α-Synuclein (α-Syn) is a small protein that, together with the closely related β- (β-Syn) and γ-Synucleins (γ-Syn), constitutes one of the most abundant proteins in the brain1,2,3,4. α-Syn plays a central role in Parkinson’s Disease (PD) pathogenesis since α-Syn mutations and multiplications cause PD5,6,7,8,9, genome-wide association studies link α-Syn to sporadic forms of PD10,11, and the brains of PD patients invariably contain Lewy bodies composed of α-Syn aggregates12. However, the physiological function of α-Syn, and that of other synucleins, remains largely unknown.
Synucleins possess a conserved N-terminal domain that binds to phospholipids13,14,15, underlying α-Syn’s affinity for membranes such as synaptic vesicles16,17. Overexpression of α-Syn in vitro and in vivo inhibits exocytosis, possibly through impairments in synaptic vesicle endocytosis, recycling, and dilation of the exocytotic fusion pore17,18,19,20. By contrast, deletion of α-Syn produces little to no effect on synaptic transmission, with α-Syn-KO mice exhibiting only slight reductions in dopamine (DA) levels and displaying modest behavioral phenotypes21,22. Moreover, synuclein double and triple knockout mice displayed no detectable changes in synaptic strength or short-term plasticity23,24. Thus, it has been difficult to reconcile α-Syn’s abundance and highly penetrant role in PD with its seemingly subtle endogenous function. Strikingly, even modest transgenic α-Syn overexpression completely prevents the lethality and neurodegeneration of CSPα KO mice25, suggesting an essential role for α-Syn in protection against neurodegeneration, which is counterintuitive given its causal involvement in PD.
The striatum, the input nucleus of the basal ganglia, is one of the most severely affected areas in PD, as the loss of DA signaling in the striatum and the degeneration of synapses on striatal spiny projection neurons (SPNs) greatly alter the striatal circuitry and underlie many of the motor and cognitive impairments observed in PD26,27,28. One particularly detrimental consequence of PD is the loss of endocannabinoid- (eCB-) dependent plasticity at corticostriatal synapses29,30,31,32, which is central to striatum-dependent learning and habit formation33,34. In eCB-dependent plasticity, eCBs are synthesized and released postsynaptically in an activity- and Ca2+-dependent manner. eCBs then retrogradely bind to presynaptic CB1 receptors (CB1Rs) to decrease the presynaptic release probability35,36,37,38. However, little is known about how eCBs are released from postsynaptic neurons. eCBs are amphiphilic molecules derived from phospholipids that are unlikely to diffuse passively out of the postsynaptic neurons and across the synaptic cleft39,40. Thus, how eCBs reach presynaptic CB1Rs during synaptic plasticity, an essential step to understanding striatal function, is unclear.
Normal basal synaptic transmission in Syn-tKO mice
Given the strong association of corticostriatal dysfunction with PD, we directly measured basal corticostriatal synaptic transmission and eCB-dependent plasticity in α/β/γ-synuclein triple KO (Syn-tKO) mice. Previous reports suggested that α-Syn decreases neurotransmitter release by acting at presynaptic sites, with some studies showing increased synaptic transmission in single α-Syn KO mice21,41, whereas no such changes were detected in double23 or triple synuclein KO mice24. We therefore investigated if corticostriatal synaptic transmission was abnormal in Syn-tKO mice. Whole-cell patch clamp recordings from SPNs in acute slices of the dorsolateral striatum prepared from wildtype (WT) and Syn-tKO mice, combined with electrical stimulation of corticostriatal axons, allowed us to measure corticostriatal synaptic responses (Fig. 1a). We found no significant difference in the stimulus-response relationship between WT and Syn-tKO corticostriatal synapses (Fig. 1b). Because previous reports have shown that survival and behavioral deficits are revealed at older ages in Syn-tKO mice24,42, we also tested aged mice (16-18 months old). Again we observed no significant difference in synaptic strength between WT and Syn-tKO mice (Extended Data Fig. 1a,b).
