Functional mapping of brain synapses by the enriching activity-marker SynaptoZip

Ideally, elucidating the role of specific brain circuits in animal behavior would require the ability to measure activity at all involved synapses, possibly with unrestricted field of view, thus even at those boutons deeply located into the brain. Here, we introduce and validate an efficient scheme reporting synaptic vesicle cycling in vivo. This is based on SynaptoZip, a genetically encoded molecule deploying in the vesicular lumen a bait moiety designed to capture upon exocytosis a labeled alien peptide, Synbond. The resulting signal is cumulative and stores the number of cycling events occurring at individual synapses. Since this functional signal is enduring and measurable both online and ex post, SynaptoZip provides a unique method for the analysis of the history of synaptic activity in regions several millimeters below the brain surface. We show its broad applicability by reporting stimulus-evoked and spontaneous circuit activity in wide cortical fields, in anesthetized and freely moving animals.

R ecent years have witnessed the birth of a variety of molecular tools to control or sense neuronal activity (see for review [1][2][3][4] ). The latter family includes several genetically encoded fluorescent indicators that can detect changes in either membrane voltage or calcium concentration (see for review [5][6][7][8][9] ). Ideally, a complete understanding of the physiological role of a specific area or circuit would require the ability to map synaptic transmission independently from the electrical activity of neurons. Following the introduction of FM dyes in the 90s 10 , there have been significant advances in the development of methods that can probe synaptic communication [11][12][13][14][15][16][17] , which include elegant genetically encoded fluorescent indicators of vesicle exo-endocytosis 11,12,14,16,18 . The most important inherent constraint of these techniques is their limited applicability to the analysis of the living mammalian brain. This is due to the transient nature of generated signals, whose detection requires online optical imaging. Besides the requirement of expensive and complex instrumentations, it can only access, with the necessary optical resolution, synapses that are no more than a millimeter below the brain surface, because of reduced light penetration in tissue 19 . This warrants the development of alternative approaches to monitor changes in synaptic activity that could be applied to brain circuits. Here, we have expanded the repertoire of available synaptic methods to include a reporter that, due to the binding of a peptide ligand 20 , generates a very specific and stable coiled-coil interaction 21 . The output relates to presynaptic bouton activity and is a long-lasting integration signal reporting the history of vesicle cycling, hence well-suited for both online and retrospective circuit analysis. We provide proof of principle of this strategy and show here its effectiveness in evaluating activity changes at the level of individual brain synapses, following in vivo experimental interventions.

Results
Design and validation of SynaptoZip (SZ). To develop Synap-toZip, one component of the Velcro coiled-coil heterodimer 20 , the Zip module, was fused to the intraluminal C-terminus of VAMP-2/Synaptobrevin2, followed by a Myc tag. Its fluorescent variant, eGFP-SynaptoZip (GZ), was generated by adding eGFP to the cytosolic N-terminus (Fig. 1a). The other component of Velcro was chemically synthetized for use as soluble SynaptoZip binder, Synbond (SB; 3.9 KDa; Fig. 1a). Both SynaptoZip variants (GZ and SZ) expressed in Hela cells were found to be localized in the endoplasmic reticulum, in the Golgi complex, and in a population of highly motile vesicles, presumably endocytic vesicles/recycling endosomes, due to their colocalization with internalized transferrin (data not shown; see below). Immunoblotting of Hela cells expressing SynaptoZip revealed a major band of 28 KDa for SZ and 55 KDa for GZ, which were recognized by anti-VAMP2, as well as by anti-Myc antibodies ( Fig. 1b; Supplementary Fig. 1a). Interestingly, fluorescent SB directly labeled the appropriate M.W. bands on blot membranes, and consistent results were obtained with neuronal extracts (Fig. 1b; Supplementary Fig. 1a, b). To test whether extracellular SB molecules were competent for SynaptoZip binding and uptake during cycles of constitutive vesicular recycling, SynaptoZip-expressing Hela cells were incubated with fluorescent SB molecules (5 nM; SB-Alexa647), for periods ranging from 1 to 60 min ( Supplementary  Fig. 2a, b). SB concentration was selected based on in vitro binding affinity 20 and live binding experiments on Hela cells ( Fig. 1c; K d = 5.09 nM; N = 36-80; logistic fitting, adj. R 2 = 0.98). As shown in Fig. 1d, in a mixture of expressing and nonexpressing cells, SB was selectively internalized in expressors, following ongoing or constitutive vesicular exo-endocytosis ( Fig. 1d; GZ left; SB and DIC image, right).
