPSP (mEPSP) amplitude and frequency were mildly reduced (Supplementary Figure S2A). The presence of normal EPSPs despite smaller mEPSPs can likely be attributed to a homeostatic increase in quantal content (Davis and Bezprozvanny, 2001; Heckscher et al, 2007). The number of synaptic boutons also appeared unchanged by RalCA expression in the muscle (control=131±6 versus muscle RalCA=133±5) but immunocytochemical characterization of synaptic proteins indicated shifts in synaptic composition. The presynaptic markers Bruchpilot (Brp) and Synaptotagmin I (SytI), which reside in active zones and synaptic vesicles, were similar in pattern and intensity in control and RalCA‐expressing muscles (Figure 2A). To characterize postsynaptic composition, we examined the distribution and quantified the intensity of immunolabelling for the glutamate receptor subunits GluRIIA and GluRIIB, the scaffolding protein Discs Large (Dlg), the cell‐adhesion molecule Fasciclin‐II, the postsynaptic signalling proteins Par‐1, Pak, and Pix, and the structural proteins α‐Spectrin and Syndapin‐1. Many of these components were not significantly altered by the expression of RalCA (Figure 2A). There were, however, some differences: GluRIIB receptor and Par‐1 levels were increased by 22 and 35%, respectively, and Dlg was reduced by 20% (Figure 2A; Supplementary Figure S2B and C). The greatest change, however, was observed in levels of the F‐bar protein, Syndapin‐1, which increased by ∼70% of control levels (Figure 2A–C).
Syndapin is a peripheral membrane protein located in the SSR and its F‐bar domain may contribute to the folding and tubulation of this postsynaptic membrane network (Kumar et al, 2009; Wang et al, 2009; Rao et al, 2010). Overexpression of Syndapin promotes the growth of the SSR (Kumar et al, 2009), and the increase in Syndapin at the NMJ therefore suggested that exocyst recruitment by Ral might have caused SSR growth.
Postsynaptic expression of RalCApromotes SSR expansion
To examine the involvement of Ral and the exocyst in SSR growth, we performed transmission electron microscopy (TEM) on control and RalCA‐expressing muscles. The extent of the SSR can be estimated by measuring the width of the zone of complex membrane infoldings surrounding the presynaptic boutons, from where it contacts the bouton to where the normal cytoplasm of the muscle begins (Figure 2D). Measurements of SSR width by TEM demonstrated Ral‐induced SSR growth. The SSR of muscles expressing RalCA was obviously larger in electron micrographs than that of controls (Figure 2D and E), increasing in width by 56% from 695±27 nm (n=58) to 1088±50 nm (n=51) (Figure 2D and G).
The increase in the SSR was apparent in each RalCA larva examined and frequently changed the overall appearance of the SSR. Whereas a discrete zone of SSR was normally found around each bouton of a control larva, we encountered several instances in RalCA‐expressing muscles where the SSR formed a continuous zone engulfing multiple boutons (Figure 2F). Also, RalCA expression gave rise to areas of muscle in which the SSR extended from the synapse deep into the muscle cell, well beyond the immediate vicinity of the presynaptic bouton. RalCA expression selectively evoked membrane overgrowth of the SSR; total muscle surface area was not changed, invaginations of the membrane were not encountered elsewhere along the muscle surface and other synaptic parameters quantified by TEM were also unchanged (Supplementary Table S1). Localized membrane overgrowth of the SSR is consistent with RalCA recruitment of the exocyst selectively to the synaptic region.
By using light microscopy, we confirmed the widespread SSR expansion surrounding type I boutons. We expressed CD8‐GFP in muscles, an exogenous membrane protein that indiscriminately labels the entire plasma membrane. Because of the abundance of membrane that forms the folds of the SSR, CD8‐GFP is brightest where SSR is present. The intensity of this fluorescence surrounding boutons and its thickness thereby reflects the packing density of SSR membrane and SSR thickness. We found that both parameters were increased at the synapse when RalCA was expressed (Supplementary Figure S2D–I), as expected from the EM. The colocalization of Sec5 and CD8‐GFP in muscles expressing RalCA (Supplementary Figure S2F and G) further supported the possibility that exocyst recruitment to the synapse reflected increased membrane addition for SSR growth. The six‐fold increase in Sec5 intensity caused by RalCA was far greater than the 1.5‐fold change in CD8‐GFP; thus, the exocyst was actively recruited to the postsynaptic region and not a passive consequence of the presence of more membrane.
