Postsynaptic regulation of synaptic plasticity by synaptotagmin 4 requires both C2 domains

Information transfer at neuronal synapses requires release of small, diffusible neurotransmitters after presynaptic vesicle fusion. Synaptic signaling can also occur in the reverse direction, with postsynaptic targets transmitting retrograde signals that alter the function or structure of presynaptic terminals (Tao and Poo, 2001). Retrograde signals can be released constitutively or in an activity-regulated fashion to modulate synaptic development and acute plasticity (Fitzsimonds and Poo, 1998). Ca²⁺-regulated vesicle trafficking also occurs in the postsynaptic compartment, suggesting that retrograde signaling may use vesicular trafficking mechanisms similar to those found presynaptically (Maletic-Savatic and Malinow, 1998; Adolfsen et al., 2004; Kennedy and Ehlers, 2006). One system in which retrograde signaling has been well characterized is at Drosophila melanogaster neuromuscular junctions (NMJs), where a strong correlation between neuronal activity levels and synaptic growth has been observed (Budnik et al., 1990; Sigrist et al., 2003; Guan et al., 2005). During the larval stages of development, an ∼100-fold increase in muscle surface area is accompanied by enhanced synaptic strength and growth of the innervating motor neuron. Retrograde release of molecules from the postsynaptic muscle controls both synaptic growth and homeostatic plasticity during larval development (Paradis et al., 2001; Marqués et al., 2002; McCabe et al., 2003; Goold and Davis, 2007). Although retrograde signals are required for plasticity and synaptic growth at Drosophila NMJs, the underlying mechanisms mediating retrograde release remain largely unknown.

Synaptotagmins are a conserved family of vesicular Ca²⁺ sensors that share a common structure with a single transmembrane domain and two cytoplasmic Ca²⁺-binding C2 domains (Adolfsen and Littleton, 2001). Only two of the seven Drosophila synaptotagmins, synaptotagmin 1 (Syt 1) and Syt 4, are present at most if not all synapses (Adolfsen et al., 2004). Unlike the synaptic vesicle–associated Syt 1, Syt 4 is expressed postsynaptically at the NMJ (Adolfsen et al., 2004), and mutants lacking Syt 4 show abnormal development and function of the embryonic NMJ (Yoshihara et al., 2005). In this study, we show that Syt 4 fusion-competent vesicles localize postsynaptically in Drosophila central nervous system (CNS) neurons. Syt 4 expression is bidirectionally regulated by neuronal activity and modulates synaptic growth and plasticity. Gain and loss of function experiments demonstrate that Ca²⁺-binding sites in both C2 domains are required for Syt 4 activity. In addition, Syt 4 modulates activity-dependent structural plasticity at NMJs, including seizure-induced and temperature-dependent enhancements of synaptic growth. These findings suggest a role for Syt 4 in regulating retrograde vesicular trafficking that underlies activity-dependent synaptic growth at Drosophila NMJs.

Based on studies of the synaptic vesicle isoform Syt 1, synaptotagmins are postulated to mediate the Ca²⁺ dependency of regulated exocytosis (Yoshihara et al., 2003; Chapman, 2008). Given their expression pattern in Drosophila, we hypothesize that Syt 1 and 4 form the core evolutionary conserved synaptotagmin family present at synapses, with Syt 4 mediating Ca²⁺-dependent trafficking of postsynaptic cargos, which is similar to the presynaptic role of Syt 1 in synaptic vesicle fusion. To analyze the emergence of the synaptotagmin family in relation to synapse evolution, we took advantage of recent genomic sequencing of some of the early branch points in metazoan evolution, beginning in single cell eukaryotes with the genomes of the amoeba Dictyostelium discoideum and the yeast Saccharomyces cerevisiae (Fig. 1 A). D. discoideum contains several single C2 domain proteins that do not show strong homology to known proteins in other species. In addition to PKC, the S. cerevisiae genome encodes several members of the extended Syt (E-Syt; also termed tricalbins) family, which encode N-terminal transmembrane proteins followed by three or more C2 domains that are predicted to function in membrane trafficking (Creutz et al., 2004; Min et al., 2007). The choanoflagellate genome of Monosiga brevicollis, thought to be the closest living single cell relative to animals (King et al., 2008), lacks synaptotagmins, though it does encode members of the PKC and E-Syt families. The multiple C2 domain and transmembrane region proteins, another evolutionarily conserved family with multiple C2 domain repeats (Shin et al., 2005), first appear in the choanoflagellates.