Since the pool of SB binding sites in Hela cells should be fairly stable (within a short-time window), we evaluated the behavior of SB uptake over time ( Supplementary Fig. 2a, b). Specific SB internalization in GZ-expressing Hela cells could be detected already at 1-2 min from its application (1 min: N expr = 64, SB expr = 0.018 ± 0.017, N non-expr = 49, SB non-expr = 10 −4 ± 0.0022; mean ± SD; p < 10 −4 Wilcoxon rank sum test), and at later time points its intracellular accumulation increased approaching a plateau level ( Supplementary Fig. 2a, b), presumably because of the progressive saturation of the cycling sites by elevated constitutive exo-endocytosis. When cells were magnified, the red SB labeling could be fully attributed to eGFP-fluorescent endocytic vesicles expressing SB binding sites (Fig. 1d, insets). Consistent results were obtained with the eGFP lacking variant SZ (see below). SBlabeled vesicles showed physiological dynamics characteristic of endocytic vesicles/recycling endosomes (Supplementary Movie 1). SB uptake was not seen at 4°C (data not shown), suggesting the requirement for constitutive exo-endocytosis 22 in SB internalization. Chemical fixation (10-15 min, 4°C, 4% paraformaldehyde, PFA) did not alter SB retention and a comparable estimate of the amount of internalized SB could be obtained before and after fixation (data not shown).
As shown by the histograms (Fig. 1e), SB uptake occurred essentially by a specific interaction of SB with the luminal epitope of GZ, and not by unspecific fluid phase endocytosis, because only GZ-positive cells showed a strong and fast SB uptake signal (Fig. 1d, e; SB 5 nM, 60 min incubation; N = 320, non-expressing cells; N = 636, GZ-expressing cells; SB expr = 0.284 ± 0.089, SB non-expr = 0.063 ± 0.014, mean ± SD; p < 0.01, Kolmogorov-Smirnov test (KS)). Co-application of SB and antibodies against the intraluminal Myc tag resulted in the co-localization of both molecules in the same GZ-positive vesicles (data not shown). After long-term exposure to SB, its uptake linearly correlated with GZ expression level, both assayed in the sub-membrane region of Hela cells ( Supplementary Fig. 2c), where constitutive endocytic vesicle recycling takes place. Similar results were obtained by analyzing individual vesicles (Supplementary Fig. 2d, e). Altogether these results confirm that SynaptoZip is correctly inserted and oriented in the vesicle membrane. Exocytosis exposes to the extracellular environment the C-terminal Zip module which then binds extracellular SB, leading to its very specific entrapment in the endocytic vesicle.
Despite continuous constitutive vesicle recycling, pre-bound SB-Alexa647 (5 nM, O/N incubation, 37°C) was neither significantly displaced by subsequent application of an excess of SB-Alexa488 (Fig. 1f, left panel; time lapse imaging; SB-Alexa488, 100 nM, 37°C, N = 6 cells; p = 0.84, Wilcoxon signed-rank test; analyzed windows 0-10 vs. 50-60 min), nor spontaneously lost up to 48 h after its vesicular uptake (Fig. 1f, right panel; N = 38-74 cells per data point; Pearson correlation: ρ = −0.047, p = 0.26). This suggests that vesicular SB is not free but it is bound to GZ in a stable complex, as predicted by Velcro in vitro stability 20 . Sequential brief incubations (10 min, 37°C) with three different SB fluorescent variants (SB-Alexa488, SB-Alexa568, and SB-Alexa647, 5 nM each) resulted in comparable macroscopic patterns of vesicular staining (Fig. 1g). When cumulative fluorescence distributions from the three sequential epochs were plotted, these nearly coincided ( Fig. 1h; average fluorescence; N = 92 cells; p ≥ 0.26; KS test). In the above experimental conditions, the absence of a significant distortion of the SB uptake distribution, over time and following repeated incubations, indicates good assay reproducibility, without saturation of available SB-binding sites. As a whole, these results show that the amount of the SZ-SB complex can effectively and quantitatively report vesicular exo-endocytosis in a cultured cell line, generating a persistent and reliable functional signal.
SynaptoZip reports synaptic activity. To test whether this approach could be applied to monitor synaptic exo-endocytosis, we expressed GZ in cultured hippocampal neurons and in various brain regions (hippocampus, visual thalamus, primary visual, and prefrontal cortices). In all preparations investigated, clear eGFP fluorescent presynaptic varicosities could be seen. These were confirmed as bona fide synapses, because of the large extent of colocalization with pre-synaptic and post-synaptic markers ( Supplementary Fig. 3a, b).
