Photocatalytic Proximity Labeling Maps RNA and Protein Subcellular Landscapes

Abstract

Mapping the molecular landscapes within subcellular compartments is critical for understanding cellular function. Photocatalytic proximity labeling has emerged as an increasingly diversified chemical strategy for interrogating biomolecular microenvironments in living systems. However, a modular small-molecule platform that can be readily retargeted and used to capture both protein and RNA neighborhoods within specific organelles remains underdeveloped. Existing photocatalytic approaches have employed distinct probe-activation modes for biomolecular labeling. Here, PhotoPL adopts an oxidative activation design based on established selenide photooxidation chemistry within a modular organelle-targeted platform. This method employs an organelle-targeted photocatalyst and an activatable selenide probe. Upon visible light irradiation, the localized catalyst generates short-lived reactive intermediates that covalently tag proximal biomolecules. We successfully applied PhotoPL to profile the proteomes and transcriptomes of mitochondria and the endoplasmic reticulum (ER), validating the high-fidelity capture of organelle-specific molecules. Unlike enzyme-based methods, PhotoPL utilizes small-molecule components, minimizing cellular perturbation. Its key advantage is a modular, “plug-and-play” design that allows for rapid retargeting to new subcellular locations without laborious genetic cloning. This modular and scalable platform expands the current chemical toolbox for dissecting the molecular organization of cells and exploring the interplay between proteins and RNAs in health and disease.

Introduction

Diverse biological functions of cellular life rely on the precise compartmentalization and dynamic organization of biomolecules within organelles and subcellular structures. , Understanding how these functions are executed requires detailed knowledge of the molecular constituentsproteins and nucleic acidsresiding within or dynamically associating with each compartment. , Mitochondria (Mito) and the endoplasmic reticulum (ER) are prime examples, serving as hubs for energy production, protein synthesis and folding, lipid metabolism, and calcium signaling. The coordinated localization and interaction of proteins and RNAs within these compartments are fundamental to their function and overall cellular homeostasis. , Dysregulation of this spatial organization is linked to numerous diseases, including neurodegeneration, metabolic disorders, and cancer. −

While techniques like subcellular fractionation coupled with omics have provided valuable inventories, they often lack the spatial resolution to capture transient interactions or map molecules within specific suborganellar domains, and sample processing can introduce artifacts. Imaging methods offer high spatial resolution but typically target only a few molecules at a time. ,

Proximity labeling (PL) techniques have emerged as powerful tools for mapping molecular neighborhoods in living cells. , Enzyme-based PL methods, such as APEX (peroxidase) − and BioID/TurboID (biotin ligase), − utilize genetically encoded enzymes fused to a protein or targeting sequence to catalyze the formation of reactive intermediates (e.g., phenoxyl radicals, biotinoyl-5′-AMP) that covalently label proximal biomolecules, primarily proteins. These methods have significantly advanced our understanding of protein interactomes and organelle composition. Furthermore, recent adaptations of these systems, CAP-seq, and Halo-seq, have elegantly expanded proximity labeling to map subcellular transcriptomes. However, they often require stable expression of large fusion proteins, which can perturb endogenous systems, and the reactive species or cosubstrates such as H₂O₂ could influence redox-sensitive proteins or pathways, interfering with cellular activities. Furthermore, redirecting these enzymatic systems to new targets invariably requires de novo generation of genetic constructs, a process that can be laborious and is not readily scalable for exploring diverse cellular locations. Efficiently adapting these enzymatic systems for robust, simultaneous mapping of both proteins and nucleic acids within the same compartment remains a challenge. Beyond enzyme-based proximity labeling, photocatalytic proximity labeling has rapidly expanded into a diversified chemical toolkit. Prior studies have established photocatalytic strategies for protein labeling, cell-surface interaction capture, organelle-resolved proteomics, and RNA profiling using both Ir- and Ru-based photocatalysts. These advances demonstrate that photocatalytic proximity labeling is already a broad and active area. Within this context, an unmet need remains for a modular platform architecture that decouples subcellular targeting from probe design and enables unified capture of both proteins and RNAs with high spatiotemporal control.