sec5 and ral mutants have defects in SSR development
To address whether SSR development requires Sec5 and Ral, we examined the structure of the SSR in sec5 and ral mutant larvae. We analysed sec5 null mutants of which a few can survive for up to 96 h due to maternally contributed protein, but the majority die earlier (Murthy et al, 2003). We also analysed two ral mutants: ralPG89, which has been characterized (Balakireva et al, 2006) and survives to 60 h and ralG0501, which survives to late pupae but that has not previously been characterized. Both ral mutants lacked detectable protein expression by western blot (Supplementary Figure S3A) and were therefore processed for TEM. The SSR of wild‐type muscles starts to form around 48 h (early second instar) and continues to grow through the third instar (Supplementary Figure S3B–D). At 60 h, the SSR is reliably present but slightly thinner than that of the third instar and 60 h was therefore chosen as a suitable time point at which to examine the influence of Sec5 and Ral on the rate of its development. sec5 mutant muscles had only a rudimentary SSR, considerably thinner (149±9 nm in sec5−/−) than that of controls (580±33 nm) (Figure 3A, B). In these mutants, only one or two SSR folds were visible around most boutons (Figure 3B); among 34 boutons examined from three sec5 larvae, only one bouton had a nearly normal thickness of SSR, but it extended only around half the bouton, with the other half lacking SSR (Supplementary Figure S3E). SSR size and complexity were more variable in ral mutants (ralPG89=240±20 nm and ralG0501=260±27 nm) than in sec5 mutants, but their SSR was significantly smaller than control larvae (Figure 3C, D; Supplementary Figure S3F and G). Expression of Sec5 or RalWT back into the mutant larvae fully rescued SSR development (Supplementary Figure S4C–E), confirming that the observed defects are specific to these genes. As in third‐instar larvae, RalCA expression enlarged the SSR of control larvae at 60 h (909±70 nm) (Figure 3E and J). However, when RalCA was expressed in muscles of sec5 mutants, no SSR expansion was seen: SSR thickness (158±11 nm) was comparable to that in sec5 mutants without RalCA (149±9 nm) (Figure 3F and J). Conversely, expression of UAS‐Sec5 in muscles of ralPG89 was also not sufficient to rescue SSR formation (Supplementary Figure S4A and B). Thus, development of the SSR is dependent on both Ral and Sec5 and the extent of its growth can be influenced by activated Ral in a Sec5‐dependent manner.
We also analysed the phenotypes of third‐instar larvae in genotypes that survived to that stage: ralG0501 mutants and larvae expressing Sec5 RNAi in the muscle (muscle Sec5‐IR) (Supplementary Figure S5). ralG0501 third‐instar larvae had a very small SSR compared to control and indeed the SSR at this stage was not significantly different from ralG0501 mutants (237±27) at 60 h. The decreased SSR therefore is not due to a generalized developmental delay (these larvae survive to pupation) but rather to a selective arrest of SSR growth. Likewise, in the third‐instar muscle Sec5‐IR, only a rudimentary SSR was observed, which was not different from sec5−/− (Dicer control: 698±44 nm versus muscle Sec5‐IR: 167±20 nm) (Figure 3H–J).
In addition, to determine if the synaptic localization of Ral (Figure 1A and B) required Sec5, we immunostained control and sec5−/− muscles expressing RalCA. RalCA surrounded the boutons of both genotypes even though there was little or no SSR in the sec5−/− muscles (Figure 3K–L). In control animals, expression of RalCA induced Sec5 recruitment starting at 48 h and Sec5 levels increased as the SSR expanded at 60 h (Figure 3E). Thus, Ral localization to the synaptic area is upstream of Sec5 activation and does not require Sec5, whereas Sec5 recruitment by RalCA parallels the growth of the SSR and is required for that growth.
Ca2+influx recruits Sec5 to the synapse
Two pathways that can activate Ral are Ras activation of Ral‐GEF (Hofer et al, 1994; Kikuchi et al, 1994) and direct Ca2+‐calmodulin binding to Ral (Hofer et al, 1998; Wang and Roufogalis, 1999). To determine if Ras acts in Drosophila muscle to recruit Sec5 to the NMJ, we expressed in the muscle a previously characterized transgene for constitutively active Ras (RasCA) (Bergmann et al, 1998). RasCA did not cause the synaptic redistribution of Sec5 that was seen with RalCA expression (Supplementary Figure S6). To examine Ca2+/calmodulin‐dependent activation of Ral, we elevated intracellular Ca2+ with the Ca2+ ionophore calcimycin. Wild‐type third‐instar larvae were dissected in saline and then treated with calcimycin for 5 min in either 0 or 1 mM Ca2+ saline. Calcimycin was removed and the preparations were rested for 15 min in 0 Ca2+ before fixation. Ca2+ influx via calcimycin treatment significantly recruited Sec5 to the vicinity of the synaptic boutons in a manner that closely resembled the effect of RalCA expression (Figure 4A–C).
To test whether the Ca2+‐induced Sec5 recruitment is mediated by Ral, we compared calcimycin effects on wild‐type and ralPG89 mutant larvae, size and age matched at ∼60 h after egg laying (AEL). In wild‐type larvae, Ca2+ influx evoked Sec5 recruitment to the NMJ as had been seen in third‐instar larvae (Figure 4D, E). In the absence of Ca2+ influx, ralPG89 larvae were identical to controls; Ral is therefore not necessary for the expression and resting distribution of Sec5. In the ral mutant larvae, however, calcimycin treatment failed to recruit Sec5 to the NMJ (Figure 4F–H). Thus, the Ca2+‐induced Sec5 recruitment to the NMJ is Ral dependent.
