Creative approaches using proximity labeling to gain new biological insights

Abstract

At its most fundamental level, life is a collection of synchronized cellular processes driven by interactions among biomolecules. Proximity labeling has emerged as a powerful technique to capture these interactions in native settings, revealing previously unexplored elements of biology. This review highlights recent developments in proximity labeling, focusing on methods that push the fundamental technologies beyond the classic bait-prey paradigm, such as RNA-protein interactions, ligand/small molecule-protein interactions, cell surface protein interactions, and subcellular protein trafficking. The advancement of proximity labeling methods to address different biological problems will accelerate our understanding of the complex biological systems that make up life.

Proximity labeling to study biomolecular interactions

Interactions among biomolecules govern all cellular processes [1]. In particular, protein-protein interactions (PPI) and protein-nucleic acid interactions (PNI) facilitate fundamental processes such as signal transduction and gene expression. Our understanding of these interactions has laid the foundations for modern biology and drug discovery [1,2]. The most prevalent detection method for these interactions remains co-immunoprecipitation (co-IP) and its derivatives [3]. These techniques use an antibody to enrich a protein of interest (POI) along with its robust binding partners from cell lysate [4]. While transformative, co-IP suffers from several well-established drawbacks, such as detecting transient and weak interactions, particularly when harsh conditions are required for cell lysis and protein solubilization [4]. The dependency on effective and selective antibodies, which are not always available, further hinders this approach. Over the last decade, numerous proximity labeling (PL, see Glossary) methods have emerged to address these limitations [5].

In a prototypical PL experiment, the POI is fused to a catalyst that, upon delivery of an external stimulus, such as biotin, hydrogen peroxide, or light, can convert nearby small molecule probes into reactive intermediates that crosslink with nearby proteins [5]. These methods are usually carried out in live cells, and the tagged proteins can be enriched and analyzed via chemoproteomics, generating a list of candidate interacting proteins (Figure 1, top). A pioneering example of PL in mammalian cells was described by Roux et al. in 2012. BirA* was engineered from an E. coli biotin ligase to generate activated esters of biotin that readily react with Lys residues on nearby proteins [6]. In the following years, a slew of alternative PL platforms were described, including APEX, TurboID, and μMap [7–9]. While these tools have significant utility in the appropriate settings, they are primarily employed in the same paradigm—i.e., measuring the protein interactome around a single POI (Figure 1). This review will discuss several emerging areas of PL research that creatively repurpose the same tools to gain new biological insights. Specifically, we will cover recent work that uses PL to study RNA (Figure 1A), how small molecules interact with the cell (Figure 1B), the cell surface (Figure 1C), and the taxis of biomolecules in the cell.

Figure 1. Classic and creative uses of proximity labeling.

Top. In traditional proximity labeling methods, a catalyst is localized to a protein of interest and generates reactive probes that tag nearby proteins. These proteins are enriched and digested for analysis and subsequent identification by mass spectrometry. Bottom. (A) A proximity labeling catalyst is used to label RNA molecules. (B) A catalyst is localized to the cell membrane for subsequent labeling of membrane-bound proteins. (C) A small molecule is used to localize a proximity labeling catalyst to a protein of interest, which then labels nearby biomolecules. (D) Two orthogonal proximity labeling enzymes are deployed in different cellular locations to measure intracellular protein trafficking or in different cell types to determine intercellular protein trafficking.

Proximity labeling to study RNA-protein interactions and RNA localization

While 2% of the human genome encodes for proteins, between 70% and 80% encodes for RNA—most of which is underexplored [10]. In recent years, the prospects of using and targeting RNA for therapeutic benefit have had a meteoric rise [11]. This necessitates the creation of PL methods to investigate RNA-protein interactions within the cell, especially since fewer tools are available to study RNAs compared to proteins.