Chemical approaches to PL offer potential advantages, including smaller labeling machinery and tunable reactivity. Crucially, a key advantage of a chemical strategy is its potential for a modular design that promises superior versatility and scalability. Photocatalysis, driven by visible light, provides exquisite spatiotemporal control and operates under biocompatible conditions. Recent studies have demonstrated photocatalytic strategies for protein labeling, including ligand-directed Ru­(bpy)₃ labeling, affinity bead-based labeling, MAUra tyrosine labeling, lectin proximity labeling, singlet-oxygen-driven histidine labeling, cell-surface capture and protein labeling, Nucleic acid cross-linking and profiling of specific RNA condensates have also been achieved via photocatalytic approaches. Recently, a pioneering study (CAT-seq) elegantly demonstrated photocatalytic RNA profiling and synchronous multiomics using an Iridium-based photocatalyst to activate quinone methide (QM) probes via a reductive mechanism. While powerful, this reductive pathway requires endogenous electron donors (e.g., NADH). Furthermore, to prevent probe diffusion, that system requires covalent attachment of organelle-targeting moieties directly onto the probe, complicating retargeting to new subcellular compartments. Additionally, simultaneous multiomics required the coadministration of two distinct orthogonal probes. More broadly, although singlet-oxygen-mediated oxidation of selenides and subsequent selenoxide-elimination chemistry are established, their integration into a modular organelle-targeted platform for unified RNA and protein proximity labeling in living cells remains underdeveloped. Motivated by this opportunity, we sought to adapt this established reactivity into a plug-and-play live-cell mapping strategy.

Here, we develop PhotoPL as a modular platform that adapts established singlet-oxygen-mediated selenide oxidation and selenoxide-elimination chemistry to organelle-targeted proximity labeling in living cells. In this design, visible-light excitation of organelle-targeted ruthenium complexes generates localized singlet oxygen, which is proposed to activate a selenide probe and thereby form reactive quinone methide (QM) intermediates. The short-lived intermediates rapidly cross-link proximal proteins and nucleic acids within nanometers of the catalyst. We demonstrate precise targeting of PhotoPL to mitochondria and the endoplasmic reticulum (ER), enabling high-resolution mapping of their respective proteomes and transcriptomes. Comparative analysis reveals distinct molecular signatures and coenrichment patterns, providing insights into organelle-specific functions, RNA localization mechanisms, and protein-RNA interplay under basal and stress conditions. This chemically controlled, light-triggered PhotoPL strategy offers a powerful, versatile, and easily scalable toolbox for dissecting the molecular organization of living cells with high spatiotemporal resolution.

Results and Discussion

Principle and Validation of the Photocatalytic Proximity Labeling (PhotoPL) Strategy

Our PhotoPL strategy employs a targeted chemical system to map the molecular neighborhood of specific cellular locations using visible light (Scheme ). The core principle involves three components: (1) a photocatalyst (Cat), typically a Ru (II) polypyridyl complex, engineered with a targeting moiety to localize it to a specific subcellular region (e.g., mitochondria or ER); (2) a bioorthogonal selenide-based chemical probe (Probe) bearing a downstream functional handle (e.g., an azide); and (3) visible light (e.g., blue light, ∼450–470 nm) as the trigger.

1. Photocatalytic Proximity Labeling (PhotoPL) Platform.

Upon illumination, the localized photocatalyst absorbs light and enters an excited state. Guided by prior reports of singlet-oxygen-mediated oxidation of selenides and subsequent selenoxide-elimination chemistry, we hypothesized that the excited catalyst would oxidatively transform the nearby probe via a selenoxide intermediate. This established reaction logic is proposed to generate a highly reactive, short-lived quinone methide intermediate in situ, which can be trapped by proximal biomolecules. Due to its short lifetime and high reactivity, the QM intermediate covalently reacts almost immediately with nucleophilic residues present on endogenous biomolecules within nanometer proximity of the catalyst. The molecular basis for this dual protein-RNA labeling stems from the inherent electrophilicity of the o-QM intermediate, which drives rapid Michael-type nucleophilic addition. Structurally, on protein surfaces, the o-QM reacts highly efficiently with nucleophilic amino acid side chains, particularly the thiol of cysteine (Cys), the primary amine of lysine (Lys), and the imidazole nitrogen of histidine (His). Concurrently, while nucleic acids are generally less nucleophilic than protein thiols, guanine (G) residues uniquely possess highly nucleophilic centers. Specifically, the exocyclic amino group (N²) and the N⁷ position of guanine serve as classic structural targets for electrophilic QM alkylation (Figure S1). − , The azide handle on the now covalently attached probe enables subsequent bioorthogonal ligation (e.g., copper-free Click chemistry) to reporter tags like biotin (for enrichment) or fluorophores (for imaging). This combination of targeted catalysis and light activation provides precise spatiotemporal control over the labeling process.

