Mechanisms and Applications of Photocatalytic Proximity Labeling

Abstract

Photocatalytic proximity labeling (PPL) is a light-controlled technique for spatially and temporally precise mapping of biomolecular interactions. This review classifies PPL platforms into small-molecule catalysts (transition metal complexes and organic chromophores) and genetically encoded photosensitizers, detailing their mechanisms and biological applicationsfrom subcellular proteomics, cell–cell interface tagging to in vivo labeling. We compare PPL with enzyme-catalyzed methods, highlighting its advantages in spatiotemporal control, minimal genetic manipulation, and suitability for primary cells. Current challenges in PPL are also discussed. As PPL methodologies continue to evolve from mechanistic innovation toward essential biological problem-solving, they hold great promise for elucidating dynamic interaction networks with unprecedented resolution in living systems.

1. Introduction

Cell is the fundamental structural unit of multicellular organisms. The precise regulation of cellular signaling networks relies on orchestrations among intracellular biomolecules such as proteins and nucleic acids, as well as intercellular communications. Processes governed by the central dogmaincluding DNA replication, transcription, and translationdepend on dynamic interactions among various biomolecules. Proteins, as the primary executors of biological functions, engage in interactions with other proteins throughout their lifecycle, from translation initiation to eventual degradation, thereby forming intricate signaling networks. Cells recognize specific ligands via receptors on the membrane to facilitate intercellular communication, maintaining homeostasis at the tissue and organism levels. , Elucidating the dynamic interactions of intracellular macromolecules and intercellular communications with high spatial and temporal resolution is essential for deepening our understanding of the dynamic and complex nature of interaction networks in living systems.

Proximity labeling (PL) has emerged as a powerful tool for mapping biomolecular interactions with high spatiotemporal resolution PL technologies, such as APEX2, BioID and TurboID, rely on the in situ generation of highly reactive intermediates, which covalently modify neighboring biomolecules, enabling their subsequent enrichment and identification. Due to the high reactivity of these intermediates, the labeling radius is typically short, allowing efficient and specific tagging of proteins or nucleic acids within localized microenvironments. PL has been widely applied to profile subcellular compartments and cell–cell interfaces.

In recent years, photocatalytic proximity labeling (PPL) has advanced rapidly, leveraging the excellent spatiotemporal controllability and noninvasive nature of light. In this review, we will summarize the development of PPL platforms over the past decades, including catalysts classification, reactive species, active light wavelength, half-life of reactive intermediates and primary applications (Table ). According to the type of photocatalyst used, we further classify these different platforms into two categories: small-molecule photosensitizer photocatalytic system and genetically encoded photosensitizer photocatalytic system. All the biological applications and key findings of each approach are discussed. Finally, we compare PPL with enzyme-based PL techniques from various dimensions and conclude with a future outlook.

1. Summary of Recent Photocatalytic Proximity Labeling Platforms .

classification	name/acronym	catalyst	reactive intermediate (precursor)	light wavelength	half-life	primary application
transition metal complex	microenvironment mapping (μMAP)	Ir	carbene (diazirine)	blue (450 nm)	2 ns	profile immunosynapse proteins, identify small-molecule protein targets
	CAT-Prox	Ir	quinone methide (para-azidobenzyl caged-quinone methide)	blue (450 nm)	<1 s	mitochondria proteins profiling
	Ru-O2-hydrazide system	Ru	singlet oxygen (O2)/phenol radical (phenol)	blue (450 nm)	4 μs, 100 μs	immune cells interactions
	μMAP-Red	Sn	aminyl radical (phenyl azide)	red (660 nm)	50 μs	EGFR interactome profiling
	targeted aryl azide activation via deep red-light	Os	nitrene (aryl azide)	red (660 nm)	200 μs	profile local protein environment in epithelial cell adhesion molecule (EpCAM)
organic chromophore	Halo-seq	4′,5′-dibromofluorescein	singlet oxygen (O2)	green (500 nm)	4 μs	RNA labeling and sequencing
	rhodamine activation via blue light	Rhodamine	singlet oxygen (O2)	blue (450 nm)	4 μs	mitochondria-selective proteins labeling
	Silicon-rhodamine-enabled identification (SeeID)	silicon-rhodamine	singlet oxygen (O2)	red (660 nm)	4 μs	KRAS-interactome; hippocampus in vivo labeling
	chol-POCA	Janelia Fluor	singlet oxygen (O2)	yellow (570 nm)	4 μs	cholesterol-specific interactions
	MultiMAP	Eosin Y	nitrene (aryl azide), phenol radical (phenol), carbene (diazirine)	blue (450 nm)	200 μs, 100 μs	EGFR interactome; cell–cell interactions
					2 ns
	PhoTag	Flavin	phenoxyl radical (phenol)	blue (450 nm)	100 μs	profile immunosynapse proteins interaction
	CAT-Tissue	Chlorin e6	singlet oxygen (O2)	red (675 nm)	4 μs	Tumor-immune interactions in primary tissues
	PATCH	porphyrin	phenoxyl radical (phenol)	red (660 nm)	100 μs, 4 μs	antigen-induced cellular response amplifying
			singlet oxygen (O2)
genetically encoded photosensitizer	CAP-seq	miniSOG protein	singlet oxygen (O2)	blue (450 nm)	4 μs	subcellular RNA sequencing
	PDPL	miniSOG protein	singlet oxygen (O2)	blue (450 nm)	4 μs	protein interactome labeling
	LITag	LOV domain	singlet oxygen (O2)	blue (450 nm)	4 μs	protein interactome labeling
	RinID	miniSOG protein	singlet oxygen (O2)	blue (450 nm)	4 μs	subcelluar protein interactome labeling
	BRET-ID	SOPP3	singlet oxygen (O2)	blue (450 nm)	4 μs	in vivo protein interactome profiling
	bioluminescence-activated proximity labeling (LAP)	miniSOG protein	singlet oxygen (O2)	blue (450 nm)	4 μs	protein interactome labeling, cell–cell interaction, transcriptome profiling

^aThis table summarizes PPL platforms developed recently and details of each technique. For PPL methods belonging to the same category, the earliest published method in the same series is selected as the representative of that category.

2. Small-Molecule Photocatalysts

2.1. Transition Metal Complexes

Transition metal complexes are among the earliest developed small molecules for PPL, owing to their excellent photophysical properties and stability. , The mechanism by which transition metal complexes enable PL typically relies on their photoexcitation under visible light, followed by activation of nearby substrates (such as azide, diazirine, quinone methide, etc.) via energy transfer or electron transfer, thereby achieving nanometer-scale labeling of surrounding molecules (Figure A).

1.

Transition-metal–based photocatalytic proximity labeling. (A). General excited-state energy- and electron-transfer pathways of transition-metal photocatalysts. PC: photocatalyst, ET: electron transfer, EA: electron acceptor, EnT: energy transfer. (B). Ir­(III)-mediated diazirine activation generates short-lived carbene intermediates upon blue-light irradiation. (C). Antibody-directed μMap enables spatially confined labeling at cell–cell contact interfaces. Reproduced with permission from ref (10). Copyright 2020 The American Association for the Advancement of Science.

2.1.1. Ir Complex

In 2020, Jacob B. Geri et al. developed the μMAP photocatalytic labeling strategy. Under light irradiation, an electron in the central Ir is excited from the ground state to a singlet excited state. Due to the strong spin–orbit coupling of heavy metal atoms, the excited electron rapidly undergoes intersystem crossing (ISC) from the singlet (S₁) to the slightly lower-energy but longer-lived triplet state (T₁). At this stage, Ir complexes engage in photoredox catalysis via energy transfer, delivering energy to a spin-matched diazirine.