Labeling RNA using APEX

Early strategies to probe native RNA-protein interactions couple classical PL with RNA immunoprecipitation (RIP) and cross-linking and immunoprecipitation (CLIP) [12,13]. Specifically, the Ting lab at Stanford developed APEX-RIP to map RNA molecules to their native cellular regions [14,15]. In this technique, APEX2 enzymes are genetically targeted to specific cellular compartments where they catalyze PL. These peroxidases create short-lived phenoxy radicals from H₂O₂ and a biotin-bearing phenol that crosslinks with surface-exposed tyrosine residues on nearby proteins [15]. Subsequent formaldehyde crosslinking leads to RNA molecules bound to biotinylated proteins proximal to the POI. Enrichment of tagged proteins followed by elution of crosslinked RNA identifies RNAs proximal to the POI.

Another foundational method developed by the Ting lab is APEX-seq (Figure 2A) [16]. In contrast to APEX-RIP, APEX-seq directly labels RNA for subsequent enrichment, obviating the requirement to capture a specific RNA-protein interaction [16]. As the phenoxy radicals generated by APEX2 readily insert into guanine nucleobases, the authors demonstrated that it is possible to isolate and sequence biotinylated RNAs following PL. In this study, APEX2 expression was confined to the outer mitochondrial membrane, nuclear lamina, and other organelles, which provided spatiotemporal information about the organization of RNAs in live cells [16]. Recently, the sensitivity of APEX-seq was improved through metabolic incorporation of electron-rich ribonucleoside analogs, namely 6-thioguanosine (s⁶G) and 4-thiouridine (s⁴U), into nascent RNAs [17]. The heightened reactivity between the modified ribonucleosides and the APEX-generated phenoxyl radicals increases labeling sensitivity, enabling the detection of low-abundance transcripts [16].

Figure 2. RNA-based proximity labeling.

(A) In APEX-seq, APEX is targeted to an organelle where it directs biotinylation to nearby RNA molecules. (B) In CAP-seq and HALO-seq, a photosensitizer generates singlet oxygen, which oxidizes guanine and renders it reactive to propargylamine probes. This allows for the covalent capture of oxidized transcripts. (C) The RNA binding protein dCas13 is used to localize a proximity labeling catalyst to RNA molecules. (D) O-MAP and HyPro methods utilize antisense oligonucleotide probes that anneal to the RNA transcript of interest and serve as an attachment point for peroxidase-based PL enzymes.

Labeling RNA by singlet oxygen photosensitizers

To label RNA in situ, the alternative RNA PL techniques, CAP-seq [18] and HALO-seq [19], employ singlet oxygen generated by the flavoprotein, miniSOG, and by a dibromofluoroscein sensitizer, respectively (Figure 2B). In both cases, singlet oxygen is generated by photoirradiation of a protein-localized sensitizer, which oxidizes guanines in proximity to the POI. These oxidized RNA nucleobases are rendered electrophilic and can react with propargylamine probes, thereby introducing an alkyne handle into RNA for covalent capture by click chemistry. The precise spatial resolution of HALO-seq was highlighted by Engel et al. in 2022 [19]. This study localized HaloTag domains to the nucleolus and chromatin by expressing Fibrillarin-HaloTag and H2B-HaloTag fusion proteins, respectively. The authors found that 7SL and SNORA68 RNAs—transcripts known to localize to the nucleolus—were enriched in the Fibrillarin pulldown but depleted in the H2B pulldown. Furthermore, the authors also showed that unspliced pre-mRNAs were more enriched in the H2B pulldown, implicating the nucleolus as a site of RNA processing [19].

RNA-directed proximity labeling

Understanding how specific RNAs regulate cellular processes (Figure 2C) is a significant challenge that has been difficult to address for proximity labeling. As most PL methods are deployed via expressed fusion proteins (e.g., POI-APEX, POI-BioID), localization to a single RNA requires an alternate approach. An early method for localizing PL to RNA molecules was developed in 2020 by Han et al., who used catalytically inactive Cas13 (dCas13) fused to an APEX2 enzyme to direct PL to human telomerase RNA [20]. Using gRNA, dCas13 can be guided to endogenous transcripts with high specificity. Several other groups have also reported similar approaches using alternate PL enzymes fused to RNA-binding proteins (RBPs) [21–23]. Yet, these approaches are limited by the requirement of highly abundant or exogenously expressed target transcripts.