The representative images of Fig. 2 Fig. 2a (top), in acute hippocampal slices from animals transduced in the CA3 region, Schaffer collaterals labeling indicated a very precise colocalization between GZ-positive synapses and SB uptake sites. The VAMP-2-based design and the lumenal localization of the SB-binding site (Fig. 1a) suggest that the most likely mechanism of SB synaptic labeling relates to the exo-endocytosis of synaptic vesicles. In order to confirm this hypothesis, we used FM1-43, an established in vitro approach to study exo-endocytosis of synaptic vesicles 23 . Separately we evaluated SB localization inside synapses by super-resolution microscopy (ground-state depletion, GSD 24 ). At FM1-43 preloaded GZ-expressing synaptic varicosities, evoked presynaptic activity was found to induce SB labeling and parallel unloading of FM1-43, thus indicating that both molecules use the exoendocytic pathway (N = 6 experiments; an exemplar experiment is shown in Supplementary Fig. 4a-c). By super-resolution microscopy we found that the SB synaptic uptake resulted in the labeling of distinct vesicular structures ( Fig. 2b-d), whose diameter distribution and average value (41.2 ± 16.2, mean ± SD; N = 428 vesicles from 24 synapses) were found to be fully consistent with those of synaptic vesicles 25 . We then investigated the relation between the level of circuit activation and SB synaptic uptake. Synaptic uptake of SB was monitored at synapses belonging to individual axons of GZ expressing cultured hippocampal neurons, which were stimulated and recorded from in the whole-cell mode (WC; N = 11 independent experiments). As depicted in the representative experiment shown in Fig. 2e, f, the elicitation of action potentials (10 Hz trains, 10 APs each, 4 s inter-train intervals) induced clear synaptic uptake at GZ-positive boutons in the presence of extracellularly applied SB-Alexa647 (pale-blue shaded areas in Fig. 2f; SB 5 nM; 24°C; N = 18 boutons; see also Supplementary Movie 2). Without stimulation, spontaneous synaptic uptake remained fairly low. This is clearly indicated by the drop in the uptake rate when the elicitation of action potentials was interrupted (white-shaded areas in the presence of SB in Fig. 2f). The activity-dependent nature of SB synaptic uptake was further confirmed in experiments where neurons were bathed in tetrodotoxin (TTX; 1 µM), which was found to strongly reduce SB uptake rate (data not shown). Importantly, application of SB to the extracellular medium or its AP-dependent uptake did not perturb action potential firing and the electrophysiological synaptic behavior. The latter conclusion arises from the observed stability of evoked synaptic currents ( Supplementary Fig. 5a, b), as well as amplitude and frequency of miniature events (mini frequency and amplitude before and during SB, p = 0.31 and p = 0.12, respectively; TTX, 1 µM; Supplementary Fig. 5c, d).
Since synapses integrate pre-synaptic and post-synaptic compartments to evaluate the relation between the strength of transmission and the amount of SB uptake, we conducted a series of dual electrophysiological recordings from synaptically connected pairs of neurons (N = 5 pair recordings; Fig. 3a). To modify the amount of transmitter release and therefore SB uptake, we varied the presynaptic AP stimulation frequency (0.2-40 Hz). As evident from the representative experiment shown in Fig. 3a-f, the increase in stimulation frequency induced a clear facilitation of excitatory postsynaptic currents (EPSCs) (Fig. 3b, d) and larger EPSCs were found to correlate with higher increments in synaptic SB uptake (Fig. 3e). This relationship diverges from a linear approximation just for very large synaptic currents, indicating a reduced sensitivity for strong release regimes. In summary, the vesicular localization of SB and the uptake dependence upon presynaptic activation strongly support the validity of SynaptoZip as a tool to measure the degree of synaptic transmission.
SynaptoZip reports evoked activation of brain circuits. The clear functional signal obtained from synapses in acute brain slices (Fig. 2a) highlights that SB diffuses inside the brain tissue, reaching the synaptic cleft. This prompted us to test this method in the living brain. SynaptoZip was expressed in rat V1 cortex and medial prefrontal cortex (mPFC) using viral vectors, followed by the local nano-rate delivery of SB through a glass micropipette (Supplementary Fig. 6a-d). Thanks to the long-term stability of the SB-GZ pair (Fig. 1f), the amount of SB uptake could be determined ex vivo, thus grasping signals even from boutons that could not be reached with the today available optical techniques. We found clear evidence for in vivo SB uptake at synapses (Figs. 4,5). The efficacy of this approach likely relates to the small M.W. and therefore diffusibility of the SB peptide, since larger M. W. tracers (anti-Myc antibodies targeted to the C-terminal luminal epitope of SZ) did not work in similar conditions (Supplementary Fig. 7a-e). Great care was taken to improve the assay reproducibility, with a precise location and timing of SB application, and by selecting the most appropriate region for image acquisition based on SB tissue distribution profile (Supplementary Fig. 8a-c, e; see Methods).
We began by exploiting the rat primary visual cortex (V1), where circuit activity can be easily controlled by light (Supplementary Fig. 9a-e). With this aim in mind, the visual thalamus (LGN) was stereotaxically transduced in its dorsal portion (dLGN; see methods for details), resulting in clear GZ expression in thalamic neurons and ipsilateral V1 thalamo-cortical synapses ( Fig. 4a-f). These eGFP fluorescing presynaptic varicosities were located in layer IV and to a smaller extent in deeper as well as in more superficial cortical layers of the V1 cortex ( Fig. 4c-f). In these experiments, after GZ cortical expression, SB was delivered into the V1 superficial layers (I-II) (Fig. 4g, Supplementary  Fig. 6a, b).