Subsequently, the triplet diazirine undergoes elimination of N₂ to release a triplet carbene by short-range Dexter energy transfer mechanism. These carbene species rapidly cross-link with neighboring biomolecules (Figure B). Due to efficient quenching by water molecules and an ultrashort diffusion radius of approximately 4 nm (with a half-life of only about 2 ns in aqueous environments), μMAP achieves exceptionally high labeling precision. In a coculture model using PD-L1-expressing B cells and Jurkat T cells expressing PD-1 and the CD45RO isoform, the researchers conjugated the Ir catalyst to nanobodies targeting PD-L1 and CD45RO, respectively. By leveraging the specificity of these nanobodies, the catalysts were precisely localized to the cell surface. Their experiments demonstrated that PD-L1-targeted μMAP successfully profiled the microenvironmental proteome at the T-cell-B-cell interaction interface mediated by PD-L1/PD-1 binding, while CD45RO-targeted μMAP specifically labeled proximal proteins on the membrane surface of T cells expressing the CD45 isoform (Figure C).

Following their initial development, Saori Suzuki et al. further extended the μMAP system to a broader range of biological contexts. They conjugated the Ir catalyst to the spike protein of SARS-CoV-2. During coronavirus infection of host cells, this approach enabled labeling of host cell surface receptor proteins, facilitating the identification of novel coreceptors that assist viral entry and thereby deepening the understanding of virus–host interactions. In another application, human IgG antibodies were conjugated to carboxy beads to mediate opsonization, followed by functionalization with an iridium photocatalyst for subsequent labeling. When macrophages were incubated with these functionalized beads, phagocytosis was initiated and progressed over time. By applying blue light irradiation at different time points, the dynamic proteomic changes at the macrophage–bead interaction interface could be profiled throughout various stages of the phagocytic process.

The μMAP system demonstrates unique advantages for in situ analysis of cell membrane proteomes, yet it is not without limitations. The first generation Ir catalyst, containing two carboxyl groups on its ligand , exhibited strong hydrophilicity and poor membrane penetration. To address these challenges, Aaron D. Trowbridge et al. rationally redesigned the structure of the Ir complex. In the second-generation Ir catalyst, the carboxyl groups were removed and replaced with hydrophobic aryl moieties and positive charges, significantly enhancing its membrane permeability (Figure A). This modification substantially broadened the applicability of the Ir catalyst. By conjugating the second-generation catalyst with JQ-1 (a BRD4 inhibitor), desHEP-Dasatinib (a Src/Abl tyrosine kinase inhibitor), and MM-206 (a STAT3 inhibitor), novel target proteins were identified in HeLa and THP-1 cells, along with active residue sites within drug-binding pockets (Figure A). Introducing a chloroalkane group allowed covalent linkage to a HaloTag-G3BP1 fusion protein, enabling dynamic labeling and regulatory protein identification within stress granule (SG)-a key membraneless organelle (Figure B). This approach, named HaloMap, revealed critical ubiquitination pathways involved in granule disassembly. Additionally, simultaneous labeling of nuclear RNA and its associated proteins was achieved, providing parallel transcriptomic and proteomic data for deciphering RNA regulatory networks.

2.

Iridium photocatalysts for ligand- and protein-directed labeling. (A). Modular Ir­(III) photocatalysts conjugated to small-molecule ligands enable target-centric labeling. (B). HaloTag-tethered Ir­(III) photocatalysts support intracellular proximity labeling.

In addition to the μMAP system, Zongyu Huang et al. have leveraged iridium catalysts to develop a series of PPL strategies centered on a “decaging” mechanism (Figure A). Under blue light (450 nm), Ir complexes catalyze the conversion of aromatic azides to anilines, which subsequently trigger cleavage of a para-azidobenzyl (PAB) group, generating a highly reactive quinone methide (QM) intermediate. As a Michael acceptor, QM undergoes Michael addition with nucleophilic amino acid residues (e.g., lysine, serine). Based on this principle, they established the CAT-Prox strategy, employing an Ir­(ppy)₂bpy catalyst for bioorthogonal photodecaging. Notably, Ir­(ppy)₂bpy was found to localize to mitochondria, enabling high-throughput profiling of the mitochondrial proteome in live cells (Figure B). CAT-Prox showed robust labeling efficiency and specificity in both HeLa cells and hard-to-transfect RAW264.7 macrophages, revealing previously unannotated localized proteins such as PTPN1 and SLC35A4. Furthermore, CAT-Prox was successfully applied to quantitatively monitor dynamic changes in the mitochondrial proteome of macrophages under inflammatory stimulation.

3.

Photocatalytic decaging strategies for PPL. (A). Ir-catalyzed blue-light decaging activates PAB-caged molecules (CAT-Prox). (B). Organelle-targeted photocatalysts confine labeling reactions within mitochondria. Reproduced with permission from ref (20). Copyright 2021 American Chemical Society.

Extending beyond subcellular organelle profiling, CAT-Ex conjugates the Ir complex to Herceptin (targeting HER2) and Avelumab (targeting PD-L1) to perform decage reactions extracellularly. This approach identified a set of functionally related proteins (e.g., GLUT1, MRP1, ATP1A1) surrounding HER2, suggesting the formation of membrane microdomain clusters. Moreover, light-triggered decaging of PAB-protected monomethyl auristatin E (MMAE) on the surface of HER2-overexpressing MDA-MB-231 breast cancer cells enabled selective cancer cell killing. CAT-Cell further achieves localized labeling at cell–cell contact interfaces. It has been successfully employed for quantitative assessment of TCR-pMHC interaction strength and for highly sensitive detection of antigen-specific T cells in primary mouse splenocytes and tumor-infiltrating lymphocytes. The CAT-S platform replaces the oxygen atom linked to the PAB group in the original QM probe with a sulfur atom, exploiting the heavy-atom effect to develop a thio-QM probe with enhanced labeling efficiency. Using thio-QM probe, in situ mitochondrial protein labeling was accomplished in primary cells derived from lung and peripheral blood, marking the first proteomic mapping of mitochondria in clinical samples.

Except for mitochondria targeted protein profiling, CAT-ER integrated an ER-targeted iridium photocatalyst with thio-QM probe to enable spatiotemporally resolved labeling of ER proteome. It achieved specificity comparable to enzymatic approaches while overcoming limitations associated with hard-to-transfect cells, including various immune and cancer cell lines. Applied to thapsigargin induced stress, CAT-ER revealed dynamic proteomic rearrangements during the transition from the unfolded protein response to apoptosis, identifying key regulators such as NFIP2, which attenuates ER stress by suppressing translation, and EMC2, whose impairment delays apoptotic progression.

In addition to its application in proteomics, the CAT-series platform has been expanded to enable subcellular-resolved transcriptomic analysis through CAT-seq. This system employs QM probe to achieve nongenetic, in situ labeling of mitochondrial RNA in living cells. Using CAT-seq, researchers have elucidated mitochondrial RNA dynamics during macrophage M1 polarization, revealing that downregulation of tRNA modifications (m⁵C and τm⁵U) impairs mitochondrial translation and contributes to oxidative phosphorylation (OXPHOS) remodeling. Moreover, applied to the maturation of primary CD8⁺ T cells, CAT-seq uncovered a mitochondrial transcription-dependent pathway underpinning OXPHOS upregulation, demonstrating its utility in dissecting complex immune metabolic reprogramming.

2.1.2. Ru Complex

In addition to Ir, Ru complexes have emerged as another type of frequently employed transition metal complexes in the field of PPL in recent years.

