Fos and Jun potentiate individual release sites and mobilize the reserve synaptic-vesicle pool at the Drosophila larval motor synapse

Abstract

In all nervous systems, short-term enhancement of transmitter release is achieved by increasing the weights of unitary synapses; in contrast, long-term enhancement, which requires nuclear gene expression, is generally thought to be mediated by the addition of new synaptic vesicle release sites. In Drosophila motor neurons, induction of AP-1, a heterodimer of Fos and Jun, induces cAMP- and CREB-dependent forms of presynaptic enhancement. Light and electron microscopic studies indicate that this synaptic enhancement is caused by increasing the weight of unitary synapses and not through the insertion of additional release sites. Electrophysiological and optical measurements of vesicle dynamics demonstrate that enhanced neurotransmitter release is accompanied by an increase in the actively cycling synaptic vesicle pool at the expense of the reserve pool. Finally, the observation that AP-1 mediated enhancement eliminates tetanus-induced forms of presynaptic potentiation suggests: (i) that reserve-pool mobilization is required for tetanus-induced short-term synaptic plasticity; and (ii) that long-term synaptic plasticity may, in some instances, be accomplished by stable recruitment of mechanisms that normally underlie short-term synaptic change.

Although the consolidation of long-term behavioral and synaptic changes has been shown to require transcription factors such as CREB, Fos, Jun, and SRF, the exact molecular strategies recruited to accomplish persistent synaptic changes remain poorly understood (1–3). The favored hypothesis is that formation of new stable synapses underlies enduring forms of synaptic potentiation (4). However, other candidate mechanisms, e.g., those that lead to sustained activation of pathways involved in short-term plasticity, have also been proposed (5, 6).

Among the first activity-regulated proteins discovered, Fos and Jun are basic leucine zipper family transcription factors that dimerize to form the transcriptional activator complex AP-1 (2). Recruited downstream of Ca²⁺-dependent signaling pathways (7), Fos and Jun are required for encoding persistent forms of behavioral plasticity including aspects of drug addiction and memory (2, 3, 8, 9). Indeed, stable expression of a Fos isoform in the striatum, observed after repeated cocaine exposure, underlies long-term sensitization to cocaine (3, 8). However, the synaptic mechanisms recruited by AP-1 remain poorly explored.

Drosophila larval motor synapses show increased synaptic strength when AP-1 is overexpressed in motor neurons (10). This synaptic enhancement is accompanied by increases in the quantal content of neurotransmitter release, and increases in the number of presynaptic varicosities (10). Here, we ask whether AP-1 mediated synapse enhancement can be explained by increases in synapse number, Ca²⁺ influx, Ca²⁺ sensitivity of vesicle fusion or synaptic vesicle number. Our observations support a model in which: (i) AP-1 induced synaptic enhancement occurs without an accompanying increase in synapse number; (ii) AP-1 increases the size of the cycling synaptic vesicle pool through mobilization of the reserve pool; (iii) that AP-1 causes persistent synaptic change by stably recruiting a cellular mechanism transiently used for posttetanic potentiation, a ubiquitous but poorly understood form of short-term synaptic facilitation. The relevance of these findings for mechanisms of Fos and Jun function and for cellular mechanisms of long-term synaptic plasticity is discussed.

Results

Increased Quantal Content of Transmitter Release from AP-1 Over-expressing Motor Neurons.

Previous studies have shown AP-1 overexpression in Drosophila motor neurons enhances glutamate release from motor terminals in a manner that is accompanied by an increase in bouton number (10). We confirmed these conclusions using failure frequency analysis, which, under conditions of very low Ca²⁺, measures frequency of “failure” to release even a single quantum of neurotransmitter. At 0.3 mM Ca²⁺, frequencies of failure events are reduced in C155/+;UAS Fos/+;UAS Jun/+ (hereafter referred to as “AP-1”) compared with control C155/+ hereafter “control”) synapses (Fig. 1 A–D). Therefore, this analysis confirmed quantal content (m = ln [number of events/number failures]) is significantly increased in motor synapses from AP-1 animals (Fig. 1 A–C). Similar results were obtained under nonfailure conditions where quantal content is calculated by m = EJP/mEJP (Fig. 1 D–F). Because quantal amplitude is not increased by AP-1 (Fig. S1), presynaptic mechanisms completely account for the measured synaptic strengthening. These observations, taken together with previous work (10), show that AP-1 increases quantal content of transmitter release at both low and physiological Ca²⁺ concentrations.