To test whether SB uptake could be induced by evoked activation of thalamo-cortical synapses, these animals were either exposed to visual stimulation with brief light pulses (light stimulated; pulse width 250 ms, intensity 340 µW/cm 2 , at 2 Hz for 15 min) or kept in the dark for the same length of time (ctrl; Fig. 4h). The light-stimulation protocol was devised in a series of preliminary experiments aimed at eliciting reproducible responses with a good balance between ON and OFF phasic responses, to maximize synaptic cortical activation under sevoflurane anesthesia ( Supplementary Fig. 9). In unstimulated animals, the SB functional signal obtained from GZ-positive boutons located in layer IV was weak, with few clearly SB fluorescing synapses, evidently reflecting a small degree of spontaneous tonic OFF activity (a representative experiment is shown in Fig. 4i, j). This is consistent with the strong inhibition of spontaneous cortical activity by sevoflurane in V1 cortex 26 . Following repeated light ON-OFF stimulations, the synaptic SB uptake was augmented, with a clear increase in the number of SB fluorescing thalamo-cortical layer IV GZ-positive boutons (a representative experiment is shown in Fig. 4k, l; see also Supplementary Movie 3). Our experimental protocol and analysis approach allowed selecting FOVs for synaptic uptake analysis located on the flat portion of SB spatial concentration profile (600-800 µm from microperfusion sites) resulting in low extracellular SB concentration variability and comparable levels in non-synaptic areas among the different experimental groups (Supplementary Fig. 8d; see Methods). Fluorescence histograms obtained from the above experiments clearly indicate that in the absence of stimulation the synaptic SB uptake is not distinguishable from tissue background (   Overall, this suggests that GZ can be effectively used to report stimulus evoked activation of brain circuits occurring in vivo. SynaptoZip detects enduring activity changes by ketamine. To further test the applicability of this approach in freely moving animals, we examined the effect of ketamine on the mPFC. Previous micro-dialysis experiments have provided evidence that a single sub-anesthetic dose of ketamine induces neurotransmitter accumulation in the extracellular space 27,28 , although the involvement of synaptic exocytosis and the contribution of the intrinsic mPFC circuitry are not established. Prelimbic mPFC of adult rats was stereotaxically transduced with GZ in layer V (Fig. 5a), resulting in the appearance of eGFP fluorescing neurons and axonal synaptic boutons belonging to the local mPFC circuitry (layers I-VIa; Supplementary Fig. 10). After 2-3 weeks from transduction, a bolus of SB was micro-perfused at low rate, ∼700 µm above the GZ-LV injection site (Fig. 5a, Supplementary  Fig. 6c). During this SB perfusion epoch (30 min the mPFC ( Supplementary Fig. 6d), with negligible synaptic uptake ( Supplementary Fig. 11a, b).  Fig. 5c-h), evidently a reflection of the strong spontaneous activity of local prefrontal circuits in the awake state 29 . Interestingly, this synaptic labeling was still detectable 7 days after SB uptake ( Supplementary Fig. 11c, d), suggesting that, in this time frame, the previously generated synaptic GZ-SB complex remains stable. Compared to controls, both ketamine-treated groups showed a clear augmentation of synaptic SB uptake. At early and late time points from ketamine treatment, changes in synaptic uptake affected the entire range of activity levels, with an homogeneous shift of fluorescence cumulative distributions ( Fig. 5i; N ctrl = 216, N ket = 275, N 72hket = 227; p ctrl-ket < 0.0001, p ctrl-72hket < 0.0001; p ket-72hket = 0.35; KS test, Bonferroni-Holm (B-H) correction; data from same experiments presented in Fig. 5c-h). As for V1 cortex, in the mPFC areas where synapses were analyzed (600-800 µm from the SB injection site) the estimates of SB concentration in non-synaptic areas indicated that synapses were exposed to similar SB concentrations (Supplementary Fig. 8f). Population data from these experiments showed that at early times, compared to controls, ketamine almost doubled synaptic SB uptake, an activity change that is still observable 72 h after ( Ketamine is known to increase animal motility 30 , an effect that could indirectly drive mPFC activity by enhanced input from several sensory modalities. At early time points, ketamine was found to induce a clear increase in locomotion ( Supplementary  Fig. 13a-d, g, j, k), characterized by ataxic movements (Supplementary Fig. 13h, i). However, this behavior was apparently lost 72 h after ketamine injection ( Supplementary  Fig. 13e-k), while the potentiation of synaptic SB uptake persisted (Fig. 5i, j). Thus, the long-lasting change in mPFC circuit activity induced by ketamine is not explained by increased animal motility.

Discussion
We have developed a novel synaptic activity reporter, which operates both in vitro and in vivo, whose integrated signal is durable and can be acquired not only online but most importantly by retrospective analysis. Signal integration arises from specific binding and trapping into the synaptic vesicle of the SynaptoZip partner SB, whose synaptic enrichment reflects the number of synaptic exo-endocytic events occurring during an experimental epoch. This method, which generates reliable activity estimates, fulfills the demand for a thorough post hoc analysis of brain synapses. Its application is cost-effective, and can reach synapses that are outside the grasp of standard optical imaging methodologies 19 , hence potentially usable to map the activity of the whole brain. Our approach is compatible with immunolabeling and possibly clearing methods (see for review 31 ). Importantly, SynaptoZip could be combined with neuronal activity integrators [32][33][34] , to solve the functional contribution of upstream and downstream circuital elements.