In 2023, Shuang Qiu et al. developed a Ru­(bpy)₃ ²⁺-based labeling method for the efficient identification of cell–cell interactions. In this system, the Ru catalyst is conjugated to lipids, sugars, or surface proteins to achieve precise anchoring on the cell membrane. Upon 450 nm visible light irradiation, Ru catalyst undergoes ISC to form the triplet state. At this stage, Ru complexes exhibit dual reactivity in photocatalysis: they can either engage in photoredox catalysis via single-electron transfer to an electron acceptor or they can act as photosensitizers through energy transfer, delivering energy to a spin-matched substrate and promoting it to a high-energy state (such as converting ground-state triplet oxygen to high-energy ¹O₂), which enables dual-radius labeling with the same catalyst (Figure ). These two reaction pathways are not mutually exclusive; in many cases both can occur simultaneously. The researchers further established a T cell–dendritic cells (DCs) coculture system, where the Ru complex was anchored on the DC membrane via glycoconjugate labeling. When DCs presented the OVA_257–264 antigen and engaged with OT-I CD8⁺ T cells, irradiation in the presence of either a hydrazide probe or a phenol probe allowed labeling of the interacting T cells. This work provides a new tool for dynamic monitoring and quantitative study of immune cell interactions.

4.

Ru­(bpy)₃ ²⁺-mediated dual-pathway photocatalysis. Photoexcited Ru­(bpy)₃ ²⁺ engages electron-transfer and singlet-oxygen pathways in parallel. Reproduced with permission from ref . Copyright 2023 John Wiley and Sons.

Meanwhile, Ashley N. Ogorek et al. reported a “switchable DNA catalyst” strategy based on Ru­(bpy)₃ ²⁺for controllable PL at protein–protein interaction sites. The Ru photocatalyst and a quencher moiety were covalently attached to a conformation-switching DNA oligonucleotide anchored to a target protein. Only when a second molecule (such as an interacting protein or a specific ligand) induced a DNA conformational change, leading to dissociation of the quencher, could the Ru catalyst be activated under blue light to enable localized protein labeling. Using the HER2–HER3 heterodimer and the c-Met homodimer as models, the spatial specificity of this method for detecting protein–protein interactions was validated in live SKBR3 and HEK293T cells. Xu Zhou et al. further extended this technology to detect cell–cell interactions. Leveraging two orthogonal chemical labeling systemsSNAP-tag/Benzylguanine and SpyTag/SpyCatcherdistinct DNA strands were anchored on two different cell surfaces. At the cell–cell contact interface, the two DNA strands hybridized via complementary sequences, a structural change that displaced the quencher strand bound to Ru­(bpy)₃ ²⁺, thereby activating the photocatalyst. Subsequent blue-light irradiation enabled phenol-probe-based labeling of proximal proteins at the contact interface, simultaneously allowing imaging visualization and protein tagging.

While Ru catalysts have demonstrated remarkable utility in various applications, several limitations remain. Ru complexes are typically activated by blue light, which exhibits limited tissue penetration, thereby constraining their effectiveness for in vivo or deep-tissue labeling. Furthermore, blue light irradiation carries inherent risks such as phototoxicity and elevated background noise during labeling. Additionally, the membrane permeability of Ru catalysts requires further enhancement; current systems are largely confined to labeling extracellular proximal proteins and lack the capability to interrogate intracellular protein interactions. One potential strategy to improve transmembrane uptake involves modifying the ligand structure with hydrophobic groups while reducing overall molecular charge. This approach could enhance lipid solubility and mitigate electrostatic repulsion with the phospholipid bilayer, thereby increasing the permeability of catalysts.

2.1.3. Other Metal Complexes

Since in vitro cultured cells lack the physiological context of in vivo microenvironments and cannot fully recapitulate living systems, Benito F. Buksh et al. and Nicholas Eng Soon Tay et al. independently developed photocatalysts centered on Sn and Os, respectively, which are activated by near-infrared light at 660 nm , (Figure A,B). These have been successfully applied to probe cell membrane microenvironments. Changing the metal center also altered the underlying chemical mechanism: instead of the Dexter energy transfer pathway, the new systems operate via an electron-transfer process. In the case of Os complexes, photoactivation excites the Os catalyst into a triplet state, after which a single electron transfer (SET) occurs to generate the excited-state Os catalyst. N₂ is then dissociated upon reduction to form the reduced nitrene. Finally, reoxidation of the reduced nitrene by Os⁺ yields the triplet nitrene. The catalytic process in the Sn-complex system is quite similar to that of the Os complex, though it involves NADH. It begins with the reductive quenching of the excited-state Sn photocatalyst by NADH, forming a highly reducing ground-state species. The azide compound oxidizes the reduced Sn by SET process, converting it into an azide anion, which then releases a molecule of nitrogen gas to generate an aminyl radical for PL.

5.

Red-shifted photocatalysts for improved tissue penetration and in vivo application. (A). Structure of a red-light-responsive μMap photocatalyst. (B). Os-based photocatalyst enables deep-red-light-activated proximity labeling. (C). Antibody-guided labeling using red-shifted photocatalysts at the cell surface. Reprduced with permission from ref . Copyright 2022 Springer Nature.

For specific biological application, μMAP-red was applied in whole blood samples and conjugated to a TER119 antibody. After the antibody binds to TER119 on the surface of blood cells, irradiation at 660 nm is applied to label proteins within the TER119 microenvironment. Interestingly, μMAP-red demonstrated notable tissue penetration efficiency. μMAP-red retains the ability to activate azide substrates and generate labeling even when tissue depth exceeds 10 mm, demonstrating its unique potential for in vivo applications. In another case, Os photocatalytic system was applied in different colon cancer cell systems for targeted protein environment labeling of epithelial cell adhesion molecule (EpCAM) (Figure C). They identified a small subset of proteins with previously known and unknown association to EpCAM, including CDH3, a clinically relevant protein that shares high tumor-selective expression with EpCAM.

Both aforementioned methods have portrayed satisfactory labeling performance in vitro and in deep-tissue systems; however, they have not been substantively applied in vivo. To further leverage the favorable optical property of near-infrared lightspecifically its superior in vivo compatibility, CAT-NIR employs an Os­(II)-polypyridine complex that, in synergy with NADH, undergoes a reductive quenching photocatalytic cycle to convert triplet ground state oxygen into superoxide. The superoxide then drives a deboronative hydroxylation reaction, resulting in decaging of boronate-ester protecting groups. Combined with genetic code expansion, a boronate-protected tyrosine analog (pBoF) was site-specifically incorporated into ZHER. Upon NIR illumination, pBoF is decaged to restore tyrosine, thereby recovering ZHER’s binding ability to HER2 and enabling cell-selective labeling. In the context of targeted drug release, prodrugs Dox-Bpin and 5-FU-Bpin were designed and synthesized. In MDA-MB-231 and HCT116 cell models, CAT-NIR efficiently activated these prodrugs and inhibited cell growth. In an MDA-MB-231 tumor xenograft model, intratumoral injection of Dox-Bpin together with the Os catalyst, followed by external 740 nm NIR irradiation, led to significant suppression of tumor growth. It broadens the scope of applications for PPL in vivo without the requirement of genetic modification.

2.2. Organic Chromophores

Transition metal complexes offer advantages such as good photostability, rich triplet excited states, and high photocatalytic efficiency, which can enhance labeling efficiency on cell surfaces. However, within intracellular environments, they are susceptible to interference from endogenous metal chelators and highly nucleophilic compounds like glutathione. In contrast, organic chromophores demonstrate superior bioorthogonality, remaining unaffected by the reducing intracellular milieu while providing structural simplicity and high membrane permeability. Consequently, organic chromophores have rapidly emerged as a prominent class of molecular tools for PPL.