Fig. 1.

AP-1 (Fos and Jun) increases evoked transmitter release. (A) Distribution of synaptic responses to stimulation under low release probability conditions (0.3 mM Ca²⁺ HL3) from individual C155/+ Control (Left) and C155/+; Fos/+; Jun/+ AP-1 (Right) Drosophila larval synapses on muscle 6/7. Inset traces display sample recordings from each genotype. Top traces illustrate failure events. Lower traces illustrate both failures and synaptic responses. (B) Failure frequences. Filled bars, control: 0.35 ± 0.01 failures; open bars, AP-1: 0.14 failures ± 0.004; P < 0.0001; n = 10 animals per genotype. An event was counted as a failure if its amplitude measured below 2 standard deviations of the background noise in the recording. For each genotype, the average background noise amplitude ± standard deviation were: control: 0.1 ± 0.27; AP-1: 0.1 ± 0.23. For each genotype, variance was: control = 0.07; AP-1 = 0.05. (C) Failure analysis confirms a presynaptic locus for the increase in synaptic strength resulting from AP-1 induction. Quantal content, m = ln[failures/total events]: filled bars, m_Control = 1.05 ± 0.05; open bars, m_AP-1 (white) = 1.99 ± 0.12; P < 0.0001; n = 10 animals per genotype. (D) Representative EJPs from individual control and AP-1 larvae under high release probability conditions (0.5 mM Ca²⁺ HL3 saline). (E) EJP amplitudes are greater in AP-1 motor terminals: control: 3.78 ± 0.49 mV; AP-1: 8.80 ± 0.67 mV, n = 8; P = 0.001). Quantal size, however, was not significantly different (see Fig. S1). (F) Quantal content (m = EJP/mEJP), as estimated by a second independent method of measurement, is significantly greater in AP-1 motor terminals in high release probability conditions (0.5 mM Ca²⁺ HL3): m_C155 = 3.08 ± 0.67, n = 7 larvae; m_AP-1: 6.15 ± 0.31, n = 8 larvae; P = 0.001. A Martin correction factor was applied to both EJP amplitudes and quantal content to correct for nonlinear summation during current clamp recordings. For these calculations, the reversal potential for the Drosophila postsynaptic muscle was estimated to be approximately zero (refer to SI Methods).

AP-1 Causes an Increase in the Probability of Vesicle Fusion per Release Site.

Although AP-1 overexpression increases the number of presynaptic boutons (10), the average bouton size is significantly reduced (Table S1). For this reason, and because individual boutons contain multiple release sites, bouton number is not necessarily a reliable measure of synapse number. We used the following strategy to assess whether AP-1-terminals have more functional synapses, which we define as presynaptic release sites apposed to postsynaptic receptor clusters. In wild-type neuromuscular junctions (NMJ), ≈95% of GluR clusters are coupled to Bruchpilot (brp/CAST) immunopositive presynaptic puncta (11). This fraction is not altered by AP-1 expression (96% ± 1% control vs. 95% ± 1% in AP1). Thus, the number of Brp-positive puncta provides a measure of synapse number in AP-1 synapses.

As individual Brp spots are clearly resolved, we could count and analyze them with a spot-detection/analysis program (see SI Methods). This method yielded values that were in good agreement with those derived from previous serial EM studies of wild-type NMJs (see Table S1). Surprisingly, total Brp positive puncta (per NMJ) decreased by 21% in AP-1 synapses (Fig. 2 A–D). AP-1 induction did not detectably alter the distribution of T-Bar or synapse size assessed by quantitative fluoresence and electron microscopy respectively (Figs. S2 and S3). Thus, we conclude that although AP-1 increases total bouton number, the number of functional synapses is significantly reduced (see Fig. 2C and Table S1). Because the quantal content of neurotransmitter release is N × p (where N is synapse number and p is the average probability of vesicle release per synapse), our observations point to an increase in p at AP-1 terminals.

Fig. 2.

