Stable excitatory-inhibitory synapse balance despite dynamic turnover

Synapse assembly, maturation, and elimination require spatial and temporal coordination of several cell biological processes. These events are initiated by pre- and postsynaptic recognition of combinatorial networks of cell surface receptors, adhesion molecules, and secreted factors (Südhof, 2018). These extracellular networks subsequently initiate bi-directional signaling cascades that stabilize nascent synapses and recruit the corresponding functional machinery essential for diverse physiological synaptic parameters. Synapse elimination likely involves termination of these signals and disassembly of trans-synaptic adhesion complexes (Südhof, 2018; Südhof, 2021). While increasing evidence supports this model, the dynamics of mammalian synapse assembly, together with the interplay between excitatory and inhibitory synapses during these processes, remain incompletely understood due to the paucity of approaches to visualize these events in real time (Lichtman and Fraser, 2001).

A dynamic relationship between excitatory and inhibitory synaptic function is critical for computation by neural circuits. Developmental imbalances between excitatory and inhibitory synapses contribute to neurological disorders, including autism (Rubenstein and Merzenich, 2003). Moreover, the pathophysiology of Rett syndrome, a severe neurodevelopmental disorder, involves a reduction in spontaneous activity caused by an increase in synaptic inhibition over excitation (Dani et al., 2005). Interestingly, a structural coordination between excitatory and inhibitory synapses and functional interactions between these synapses can maintain a constant excitatory/inhibitory ratio along the same dendritic tree (Liu, 2004). This raises the hypothesis that excitatory and inhibitory synapse assembly is coordinated to maintain a functional balance during development.

Testing this hypothesis requires approaches to simultaneously visualize interactions between mammalian pre- and postsynaptic compartments over relatively long timescales with high spatiotemporal resolution (Kaech et al., 2012; Okabe, 2013; Green et al., 2019). Long-term imaging studies have revealed dynamic structural changes associated with circuit remodeling during development. Invertebrate systems enable global and long-term time-lapse imaging of circuit assembly in vivo (Andlauer and Sigrist, 2012; Özel et al., 2019), including high content, long-term visualization of growth cones and filopodial dynamics during Drosophila visual system development (Özel et al., 2015; Agi et al., 2024). In mammalian systems, dendritic spines, structures associated with excitatory postsynaptic compartments, can also be visualized long-term during their formation and dynamics (Kwon and Sabatini, 2011). Studies in the CA1 region of live mice provided evidence for a remarkably dynamic turnover of dendritic spines (Attardo et al., 2015). This was in contrast to studies in the neocortex, which found a higher level of spine stability, suggesting that different cell types may display different synaptic turnover rates to support their network functions (Holtmaat et al., 2005; Xu et al., 2009; Yang et al., 2009). Despite these advances, methods to study both excitatory and inhibitory synapses concomitantly are limited.

To address the dynamic interplay between excitatory and inhibitory synapse assembly in mammalian neurons, we generated long-term time-lapse imaging approaches and a computational analysis pipeline to simultaneously monitor the dynamics of presynaptic Synaptobrevin-2/VAMP2 (Syb2), excitatory postsynaptic Homer1c, and inhibitory postsynaptic Gephyrin, which are routinely used to label pre- or postsynaptic compartments. These strategies enable examination of the spatial and temporal dynamics of excitatory and inhibitory synaptic compartments, and their co-clustering into contacts. By applying these approaches, we observed that subsets of synapses are forming and eliminating even in mature hippocampal neurons. Despite this, excitatory and inhibitory synapse levels remain balanced over time. Collectively, these approaches provide new insights into the dynamics and balance of inhibitory and excitatory synapses during mammalian circuit establishment.

While canonical synaptic proteins are routinely used as markers of pre- or excitatory postsynaptic compartments, long-term live imaging of both simultaneously has been challenging. Thus, we developed live imaging procedures to visualize excitatory synapse dynamics over relatively long timescales (Figure 1). We first generated a set of lentiviral reporters based on routinely used synaptic markers, including mClover3-Homer1c and HaloTag-Syb2 fusion constructs. mClover3 is a brighter and more photostable version of the green fluorescent protein Clover (Bajar et al., 2016), and HaloTag is a self-labeling tag with a panel of versatile and highly stable fluorescent dyes available, including far-red dyes (Los et al., 2008). To test if these reporters could be used for labeling synapses, we validated their proper localization via co-staining hippocampal neurons for other endogenous synaptic markers (Figure 1A–F). Hippocampal neurons transduced with HaloTag-Syb2 were live labeled with cell-permeant JF646 HaloTag ligand (Grimm et al., 2017), and subsequently immunolabeled for endogenous synaptic markers. As expected, HaloTag-Syb2 co-localized with presynaptic Syn1/2 (Synapsin1/2), and co-clustered together with postsynaptic Homer1 (Figure 1A–C). Additionally, mClover3-Homer1c exhibited higher co-localization with the excitatory postsynaptic marker, SHANK2, compared to the inhibitory postsynaptic marker Gephyrin (Figure 1D–F).

Figure 1 with 1 supplement see all

Download asset Open asset

We next tested their efficacy ex vivo in the hippocampal CA1 circuit (Figure 1G–I). We injected AAV encoding HaloTag-Syb2 into the CA3 to label Schaffer collateral inputs into the CA1 and injected AAV expressing mClover3-Homer1c into the CA1. This approach yielded robust labeling of pre- and postsynaptic Schaffer collateral synapses in the stratum oriens and radiatum regions (Figure 1G–I). These results support that our fluorescently labeled reporters serve as reliable proxies for their respective synaptic compartments in native circuits and demonstrate their utility to label pre- and postsynaptic compartments simultaneously.

