Neurotrophic factors and target-specific retrograde signaling interactions define the specificity of classical and neuropeptide cotransmitter release at identified Lymnaea synapses

Many neurons concurrently and/or differentially release multiple neurotransmitter substances to selectively modulate the activity of distinct postsynaptic targets within a network. However, the molecular mechanisms that produce synaptic heterogeneity by regulating the cotransmitter release characteristics of individual presynaptic terminals remain poorly defined. In particular, we know little about the regulation of neuropeptide corelease, despite the fact that they mediate synaptic transmission, plasticity and neuromodulation. Here, we report that an identified Lymnaea neuron selectively releases its classical small molecule and peptide neurotransmitters, acetylcholine and FMRFamide-derived neuropeptides, to differentially influence the activity of distinct postsynaptic targets that coordinate cardiorespiratory behaviour. Using a combination of electrophysiological, molecular, and pharmacological approaches, we found that neuropeptide cotransmitter release was regulated by cross-talk between extrinsic neurotrophic factor signaling and target-specific retrograde arachidonic acid signaling, which converged on modulation of glycogen synthase kinase 3. In this context, we identified a novel role for the Lymnaea synaptophysin homologue as a specific and synapse-delimited inhibitory regulator of peptide neurotransmitter release. This study is among the first to define the cellular and molecular mechanisms underlying the differential release of cotransmitter substances from individual presynaptic terminals, which allow for context-dependent tuning and plasticity of the synaptic networks underlying patterned motor behaviour.

distinct cotransmitter composition and release characteristics, and that this can be dynamically regulated in order to meet the spatiotemporal functional requirements of each synapse [5][6][7][8][9][10][11][12][13] . However, our understanding of the cellular and molecular mechanisms that allow a neuron to regulate the use of its cotransmitters at distinct synaptic sites is considerably lacking.
The concurrent use of classical small molecule and peptide neurotransmitters is common throughout the nervous system and across evolution. Neuropeptides are important modulatory signals that regulate numerous aspects of synaptic development, plasticity and network function, as well as animal physiology and behaviour 3,14 .
Although peptide neurotransmitters are traditionally viewed as slow-acting neuromodulators released from perisynaptic sites in response to Ca 2+ diffusion during high-frequency bursting 15 , their release has been demonstrated to occur over a wide range of firing frequencies [16][17][18] . Peptidergic large dense-core vesicles (LDCVs) have also been shown to cluster directly at presynaptic active zones and be released by single action potentials 19,20 . Furthermore, several studies have indicated that postsynaptic target identity influences the selective use of classical and peptide neurotransmitters at the different presynaptic terminals of a given neuron 6,19,21,22 . Taken together, these observations suggest that the peptidergic characteristics of cotransmitting terminals can be highly variable, although the molecular mechanisms that give rise to such differences are yet to be defined.
In this study we sought to identify the cellular and molecular mechanisms underlying presynaptic specificity, with a focus on two unresolved questions: (i) how does a neuron establish functionally distinct presynapses with specific postsynaptic targets, and (ii) how does a presynaptic neuron generate the appropriate patterns of cotransmitter release that are required for the coordinated actions of neurons in behaviourally-relevant networks? To address these outstanding questions, we used cardiorespiratory neurons from the central nervous system (CNS) of the invertebrate mollusc Lymnaea stagnalis to study how the selective use of cotransmitters at synapses with distinct postsynaptic targets influences the assembly and function of the neuronal circuits involved in cardiorespiratory regulation. Reductionist approaches using the simple nervous systems of invertebrate models have enabled fundamental insights into cotransmission because their large identified neurons with known transmitter phenotypes participate in behaviourally-defined synaptic networks that can be studied directly in situ or reconstructed and manipulated in vitro 23,24 . Moreover, invertebrate models are especially valuable for functional studies on peptidergic transmission because they are one of the few systems in which electrophysiologically measurable synaptic responses can be directly attributed to the actions of identified peptide neurotransmitters at individual synapses.
Here, we found that presynaptic neuropeptide release competency was regulated in a target-and contextdependent manner. This involved an interplay between extrinsic neurotrophic factors (NTF) and synapse-specific retrograde arachidonic acid (AA) signaling interactions, which converged on glycogen synthase kinase 3 (GSK-3) activity. In this context, we identified a surprising role for the Lymnaea synaptophysin (Syp) homologue in the inhibitory regulation of peptide neurotransmitter release. These findings on the regulation of cotransmitter use at individual synapses uncover a previously undefined mechanism for presynaptic cotransmitter specificity in synaptic networks, with implications for the appropriate expression and plasticity of patterned motor behaviours.

Synaptic transmission via classical and peptide cotransmitters is target-specific.