Synapse number does not account for Fos and Jun induced synaptic strengthening. (A) Individual release sites were quantified with automated spot analysis software Progenesis. Yellow outlines delineate boundaries of each identified spot. (Scale bar, 10 μm.) (B) Larval motor terminals immunostained for the active zone protein Bruchpilot (Brp) (green) and Drosophila glutamate receptor III (DGluRIII) (magenta). Single synaptic Type Is and Ib boutons contain on average 5 to 9 Brp puncta, respectively. This is consistent with previously published SSEM estimates of T-bar number of a wild-type NMJ (see Table S1). In both control and AP-1 synapses, all Brp puncta are found apposed to GluRIII clusters whereas only 95% of GluRIII clusters are found with Brp puncta. (C) Plot of release site number as measured by number of bruchpilot (Brp) immuno-reactive puncta, and varicosity number, as measured by synaptotagmin (Syt) immuno-reactive puncta. Each point represents Brp and Syt values from motor synapses (A2, M6/7) of individual larvae. (D) Total number of release sites is decreased by 21% in AP-1 synapses (control: 1,524 ± 29, n = 18 larvae; AP-1: 1,201 ± 40, n = 12 larvae; P < 0.0000005). The decrease may arise partly from homeostatic mechanisms that act to maintain transmitter release within a range appropriate for biological function.

Tetanus-Induced Potentiation but Not Paired-Pulse Facilitation Is Altered by AP-1 Overexpression.

If AP-1 overexpression leads to changes in the probability of release, we reasoned that forms of short-term plasticity, which also alter p, might be altered at these motor terminals. To test this idea, we measured 2 separable forms of short-term plasticity observed at the Drosophila larval NMJ at low Ca²⁺ concentrations. The first form, paired-pulse facilitation (PPF) is short-lived and decays within milliseconds. This is easily distinguished from longer-lived presynaptic plasticity, observed during and after tetanic stimulation, which decays more slowly (10s of seconds to minutes). Although multiple processes (e.g., augmentation and posttetanic potentiation) could contribute to this longer-lived form of plasticity, we refer to the phenomenon by a single term, tetanus-induced potentiation (TIP) (6, 12).

At interstimulus intervals (ISI) of 25 ms, 50 ms, 100 ms, and 1,000 ms, the paired pulse ratios exhibited by control and AP-1 motor terminals did not differ significantly (Fig. 3 A and B and Table S2) (recordings at 0.15 mM Ca²⁺ HL3.1 saline). The site of action for residual Ca²⁺ during paired pulse facilitation (PPF) has been demonstrated in previous studies to be located in the Ca²⁺ microdomain immediately surrounding clustered Ca²⁺ channels and vesicle release sites (6, 13). The observation that PPF is normal in AP-1 synapses suggests that Ca²⁺ dynamics in this microdomain are not significantly altered by AP-1.

Fig. 3.

Tetanus induced potentiation (TIP), but not paired-pulse facilitation (PPF) is lost after AP-1 induction. (A) Sample trace of a motor response to paired stimulus pulses delivered 100 ms apart. The amplitude of the response to the second pulse (P2) was calculated by subtracting the decay phase contribution of the postsynaptic response to the first pulse. (B) Graph showing average motor terminal responses to paired-pulse stimuli of different interstimulus intervals (ISI): 25 ms, 50 ms, 100 ms, 1,000 ms. Control and AP-1 animals exhibit similar levels of facilitation—as measured by the paired pulse ratio (PPR) at all ISIs. PPR = (P2 − P1)/P1 where P2 = amplitude of response to second pulse; P1 = amplitude of response to first pulse. See Table S1 for actual values. (C) TIP is occluded after Fos, Jun induction. Traces from individual control or AP-1 larval muscle recordings depicting facilitated motor responses to tetanic stimuation (10 Hz, 2 min). (D) Stimulation at 10 Hz for 2 min induces potentiation in control preparations. This is greatly diminished in AP-1 synapses. The potentiation factor immediately after tetanus (PF₀) is 2.53 ± 0.13 for control and 1.15 ± 0.10 for AP (P < 0.0001). Potentiation factors at the latest time point observed (PF_2.75 measured 2.75 min after stimulation cessation) = are 1.54 ± 0.14 for control and 0.93 ± 0.11 for AP-1 (P = 0.010). (E) Plot of EJP amplitudes during 10-Hz stimulation show that control motor terminals dramatically increase output during tetanic stimulation whereas AP-1 motor terminals do not. A Martin correction factor was applied to EJP amplitude data to correct for nonlinear summation. Graph shows average EJP amplitudes before, during, and after a 10-Hz TIP protocol: filled circles, control; open circles, AP-1.