We next tested their compatibility and efficacy for long-term time-lapse imaging. Pre- and postsynaptic combinations (HaloTag-Syb2 with mClover3-Homer1c) were lentiviral-transduced into primary hippocampal neurons together with lentiviral-delivered mTagBFP2 as a cell marker. Cells were then labeled live with JF646 HaloTag ligand as before and imaged at DIV12-14, a point when excitatory synapses mature in primary hippocampal neurons (Chanda et al., 2017; Sando et al., 2022). Cells were imaged for 15 hr total at 5 min intervals in a live cell incubation chamber with resonance scanning confocal microscopy (Figure 1J & K, Figure 1—video 1). This permitted stable imaging of labeled synaptic compartments, including co-clustered pre- and postsynaptic compartments, without detectable photobleaching (Figure 1J). This imaging strategy captured nascent synapses assembling even in relatively mature primary hippocampal neurons, such as HaloTag-Syb2-positive growth cones approaching and forming contacts with mClover3-Homer1c postsynaptic compartments along mTagBFP2-filled dendrites (Figure 1K).

We next generated custom-built analysis pipelines to enable identification, tracking, and categorization of individual puncta and co-clustered puncta over time (Figure 2, Figure 2—figure supplements 1 and 2). Due to potential microscope drift over long imaging periods, we created a new drift correction algorithm that corrects for image drift while preserving biological motion (e.g. neurite motion over time; Figure 2—video 1, see Methods). We observed that some neurites were highly mobile during the 15 hr imaging period while others were stable. Thus, we binned neurons into two categories, moving or still, based on several quantitative parameters (Figure 2—figure supplement 1). Puncta from still neurites displayed significantly longer durations, lower speed, and less total net displacement over time compared to moving neurites (Figure 2—figure supplement 1). We focused on analysis of still neurites for further tracking and quantification for simplicity.

Figure 2 with 4 supplements see all

Download asset Open asset

To better understand the temporal dynamics of synaptic puncta in cultures, we further stratified synaptic puncta into subclasses. Co-clustered, paired HaloTag-Syb2/mClover3-Homer1c puncta (defined as synapses) were identified and categorized based on their persistence and formation/elimination during the imaging period (Figure 2A & B). Stable synapses persisted over the entire imaging period, while transient synapses were short-lived. A third intermediate class of synapses was composed of either nascent, eliminated, or synapses that turnover based on their presence and/or absence at the beginning and/or end of imaging (Figure 2B). These analysis tools enabled long-term tracking and categorization of synapses over a 15-hr period (Figure 2C & D, Figure 2—video 2). Moreover, they allowed quantification of the abundance of distinct synapse populations and their distribution during the imaging period (Figure 2E & F). Accordingly, we observed sub-populations of synapses being eliminated over time, while others were newly formed (nascent) using these approaches (Figure 2F). Stable synapses displayed significantly lower movement speed compared to transient or stable-like synapses from the same cultures (Figure 2G), suggesting a relationship between temporal and spatial dynamics. These methods were also effective at tracking and quantifying unpaired HaloTag-Syb2 or mClover3-Homer1c puncta (Figure 2—figure supplement 2). These computational approaches highlight the dynamic nature of excitatory synapses over these timescales and the interplay between pre- and postsynaptic compartments in live hippocampal neurons.

Previous studies have shown that excitatory synapses are functionally maturing at approximately DIV12 in primary hippocampal neurons (Chanda et al., 2017; Sando et al., 2022). Therefore, we next examined excitatory synapses relatively early (DIV8-9) and later (DIV11-14) during maturation in culture (Figure 3, Figure 2—figure supplement 2F, G). Temporal dynamics of synapses were similar between groups (Figure 3A–F). Homer1c puncta displayed significantly higher total durations at DIV8-9 compared to presynaptic Syb2 or synapses, suggesting postsynaptic compartments are less mobile and less dynamic relative to presynaptic (Figure 3C, Figure 2—figure supplement 2B, C). Similarly, at DIV11-14, Homer1c puncta were more persistent than synapse pairs (Figure 3D–F). However, the duration of Syb2 puncta was not significantly different than Homer1c in this group, suggesting that presynaptic compartments become more stable over time (Figure 3D–F).

Figure 3

Download asset Open asset

Interestingly, despite these spatiotemporal dynamics, total puncta density remained constant over the 15 hr imaging session in both DIV8-9 and DIV11-14 neurons (Figure 3G & H). Furthermore, the density of synapse pairs was significantly lower than individual pre- and postsynaptic components at both time points, suggesting a subset of available pre- and postsynaptic sites remain even at the later maturity stage (Figure 3I). Many of these available pre- or postsynaptic compartments remain unassociated for the observable period, while other puncta tightly co-associated for a prolonged time (Figure 2—figure supplement 2D). Based on that we categorized them as either unpaired or paired (Figure 2—figure supplement 2E). We then quantified the speed and track duration of paired and unpaired puncta at the two different time points (Figure 3J–M). Unpaired Syb2 puncta displayed higher speed than paired in DIV8-9 cultures compared to DIV11-14, supporting more highly dynamic presynaptic compartments are present early during maturation (Figure 3J). Unpaired Syb2 track durations were significantly shorter than paired in both culture periods, suggesting paired puncta persist longer (Figure 3K). However, unpaired Homer1c puncta speed was comparable to paired, indicating less mobility relative to presynaptic sites (Figure 3L). Paired Homer1c puncta again persisted longer than unpaired in both culture periods (Figure 3M). We also observed similar levels of colocalization and similar total amounts of puncta in both culture periods (Figure 2—figure supplement 2F–G). These results highlight that neurons maintain an equilibrium of excitatory synapse density despite continual spatiotemporal dynamics.

We then generated analogous lentiviral tools to facilitate visualization and tracking of inhibitory synapses over time (Figure 4). To facilitate compatibility with our other tools, we generated a tdTomato-Gephyrin fusion protein. We validated that this fluorescent fusion protein was highly co-localized with the inhibitory postsynaptic marker GABARα1 but not the excitatory postsynaptic marker Homer1 (Figure 4A). We then tested the efficacy of the AAV-delivered fusion protein in the hippocampal CA1 region (Figure 4B & C). We utilized the Chrna2-Cre line, which is selective for hippocampal OLM (oriens lacunosum-moleculare) interneurons (Leão et al., 2012; Nichol et al., 2018). These neurons reside in the stratum oriens and project to the stratum lacunosum-moleculare (Figure 4B). Contrary to our experiments labeling excitatory synapses in the stratum oriens and radiatum (Figure 1G–I), this approach selectively labeled inhibitory synapses in the stratum lacunosum-moleculare (Figure 4B & C). These results support the fidelity of these tools in labeling specific excitatory or inhibitory postsynaptic compartments.