To investigate the mechanisms that differentiate cotransmitter use at individual synapses, and the consequences of this for specifying network function and behaviour, we first sought to monitor synaptic transmission by classical and peptide transmitters at functionally-defined synapses. The Lymnaea cardiorespiratory interneuron visceral dorsal 4 (VD4) was of interest for this study as it is well known to use the classical small molecule neurotransmitter acetylcholine (ACh) alongside a mixture of neuropeptides derived from the heptapeptide transcript of the FMRFamide gene (primarily G/SDPFLRFamide; discussed in subsequent text as FMRF neuropeptides) [25][26][27][28] . VD4 is integrated into a three-neuron network via reciprocal inhibitory synapses with the FMRFamidergic input 3 interneuron (IP3I) and the giant dopaminergic neuron right pedal dorsal 1 (RPeD1). These neurons and their reciprocal synapses establish Lymnaea's respiratory central pattern generator (rCPG) network in vivo, are indispensable for the expression of respiratory behaviour, and recapitulate appropriate synaptic networks and rhythmic activity when reconstructed in vitro 23,29 . VD4 is also known to form one-way excitatory synapses with a number of other identified neurons that act to coordinate cardiorespiratory behaviour, including the FMRFamidergic visceral F group cells (VF) and the serotonergic neuron left pedal dorsal 1 (LPeD1) 26,[30][31][32][33] (Fig. 1A-C). In Lymnaea neurons, both ACh and FMRF neuropeptides can elicit synaptic excitation or inhibition. For ACh, this is due to NTF-regulated expression of cationic or anionic nicotinic ACh receptors (AChR) [34][35][36] . For FMRF neuropeptides, this is due to cell type-specific coupling of FMRF G protein-coupled receptors (GPCRs) to distinct G protein complexes and metabotropic signaling cascades 37,38 . FMRF neuropeptides have also been shown to occlude transmitter release from target neurons by inhibitory regulation of Ca 2+ influx and neurosecretory mechanisms 39 . We therefore wondered whether the release of ACh and FMRF neuropeptides might be selectively tuned at VD4′s presynaptic terminals to establish the target-specific formation of one-way or reciprocal, and inhibitory or excitatory synapses that are required for the appropriate expression and modulation of cardiorespiratory behaviour.
To investigate this question, we first performed simultaneous intracellular recordings from the isolated intact CNS to determine the classical vs peptidergic nature of synaptic transmission at endogenous synapses between VD4 and the identified cardiorespiratory neurons VF and RPeD1. Synaptic transmission was assessed in response to a ~ 4 s burst induced in VD4, first under control conditions to assay cotransmission, and then in the presence of an AChR antagonist cocktail (see "Methods" section) to isolate the peptidergic component of synaptic transmission. We found that excitatory transmission with VF was mediated primarily by FMRF neuropeptides, as the mean number of action potentials induced in VF in response to VD4 stimulation was unchanged after the addition of AChR antagonists. Inhibitory transmission with RPeD1, however, was mediated primarily by ACh, as the duration of RPeD1 inhibition in response to VD4 stimulation was reduced with AChR antagonists (Fig. 1D-F; Table S1). This suggests that synaptic transmission with distinct postsynaptic targets is not equivalently mediated by VD4′s classical and peptide cotransmitters (Fig. 1G).
Target-specific regulation of cotransmission is modulated by extrinsic NTF signaling. Because the identified cardiorespiratory neurons and their synaptic networks cannot be easily resolved or manipulated in the intact brain preparation, we turned to in vitro reconstruction of these synapses to uncover the regulatory mechanisms that specify transmission by classical and peptide neurotransmitters. To this end, we isolated the identified neurons from the CNS and paired VD4 in a soma-soma configuration with RPeD1 or VF in culture. As in the intact preparation, we made simultaneous intracellular recordings from cultured neurons, first under control conditions to study cotransmission, and then in the presence of AChR antagonists to isolate peptidergic transmission.
When cultured in CNS-conditioned media (CM), a NTF-rich media that recapitulates the in vivo environment, VD4-VF exhibited mixed cholinergic-peptidergic cotransmission, whereas most VD4-RPeD1 pairs exhibited primarily cholinergic transmission ( Fig. 2A,C,D; Table S2A-B). As NTFs are well known to modulate synaptic function 40 , we next cultured pairs in NTF-deficient defined media (DM) to ask whether regulated cotransmission is solely target dependent, or whether it might also be influenced by extrinsic factors. Surprisingly, we observed a reversal of peptidergic phenotype, where most VD4-VF pairs exhibited primarily cholinergic transmission, and most VD4-RPeD1 pairs exhibited mixed cotransmission (Fig. 2B,C,D; Table S2A Table S2B), while analysis of the postsynaptic ACh biphasic response indicates that excitatory, but not inhibitory, AChR expression is regulated by NTFs ( Fig. 2F; Table S2B), in line with previous findings 27,35,[41][42][43] . Taken together, these observations suggest that cholinergic and peptidergic transmission are tuned by distinct mechanisms.