In contrast, TIP was strikingly altered by AP-1 expression. In control synapses, transmitter release increases during a 2-min train of 10-Hz stimulation, eventually reaching a plateau. Contributions from both facilitation and TIP processes underlie the potentiated response during delivery of the tetanic stimulus train. Facilitation, however, decays within a few hundred milliseconds. Thus, longer-lived components (TIP), which decay on the order of seconds to minutes, can be isolated in the potentiated response after the tetanic train ends. TIP is greatly reduced in AP-1 terminals compared with the control (Fig. 3 C and D). The potentiation factor immediately after the tetanus (PF₀) is 2.53 ± 0.13 for control and 1.15 ± 0.10 for AP-1. This early potentiation decays with time but lasts for several minutes as evidenced by the values for PF_2.75 measured 2.75 min after stimulation cessation, which are 1.54 ± 0.14 for control and 0.93 ± 0.11 for AP-1. Thus, in AP-1 appears to affect both PF₀ (P < 0.0001) and PF_2.75 (P = 0.010).

The absence of TIP components in AP-1 synapses is consistent with a model where individual release sites are “prepotentiated” in AP-1 motor terminals (Fig. 3E). Loss of TIP cannot be explained by postsynaptic receptor saturation, because EJPs of twice this magnitude can easily be detected at this motor synapse. The observation that one form of short-term plasticity (PPF) remains unaltered, whereas longer lived forms (TIP) are dramatically diminished argues that AP-1 acts through a selective and relatively specific mechanism normally used for tetanus-induced presynaptic plasticity (6).

To determine the underlying mechanism of synaptic enhancement by AP-1, we measured 3 key parameters that influence the efficiency of neurotransmitter release: (i) presynaptic Ca²⁺ entry; (ii) sensitivity of the exocytotic machinery to Ca²⁺; and (iii) the available pool of synaptic vesicles.

No Increases in Ca²⁺ Entry or Shifts in Ca²⁺ Sensitivity Are Observed at AP-1 Synapses.

A simple mechanism for increasing the probability of exocytosis from an active zone is enhanced Ca²⁺ entry, e.g., because of a decreased presynaptic potassium conductance (14) and/or an increased Ca²⁺ current (15). The highly comparable paired-pulse ratios in AP-1 and control terminals suggest presynaptic Ca²⁺ entry and, particularly, the molecular target of residual Ca²⁺ during PPF, is unchanged in AP-1 expressing motor neurons.

Direct Ca²⁺ imaging to support the above argument is difficult, because small changes in single-action potential induced Ca²⁺ entry potentially can account for the observed increase in quantal content (16). Using an indirect approach, we instead asked whether summed Ca²⁺ entry during 40-Hz nerve stimulation was increased in AP-1 expressing animals (Fig. S4 A–D).

In motor terminals expressing the genetically encoded Ca²⁺ indicator, GCaMP 1.6 (17) we imaged fluorescence during sustained 40-Hz stimulation. Values for DF/F at a plateau reached in ≈2 seconds were similar in AP-1 and control synapses (Fig. S4 A–D). Unexpectedly, Ca²⁺ rise times in AP-1 terminals were slightly slower than in the control. This cannot be ascribed to faster Ca²⁺ extrusion as GCaMP signal does not decay any faster in AP-1 synapses after stimulation cessation (Fig. S4A). Instead, these data indicate that less Ca²⁺ enters AP-1 presynaptic terminals per action potential, at least during high-frequency stimulation. Although GCaMP imaging does not provide absolute measurement of presynaptic Ca²⁺ before and after stimulation, our data argue against increased evoked Ca²⁺ entry as being the primary mechanism for AP-1's effect on transmitter release.