We next measured the use-dependent dynamics of synaptic transmission by delivering stimulus trains at varying frequencies (Fig. 1c). We measured the rate of synaptic depression resulting from repeated stimulation23 and found virtually indistinguishable depression dynamics (Fig. 1d) and steady-state response amplitudes (Fig. 1e) between WT and Syn-tKO cells across stimulation frequencies. Together, these results show that basal corticostriatal synaptic transmission in Syn-tKO mice is largely normal, including responses engaged by repeated stimuli that depend on the rates of presynaptic vesicle recycling and the sizes of the reserve vesicle pool.
Syn-tKO mice lack eCB-dependent plasticity
One of the best-characterized forms of corticostriatal synaptic plasticity is eCB-LTD43,44,45 that is required for striatal learning33,34. Importantly, impairments in corticostriatal eCB-LTD are observed in mouse models of PD32,46,47. We assayed eCB-LTD in acute slices of young-adult (3 months old) WT and Syn-tKO mice by combining slight membrane depolarization (-50 mV) with an application of a type I mGluR agonist ((S)-3,5-dihydroxyphenylglycine (DHPG; 50 µM); Fig. 1f), which results in a lasting depression of evoked corticostriatal excitatory postsynaptic currents (EPSCs) (Fig. 1g). Strikingly, we found that eCB-LTD is abolished in Syn-tKO mice (Fig. 1h). Syn-tKO cells were indistinguishable from WT cells in the presence of the CB1R antagonist, AM251 (10 µM) (Fig. 1g,i). Importantly, paired-pulse ratios (PPRs) were significantly increased in WT cells following eCB-LTD, but not in Syn-tKO cells (Fig. 1j), consistent with a selective decrease in presynaptic release probability in WT cells. We observed impaired eCB-LTD in both young-adult and aged mice (16-18 months old; Extended Data Fig. 1c-f), suggesting that the phenotype is not an age-dependent effect, but instead due to a direct loss of an endogenous synuclein function. Furthermore, we found that eCB-LTD was normally expressed in KO mice lacking α-Syn alone or both β- and γ-synuclein (βγ-Syn-KO mice), suggesting redundancy among synucleins (Extended Data Fig. 2a-d).
In order to further characterize the Syn-tKO phenotype, we measured depolarization-induced suppression of inhibition (DSI)38, a different form of eCB-dependent plasticity in the striatum. During DSI, strong depolarization of SPNs results in the Ca2+-dependent synthesis and release of eCBs that transiently suppress inhibitory inputs (Fig. 1k)35,36,37,48. Indeed, a 5-second depolarization (to 0 mV) in WT cells was sufficient to transiently inhibit spontaneous inhibitory postsynaptic currents (sIPSCs) in a CB1R-dependent manner (Fig. 1l,n). Strikingly, the same DSI protocol failed to elicit a significant reduction in sIPSCs in Syn-tKO mice (Fig. 1m,o). We observed the same results when we repeated this experiment using a stimulation-evoked IPSC protocol (Extended Data Fig. 3a-d), with WT but not Syn-tKO cells showing a significant increase in PPRs during DSI (Extended Data Fig. 3e), which reflects the presynaptic locus of the transient suppression of inhibitory inputs.
Finally, in a parallel set of experiments, we recorded DSI in pyramidal neurons of the hippocampal CA1 region (Fig. 1p)37. Here we once again found that DSI was readily inducible in WT cells, but not in Syn-tKO cells (Fig. 1q-t; Extended Data Fig. 3f-h). The observations that Syn-tKO mice exhibit impairments in two forms of eCB plasticity (eCB-LTD and DSI), across different synapse types (glutamatergic and GABAergic), and brain regions (striatum and hippocampus) suggest a broad defect in eCB signaling in Syn-tKO mice.