The next question was whether the specificity of classical and peptide cotransmission would be observed if VD4 innervated multiple postsynaptic targets in vitro. We therefore cultured VD4 simultaneously with VF and Characterization of cotransmission and isolated FMRF transmission by simultaneous intracellular current clamp recordings in soma-soma paired neurons (image inserts; scale bar, 50 μm), performed in CM or DM control (Ai,iii and Bi,iii), or with AChR antagonists (Aii,iv and Bii,iv; 5 μM MLA, 10 μM TC, 20 μM TEA) to isolate peptidergic transmission. Inserts show ACh-PSPs (i,iii; − 100 mV holding potential), inhibited by AChR antagonists (ii,iv). Mixed cholinergic-peptidergic cotransmission is observed at VD4-VF synapses in CM, with biphasic cholinergic and excitatory peptidergic responses (Ai,ii; e ACh , i ACh , e FMRF ), and at VD4-RPeD1 synapses in DM, with biphasic cholinergic and inhibitory peptidergic responses (Biii,iv; e ACh , i ACh , i FMRF ). Primarily cholinergic transmission is observed at VD4-RPeD1 synapses in CM (Aiii,iv; e ACh , i ACh , absence of i FMRF ), and VD4-VF synapses in DM (Bi,ii; i ACh , absence of e FMRF ), demonstrating that peptide cotransmitter specificity is determined by postsynaptic target identity and extrinsic NTF signaling. (C). Summary data, mean peptidergic excitation of VF (number of action potentials induced) is unchanged by AChR antagonists, and influenced by NTFs. N ≥ 9. **P ≤ 0.002 (Independent Samples Kruskal-Wallis test). (D). Summary data, mean inhibition of RPeD1 (duration in s, reflects ACh + FMRF cotransmission or isolated FMRF transmission) is reduced by AChR antagonists in CM but not DM, indicating differential peptidergic inhibition influenced by NTFs. N ≥ 10. *P ≤ 0.010; ***P < 0.001 (Independent Samples Kruskal-Wallis test). (E). Summary data, mean amplitudes of ACh-PSPs are unchanged in CM or DM control. N ≥ 9. P = 0.490 (Independent Samples Kruskal-Wallis test). www.nature.com/scientificreports/ RPeD1 in a triple-soma configuration in CM, and found that the VD4-VF synapse exhibited mixed cotransmission, while the VD4-RPeD1 synapse exhibited primarily cholinergic transmission (Fig. 3A,C,D;  Table S3B). This parallels the limited ACh transmission observed at VD4-VF synapses in situ (see Fig. 1 and Fig. S3), and suggests that ACh release in the context of synaptic competition is selectively modulated. As FMRF neuropeptides have been shown to inhibit transmitter release 39 , we hypothesized that the degree of FMRF neuropeptide transmission would be inversely correlated with the degree of dopamine (DA) transmission at the reciprocal inhibitory synapse between VD4-RPeD1. Therefore, in parallel with monitoring ACh and FMRF neuropeptide transmission from VD4, we also evaluated DA transmission from RPeD1 in the above experiments. Most RPeD1-VD4 synapses exhibited strong inhibitory DA transmission when cultured in CM, whereas in DM most did not ( Fig. 4A; Table S4). When we analyzed the duration of RPeD1 peptidergic inhibition vs the duration of VD4 dopaminergic inhibition, we found a significant negative correlation ( Fig. 4B; Pearson correlation, R = − 0.534, P = 0.001; R 2 = 0.2849). This suggests that the context-and target-dependent regulation of FMRF neuropeptide transmission observed above may reflect differences in VD4′s FMRF neuropeptide release propensity, such that high-releasing synapses would occlude DA release from RPeD1 and low releasing synapses would facilitate DA release. If so, this NTF-dependent regulation of neuropeptide release holds important implications for the assembly, plasticity, and function of the Lymnaea rCPG network, and how the rhythmic network activity required for breathing behaviour is established and modulated 23,29 . Presynaptic inhibition of neuropeptide release defines cotransmitter specificity. Considering the above observations, we postulated three scenarios to account for the apparent specificity of VD4′s FMRF neuropeptide release: (i) the translational expression of FMRF neuropeptides changes; (ii) FMRF neuropeptides are selectively targeted to specific presynaptic sites; or (iii) FMRF neuropeptides are selectively released from specific presynaptic sites. To explore these possibilities, we cultured VD4 neurons alone or with VF and/ or RPeD1 postsynaptic targets in CM or DM, using an axon-axon configuration to facilitate visualization of synaptic sites. Using immunocytochemistry (ICC) and an antibody against the RF-NH 2 moiety common to FMRF neuropeptides, we found a uniform intensity of somatic neuropeptide immunolabel amongst the culture conditions, suggesting that the translational expression of FMRF neuropeptides does not change (Fig. 5A,B; Table S5A). In paired neurons, conditions that exhibited limited peptidergic transmission (VD4-RPeD1 CM and VD4-VF DM) were characterized by a 'hyper-innervation' phenotype in which the intensity of the synaptic neuropeptide immunolabel was increased relative to the conditions that exhibited strong peptidergic transmission (VD4-VF CM and VD4-RPeD1 DM) ( Fig. 5A,C,D; Table S5B). These results reveal that FMRF neuropeptides accumulate at presynaptic terminals with low release propensity, and that an inhibitory presynaptic mechanism regulates the target-and context-dependent specificity of peptidergic transmission.