Another mechanism to enhance transmitter release is to increase sensitivity of the exocytotic machinery to free Ca²⁺ (16, 18). Our measurements, however, show Ca²⁺ cooperativity of transmitter release was not significantly altered by AP-1 expression (Fig. S4E).

Synaptic Vesicles Shift from the Reserve Pool to the Actively Cycling Pool in AP-1 Motor Terminals.

The last major parameter that influences and often correlates with quantal content is the size of the active cycling vesicle pool (also referred to as exo-endo cycling pool, ECP) available for release (19). At Drosophila motor synapses, the ECP contributes to transmitter release at low to moderate rates of nerve stimulation, e.g., 3 Hz. A second “reserve” pool of vesicles (RP) poorly accessed at 3-Hz stimulation, is efficiently mobilized during high frequency stimulation >10 Hz (20). Two independent approaches, one electrophysiological and the other, optical allow the sizes of the cycling and total synaptic vesicle pools to be compared at the Drosophila NMJ (21).

ECP sizes were compared as follows. First, AP-1 and control synapses were stimulated continuously at 3 Hz in the presence of 1 μM bafilomycin A1, a drug that pharmacologically blocks the refilling of vesicles with neurotransmitter. Initial rates of synaptic depression under these experimental conditions largely reflect depletion of the cycling pool of vesicles. The later phase in the decay plot, after significant ECP depletion, represents vescles that arise from slow mixing between RP and ECP. Fig. 4A shows that the initial phase is extended in AP-1 compared with control, consistent with a larger ECP. To quantitatively estimate ECP size, we determined Y-intercept values (Fig. 4C) by linear regression of the points from the later slow phase of depression in a cumulative plot (20). These ECP estimates were consistent with substantial enlargement of the ECP in AP-1 motor terminals (Fig. 4D) (control: 19,644 ± 2,922 quanta, n = 8 larvae; AP1: 30,063 ± 3,511 quanta; n = 7 larvae; P = 0.0057). Because these estimates derive from fitting the observed curves to a specific (previously suggested) model (20), we used a second and completely independent technique to estimate the ECP. In this technique, we used optical measurements of styryl dye uptake into individual varicosities (21). Consistent with predictions from electrophysiological measurements, varicosities at AP-1 synapses were more brightly labeled than control synapses when the ECP was loaded with FM1–43 dye by 3-Hz stimulation for 7 min, indicating a larger ECP (Fig. 4E).

Fig. 4.

AP-1 induction expands the actively cycling vesicle pool at the expense of the reserve pool. (A) Time course of synaptic depression in motor terminals stimulated at 3 Hz in the presence of 1 μM bafilomycin to deplete the “exo/endo cycling pool” (ECP) of vesicles. Martin correction factor was applied to all quantal measurements to adjust for nonlinear summation in all bafilomycin experiments. (B) Time course of synaptic depression in motor terminals stimulated at 10 Hz in the presence of bafilomycin to rapidly deplete the ECP and reserve pool (RP) of vesicles. Expansion from stimulus numbers 0 to 4,200 of Fig. 4F shows that at 10 Hz, the kinetics of synaptic depression between control and AP-1 motor terminals are fundamentally different from the kinetics of their depressions at 3 Hz (A). (C) Cumulative quantal plot of Fig. 4A. Linear regression analysis was used to backextrapolate from points between stimulus numbers 3,000 and 4,200 for control and AP-1 motor terminals. ECP estimates were obtained from the y-intercepts. (D) ECP estimates obtained from linear regression analysis of cumulative plot in C. Black bar, control: 19,644 ± 2,922 quanta, n = 8 larvae; blue bar, AP-1: 30,064 ± 3,511 quanta, n = 7 larvae, P = 0.0057. (E) ECP estimates obtained from 3 Hz 7 min loading of synaptic varicosities with FM1–43 dye. Black bar, control: 215.6 ± 9.7 a.u., n = 40 boutons from 4 larvae; blue bar, AP-1: 277.2 ± 12.9 a.u., n = 56 boutons from 4 larvae; P = 0.00025. (F) Synaptic depression plot showing depletion of total vesicle pool using 10 Hz frequency stimulation with bafilomycin. (G) Bar graph of total pool estimates obtained by integrating quantal content over stimulus number in synaptic depression plots obtained from 10-Hz stimulation. Black bar, control: 68,281 ± 5,341 quanta, n = 9 larvae. Blue bar, AP-1: 59,349 ± 3,989 quanta, n = 8 larvae; P = 0.21, n.s. (H) Total pool estimates obtained estimates obtained from 30-Hz, 7-min loading of synaptic varicosities with FM1–43 dye. Black bar, control: 380 ± 19 a.u. n = 50 boutons, 4 larvae; blue bar, AP-1: 363 ± 13 a.u. n = 50 boutons, 4 larvae; P = 0.49, n.s.