Presynaptic CB1Rs are intact in Syn-tKO mice
α-Syn is thought to function predominantly in the presynaptic terminal, suggesting that the impairment in eCB-dependent synaptic plasticity in Syn-tKO mice is likely due to a failure of CB1R signaling49. To test this hypothesis, we applied the CB1R agonist WIN55,212 (WIN; 2 µM) in acute brain slices. WIN strongly depressed corticostriatal transmission via direct activation of presynaptic CB1Rs, bypassing the postsynaptic eCB synthesis and release mechanisms engaged during eCB-LTD and DSI (Fig. 2a). We found that WIN strongly reduced evoked EPSCs in both WT and Syn-tKO mice (Fig. 2b,c). The magnitude of synaptic depression was indistinguishable between genotypes (Fig. 2d), as was the concomitant significant increase in PPRs (Fig. 2e) that would be expected for a presynaptic weakening via CB1R activation. These results were reproduced when repeated in aged mice (Extended Data Fig. 4a-d). Thus, presynaptic CB1R function is intact in Syn-tKO mice, suggesting a postsynaptic deficit upstream of CB1R activation.
Release of eCBs is impaired in Syn-tKO mice
Given the defects in eCB plasticity across different experimental contexts, we next tested whether a more upstream step in eCB signaling was impaired in Syn-tKO mice, namely the postsynaptic release of eCBs as retrograde signals. Postsynaptic release of eCBs precedes CB1R activation but is downstream of eCB synthesis40,50. Although the specific mechanisms of retrograde eCB release are not well understood50,51, the direct introduction of eCBs into a postsynaptic neuron via a patch pipette has been shown to induce a progressive release of these eCBs, resulting in synaptic depression52,53. Thus, in order to directly test eCB release, we dialyzed SPNs intracellularly with the endogenous eCB anandamide (AEA; 50 µM) through the patch-pipette (Fig. 2f). In WT cells, the intracellular application of AEA caused a progressive depression of evoked corticostriatal EPSCs that depended on CB1R function (Fig. 2g). Strikingly, in Syn-tKO cells postsynaptic AEA loading had no effect (Fig. 2h,i). Correspondingly, we observed significant PPR increases in WT cells, but not in Syn-tKO cells (Fig. 2j). Because intracellular loading with AEA bypasses the eCB synthesis pathways, these results suggest that the defect in Syn-tKO mice lies specifically in the release of eCBs from postsynaptic cells.
To directly visualize eCB release, we utilized a recently developed eCB fluorescent sensor (GRABeCB2.0)54. Viral expression of the GRABeCB2.0 sensor in the dorsal striatum of mice allowed us to image stimulation-induced release of eCBs in acute slices (Fig. 2k). Local electrical stimulation in WT slices resulted in a significant increase in GRABeCB2.0 signal, reflecting the release of eCBs (Fig. 2l). However, evoked GRABeCB2.0 signals were significantly reduced in Syn-tKO mice (Fig. 2m,n), consistent with a deficit in eCB release. Importantly, we validated GRABeCB2.0 sensor expression and function in all imaged slices. Bath application of AEA (10 µM) significantly increased GRABeCB2.0 fluorescence in both WT and Syn-tKO mice, and AM251 (10 µM) decreased GRABeCB2.0 fluorescence and blocked stimulation-induced GRABeCB2.0 activity in WT slices (Fig. 2l-o). Thus, in combination with our electrophysiology data, these results suggest that normal eCB release requires synucleins.