A Lymnaea synaptophysin homologue selectively inhibits neuropeptide release. We next sought to identify the molecular switch responsible for the selective facilitation and inhibition of peptidergic LDCV release. We focused on two synaptic vesicle protein (SVP)-dependent mechanisms that have been reported to selectively inhibit transmitter release. The first was the synaptotagmin (Syt) family of Ca 2+ sensor SVPs, considering that different isoforms selectively facilitate or inhibit small synaptic vesicle (SSV) and LDCV release characteristics [44][45][46] . This has also been reported in the invertebrate literature [47][48][49] , indicating that the use of alternative Syt isoforms as selective regulators of transmitter release is an evolutionarily conserved strategy. The second was the vesicle integral membrane protein synaptophysin (Syp), which reversibly complexes with the v-SNARE synaptobrevin (Syb) when phosphorylated, rendering Syb unable to enter the fusogenic SNARE complex [50][51][52] .
To determine the influence of the Lymnaea homologues of Syt and Syp on cotransmitter release, we first identified and cloned them from a Lymnaea CNS cRNA library (see Fig. S6A-C), then microinjected synthetic mRNA into VD4 to evaluate the effects of their overexpression on classical and peptidergic transmission. VD4-VF pairs were cultured in CM and VD4-RPeD1 pairs were cultured in DM to facilitate formation of synapses with high neuropeptide release propensity. In both cases, overexpression of Syp reduced FMRF neuropeptide transmission relative to H 2 O vehicle control (Fig. 6A,B,D,E; Table S6A), but had no effect on the amplitude of ACh-mediated PSPs (Fig. 6C,F; Table S6B), whereas Syt isoforms affected ACh, but not FMRF neuropeptide transmission (see Fig. S6D). Using ICC in axon-axon pairs, we found that Syp overexpression increased the intensity of the somatic and synaptic FMRF neuropeptide immunolabel relative to H 2 O vehicle control (Fig. 6G-I; Table S6C), while the Syt isoforms had no effect (see Fig. S6E). These data suggest that Syp acts as an inhibitory regulator of LDCVs, which tunes synaptic cotransmission by selectively attenuating neuropeptide release.
Next, if Syp mediates target-specific inhibition of FMRF neuropeptide release, we reasoned it would sort to synapses in a manner resembling the distribution patterns observed for FMRF neuropeptides. To characterize Syp localization, we generated a C-terminal mCherry tagged construct. VD4 was microinjected with synthetic Syp-mCherry mRNA and paired in a dual synapse axon-axon configuration with VF and RPeD1 in CM. We performed ICC using an α-mCherry antibody, and found that Syp synaptic fluorescence intensity was higher at synapses with RPeD1 than with VF ( Fig. 6J-L; Table S6D), mimicking the selective accumulation of FMRF neuropeptides observed at low release propensity synapses (see Fig. 5). www.nature.com/scientificreports/ Taken together, these observations support a Syp-dependent mechanism for the inhibitory regulation of  www.nature.com/scientificreports/ neuropeptide release, and suggest that presynaptic peptide cotransmitter specificity requires synapse-specific populations of release-competent and -incompetent LDCVs.
NTF and target-specific regulation of GSK-3 activity determine neuropeptide release competency. We next sought to identify the molecular signaling cascades that define FMRF neuropeptide release  To test this experimentally, we cultured VD4-VF and VD4-RPeD1 pairs in CM or DM, in the presence of a GSK-3 inhibitor (5 μM SB 216763) or vehicle control (0.1% DMSO). When cultured in CM, VD4-VF peptidergic transmission was not impacted by GSK-3 inhibition, while in DM, GSK-3 inhibition induced peptidergic transmission (Fig. 7A,B; Table S7A). This suggests that neuropeptide release characteristics at VD4-VF synapses are defined by NTF-dependent regulation of GSK-3 activity. In VD4-RPeD1 pairs, however, we found that GSK-3 inhibitors reversed neuropeptide release characteristics in both CM and DM (Fig. 7C,D; Table S7B). This suggests that GSK-3 activity at VD4-RPeD1 synapses is not regulated by the conventional NTF/RTK pathway, and implicates a target-dependent molecular signaling interaction that acts to define neuropeptide release competency.