Glutamate or nerve stimulation recruits Sec5 to the NMJ
To determine whether physiologic stimuli that elevate cytosolic Ca2+ could activate the Ca2+/Ral pathway, we examined Sec5 distribution upon glutamate application (Figure 5). In 60 h, larval muscles exposed to 30 μM glutamate, Sec5 was recruited to the NMJ, but only if Ca2+ was present in the medium and only in wild‐type but not in ralPG89mutant larvae (Figure 5A–D and I). Glutamate application also recruited Sec5 at third‐instar NMJs (Figure 5E, F). Thus, Ca2+ influx triggered by glutamate‐receptor activation can induce Ral‐dependent recruitment of Sec5 to synapses.
We used nerve stimulation for a more physiological activation of the glutamate receptors. We expressed the temperature‐activated channel dTrpA1 in motor neurons using the driver OK6‐Gal4 (Aberle et al, 2002; Hamada et al, 2008). The channel was activated with application of 37°C saline to the preparation for 5 min, which was subsequently returned to 20°C for 15 min before fixation. Sec5 immunoreactivity was compared to larvae lacking the dTrpA1 transgene but subjected to the same temperature shift. Nerve activation by dTrpA1 potently led to the accumulation of Sec5 at the synapse (Figure 5G, H). As an alternative means of nerve activation, a suction electrode was used to stimulate the nerve for 5 min at 10 Hz in 1 mM Ca2+, which was followed by a 15 min rest period before fixation and immunostaining. Sec5 immunoreactivity was significantly increased at the stimulated synapses relative to the unstimulated side (Supplementary Figure S7). The ability of stimulation by either method to recruit Sec5 to the NMJ is consistent with the ability of glutamate receptors to activate the Ca2+‐ and Ral‐dependent pathway and indicates that this recruitment is likely to occur under normal physiological conditions.
SSR size is modulated by activity
The ability of RalCA to enlarge the SSR raised the possibility that nerve stimulation over an extended period might accomplish a similar change in this structure. We therefore used the OK6‐Gal4 to express UAS‐dTrpA1 in motor neurons to increase their firing by increasing the temperature. Conversely, a transgene for a temperature‐sensitive, dominant‐negative allele of dynamin (UAS‐shits) was expressed to decrease synaptic transmission. We used two protocols: (1) shifting larvae to 30°C for 48 h and (2) placing them in tubes in a PCR machine and exposing them for 48 h to cycles of 15 min at 30°C followed by 45 min at 20°C. Larval survival was much higher with the latter method but for both methods larvae were 60 h AEL when the protocols were initiated and after the completion of the 48 h, were dissected and processed for TEM. Control larvae were exposed to the same temperature shifts. With either protocol, upregulation or downregulation of activity significantly altered SSR width. When UAS‐shits was expressed in motor neurons, the SSR was thinner than in control larvae (635±34 versus 789±41 nm for protocol 1, and 586±33 versus 862±57 nm for protocol 2). Conversely, when UAS‐dTrpA1 was expressed the SSR was significantly larger than in controls (1247±73 for protocol 1 and 1157±56 nm for protocol 2) (Figure 6; Supplementary Figure S8). We also quantified Syndapin, GluRIIB, and Dlg levels at synapses after submitting the larvae to 48 h of cycling temperatures in the PCR machine. Although these values showed modest changes of the sort that might be expected from activation of the Ral pathway, only the increase in the levels of Syndapin and GluRIIB upon TrpA1 activation reached statistical significance. The effect is weak probably because manipulations with TrpA1 and shits are less potent than constant expression of RalCA. Thus, the SSR is a plastic structure whose size can be regulated by the level of synaptic activity, and the changes in the SSR width correlated well with the ability of activity to recruit the exocyst to the synapse.
RalACAexpression in mammalian neurons increases dendritic spine density
The effect of Ral‐mediated exocyst activation on synapse morphology in Drosophila raised the possibility that Ral might also influence synaptic growth in mammalian neurons. As a preliminary inquiry into this question, we expressed RalA, the closest homologue of Drosophila Ral, in cultured hippocampal neurons. RalA and the exocyst are known to regulate early stages of neuronal development, including neurite formation (Lalli and Hall, 2005). To avoid the early developmental contribution of this pathway, we transfected neurons on the 19th day in vitro (DIV), at which time they have already differentiated and formed dendritic spines and synapses (Dotti et al, 1988). Three RalA constructs were used: RalA72L‐myc (RalACA) which is locked in the GTP‐bound state, RalA28N ‐myc (RalAinact) which is locked in the GDP‐bound state, and RalA72L/D49E‐myc (RalCAΔExo) in which a point mutation that blocks binding to the exocyst components Sec5 and Exo84 has been introduced into the RalACA‐myc construct (Lalli and Hall, 2005). These constructs were transfected with DsRed to fill the neurons and reveal their dendrite and spine morphology and PSD95‐YFP to mark postsynaptic densities. At 21 DIV, 48 h post transfection, we processed the neurons for immunocytochemistry and quantified the number of spines. All RalA constructs localized to both the cell body and dendrites (Supplementary Figure S9A). Neuronal morphology was not grossly altered by expression of these constructs, although a small but significant decrease in the number of distal branches and an increase in proximal branch density were seen upon RalACA expression relative to control neurons without a RalA transgene (Supplementary Figure S9B–D). Spine density per μm of dendrite, however, increased 32% when RalACA was expressed (control=0.30±0.016 versus RalACA=0.39±0.023 spines/μm of dendrite) (Figure 7A and B). In contrast, neither RalAinact nor RalACAΔExo altered spine density (0.27±0.023 and 0.29±0.022 spine/μm) relative to controls. When only those spines positive for PSD95‐YFP were counted, i.e., those that are likely to be part of functional synapses, RalACA expression caused a 50% increase (Figure 7D) relative to controls or the other RalA transgenes (control=0.21±0.027, RalACA=0.32±0.034, RalACAΔExo=0.18±0.029, and RalAinact=0.23±0.016 spines with PSD95/μm). The increase in PSD95‐positive spines after expression of RalACA relative to RalAinact is also present when endogenous PSD95 rather than the YFP‐tagged transgene is used to identify those spines (RalACA=0.15±0.015 and RalAinact=0.07±0.016) (Supplementary Figure S10). Although spine density changed, their average neck length and head diameter did not (Figure 7E and F). In summary, activated RalA expression in hippocampal neurons increased the density of PSD‐95‐positive spines and this increase was blocked by a mutation that prevents the interaction of RalA with Sec5.