Given the lack of synaptotagmins at the single cell stage of evolution, we analyzed primitive metazoan branches before the evolutionary emergence of neurons and synapses. Trichoplax adhaerens is a placozoan consisting of a flat disc of epithelial-like cells surrounding a layer of multinucleate fiber cells with no nervous system. To our surprise, we identified synaptotagmin homologues in the T. adhaerens genome (Srivastava et al., 2008), including isoforms of Syt 1 and 7. Four additional tandem C2 domain proteins were also identified, although they showed weak homology to the bilateria synaptotagmins. Given the role of Syt 7 in plasma membrane repair via Ca²⁺-dependent fusion of secretory lysosomes (Reddy et al., 2001), it is not surprising that this family member emerged early in evolution of multicellular eukaryotes when cell–cell tension issues arose. However, the finding of a Syt 1 homologue before the evolution of the synapse was surprising. Indeed, we identified homologues of many presynaptic and synaptic vesicle proteins in T. adhaerens, suggesting that a cell–cell communication pathway ancestral to the synapse was present early in multicellular evolution. We did not identify a Syt 4 isoform in the T. adhaerens genome. The earliest evolution of a Syt 4–like protein coincides with the emergence of a primitive nervous system, as we identified a putative ancestral Syt 4 protein in the sea anemone Nematostella vectensis (Putnam et al., 2007). The characteristic D to S substitution in the Syt 4 C2A D3 residue does not appear until Drosophila (Fig. 1 B), suggesting continued evolution of Syt 4 function from its earliest appearance. We hypothesize that synaptotagmins evolved from diversification of function from an early PKC family member, as yeast PKC is implicated in the cell wall integrity signaling pathway required to remodel the cell surface during environmental stress and during growth and morphogenesis (Levin, 2005). Evolution of synaptotagmins from a Ca²⁺-dependent cell membrane remodeling pathway is an attractive origin for a protein family underlying Ca²⁺-dependent secretion. In summary, our genomic analysis indicates that Syt 1 and 7 are the founding members of the synaptotagmin family and arose before the origin of neurons, whereas Syt 4 emerged coincident with nervous system evolution.

To determine whether Syt 4 localizes to the postsynaptic compartment in the CNS as well as muscles, we constructed transgenic UAS lines expressing YFP fused to the intravesicular N terminus of Syt 4 or monomeric RFP (mRFP) fused to the cytoplasmic C terminus of Syt 4, allowing cell type–specific expression with defined GAL4 driver strains. Imaging of Syt 4–YFP expressed under control of the muscle-specific driver Myosin heavy chain (Mhc)–GAL4 in third instar larvae showed punctate expression at the postsynaptic compartment of NMJ synapses (Fig. 2 A). To analyze Syt 4 targeting in the CNS, we took advantage of the defined polarity of CNS motor neurons in the ventral nerve cord (VNC) of Drosophila larvae and the motor neuron–specific GAL4 driver D42 (Fig. 2 B; Parkes et al., 1998). Motor neuron axons terminate at muscle NMJs, whereas their dendrites are restricted to the VNC where they receive synaptic input. Syt 4–YFP driven by D42-GAL4 prominently localized to motor neuron cell bodies and dendrites in the VNC (Fig. 2 A) where synaptic inputs are known to occur (Littleton et al., 1993). In contrast, Syt 4–YFP targeted poorly to presynaptic terminals at NMJs.

To further characterize Syt 4 localization, we generated transgenic animals coexpressing Syt 4–mRFP and Syt 1–GFP with the D42-GAL4 driver, allowing us to compare targeting of a synaptic vesicle protein with Syt 4 in the same neurons. As shown in Fig. 2 B, Syt 1–GFP was not found in dendrites, but robustly targeted to presynaptic boutons, as expected for a synaptic vesicle protein. In contrast, Syt 4–mRFP showed low expression in presynaptic terminals but strong expression in the CNS dendritic neuropil. To determine whether Syt 4 vesicles were fusion competent in dendritic compartments of CNS neurons, we analyzed transgenic animals expressing a Syt 4 transgene tagged with a pH-sensitive GFP variant (ecliptic pHluorin) whose fluorescence increases when exposed to the extracellular space after fusion (Miesenböck et al., 1998). Expression of Syt 4–pHluorin in motor neurons resulted in intense fluorescence on dendritic membranes in the VNC, with no fluorescence observed at presynaptic NMJs (Fig. 2 C), indicating that Syt 4 vesicles only fuse within the dendritic compartment of motor neurons. We did not observe activity-dependent increases in Syt 4–pHlourin signals using either application of high K⁺ or electrical stimulation, which is likely the result of the small number of postsynaptic vesicles undergoing fusion. We conclude that fusion-competent Syt 4 vesicles target to postsynaptic compartments in CNS dendrites and muscle NMJs.

Given the potential role of Syt 4 in regulating Ca²⁺-dependent retrograde signaling, modulation of Syt 4 function or expression could alter the capacity for postsynaptic targets to transmit retrograde signals. Mammalian syt 4 was originally identified as an immediate early gene up-regulated by seizure induction and neuronal depolarization (Vician et al., 1995). To determine whether activity-dependent regulation of syt 4 is evolutionarily conserved, we tested whether the Drosophila homologue is also regulated by neuronal activity. Our first indication that activity levels modulate Syt 4 expression was the observation that Syt 4 immunoreactivity is reduced in syt 1^−/− larvae (Fig. 3, A and B), which have reduced synaptic activity (Yoshihara and Littleton, 2002). We next analyzed Syt 4 levels in Drosophila temperature-sensitive (TS) mutants that enhance or reduce neuronal activity (Ganetzky and Wu, 1986; Littleton et al., 1998). Syt 4 staining at larval NMJs of wild-type, para^ts1, and sei^ts1 strains demonstrated that Syt 4 protein expression correlates with the level of neuronal activity (Fig. 3, C–H). This regulation was not restricted to NMJs, as we observed similar results in the CNS by Western blot analysis of Syt 4 levels from adult Drosophila head extracts (Fig. 3, I and J). Quantification of Syt 4 protein demonstrated a statistically significant fivefold reduction in para^ts1 and syx^3-69 mutants and a statistically significant 46% increase in sei^ts1 mutants (Fig. 3 I).