Recently, the use of functionalized antisense oligonucleotide probes to anneal to an RNA molecule of interest and display a handle for localizing a PL catalyst has gained recognition in the field [24,25]. This general strategy has been employed in two methods termed Hybridization-Proximity sequencing (HyPro-seq) [24] and oligonucleotide-mediated proximity-interactome mapping (O-MAP) (Figure 2D) [25]. The antisense oligonucleotide probes deployed in these methods anneal to an RNA molecule and present a secondary binding site, either a digoxigenin molecule (DIG; HyPro-seq) or a “landing pad” sequence (O-MAP). These points of attachment enable the localization of a peroxidase to the RNA molecule of interest, which, following treatment with biotin-phenol and hydrogen peroxide, generates reactive phenoxy radicals that crosslink with interacting proteins, effectively generating a network of protein/RNA interactions [24,25]. While O-MAP and HyPro-seq improve upon dCas13 and RBP guided PL methods and effectively target low-abundance RNAs, these methods are performed on crosslinked cells, presenting a challenge for protein identification during mass spectrometry [26,27]. Together with established PL-seq methods, these RNA-directed PL strategies will deepen our understanding of how RNAs interact with proteins within the cell.

Proximity labeling to study ligand-biomolecule interactions

Target misidentification is a common cause of clinical failure, leading to adverse patient outcomes and costing the pharmaceutical industry billions of dollars [28–30]. Therefore, new strategies to determine the interactions between a given small molecule, such as a therapeutic drug (ligand) of interest, and its protein and RNA targets are in great demand. Photoaffinity labeling (PAL) is the most employed technique to study such interactions [31]. In this approach, the ligand of interest is appended to a molecular handle (e.g., biotin) and a photoreactive moiety (e.g., diazirine) that crosslinks to the target protein upon UV irradiation, enabling enrichment and identification of the target. However, PAL approaches often suffer from low signal-to-noise ratios and cumbersome synthetic manipulations to achieve the optimal binding pose for efficient labeling [31]. Ligand-directed PL methods have been developed to address some of the limitations associated with PAL.

Ligand-directed proximity labeling

In collaboration with scientists at Merck, the MacMillan group at Princeton deployed their photocatalytic proximity labeling method (μMap) via small molecule ligands tethered to a cell-permeable iridium (Ir) catalyst (Figure 3A) [32]. Upon photoexcitation of the ligand-localized catalyst at 450 nm, diazirine-biotin (Dz-Bt) probes are activated through Dexter energy transfer, which promotes their decomposition into carbene species [9]. The high reactivity of carbenes makes μMap particularly advantageous; carbene probes can crosslink with all amino acids, eliminating sequence-specific labeling bias exhibited by other PL methods. Furthermore, water efficiently quenches carbenes, creating a tether-limited labeling radius [33]. The authors demonstrated the generality of μMap by performing target identification on ligands that bind epigenetic regulators, cytoskeletal components, kinases, and GPCRs [32].

Figure 3. Small-molecule-based proximity labeling.

(A) A proximity labeling catalyst is conjugated to a small molecule ligand of a protein of interest. This design is used in μMap and PROCID. In the former method, an iridium catalyst is recruited to a target protein. In the latter method, the small molecule is displayed by a HaloTag-TurboID fusion protein, which recruits native protein binders for labeling. (B) A small molecule disrupts the interactome of a target protein, which is detected by a nearby proximity labeling catalyst. (Left) A molecular glue recruits’ interactors to a protein of interest, which are then labeled by a fused PL enzyme. (Right) A small molecule disrupts binding of epigenetic proteins to chromatin which is detected by μMap PL.

