Synapses are highly specialized subcellular compartments of neurons and represent the fundamental computational units for neurotransmission. Each neuron connects to thousands of others via asymmetric intercellular junctions composed of presynapses and postsynapses, facilitating continuous signal transmission. In the human brain, approximately 150 trillion synapses form intricate neural circuits across various brain regions. These synapses undergo dynamic morphological and functional plasticity throughout life, influenced by environmental stimuli such as sensory experience and behavioral activity besides to the genetic programs. This remarkable plasticity underpins fundamental brain functions such as learning, memory, and emotion by supporting adaptive modifications in synaptic connectivity and strength. Structurally, synapses are composed by synaptic vesicles, the presynaptic active zone, the synaptic cleft, and the postsynaptic density (PSD), each compartment drives the above brain functions in a coordinated manner. Notably, thousands of distinct synaptic proteins orchestrate the brain functions and the structural of synapses, and define the discrete synaptic properties in the different brain regions (O’Rourke et al., 2012; Koopmans et al., 2019; Van Oostrum et al., 2023). Also, synaptic proteins include cytoskeletal proteins, receptors, neurotransmitters, adhesion molecules and scaffold proteins (O’Rourke et al., 2012; Koopmans et al., 2019), and this molecular diversity exemplifies how synaptic components are not merely structural but actively shape signaling integration, synapse specification, and plasticity. This functional complexity is made possible by the spatially confined and molecularly compartmentalized organization of synapses, which enables precise and localized biochemical processing. Thus, each synapse operates as a self-regulating biochemical microdomain capable of adaptive computation within neural circuits. Conversely, the dysfunction of synaptic protein networks leads to impairments in synapse number, morphology, and signal transmission. Accumulating evidence suggests that such synaptic dysfunctions, which are often referred to as synaptopathies, are closely associated with neurodevelopmental and psychiatric disorders as well as with the progression of neurodegenerative diseases (Grant, 2012; Lepeta et al., 2016; Hindley et al., 2023). Notably, these abnormalities often manifest in specific brain regions. Although large-scale efforts have led to the identification of over 2,000 distinct synaptic proteins through conventional proteomic approaches (Bayés et al., 2011; Loh et al., 2016; Koopmans et al., 2019), the molecular mechanisms governing synaptic specificity, diversity, and plasticity remain incompletely understood. This is due, in part, to limitations in spatial resolution, cell-type specificity, and the ability to analyze intact neural circuits in vivo.

In recent years, emerging proximity labeling (PL)-based proteomic techniques such as BioID (biotin ligase-based), APEX (ascorbate peroxidase) and HRP (horseradish peroxidase) have made it possible to profile the local proteome of synaptic compartments with high spatial resolution in living tissue (Han et al., 2018; Takano and Soderling, 2021; Ito et al., 2024). These techniques have facilitated the discovery of proteomes associated with specific neuronal populations (Uezu et al., 2016), synaptic clefts (Loh et al., 2016; Takano et al., 2020), and tripartite synapses formed by astrocyte-neuron connections (Takano et al., 2020; Takano and Soderling, 2021). In this short review, we highlight recent advances in spatial synaptic proteomics enabled by proximity labeling (PL) technologies and discuss how these approaches have advanced our understanding of the molecular mechanisms underlying synapse formation, diversity, and function. We further introduce current insights into the pathophysiology of synapse-related neurological disorders uncovered through PL-based studies and outline future directions for the therapeutic application of these technologies. By enabling precise profiling of synapse-specific molecular networks, PL-based proteomic approaches offer novel insights into brain function and hold considerable promise for the development of targeted diagnostic and therapeutic strategies for synapse-associated disorders.

Traditionally, synaptic proteins have been identified using liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of synaptic vesicles and synaptosomes, purified by differential centrifugation, density-gradient centrifugation, immune-purification and affinity-purification (Fernández et al., 2009; Morciano et al., 2009; Grønborg et al., 2010; Boyken et al., 2013; Wilhelm et al., 2014; Dieterich and Kreutz, 2016; Xu et al., 2021; Kaizuka et al., 2024). While these proteomic approaches have proven valuable for detecting synaptic proteins enriched in cultured neurons and brain tissues, they are limited by low spatial resolution and contamination from heterogeneous mixtures derived from multiple synapse types. These limitations hinder the ability to resolve the molecular characteristics of specific cell types, synapse subtypes, synaptic clefts, and tripartite synapses.