Syt 4 expression could be regulated by a direct change in protein levels through altered protein stability or degradation or via activity-dependent transcriptional control of syt 4 RNA. We used RT-PCR to compare syt 4 mRNA levels in head extracts from TS excitability mutants raised at 25°C with those subjected to a 38°C heat shock for 20 min followed by a 30-min recovery (Fig. 3 K). Relative to wild-type, sei^ts1 showed a twofold increase in Syt 4 mRNA levels at the permissive 25°C, likely reflecting the excess chronic activity differences in the mutant. In addition, sei^ts1 flies given a 20-min heat shock to induce acute seizures displayed a 2.9-fold increase in Syt 4 mRNA expression compared with sei^ts1 25°C controls and a 12-fold increase compared with heat-shocked wild-type controls. para^ts1 animals displayed a 70% reduction in syt 4 mRNA levels, which is consistent with results from Western analysis. These findings indicate that Syt 4 mRNA and protein levels are bidirectionally modified by changes in neuronal activity, suggesting that Syt 4–dependent postsynaptic function may be regulated by the prior activity history of neurons.

Our previous data suggested that Syt 4 overexpression in postsynaptic muscles stimulates synaptic varicosity formation and growth (Yoshihara et al., 2005). To characterize the effect of altered Syt 4 levels, we analyzed the requirement for C2A and C2B function in the promotion of synaptic growth. Studies of Syt 1 have demonstrated that mutations disrupting the C2B Ca²⁺–binding domain have profound effects on release (Mackler et al., 2002), whereas C2A mutants do not (Robinson et al., 2002). Multiple negatively charged aspartic acid residues within loops emerging from the two C2 domains mediate Ca²⁺ binding by forming a bridge between negatively charged lipids embedded in the plasma membrane (Chapman, 2008). We neutralized one or more of the five key aspartic acid residues (previously referred to as D1–D5; Mackler et al., 2002) in each C2 domain to asparagines to test their requirement for Syt 4 function (Fig. 4 A). Multiple transgenic lines expressing each mutant protein under control of the UAS–GAL4 system were generated, and strains overexpressing similar levels of Syt 4 were selected for analysis (Fig. S1). Third instar larvae reared at 25°C were dissected, and bouton number was quantified at muscle 6/7 of segment A3 by α-complexin (α-Cpx) immunocytochemistry. We observed a 60% increase in bouton number when wild-type Syt 4 was overexpressed postsynaptically (Fig. 4 B). Mutating aspartic acids in both C2A (D292N [D5]) and C2B (D427N [D3] and D429N [D4]) Ca²⁺–binding domains eliminated the ability of Syt 4 to facilitate synaptic proliferation (Fig. 4, B and C). To determine whether the C2B domain plays a critical role in Syt 4 function, we generated transgenic lines that overexpress Syt 4 with D427N and D429N substitutions in C2B that are required for in vitro Ca²⁺ binding (Mackler et al., 2002). Postsynaptic overexpression of this transgene failed to promote synaptic growth compared with wild-type Syt 4, indicating that the C2B Ca²⁺–binding pocket is required for Syt 4 function, which is similar to its role in Syt 1.

In contrast to C2B, the C2A domain of Syt 4 contains a conserved substitution (D284S) in a key aspartic acid residue (D3) required for coordination of Ca²⁺ binding of Syt 4 C2A in vitro (Fernández-Chacón et al., 2002). Combined with data suggesting that C2B appears to the most critical Ca²⁺-binding domain in Syt 1, we hypothesized that Ca²⁺ binding by C2A would not be required for Syt 4 function in vivo. To directly test the relevance of the C2A Ca²⁺–binding site, we generated transgenic lines expressing an aspartic acid to asparagine mutation at D292 (D5 of C2A). Surprisingly, overexpression of D292N was less efficient than wild type at promoting synaptic growth (Fig. 4, B and C), although the transgene did retain some residual function compared with C2B mutants. These data suggest that C2A Ca²⁺ binding is likely important for Syt 4 activity in vivo and that serine 284 (S284), which is conserved in Syt 4 isoforms from Drosophila to humans (Fig. 1 B), might not simply be an inactivating mutation for Ca²⁺ binding. Given that no other amino acid substitution has been allowed at S284, it would be unusual if the sole purpose were to inactivate Ca²⁺ binding. One explanation for the conserved serine is that it represents a site of phosphorylation, potentially regenerating a negative charge in the C2A Ca²⁺–binding pocket normally formed by an aspartic acid in other synaptotagmin isoforms (Fig. 4 A). To test whether S284 is required for Syt 4 function in vivo, we generated transgenic animals expressing UAS–syt 4 S284A and expressed the construct postsynaptically under the control of Mhc-GAL4 to determine whether it was capable of promoting synaptic growth. If the function of S284 is to eliminate Ca²⁺ binding, an S284A change should not alter Syt 4's ability to enhance synaptic growth. In contrast to this prediction, the C2A S284A mutant completely eliminated Syt 4 function, suggesting that this residue is essential in vivo (Fig. 4, B and C). To determine whether phosphorylation of S284 might contribute to Ca²⁺ binding by C2A, we constructed a phosphomimetic serine to aspartic acid construct (S284D), which has been previously shown to restore Ca²⁺ binding of the Syt 4 C2A domain in vitro (von Poser et al., 1997). When this construct was overexpressed postsynaptically in the muscle, it promoted NMJ growth similar to wild-type Syt 4, which is in sharp contrast to the nonfunctional serine to alanine mutation (Fig. 4, B and C). Phosphorylation of S284 is an attractive mechanism for setting thresholds for retrograde plasticity, as less-active synapses might have Syt 4 in an unphosphorylated state that is less efficient at promoting retrograde signaling. Phosphorylation of S284 in Syt 4 in an activity-dependent fashion would be expected to enhance Ca²⁺ binding, potentially requiring less synaptic input to trigger future retrograde signaling. We conclude that Ca²⁺ binding by both C2A and C2B of Syt 4 is essential for its in vivo function and that the highly conserved serine at the D3 position of C2A is critical for the in vivo properties of the Syt 4 subfamily.