Neuronal depolarization mimics RalACAexpression by promoting Sec5 trafficking into dendritic spines
To further explore whether the Ral/exocyst pathway may also regulate mammalian synaptic structure, we asked whether expression of the RalA constructs would have any impact on exocyst localization. We transfected hippocampal neurons DIV19 with HA‐Sec5, GFP, and myc‐tagged constructs for RalACA, RalAinact, or RalCAΔExo. At DIV21, neurons were processed for immunocytochemistry. The three Ral constructs expressed equivalently well in the neurons analysed (Supplementary Figure S11) and in each case the Sec5 transgene was present throughout the soma and dendrites. However, quantification of HA‐Sec5 intensity and distribution revealed that RalACA expression increased levels of Sec5 in spines, whereas RalAinact and RalCAΔExo expression did not (Figure 8A–D).
Because synaptic activity caused exocyst redistribution in a Ral‐dependent manner in Drosophila, we asked whether depolarization of the hippocampal cultures would have a similar effect on HA‐Sec5 immunostaining as RalACA expression. For this, we transfected hippocampal neurons DIV19 with HA‐Sec5 and GFP, and 48 h later depolarized the neurons with 60 mM KCl in the medium for 3.5 h. The intensity of HA‐Sec5 immunoreactivity was not significantly altered in the neuronal somas, or in dendrites but was significantly increased in spines when compared to untreated neurons (Figure 8E–H). These results suggest that spine anatomical plasticity in the hippocampus may indeed share mechanistic features with the Ral/exocyst pathway we have characterized in Drosophila.
In this study, we establish that the postsynaptic plasma membrane of the Drosophila NMJ is subject to activity‐dependent morphological plasticity and identify a pathway that can link synaptic activity to postsynaptic membrane growth: the Ca2+‐dependent activation of Ral and consequent recruitment of the exocyst (Figure 9). We show that postsynaptic activation of Ral, but not presynaptic, recruited to the NMJ Sec5, Sec15, Sec3, and Sec8 (Figure 1; Supplementary Figure S1), four components of the exocyst. Because Ral is known to promote the assembly of the exocyst complex, it is likely that functional complexes are recruited. Sec5 recruitment was also stimulated by increasing cytosolic Ca2+, a known activator of Ral. Ca2+‐dependent exocyst recruitment could be accomplished with an ionophore, by applying glutamate, or by nerve stimulation either via direct electrical excitation or activation of dTrpA1 channels. Thus, synaptic activity and the resulting influx of Ca2+ can recruit Sec5 to the postsynaptic compartment of the muscle fibres (Figures 4 and 5). In ral mutant larvae, Ca2+ failed to recruit Sec5. The ability of synaptic activity to activate this pathway predicts that activity levels during larval life determine the amount of exocyst‐dependent targeting of membrane vesicles to the synaptic region. The significance of chronic activation of the pathway in larval muscle was confirmed by the ability of activated Ral expression to promote enlargement of the membranes of the SSR in an exocyst‐dependent fashion, while other aspects of synaptic morphology and composition were unchanged (Figures 2 and 3). Finally, we established that the SSR is indeed a plastic structure whose size is regulated by levels of synaptic activity (Figure 6). Given the ability of synaptic activity to recruit Sec5 to the synapse upon acute stimulation, it seems reasonable to hypothesize that the chronic activation of the pathway during development was responsible for the activity‐dependent control of SSR growth (Figure 9).