Target-specific retrograde arachidonic acid signaling defines neuropeptide cotransmitter release. The release of FMRF neuropeptides from VD4 produces excitation in VF and inhibition in RPeD1, and the metabotropic receptors are coupled to distinct G proteins in these two postsynaptic targets (see Fig. S2B). As GPCRs activate a range of metabolic signaling cascades that act locally and trans-synaptically, we reasoned that cotransmitter specificity was elicited by differences in FMRF transmitter-receptor interactions and cell-cell signaling. Retrograde signaling by membrane-permeable second messengers is well known to influence synaptic function 56 . In particular, the generation of arachidonic acid (AA) metabolites has been linked to the actions of FMRF neuropeptides 57,58 , and was previously implicated in the function of reciprocal VD4-RPeD1 synapses 59 . Since we observed an atypical GSK-3 signature at VD4-RPeD1 synapses, we wondered whether target-specific retrograde AA might influence neuropeptide release. Therefore, VD4-VF and VD4-RPeD1 pairs were cultured in DM, to avoid NTF/RTK signaling, and supplemented with free arachidonic acid (AA, 5 μM), ETYA (10 μM, an inhibitor of AA metabolism), or vehicle control (0.1% DMSO + 0.1% Tocrisolve 100). Relative to control, AA induced peptidergic transmission at VD4-VF pairs, while the inhibitor ETYA had no effect (Fig. 8A,B; Table S8A), indicating that AA metabolism is not endogenous to FMRF neuropeptide signaling between VD4-  Table S8B), indicating that FMRF neuropeptide signaling between VD4-RPeD1 induces AA metabolism. Since AA has been shown to activate protein kinase C (PKC) family kinases 60 , which have been reported to influence GSK-3 activation 61,62 , we suspected that target-specific retrograde AA signaling influenced peptidergic transmission via PKC. VD4-VF and VD4-RPeD1 pairs were cultured in DM in the presence of the inhibitor Chelerythrine chloride (Ch Cl − ; 1 μM) to determine whether PKC is a molecular target for AA. Relative to vehicle control (0.1% DMSO), PKC inhibition prevented peptidergic transmission at VD4-RPeD1 synapses (Fig. 9A,B; Table S9A). We found no effect of PKC inhibition on peptidergic transmission at VD4-VF synapses; however, it negated the induction of peptide release via AA (Fig. 9C,D; Table S9B).
These data reveal that target-specific retrograde AA signaling acts via a PKC family kinase to define the cotransmitter release properties of innervating presynaptic terminals. Taken together, our observations on the regulation of peptide cotransmitter use at individual synapses uncover a previously undefined molecular signaling network that tunes the cotransmitter release characteristics of individual presynaptic terminals (Fig. 10).

Discussion
In this study we sought to identify the cellular and molecular mechanisms that specify the differential use of neurotransmitters at the individual presynaptic terminals formed by a cotransmitting neuron-a key mechanism for synaptic heterogeneity that underlies the selective processing and relay of information within neuronal circuits. Here, using Lymnaea cardiorespiratory neurons, we have identified (i) a new component of NTF-dependent plasticity that selectively defines peptidergic cotransmission, (ii) a novel function for the Lymnaea synaptophysin homologue in the inhibitory regulation of neuropeptide release, and (iii) a role for AA metabolites as a targetspecific retrograde signaling mechanism that establishes cotransmitter specificity. While we do not exclude the possibility that there may be concurrent changes in the expression or function of postsynaptic receptors, our observations of presynaptic neuropeptide accumulation and the inhibitory role of synaptophysin identify the presynaptic terminal as an important locus for target-and context-dependent regulation of cotransmission. Taken together, our findings define a molecular signaling network that underlies presynaptic cotransmitter specificity, and can allow for the context-dependent tuning of synaptic network function underlying the plasticity peptidergic cotransmission. One particular question in the field of cotransmission that has remained unresolved is whether the synaptic colocalization of multiple neurotransmitter substances necessarily implies their corelease 63 . Here, we show that peptide cotransmitter release at inter-neuronal synapses is regulated by context-dependent inhibition. FMRF neuropeptides accumulated at functionally non-peptidergic presynapses, while functionally peptidergic presynapses showed comparatively little localization, indicating that colocalization does not imply cotransmission 63 . Consistent with this notion, a previous study on the Drosophila neuromuscular junction showed that neuropeptide stores are low in resting terminals, where they are supplied in a use-dependent manner by the activity-dependent capture of rapidly transiting LDCVs 64 . This suggests that, in contrast to classical transmission, peptidergic transmission may be sufficiently maintained in the absence of a sizeable reserve of LDCVs in the presynaptic bouton. www.nature.com/scientificreports/ The main focus of our study was to define the molecular mechanisms that selectively regulate neuropeptide cotransmitter release, however, our observation that the release of ACh from VD4 is tuned independently from FMRF neuropeptide release warrants further investigation to advance our understanding of the dynamic regulation of cotransmitter release and presynaptic heterogeneity. We suspect that there are a number of mechanisms, likely involving target-specific signaling interactions and cross-talk amongst signaling networks, at play. First, work in Drosophila has demonstrated that neuropeptide autoreceptor signaling via presynaptic convergence adjusts cotransmitter output by enhancing the release of classical transmitters to accelerate the circadian rhythm directing sleep-wake behaviour 65 . Second, we have previously reported that ACh release competency in VD4 is scaled by presynaptic cAMP/PKA activation during competition for postsynaptic innervation 66 . Third, in the murine hippocampus, target-specific retrograde AA signaling has been found to induce presynaptic spike broadening by local inhibition of voltage-gated potassium channels, leading to synaptic facilitation 67 . Taken together, these observations indicate that a variety of mechanisms differentiate the function of individual presynaptic terminals to establish synaptic heterogeneity and enable selective information processing in neuronal circuits.