Figure 4

Download asset Open asset

We then co-transduced primary hippocampal neurons with tdTomato-Gephyrin, HaloTag-Syb2, and mTagBFP2, and conducted live imaging, tracking, and quantifications as before, again comparing DIV8-9 to DIV11-14 cultures (Figure 5, Figure 5—figure supplement 1, Figure 5—video 1). Our analysis tools effectively visualized presynaptic and inhibitory postsynaptic puncta, as well as co-clustered pairs, over the 15 hr imaging period, and captured dynamic populations of inhibitory synapses (Figure 5A–F). Postsynaptic Gephyrin puncta trended towards more stable at both culture maturation times, with a larger fraction of stable synapse subtypes (Figure 5B & E), but no significant difference in duration from Syb2 (Figure 5C & F). Additionally, synapse tracks of co-clustered puncta pairs were shorter in duration than individual Syb2 tracks (Figure 5C & F), suggesting inhibitory postsynaptic compartments may not associate as strongly with Syb2-containing pre-synapses as excitatory postsynaptic compartments. This was consistent with other measurements including broader distribution of inter-puncta distances (Figure 2—figure supplement 2D, Figure 5—figure supplement 1D), larger fraction of unpaired Syb2 (Figure 2—figure supplement 2E, Figure 5—figure supplement 1E), and low fraction of paired puncta and co-cluster density (Figure 2—figure supplement 1F, G, and Figure 5—figure supplement 1F, G). Furthermore, we saw continuous turnover of puncta that maintained a stable density along dendrites (Figure 5G–I). We saw no significant differences between inhibitory paired and unpaired puncta motion speed (Figure 5J & L). However, Gephyrin synaptic pairs were substantially longer in duration when paired with Syb2 compared to non-paired Gephyrin puncta (Figure 5M), but not Syb2 paired with Gephyrin (Figure 5K). This observation suggests the Gephyrin postsynaptic association with Syb2 pre-synapses drives its puncta stability, but Syb2 puncta stability is likely primarily driven by association with excitatory postsynaptic compartments as observed previously (Figure 3J & K). Altogether, these tools enable long-term visualization and quantitative assessment of inhibitory synapses in primary hippocampal neurons.

Figure 5 with 2 supplements see all

Download asset Open asset

Given that our fluorescent reporters rely on overexpression of the tagged proteins, as a parallel approach, we also applied recent advances in CRISPR/Cas9 knock-in strategies to introduce tags into genomic loci within postmitotic neurons (Fang et al., 2021; Figure 6, Figure 6—figure supplements 1–3). We employed the TKIT (tandem knock-in with two guides) approach that enables the introduction of tags into endogenous neuronal genomic loci with high efficacy (Fang et al., 2021). Moreover, the TKIT labeling approach avoids the concern of variable indels in coding regions following CRISPR modification as observed with other strategies (Nishiyama et al., 2017; Gao et al., 2019). Consistent with these studies, we found that this system robustly tagged endogenous GluA2 in primary hippocampal neurons (Figure 6—figure supplement 1A, B). Since we are focused on developing tagging strategies for commonly used pre- and postsynaptic markers, we developed TKIT constructs for presynaptic Bassoon, excitatory postsynaptic Homer1c, and inhibitory postsynaptic Gephyrin, and generated AAVs encoding the respective sgRNAs/donor and Cas9.

Figure 6 with 3 supplements see all

Download asset Open asset

We first tagged Bassoon in primary hippocampal cultures with an N-terminal HA tag (Figure 6A–E, Figure 6—figure supplement 2A). When we virally transduced cultures with HA-Bassoon donor and sgRNAs, we detected robust HA labeling only when Cas9 was also co-delivered. The HA staining overlapped with the staining for endogenous Bassoon at DIV12-14 (Figure 6B & C). Endogenously tagged HA-Bassoon also co-localized with presynaptic Syn1/2 and formed co-clusters with postsynaptic Homer1 (Figure 6D and E, Figure 6—figure supplement 2A). Primary neurons harboring TKIT AAVs for HA tagging of Homer1c displayed HA staining that co-localized with endogenous Homer1 only when Cas9 was co-delivered (Figure 6G & H). Furthermore, HA-Homer1c co-localized with another excitatory postsynaptic protein, SHANK2, but was excluded from inhibitory postsynaptic sites stained for Gephyrin (Figure 6I and J, Figure 6—figure supplement 2B). We used an analogous approach to tag endogenous Gephyrin with GFP, which was based on the design of Fang et al., 2021 but with further optimized sgRNAs. As expected, tagged Gephyrin co-localized to inhibitory synapses with high fidelity and was excluded from excitatory synapses (Figure 6K–O, Figure 6—figure supplement 2C). Namely, GFP-tagged Gephyrin co-localized with endogenous Gephyrin and GABARα1, but not excitatory Homer1 (Figure 6M–O, Figure 6—figure supplement 2C). The Gephyrin TKIT tagging approach was also effective in the hippocampal CA1 region, where injection of the Gephyrin TKIT components resulted in GFP-Gephyrin fusion throughout CA1 pyramidal neurons (Figure 6P–R). Finally, we found that these tools could label both Homer1c and Gephyrin in the same neuron, indicating the potential to modify two targets in the same cell albeit at low efficacy (Figure 6—figure supplement 2D).

Tagged mRNA transcripts were only detected when donor/sgRNA AAVs were co-delivered with Cas9 (Figure 6—figure supplement 2E–G). Moreover, we estimated the overall efficiency of Homer1c and Gephyrin TKIT using standard curve RT-qPCR and found they modify approximately 5% of total transcripts (Figure 6—figure supplement 2H). We subsequently assessed potential CRISPR/Cas9 off-target effects (Figure 6—figure supplement 3). We identified the top two potential off-target sites for each gRNA, PCR amplified the genomic loci and sequenced to detect modifications (Figure 6—figure supplement 3). For the Homer1c TKIT, sequencing the targeted region detected introduction of the HA tag in the desired location, confirming TKIT modification (Figure 6—figure supplement 3A). The sequence of the predicted top two off-target sites for each Homer1c sgRNA matched the consensus genomic sequence, supporting the absence of off-target modifications (Figure 6—figure supplement 3B, C). Similarly, we identified TKIT-mediated tag introduction in Gephyrin and Bassoon without evidence of off-target modifications (Figure 6—figure supplement 3D–I). These CRISPR/Cas9 approaches build upon the original TKIT toolbox and facilitate the fusion of tags with endogenous presynaptic and postsynaptic proteins.