2.2.1. Xanthene-Based Chromophores

2.2.1.1. 4′,5′-Dibromofluorescein (DBF)

Since 2017, several research teams have leveraged 4′,5′-dibromofluorescescein as a core chromophore in a series of labeling strategies, advancing methodologies from protein–protein interaction studies to the subcellular localization of RNA. Ying Li et al. first introduced an in situ RNA proximity labeling strategy using DBF (Figure A). They fused HaloTag to nuclear proteins (H2B, fibrillarin) or the cytoplasmic protein NFκB p65 and conjugated a chloroalkane-functionalized DBF to HaloTag. Upon excitation, DBF enters a triplet state and interacts with molecular oxygen via an energy-transfer pathway to generate singlet oxygen (¹O₂) locally within the nucleus or cytoplasm. This oxidizes guanosine (G) to 8-oxoguanosine (8oxoG), which is then captured by an amine-based probe for RNA enrichment. RT-qPCR validation confirmed successful enrichment of compartment-specific RNAs such as 7SK, U6, and GAPDH, demonstrating the spatial specificity of the approach.

6.

Xanthene-based organic photocatalysts. (A) Dibromofluorescein (DBF) serves as a xanthene-based photocatalyst. (B). Rhodamine and silicon-rhodamine derivatives enable organelle-targeted labeling. (C). Janelia Fluor dyes enhance singlet-oxygen generation through heavy-atom substitution. (D). Rose Bengal selectively labels protein aggregates under light activation. (E). Eosin Y supports multiscale photocatalytic labeling.

Using the same labeling strategy, Krysta L. Engel et al. systematically optimized RNA extraction and downstream bioinformatics workflows to develop Halo-seq, a higher-throughput RNA proximity labeling method (Figure ). Halo-seq applies labeled RNA to RNA-seq analysis, significantly increasing data throughput and quantitative capability. Their work revealed that RNAs containing AU-rich elements in their 3′-UTRs were enriched in the nucleus in a manner associated with HuR protein binding, with the number of HuR-binding sites correlating with nuclear enrichment. Treatment with the nuclear export inhibitor leptomycin B (LMB) impaired HuR-dependent RNA export, leading to increased nuclear retention; ultimately, 105 LMB-sensitive RNAs were identified. Notably, the authors concluded that Halo-seq exhibits higher labeling efficiency than APEX-seq and CAP-seq (based on miniSOG), particularly for detecting small RNAs and unspliced transcripts. This difference may stem from the short-lived nature of APEX-derived radicals and their known propensity to cleave RNA, which would reduce the reactive capacity of generated tyramide radicals for tagging.

7.

DBF-based RNA proximity labeling (Halo-seq). HaloTag-tethered DBF generates localized singlet oxygen to label proximal RNAs in defined subcellular compartments. Reproduced from Engel, K. L. et al., ref ; licensed under CC-BY-NC 4.0.

In 2025, Hei-Yong G. Lo et al. further refined sequencing methods and developed updated data-analysis tools, resulting in ONIC-seq technology. Instead of using an aniline probe to label 8oxoG, the oxidized guanosine itself serves as a direct tag on the transcriptome. The PIGPEN pipeline directly sequences these oxidative modifications. The primary oxidation product of guanosine is 8-oxoguanosine (8oxoG), which can be further oxidized to 5-guanidinohydantoin (GH). Reverse transcriptases misincorporate nucleotides (G → T, G → C) or skip bases when encountering 8oxoG or GH, and these conversion events are detected by high-throughput sequencing without the need for biotinylation and enrichment. The team also developed the bioinformatics tool PIGPEN to identify and quantify G → T, G → C, and G-deletion events from RNA-seq data, with the BACON module performing statistical comparisons of conversion rates under different conditions. Using four HaloTag-fusion proteins targeting distinct subcellular compartments, PIGPEN analysis showed significant enrichment of conversion events in known compartment-specific RNAs. Finally, by expressing HaloTag-NES (cytoplasmic) and HaloTag-H2B (nuclear) in live zebrafish embryos, they achieved in vivo RNA oxidative labeling; sequencing-detected G-conversion events displayed high spatial specificity.

PhoXCELL technology, in contrast, utilizes the transacylase Sortase A (SrtA) to covalently attach DBF bearing an LPETG motif to GGG-modified proteins on the surface of dendritic cells (DCs), achieving membrane-specific anchoring of DBF. Under 520 nm light, DBF-catalyzed reactive oxygen species oxidize surface proteins on T cells interacting with DCs, followed by labeling with an aniline probe, enabling high-precision spatiotemporal labeling within 10 min. Bulk RNA-seq of labeled T cells revealed upregulation of genes associated with cell differentiation, adhesion, and cytokine secretion in strongly interacting cells. When DBF was conjugated to the mitochondrial-targeting triphenylphosphonium group, proteome labeling within this subcellular compartment was achieved. Although coverage was slightly lower than with APEX2 or TurboID, this approach identified more low-abundance proteins.

2.2.1.2. Rhodamine

Rhodamine and its derivatives, as xanthene-based photosensitizers structurally analogous to DBF, also hold considerable potential for PL applications. Haoyan Wang et al. developed a photocatalytic labeling method based on Rhodamine 123 (Figure B). They initially screened various photosensitizing dyes and found that fluorescein, acridine orange (AO), and rhodamine 123 (Rh123) could activate aryl azide probes via an energy transfer mechanism. Upon photoexcitation, the probes release nitrogen gas and generate highly reactive triplet nitrene intermediates, which subsequently convert into benzazirine and ketenimine intermediates. These reactive species covalently label nucleophilic amino acid residues on nearby proteins. While AO exhibits a high ¹O₂ yield in acidic environments, its DNA-binding property compromises cell viability. In contrast, Rh123 demonstrates a favorable ¹O₂ yield and intrinsic mitochondrial targeting ability, enabling spatially specific labeling within mitochondria. Furthermore, by treating cells with rotenone to induce stress, they identified several mitochondrial stress-related proteins, including four previously unreported proteins such as PITRM1.

Leveraging the heavy-atom effect, Maik Müller et al. conjugated a thiorhodamine derivative via an N-hydroxysuccinimide (NHS) ester linkage to an anti-CD20 antibody, cardiac glycoside CG1, transferrin, and insulin. This approach enabled the labeling of proximal protein domains around antibody targets, small-molecule drug ligands, and other biologically important molecules.

To achieve highly specific labeling and develop an in vivo-compatible PPL technology, we developed an ultralow-background near-infrared labeling strategy termed SeeID (silicon-rhodamine-enabled Identification). This method utilizes a silicon-rhodamine chloralkane (SiR-CA) that becomes fluorescent upon covalent binding to HaloTag while simultaneously serving as a photocatalyst (Figures B and A). Under 660 nm irradiation, SiR generates ¹O₂, which oxidizes protein residues to form 2-oxohistidine; this intermediate is subsequently labeled by nucleophilic addition of 3-ethynylaniline. In HeLa (KRAS-WT) and SW1573 (KRAS-G12C) cells, SeeID identified 427 KRAS-interacting proteins, 212 of which were novel interactions. Newly identified interactors such as AXL, OSBPL5, and PHLDA1 were validated by PLA and coimmunoprecipitation. Combining live-cell imaging with SeeID, the dynamic interaction changes of Parkin during BL-918-induced mitophagy were investigated. The proteasome complex showed enhanced interaction with Parkin in early mitophagy stages, suggesting its role in early mitochondrial protein degradation. Furthermore, by delivering Halo-TM via AAV into the mouse hippocampus and visual cortex, near-infrared-controlled labeling was achieved in vivo. The labeling depth reached 2 mm, and 62% of the enriched proteins were associated with membranes, vesicles, or the extracellular matrix, including neurotransmitter receptors and cell adhesion molecules.