The SSR is a membrane system conserved in several arthropods (Jahromi and Atwood, 1974; Rheuben, 1985; Feeney et al, 1998) that harbours postsynaptic machinery (Ataman et al, 2006). Its function is not understood but, given its architecture, it is possible that it functions in ways akin to those that are ascribed to dendritic spines: sequestering signalling molecules for localized responses to receptor activation and sculpting the electric signals that are propagated into the postsynaptic cell. Our observation that RalA can also act via the exocyst to alter spine number in hippocampal cells represents an additional and provocative similarity. SSR development starts at 48 h during the second instar and continues through the third‐instar stage. During this interval, larvae grow considerably and synaptic boutons are added at the NMJ to match the enlargement of the muscle fibres (Ataman et al, 2006; Prokop, 2006; Collins and DiAntonio, 2007). Our data indicate a necessary role for the Ral/exocyst pathway in the formation of the SSR that accompanies this growth. Mutations in either ral or sec5 interfered with formation of the SSR and expression of RalCA in the muscle increased the SSR (Figures 2 and 3). The action of Ral was upstream of the exocyst, because RalCA could not promote SSR growth in a sec5 mutant background. Activation of the pathway alone, however, was not sufficient to induce SSR growth: RalCA expression prior to 48 h AEL, the time point at which the SSR normally starts developing (Supplementary Figure S3B; Guan et al, 1996), neither recruited Sec5 to the NMJ nor caused the premature formation of SSR. We hypothesize that components of the SSR are not present in the muscle prior to 48 h and therefore cannot be trafficked to the synapse in response to Ral. It is likely that SSR development entails two components: an unknown developmental signal that initiates transcription and translation of the building blocks and a membrane trafficking mechanism, dependent on Ral and the exocyst, that mediates and modulates the assembly of those components at the synapse. Because the exocyst can direct membrane addition to particular membrane domains (Wu et al, 2008; He and Guo, 2009), the Ral/exocyst pathway at the NMJ likely regulates the rate and location of membrane addition, rather than other aspects of SSR development.
The localization of both Ral and the exocyst components to the region surrounding the bouton was a notable feature of this pathway. However, Ral localization to the NMJ was independent of Sec5 and of the nucleotide‐bound state of Ral. In addition, expressed Ral concentrated at the synapse prior to SSR growth and Ral is thus well situated to direct vesicle addition to postsynaptic membranes upon activation by Ca2+ and when appropriate cargo vesicles have been produced. Sec5, in contrast, was recruited to this region only when Ral was activated. Two lines of evidence demonstrate that this recruitment was a cause and not a consequence of the enlargement of the SSR. First, the growth of the SSR does not occur in sec5 mutants or when Sec5 is selectively knocked down in the muscle by RNAi. Second, whereas CD8‐GFP is always abundant surrounding boutons due to the concentration of membrane in that region, Sec5 moves to the SSR only in response to the activation of Ral.
SSR development is also under the likely control of other signals, including Neuroligin/neurexins (Banovic et al, 2010), Pix and Pak (Parnas et al, 2001; Albin and Davis, 2004), Dlg and Spectrin (Lahey et al, 1994; Budnik et al, 1996; Guan et al, 1996; Pielage et al, 2006), Par‐1 (Zhang et al, 2007b), and CamKII (Koh et al, 1999). In particular, the wingless/wnt pathway modulates SSR development through nuclear signalling and control of the export of nuclear transcripts (Packard et al, 2002; Ataman et al, 2008; Mosca and Schwarz, 2010; Speese et al, 2012). Because the Wnt‐ligand Wingless is secreted from the innervating boutons in an activity‐dependent manner (Packard et al, 2002; Ataman et al, 2008) Wnt signalling may act in parallel to the Ca2+/Ral pathway with the former increasing the synthesis of necessary components and the latter stimulating their incorporation into the postsynaptic membrane.
The acquisition of proper synaptic morphology has multiple components and represents an important late stage of synaptic maturation (Bhatt et al, 2009; Newpher and Ehlers, 2009). It includes not only the elaboration of postsynaptic specializations such as spines, folds, or the SSR, but also determination of presynaptic bouton number and size, two processes that are also carefully controlled at the fly NMJ (Griffith and Budnik, 2006; Marques and Zhang, 2006). Notably, the Ca2+/Ral/exocyst pathway altered postsynaptic morphology independently of other properties of the NMJ: presynaptic properties including bouton number and size were unaffected when SSR growth was promoted and glutamate receptors were largely unaltered. The increase in the SSR may occur primarily in its deeper folds and this zone is marked by Syndapin (which was increased surrounding boutons by RalCA expression) but lacks Dlg (which did not increase). The ability to regulate independently the growth of the postsynaptic membrane is reminiscent of spine dynamics at mammalian synapses, where synaptic contacts can be converted from shaft synapses to spine synapses and where the volume and shape of spines can change in response to activity without necessarily changing synapse number (Alvarez and Sabatini, 2007; Bourne and Harris, 2007; Bhatt et al, 2009; Newpher and Ehlers, 2009). We found that spine density was increased by activated RalA and that both activated RalA and neuronal depolarization increase Sec5 in dendritic spines. These findings raise the possibility that the mechanism by which the postsynaptic membrane expands with spine growth in response to glutamate receptor activation (Kwon and Sabatini, 2011) may resemble the Ral/exocyst pathway by which the glutamatergic synapse at the fly NMJ expands its postsynaptic membrane. Indeed spine growth is likely to be a multifaceted event involving regulated changes in cytoskeletal elements, membrane addition, and receptor trafficking under the control of multiple small GTPases (Park et al, 2004; Murakoshi et al, 2011; Murakoshi and Yasuda, 2012).