We subsequently used a Gephyrin donor to label this inhibitory postsynaptic protein with tdTomato for live imaging (Figure 7, Figure 7—figure supplement 1). We then conducted time-lapse imaging as described before in CRISPR-modified neuronal cultures at DIV11-14 (Figure 7A, Figure 7—video 1). To concurrently monitor presynaptic terminals and endogenous postsynaptic inhibitory compartments, we introduced tdTomato-Gephyrin TKIT AAVs together with lentiviral HaloTag-Syb2 and mTagBFP2 (Figure 7B–H, Figure 7—figure supplement 1). As a result, we detected and quantified the distinct subpopulations of synapses (Figure 7B–D). Furthermore, we quantified the individual and the paired HaloTag-Syb2:tdTomato-Gephyrin density along mTagBFP2-labeled dendrites during the imaging period (Figure 7E & F). As with lentiviral pre- and postsynaptic tracking, a steady-state density of puncta was maintained over time despite dynamic sub-populations (Figure 7B–F). Our analyses tools allowed us to characterize the temporal and spatial properties of inhibitory synapses and the non-paired synaptic compartments in neurons with fluorescently fused endogenous proteins. As a result, we demonstrated their similarity to the properties observed in cultures with overexpressed reporters (Figure 7J–M, Figure 7—figure supplement 1). Collectively, this molecular toolbox and tracking approaches enable quantification of the spatiotemporal dynamics and densities of excitatory and inhibitory synapses in live cells.

Figure 7 with 2 supplements see all

Download asset Open asset

We next analyzed excitatory and inhibitory synapses concurrently in neurons transduced with mClover3-Homer1c, tdTomato-Gephyrin, HaloTag-Syb2, and mTagBFP2 lentiviral reporters (Figure 8A). We quantified the density of Homer1c:Syb2 excitatory and Gephyrin:Syb2 inhibitory pairs at DIV8-9 and DIV11-14 (Figure 8B & C). Excitatory synapse density was significantly higher than inhibitory at both times (Figure 8B). We next quantified the ratio of total Homer1c:Syb2 co-clustered puncta counts relative to Gephyrin:Syb2 co-clusters from the same cultures and found that the overall ratio of excitatory to inhibitory synapses remained unchanged over this period (Figure 8C). These results suggest that despite the dynamic environment, hippocampal neurons maintain a balanced equilibrium of excitatory and inhibitory synapses during this period of culture maturity.

Figure 8

Download asset Open asset

To examine functional maturation of developing excitatory and inhibitory synapses, we recorded spontaneous mEPSCs and mIPSCs during synaptogenesis (Figure 8D–N). We monitored mEPSCs and mIPSCs from the same neurons by adjusting the holding potential (mEPSCs vH = –70 mV; mIPSCs vH = 0 mV) and analyzed their frequency and amplitude ratios at different maturation stages of the culture (DIV6, 10, and 16; Figure 8D–N). Cellular capacitance increased and membrane resistance decreased over time, consistent with previous reports and supporting maturation over culture time (Figure 8E & F). While mEPSC amplitude modestly increased over culture time (Figure 8G), mEPSC frequency increased and reached peak interevent interval by approximately DIV10 (Figure 8H). mIPSC amplitude was stable between DIV6 and DIV10 and increased as neurons matured by DIV16 (Figure 8I). mIPSC frequency progressively increased over time (Figure 8J). We then calculated the ratios of mEPSC/mIPSC amplitude and frequency ratios (Figure 8K–N) and observed no statistically significant differences between average mEPSC/mIPSC ratio measurements during culture maturation, suggesting that the development of excitatory and inhibitory spontaneous release is balanced as neurons mature.

Using live imaging approaches, we observed the dynamics between pre- and postsynaptic compartments in primary neuron cultures over extended time periods. These results indicate that subsets of synapses are spatially and temporally dynamic and continue to form and turnover even in relatively mature hippocampal neurons. Despite this turnover, excitatory and inhibitory synapses remain at an equilibrium over the time periods we examined. This balance is also preserved at the functional level, where the overall E/I ratios of spontaneous miniature release are maintained during maturation in cultured neurons. We postulate that intrinsic mechanisms possibly synchronize excitatory and inhibitory synapse assembly during circuit development. Two potential models may explain this synchronization. Temporal synchrony of intrinsic genetic programs may independently drive excitatory and inhibitory synapse maturation in parallel. Alternatively, regulation of excitatory and inhibitory synapse ratios may occur concurrently whereby signaling crosstalk between excitatory and inhibitory synapses dynamically balances excitation/inhibition and supports the functional outputs of neural circuits. These hypothetical mechanisms may help balance excitation/inhibition and support the functional outputs of neural circuits. Previous studies have demonstrated the role of synaptic plasticity and homeostatic plasticity on maintaining excitation/inhibition balance, demonstrating the intrinsic capacity for stimulus-dependent dynamic regulation (Turrigiano et al., 1998; Turrigiano and Nelson, 2004). Moreover, the activity state of a neuron/synapse including plasticity-inducing stimuli might alter the dynamic properties we observed and shift the balance of E/I synapse density. By tracking both excitatory and inhibitory synaptic structures in developing circuits, our approaches can be used in future studies to test these models.