To test whether this increased ECP in AP-1 synapses occurs at the expense of the reserve pool, we measured the total vesicle content in AP-1 and control terminals, by stimulating them to depletion at 10-Hz frequency in the presence of Bafilomycin. We estimated total vesicle pool size by integrating the complete depression curve of quantal content versus stimulus number (Fig. 4F). This direct electrophysiological estimate showed a slightly smaller total pool size in AP-1 terminals (Fig. 4 F and G) (C155: 68,281 ± 5,341, n = 9 animals; C155;Fos;Jun: 59,349 ± 3,989, n = 8 animals; P = 0.2). To independently assess the sizes of the total vesicle pool we measured FM1–43 uptake into presynaptic boutons after 7 min of 30-Hz stimulation, conditions that should label both ECP and RP. Remarkably, both control and AP-1 terminals were labeled to very similar levels under these conditions, with AP-1 showing slightly lower labeling (Fig. 4H). This indicates that the total number of synaptic vesicles is similar in control and AP-1 synapses. Thus, 2 independent approaches–electrophysiological and optical establish that AP-1 increases the actively cycling vesicle pool by partially mobilizing the reserve pool of synaptic vesicles. EM analyses of synaptic-vesicle density in AP-1 and control nerve terminals are also conistent with this conclusion (Fig. S5 and Table S3).

Link Between Vesicle Mobilization and Tetanus-Induced Potentiation.

Based on our observations, AP-1 synapses show 2 major differences from the wild-type. First, they have a larger fraction of actively cycling vesicles. Second, they exhibit highly reduced TIP. These 2 phenotypes can be linked if one proposes that mobilization from the reserve vesicle pool is required for TIP. In such a model, AP-1 synapses cannot be further potentiated because the RP has already been mobilized. We therefore asked whether tetanus-induced potentiation requires RP mobilization.

Previous work has established that RP mobilization depends on activity of the myosin light chain kinase (MLCK) in Drosophila motor terminals (22). Blocking the activity of this enzyme results in failure to recruit vesicles from the inactive pool under high frequency stimulation (22). Strikingly, the MLCK inhibitor ML-7 also inhibited tetanus-induced potentiation [potentiation factor PF₀: control: 1.70 ± 0. 10; ML-7: 1.2 ± 0.08; P < 0.004 (Fig. 5 A and B); PF was not examined at later time points because in relevant control preparations, the small amount of DMSO required to dissolve MLCK increased the rate of decay of TIP]. Taken together, the above experiments indicate that (i) TIP requires synaptic-vesicle mobilization from the reserve pool; and, by inference, (ii) AP-1 driven prepotentiation of transmitter release is accompanied by a stable expansion of the cycling pool of vesicles through reserve pool mobilization.

Fig. 5.

Reserve pool mobilization (RP) is required for tetanus-induced potentiation. (A) Traces showing sample responses from wild-type larval NMJ preparations before, during, and after a 10 Hz, 2 min TIP induction protocol. NMJs had been incubated for 30 min in either a control (0.1% DMSO) (Upper) or ML-7 (15 μM) (Lower), an inhibitor of myosin light chain kinase (MLCK). Pharmacological blockade of MLCK has been shown to block RP mobilization (25) at Drosophila NMJs. (B) Graph shows average EJP amplitudes before, during, and after a 10 Hz, 2 min after incubation of preps with either 0.1% DMSO (vehicle alone) or 15 μM ML-7 (and 0.1% DMSO). Filled circles show average EJPs from preparations incubated with the vehicle alone (DMSO). Open circles show average EJPs from synapses incubated with the ML-7.

Discussion

Long-Term Plasticity Without Increases in Synapse Number.