eCB plasticity requires postsynaptic α-Syn
Thus far, our results suggest that synucleins are required for the postsynaptic release of eCBs. To further test this conclusion, we sparsely infected SPNs in the dorsolateral striatum of Syn-tKO mice with adeno-associated viruses (AAVs) that co-express GFP and α-Syn (Fig. 3a, top). Recordings of corticostriatal eCB-LTD from GFP+ or GFP- cells allowed us to directly test whether postsynaptic exogenous α-Syn can rescue the Syn-tKO phenotype (Fig. 3a, bottom). As expected, eCB-LTD was not observed in GFP- cells (Fig. 3b). Remarkably, almost all GFP+ cells expressing α-Syn exhibited significant eCB-LTD (10 out of 11) (Fig. 3c,e). The presence or absence of α-Syn in recorded cells was confirmed by immunocytochemistry (Fig. 3b,c, top). Moreover, viral expression of C-terminally truncated α-Syn (residues 1-95) also rescued eCB-LTD in Syn-tKO cells (Fig. 3d,e). The rescued eCB-LTD in GFP+ cells was accompanied by a significant increase in PPRs, which was not observed in uninfected GFP- cells (Fig. 3f). Finally, postsynaptic rescue of α-Syn also restored striatal DSI in Syn-tKO cells (Fig. 3h,i). Together, these results demonstrate that not only are synucleins required for eCB plasticity, but also that the role they play is a postsynaptic one.
Membrane-binding domains of α-Syn are required for eCB plasticity
In order to dissect the mechanism of synuclein function in postsynaptic eCB release, we sparsely expressed α-Syn in the striatum of Syn-tKO mice as before, but included mutations in the α-Syn rescue sequence to determine which regions (and therefore functions) of α-Syn are required for eCB-dependent plasticity. Although we previously observed that C-terminal truncation of α-Syn had no effect on eCB-LTD (Fig. 3d), we asked if C-terminal serine 129, a site previously implicated in Ca2+-binding affinity and regulating PD neurodegeneration55, could modulate eCB-LTD. However, we found that phosphorylation at serine 129 was not relevant for α-Syn’s function within eCB-LTD, as neither alanine (S129A, phosphorylation-deficient) nor aspartate substitutions (S129D, phosphorylation-mimic)56 affected the viral rescue of eCB-LTD in Syn-tKO mice (Extended Data Fig. 5).
Our results thus indicate that the N-terminal domain of α-Syn is required for eCB release. The major biochemical activity of α-Syn consists of phospholipid membrane binding that is mediated by its N-terminal domain13,15,57. To test whether membrane binding by α-Syn is required for eCB-LTD, we virally expressed α-Syn mutants carrying A11P and V70P (A11P/V70P) substitutions that ablate membrane binding by α-Syn but do not impair its synaptic localization15. Remarkably, A11P/V70P- mutant α-Syn failed to rescue eCB-LTD in Syn-tKO mice (Fig. 4a-c), suggesting that membrane binding of α-Syn is required for eCB-LTD. To strengthen this hypothesis, we repeated these experiments in cells infected with A30P mutant α-Syn, a PD mutation that also decreases lipid binding by α-Syn15. A30P-mutant α-Syn also did not rescue the loss of eCB-LTD in Syn-tKO mice (Fig. 3d,e). Correspondingly, PPRs were increased in cells expressing WT α-Syn but not in cells expressing A11P/V70P- or A30P-mutant α-Syn (Fig. 3f). Together, these results demonstrate that in postsynaptic neurons, α-Syn enables eCB-LTD by binding to phospholipid membranes, likely by mediating the postsynaptic release of eCBs.
Postsynaptic SNAREs are required for eCB release
α-Syn has been shown to act as a SNARE chaperone that facilitates SNARE complex assembly during vesicular exocytosis by binding to phospholipid membranes4,24,58. SNARE proteins not only mediate presynaptic vesicle exocytosis but are also essential for postsynaptic exocytosis of AMPA receptors and other proteins59,60,61. Thus, the fact that eCB release requires postsynaptic α-Syn that is competent to bind to phospholipid membranes suggests that eCBs are released by synuclein-dependent exocytosis. To investigate this possibility, we tested if postsynaptic SNAREs are involved in eCB-dependent plasticity and eCB release.