Synapse-specific inhibition of neuropeptide release by synaptophysin. Considering our obser-
vations on role of the Lymnaea Syp homologue in inhibiting LDCV release, it is important to note that mammalian Syp has been widely regarded to be specifically localized to SSVs, and absent from LDCVs 68 . If the mechanism described here is conserved, one possibility is that inhibitory regulation of LDCVs in mammals is mediated by another member of the Syp protein family. Another possibility raised by studies in neuroendocrine cells is that Syp is present at varying concentrations in LDCVs, and at lower concentrations than in SSVs 69,70 . This is consistent with a study on the protein composition of single SVs reporting that some SVPs, including Syt I, are sorted with high precision, whereas others, including Syp and Syb, show significant intervesicle variability 71 . In view of our findings on the inhibitory regulation of neuropeptide release by Lymnaea Syp, it seems that synaptic cotransmitter specificity might be achieved by two complementary mechanisms: the differential trafficking of release competent (Syp-) or release incompetent (Syp +) LDCVs to presynaptic terminals; and synapse-delimited PTM of Syp that alters the release competency of Syp + LDCVs.
Regarding the differential trafficking of Syp + and Syp− LDCVs, N-glycosylation, a PTM that occurs in the endoplasmic reticulum and affects protein sorting and trafficking, has been shown to be essential for targeting Syp, but not other SVPs such as Syt, out of the cell body and to the synapse 72 . This suggests that N-glycosylation might act as a selectivity filter to govern the amount of Syp incorporated during LDCV biogenesis, which could regulate the release competency of LDCVs and also account for the controversial and variable presence of Syp on LDCVs. Downstream of biogenic sorting, the synapse-specific targeting of cargo, including LDCVs, www.nature.com/scientificreports/ is a second highly regulated process that differentiates the composition and function of individual presynaptic terminals during synaptogenesis and synaptic plasticity 64,73,74 . Considering our findings, it may be that NTF/ RTK and retrograde AA signaling establish synapse-specific populations of release-competent and -incompetent LDCVs by influencing Syp N-glycosylation and the selective trafficking or capture of Syp + vs Syp-LDCVs at presynaptic terminals. Regarding synapse-delimited PTM of Syp, Syp-Syb affinity has been reported to be modulated by GSK-3 activity, where phosphorylation of Syp by GSK-3 inhibits Syb dissociation and attenuates transmitter release 53 . Here, we found that target-dependent regulation of neuropeptide release involved the regulation of GSK-3 activity by NTF/RTK signaling, and PKC family kinase activity by retrograde AA signaling. The inhibition of GSK-3 by RTKs is known to involve the Par3-Par6-aPKC cell polarity complex 61 . During CNS development, the asymmetric accumulation of these polarity proteins mediates neurogenic divisions by neural progenitors, and regulates microtubule dynamics underlying axogenesis and directed outgrowth 61 . An attractive mechanism for the local regulation of neuropeptide release by GSK-3-dependent phosphorylation of Syp would be the differential localization of the Par3-Par6-aPKC cell polarity complex underlying the functional heterogeneity of individual presynaptic terminals.
Functional implications for context-dependent tuning of peptidergic transmission. In Lymnaea, NTF-dependent synaptic network plasticity has been proposed to mediate the seasonal regulation of heart rate and breathing behaviour during aestivation 75 . The regulation of FMRF neuropeptide release described here identifies a new mechanism with significant potential to underlie the plasticity of cardiorespiratory behaviour induced by NTFs. Increases or decreases in NTF/RTK signaling would differentially affect FMRF neuropeptide release at VD4′s excitatory synapse with VF to promote or inhibit cardiac drive, and the reciprocal inhibitory synapse with RPeD1 to facilitate or attenuate rCPG rhythmogenesis and breathing behaviour.
In the stomatogastric ganglion (STG) of the crab Cancer borealis, a cotransmitting projection neuron (MPN) selectively activates and inhibits the expression of pyloric and gastric mill motor patterns using its neuropeptide cotransmitter at synapses within the STG network and its small molecule cotransmitter at synapses with projection neurons in the commissural ganglia 6 . Thus, it seems that the selective use of neuropeptide and classical small molecule cotransmitters with distinct postsynaptic targets is a common mechanism underlying the expression of patterned motor behaviour in invertebrates. It would next be pertinent to determine whether an analogous mechanism involving target-specific retrograde AA signaling and Syp-dependent inhibition of LDCV fusion may be involved in expressing gastric mill motor patterns in the Cancer STG, and whether this may be a more general mechanism for the target-specific regulation of cotransmitter release underlying the appropriate expression and context-dependent plasticity of patterned motor behaviours across species.
The regulated complexing interactions of Syp and Syb have been suggested to act in establishing a reserve pool of SVs that can be recruited by synaptic activity 76 . Applied to the regulated secretion of neuropeptides, this adjustable reserve function would have interesting implications for our understanding of the dynamic propensity of neurons to modulate the release of their neuropeptides. In Lymnaea, release of the ovulation hormone requires priming by a prolonged period of coordinated network spiking activity 77 . Activity-dependent dephosphorylation of Syp might switch release-incompetent LDCVs to release-competent ones, perhaps involving activity-dependent NTF/RTK signaling. This mechanism would allow resting secretory terminals to accumulate neurohormones to levels sufficient for sustaining endocrine signaling, and maintain the system in an energetically favorable 'primed-but-off ' state that would decouple neurohormone release from slower somal-dependent processes.