8.

Rhodamine- and Janelia Fluor–based photocatalytic protein labeling. (A) Halo-conjugated rhodamine derivatives function as both fluorescent probes and photocatalysts for proximity labeling (SeeID). Reproduced from Wang, W. et al., ref ; licensed under CC-BY–NC–ND 4.0. (B) Janelia Fluor–based cholesterol probes enable photocatalytic capture of lipid-associated protein interactomes. (C) Light-dependent labeling followed by mass spectrometry identifies proximal protein networks.

2.2.1.3. Janelia Fluor

In 2020, Thomas C. Binns et al. introduced JF567 and JF570, two photosensitizers synthesized based on the rhodamine scaffold by incorporating heavy atoms (iodine or sulfur) to enhance singlet-oxygen quantum yields and achieve longer excitation wavelengths, making them more suitable for in vivo imaging (Figure C). In transgenic casper zebrafish embryos, JF567 and JF570 were respectively conjugated to a mitochondria-targeted HaloTag-TOMM20 fusion protein, enabling neuron-specific light-induced cell ablation.

To achieve the identification of cholesterol-interacting proteome, Andrew P. Becker et al. selected the cell-permeable photosensitizer JF570 to construct two probes: cholesterol-conjugated JF570 (chol-POCA) for capturing cholesterol-interacting proteins, and palmitate-conjugated JF570 as a control to distinguish cholesterol-specific interactions (Figure B,C). Chol-POCA successfully identified multiple known cholesterol-binding proteins, such as NPC1, NPC2, OSBP, VDAC1/2, and LAMP1/2. Compared with palmitate-JF570, chol-POCA-enriched proteins were more localized to cholesterol-rich compartments including the endoplasmic reticulum, Golgi apparatus, and plasma membrane. When benchmarked against diazirine-based probes, POCA covered a similar set of cholesterol-binding proteins and captured more protein complexes. Applying an HDL-cholesterol probe in hepatocytes via POCA identified proteins involved in HDL uptake, such as SR-B1, PLIN2, and CES1. By covalently anchoring JF570 to a HaloTag fused to target proteins (Aster-B or NUP153), the researchers investigated the cholesterol transporter Aster-B and revealed its ER-domain-dependent oligomerization, which was further validated by photo-cross-linking. POCA also captured the colocalization of Aster-B with the lipid raft protein FLOT1, suggesting that Aster-B may mediate cholesterol transport through lipid rafts.

Fluorescein derivatives are anionic at physiological pH and typically require ester-masking groups to facilitate cellular entry. For halogenated derivatives, their ester bonds are notoriously unstable in aqueous solutions, especially. Thus, in vivo application of fluorescein derivatives often necessitates direct injection of DMSO solutions into tissues. In contrast, simple rhodamine dyes are net neutral and exist in an equilibrium between a lipophilic lactone form and a polar zwitterion. Consequently, rhodamine-based Janelia dyes are inherently more compatible with physiological conditions.

2.2.1.4. Rose Bengal

In 2024, Huan Feng et al. collaborated to develop a photocatalytic proximity labeling method specifically targeting aggregated proteins under thermal stress. Using dihydrofolate reductase as a model protein, they screened and identified Rose Bengal as a high-affinity binder selective for protein aggregates (Figure D). The interaction is driven by hydrophobic contacts between exposed hydrophobic patches on the aggregating proteins and the hydrophobic moieties of Rose Bengal. Upon illumination, Rose Bengal generates reactive oxygen species (ROS), which covalently modify neighboring proteins. Subsequent mass spectrometry analysis of the enriched proteins revealed the regulatory role of the HSP70 family chaperone HSPA1B within the aggregates. Moreover, the study demonstrated that under proteostasis imbalance, aggregated proteins are primarily degraded via the BAG3-mediated autophagy pathway rather than through the proteasome system.

2.2.1.5. Eosin

In their 2017 study, Ying Li et al. demonstrated that Eosin Y can oxidatively label RNA via ¹O₂ generation, but comparative analysis indicated that DBF exhibits lower nonspecific background. In 2024, Zhi Lin et al. developed MultiMap, a multiscale PPL method using Eosin Y as the catalyst (Figure A). This system activates three distinct classes of reaction probes (aryl azides, phenols, and diazirines) to achieve tunable labeling radius, allowing flexible adjustment of spatial resolution. In studies of membrane proteins such as EGFR, MultiMap uncovered multiple functionally associated proteins, with results validated by AlphaFold-based structural predictions. Furthermore, the method was extended to investigate immune cell–cell interactions, enabling efficient profiling of the proximal interactome at intercellular contact interfaces.

9.

Eosin Y–based multiscale and organelle-specific labeling strategies. (A). Antibody-anchored Eosin Y activates distinct probe chemistries to achieve multiscale cell–surface interactomics (MultiMap). (B). Lysosome-targeted Eosin Y enables in situ labeling of lysosomal proteomes (CAT-Lyso). Reproduced with permission from ref . Copyright 2024 The American Association for the Advancement of Science.

Building on the versatility of Eosin Y based photocatalytic platforms, strategies have emerged to combine such labeling mechanisms with organelle-specific targeting for deeper subcellular proteome insights. CAT-Lyso combines the thio-QM probe with a morpholine moiety for lysosomal targeting, enabling the revelation of heterogeneous lysosomal proteomes in live cells and uncovering key regulators associated with the mTOR pathway (Figure B).

2.2.2. Isoalloxazine-Based Chromophores

2.2.2.1. Flavin

Flavins are a class of compounds featuring an isoalloxazine ring scaffold, which is also present in essential human biomolecules such as vitamin B₂ and the coenzyme factors flavin mononucleotide (FMN) and flavin adenine dinucleotide (FAD) (Figure A). Under irradiation at 450 nm, flavins can transit to an excited state and subsequently catalyze the generation of ¹O₂ via energy transfer or activate other substrates via single-electron transfer to produce radicals, making them suitable for photocatalytic labeling applications.

10.

Flavin-based photocatalytic labeling. (A). Flavin scaffolds act as photocatalytic chromophores. (B). Riboflavin derivatives support phenol–tyrosine coupling reactions. (C). PhoTag maps protein microenvironments across immune synapses with defined spatial resolution.

Rob C. Oslund et al. first employed flavins to establish a PPL platform termed photocatalytic cell tagging (PhoTag). They identified riboflavin tetraacetate (RFT) as an efficient catalyst (Figure B). Under 455 nm illumination, RFT activates phenolic substrates to label tyrosine residues via a photoinduced electron-transfer process, exhibiting significantly higher efficiency than mechanistically similar Ru complexes. By conjugating RFT to single-domain antibodies (VHHs) targeting PD-L1, they achieved localized presentation of RFT on the cell membrane. In a coculture system of Raji PD-L1⁺ cells and Jurkat PD-1⁺ cells, the method enabled specific labeling of interacting cells bearing complementary ligands (Figure C). When applied to mixed populations of peripheral blood mononuclear cells (PBMCs), PhoTag allowed multiomics single-cell sequencing of T-cell subsets that transiently interacted with PD-L1-expressing Raji B cells. The analysis revealed that not all PD-1⁺ T cells interacted with Raji B cells at the same frequency; naïve CD4⁺ T cells and terminally differentiated effector memory CD8⁺ T cells engaged more frequently with PD-L1⁺ Raji B cells. The authors further extended PhoTag to profile the protein microenvironment at intercellular interfaces during immune synapse formation, identifying previously unreported immune synapse-associated proteins such as DSG2, EPHA3, PLXNB2, and ITGAV, thereby opening new avenues for investigating the composition and functional regulation of immune synapses.