Our live imaging studies enabled us to observe the motion and lifetime of synaptic compartments at key developmental timepoints. In addition to co-clustered pre- and postsynaptic compartments, we observed abundant puncta without a nearby corresponding pre- or postsynaptic compartment, recapitulating what has been extensively observed in immunostaining of endogenous synaptic proteins. There are two possibilities for the molecular identity of these unpaired synaptic compartments. First, the absence of a cognate pre- or postsynaptic marker may be indicative of a synapse lacking the specific tagged proteins. For example, certain synaptic proteins may accumulate at specific synapses (Metzbower et al., 2023), such that a given synaptic marker lacks the capacity to label all excitatory or inhibitory synapses. A second possible identity of unpaired pre- or postsynaptic proteins could be synapse precursor elements, which are subsequently stabilized upon recruitment to mature synaptic junctions. Indeed, we found that unpaired puncta exhibited distinct spatial and temporal characteristics, including shorter durations and faster speeds. This suggests recruitment at synapses could physically stabilize these protein clusters, prolonging their observable lifetime and physically restraining their motion. Available pre- or postsynaptic sites could also exist waiting for a viable partner but are eliminated without additional input, reducing average lifetime. Puncta lifetimes were also marker-specific, as seen with longer puncta duration of tagged Homer1c as compared to Syb2, indicating pre- and postsynaptic compartments' lifetimes could be partially independent. Further studies are necessary to understand the full relationship between observed spatial dynamics and the functional consequences on developing circuits.

Our approaches also display several inherent limitations which will require future advances to overcome. While we focused on canonical and widely used pre- and postsynaptic markers as proxies for synaptic compartments, our tools do not assess the adhesive interaction between pre- and postsynaptic compartments. Combining our approaches with other systems relying on adhesion between pre- and postsynaptic cell adhesion molecules, such as SynView, GRASP, or SynapseShot (Tsetsenis et al., 2014; Yamagata and Sanes, 2012; Son et al., 2024), will be important toward understanding further details of the dynamics of synapse assembly. Our approaches complement those based on trans-synaptic adhesion, considering the latter systems only exhibit signal after cell adhesion molecules contact, while our reporters can visualize synaptic components prior to forming a synapse. More generally, our observations support that some of the functional components of synapses are present in nascent pre- or postsynaptic compartments prior to their contact, including within growth cones. The live imaging tags in our tools can also be easily exchanged for a given experimental purpose or to introduce more robust labels as they are developed. Future studies combining our approaches with trans-synaptic adhesion reporters or genetically encoded activity indicators including GEVIs (Bando et al., 2019; Hao et al., 2024) will help unravel the temporal sequences of synapse assembly and maturation.

Our live imaging was conducted in neuronal culture given its broad applicability and tractability for molecular studies. While this system provides a powerful and widely used reduced model for mechanistic questions, future advances will be necessary to employ these tools in combination in vivo. Our proof-of-concept studies support the utility of these tools ex vivo, establishing a foundation for future optimization of their use in neural circuits. Considering our approaches relied on reporter expression or CRISPR/Cas9 genome modifications, both of which require time for protein buildup to detectable levels, we were unable to visualize the very early stages of postnatal synaptogenesis. Synapses form in excess during postnatal development in vivo, followed by refinement and elimination of connections (Südhof, 2017). Future advances will be required to assess the events at each stage of postnatal synaptogenesis.

While tagged reporters for Homer1, Syb2, and Gephyrin are routinely used, our approaches provide several new features and applications compared to existing live imaging approaches, including intrabodies (Dong et al., 2019; Trimmer, 2022). The lentiviral reporters and image tracking approaches are optimized for simultaneous monitoring of pre- and postsynaptic compartments, while currently reliable intrabody labeling approaches (FingRs) are available for postsynaptic proteins (Rimbault et al., 2024; Gross et al., 2013; Son et al., 2016; Cook et al., 2019). Furthermore, our viral approaches are designed to simultaneously image both excitatory and inhibitory synapses in the same neurons. While our TKIT studies focused on endogenous Gephyrin, the TKIT system can tag multiple different proteins in the same neuron. This raises the possibility of introducing distinct live imaging tags into different synaptic proteins in the same cell and monitoring them simultaneously. Ultimately, a combination of systems including viral tagging and intrabody labeling will be optimal for comprehensive evaluation of synapse dynamics during synaptogenesis.

Our work introduces new tools and methods that enable visualization of pre- and postsynaptic dynamics at high spatiotemporal resolution. We leveraged these new approaches to observe the spatial and temporal dynamics of individual synapses in neuronal cultures. Despite their continual temporal dynamics and turnover, excitatory and inhibitory synapse density remained balanced, which was further reflected by the functional balance of excitatory and inhibitory spontaneous synaptic transmission. Future work using these tools and others will continue to illuminate the molecular mechanisms used by neural circuits to maintain steady state levels of synapses, and how these structures are dynamically generated and removed.

Code generated in this study is available at the following deposit: https://github.com/datajamesallen/Garbett_2026_eLife (copy archived at Allen, 2026). Live imaging data generated in this study is available at Dryad: https://doi.org/10.5061/dryad.5qfttdzp5. All materials generated in this study will be shared by the lead contact upon request.

1. 1. Garbett KA
   2. Allen JP
   3. Lopez JM
   4. Smith CM
   5. Sando RC

   (2026) Dryad Digital Repository

   Data from: Stable excitatory-inhibitory synapse balance despite dynamic turnover.

   https://doi.org/10.5061/dryad.5qfttdzp5

1. 1. Agi E
   2. Reifenstein ET
   3. Wit C
   4. Schneider T
   5. Kauer M
   6. Kehribar M
   7. Kulkarni A
   8. von Kleist M
   9. Hiesinger PR

   (2024) Axonal self-sorting without target guidance in Drosophila visual map formation

   Science 383:1084–1092.

   https://doi.org/10.1126/science.adk3043

   • PubMed
   • Google Scholar
2. Software

   1. Allen J

   (2026) UniSynapse, version swh:1:rev:08be6e0626df4dd4c9a9869b350dce35b416808a

   Software Heritage.

   https://archive.softwareheritage.org/swh:1:dir:012806aa0baf7935851fc50dcd6087891bd6aa49;origin=https://github.com/datajamesallen/Garbett_2026_eLife;visit=swh:1:snp:80b2fbaca26be448fa250afb8a187027f9eba74d;anchor=swh:1:rev:08be6e0626df4dd4c9a9869b350dce35b416808a
3. 1. Andlauer TFM
   2. Sigrist SJ

   (2012) In vivo imaging of Drosophila larval neuromuscular junctions to study synapse assembly

   Cold Spring Harbor Protocols 2012:407–413.

   https://doi.org/10.1101/pdb.top068577

   • PubMed
   • Google Scholar
4. 1. Attardo A
   2. Fitzgerald JE
   3. Schnitzer MJ