Here, we have applied SynaptoZip to the V1 cortex, obtaining reliable estimates of light-evoked synaptic activity in the anaesthetized rat (Fig. 4). Indeed, like other exo-endocytosis indicators 10-17 , SynaptoZip generates a synaptic signal that depends on presynaptic action potential firing (Fig. 2), whose frequency is strongly modulated at thalamo-cortical synapses by phasic ON-OFF light stimulations ( Supplementary Fig. 9). In the mPFC this approach was applied to reveal drug-induced changes in spontaneous synaptic activity in freely moving animals (Fig. 5).
In these experiments we used ketamine, a drug that at subanesthetic doses exerts a rapid antidepressant effect, whose mechanism and site of action are still highly debated 28,35 . We found that the activity of the intrinsic mPFC synaptic network is rapidly enhanced by a single ketamine administration, a change that persisted for at least 72 h. This finding agrees with previous micro-dialysis experiments, which provided evidence for ketamine-induced glutamate accumulation in the extracellular space 27,28 . By use of SynaptoZip, we show that an enhanced vesicular release of glutamate at the level of layer V to layers II-III mPFC synapses is the most likely mechanism behind glutamate accumulation, which could induce and/or maintain subsequent plastic changes 36 in the mPFC circuitry.
Regarding the design of the SynaptoZip-SB pair, we used the Velcro coiled-coil heterodimer 20 for its very high-binding specificity and pair stability 20,21 , but also because the soluble peptide partner (SB) is small (3.4 KDa), thus favoring an efficient diffusion in brain extracellular spaces and into the synaptic cleft ( Supplementary Fig. 7). In the future it might be interesting to evaluate alternative peptide pairs with different properties and binding affinities 37 . VAMP-2, SynaptoZip vesicular scaffold, has been previously used in other synaptic activity sensors 11,12,14 . VAMP-2 is the most abundant vesicular protein 38,39 , with a single transmembrane domain and a very short, presumably inert, intraluminal segment 40 . Conceivably, both factors have guaranteed the high level of vesicular expression found here, which is important to avoid a fast saturation of the integrated SB signal. In addition, because VAMP-2 is a ubiquitous synaptic vesicle protein, it is present at all synapses independently of their neurochemical phenotype, and SynaptoZip would be well integrated with the native fusion machinery.
As demonstrated here, our approach works well in vivo, although some methodological aspects would benefit from ad hoc future developments. At present, the major technical difficulty relates to the SB delivery phase into the brain tissue, a procedure that involves extreme accuracy during the experimental and subsequent analysis phase. Clearly, at a given stimulation rate, the cumulative index of synaptic activity at each bouton would depend upon the number of exo-endocytotic events along the experimental epoch, the experimental parameter that is important to extract, but also upon the average number of SB molecules successfully bound at each fusion event. The latter would be influenced by SB concentration in the extracellular space (Supplementary Fig. 12d-i), by fusion kinetics 16,18 and its modality, but also by the average number of available binding sites at each individual vesicle that can be extracted by the expression signal. In our experimental conditions, the extracellular tracer concentration at regions selected for analysis was found to be fairly constant and comparable among the different groups (Supplementary Fig. 8d, f). Regarding the dynamics of fusion and retrieval of synaptic vesicles, known to display complex kinetics, a dissimilar coverage of the different modalities 41-44 might occur. Such limitation might contribute to the shape of the relationship between SB uptake and synaptic transmission, with some hints of saturation at very high stimulation frequencies (Fig. 3f). In the future, it would be important to find alternative ways to deliver SB locally at synapses, but also to test smaller versions of these tracers. Despite these caveats, since the integrated output of SynaptoZip clearly correlates with the degree of synaptic activation (Figs. 2-4; Supplementary Fig. 4 and Supplementary Movie 2), this ensures that even at this stage, a reliable mapping of brain synaptic activity can be achieved, a result that cannot be obtained with any other presently available technique.
Concerning behavioral studies, if more effective ways for the chronic and timed-controlled delivery of SB to brain synapses were to be found, the analysis of time-dependent changes in synaptic activity in freely behaving animals would become feasible. Interestingly, the long-term stability of the vesicular SynaptoZip-SB complex for at least 1 week ( Supplementary  Fig. 11c, d) suggests that multiple chasing with different SB fluorescent versions could track the fate of synapses and synaptic vesicles, thus helping to address important queries in behavioral neuroscience but also in neurobiology 45 , and in the field of circuit plasticity 36 . This technology could also be used for the selective delivery of molecules to active terminals, to tag or remove circuits involved in specific behaviors. In the future, following ad hoc labeling of SB with radioactive isotopes or paramagnetic molecules, our approach could be adapted for in vivo imaging by animal PET and fMRI. Since antibodies targeting natural intraluminal vesicular epitopes 46,47 cannot be applied as in vivo functional probes (Supplementary Fig. 7), we envisage that the search for small ligands for native synaptic vesicle intraluminal epitopes may be an important and rewarding future avenue, capable of extending our approach to naïve animals.