One important conclusion from this work is that Fos and Jun enhance synaptic strength, not by increasing synapse number, but rather by increasing the average probability of release from individual active zones. This conclusion is based on the following: (i) increases in synaptic strength in AP-1 motor terminals can be completely accounted for by increased transmitter release; and (ii) light microscopic studies show no increases in the number of release sites. Thus, there is an increase in the average probability of synaptic vesicle fusion at release sites of AP-1 motor neurons. There are some caveats to this argument. First, our definition of functional synapses as Brp-positive puncta is based on the assumptions that Brp puncta: (i) mark the large majority of release sites; and (ii) are mostly capable of transmitter release postsynaptic stimulation. These assumptions are supported by the tight colocalization of presynaptic Ca²⁺ channels and postsynaptic receptors with Brp puncta (23) (Table S1).

Fos and Jun Enhance Reserve Pool Mobilization.

Increased vesicle-release probability from active zones could conceivably be explained by several different mechanisms. In AP-1 synapses, we demonstrate a large increase in the size of the actively cycling synaptic vesicle pool (ECP), which arises at the expense of the reserve pool (RP). An increased ECP can account for the observed synaptic enhancement in AP-1 motor terminals if it increases the number of synaptic vesicles immediately available for fusion (19). RP mobilization has been associated with specific instances of short-term plasticity: e.g., cocaine-induced increases in dopamine release from rat striatal neurons (24). We here show that this process can be initiated by nuclear gene expression.

How widely might RP mobilization be deployed for synaptic change in vivo? Studies of the reserve pool in hippocampal synapses do not easily support the idea that vesicle trafficking from this source controls synaptic vesicle availability within these small axonal terminals (25, 26). However, RP mobilization can regulate the output from larger synapses such as neuromuscular junctions or the calyx of Held, where inhibition of the myosin light chain kinase required for vesicle mobilization has been shown to reduce the stability of the synaptic firing during repetitive stimulation (22). Because the actual mechanism of RP mobilization is poorly understood (25–28), more experiments will be required to understand exactly how Fos and Jun regulate this process. In one model, phosphorylation of synapsin, which tethers synaptic vesicles to an actin-based cytoskeleton in the central domain of synaptic boutons, may mobilize the reserve pool by triggering the dissociation of vesicles from the cytoskeleton, and their transport/diffusion to peripheral release sites (28, 29). In Drosophila NMJs, mitochondria within the presynaptic terminal are required for sustained release of neurotransmitter under high frequency stimulation (22, 30). One study suggests ATP production from mitochondria is required to fuel MLCK-activated, myosin-propelled transport of reserve vesicles from central to peripheral sites (22). Although such processes might be involved, it is also possible that different pathways are used for Fos and Jun mediated RP mobilization. For instance, smaller sized boutons, such as those observed in AP-1 animals (Table S1), may be less efficient at holding a central pool of reserve vesicles. In such a scenario, bouton geometry, rather than a specific regulatory protein, may prove to be the relevant target of AP-1 activity.

AP-1 Stably Recruits a Process Transiently Deployed for Posttetanic Potentiation.

The simplest interpretation of our observations is that stable reserve pool mobilization underlies the observed loss of tetanus-induced potentiation in AP-1 synapses. Other work at the Drosophila neuromuscular junction has associated mobilization of the reserve pool with the expression of TIP (12, 28).

Cytosolic Ca²⁺ accumulation and signaling is required for the induction of TIP (6). However, links between Ca²⁺ signaling and the expression of TIP are poorly defined; indeed, tetanus induced potentiation could include multiple Ca²⁺-dependent processes including augmentation and posttetanic potentiation/PTP (6). If TIP expression requires RP mobilization, then it would be occluded in 1 of 2 ways. Either: (i) by preexisting mobilization of the reserve pool; or (ii) by inhibition of reserve pool mobilization. Drosophila dnc mutants with enhanced cAMP signaling, and rut mutants with reduced cAMP signaling respectively illustrate these 2 different mechanisms of inhibition. Both mutants do not show tetanus-induced potentiation (12). Although dnc mutants show a greatly increased ECP and already enhanced transmitter release, rut mutants show a large, stable reserve pool that cannot be mobilized by tetanic stimulation (28). Our data indicate that AP-1 synapses behave like dnc mutants in which reserve vesicles have already been mobilized.