In recent years, PL technologies such as BioID, APEX, and HRP have emerged as powerful biochemical tools for spatially resolved synaptic proteomics (Ito et al., 2024; Table 1). These approaches rely on enzyme-mediated biotinylation of proteins located in the immediate vicinity of a target protein fused to a biotin ligase or peroxidase. Biotinylated proteins are subsequently purified using streptavidin, NeutrAvidin, anti-biotin antibody and Tamavidin 2-REV-coated beads, followed by identification using LC-MS/MS (Figure 1A). BioID, the first biotin ligase-based PL method, uses a mutant Escherichia coli biotin ligase (BirA*-R118G) that generates reactive biotin (biotinoyl-5’-AMP) and biotinylates lysine residues of nearby proteins in the presence of biotin (typically within ∼10–20 nm) (Han et al., 2018; Ito et al., 2024; Table 1). Since the original development of BioID, a broad spectrum of proximity-labeling ligases has been engineered to enhance properties such as molecular size, catalytic efficiency, labeling kinetics, and specificity under various physiological and pathological condition. BioID2 is a truncated variant of BioID that retains proximality labeling capability while offering improved efficiency due to its smaller size (Kim et al., 2016). BASU enables more than 1,000 times faster kinetics and more than 30 times increased signal-to-noise ratio over the prior BioID (Ramanathan et al., 2018). TurboID and miniTurbo show much greater efficiency than BioID and BioID2, and biotinylate proteins for 10 min (Branon et al., 2018). Labeling speed of TurboID (∼1 h) is much faster than BioID (∼12–16 h), but TurboID has strong biotinylation activity, which may cause non-specific labeling (Table 1). Split-BioID is splitting BirA into two parts, fusing each fragment with a different protein, and reactivating the BirA enzyme when the complex is formed (De Munter et al., 2017; Schopp et al., 2017; Cho et al., 2020; Takano et al., 2020). The microID, a truncation variant of BioID, is a small-sized biotin ligase and shows efficient Biotinylation at short labeling times (Kubitz et al., 2022). The ultraID is also the directed evolution of microID (Kubitz et al., 2022). MicroID2 is a modified BioID2 and enables lower background labeling than TurboID (Johnson et al., 2022). AirID was engineered by in silico design and shows low ability to biotinylate proteins non-specifically (Kido et al., 2020; Table 1). In this way, researchers can select appropriate tools based on specific experimental needs. In contrast, APEX and HRP are peroxidase-based PL methods that catalyze biotinylation of tyrosine residues using reactive radicals generated from biotin-phenol and hydrogen peroxide (H₂O₂) (Table 1). HRP is mainly used for biotinylation of extracellular proteins, because HRP requires intramolecular disulfide bonds, but disulfide bond formation is basically difficult inside cells. On the other hand, APEX can use intracellular labeling because it does not require disulfide bonds. These two peroxidase-based PL methods enable rapid (seconds to minutes) and extensive labeling (APEX: 20 nm, HRP: 200–300 nm) than BioID (∼10 nm). However, the important point to note is that these approaches are mainly restricted to in vitro or ex vivo applications due to the cytotoxicity of H₂O₂ (Figure 1A and Table 1).

Notably, BioID approaches have enabled the mapping of synaptic proteins in the brain (Uezu et al., 2016). This in vivo BioID (iBioID) technique, in combination with genetic tools such as adeno-associated virus (AAV) vectors and transgenic mice, permits targeted profiling of synaptic proteomes in specific circuits and cell types (Uezu et al., 2016; Ito et al., 2024). However, unlike in vitro, it is necessary to administer biotin continuously for several days. More recently, Cho et al. (2025) introduced a membrane-tethered version of HRP (HRP-TM) that utilizes endogenously generated H₂O₂ for cell surface biotinylation in vitro, offering potential for application without the need for exogenous H₂O₂. In another advancement, Zhang et al. (2025) reported TyroID, a novel tyrosinase-based PL technique, that enables non-toxic labeling of various nucleophilic residues both in vitro and in vivo, using reactive o-quinone intermediates derived from phenol-based probes such as alkyne-phenol or biotin-phenol (Zhang et al., 2025). These tools have not yet been applied to spatial synaptic proteomics, but they are expected to be novel in vivo PL tools that compensate for the poor temporal resolution of iBioID. Collectively, PL technologies are rapidly evolving to support in vivo applications and, when combined with virus-based chemogenetic tools such as AAV and transgenic mouse systems, offer a powerful platform for dissecting the molecular mechanisms of synapse formation and function at high spatial resolution.