A second ligand-directed PL strategy was reported by the Rhee group [34] which utilizes chimeric bifunctional molecules to recruit a TurboID fusion protein to a ligand of interest. The method, named proximity-based compound-binding protein identification (PROCID) [34] localizes a HaloTag-TurboID fusion protein to the ligand by appending a hexyl chloride tag which forms a covalent bond with the HaloTag protein (Figure 3A, right). This method directly brings the ligand and its associated interactome into proximity of the labeling enzyme, which, upon treatment with biotin, labels the protein targets of the small molecule ligand. For example, the authors examined the interactome of the kinase inhibitor, dasatinib, in K562 cells, where they demonstrate the effectiveness of PROCID by its capture of known targets Abl, CSK, and BTK, and a chromatin remodeler previously unannotated as a target, SMARCA2 [34,35]. Recently, Tao et al. has applied a similar approach, termed BioTAC, in which they use bifunctional molecules to recruit a chimeric miniTurbo-FKBP12^F36V proximity labeling enzyme to the intracellular target(s) of small molecules [36]. These publications illustrate the increasing adoption and effectiveness of chimera-based strategies to characterize a ligand’s interactome and better understand their mechanism of action.

Ligand-disrupted proximity labeling

Another approach to studying the interactome of small molecule ligands is through comparative proteomics. The groups of Rhee [37], MacMillan and Muir [38], and the Sawasaki group [39] did this by fusing the PL catalyst or enzyme to a POI and measuring its interactome in the presence and absence of a small molecule ligand. Comparative proteomics can be used to elucidate how a ligand disrupts or promotes PPIs (Figure 3B). In 2016, the Rhee group reported the first example of using PL to identify the target of the molecular glue rapamycin [37,40]. To this end, BirA* was fused to the rapamycin binding domain of mTOR, FRB, and PL was carried out in live cells with and without rapamycin present. This experiment was designed to identify proteins that enter the FRB interactome only when the ligand is present and it revealed a novel rapamycin-induced FRB-FKBP25 interaction [37].

Building upon this work, the Sawasaki group used a fusion protein with the E3 ligase, cereblon, and a modified BirA* enzyme, AirID, to identify PPIs induced by PROTACs and molecular glues (Figure 3B) [39]. The binding of molecular glues can be profoundly changed with only minor structural changes to the glue [41]. The authors investigated the interactome of the E3 ligase in three cell lines, with and without four different immunomodulatory drugs (IMiDs) that are known to bind cereblon (CRBN), revealing how the ligands influence PPIs with CRBN. In addition to established neosubstrates of CRBN, such as GSPT1 and CK1α, [41] the data suggested that several proteins containing a C2H2 zinc finger (C2H2-ZNF) domain (ZNF536, ZNF687, ZMYM2) may also be IMiD-dependent neosubstrates. Follow-up studies demonstrated the degradation of ZMYM2 is pomalidomide-dependent in HEK293T and IMR32 cell lines, validating the proteomics data. This novel interaction is noteworthy, as ZMYM2 has been characterized as a driver of stem cell leukemia-lymphoma (SCLL) via a chromosomal translocation to FGFR1 [42]. The resulting FGFR1-ZMYM2 fusion protein is responsible for aberrant signaling and subsequent cellular proliferation [42]. The authors showed that pomalidomide degrades this oncogenic fusion protein in a CRBN dependent manner, attenuating STAT1/3 and ERK phosphorylation in HEK293T cells [39]. These data suggest pomalidomide treatment may be a valuable therapeutic strategy for cancers driven by ZMYM2 translocations, such as stem cell leukemia lymphoma.