To further examine how Syt 4 regulates synaptic function, we compared synaptic growth, neurotransmitter release, and synaptic plasticity in response to high frequency stimulation in third instar larvae of syt 4 mutants and Mhc-GAL4, UAS–syt 4 overexpression lines. Immunocytochemistry with α-Cpx antisera revealed that syt 4–null mutants reduce synaptic growth at the NMJ with a 24% decrease in varicosity number compared with controls (Fig. 5 A). syt 4 mutants in trans to Df(3R)rn¹⁶, which removes the syt 4 locus, showed a 28% reduction in synaptic bouton number (Fig. 5 A). Expression of Syt 4 using the muscle-specific Mhc-GAL4 driver rescued the synaptic growth defect (Fig. 5 A). To determine whether C2A and C2B function were required for Syt 4–dependent synaptic growth, we expressed Ca²⁺-binding domain mutants in C2A (S284A) or C2B (D427N and D429N) in the null background and tested their ability to rescue the synaptic growth defect. Neither mutant transgene was able to restore synaptic growth to the levels observed with wild-type transgene expression, with C2A rescue mutants having a more severe defect than their C2B counterparts (Fig. 5 B). To analyze corresponding changes in the number of presynaptic release sizes (active zones), we calculated the active zone to bouton ratio at NMJs of controls, syt 4–null mutants, and Syt 4 overexpression lines by costaining with α-Cpx to identify the number of varicosities and with α-Bruchpilot (α-Brp; nc82) to label active zones. Quantification revealed that active zone number changed parallel to bouton number, resulting in a similar active zone to bouton ratio in syt 4 mutants and Mhc-GAL4; UAS–syt 4 overexpression lines compared with controls (Fig. 5, C–G).

We next examined physiological properties of synapses in syt 4 mutants and postsynaptic overexpression lines under voltage clamp conditions at Drosophila third instar larval NMJs. syt 4 mutants reduced spontaneous neurotransmitter release in the absence of nerve stimulation, whereas postsynaptic overexpression of Syt 4 enhanced spontaneous release (Fig. 6 A), which is consistent with the changes in varicosity number. Evoked responses to nerve stimulation in moderate extracellular Ca²⁺ (0.5 mM) showed a similar pattern, with the loss of Syt 4 reducing evoked neurotransmission, whereas postsynaptic overexpression of Syt 4 enhanced neurotransmitter release (Fig. 6 B). These changes in evoked responses were not observed in 2.0 mM extracellular Ca²⁺ (Fig. 6 C), suggesting that alterations in release probability or vesicle pool size may compensate for the altered number of release sites. To examine the basis for the response in higher Ca²⁺, we tested short-term depression to 10-Hz stimulation. Fast and slow components of synaptic depression have been described at Drosophila NMJs (Delgado et al., 2000; Wu et al., 2005). The fast component represents the immediately releasable pool, whereas the slow component is correlated with delivery of synaptic vesicles from the readily releasable pool. The fast component of release during the initial stimulation was unchanged in syt 4 mutants or overexpression lines, but distinct depletion profiles for the slow component of release were observed (Fig. 6 D). syt 4 mutants displayed a lower depletion of synaptic vesicles and maintained a higher baseline-evoked response during steady-state stimulation. In contrast, Syt 4 overexpression larvae showed more depression and a smaller steady-state response during the train. Previous experiments have shown that retrograde signaling mechanisms can trigger homeostatic plasticity at the NMJ to compensate for altered synaptic structure (Davis et al., 1996). Given the changes in synaptic vesicle pool size we observed, Syt 4 may not be essential for retrograde signaling that mediates homeostatic compensation. Alternatively, Syt 4–dependent retrograde signals may act to repress development of the readily releasable vesicle pool.