To elucidate the mechanistic basis of flavin-mediated phenol-tyrosine coupling, Taylor O. Hope et al. conducted a systematic study of the process by which RFT activates phenolic substrates to label tyrosine residues. Contrary to the earlier proposed radical addition hypothesis, their work demonstrated that tyrosyl radicals and phenol radicals actually couple via a radical–radical combination mechanism. The latter pathway exhibits a significantly lower reaction energy barrier (−1 to +3 kcal/mol) compared to the former (>+34 kcal/mol).

2.2.3. Other Chromophores

In 2016, Heather L. Glasgow et al. conjugated methylene blue (MB) to the nerve-binding peptide NP41 (Figure A). Upon irradiation with 650 nm red light, the excited methylene blue generated ¹O₂, which oxidized protein residues in the vicinity of NP41 bound to the surface of mouse sciatic nerve tissue. The oxidized residues were then covalently labeled using a biotin-hydrazide probe. This work pioneered the use of peptide-conjugated photosensitizers for proximity-based proteomic labeling at the tissue level, paving the way for subsequent research in PPL.

11.

Alternative organic photocatalysts for proximity labeling. (A). Methylene blue serves as a red-light-activated organic photocatalyst capable of generating reactive oxygen species for proximity labeling. (B). BODIPY derivatives function as tunable photosensitizers with modifiable photophysical properties. (C). TTMAPP acts as a cationic porphyrin photocatalyst enabling near-infrared light–driven labeling reactions. (D). Pyropheophorbide-a (PPa) is employed as a photosensitizer for NIR-activated proximity labeling and spatial transcriptomics. (E). Chlorin e6 enables deep-red-light-triggered singlet-oxygen generation for in situ protein labeling. (F). The helical carbenium ion n-Pr-DMQA⁺ supports single-electron-transfer-based proximity labeling under red light.

In 2022, Keita Nakane et al. first reported a novel application of the classic fluorescent probe BODIPY in PL (Figure B). Iodinated BODIPY, when excited by 455 nm blue light, efficiently produced ¹O₂ and selectively labeled histidine residues. In the same year, Kazuki Miura et al. discovered that acriflavine serves as a cell-permeable photocatalyst. By covalently linking it to a HaloTag-H2B fusion protein, they localized the catalyst to chromatin regions and subsequently captured tyrosyl radicals using an MAUra-DTB probe, enabling the identification of proteins interacting with Histone 2B.

The photosensitizers mentioned above are typically excited by short-wavelength light around 500 nm, which exhibits limited tissue penetration. To overcome this limitation, Keun Ah Ryu et al. developed a near-infrared photocatalytic protein labeling strategy based on fluoroalkylation. When the 5,10,15,20-tetrakis­(4-trimethylammoniophenyl)­porphyrin tetra­(p-toluenesulfonate) (TTMAPP) and the helical carbenium ion N′-di-n-propyl-1,13-dimethoxyquinacridinium tetrafluoroborate (n-Pr-DMQA⁺) are exposed to 660 nm red light, they are excited and reduce iodo-perfluoroalkyl reagents via a single-electron transfer mechanism, leading to homolytic cleavage of the carbon–iodine bond and generating fluoroalkyl radicals (Figure C,F). These radicals subsequently attack electron-rich aromatic amino acid residues such as tryptophan, enabling C–H fluoroalkylation labeling. The entire reaction proceeds in physiologically compatible aqueous buffers, offering spatiotemporal controllability and excellent biocompatibility. To achieve selective labeling of cell–surface proteins and introduce fluorescent tags via click chemistry for visualization, TTMAPP is conjugated to antibodies via NHS ester chemistry. In human colorectal cancer tissue, mouse brain tissue, and esophageal tissue, localized irradiation combined with antibody targeting enabled efficient labeling of surface proteins on specific cell populations within the tissues.

In addition to applying chromophores for protein labeling, Tao Deng et al. developed CINTER-seq, which employs pyropheophorbide-a (PPa) to generate reactive oxygen species under 675 nm near-infrared light (Figure D). Biotin-aniline covalently couples to oxidized biomolecules, and the biotin signal is converted into sequencable information using DNA-barcoded streptavidin for enrichment. Combined with clickTag and single-cell RNA-seq, CINTER-seq simultaneously captures interaction strength, transcriptome data, and sample origin. In mouse spleen, CINTER-seq successfully captured antigen-specific interactions between B cells and CD8⁺ T cells and validated the stabilizing role of molecular pairs such as LAG3-MHCII in these interactions. Furthermore, correlating interaction strength (biotin signal) with single-cell transcriptomes revealed that Lag3 expression in CD4⁺ T cells strongly correlated with interaction intensity, suggesting that LAG3-MHCII directly mediates intercellular adhesion and highlighting a potential target for immunotherapy. Interestingly, they observed that neutrophils interacting with tumor cells acquired a pro-tumor phenotype, characterized by high expression of PD-L1, Icam1, Ccl4, and other genes, and suppressed T-cell function, thereby promoting tumor immune escape. In tumor-draining lymph nodes of hCD20 transgenic mice, they found that tumors induced B-cell infiltration into T-cell zones and enhanced B-T-cell interactions.

Chlorin e6 shares structural and physicochemical similarities with PPa (Figure E). Zhizheng Lou et al. conjugated chlorin e6 to nanobodies that specifically target tumor-cell surface antigens such as PD-L1 and HER2. Under 660 nm deep-red light irradiation, Ce6 generated short-lived ¹O₂, which oxidized proximal histidine residues on proteins; these were then covalently labeled with a biotin-aniline probe. In coculture systems, the method detected and quantified key immune interactions including CD40L-CD40 and TCR-pMHC, with labeling signal intensity correlating with interaction affinity. On primary tumor sections from mice, the approach specifically labeled and sorted CD8⁺ T cells, NK cells, and CD4⁺ T cells interacting with MC38-hPD-L1 tumor cells. They discovered that CD8⁺ T cells adjacent to tumors exhibited a distinct transcriptional profile, with high expression of activation, effector function, and exhaustion-related genes (e.g., PD-1, TIM-3, Granzyme-B), whereas CD8⁺ T cells distant from tumor cells were more naïve or memory-like.

Additionally, Feifei Tong et al. designed an α-sulfonyl diazo ester probe that, in the presence of chlorin e6 as a photocatalyst under 660 nm red light, generated carbene intermediates to covalently label protein residues. This system provided strong, rapid labeling with a radius of approximately 4 nm, comparable to μMAP. Simultaneously, leveraging high singlet-oxygen yield of Ce6, they used aniline and hydrazide probes for labeling, achieving a larger labeling radius (about 30 nm) and an 8-fold higher signal intensity compared to the carbene system. By covalently conjugating Ce6 to the anti-EGFR antibody cetuximab to form a Ctx-Ce6 complex, they achieved EGFR-specific labeling on A549 cells, confirming the system’s selectivity, low toxicity, and applicability. Tissue samples from different regions of the same patient’s glioblastoma (GBM) tumor were analyzed. Using carbene and singlet-oxygen probes separately, EGFR interactomes were profiled by mass spectrometry. Integrated bioinformatics analysis revealed GBM-specific, location-dependent EGFR interactors. The interactomes differed significantly across regions, with only 11 proteins commonly enriched in all areas.