A subsequent report from MacMillan and Muir presented a photocatalytic strategy to assess the impact of epigenetic compounds on nucleosomal PPIs [38]. This class of ligands has been used in several disease areas, such as neurodegenerative diseases, metabolic diseases, and cancers [43,44]. In this study, a μMap photocatalyst was fused to the C-terminus of histone H2A using ultrafast split intein splicing in biochemically intact nuclei. Subsequently, the authors measured the impact of the BET inhibitor, JQ1, on the interactome of engineered nucleosomes using μMap PL. The data show enrichment of direct JQ1 targets in untreated samples, demonstrating that JQ1 blocks BET proteins from associating with chromatin [38,45].

These methods offer distinct advantages and disadvantages. The chimeric small molecule approaches (ligand-directed proximity labeling) provide direct target identification data (Figure 3A) but require synthetic modifications to the ligand, which is not possible in all cases. Appealingly, the indirect methods (ligand-disrupted proximity labeling) do not demand synthetic manipulation of the ligand (Figure 3B); however, these approaches only measure the changes induced by the ligand at a specific site within the cell and cannot provide a comprehensive assessment of the cellular effects induced by a given ligand. A significant drawback of these approaches is that all require some degree of cellular or chemical engineering, which limits their application in more biologically relevant samples. Overall, using PL to understand the bioactivity of ligands in biology is an emerging field and has shown to be an effective method for target identification across various ligand classes.

Proximity labeling to study the cell surface

The composition of cell surface proteins (CSPs), collectively known as the surfaceome [46], guides fundamental processes, including signal transduction and intercellular communication. Traditional efforts to enrich and identify CSPs have drawbacks. For instance, centrifugation methods, which rely on discrepancies among the sedimentation rates of subcellular species, cannot easily differentiate between proteins embedded in organellar and plasma membranes [47]. IP mass spectrometry is another common approach for detecting PPIs, but the relatively low abundance of CSPs makes them difficult to detect [47,48]. Various innovative PL methods have recently been developed to map the cell surface and its interactions.

Labeling localized with a membrane-bound anchor

In 2020, the Luo group profiled the membrane proteomes of developing and mature Drosophila olfactory projection neurons (PNs) in situ. The authors induced transgenic expression of extracellular-facing horseradish peroxidase (HRP) fused to the transmembrane protein CD2 (Figure 4A) [49]. As with the APEX experiments described above, HRP converts a cell-impermeable biotin phenol (BxxP) derivative to phenoxyl radicals following treatment with H₂O₂ that biotinylate surface exposed tyrosine residues. Compared to the proteome of developing PNs, mature PNs exhibited upregulation of proteins associated with signaling and localization and presented downregulation of circuit-wiring proteins [49].

Figure 4. Modes of localizing proximity labeling to the cell surface.

(A) A proximity labeling enzyme may be directed to the plasma membrane by a fusion to a membrane protein, such as CD2. (B) A proximity labeling enzyme may be localized to membrane bound cholesterol conjugates using sortase protein ligations. (C) Peroxidase enzymes can be localized to glycosylated proteins using glycan binding proteins. (D) A proximity labeling catalyst may be clicked onto a modified sialic acid metabolically incorporated into the cell surface.

The group took a similar approach to develop in situ cell surface proteome extraction by extracellular labeling (iPEEL) [50]. The authors created transgenic mice expressing HRP under Cre recombinase control to localize labeling to cerebellar Purkinje cells. The authors acquired the surfaceomes of developing and mature cells by labeling murine tissues ex vivo. They saw developing Purkinje cells were enriched in post-translational protein processing elements, such as proteolytic enzymes and trafficking proteins [50].

In 2023, the Huang lab localized APEX2 to the cell surface using a cholesterol-PEG2000-CYGGG conjugate that incorporates into the cell membrane and then is covalently linked to recombinant APEX2-LPETG on live cells using an engineered sortase enzyme (Figure 4B) [51]. Here, cells decorated with extracellular CYGGG peptides (attached to membrane bound cholesterol) are incubated with recombinant sortase and recombinant APEX2-LPETG proteins. The sortase enzyme catalyzes the ligation of APEX2 with the membrane bound CYGGG peptide, effectively adorning the cell membrane with APEX2 enzymes [52].