Previous studies at embryonic NMJs of syt 4 mutants demonstrated that loss of Syt 4 disrupts retrograde signaling required for enhanced spontaneous presynaptic release after high frequency stimulation (Yoshihara et al., 2005). Whether this pathway is unique to developing synapses or also functions at mature synapses was unknown. Likewise, it is unclear whether the enhanced synaptic growth caused by postsynaptic overexpression of Syt 4 is linked to alterations in short-term synaptic plasticity. To examine synaptic plasticity at mature NMJs of third instar animals, we tested the effects of high frequency stimulation (four 1-s 100-Hz stimuli spaced 2 s apart) under voltage clamp conditions. High frequency stimulation triggered a long-lasting enhancement of spontaneous release at mature third instar NMJs (Fig. S2), although the fold increase in mini frequency is smaller than observed at embryonic NMJs. There is no change in quantal size after stimulation (Fig. S2 C) but rather in the appearance of multiquantal events, indicating increased presynaptic release probability as opposed to postsynaptic effects. Spontaneous release after high frequency stimulation remained elevated for ∼5 min, indicating a long-lived effect on presynaptic release properties. In contrast to controls, syt 4–null mutants failed to show increased spontaneous release after stimulation (Fig. 7, A–D), suggesting that stimulation-dependent enhancement of presynaptic release requires Syt 4 function. Overexpression of Syt 4 in postsynaptic muscles triggered an elevation in the frequency of spontaneous fusion after high frequency stimulation compared with wild-type synapses (Fig. 7, A–D), with spontaneous release rates often >15 Hz. To determine whether C2A and C2B function were required, we expressed Ca²⁺-binding mutants in C2A (S284A) or C2B (D427N and D429N) in the null background and tested their ability to support stimulation-dependent enhancements in spontaneous release. Mutations in C2A or C2B eliminated the ability of Syt 4 to induce presynaptic release (Fig. 7 E), which is similar to their effects on synaptic growth (Fig. 4 B). We next tested whether the phosphomimetic serine to aspartic acid construct (S284D) could rescue release in syt 4 mutants. When expressed postsynaptically, Syt 4 S284D promoted enhanced spontaneous presynaptic release to an even greater extent than wild-type Syt 4 (Fig. 7 E). In summary, our findings indicate that the loss of Syt 4 disrupts short-term enhancement of neurotransmitter release and synaptic growth, whereas overexpression of Syt 4 in the postsynaptic compartment enhances presynaptic plasticity and subsequent synapse proliferation.

Enhanced synaptic growth in response to increased neuronal activity has been well characterized at Drosophila NMJs (Barber and Littleton, 2007) and can be triggered by seizure episodes (Guan et al., 2005), enhanced larval locomotion at elevated rearing temperatures (Sigrist et al., 2003), or by genetic mutations that increase neuronal excitability (Budnik et al., 1990). We hypothesized that increased neuronal activity could trigger Syt 4–dependent retrograde signaling secondary to Ca²⁺ influx post-synaptically during elevated motor neuron firing. To test this model, we quantified growth after enhanced neuronal activity in syt 4–null mutants and precise excision lines. We first examined structural plasticity at larval NMJs that occurs after increased larval locomotion stimulated by rearing at elevated temperatures (Sigrist et al., 2003). Although control larvae showed a 33% increase in bouton number when reared at 31°C compared with 25°C, animals lacking Syt 4 displayed no statistically significant increase in bouton number at 31°C (Fig. 8, A and B). Next, we examined synaptic overgrowth that occurs secondary to mutations that enhance neuronal activity. When sei^ts1 mutants lacking the Drosophila ERG potassium channel are subject to developmental heat shocks, increased NMJ activity causes expansion of the larval NMJ (Guan et al., 2005). To test the requirement for Syt 4, the sei^ts1 allele was moved into the syt 4^−/−, and precise excision background and larvae were reared at 25 and at 31°C (Fig. 8 C). sei^ts1 mutants in the precise excision background displayed a 60% increase in bouton number at 31°C compared with 25°C. In contrast, sei^ts1 in the syt 4^−/−–null background showed no significant change in bouton number between 25 and 31°C.

To determine whether neuronal activity is also required for the enhanced synaptic growth observed when Syt 4 is overexpressed postsynaptically, we examined the effects of reducing synaptic activity in Syt 4 overexpression strains. Previous studies have shown that mutations that reduce sodium channel function (para^ts1) suppress synaptic overgrowth induced by hyperexcitability mutants (Budnik et al., 1990) or elevated temperature rearing (Sigrist et al., 2003). We compared synaptic growth properties in Syt 4 overexpression strains in the para^ts1 genetic background at permissive and nonpermissive temperatures (Fig. 8 D). Similar to previous observations (Sigrist et al., 2003), para^ts1 mutants in the Mhc-GAL4 genetic background suppressed temperature-induced synaptic overgrowth. In addition, para^ts1 was able to dramatically reduce Syt 4–induced synaptic overgrowth when animals were reared at 31°C compared with 25°C. In summary, these findings indicate that synaptic proliferation triggered by postsynaptic Syt 4 overexpression at Drosophila NMJs occurs in an activity-dependent fashion.