Proximity labeling approaches have significantly advanced the field of synaptic proteomics by enabling molecular profiling of subcellular compartments with high spatial precision. Applications of BioID, APEX, and HRP have facilitated the identification of proteomes in various synaptic environments, including excitatory and inhibitory synapses (Uezu et al., 2016), cholinergic (Rosenthal et al., 2025) and electrical synapses (Tetenborg et al., 2025), as well as glial interfaces such as tripartite synapses (Takano et al., 2020). These studies have expanded the catalog of synaptic proteins and provided valuable insights into how synapses are assembled, maintained, and modified in both healthy and diseased brains (Figure 1 and Table 2).

Nevertheless, several technical and conceptual challenges must be addressed to fully harness the potential of PL-based approaches. One major limitation is the low temporal resolution of current labeling systems. BioID-based methods typically require several days of biotin supplementation to achieve effective labeling in vivo, which makes it difficult to capture rapid or transient protein interactions that occur in response to neuronal activity or environmental changes. In contrast, APEX and HRP allow for much faster labeling (≤1 min) but depend on hydrogen peroxide, which is cytotoxic and unsuitable for applications in intact brain tissue. Although new methods such as HRP-TM, which utilizes endogenous hydrogen peroxide (Cho et al., 2025), and TyroID, which employs non-toxic o-quinone chemistry (Zhang et al., 2025), offer promising alternatives, their specificity and applicability in complex brain tissue remain to be fully validated. Further in vivo studies are expected to assess their performance under physiological and pathological conditions, particularly in identifying activity-dependent or circuit-specific proteomic changes within intact neural networks. Another important issue is the limited spatial resolution of current labeling methods. In the brain, synapses are highly compact structures where proteins from presynaptic neurons, postsynaptic neurons, and surrounding glial cells are densely intermingled. Although, TurboID can label proteins within a radius of approximately 10 nm, its biotynilation activity is so potent compared to BioID (Branon et al., 2018). Therefore, it may biotinylate not only the intended molecular targets but also nearby proteins from adjacent compartments. This overlap makes it difficult to determine exactly where the labeled proteins are localized within the synapse. To overcome this limitation, recent methods such as Split-HRP (Martell et al., 2016) and Split-TurboID (Cho et al., 2020; Takano et al., 2020) have been developed. This technique divides the labeling enzyme into two inactive fragments that only reconstitute and become active when two different cell types are in direct contact. Using this strategy, we successfully identified molecules such as NRCAM that localize specifically to astrocyte-neuron interfaces and play a critical role in organizing inhibitory synapses (Takano et al., 2020). These findings highlight how cell-contact-dependent labeling can improve spatial precision and uncover new mechanisms of synaptic regulation.

Quantitative interpretation of PL data also remains a challenge. Currently, there is no widely accepted standard for normalization, statistical comparison, or integration of PL proteomes across different developmental stages or disease models. Combining PL-based proteomic data with complementary approaches such as single-cell transcriptomics (Yao et al., 2023; Zhang et al., 2023), spatial transcriptomics (Lein et al., 2017; Yuan et al., 2025), and high-resolution imaging (Newman et al., 2022; Unterauer et al., 2024) will likely be necessary to interpret the data in a biologically meaningful context. For instance, a recently published single-cell mass cytometry-based atlas of the developing mouse brain provides a valuable resource for anchoring synaptic proteomic data within a broader cellular and developmental framework (Van Deusen et al., 2025). The combined analysis of this advanced technology and PL approaches could also offer a key resource for future novel “single-synapse proteome” research field. Additionally, PL methods have provided important insights into the molecular mechanisms of neurological and psychiatric disorders (Table 2). Recent studies using disease models or patient-derived mutations have shown how alterations in protein–protein interactions can impair synaptic signaling and lead to behavioral and cognitive deficits. For example, disrupted interactions between Densin-180 and PP1α (Willim et al., 2024) as well as changes in the Scn2a-associated proteome (Gao et al., 2024) have been linked to neurodevelopmental disorders. These findings highlight the utility of PL techniques for mechanistic investigations and therapeutic target identification besides for mapping molecular discovery (Figure 1B). In summary, proximity labeling-based synaptic proteomics represents a powerful platform for investigating the molecular logic of synapse formation, function, and dysfunction. Also, synaptic proteomics has greatly advanced our understanding of molecular diversity within synapses, and revealing a number of unknown molecular mechanisms involved in this diversity may provide cues to decoding the intricate brain functions induced by diverse neural circuits. Future improvements in the temporal control, spatial accuracy, and quantitative robustness of these technologies will be crucial for advancing both basic neuroscience and clinical applications. By integrating molecular, cellular, and circuit-level information, PL approaches have the potential to reshape our understanding of the brain and inform the development of targeted therapies for complex brain disorders.