   (2015) Impermanence of dendritic spines in live adult CA1 hippocampus

   Nature 523:592–596.

   https://doi.org/10.1038/nature14467

   • PubMed
   • Google Scholar
5. 1. Bajar BT
   2. Wang ES
   3. Lam AJ
   4. Kim BB
   5. Jacobs CL
   6. Howe ES
   7. Davidson MW
   8. Lin MZ
   9. Chu J

   (2016) Improving brightness and photostability of green and red fluorescent proteins for live cell imaging and FRET reporting

   Scientific Reports 6:20889.

   https://doi.org/10.1038/srep20889

   • PubMed
   • Google Scholar
6. 1. Bando Y
   2. Grimm C
   3. Cornejo VH
   4. Yuste R

   (2019) Genetic voltage indicators

   BMC Biology 17:71.

   https://doi.org/10.1186/s12915-019-0682-0

   • PubMed
   • Google Scholar
7. 1. Chanda S
   2. Hale WD
   3. Zhang B
   4. Wernig M
   5. Südhof TC

   (2017) Unique versus redundant functions of neuroligin genes in shaping excitatory and inhibitory synapse properties

   The Journal of Neuroscience 37:6816–6836.

   https://doi.org/10.1523/JNEUROSCI.0125-17.2017

   • PubMed
   • Google Scholar
8. 1. Cook SG
   2. Goodell DJ
   3. Restrepo S
   4. Arnold DB
   5. Bayer KU

   (2019) Simultaneous live imaging of multiple endogenous proteins reveals a mechanism for Alzheimer’s-related plasticity impairment

   Cell Reports 27:658–665.

   https://doi.org/10.1016/j.celrep.2019.03.041

   • PubMed
   • Google Scholar
9. 1. Dani VS
   2. Chang Q
   3. Maffei A
   4. Turrigiano GG
   5. Jaenisch R
   6. Nelson SB

   (2005) Reduced cortical activity due to a shift in the balance between excitation and inhibition in a mouse model of Rett syndrome

   PNAS 102:12560–12565.

   https://doi.org/10.1073/pnas.0506071102

   • PubMed
   • Google Scholar
10. 1. Dong J-X
    2. Lee Y
    3. Kirmiz M
    4. Palacio S
    5. Dumitras C
    6. Moreno CM
    7. Sando R
    8. Santana LF
    9. Südhof TC
    10. Gong B
    11. Murray KD
    12. Trimmer JS

    (2019) A toolbox of nanobodies developed and validated for use as intrabodies and nanoscale immunolabels in mammalian brain neurons

    eLife 8:e48750.

    https://doi.org/10.7554/eLife.48750

    • PubMed
    • Google Scholar
11. 1. Fang H
    2. Bygrave AM
    3. Roth RH
    4. Johnson RC
    5. Huganir RL

    (2021) An optimized CRISPR/Cas9 approach for precise genome editing in neurons

    eLife 10:e65202.

    https://doi.org/10.7554/eLife.65202

    • PubMed
    • Google Scholar
12. 1. Gao Y
    2. Hisey E
    3. Bradshaw TWA
    4. Erata E
    5. Brown WE
    6. Courtland JL
    7. Uezu A
    8. Xiang Y
    9. Diao Y
    10. Soderling SH

    (2019) Plug-and-play protein modification using homology-independent universal genome engineering

    Neuron 103:583–597.

    https://doi.org/10.1016/j.neuron.2019.05.047

    • PubMed
    • Google Scholar
13. 1. Garbett K
    2. Tosun B
    3. Lopez JM
    4. Smith CM
    5. Honkanen K
    6. Sando RC

    (2024) Synaptic Gα12/13 signaling establishes hippocampal PV inhibitory circuits

    PNAS 121:e2407828121.

    https://doi.org/10.1073/pnas.2407828121

    • PubMed
    • Google Scholar
14. 1. Green MV
    2. Pengo T
    3. Raybuck JD
    4. Naqvi T
    5. McMullan HM
    6. Hawkinson JE
    7. Marron Fernandez de Velasco E
    8. Muntean BS
    9. Martemyanov KA
    10. Satterfield R
    11. Young SM Jr
    12. Thayer SA

    (2019) Automated live-cell imaging of synapses in rat and human neuronal cultures

    Frontiers in Cellular Neuroscience 13:467.

    https://doi.org/10.3389/fncel.2019.00467

    • PubMed
    • Google Scholar
15. 1. Grimm JB
    2. Muthusamy AK
    3. Liang Y
    4. Brown TA
    5. Lemon WC
    6. Patel R
    7. Lu R
    8. Macklin JJ
    9. Keller PJ
    10. Ji N
    11. Lavis LD

    (2017) A general method to fine-tune fluorophores for live-cell and in vivo imaging

    Nature Methods 14:987–994.

    https://doi.org/10.1038/nmeth.4403

    • PubMed
    • Google Scholar
16. 1. Gross GG
    2. Junge JA
    3. Mora RJ
    4. Kwon H-B
    5. Olson CA
    6. Takahashi TT
    7. Liman ER
    8. Ellis-Davies GCR
    9. McGee AW
    10. Sabatini BL
    11. Roberts RW
    12. Arnold DB

    (2013) Recombinant probes for visualizing endogenous synaptic proteins in living neurons

    Neuron 78:971–985.

    https://doi.org/10.1016/j.neuron.2013.04.017

    • PubMed
    • Google Scholar
17. 1. Hao YA
    2. Lee S
    3. Roth RH
    4. Natale S
    5. Gomez L
    6. Taxidis J
    7. O’Neill PS
    8. Villette V
    9. Bradley J
    10. Wang Z
    11. Jiang D
    12. Zhang G
    13. Sheng M
    14. Lu D
    15. Boyden E
    16. Delvendahl I
    17. Golshani P
    18. Wernig M
    19. Feldman DE
    20. Ji N
    21. Ding J
    22. Südhof TC
    23. Clandinin TR
    24. Lin MZ

    (2024) A fast and responsive voltage indicator with enhanced sensitivity for unitary synaptic events

    Neuron 112:3680–3696.