In mice Syp knockout does not result in any readily apparent structural or synaptic deficits, however, changes in short-and long-term plasticity and hippocampus-dependent behaviour have been reported [78][79][80] . These findings suggest that mammalian Syp is an essential regulator of synaptic plasticity, although previous studies have been unable to elucidate the underlying mechanism. Moreover, Syp is one of the genetic mutations implicated in human X-linked intellectual disability 81 . A compelling possibility raised by our work is that the synaptic plasticity deficits linked to mammalian Syp family proteins may involve the aberrant release of peptide cotransmitters. Further characterization of this Syp-dependent mechanism for regulated neuropeptide release, as it pertains to synaptic specificity and plasticity, is therefore likely to offer new fundamental insights into neurological disorders.

Methods
Animals and neuronal cell culture. Lymnaea stagnalis were raised under standard conditions in freshwater aquaria at room temperature (~ 22 °C) on a diet of lettuce. Animals (6-8 weeks old) were anesthetized with a 10% Listerine solution (10 min) and sacrificed by dissection of the CNS in normal Lymnaea saline. Identified neurons VD4, VF and RPeD1 were isolated from trypsinized CNS by suction applied through a glass pipette and then cultured on poly-L-lysine coated glass culture dishes, as previously described in detail 23,82 . Neurons were maintained overnight (15-20 h) in defined media (DM, NTF-deficient media; L-15; Life technologies; special order) or NTF-rich CNS conditioned media (CM, CNS-incubated DM), as described elsewhere 41 .
Molecular biology. The Lymnaea Syp homologue was identified from the Lymnaea EST library 83 using a BLAST search (blastn, NCBI) against mouse Syp (Accession number NM_009305), which revealed a single hit containing a complete protein-coding sequence (Accession number ES572211). We screened clones of the Lymnaea Syt I homologue (Accession number AF484090) 84 , and identified C2B-α (Syt-α) and C2B-β (Syt-β) isoform splice variants, each with a frequency of ~ 50%, suggesting that both transcripts are abundant in the CNS. The coding sequences for Syp, Syt-α, and Syt-β were cloned from Lymnaea CNS cDNA using Platinum Taq DNA polymerase (Invitrogen), and then subcloned into pBlueScript SK-(Clontech), as described elsewhere 30 (Table S10). Synthetic mRNA was made with the mMESSAGE mMACHINE T7 Ultra transcription kit, according to manufacturer's instructions (Ambion). Purified synthetic mRNA was dissolved in molecular-grade water (3-6 μg/ μL) and microinjected into VD4 neurons using a sterile low resistance glass electrode (10 pulses, 250 ms, 10 PSI). Molecular-grade water was used as a vehicle control. Microinjections were performed shortly after neurons were plated, allowing time for the cells to adhere to the poly-L-lysine coated coverslip (~ 2 h). Synapses developed overnight in CM or DM.
Immunocytochemistry and imaging. Neurons were fixed for 30 min with 1% paraformaldehyde and 0.2% picric acid in 1xPBS, then permeabilized for 1 h with incubation media (IM) containing 0.2% Triton, 5% fetal calf serum, and 0.25% fish gelatin in 1xPBS. Primary antibodies against FMRF neuropeptides (detects the common Arg-Phe-NH 2 moiety 85 ), or mCherry (BioVision, 5993) were used at 1:500 in IM for 1 h. The secondary antibody (Invitrogen, Alexa Fluor 546 Goat α Rabbit) was used at 1:100 in IM for 1 h. Three 15 min washes in 1xPBS were performed between each incubation, and all incubations were performed at room temperature. Cells were mounted with MOWIOL containing DAPI.
Samples were imaged with an A1R MP microscope using a CFI Plan Fluor 20x/0.75 MI objective (Nikon). Laser excitation wavelengths were 402 and 561, in series, and emission wavelengths were collected through 450/50 and 595/50 filter cubes (Nikon). For quantitative analysis, imaging parameters were kept the same amongst relevant samples. Images were acquired with NIS Elements v4.13.00 (Nikon), and intensity analysis was performed with ImageJ (NIH). To determine the synaptic localization of FMRF neuropeptides and Syp-mCherry, we measured the mean fluorescence intensity at 3 presynaptic varicosities per pair. Presynapses were defined by their distinct morphology coincident with FMRF or mCherry immunoreactivity. We have previously combined this approach with FM1-43 dye uptake to verify that morphologically defined varicosities are functional presynaptic terminals 35 . A small number of samples in which presynaptic varicosities were not apparent were excluded from analyses.