The authors used this method to study the surfaceome of 2D and 3D pancreatic adenocarcinoma cell cultures. Notably, the non-glycosylated proteins OCLN, ANXA5, and ANXA6, which are known to localize to tight junctions and the extracellular matrix, were enriched in 3D spheroids [51]. The Huang group also applied this method to investigate the role of basigin (Bsg) in transporting CSPs [53]. This glycoprotein provides a non-canonical route for proteins to travel from the Golgi apparatus to the surface [54]. Surfaceome PL under Bsg knockdown conditions reduced the enrichment of the monocarboxylate transporters MCT1 and MCT4, suggesting they are brought to the cell surface via a mechanism dependent on Bsg [53].

Labeling localized to cell surface glycans

Beyond using membrane bound PL catalysts, researchers have leveraged the extensive glycosylation of the cell surface to recruit PL machinery and label CSPs. The glycocalyx coats the cell membrane, presenting an ensemble of carbohydrate modifications on proteins and lipids at the extracellular surface [55]. These glycans have provided researchers with a chemical vector for studying the cell surface.

In 2020, the Huang group studied the cell surface by fusing an APEX2 enzyme to the glycan binding protein, galectin-3 (Gal-3) (Figure 4C). Following PL and enrichment of biotinylated proteins, the authors identified novel binding partners for Gal-3, including CD9, CD47, and CD81. Interestingly, CD81 is not known to be glycosylated, and its binding to Gal-3 was maintained in the presence of the Gal-3 inhibitor TD139, suggesting that secondary PPIs mediate its interaction [56].

Building upon this work from Huang, Kirkemo et al. employed membrane-localized PL in Myc-induced prostate cancer cells to uncover proteomic changes at the cell surface and in extracellular vesicles [57]. The group used commercially available HRP conjugates fused to the glycan binding domain of wheat germ-agglutinin (WGA-HRP), which bind to N-acetylglucosamine (GlcNAc) and sialic acid glycans that are presented by cell surface proteins. Compared to normal epithelial prostate cells, the authors identified several membrane proteins significantly enriched in the Myc-induced prostate cancer cell line, such as ANPEP and FN1 [57].

In contrast to these methods where PL is directed to the cell surface by a glycan binding protein, the Lebrilla group developed the Protein Oxidation of Sialic Acid Environment (POSE) technique to study the interactions between sialic acid and cell surface proteins. POSE utilizes metabolic incorporation of N-azidoacetyl sialic acid to display azide-bearing glycans on cell surface proteins [58]. These azido sialic acid groups are then used to anchor an iron catalyst via a DBCO strain-click (Figure 4D). The catalyst enables the conversion of exogenous H₂O₂ to hydrogen peroxide radicals that oxidize nearby methionine and cysteine residues, which can be detected by mass spectrometry.

In 2022, the MacMillan group developed GlycoMap, which also uses metabolic incorporation of N-azidoacetyl sialic acid to attach a DBCO-containing iridium photocatalyst to the cell membrane. Subsequent photocatalyzed PL using a biotin-diazarine probe led to temporally controlled labeling of CSPs (Figure 4D) [59]. Since sialylation is upregulated in cancer, the authors compared the interactomes of sialylated cell surface glycoproteins, or sialomes, in primary cervical cells (PCC) and HeLa cells [60]. The sialome of HeLa cells was enriched in solute carrier proteins involved in transporting ethanolamine, carnitine, and zinc compared to PCC, providing mechanistic insight into the role of hypersialylation in cancer.

Collectively, these methods have led to an increased understanding of the cellular membrane and its constituents by localizing PL enzymes and catalysts stochastically rather than to a specific POI.