Over the last decade, there has been a reversal in our appreciation of retrograde signaling at synapses, as experimental evidence has challenged the well-established view that synaptic information flows unidirectionally from the presynaptic terminal to the postsynaptic side. Retrograde signaling can be mediated by cell–cell contact through surface-attached adhesion complexes (Knöll and Drescher, 2002), membrane-permeable factors (Hawkins et al., 1994), and conventional neurotransmitters (Ludwig and Pittman, 2003; Regehr et al., 2009). Recent experiments have also established the presence of membrane trafficking components in the postsynaptic compartment, including ER, Golgi, and postsynaptic vesicles (Kennedy and Ehlers, 2006). In contrast to the well-described mechanisms for presynaptic vesicle trafficking, how retrograde signals are released from the postsynaptic compartment is poorly understood. In this study, we show that the Syt 1 homologue, Syt 4, plays a postsynaptic role in regulation of synaptic growth and plasticity in an activity-dependent manner at Drosophila synapses.

Our phenotypic analysis of syt 4–null mutants reveals a decrease in the number of NMJ synaptic varicosities, a reduction in neurotransmission, and defective-enhanced spontaneous release after high frequency stimulation. Our results at mature third instar synapses are qualitatively similar to those we observe at newly formed synapses (Yoshihara et al., 2005). However, although high frequency stimulation can induce an approximately two- to threefold increase in mini frequency at mature NMJs, similar stimulation of embryonic synapses results in an ∼100-fold change, indicating that newly formed connections are exquisitely sensitive to Syt 4–dependent retrograde signaling. In contrast to the loss of Syt 4 function, overexpression of Syt 4 in the postsynaptic compartment has the opposite effect, inducing bouton proliferation and enhancing spontaneous release after high frequency stimulation. Whether Syt 4 overexpression increases the likelihood of individual postsynaptic vesicle fusion events by altering the Ca²⁺-dependent release machinery or results in an increased number of postsynaptic vesicles is currently unknown. Similar increases in spontaneous release frequency have been observed after repetitive pulses of high K⁺ saline at the NMJ (Ataman et al., 2008). Although the link between increased mini frequency after strong stimulation and synaptic growth triggered by Syt 4 postsynaptic overexpression is unclear, it is tempting to speculate that increased Syt 4 levels enhance stimulation-dependent release of retrograde signals that act on presynaptic release and the synaptic growth machinery. We hypothesize that enhanced spontaneous release may function to prolong postsynaptic Ca²⁺ transients that normally occur after high frequency neuronal firing, enhancing generation of second messengers that promote synaptic proliferation. Syt 4–dependent synaptic growth requires neuronal activity, as the effect can be blocked by decreased action potential firing in para^ts1 mutants. In addition, Syt 4 contributes to activity-induced synaptic growth triggered by rearing animals at higher temperatures or by seizure-inducing mutations. By imaging tagged versions of Syt 4 in CNS motor neurons, we find that Syt 4 fusion-competent vesicles are also found in CNS dendrites, suggesting that the protein may play a general role in Ca²⁺ regulation of postsynaptic membrane traffic.

What is the role of Syt 4 in the postsynaptic compartment? We hypothesize that Syt 4 functions as a postsynaptic Ca²⁺ sensor for regulating retrograde signaling by activating postsynaptic vesicle fusion. This would result in release of a diffusible ligand from the postsynaptic compartment or insertion of a transmembrane protein that would trigger bidirectional signaling. Given the role of Syt 1 in presynaptic neurotransmitter release and the acute effects on synaptic plasticity we observed with manipulation of Syt 4 levels, we favor release of a diffusible signal. Alternatively, Syt 4 might be required for endocytosis of a transsynaptic signaling complex that would regulate short-term changes in bidirectional signaling. A role for Syt 4 solely in endocytosis is unlikely, as the protein is found on a vesicle compartment and not associated with the plasma membrane at rest (Adolfsen et al., 2004), where it presumably would be required for endocytotic reactions. In addition, interactions between Syt 1 and the clathrin adapter AP-2 do not require Ca²⁺ binding by Syt 1. We have demonstrated that Syt 4's ability to drive synaptic growth and enhance presynaptic release requires C2 domain Ca²⁺–binding residues. Finally, Syt 4 could have a unique postsynaptic function that is unrelated to vesicle trafficking. Given the established role of synaptotagmins in regulating membrane trafficking, we view this as a less likely option. The specific cargo contained in Syt 4 vesicles is currently unknown. Similar to the diverse set of cell type–specific neurotransmitters found in Syt 1–containing synaptic vesicles, we expect a similar diversity of Syt 4 postsynaptic vesicle cargos. Several target-derived retrograde factors function in synaptic growth, including FGF, Wingless/Wnt, TGF-β peptides, and the NGF family of neurotrophic factors (Hall et al., 2000; Umemori et al., 2004; Salinas, 2005; Zweifel et al., 2005). Recent work indicates that mammalian Syt 4 is localized to postsynaptic vesicles in hippocampal neurons that contain brain-derived neurotrophic factor where it regulates their release and subsequent effects on long-term potentiation (Dean et al., 2009), supporting an evolutionarily conserved role of Syt 4 as a regulator of postsynaptic fusion events.