Methods
Research and animal procedures. Research and animal care procedures were approved by our Institutional Animal Care and Use Committee for Good Animal Experimentation in accordance with Italian MIUR code of practice for the care and use of animals for scientific purposes (IACUC numbers: 576, 541, and 543). Experiments were performed on Sprague Dawley male rats (150-350 g). Animal group allocation was known by the investigators. All efforts were devoted to minimize animal's distress, pain, and suffering during the entire course of the experimentation. All animals were caged with free access to food and water ad libitum and were exposed to 12 h light/dark cycles at 23°C constant room temperature.
Constructs and expression vectors. SynaptoZip (SZ) was made starting from a cDNA clone in pBluescript KS provided by Elferink and colleagues 48  ARTICLE fragment (174 bp) was inserted in frame that codes, in the order, SVPEG (an adapter) and KGVEPKTYCYYSS (a spacer 49 ), AQLEKELQALEKENAQLE-WELQALEKELAQ (Acid-p1 20 ), EQKLISEEDI (c-Myc tag), ending with an ad hoc stop codon (TGA). The remaining 3′UTR (1300 bp) was left intact and in place. For the expression vector of the non-fluorescent variant SynaptoZip (SZ) a HindIII-SacI fragment was transferred into pBeta-Actin, kindly provided by Andrew Matus. The fluorescent variant eGFP-SynaptoZip (GZ) was obtained by in-frame fusion to eGFP at the EcoR1 site of pEGFP-C2 (Clontech); it comprises 13 foreign codons from the vector MCS and lacks the first VAMP-2 codon. For insertion into the lentiviral (LV) transfer vector, kindly provided by L. Naldini 50 , an Age1-Sal1 fragment containing GZ was used to replace the pre-existing eGFP, resulting in GZ-LV transfer vector. LV particles were produced as previously described 50 and stock titers (~10 9 TU ml −1 ) determined. All constructs were confirmed by DNA sequencing (Eurofins Genomics, Italy). SB peptide was produced by synthesis (CGGAQLKKKLQALKKKNAQLKWKLQALKKKLAQ; JPT Peptide Technologies GmbH, Berlin, Germany) and fluorescent dyes were conjugated to the terminal Cys (Alexa Fluor Dyes®: 488, 568, 647; Thermo Fisher Scientific).

Experiments with Hela cells. Hela cells (ATCC ® CCL-2™;
LGC Standards S.r.l., Italy; STR authenticated), regularly tested to be mycoplasma free, were grown on glass coverslips (25 mm diameter) or plastic petri dishes (10 cm diameter) at 37°C, in 5% CO 2 humidified incubator in DMEM supplemented with 10% (v/v) FCS, GlutaMax, and antibiotics. For SZ expression, cultures were transfected with vector DNA using Lipofectamine 2000 (Invitrogen) or transduced with GZ-LV vector prediluted in culture medium (final titer for transduction in vitro:~10 6 TU ml −1 ). Incubation with SB (0.1-100 nM) was run either in DMEM at 37°C 5% CO 2 or in oxygenated Tyrode solution at 23°C. For long-term experiments, DMEM was supplemented with 1% FCS (v/v). For fixed samples, at the end of incubation, cells were extensively washed first with cold Tyrode solution containing 1% BSA, followed by PBS and PFA 4% (30 min, 4°C).
Acute hippocampal slices. Acute hippocampal slices were obtained from Sprague Dawley rats (150-200 g) previously transduced (2 weeks before) with GZ-LV in the right hippocampus CA3 (stereotaxic coordinates: 3.8 mm mediolateral, ML, from the sagittal axis; −3.55 mm rostrocaudal, RC, from bregma; 3.8 mm dorsoventral, DV). For brain acute slice preparation, after a lethal injection of thiopental (50 mg i.p.; RotexMedica, GMBH, Germany), rats were intracardially perfused with icecold, modified Ringer solution adapted for cutting (119 mM NaCl, 2.5 mM KCl, 1 mM CaCl 2 , 3 mM MgSO 4 , 26.2 mM NaHCO 3 , 1 mM NaH 2 PO 4 , 11 mM D-glucose) bubbled with 95% O 2 and 5% CO 2 , and containing 5000 IU l −1 heparin (Pharmatex, Milano, Italy). After decapitation, brains were quickly removed and transferred to ice-cold modified Ringer. After dissection of the right hippocampus, transverse slices (450 μm thick) were cut using a tissue chopper (Stoelting, Wood Dale, IL), submerged in the above solution adapted for recording (2 mM CaCl 2 and 1.3 mM MgSO 4 ), warmed at 33.6°C for 1 h and then maintained at room temperature up to 6 h. For SB uptake experiments, acute slices were placed on the slice recording chamber positioned under the microscope and extracellularly perfused with the above recording solution (24°C, bubbled with 95% O 2 and 5% CO 2 ). SB (5 nM) was directly added to the extracellular solution, modified by the addition of 30 mM KCL (isosmotic).