If reserve pool mobilization is required for TIP, then mutations or drugs that inhibit reserve pool mobilization would also be expected to block TIP. Consistent with this prediction, we find application of an MLCK inhibitor, which blocks reserve pool mobilization (22), dramatically inhibits TIP induction in wild-type motor synapses (Fig. 5 A and B). Thus, although we do not rule out alternative contributing mechanisms, we show that tetanus induced presynaptic potentiation is tightly linked to reserve pool mobilization.

Implications for Mechanisms of Long-Term Plasticity.

It is possible that many different direct and indirect targets of AP-1 contribute to various observed AP-1 dependent neuronal phenomena: e.g., increased bouton number, reduced bouton size, increased dendritic growth, elevated evoked transmitter release and increased ECP size (31). In addition, AP-1 may have effects on some phenomena that are not yet measured, e.g., kinetic and spatial features of synaptic Ca²⁺ dynamics. Nonetheless, our work shows functions of Fos and Jun in neurons, and provides substantial evidence for a model in which transcription-dependent changes in synaptic function occur through stable recruitment of mechanisms used in short-term plasticity. Recent observations that short-term forms of presynaptic plasticity are altered following synaptic enhancements induced by either BDNF or postsynaptic PSD-95 overexpression suggest that this could be a viable strategy for long-term information storage in central synapses (32). If long-term plasticity requires stable recruitment of short-term plasticity mechanisms, then the lability of long-term memory traces, as observed in studies of reconsolidation, may not require the elimination of stable synaptic connections representing the stored memory (5).

Methods

For detailed experimental procedures, see SI Methods.

Electrophysiology.

For physiological recordings, larvae were dissected in Ca²⁺ free haemolymph-like saline. Intracellular recordings from muscle 6 of abdominal segments 2 or 3 of wandering 3rd instar larvae were collected using sharp microelectrodes (50–75MΩ) pulled from borosilicate capillary tubes (OD 1.5 mm, ID 0.84 mm) filled with a 1:1 mixture of 3MKAc:3MKCl. Signals were amplified with Axoclamp 9.0, digitized in Digidata 1350, and recorded with pClamp 9.2. Recordings were done in either HL3 or HL3.1with indicated Ca²⁺ concentrations in text. Signals were collected in pClamp 9.2 and analyses performed in Clampfit 9.2. Recording were only used for analysis if the resting membrane potential was between −60 mV and −90 mV and muscle input resistance was < 5 MΩ. Under conditions of high frequency stimulation, recordings were discarded if Rin or Vm changed by >20%. Bafilomycin A1 (Sigma) and 1-(5-Iodonaph-thale0ne-1-sulfonyl)-1H-hexahydro-1,4-diazepine hydrochloride (ML-7) (Sigma) were both prepared as stock solutions and diluted in HL3 saline just before recording sessions to 1 μM and 15 μM, respectively. NMJ preparations were incubated for 15 min in 1 μM bafilomycin or 30 min in 15 μM ML-7 before recording sessions at room temperature. All statistical tests were done with a Student's t test unless otherwise indicated in the figure legend. For additional details on data analysis, refer to SI Methods.

Immunohistochemistry and Imaging.

Larval dissections and immunostaining were carried out as described in ref. 11 using the following antibodies: a-DSyt 1:100, a-Brp (1:50), a-DGluRIII (1:1,500). The release site counting, process was automated using Progenesis image analysis software. For additional details on release site quantification, refer to SI Methods.

FM1–43 Imaging.

Experiments were carried out as described in ref. 23. For additional details, see SI Methods.

Drosophila Strains and Genetic Controls.

The following strains were used in our analyses: control [Oregon R; D. Brower (Deceased, University of Arizona)]; UAS-Fos, UAS-Jun, transgenes [M. Bienz (MRC Laboratory of Molecular Biology, Cambridge, United Kingdom)]; neural Gal4 line, elav^C155 (C. Goodman); UAS-GCaMP 1.6 [D. Reiff (Max-Planck-Institute of Neurobiology, Martinsried, Germany)]. C155 control larvae are the female progeny of C155/Y males crossed to Oregon-R.