Both glutamate receptor activation and Ca2+ influx through calcimycin promoted Sec5 recruitment selectively to the vicinity of the synapse (Figures 4 and 5). Calcimycin, however, permeabilizes the entire muscle membrane to Ca2+. That Sec5 was recruited selectively to the NMJ and not uniformly across the muscle surface suggests that a synaptic targeting mechanism must be present independent of the site of Ca2+ entry and regardless of whether the Ca2+ enters through glutamate receptors or voltage‐dependent channels. We propose instead that the targeting is determined by the localization of Ral to the membrane surrounding the boutons, which in turn recruits and activates the exocyst. The subcellular localization of the exocyst in these activated muscles is consistent with the known function of the exocyst in directing post‐Golgi vesicles to target membranes for localized membrane addition (Zhang et al, 2001; Dupraz et al, 2009; He and Guo, 2009; Murthy et al, 2010).
Our studies also indicated that some aspects of exocyst function do not require Ral. Although loss of function mutations in ral prevented exocyst‐dependent SSR growth, other exocyst‐catalysed processes persisted. Sec5 is required for neurite outgrowth and bouton addition at the fly NMJ (Murthy et al, 2003), but, even though ral mutants have reduced synaptic arbors, NMJs were present in ral−/− embryos and the ralG0501 allele survives to pupation though no Ral protein is detectable. Moreover, expression of RalCA in neurons did not alter presynaptic morphology. Similarly, in mammalian neurons, both Ral‐dependent and ‐independent exocyst functions are found. Ral, via the exocyst, promotes neurite branching and participates in establishing neuronal polarity (Lalli and Hall, 2005; Lalli, 2009) but exocyst‐mediated AMPA receptor targeting, insertion, and recycling (Sans et al, 2003; Gerges et al, 2006; Mao et al, 2010) are thought to be independent of Ral. Conversely, Ral also has exocyst‐independent neuronal functions, including regulation of NMDA receptor endocytosis in response to LTD, via RalBP1 (Han et al, 2009). The complex interplay of Ral and the exocyst clearly requires further elucidation.
The identification of a pathway by which synaptic activity can promote growth of postsynaptic membranes at the fly NMJ and the observations that RalACA expression in hippocampal neurons increases spine density and both RalCA and depolarization recruit Sec5 to spines, raise the question as to whether the same Ca2+/Ral/exocyst pathway functions at mammalian synapses. Ca2+ entry is a consequence of excitatory transmission in the mammalian brain and mammalian neurons contain both Ral and the exocyst (Brymora et al, 2001). While RalA activity promotes the outgrowth and branching of neurites in newly dissociated neurons in a Sec5‐dependent manner (Lalli and Hall, 2005), at later stages, it may regulate aspects of dendritic spine dynamics. In particular, LTP increases spine volume, a process that requires the addition of membranes whose source remains largely unknown (Cooney et al, 2002; Park et al, 2006; Bourne and Harris, 2007). As with the growth of the SSR, spine dynamics represent changes in postsynaptic architecture that are not necessarily coupled with the formation of new synaptic contacts. The glutamatergic NMJ of Drosophila has provided insight into molecular pathways of synaptic development and function (Collins and DiAntonio, 2007) that are shared with mammalian synapses. Further studies may establish if a mechanism similar to the Drosophila pathway reported here might also figure in spine growth or remodelling in mammals.
Materials and methods
Fly stocks and genetics
Flies were raised at 25°C in cages with grape plates and yeast paste for timed collections. The wild‐type strain used was w1118, control larvae were obtained by crossing MHC‐Gal4 or OK6‐Gal4 with either w1118, y, w or FRT40A. Stocks used were UAS‐RalWT, UAS‐RalCA, and UAS‐Ralinact (Mirey et al, 2003), ralPG89 (Balakireva et al, 2006), ralG0501 (Bloomington Stock Center), UAS‐RasCA (Bergmann et al, 1998), MHC‐Gal4, MHC‐Gal4‐Gene‐Switch, OK6‐Gal4 (Aberle et al, 2002), G14‐Gal4, UAS‐Rab11CA‐YFP (Zhang et al, 2007a), y,w; sec5E10 (Murthy et al, 2003), y,w; sec5ts1 (Murthy et al, 2010), UAS‐Sec5, UAS‐Sec5‐IR (VDRC‐ w1118; P(GD13789)v28873), UAS‐dTrpA1 (Hamada et al, 2008), and 2XUAS‐shits (Bloomington Stock Center).