    https://doi.org/10.1016/j.neuron.2024.08.019

    • PubMed
    • Google Scholar
18. 1. Holtmaat AJGD
    2. Trachtenberg JT
    3. Wilbrecht L
    4. Shepherd GM
    5. Zhang X
    6. Knott GW
    7. Svoboda K

    (2005) Transient and persistent dendritic spines in the neocortex in vivo

    Neuron 45:279–291.

    https://doi.org/10.1016/j.neuron.2005.01.003

    • PubMed
    • Google Scholar
19. 1. Kaech S
    2. Huang CF
    3. Banker G

    (2012) Long-term time-lapse imaging of developing hippocampal neurons in culture

    Cold Spring Harbor Protocols 2012:335–339.

    https://doi.org/10.1101/pdb.prot068239

    • PubMed
    • Google Scholar
20. Preprint

    1. Kumar S
    2. Coste GI
    3. Premathilaka D
    4. Huganir RL
    5. Graves AR
    6. Charles AS
    7. Miller MI

    (2025) A space-time hidden markov model for in vivo nanoscale synaptic plasticity tracking

    bioRxiv.

    https://doi.org/10.1101/2025.10.10.681691

    • Google Scholar
21. 1. Kwon HB
    2. Sabatini BL

    (2011) Glutamate induces de novo growth of functional spines in developing cortex

    Nature 474:100–104.

    https://doi.org/10.1038/nature09986

    • PubMed
    • Google Scholar
22. 1. Leão RN
    2. Mikulovic S
    3. Leão KE
    4. Munguba H
    5. Gezelius H
    6. Enjin A
    7. Patra K
    8. Eriksson A
    9. Loew LM
    10. Tort ABL
    11. Kullander K

    (2012) OLM interneurons differentially modulate CA3 and entorhinal inputs to hippocampal CA1 neurons

    Nature Neuroscience 15:1524–1530.

    https://doi.org/10.1038/nn.3235

    • PubMed
    • Google Scholar
23. 1. Lichtman JW
    2. Fraser SE

    (2001) The neuronal naturalist: watching neurons in their native habitat

    Nature Neuroscience 4 Suppl:1215–1220.

    https://doi.org/10.1038/nn754

    • PubMed
    • Google Scholar
24. 1. Liu G

    (2004) Local structural balance and functional interaction of excitatory and inhibitory synapses in hippocampal dendrites

    Nature Neuroscience 7:373–379.

    https://doi.org/10.1038/nn1206

    • PubMed
    • Google Scholar
25. 1. Los GV
    2. Encell LP
    3. McDougall MG
    4. Hartzell DD
    5. Karassina N
    6. Zimprich C
    7. Wood MG
    8. Learish R
    9. Ohana RF
    10. Urh M
    11. Simpson D
    12. Mendez J
    13. Zimmerman K
    14. Otto P
    15. Vidugiris G
    16. Zhu J
    17. Darzins A
    18. Klaubert DH
    19. Bulleit RF
    20. Wood KV

    (2008) HaloTag: a novel protein labeling technology for cell imaging and protein analysis

    ACS Chemical Biology 3:373–382.

    https://doi.org/10.1021/cb800025k

    • PubMed
    • Google Scholar
26. Preprint

    1. Metzbower SR
    2. Dharmasri PA
    3. Levy AD
    4. Anderson MC
    5. Blanpied TA

    (2023) Distinct SAP102 and PSD-95 nano-organization defines multiple types of synaptic scaffold protein domains at single synapses

    bioRxiv.

    https://doi.org/10.1101/2023.09.12.557372

    • Google Scholar
27. 1. Nichol H
    2. Amilhon B
    3. Manseau F
    4. Badrinarayanan S
    5. Williams S

    (2018) Electrophysiological and morphological characterization of Chrna2 cells in the subiculum and CA1 of the hippocampus: an optogenetic investigation

    Frontiers in Cellular Neuroscience 12:32.

    https://doi.org/10.3389/fncel.2018.00032

    • PubMed
    • Google Scholar
28. 1. Nishiyama J
    2. Mikuni T
    3. Yasuda R

    (2017) Virus-mediated genome editing via homology-directed repair in mitotic and postmitotic cells in mammalian brain

    Neuron 96:755–768.

    https://doi.org/10.1016/j.neuron.2017.10.004

    • PubMed
    • Google Scholar
29. 1. Okabe S

    (2013) Fluorescence imaging of synapse formation and remodeling

    Microscopy 62:51–62.

    https://doi.org/10.1093/jmicro/dfs083

    • PubMed
    • Google Scholar
30. 1. Özel MN
    2. Langen M
    3. Hassan BA
    4. Hiesinger PR

    (2015) Filopodial dynamics and growth cone stabilization in Drosophila visual circuit development

    eLife 4:e10721.

    https://doi.org/10.7554/eLife.10721

    • PubMed
    • Google Scholar
31. 1. Özel MN
    2. Kulkarni A
    3. Hasan A
    4. Brummer J
    5. Moldenhauer M
    6. Daumann IM
    7. Wolfenberg H
    8. Dercksen VJ
    9. Kiral FR
    10. Weiser M
    11. Prohaska S
    12. von Kleist M
    13. Hiesinger PR

    (2019) Serial synapse formation through filopodial competition for synaptic seeding factors

    Developmental Cell 50:447–461.

    https://doi.org/10.1016/j.devcel.2019.06.014

    • PubMed
    • Google Scholar
32. 1. Rimbault C
    2. Breillat C
    3. Compans B
    4. Toulmé E
    5. Vicente FN
    6. Fernandez-Monreal M
    7. Mascalchi P
    8. Genuer C
    9. Puente-Muñoz V
    10. Gauthereau I
    11. Hosy E
    12. Claverol S
    13. Giannone G
    14. Chamma I
    15. Mackereth CD
    16. Poujol C
    17. Choquet D
    18. Sainlos M

    (2024) Engineering paralog-specific PSD-95 recombinant binders as minimally interfering multimodal probes for advanced imaging techniques

    eLife 13:e69620.

    https://doi.org/10.7554/eLife.69620

    • PubMed
    • Google Scholar
33. 1. Rubenstein JLR
    2. Merzenich MM

    (2003) Model of autism: increased ratio of excitation/inhibition in key neural systems