PPa and chlorin e6 share a core porphyrin scaffold, highlighting the significant potential of porphyrin-based structures in PPL. Building on this foundation, porphyrin-derived chromophores have recently been repurposed to engineer cell surfaces and enhance immune recognition. Shuojun Li et al. introduced an in vivo labeling strategy termed PATCH (Proximity Amplification and Tagging of Cytotoxic Haptens) that uses a porphyrin-based nanozyme activated by either red light or ultrasound to densely and covalently deposit fluorescein probes on target tumor cells. By combining this localized, catalytic amplification of synthetic antigens with a fluorescein-binding bispecific T-cell engager (FITC-BiTE), they achieved potent T-cell receptor clustering, enhanced cytotoxicity, and complete tumor eradication in multiple syngeneic mouse models. Importantly, the treatment further stimulated systemic antitumor immunity and immune memory against rechallenge. This work demonstrates how red-light photocatalytic chemistry can be leveraged not only for mapping molecular neighborhoods but also for reprogramming cell–surface antigen densitythereby offering a new therapeutic paradigm to overcome insufficient receptor activation in cancer immunotherapy.

3. Genetically Encoded Photosensitizer

While small-molecule photocatalysts enable initial applications in PL, their diffusible nature limits precise spatial and temporal control. This constraint is elegantly overcome by genetically encoded photosensitizer, which can be targeted to specific cellular compartments or fused to proteins of interest. These genetically encoded tools offer enhanced specificity, minimal background, and the capacity for dynamic, spatiotemporally resolved labeling in living systemsthereby enabling more physiologically relevant interrogation of protein interactions and local environments.

miniSOG originates from the blue-light-sensing LOV2 (Light, Oxygen, Voltage-sensing) domain of Arabidopsis phototropin 2 (AtPhot2). AtPhot2 retains the ability to bind flavin mononucleotide (FMN) characteristic of LOV-family proteins, and upon photoexcitation it efficiently generates triplet-state FMN (³FMN), which then transfers energy to ground-state molecular oxygen via an energy-transfer mechanism to produce highly reactive ¹O₂ (Figure A). However, Cys 426 in native AtPhot2 can form a covalent bond with photoexcited FMN, partially diverting energy that would otherwise be transferred to oxygen. Tsien et al. mutated this cysteine to glycine, thereby improving oxygen accessibility to the excited FMN, and performed saturation mutagenesis around the chromophore-binding pocket, ultimately isolating a brighter variant designated miniSOG. The singlet-oxygen quantum yield (ΦΔ) of miniSOG is 0.03 ± 0.01, measured using free FMN (ΦΔ = 0.65 ± 0.04) as the gold standard. Subsequent studies indicated that miniSOG can also generate superoxide anion via a photoinduced electron-transfer (PET) pathway, which likely contributes to its lower ¹O₂ yield compared to free FMN.

12.

Mechanism and engineering of genetically encoded singlet-oxygen–generating photocatalysts. (A). Blue-light excitation of flavin mononucleotide (FMN) produces triplet-state FMN, which transfers energy to molecular oxygen to generate singlet oxygen. (B). Directed mutagenesis of the LOV2 domain from Arabidopsis thaliana phototropin 2 yields miniSOG and its higher-efficiency variants SOPP, SOPP2, and SOPP3. (C). Singlet oxygen oxidizes proximal RNA bases, enabling nucleophilic trapping and covalent labeling. (D). Singlet-oxygen-mediated oxidation of amino-acid side chains allows covalent labeling of nearby endogenous proteins.

Michael Westberg et al. introduced point mutations to modulate the electron affinity of FMN in miniSOG. The Q102L mutant, named SOPP, exhibited an increased ΦΔ of 0.19 ± 0.03. To further enhance reactive-oxygen-species production, they systematically optimized SOPP and identified the double mutant W81L/L103 V, termed SOPP2, which achieved ΦΔ = 0.51–0.55 under air-saturated conditions. To fully eliminate PET, residues that could act as alternative electron donors after removal of Trp 81 (H85, M89, Y98) were additionally mutated. The resulting quintuple mutant W81L/H85N/M89I/Y98A/L103 V, designated SOPP3, displayed a stable ΦΔ = 0.6 under air saturation at physiological temperature approximately, representing the highest singlet-oxygen quantum yield reported for a genetically encoded photosensitizer to date (Figure B).

In 2019, Pengchong Wang et al. reported CAP-seq (chromophore-assisted proximity sequencing), which employs miniSOG to generate ¹O₂ under blue light. With a short diffusion radius (<70 nm), the reactive oxygen species selectively oxidize proximal guanosine residues. An alkyne-containing amine probe (e.g., propargylamine, PA) is then covalently attached to the oxidized guanosine, enabling biotin introduction via click chemistry for affinity purification and high-throughput sequencing (Figure C). CAP-seq was successfully applied to multiple subcellular compartments: in the mitochondrial matrix, it specifically captured all mitochondrially encoded transcripts; on the endoplasmic reticulum membrane, it efficiently enriched 96% of mRNAs encoding secretory-pathway proteins; and at the outer mitochondrial membrane (OMM), it revealed significant enrichment of 110 mRNAs encoding mitochondrial proteins (including 33 oxidative-phosphorylation subunits) and 62 mRNAs encoding cytosolic ribosomal subunits, supporting the hypothesis of localized translation near the OMM that coordinates with mitochondrial import. Further pharmacological perturbations showed that the protein-synthesis inhibitor cycloheximide enhanced the enrichment of mitochondrial mRNAs at the OMM, whereas the uncoupler FCCP reduced the localization of secretory-protein mRNAs, suggesting distinct ribosome-mediated and membrane-potential-dependent targeting pathways.

CAP-seq was subsequently used to systematically profile the proximal transcriptome of stress granules (SGs) in living cells (Figure A). The study found that SG RNA localization positively correlated with transcript length and AU content but negatively correlated with translation efficiency. SGs induced by different stressors (e.g., arsenite and sorbitol) exhibited markedly distinct RNA compositions. Moreover, the core SG protein G3BP1 pre-existed in a network with AU-rich RNAs even under nonstress conditions, acting as a “seed” for SG assembly. Time-resolved CAP-seq analysis of SG disassembly revealed subpopulations of RNAs with different dissociation kinetics and sequence features, and showed that even after stress removal and SG dissolution, a subset of RNAs remained proximal to G3BP1, forming a “memory”-like nanostructure. This work not only confirmed the high spatial specificity and dynamic analysis capability of CAP-seq for studying membraneless organelles, but also provided novel mechanistic insights into SG assembly, disassembly, and functional regulation.

13.

Applications of miniSOG-based photocatalytic proximity labeling. (A). miniSOG-mediated RNA proximity labeling enables selective enrichment and sequencing of RNAs localized within stress granules. (B). Protein-centric miniSOG labeling followed by click chemistry and affinity enrichment supports quantitative proteomic analysis. (C). Bioluminescence resonance energy transfer (BRET) from NanoLuc activates miniSOG/SOPP3 without external light to enable proximity labeling in living systems. Reproduced from Ren, Z. et al., ref ; licensed under CC-BY-NC 4.0. Reproduced from Zheng, F. et al., ref ; licensed under CC-BY-NC 4.0. Reproduced from ref . Copyright 2025 American Chemical Society.