To identify an optimal photocatalyst for this transformation, we screened a panel of 18 photosensitizers (Figure S2). Using the conversion of a model selenide probe which could be transformed into a reactive QM intermediate in response to our photocatalytic reaction to react with neighbor H₂O, we found that ruthenium­(II) complex (hereafter Cat-10) demonstrated superior catalytic efficiency under blue light irradiation (450–470 nm), achieving near-quantitative conversion of the substrate within 30 min (Figures A,B and S2). The photocatalytic activation is proposed to proceed via a photosensitization mechanism, wherein the excited Ru (II) complex generates singlet oxygen (¹O₂) that oxidizes the selenide probe. This hypothesis predicts a strict dependence on molecular oxygen. To test this, we monitored the reaction under both anaerobic (N₂) and oxygen-saturated (O₂) conditions. As expected, catalysis was completely abrogated in the absence of oxygen. Conversely, under pure O₂, the reaction was dramatically accelerated, with near-complete substrate conversion observed in just 12 mina rate markedly faster than under ambient air (Figure C). To further validate that singlet oxygen (¹O₂) is the specific reactive oxygen species driving this transformation, we performed a specific quenching experiment. The addition of 1,4-diazabicyclo[2.2.2]­octane (DABCO, 10 equiv), a well-established ¹O₂ scavenger, markedly suppressed product formation (Figure S3). Labeling was strictly dependent on the presence of all three components: catalyst, probe, and light, and was suppressed by the addition of singlet oxygen quenchers, confirming that the labeling proceeds through the proposed photocatalytic selenide activation pathway rather than direct nonspecific oxidation by ¹O₂.

1.

Development of PhotoPL strategy to map both RNA and protein. (A) Photocatalysts screening for template reaction of o-(phenylseleno)­methyl phenol oxidation to yield quinone methide intermediate. (The reaction was carried out in a H₂O and DMSO (1:1) system with 1.0 equiv of probe and 0.1 equiv of photocatalyst under 365 nm blue LED irradiation for 30 min). (B) Chromatograms of the template reaction substrates and products on a liquid chromatography column. (Mobile phase conditions are provided in Supporting Information Table 1). (C) Monitoring the reaction progress over time under oxygen conditions by HPLC. (D) Schematic diagram of probe molecule optimization and conjugation of photocatalysts with organelle-targeting groups. (E) Cytotoxicity assessment of the photocatalysts. (F) In vitro effect detection of the labeling system using BSA and DNA strands. (G) Results of labeling efficiency detection for BSA and template strands in vitro. (a)­Template DNA was detected by Fam fluorescence and GelRed staining to identify labeled DNA strands and total DNA strands, respectively; (b) BSA was detected by Fam fluorescence and Coomassie Brilliant Blue to identify labeled BSA and total BSA, respectively.

Having established the core photocatalytic transformation, we next engineered the system’s components for targeted biological applications. To enable downstream capture and analysis of labeled biomolecules, we functionalized the phenylselenomethyl phenol probe with a bioorthogonal azide handle through carboxyl group at the phenol para-position. Electron-withdrawing substituent at the para-position could progressively decrease QM lifetime and promote nucleophilic addition to the nascent QMs. This azide handle permits subsequent copper-free Click chemistry ligation to reporter tags, such as biotin for affinity purification or fluorophores for imaging. To achieve organelle-specific labeling, we synthesized targeted versions of the Ru (II) photocatalyst by conjugating it to either a triphenylphosphonium (TPP) moiety for mitochondrial delivery (Mito-Cat) or a p-toluenesulfonamide group for endoplasmic reticulum targeting (ER-Cat) (Figures D and S4).