Of these methods, those that utilize exogenously incorporated PL enzymes (e.g., cholesterol localization) are particularly powerful in comparison to techniques that rely on chemically functionalized glycans or transgenes, which are subject to differential expression between cells. However, this benefit must be weighed against the experimental advantages of using transgenes or commercial small molecule catalysts, rather than recombinant proteins. Methods for transcellular membrane labeling have also emerged in recent years [61–64] that function through similar principles as those described above, however these are outside the scope of this review.

Proximity labeling to study protein trafficking

Many proteins perform multiple context-dependent functions in different cellular compartments. Methods to study the movement of biomolecules in response to cellular stimuli are limited to techniques that require prior knowledge of the biomolecules, such as live cell imaging. At first glance, PL is not well suited to study protein trafficking, as the prototypical experiment reveals only the proteins near a specific bait, with no way to extrapolate biomolecular movement.

This problem was addressed in a report from Droujinine et al. where the authors expressed the generation 3 BirA* enzyme in a specific organ in Drosophila (e.g., head, leg, muscle) [65]. They then measured biotinylation in a distal organ to assess inter-organ protein trafficking. While powerful, this method necessitated the use of whole organism models, limiting its use for studying protein movement in simpler in vitro models [65].

In 2023, this concept was further developed by the Ting lab, which reported TransitID [66]. This method combines multiple PL techniques to study protein trafficking within and between cells [9,66,67]. In this study, TurboID and APEX2 were localized to different parts of the cell (or other cell types) and stimulated with biotin, then, an hour later to allow for protein trafficking, a phenol alkyne probe and H₂O₂ (Figure 5A). Proteins labeled with both biotin and an alkyne are considered to have traveled between the cellular locations (Figure 5B). The authors demonstrate the utility of this method by revealing the oncoprotein, JUN, moves between the nucleolus and stress granules following stress induction [65]. This is the first evidence that JUN is trafficked to stress granules following cellular stress (caused by arsenite in this case), preventing irreversible aggregation and dysfunction. While only one of the many insights derived from this study are discussed here, TransitID provides a valuable tool for monitoring biomolecular trafficking and illustrates the variety of biological insights that can be gained through the combinatorial application of proximity labeling tools [65].

Figure 5. TransitID.

(A) Two different proximity labeling enzymes were fused to proteins in the nucleolus or stress granules, respectively. After the constructs were transfected, orthogonal labeling experiments were performed to capture proteins that were localized to each compartment after cellular stress induced by arsenite. (B) Proteins in the nucleolus (Group 1) were labeled with biotin-AMP probes generated by localized TurboID. An hour later, proteins in cytosolic stress granules (Group 2) were labeled with alkyne-containing phenoxy probes generated by localized APEX2. Proteins that were labeled by both probes (Group 3) were those that transited from the nucleolus to stress granules upon arsenite treatment.

Concluding remarks

Over the last decade, the field of proximity labeling has grown exponentially and, as a result, has impacted numerous areas of biology. The studies highlighted here show four ways PL has been used beyond the classic bait-prey PPIs that initiated the field; the tools commonly used to explore the interactomes of specific POIs are being creatively repurposed to elucidate complex biological systems.

Despite the recent advances in PL, the current technologies have inherent limitations that can be improved upon in the coming years. For example, PL techniques that exogenously manipulate cells likely perturbate native cellular physiology, which may confound results. This continues to motivate efforts to develop PL tools that minimize alteration of native cell states. The development of PL has enabled the study of biomolecular interactions to transition from cell lysate to live cells and, in some cases, in vivo. Future research should continue applying PL in innovative biological contexts to diversify the scope of PL applications. In the coming years, we anticipate the community will explore the distinct areas discussed in this review (RNA, small molecule target identification, surfaceomics, and dual-labeling approaches) by using PL enzymes and catalysts in other innovative ways that have yet to be considered (see Outstanding Questions).

Outstanding Questions.

• Many PL methods rely on overexpressing fusion proteins. To what extent does this affect native cellular physiology? How can we minimize artifacts from this genetic manipulation in our findings?