Yansheng Zhai et al., Nir Hananya et al., and Fu Zheng et al. independently developed proximity-labeling techniques based on miniSOG and its engineered variants, namely photoactivation-dependent proximity labeling (PDPL), light-induced interactome tagging (LITag), and reactive-oxygen-species-induced protein labeling and identification (RinID) (Figure B). All three methods rely on blue-light excitation of miniSOG-type photosensitizing proteins to generate ¹O₂, but they employ different nucleophilic probes: PDPL uses a meta-alkynylaniline probe, LITag utilizes a phenol probe, and RinID employs a propargylamine probe (Figure D). In mass-spectrometry-based residue analysis, the phenol probe in LITag primarily labels tyrosine residues, whereas the amine-type probes in PDPL and RinID predominantly modify histidine residues. This divergence likely stems from the tendency of phenol probes to capture tyrosyl radicals via a single-electron-transfer pathway, while amine probes preferentially trap oxidized intermediates.

For further eliminating the need for exogenous light and avoiding nonspecific labeling caused by endogenous photosensitizer, our lab developed a bioluminescence-resonance-energy-transfer-based proximity-labeling platform termed BRET-ID (Figure C). The core design fuses SOPP3 with Nanoluc; upon substrate addition, bioluminescence resonance energy transfer (BRET) locally activates the photosensitizing protein to generate ¹O₂, which labels neighboring proteins for subsequent enrichment and systematic interactome profiling. Applying BRET-ID to capture interaction remodeling during internalization of the human μ-opioid receptor (hMOR) identified novel hMOR interactors such as ZYG11B and Septin7. Further, BRET-ID was used to dissect dynamic protein composition changes in stress granules, revealing the mTORC2 subunit RICTOR as a novel SG component and providing preliminary evidence for its regulatory role in SG assembly. Finally, leveraging the advantages of BRET-IDincluding negligible phototoxicity, high spatiotemporal resolution, and deep-tissue compatibilitywe performed the first in vivo profiling of the G3BP1 interactome in living animals.

Meanwhile, Ruixiang Wang et al. reported a bioluminescence-activated proximity labeling (LAP) platform (Figure 13C). Although this system similarly employs BRET to activate a photocatalyst, it substantially enhances catalytic efficiency through yeast display-directed evolution and successfully achieves multidimensional spatial profiling of the proteome, transcriptome, and cell–cell interactions. In contrast to emphasis of BRET-ID on high spatiotemporal resolution, the LAP platform places greater emphasis on versatility and integrative multiomics capabilities both in vitro and in vivo, demonstrating unique value particularly in studies of the secretome within the tumor microenvironment and ligand–receptor-mediated cellular interactions.

4. Comparison with Enzymatic Proximity Labeling Methods

Traditional enzyme-catalyzed proximity labeling methods, such as APEX2 and TurboID, have been widely adopted in recent years. Here, we provide a comparative analysis between enzyme-catalyzed proximity labeling and photocatalytic proximity labeling across various dimensions (Table ).

2. Comparison of Enzyme-Catalyzed and Photocatalytic Proximity Labeling .

feature dimension	enzyme-catalyzed proximity labeling	photocatalytic proximity labeling
core mechanism	uses engineered enzymes to catalyze substrates into short-lived radicals that covalently label neighboring proteins	relies on photocatalysts to generate reactive species under specific-wavelength light for labeling
genetic engineering dependency	required	optional. For targeting, antibodies, reversible binding proteins, or covalent conjugation (e.g., HaloTag) can be used
key advantages	1. high organelle specificity. Fusion proteins enable precise subcellular localization	1. no genetic manipulation required. suitable for hard-to-transfect immune cells and precious clinical samples
	2. high sensitivity and labeling efficiency. Enzymatic amplification provides high labeling density	2. ultrahigh spatiotemporal resolution. light control enables second-scale ON/OFF switching and submicron spatial precision
		3. accessible to diverse reactive species. PPL can cover various labeling substrate, such as diazirine, phenol and aniline
key disadvantages	1. requires genetic engineering	1. risk of phototoxicity
	2. enzyme expression may interfere with target protein localization, folding, and native interactions	2. rely on antibody conjugation or chemical tags which lead to targeting strategies complexity
	3. background may be elevated due to endogenous enzymatic activity	3. limited ability of light to penetrate tissues or thick specimens
spatiotemporal control	time: depends on substrate concentration and diffusion rate; cannot be turned off instantaneously	time: precisely controlled by illumination; suitable for kinetic studies
	space: determined by fusion protein expression site; precision limited by diffusion radius of reactive substrates	space: achieves subcellular or even suborganelle resolution using microscopy techniques

^aThis table compares enzyme-catalyzed proximity labeling and photocatalytic proximity labeling across core mechanism, genetic engineering dependency, key advantages and disadvantages, and spatiotemporal control.

In the developmental trajectory of PL technologies, the two dominant strategiesenzyme-catalyzed and photocatalytic labelingare evolving from parallel progression toward an integrated and convergent trend. Optimization of enzyme-catalyzed methods has primarily focused on enhancing the activity and specificity of the tool proteins. Examples include engineering smaller and more catalytically efficient enzyme variants to minimize interference with target protein function, as well as exploring their combination with efficient delivery systems (e.g., viral vectors) to extend their utility in hard-to-transfect cell types.

Simultaneously, photocatalytic proximity labeling technology is advancing to overcome current limitations. Major efforts are directed toward designing novel photosensitizing probes, including molecules with higher singlet-oxygen quantum yields, improved biocompatibility, and near-infrared absorption wavelengths. These improvements aim to boost labeling efficiency, reduce phototoxicity, and enhance tissue penetration, thereby enabling in vivo and deep-tissue in situ interaction studies.

Notably, the boundary between these two approaches is becoming increasingly blurred, giving rise to emerging “hybrid” strategies. For instance, “photoactivatable-enzyme” systems have been developed by coupling photocontrollable elements (such as photolabile caging groups or light-sensing domains) with highly efficient biotin ligases or peroxidases. These systems elegantly combine the transient, spatially precise control offered by light with the high efficiency and sensitivity inherent to enzymatic amplification, representing the next generation of proximity labeling tools. , This evolution marks a paradigm shiftfrom merely “mapping static interaction networks” toward “intervening in and dissecting dynamic biological processes with high spatiotemporal precision”.

5. Summary and Outlook

PPL methods are progressively shifting from mechanistic development toward addressing core biological questions. Whether relying on oxygen-dependent ROS mechanisms or energy-transfer pathways under hypoxic conditions, different labeling platforms can now be customized for diverse application scenarios. Compared to metal complexes, organic chromophores offer greater flexibility for chemical modification to achieve targeting specificity and tissue selectivity, and they integrate more readily with near-infrared excitation systems and optical imaging technologies.

However, a primary limitation of small-molecule photocatalysts lies in achieving precise subcellular and tissue localization. Current strategies that depend on HaloTag fusion or reversible protein interactions confine their utility to easily transfectable cell models or cell–surface studies. More broadly, most of the PPL platforms grapple with trade-offs between excitation efficiency, in vivo stability and high background noise. Inherent to light-driven methods, phototoxicity necessitates careful mitigation through strategies such as shifting to longer-wavelength (e.g., NIR) excitation, engineering catalysts with higher quantum yields and shorter-lived reactive species, optimizing irradiation parameters, and developing exogenous-light-free activation systems like bioluminescence resonance energy transfer (BRET). Successfully balancing labeling efficacy with biocompatibility thus remains a central design criterion for next-generation PPL platforms.

further progress along several promising directions.

• (1)designing smarter catalysts with improved specificity and photophysical properties;
• (2)integrating PPL with multimodal omics and imaging to capture dynamic network biology;
• (3)transitioning from analytical mapping to therapeutic modulationusing photocatalytic chemistry not only to observe but also to perturb and engineer cellular interactions in living systems. Through continued innovation, PPL is poised to become an indispensable tool for both fundamental discovery and translational intervention.

#.

E.G. and Y.X. contribute equally to this work.

The authors declare no competing financial interest.