A critical prerequisite for live-cell applications is biocompatibility. We confirmed that both the probe and the targeted photocatalysts exhibited minimal cytotoxicity at the working concentrations, ensuring that the PhotoPL system operates without perturbing cellular health (Figures E, and S5). We then validated the system’s capacity to label both proteins and nucleic acids in vitro. Upon brief (15 min) blue light irradiation, both bovine serum albumin (BSA) and a model DNA oligonucleotide were efficiently modified by the probe. Subsequent bioorthogonal conjugation with a DBCO-fluorophore and analysis by gel electrophoresis revealed robust fluorescent signals for both biomolecule classes (Figure F,G). Crucially, labeling was strictly dependent on the presence of all three componentscatalyst, probe, and lightdemonstrating the system’s precise chemical and temporal control and confirming its feasibility for dual RNA-protein mapping.

Identification of Labeling Sites and Residue Selectivity

To confirm that the observed labeling arises from QM-mediated capture rather than nonspecific ROS chemistry, and to assess the residue selectivity of PhotoPL, we performed LC-MS/MS analysis of BSA labeled under optimized PhotoPL conditions. To avoid combinatorial explosion from searching all possible residue types simultaneously, we performed two parallel searches: (i) KHR search (modification allowed on Lys, His, and Arg), and (ii) WCM search (modification allowed on Trp, Cys, and Met).

Across three independent replicates, we achieved 88.5% (KHR search) and 88.1% (WCM search) sequence coverage, identifying a total of 559 and 518 BSA peptides, respectively. High-confidence modification sites (localization probability >0.9, detected in at least two replicates) were identified on Lys, His, Arg, Trp, and Cys residuesthe expected nucleophilic targets of QM electrophiles (Tables S4–S5).

To distinguish whether labeling occurs on adjacent residue pairs (e.g., K235 vs W237 in ALK-AWSVAR; H311 vs C312 in SHCIAEVEK), we cross-referenced the KHR and WCM search results. For each peptide containing adjacent potential modification sites, the modification ratios from both searches were identical, indicating that the same peptide molecule carries a single modification at one of the adjacent residues (Table S6). Based on the relative nucleophilicity (Cys > His > Lys > Trp) and solvent accessibility, we favor Cys and His as the primary targets in these adjacent pairs

Importantly, we did not observe elevated levels of methionine oxidation or other oxidative damage markers that would be expected from direct ¹O₂-mediated protein oxidation. The modification pattern was highly specific to the +350.159 Da mass shift, and the wide dynamic range of modification ratios (0.001 to 60.3) reflected differences in nucleophile reactivity, solvent accessibility, and local microenvironment. Collectively, these data provide direct molecular evidence that PhotoPL labeling proceeds through QM-mediated capture at specific nucleophilic residues rather than diffuse nonspecific oxidative damage.

High-Fidelity Subcellular Targeting and In-Cell Labeling

To apply this PhotoPL platform in the complex environment of living cells, we first optimized the labeling parameters. Experiments established optimal working concentrations of 4 μM for the photocatalyst and 400 μM for the probe, with a 15 min light exposure providing robust labeling of both RNA and proteins (Figures A and S6). Further increasing the concentration to 4000 μM resulted in a slight signal decrease (Figure S7), likely due to solubility limitations at millimolar levels. Importantly, this empirically derived optimum conceptually aligns with the ideal 100:1 stoichiometric ratio of probe to catalyst typically required for efficient turnover in this type of photocatalytic reaction. During protein labeling optimization, we observed a low level of background signal from the DBCO-fluorophore reporter. This was effectively suppressed by preblocking nonspecific reactions with iodoacetamide (IAM), significantly improving the signal-to-noise ratio for both in vitro and cellular contexts (Figures B and S8–S9).

2.

Development, optimization, and subcellular imaging validation of PhotoPL platform. (A) RNA probe, photocatalyst concentration, and time-dependent labeling. Optimization of protein labeling conditions is shown in Supporting Figure S3. (B) Elimination of false-positive background in protein labeling. (C) Fluorescence spectra of photocatalyst. (The maximum emission wavelength is 640 nm, and the maximum excitation wavelength is 260 nm.) (D) Confocal microscopy images of cells treated with photocatalyst and subcellular markers. Confocal microscopy images of cells treated with photocatalyst (2 μM, green) and MitoTracker Deep Red (or ER Tracker Green, 100 nM, red) are shown. Scale bar, 20 μm. (E) Pearson correlation coefficient calculation. The Pearson correlation coefficient was calculated to evaluate the colocalization of photocatalyst.