An interesting question raised by our experiments is how similar the biochemical activities of Syt 1 and 4 are in vivo. Syt 1 and 4 cannot functionally substitute for each other at synapses (Adolfsen et al., 2004; Yoshihara et al., 2005), but this reflects sorting to distinct vesicle populations. Syt 1 preferentially requires C2B Ca²⁺ binding for activity, with C2A playing a more minor role. Syt 4–dependent retrograde signaling at NMJs shows Ca²⁺ dependence in vivo (Fig. 4; Yoshihara et al., 2005), although in vitro methods have produced conflicting evidence about its potential to bind Ca²⁺ (Fukuda et al., 1996; Dai et al., 2004). By examining the ability of mutant Syt 4 proteins to promote synaptic growth and to rescue growth and plasticity defects in syt 4 mutants, we demonstrate that Ca²⁺-binding sites in both the C2A and C2B domains of Syt 4 are necessary. In particular, S284 of the C2A domain is required for Syt 4 activity in vivo. This was an unexpected finding, as the serine replaces an essential aspartic acid residue required for C2 domain Ca²⁺ binding in other synaptotagmin isoforms. This aspartic acid to serine substitution is conserved from Drosophila to humans, suggesting an evolutionarily important role for the residue, previously hypothesized to be inactivation of the C2A Ca²⁺–binding site. Our finding that S284 is required for Syt 4 function in vivo raises the possibility that the serine may be phosphorylated, reintroducing a negative charge into the Ca²⁺-binding pocket and allowing it to functionally substitute for the aspartic acid present in Syt 1's C2A Ca²⁺–binding pocket. Indeed, substitution of an aspartic acid at S284 restores Ca²⁺ binding by Syt 4 in vitro (von Poser et al., 1997) and can promote synaptic growth when overexpressed in vivo (Fig. 4). Phosphorylation of Syt 4 could potentially explain the conflicting in vivo and in vitro data concerning the protein's ability to bind Ca²⁺. There is precedent for phosphorylation of synaptotagmins regulating Ca²⁺ binding, as Syt 2 is phosphorylated by WNK1, reducing its Ca²⁺-dependent, phospholipid-binding affinity (Lee et al., 2004). Current studies are under way to determine whether S284 of Syt 4 is phosphorylated in vivo and which kinases are involved.

Two orthologues of Drosophila Syt 4 are found in mammalian genomes encoded by the Syt 4 and Syt 11 genes. Although mammalian Syt 4 has been implicated in retrograde signaling (Dean et al., 2009), little is known about Syt 11 other than it is abundantly expressed in the brain (von Poser et al., 1997). However, Ca²⁺-dependent exocytosis of postsynaptic vesicles in mammalian neurons has been documented (Maletic-Savatic and Malinow, 1998; Li et al., 2005). Ca²⁺-dependent fusion of somatodendritic vesicles containing the VMAT2 monoamine transporter has also been demonstrated using immunocytochemistry and amperometric recording of dopamine release (Li et al., 2005). Surprisingly, the kinetics of somatodendritic dopamine release is similar to quantal events recorded at presynaptic sites, with release occurring in <1 ms. Our observation that Syt 4 mRNA and protein levels are modulated by neuronal activity, similar to findings in mammals (Vician et al., 1995) and birds (Poopatanapong et al., 2006), suggests that not only has Syt 4's primary sequence been conserved but also its transcriptional modulation by neuronal activity. Given the profound effects we observe on synaptic growth when postsynaptic Syt 4 levels are altered, it is exciting to speculate that activity-dependent modulation of Syt 4 and Syt 11 may be a general mechanism for regulating synaptic growth and plasticity at activated synapses.

Drosophila were cultured on standard media at the indicated temperatures. The syt 4–null mutant was generated through imprecise P-element excision of insert EY09259, 100-bp 5′ of the syt 4 transcription start site (Adolfsen et al., 2004). UAS–syt 4 mutant lines were constructed in the syt 4 cDNA by site-directed mutagenesis of pUAST–syt 4 and injected into w¹¹¹⁸ flies at the Duke Model Systems Transgenic Facility. YFP–syt 4 cDNA fusions were generated by joining cDNA-encoding YFP to the 5′ beginning of the syt 4–coding sequence in pUAST and generating transgenic animals by injection into w¹¹¹⁸ flies. Syt 4–mRFP lines were generated in a similar fashion by fusing the mRFP cDNA to the 3′ end of the coding region of syt 4 and deleting the stop codon and were injected into w¹¹¹⁸ flies by Genetic Services (Cambridge, MA). All constructs were verified by DNA sequencing before injection.

The following proteins were used in the analysis shown in Fig. 1: Homo sapiens Syt 1 (NCBI Protein database accession no. NP_005630), Mus musculus Syt 1 (NCBI Protein database accession no. NP_033332), Drosophila Syt 1 (NCBI Protein database accession no. NP_523460), Caenorhabditis elegans snt-1 (NCBI Protein database accession no. NP_495394), H. sapiens Syt 4 (NCBI Protein database accession no. NP_065834), M. musculus Syt 4 (NCBI Protein database accession no. NP_033334), Drosophila Syt 4 (NCBI Protein database accession no. NP_477464), C. elegans snt-4 (NCBI Protein database accession no. NP_491853), and N. vectensis putative Syt 4 (NCBI Protein database accession no. XP_001633294). We used standard parameters in the field for identifying a synaptotagmin: a single transmembrane-containing protein lacking a signal peptide with tandem cytoplasmic C2 domains. For distinguishing Syt 1 and 4 family members, we relied on overall blast similarity to previously characterized family members and the presence of the aspartate to serine substitution at the C2A D3 residue found in known members of the Syt 4 family.