In vivo SB uptake experiments. A balance between subjects design without randomization was adopted for all in vivo SB uptake experiments. In most experiments, two animals were tested a day and their temporal order was reversed the next session (for example, control animal before/after light stimulated animal). For experiments on the visual system, all phases were run with animals fully anaesthetized as described above. 2-3 weeks from viral transduction, in the preparatory phase, the animal was blindfolded bilaterally in order to prevent exposure to ambient light during surgery and micro-pipette placement. Stereotaxic coordinates for SB delivery were: V1, 2.85 mm ML; −7.35 mm RC; 0.8 mm DV. For these experiments we used SB-Alexa647, 13.3 μM dissolved in standard Tyrode solution (30 min; 100 nl min −1 ). During SB perfusion control animals were maintained in complete darkness. Light-stimulated animals were initially kept in complete darkness (15 min), then their left eye was unblinded and light pulses were administered for the subsequent 15 min. Light stimulation was delivered using a white LED matrix placed at a distance of 1.2 cm from the eye (250 ms pulse duration; 340 μW irradiance; 2 Hz stimulation rate). For mPFC experiments, animals fully anaesthetized as described above were locally perfused with a bolus of SB (SB-Alexa647, 1.9 μM dissolved in standard Tyrode solution; 30 min; 100 nl min −1 ) at coordinates 0.65 mm ML, +3.5 mm RC, 2.8 mm DV; after micro-pipette removal and re-suturing (see above). Animals were waken up and left freely moving in cages for 1.5 h. For drug treatments, ketamine (15 mg/kg) or vehicle were i.p. administered either immediately after the end of micro-perfusion of SB-Alexa647 (animal was still under anesthesia), or 72 h before micro-perfusion of SB-Alexa647. At the end of perfusion, animals still under general anesthesia were sacrificed with a lethal dose of Thiopental (Thiopental, Vuab-Pharma) and then intracardially perfused with cold saline supplemented with heparin (5000 IU l −1 ), followed by PFA 4% in 120 mM phosphate buffer (pH 7.4 at 4°C). The brain was removed, submerged in fixative O/N (4°C), embedded in 4% agar. 35-40 µm thick slices were cut with a vibratome (VT1000S, Leica, Germany), when required processed for immunofluorescence, then mounted onto glass slides (Fluorsave, EMD Millipore). For detection of PSD95A, we used the method by Schneider et al. 54 on unfixed 15 µm thick slices (Cryostat sectioning; CM 1850, Leica, Germany).
Behavioral analysis. For behavioral experiments a balanced within subjects design without randomization was adopted. Rats were placed in a behavioral cage and their motility monitored after their awaking from sevoflurane anesthesia, in conditions fully matching in vivo SB experiments. The behavioral cage was composed of the rat's home cage bottom (38 × 22 cm 2 ) and a custom-built roof hosting a digital camera for video recording (1280 × 720 pixel 2 resolution, 25 fps, NoIR v2 camera controlled by Raspberry Pi 3, Raspberry Pi Foundation). Food and water were always provided during sessions (water dispenser in the top-right corner; room temperature 23°C, low light illumination). Rats were subjected to four video recording sessions each lasting 90 min (day 0: preconditioning to behavioral cage; day 1: control recording; day 2: recording after ketamine administration; day 5: recording 72 h after ketamine administration). Ketamine was administered during anesthesia (i.p., 15 mg/kg). Tracking and analysis of rat movement were performed using custom software (Python and Matlab). A pause was considered as a period of time where rat barycenter remained confined for more than 10 s in a circular area (2 cm radius). The central area crossing index used an area at the center of the cage with dimensions 19 cm × 11 cm.
Image acquisition and data analysis. All fluorescence images (excluding superresolution) were obtained using a confocal microscope (LSM510, Zeiss) equipped with 488, 543, and 633 nM laser lines and standard filter sets for FITC, TRITC, and CY5 excitation/emission spectra. In preliminary experiments we evaluated the range of acquisition parameters (pinhole size, laser power, and detector gain) best suited to each experimental condition, in order to favor signal to noise ratio while preserving linearity and avoid signal saturation. Typhoon800 data acquisition was performed on all brain sections ( Supplementary Fig. 6) to identify the SB injection site. At the beginning of confocal imaging acquisition, low-magnification scanning (10× objective) was carried out on slices to evaluate GZ expression and locate the highest SB intensity spot and select the appropriate areas for the highmagnification acquisition step (63× objective; Supplementary Fig. 8). No relevant information about SB uptake by synapses could be discerned at low magnification. The high-magnification step to acquire images from synaptic areas (as well as later selection of mPFC axons) was performed using the GZ signal as reference, so that no information about the functional signal (SB uptake) was known by experimenter. Image analysis was performed using either ImageJ (NIH) or custom code developed in Matlab (Mathworks). All numerical analyses were then performed in Matlab (Mathworks) or Origin (OriginLab).