A prerequisite for high-resolution spatial mapping is the precise localization of the photocatalyst. We confirmed the subcellular distribution of Mito-Cat and ER-Cat in HeLa cells using confocal microscopy. The intrinsic fluorescence of the ruthenium complex (λ_ex/em ≈ 460/640 nm) allowed for direct visualization (Figure C). As designed, Mito-Cat showed a distinct filamentous network pattern that strongly colocalized with the mitochondrial stain MitoTracker Deep Red (Pearson’s coefficient = 0.82). Similarly, ER-Cat displayed a reticular pattern characteristic of the endoplasmic reticulum, colocalizing with ER-Tracker (Pearson’s coefficient = 0.79) (Figures D,E and S10). These results confirm the high-fidelity targeting of our photocatalysts, setting the stage for organelle-specific molecular mapping.

PhotoPL Captures Organelle-Specific Transcriptomes and Proteomes

With a validated PhotoPL system in hand, we proceeded to map the RNA and protein landscapes of mitochondria and the ER in living HEK293T and HeLa cells. After PhotoPL, cell lysates were subjected to Click chemistry with a biotin-azide tag, and labeled biomolecules were enriched using streptavidin-coated magnetic beads. We targeted the mitochondria and endoplasmic reticulum of HEK293T cells with Mito-Cat and ER-Cat, respectively, and retrieved the corresponding RNA. The samples were then reverse-transcribed into first-strand cDNA and directly subjected to qPCR amplification. Using total RNA extracted from 293T cells as a control, we compared the relative expression levels of different genes based on the qPCR results. As shown in Figure A, compared with total RNA, the Mito-PhotoPL samples showed significant enrichment of known mitochondrial-encoded transcripts (e.g., MTND1, MTCO2). Crucially, to quantitatively assess specificity and rule out off-target cytosolic labeling, we monitored the cytosolic house-keeping gene GAPDH. The relative expression levels of GAPDH and the ER-resident gene SSR2 were significantly de-enriched in the Mito-PhotoPL samples, consistent with the expression characteristics of mitochondria. − Conversely, ER-PhotoPL enriched for canonical ER-resident transcripts like SSR2, TMX1, and SFT2D2 (Figure A). Overall, the results were consistent with expectations. The enriched mitochondria and endoplasmic reticulum from HEK293T cells, along with

3.

PhotoPL in living cells. (A) qPCR validation of RNA enrichment in mitochondria and the endoplasmic reticulum in HEK293T cells. (B) WB validation of protein labeling in HeLa cells.

total RNA as a reference, were processed for library construction according to the protocol provided by the NEBNext UltraExpress RNA Library Prep Kit. This involved steps such as reverse transcription, adapter ligation, and PCR amplification to generate cDNA libraries for the enriched mitochondrial and endoplasmic reticulum RNA. The library samples were then sent to a sequencing facility for further analysis.

In Figure B, lanes 1–4 represent samples prepared from protein solutions after magnetic bead enrichment, while lanes 5–8 correspond to the magnetic bead samples. To rigorously rule out nonspecific background labeling, strict control groups were evaluated (lanes 6, 7, and 8), including probe-only treatments (minus light or minus catalyst). These controls exhibited completely clean backgrounds, demonstrating that the selenide probe remains entirely inert until activated by the localized photocatalyst. Among the four enriched samples, lane 5 (experimental group) shows a strong contrast against lanes 6, 7, and 8 (control groups), demonstrating good labeling efficiency in the in vivo experiments. We also validated Mito-Cat using the same approach and obtained similar results. Moving forward, we will perform proteomic analysis on the magnetic bead samples to identify the localization and functions of the enriched proteins.