Antibody production and immunocytochemistry
A polyclonal antibody against Drosophila Ral protein was raised in guinea pig by injection of the entire Ral protein (produced in bacteria as a GST fusion) (Eurogentec, Belgium). For immunocytochemistry, third‐instar larval fillets were dissected and fixed in 4% paraformaldehyde (Sigma‐Aldrich) for 20 min or in Bouin's fixative solution (Ricca Chemical Company) for 5 min. For Ral immunostaining in 60 h larvae, animals were fed 5 μg/ml RU486 to increase Ral expression. Antibody staining was performed in PBS containing 0.3% Triton X‐100 and 10% normal donkey serum. Larvae were incubated overnight at 4°C in primary antibodies, washed for at least 1 h, blocked for 30 min to 1 h and incubated for 2 h at room temperature in secondary antibodies, diluted in blocking solution. Larval fillets were mounted in Vectashield (Vector Labs). The following primary antibodies were used: guinea pig anti‐dRal (1:500), mouse anti‐Sec5 (1:40) (Murthy et al, 2003), guinea pig anti‐Sec15 (1:1500) (Mehta et al, 2005), guinea pig anti‐Syndapin (1:1000) (gift of V Kumar and M Ramaswami), rabbit anti‐Dlg (1:10 000) (Koh et al, 1999), rabbit anti‐GluRIIB (1:2000) (Marrus et al, 2004), mouse anti‐GluRIIA, (1:50) (Developmental Studies Hybridoma Bank‐DSHB), mouse anti‐Brp (1:100) (Hofbauer et al, 2009), rabbit anti‐Synaptotagmin I (1:4000) (Mackler et al, 2002), and rabbit anti‐Par1 (1:200). Cyanine 5 (Cy5)‐, cyanine 3 (Cy3) or FITC‐conjugated goat anti‐HRP (1:200; Jackson ImmunoResearch) was used to label neuronal membranes; for Sec5 staining, we used minimum crossreactivity DyLight 549‐conjugated secondary antibodies (1: 200; Jackson ImmunoResearch). For all remaining reactions, we used minimum crossreactivity secondaries that were A488‐, Cy2‐, DyLight 488‐, A568‐ or DyLight 549‐conjugated antibodies (1:200, Jackson ImmunoResearch).
Larvae were dissected in Ca2+‐free saline (5 mM Mg2+) (Jan and Jan, 1976). For unstimulated control preparations, this saline included 0.5 mM EGTA. Dissected larvae were then pre‐incubated in 50 μM calcimycin/0 mM Ca2+ saline for 10 min, after which they were either placed for 5 min in calcimycin/1 mM Ca2+ saline or left in the calcimycin/0 mM Ca2+ as a control. The larvae were then washed in 0 mM Ca2+ saline and left at rest for 15 min. When comparing wild‐type and ral mutants, 60 h larvae of each genotype were dissected on the same Sylgard‐coated slide and processed together to ensure equivalent treatments.
Larvae were dissected in Ca2+‐free saline (Jan and Jan, 1976) and then exposed for 5 min to 1 mM Ca2+ saline with or without 30 μM Glutamate (Glutamic Acid, Sigma Aldrich), followed by 15 min rest in 1 mM Ca2+ saline. Larvae were then fixed and processed for immunocytochemistry.
Stimulation of motorneurons using UAS‐dTrpA1.
OK6‐Gal4 larvae were crossed to w1118 or to UAS‐dTrpA1 flies and kept at 22°C. Third‐instar larvae from the two sets of crosses were dissected at room temperature in Ca2+‐free saline (Jan and Jan, 1976) and then incubated for 5 min in 1 mM Ca2+ saline prewarmed to 37°C. The warm solution was then replaced by a 20°C solution for 15 min prior to fixation.
Manipulations of activity of larval motorneurons.
For the temperature cycling paradigm, 60 h larvae of control (OK6‐Gal4/+), OK6‐Gal4/UAS‐dTrpA1, or OK6‐Gal4/UAS‐shits were placed in perforated PCR tubes with yeast paste, and subjected to a regimen of 15 min at 30°C followed by 45 min at 20°C for a total of 48 h in a PCR machine (Eppendorf Mastercycler Pro). Larvae (now at the third instar) were dissected and processed for TEM. For the constant temperature shift, the same genotypes were transferred to 30°C for a 48 h period, after which the third‐instar larvae were dissected and processed. Few OK6/TrpA1 larvae survived the sustained 30°C for 48 h, but in the cycling protocol viability was good and equivalent for each genotype.
Larvae aged 48±1 h, 60±1 h, or 66±1 h, or wandering third instars were dissected in cold 0.1 M cacodylate buffer. Third‐instar larvae were fixed at 4°C overnight in 2.5% paraformaldehyde, 5.0% glutaraldehyde, 0.06% picric acid in 0.1 M cacodylate buffer. Second‐instar larvae were processed similarly but fixed for 1 h at room temperature. For the analysis of second instars, each genotype was collected at 60 h but sec5−/− larvae were small, resembling 48 h controls. Larvae were processed for TEM as previously reported (Mosca and Schwarz, 2010). Ultrathin sections were taken parallel to the surface of the muscles and were mounted on single slot grids, stained with lead and uranyl acetate, and imaged on a Tecnai G2 Spirit BioTWIN (FEI Company) electron microscope. A minimum of three larvae were quantified per genotype.