Electrophysiology. Simultaneous intracellular current clamp recordings and Lucifer yellow (LY) dye injections were performed as previously described for in vitro 41 and in situ 86 preparations. Briefly, for cultured neurons, cells were maintained in CM or DM (in mM: 40 NaCl; 1.7 KCl; 4.1 CaCl 2 ; 1.5 MgCl 2 ; 5 HEPES; pH 7.9), and glass microelectrodes (resistance 20-50 MΩ) were filled with a saturated solution of K 2 SO 4 . Cells were visualized under an inverted microscope (Axiovert; Zeiss), and impaled using micromanipulators (Narishige). VD4, VF and RPeD1 neurons were identified by size, plating position, and electrophysiological signature. Signals were amplified with a Neuro data dual channel intracellular recording amplifier (Neuron Data Instruments), digitized (Digidata 1440; Molecular Devices), and recorded with Axoscope 10.2 (Molecular Devices). For in situ recordings, Lymnaea CNS were dissected as above and briefly treated (5-10 s) with focally applied Protease Type XIV (Sigma-Aldrich) to soften and remove the inner sheath. Isolated CNS were maintained in normal Lymnaea saline (in mM: 51.3 NaCl; 1.7 KCl; 4.1 CaCl 2 ; 1.5 MgCl 2 ; 5 HEPES; pH 7.9), and neurons were visualized under a Wild stereoscope (Wild Leitz AG), identified by size, position and color, then impaled with K 2 SO 4 -filled glass microelectrodes (resistance 20-50 MΩ) using micromanipulators (Sutter Instruments). Signals were amplified with a Neuro data amplifier, digitized (NI DAQPad; National Instruments), and recorded using custom Electro-Lite software.
To determine ACh release characteristics from VD4, the mean amplitude of 5 consecutive PSPs, recorded in VF or RPeD1 neurons at a holding potential of − 100 mV was measured in response to single action potentials elicited by brief depolarizing current injection in VD4. Peptidergic transmission via FMRF neuropeptides was isolated pharmacologically using an AChR antagonist cocktail containing 5 μM methyllcaconitine (MLA), 10 μM tubocurarine (TC), and 20 μM tetraethylammonium (TEA), incubation time 5-10 min 87 . To determine FMRF neuropeptide release characteristics from VD4, the excitation of VF at holding potentials of − 60 mV in vitro or − 70 mV in situ (below firing threshold), or inhibition of RPeD1 at − 50 mV in vitro or − 60 mV in situ (at firing threshold), was measured in response to a ~ 4 s burst elicited by depolarizing current injection in VD4. These holding potentials were used for in vitro and in situ recordings because ionic differences between the external recording solutions shift neuronal firing thresholds. Peptidergic transmission was deemed to have occurred if stimulation of VD4 resulted in the induction of action potentials in VF (exhibits an excitatory response to FMRF neuropeptides from VD4) or the cessation of firing and hyperpolarization of RPeD1 (exhibits an inhibitory response to FMRF neuropeptides from VD4). To determine DA release characteristics from RPeD1, the inhibition of VD4 at a holding potential of − 50 mV was measured in response to a ~ 4 s burst elicited by depolarizing current injection in RPeD1. Dopaminergic transmission was deemed to have occurred if stimulation of RPeD1 resulted in the cessation of firing and hyperpolarization of VD4 (exhibits an inhibitory response to DA from RPeD1). In presynaptic neurons neither the duration nor the number of action potentials per burst were significantly different amongst experimental conditions (see Supplementary Information). Data analysis was performed using AxoScope 10.3 (Molecular Devices). chemicals. TC, MLA, TEA and pertussis toxin (PTX) were purchased from Sigma-Aldrich. AA, Chelerythrine chloride (Ch Cl -), 3-(2,4-Dichlorophenyl)-4-(1-methyl-1H-indol-3-yl)-1H-pyrrole-2,5-dione (SB 216763) and Tocrisolve 100 were purchased from Tocris. 5,8,11,14-Eicosatetraynoic acid (ETYA) was purchased from Santa Cruz Biotech. Drugs were dissolved in dimethyl sulfoxide (DMSO) or DM according to solubility, and then added to culture media shortly after neurons were plated (~ 1 h). The next day, three washes with DM fol-Scientific RepoRtS | (2020) 10:13526 | https://doi.org/10.1038/s41598-020-70322-5 www.nature.com/scientificreports/ lowed by replacement with fresh DM or CM was performed to remove drugs prior to experiments (~ 1 h). 0.1% DMSO and Tocrisolve 100 vehicle controls were performed for all experiments, as appropriate.
Experimental design and statistical analysis. All reported results are derived from ≥ 2 independent experiments, to ensure reliability and replicability of the data. Sample sizes were limited by uncontrollable factors inherent to the cell culturing techniques used. Data analysis was performed blinded by acquisition file number. Statistical analyses were performed using SPSS Statistics v26 (IBM). Data sets were assessed with Shapiro-Wilk test for normality, and parametric (P > 0.05) or non-parametric (P < 0.05) statistical tests were performed as appropriate. Non-parametric data sets were analyzed with Wilcoxon Signed Ranks tests, Mann-Whitney U tests, or Kruskal-Wallis tests. Parametric data sets were analyzed with Student's Paired or Independent Samples T-tests, or One-way ANOVA with Games-Howell (unequal variance; Levene statistic P < 0.05) or Tukey's HSD (equal variance; Levene statistic P > 0.05) post hoc tests, as appropriate (see Supplementary Information).

Data availability
All data generated or analyzed during this study are included in this published article and its Supplementary Information file. Materials generated during the current study are available from the corresponding authors on reasonable request.