To obtain a global view of the organellar molecular landscapes, we subjected the enriched samples to RNA-sequencing and mass spectrometry-based proteomics. The RNA-seq data from Mito-PhotoPL and ER-PhotoPL revealed a strong enrichment of transcripts known to reside in each respective compartment. For instance, mitochondrial genes, such as MT-TL2, and endoplasmic reticulum-related genes, including GPAA1 and HSPA5, were captured by Mito-PhotoPL and ER-PhotoPL, respectively (Figure S11). Taking the comparison results of the enriched endoplasmic reticulum RNA as an example, five endoplasmic reticulum genes, including GPAA1, were selected, all showing more than 2-fold enrichment. To quantitatively benchmark the performance of PhotoPL against the gold-standard APEX method, we conducted an in-depth bioinformatic analysis of the ER-enriched transcriptomes. First, a volcano plot comparing ER-PhotoPL against total RNA revealed significant enrichment, successfully identifying 679 potential ER-localized RNAs, demonstrating a robust signal-to-noise ratio (Figure S12). We next evaluated the accuracy by comparing these 679 transcripts with the 2468 ER-localized RNAs identified by APEX under identical screening criteria. A Venn diagram revealed a high-confidence intersection of 121 RNAs (Figure S13). While PhotoPL yields a more focused subset compared to the enzymatically amplified APEX data set, it maintains remarkable spatial specificity. To rigorously quantify this, we assessed the false-positive rate by cross-referencing our data with 674 strictly nuclear-localized transcripts (localization score >0.7) from the RNALocal database. Strikingly, PhotoPL exhibited a notably low false-positive rate of only 1.18% (8 overlapping nuclear RNAs), which is highly comparable to the 0.49% false-positive rate of APEX (12 overlapping RNAs). These metrics confirm that PhotoPL achieves exceptional spatial fidelity and high accuracy in mapping organellar transcriptomes.

After obtaining the proteomics results, we analyzed the data using the endoplasmic reticulum as an example. The volcano plot of proteins revealed significant differences between the experimental and control samples, which were consistent with the Western blot (WB) results (Figure S14). After extracting the significantly enriched protein sequences, we performed Gene Ontology (GO) analysis on these data using bioinformatics tools (Figure S15) to understand the biological functions, cellular component localization, and biological processes of the enriched proteins. Analysis of the cellular component (CC) data showed that the number of proteins located in the endoplasmic reticulum lumen was the highest and far exceeded the distribution in other sites, such as focal adhesions or the cytoskeleton. This indicates that the majority of the enriched proteins are related to the endoplasmic reticulum, further validating the good targeting effect of the photosensitizer and the good reactivity of the synthesized selenide probe under photocatalysis for proteins.

Functional Insights into the ER Proteome and Stress Response

The biological process enrichment dot plot and bubble plot illustrate the biological processes associated with the enriched proteins. In Figure S16A, significant enrichment of protein hydroxylation and response to endoplasmic reticulum (ER) stress is observed, which is closely related to our primary research interest in ER-associated proteins. Biological processes involving protein folding and post-translational modifications, including hydroxylation, phosphorylation, acetylation, and ubiquitination, also account for a large proportion. These processes have been widely validated as important biological functions of the ER. Figure S16B displays the enrichment analysis results for a range of biological processes, with significance used to measure the importance of each process. The distances between bubbles in the plot indicate the similarity or correlation between biological processes, reflecting the high degree of shared proteins and the potential functional interrelatedness of these proteins. These results provide clues for further experimental research, such as exploring the specific mechanisms of ER stress or investigating the roles of protein modifications in cellular functions.

In the molecular function association map of the enriched endoplasmic reticulum (ER)-related proteins (Figure S17), 26 proteins, including PDIA4, P4HB, and TXNL1, exhibit disulfide isomerase activity. , Disulfide bonds are one of the critical stabilizing factors for the three-dimensional structure of proteins. During protein synthesis, the correct formation of disulfide bonds is essential for protein folding and function. Proteins such as PDIA4, P4HB, and TXNL1, through their disulfide isomerase activity, can recognize and catalyze the formation and cleavage of disulfide bonds in proteins, thereby assisting in proper protein folding. This process is vital for maintaining cellular protein homeostasis. These proteins not only provide important insights into the molecular mechanisms of protein folding but also offer potential therapeutic targets for related diseases. For instance, PDIA4 has been found to be significantly overexpressed in colorectal cancer and is closely associated with tumor invasiveness, metastasis, and resistance to chemotherapeutic drugs. The correlation between high expression of PDIA4 and chemoresistance in ovarian cancer has been confirmed. PDIA4 helps cancer cells resist the toxicity of chemotherapeutic drugs by modulating the ER stress response, thereby contributing to drug resistance. Inhibition of PDIA4 expression can significantly enhance the sensitivity to chemotherapeutic drugs. Additionally, three proteinsGLRX, SELENOT, and SUMF1exhibit disulfide oxidoreductase activity. They catalyze redox reactions involving sulfur atoms, thereby playing a key role in maintaining redox balance within the cell. − This redox balance is crucial for normal cellular physiological functions and for coping with various stress conditions.