    Genes, Brain, and Behavior 2:255–267.

    https://doi.org/10.1034/j.1601-183x.2003.00037.x

    • PubMed
    • Google Scholar
34. 1. Sando R
    2. Ho ML
    3. Liu X
    4. Südhof TC

    (2022) Engineered synaptic tools reveal localized cAMP signaling in synapse assembly

    The Journal of Cell Biology 221:e202109111.

    https://doi.org/10.1083/jcb.202109111

    • PubMed
    • Google Scholar
35. 1. Son J-H
    2. Keefe MD
    3. Stevenson TJ
    4. Barrios JP
    5. Anjewierden S
    6. Newton JB
    7. Douglass AD
    8. Bonkowsky JL

    (2016) Transgenic FingRs for live mapping of synaptic dynamics in genetically-defined neurons

    Scientific Reports 6:18734.

    https://doi.org/10.1038/srep18734

    • PubMed
    • Google Scholar
36. 1. Son S
    2. Nagahama K
    3. Lee J
    4. Jung K
    5. Kwak C
    6. Kim J
    7. Noh YW
    8. Kim E
    9. Lee S
    10. Kwon H-B
    11. Heo WD

    (2024) Real-time visualization of structural dynamics of synapses in live cells in vivo

    Nature Methods 21:353–360.

    https://doi.org/10.1038/s41592-023-02122-4

    • PubMed
    • Google Scholar
37. 1. Südhof TC

    (2017) Molecular neuroscience in the 21^st century: a personal perspective

    Neuron 96:536–541.

    https://doi.org/10.1016/j.neuron.2017.10.005

    • PubMed
    • Google Scholar
38. 1. Südhof TC

    (2018) Towards an understanding of synapse formation

    Neuron 100:276–293.

    https://doi.org/10.1016/j.neuron.2018.09.040

    • PubMed
    • Google Scholar
39. 1. Südhof TC

    (2021) The cell biology of synapse formation

    The Journal of Cell Biology 220:e202103052.

    https://doi.org/10.1083/jcb.202103052

    • PubMed
    • Google Scholar
40. 1. Trimmer JS

    (2022) Genetically encoded intrabodies as high-precision tools to visualize and manipulate neuronal function

    Seminars in Cell & Developmental Biology 126:117–124.

    https://doi.org/10.1016/j.semcdb.2021.11.004

    • Google Scholar
41. 1. Tsetsenis T
    2. Boucard AA
    3. Araç D
    4. Brunger AT
    5. Südhof TC

    (2014) Direct visualization of trans-synaptic neurexin-neuroligin interactions during synapse formation

    The Journal of Neuroscience 34:15083–15096.

    https://doi.org/10.1523/JNEUROSCI.0348-14.2014

    • PubMed
    • Google Scholar
42. 1. Turrigiano GG
    2. Leslie KR
    3. Desai NS
    4. Rutherford LC
    5. Nelson SB

    (1998) Activity-dependent scaling of quantal amplitude in neocortical neurons

    Nature 391:892–896.

    https://doi.org/10.1038/36103

    • PubMed
    • Google Scholar
43. 1. Turrigiano GG
    2. Nelson SB

    (2004) Homeostatic plasticity in the developing nervous system

    Nature Reviews. Neuroscience 5:97–107.

    https://doi.org/10.1038/nrn1327

    • PubMed
    • Google Scholar
44. Preprint

    1. Woods H
    2. Günar Ü
    3. Gillard SC
    4. Notman B
    5. Yuan K
    6. Dominic D
    7. Robson E
    8. Varga G
    9. Komiyama NH
    10. Qiu Z
    11. Sengpiel F
    12. Grant SGN

    (2025) Experiences are encoded by brainwide reprogramming of synaptome architecture

    bioRxiv.

    https://doi.org/10.64898/2025.12.12.693919

    • Google Scholar
45. 1. Xu T
    2. Yu X
    3. Perlik AJ
    4. Tobin WF
    5. Zweig JA
    6. Tennant K
    7. Jones T
    8. Zuo Y

    (2009) Rapid formation and selective stabilization of synapses for enduring motor memories

    Nature 462:915–919.

    https://doi.org/10.1038/nature08389

    • PubMed
    • Google Scholar
46. 1. Yamagata M
    2. Sanes JR

    (2012) Transgenic strategy for identifying synaptic connections in mice by fluorescence complementation (GRASP)

    Frontiers in Molecular Neuroscience 5:18.

    https://doi.org/10.3389/fnmol.2012.00018

    • PubMed
    • Google Scholar
47. 1. Yang G
    2. Pan F
    3. Gan WB

    (2009) Stably maintained dendritic spines are associated with lifelong memories

    Nature 462:920–924.

    https://doi.org/10.1038/nature08577

    • PubMed
    • Google Scholar
48. Conference

    1. Zhang L
    2. Yuan L
    3. Nevatia R

    (2008) Global data association for multi-object tracking using network flows

    2008 IEEE Conference on Computer Vision and Pattern Recognition. pp. 1–8.

    https://doi.org/10.1109/CVPR.2008.4587584

    • Google Scholar

Author details

1. Krassimira A Garbett

   Department of Pharmacology, Vanderbilt Brain Institute, Vanderbilt University, Nashville, United States

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   Contributed equally with

   James P Allen

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-3969-3919

2. James P Allen

   Department of Pharmacology, Vanderbilt Brain Institute, Vanderbilt University, Nashville, United States

   Contribution

   Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   Contributed equally with

   Krassimira A Garbett

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2124-6881

3. Jaybree M Lopez

   Department of Pharmacology, Vanderbilt Brain Institute, Vanderbilt University, Nashville, United States

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Visualization, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

4. Cassandra M Smith

   Department of Pharmacology, Vanderbilt Brain Institute, Vanderbilt University, Nashville, United States

   Contribution

   Data curation, Formal analysis, Validation, Investigation, Visualization, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

5. Richard C Sando

   Department of Pharmacology, Vanderbilt Brain Institute, Vanderbilt University, Nashville, United States

   Contribution

   Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   richard.sando@vanderbilt.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1797-2346

• 721

  views

• 27

  downloads

• 0

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.