• The development of PL has enabled the study of biomolecular interactions to transition from cell lysate to in cello, and in some cases, ex vivo tissues. How can this technique be leveraged to study such interactions in animals?

• How can we incorporate endogenous switches to trigger PL?

Highlights.

• Proximity labeling can be used to study biological systems that are challenging to study through established paradigms, such as the cell membrane.

• In recent years, scientists have taken advantage of novel proximity labeling systems to reveal biological mechanisms that span from cell-cell interactions to signal transduction and ligandbinding events.

• Advances in molecular and chemical biology, have enabled the creation of novel localization strategies for PL. This presents opportunities for specific subcellular and context-dependent labeling of protein and RNA.

• When deployed appropriately, PL can be used to understand small molecule mechanism of action.

Glossary

APEX

(ascorbate peroxidase) In the presence of exogenous hydrogen peroxide, this enzyme turns over biotin-phenol probes into biotin-phenoxy radicals that covalently react with electron-rich amino acids.

APEX2

Engineered version of APEX that has improved activity in cells, allowing superior enrichment and detection of proteins.

HRP

(horseradish peroxidase) A proximity labeling enzyme that activates biotin-phenol probes in the presence of hydrogen peroxide.

Proximity Labeling

A technique used to tag biomolecules physically near a point of interest. The tagged molecules may be subsequently enriched and identified.

TurboID

An engineered proximity labeling enzyme. In the presence of ATP, APEX generates a reactive biotin probe (biotinyl-5’-AMP) without the requirement of cytotoxic hydrogen peroxide. Notably, the labeling time of TurboID is ten minutes, which is orders of magnitude faster than previous mutants of the BirA enzyme.

μMap

A photocatalytic proximity labeling technique. μMap is known for its short labeling radius (~ 4 nm). In this method, blue light irradiation excites a localized iridium catalyst, which performs Dexter energy transfer to nearby diazirine probes. These probes decompose into carbene species, which quickly react with nearby biomolecules.

HaloTag domains

HaloTag is a self-labeling protein tag derived from a bacterial enzyme called Haloalkane dehydrogenase. This 33kDa protein reacts covalently with a hexyl chloride ligand. Fusion of this domain to proteins of interest allows effective localization of a variety of small molecules to a protein of interest.

Antisense oligonucleotide probes

Antisense oligonucleotides are single-stranded DNA sequences, typically 15–25 nt in length, that bind to complementary sites in RNAs. Attachment of molecular probes to these sequences allows biological interrogation of the RNA of interest.

miniSOG

mini–Singlet Oxygen Generator is a 106 amino acid flavin-binding protein that generates singlet oxygen under exposure to blue light.

Dibromofluoroscein sensitizer

Small molecule dye of the fluorescein family that can generate singlet oxygen through triplet sensitization following irradiation. Singlet oxygen reacts with nearby biomolecules, facilitating detection.

Comparative proteomics

Comparative proteomics is used to analyze proteome changes in response to development, disease, or environment.

PROTAC

Proteolysis Targeting Chimera. Heterobifunctional molecules that degrade protein targets via recruitment of the ubiquitin proteosome system. PROTACs typically consist of a ligand covalently linked to a ligand that binds an E3 ubiquitin ligase.

Molecular glue

A small molecule that brings two biomolecules together (usually a protein-protein interaction) that would not usually interact.

Surfaceome

Collective term for all biomolecules that reside at the cell surface.

TurboID fusion protein

Protein of interest fused to the proximity labeling enzyme TurboID. Allows enrichment of proteins localized to a specific sub compartment of the cell or in proximity to the fusion protein of interest.

Generation 3 BirA* enzyme

Engineered version of BirA* (BioID) to improve reactivity and enzyme kinetics.

BET inhibitor

A class of drugs that reversibly bind the bromodomains of Bromodomain and Extra-Terminal motif (BET) proteins such as BRD2, BRD3, and BRD4.