By applying the photooxidation catalysis system to organelles such as the endoplasmic reticulum (ER), we successfully identified subcellular-localized proteins and explored their potential applications through bioinformatics analysis. To further validate the subcellular specificity of the photooxidation-catalyzed proximity labeling method and to uncover the biological significance of the identified proteins, we established a tunicamycin-induced ER stress model. Tunicamycin (TU) inhibits N-linked glycosylation, leading to the accumulation of unfolded proteins within the ER lumen and subsequent activation of the unfolded protein response (UPR), a canonical cellular stress pathway. , Using the ER membrane protein calnexin (CANX, 90 kDa) as a reference, we selected the disulfide isomerase PDIA4 (72 kDa) as our target for investigation. Western blotting analysis of the tunicamycin-induced system revealed that the expression of CHOP (29 kDa), a core marker of ER stress, was significantly upregulated after 8 h of tunicamycin treatment, whereas CANX levels remained stable, confirming successful induction of ER stress (Figure A). We subsequently analyzed the dynamic changes of PDIA4 under this stress condition. Photooxidation-catalyzed labeling was performed on both tunicamycin-treated and control groups. Prior to streptavidin bead-mediated enrichment, tunicamycin treatment resulted in only a modest increase in total PDIA4 protein levels. However, upon specific enrichment, a marked difference in signal intensity emerged between the two groupsthe tunicamycin-treated group displayed significantly enhanced PDIA4 labeling signals (Figure B). These results demonstrate that, under the catalysis of ER-targeted photosensitizers, the probe effectively captures protein dynamics associated with organellar microenvironmental changes, providing a high spatiotemporal resolution analytical tool for dissecting the functional regulatory mechanisms of organelle-specific proteins under pathological conditions such as stress.

4.

PhotoPL captures ER-resident protein dynamics under tunicamycin-induced ER stress. (A) Validation of ER stress induction. HeLa cells were treated with tunicamycin (4 μg/mL, 8 h). Whole-cell lysates were analyzed by Western blotting with antibodies against CHOP (29 kDa, ER stress marker) and CANX (90 kDa, ER membrane loading control). CHOP expression was significantly upregulated upon tunicamycin treatment, confirming successful ER stress induction, while CANX levels remained unchanged. (B)­PhotoPL captures stress-induced PDIA4 enrichment. Cells were subjected to PhotoPL labeling (Mito-Cat/ER-Cat, 4 μM; selenide probe, 400 μM; blue light, 450–470 nm, 15 min) under control or tunicamycin-treated conditions. After streptavidin bead enrichment, PDIA4 (72 kDa) levels were analyzed by Western blotting.

In the analysis of mitochondrial proteins, several functional proteins were identified. For example, UQCRC1 (Ubiquinol-cytochrome c reductase core protein 1), which is involved in the mitochondrial electron transport chain, is a key component of cellular respiration and energy metabolism. ATP5F1B (ATP synthase subunit β, mitochondrial) is the β-subunit of mitochondrial ATP synthase, involved in ATP synthesis and a core component of mitochondrial energy metabolism. SUCLG2 (Succinyl-CoA ligase, GDP-forming, mitochondrial) participates in the tricarboxylic acid cycle within the mitochondria, catalyzing the synthesis of succinyl-CoA, a key enzyme in energy metabolism.

Conclusion

In this study, we have developed PhotoPL, a versatile chemical platform for the simultaneous mapping of protein and RNA landscapes. To clearly delineate the novelty and advancement of our approach, it is crucial to compare it with state-of-the-art photocatalytic proximity labeling methods, such as CAT-ER and CAT-Prox. As summarized in Supporting Table 2, PhotoPL is differentiated not by introducing an underlying oxidation/elimination chemistry, but by integrating established photooxidative selenide activation into a modular organelle-targeted platform architecture for proximity labeling. By employing targeted photocatalysts and activatable probes, PhotoPL achieves high spatiotemporal resolution and operates under biocompatible conditions, offering an alternative to enzyme-based PL methods. We successfully applied PhotoPL to map the molecular constituents of mitochondria and the ER in living cells, providing comprehensive proteomic and transcriptomic snapshots of these vital organelles.