We sparsely infected SPNs in the dorsolateral striatum of WT mice with lentiviruses that co-express GFP and tetanus-toxin light chain (TeNT), which inactivates synaptobrevin-2, a SNARE protein involved in most forms of exocytosis. We confirmed that postsynaptic TeNT expression did not disrupt basal synaptic properties of infected SPNs, as previously shown for hippocampal neurons62,63 (Extended Data Fig. 6). Next, we measured eCB-dependent plasticity, comparing GFP+ (TeNT-expressing) cells to adjacent GFP- controls. Strikingly, TeNT significantly impaired eCB-LTD (Fig. 5a-c) and blocked DSI (Fig. 5d-f), an effect that was not revealed in previous studies using acute neurotoxin dialysis through the patch-clamp recording pipette37. Together, the impaired eCB-LTD and DSI results mirror the Syn-tKO phenotypes and suggest that postsynaptic SNAREs are also required for eCB-dependent plasticity. Lastly, to further explore the specificity of the effect of TeNT in impairing the release of eCBs, we performed the AEA-loading experiment as before. AEA-loading of GFP+ cells expressing TeNT failed to induce progressive synaptic depression, whereas loading of GFP- control cells robustly suppressed synaptic transmission (Fig. 5g-i). Thus, in addition to synucleins, SNAREs are required postsynaptically for the active release of eCBs (Fig. 5j).
Summary
Here we show that eCBs are released by postsynaptic vesicular exocytosis, in a process that requires synucleins. Thus, we report an unexpected convergence of two puzzling questions in neuroscience, namely the questions of the function of synucleins and of the mechanism of eCB release. We show that mice lacking all three synuclein isoforms have apparently normal basal synaptic properties but exhibit significant defects in multiple forms of eCB-dependent plasticity spanning different time frames (eCB-LTD and DSI), synapse types (glutamatergic and GABAergic), and brain regions (striatum and hippocampus). Using direct measurements of eCB release, we demonstrate that synuclein-deficient neurons suffer from a loss of eCB release, but retain normal CB1R function. Strikingly, bypassing the Ca2+-dependent eCB synthesis processes via postsynaptic loading of neurons with AEA, an endogenous eCB, revealed that the export of AEA from the postsynaptic cell is impaired by the synuclein deletion. Mechanistically, we identify the N-terminal membrane-binding domain of α-Syn, as well as postsynaptic synaptobrevin SNAREs, as required for eCB release. Together, these results point towards vesicular exocytosis as the process underlying eCB transmission.
Our results are surprising given that α-Syn is known to function presynaptically, and do not preclude additional presynaptic roles for α-Syn. Our viral α-Syn rescue experiments take advantage of the corticostriatal circuit’s compartmentalization of pre- and postsynaptic cells to demonstrate that the postsynaptic expression of α-Syn is sufficient to restore eCB-dependent plasticity in Syn-tKO mice. Indeed, synuclein-dependent release of eCBs adds to our growing understanding of the roles played by postsynaptic SNAREs during synaptic plasticity60,61. Furthermore, our results provide a potential link between eCB signaling and PD. In particular, our finding that A30P PD mutant α-Syn is unable to rescue eCB-LTD suggests that eCB release and eCB-dependent plasticity may be aberrant in PD, potentially contributing to the cognitive deficits observed in PD pathology. Additionally, our results are unexpected since an early report suggested that SNAREs are not involved in eCB release37. However, the previous study relied on acute Botulinum toxin light-chain dialysis, which may be temporally insufficient to fully access SNAREs and block eCB-release. In our study, we achieved neurotoxin expression (e.g., Lenti-GFP-TeNT) that enables TeNT action for multiple days prior to experiments. Together, our results demonstrate a novel postsynaptic function of endogenous synucleins in regulating eCB release and synaptic plasticity, and reveal that eCBs are released postsynaptically via synuclein-dependent vesicular exocytosis.