For experiments with Hela cells, cell-contours were segmented based on GZ fluorescence (or on SB-Alexa647 fluorescence when expressing SZ lacking eGFP). Hela vesicles were identified by the Particle Analyzer plugin of ImageJ using thresholding parameters based on the average and SD of vesicle size, roundness, and fluorescence. For Hela vesicles fluorescence thresholding, background noise was measured in vesicle-free regions near the plasma membrane using a circular ROI with similar radius and fluorescence thresholds were then set to mean background fluorescence plus 2 SD. For cultured neurons experiments, confocal time-lapse recordings (interval 10 s) were filtered by median and Kalman filters (equalized for producing color maps). Synapses belonging to the same axon were manually selected on the GZ channel and automatically segmented in each frame. SB and GZ fluorescence were averaged at each synapse and their ratio taken as normalized SB uptake. The time series produced was filtered (moving average on four samples) after removing the initial baseline (no SB in the bath). For superresolution localization of SB uptake in synaptic vesicles, we used a GSD microscope (SR GSD 3D, Leica Microsystems, Buffalo Grove, IL), the PALMsiever Matlab toolbox was then used to generate images from localization lists (kernel density estimation, pixel size set to 4.12 nm). Standard Matlab functions were utilized for circle segmentation and measurements of vesicle diameters (synapse contours were manually traced using the eGFP low-resolution image).
For in vivo experiments, synapses were automatically identified and segmented based on GZ fluorescence. A few animals were not included in the analysis when GZ viral transduction was below detection (lack of a significant eGFP signal in neuronal somata at the site of transduction; N excluded /N tot : V1, 5/15; mPFC, 6/24). Due to discrepancies in tissue morphology, different approaches were used for V1 and mPFC experiments. In the former case, synapses were segmented using a custom algorithm based on image thresholding, Voronoid tessellation and then selected based on shape and fluorescence intensity (threshold: 2 SD of background noise). In V1 experiments, synapses were considered active when average SB fluorescence was higher than background SB fluorescence plus 1 SD. In mPFC experiments, synapses were segmented using the median filtered GZ channel by a custom algorithm based on image local contrast equalization, image thresholding, and then selected based on their radius. In these mPFC experiments, to avoid any misdetection due to the sporadic presence of GZ expressing somata (local transduction), synapses were accepted for quantitative estimates only when positioned along putative axons (visually identified as short sequences of GZ fluorescing boutons connected by an inter-bouton axonal structure). The normalized SB uptake index was computed at the level of each individual synapse as the ratio of SB uptake and GZ expression, both background subtracted. SB and GZ synaptic fluorescence values were expressed as averages over segmented areas, while the corresponding signal background referred to the surrounding fluorescence (circular region around each segmented synapse). For population comparisons, the index of interest (V1: number of active synapses; mPFC: normalized SB uptake) was averaged among different fields of view each coming from the same subject (V1: 2-5 fields/rat; mPFC: 3 fields/rat). In most experiments acquisition parameters were kept constant. For those experiments where they had to be modified, a linear de-trend was applied to the index of interest (trend estimated on the same experimental group; independent variable: detector gain or detector gain times laser power). All pseudo-color representations of SB fluorescence for the in vivo experiments (color maps) here presented have not been equalized for the sake of comparability. All custom software used for the analysis can be provided upon request.
Statistical analysis. All statistical analysis was done by routines written in Matlab (Mathworks) and R or using Origin (OriginLab). Student's t-test (paired or unpaired; always two-tailed) was used only after testing for normality by onesample KS test. Two-sample KS test was used to evaluate differences between distributions of fluorescence data sets. Minimum level of significance was set at 5%. Permutation tests for independent samples (10 4 permutations; two-tailed) were performed as previously described 55 . B-H correction of p-values was applied for multiple comparisons. For both in vitro and in vivo investigations, preliminary experiments were run to evaluate the expected degree and variability of SB uptake and establish the most adequate sample size, with the aim of both reaching a meaningful statistical power and minimizing the number of animals involved in the study. The final set of in vitro and in vivo results showed large SB uptake to noise ratios, indicating that the initial estimates were appropriate (for example, the normalized effect size for V1: 4.0; mPFC acute ketamine: 3.56; power > 0.999). Error bars plot SD or s.e.m. as indicated in the text.
Data availability. The eGFP-SynaptoZip (GZ) cds sequence used in this study has been deposited at GenBank ® with the accession code MF797884 (http://www.ncbi. nlm.nih.gov/Genbank). DNA constructs and data that support the findings of this study are available from the corresponding author upon reasonable request.