Immunostaining was performed on wandering third instar larvae reared on standard media at 25 or 31°C as indicated in Figs. 1–8,^2345678. Larvae were dissected in HL3.1 saline (70 mM NaCl, 5 mM KCl, 10 mM NaHCO₃, 4 mM MgCl₂, 5 mM trehalose, 115 mM sucrose, and 5 mM Hepes, pH 7.2). Dissected larvae were fixed with 4% formaldehyde in HL3.1 for 45 min and washed three times for 5 min in PBS with 0.1% Triton X-100 (Huntwork and Littleton, 2007). Preparations were blocked for 1 h at room temperature with gentle agitation in 1% BSA and 0.1% Triton X-100 in PBS before primary antibodies were added. The following primary antibody dilutions were used: α-Cpx (Huntwork and Littleton, 2007) for bouton counting (1:1,000), nc82 α-Brp (Developmental Studies Hybridoma Bank) for active zone counting (1:50), and α–syt 4 (1:250; Adolfsen et al., 2004). Larvae were incubated with primary antibodies overnight at 4°C and washed three times for 5 min in PBS with 0.1% Triton X-100. The secondary antibodies, Cy2-conjugated goat α–rabbit and Cy5-conjugated goat α–mouse (Jackson ImmunoResearch Laboratories), were used at 1:250 dilutions and incubated for 4 h at room temperature. Before mounting, the dissections were washed three times in PBS with 0.1% Triton X-100 and three times in PBS.

Samples were mounted in 70% glycerol and imaged with a 40× NA 1.3 Plan Neofluar oil immersion lens (Carl Zeiss, Inc.) at room temperature. Images were captured with a laser-scanning confocal microscope (Pascal; Carl Zeiss, Inc.) using the accompanying software. Bouton and active zone number were quantified from confocal stacks taken of the entire NMJ at muscle 6/7 of segment A3 (Huntwork and Littleton, 2007). Published images were not subject to additional image processing. All error measurements indicate SEM. Statistical significance was calculated with single-factor analysis of variance (ANOVA), with Student's t tests used for secondary analysis. Statistical significance is noted as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

α–Syt 4 antibody (Adolfsen et al., 2004) was used at 1:750, and α-Dlg antibody 4F3 (Developmental Studies Hybridoma Bank) was used at 1:2,000. Secondary antibodies were used at 1:10,000, including goat α–rabbit IgG Alexa Fluor 680 (Invitrogen) and goat α–mouse IR dye 800 (Rockland). Western blots were visualized on an infrared imaging system (Odyssey; Li-COR Biosciences), and protein expression was quantified with the accompanying software.

RT-PCR was performed on male Drosophila head RNA preparations after a single acute 20-min heat shock at 38°C followed by a 30-min recovery as described previously in Guan et al. (2005). In brief, RNA was isolated from heads of males aged 3–4 d after eclosion. All flies were sacrificed between 2 and 4 p.m. to reduce any circadian-dependent transcriptional changes. After seizure induction and recovery, Drosophila were frozen in liquid nitrogen and vortexed to separate heads from bodies. From purified total RNA from heads, mRNA was isolated using an mRNA extraction kit (Oligotex; QIAGEN). cDNA was created with T7-(dT)₂₄ primer (Genset Corp.) using a cDNA kit (Invitrogen) and purified by Phase Lock Gel extraction (Eppendorf). Primers used for RT-PCR quantification include: GAPDH1 primers (5′ end) 5′-AATCAAGGCTAAGGTCGAGGAG-3′ and (3′ end) 5′-TAACCGAACTCGTTGTCGTACC-3′; and syt 4 primers (5′ end) 5′-CATTGATTCGGAGAGG-3′ and (3′ end) 5′-CTTATTGCGTTCGGTTTGAG-3′.

Evoked postsynaptic currents were recorded from ventral longitudinal muscle 6 at abdominal segment A3 in third instar larvae using two-microelectrode voltage clamp (OC725; Warmer Instruments) at −80 mV holding potential (Acharya et al., 2006). All experiments were performed in modified HL3 solution consisting of 70 mM NaCl, 5 mM KCl, 4 mM MgCl₂, 10 mM NaHCO₃, 115 mM sucrose, and 5 mM Hepes-Na, pH 7.2. Extracellular Ca²⁺ concentration was adjusted to final values as indicated in Figs. 1–8,^2345678. For stimulation, nerves were cut close to the ventral ganglion and sucked into a pipette filled with working solution. The nerve was stimulated at frequencies indicated in each experiment using a programmable stimulator (Master-8; AMPI). Data acquisition and analysis were performed using pClamp software (Axon Instruments).

Fig. S1 shows expression of Syt 4 in transgenic strains. Fig. S2 shows analysis of posttetanic induction of spontaneous release.

We thank Shuai Chen for help with data analysis and Bill Adolfsen, Kathy Galle, and Robin Stevens for help with generation and localization of transgenic fluorescently labeled proteins. The nc82 antibody against Brp developed by Erich Buchner and the 4F3 antibody against Dlg developed by Corey Goodman were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the National Institute of Child Health and Human Development and were maintained by the Department of Biological Sciences at the University of Iowa (Iowa City, IA).

This work was supported by the National Institutes of Health (grant NS40296) to J.T. Littleton.