Larvae were imaged on a Zeiss LSM 510 Meta laser scanning confocal microscope (Carl Zeiss, Oberkochen, Germany) with a 63 × 1.4 NA objective. Optical sections were 1 μm except for Figures 1A, B, 3K and Supplementary Figure S2F and G, which were 0.5 μm. Images were processed in separate channels using the LSM software or Adobe Photoshop CS4 (Adobe Systems, San Jose, CA). Each experiment was repeated at least three independent times and quantified data are always from at least four different larvae, segments A2 and A3, and a minimum of eight NMJs. For each set of quantification, data were acquired with identical settings and analysed as 8‐byte images, maximum projection, with the same settings, using Fiji/ImageJ (NIH, Bethesda, Maryland).
Quantification of Sec5 and other synaptic markers.
To define the perisynaptic region, anti‐HRP staining was thresholded so as to outline synaptic boutons. For quantification of Sec5, threshold settings and parameters were kept constant for all data points. After thresholding, the HRP‐positive area was expanded by 2 μm (5 iterations on ImageJ) to form a synaptic region of interest that contained the postsynaptic region and within which the intensity of the Sec5 immunostaining could be quantified. The integrated fluorescence intensity was normalized to the area of the region of interest for total Sec5 at NMJ/μm2. To determine the degree of enrichment at the synapse, intensity/area of Sec5 immunostaining was also measured in a non‐synaptic region of the muscle in an area of at least 10 × 10 μm. Synaptic Sec5 enrichment was defined as the difference between the normalized intensities of the synaptic and non‐synaptic regions. To quantify other postsynaptic markers (Figure 2), we also measured intensity of immunofluorescence within an expanded region around the boutons, but for pre‐synaptic markers, the anti‐HRP‐positive area was not expanded. All values in Figure 2A are expressed as a percentage of the intensity in control larvae.
TEM analysis of SSR thickness.
To quantify SSR thickness, micrographs were analysed using Fiji/ImageJ (NIH, Bethesda, Maryland). Boutons were identified by the presence of synaptic vesicles and active zones and eight straight lines, at 45° angles from one another, were drawn from the centre of the bouton. The width of the SSR along each of these radii was measured and a value for each bouton was derived from the average of the eight radii.
Statistical analysis was conducted using Excel (Microsoft Corporation, 2007) or GraphPad Prism 6 software (Graphpad Software, La Jolla, CA) and significance values calculated using a two‐tailed unpaired t‐test or, when more than two samples were compared, a one‐way analysis of variance (ANOVA) test followed by post hoc Bonferroni test. If our measurements of SSR thickness did not clearly fit a normal distribution (as visualized with a qqplot), and were instead positively skewed, then we applied a log transformation to those data sets to convert them to a symmetrical distribution. The transformed data could then be processed using parametric t‐tests or ANOVA. All histograms and measurements are shown as mean±s.e.m.; sample size (n) is described either in the figure legend or in Results section.
Hippocampal culture, transfection, and immunostaining
Primary hippocampal cultures were prepared from embryonic day 18 (E18) Long‐Evans rat brains as described previously (Xia et al, 1996). Cells were plated on coverslips (Bellco Glass, Inc., Vineland, NJ) coated with poly‐ornithine and laminin at a density of 1 × 105/cm2. Transfections were done using Lipofectamine™ 2000 (Life Technologies, Carlsbad, CA) at DIV19. Two days post transfection (DIV21) neurons were fixed and immunostained as described previously (Tolias et al, 2007). Images were acquired with a Zeiss LSM510 confocal microscope (Zeiss, Thornwood, NY) using a 63 × 1.4 NA objective for spine analysis and 40 × 1.3 NA objective for Sholl analysis. For details of constructs, antibodies, image acquisition parameters, and analysis, see Supplementary data.
Conflict of Interest
The authors declare that they have no conflict of interest.
Supplementary Information [emboj2013147-sup-0001.pdf]
This work was supported by National Institutes of Health (NIH) Grants RO1 NS041062 and MH075058 (TLS), the IDDRC Imaging Core (P30HD18655), Harvard Neurodiscovery Imaging Center, and the HMS Conventional Electron Microscopy Facility. We thank X Chen, A Saltiel, G Lalli, H Bellen, V Kumar, M Ramaswami, H Stellar, Vivian Budnik, A DiAntonio, Y Jan, L Zipursky, P Garrity and the Developmental Studies Hybridoma Bank and Bloomington Drosophila Stock Center for DNA constructs, antibodies, and fly stocks; Ann Goldstein for fruitful discussions and comments; and David Van Vactor, Elizabeth McNeill, and members of the Schwarz laboratory for critical comments on the manuscript.
Author contributions: ROT and TLS designed all experiments and wrote the manuscript. GP, ROT, AN, and TLS designed and analysed the experiments for Figures 7 and 8; Supplementary Figures S9–S11, GP and AN conducted these experiments. MEHK conducted and analysed the electrophysiology experiments in Supplementary Figure S2A. MB and JC designed and produced the anti‐dRal antibody. IGM conducted and analysed the data for Supplementary Figure S7. ROT conducted and analysed all other experiments.
- Copyright © 2013 European Molecular Biology Organization