The PhotoPL strategy offers several advantages. First, the use of small-molecule catalysts and probes minimizes potential steric hindrance or functional perturbation associated with large enzyme fusions. Second, visible light activation allows for precise temporal control and potentially deeper tissue penetration compared to UV-based methods. Third, and perhaps most significantly, the platform is inherently modular and highly scalable. Unlike systems that require complex synthesis of targeted probes, PhotoPL utilizes a universal probe. Redirecting the labeling to new subcellular landscapes simply requires conjugating the catalyst to a different known targeting ligand. This ’plug-and-play’ chemical approach bypasses the laborious process of genetic cloning required for enzyme-based systems and opens the door to rapidly exploring a multitude of cellular targets. Crucially, the ability to capture both proteins and RNAs using a unified single-probe approach significantly streamlines the simultaneous mapping of molecular ecosystems within organelles.

Our results confirm the expected protein and RNA populations within mitochondria and the ER, validating the specificity of the PhotoPL approach. The MS-based identification of specific modification sites on BSA provides direct molecular evidence for the QM-mediated labeling mechanism. The observation of modifications exclusively on nucleophilic residues (Cys, Lys, His, Arg, Trp)with modification ratios spanning 4 orders of magnitude (0.001 to 60.3)is consistent with proximity-driven QM trapping that depends on residue-specific nucleophilicity, solvent accessibility, and local microenvironment, rather than diffuse nonspecific oxidative damage. The absence of elevated methionine oxidation or other oxidative stress markers further supports this conclusion. Notably, the exceptionally high modification at K235/W237 (ratio >19) suggests that this surface-exposed region of BSA serves as a particularly favorable site for QM capture, potentially due to a combination of high solvent accessibility and favorable local electrostatic environment. These mechanistic insights, enabled by direct adduct site identification, validate the chemical selectivity of the PhotoPL platform and reinforce its utility for high-fidelity proximity labeling. Furthermore, the identification of known and potentially novel organelle-associated factors highlights the discovery potential. Comparative analysis of the mitochondrial and ER data sets underscores the distinct molecular environments tailored to their specific functions.

While PhotoPL offers significant advantages, potential limitations include the need for efficient probe delivery and the possibility of off-target labeling by diffusible intermediates, although the short lifetime of QMs and the localized generation aim to minimize this. Careful optimization and control experiments, such as those performed here (catalyst/probe/light dependence, IAM blocking), are essential. A defining strength of the PhotoPL platform is its structural modularity, which provides a clear pathway for generalizability to other organelles. Future developments will directly leverage this “plug-and-play” capability. For instance, by replacing the mitochondrial TPP moiety with DNA-binding ligands (such as Hoechst derivatives) or nuclear localization sequences (NLS), the platform can be readily adapted to interrogate the proteome and transcriptome of the nucleus and distinct chromatin domains. Similarly, conjugating the core Ru­(II) catalyst to morpholine or ceramide-based chemical moieties will enable mapping of the lysosomes and the Golgi apparatus, respectively. Furthermore, conjugating catalysts to small-molecule inhibitors could allow researchers to map the highly localized interactome of specific endogenous proteins. This adaptability, combined with developing catalysts responsive to longer wavelengths (e.g., red/NIR light) for deeper tissue studies, will further expand the utility of PhotoPL in complex biological systems.

In conclusion, photocatalytic proximity labeling provides a powerful and versatile chemical tool for dissecting the complex molecular organization of living cells. Its ability to map both protein and RNA landscapes with high spatiotemporal control, combined with its modular design and scalability, opens new avenues for exploring organelle biology, protein-RNA interactions, and their roles in health and disease.

All data sets have been deposited in the Gene Expression Omnibus under accession number GSE304177. Other data and materials are available from the authors upon reasonable request.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.6c01018.

• Materials and methods, chromatography mobile phase condition, photoactivation of probe generates QM in situ table, cytotoxicity measurement experiments of probes and photocatalyst molecules, Venn diagram illustrating the overlap of identified ER-localized RNAs, and sequences, and NMR spectra (PDF)

†.

M.H., L.R., X.X. and R.Z. contributed equally to this work. The manuscript was written through contributions of all authors.

This work was supported by the Natural Science Foundation of China (22277091).

The authors declare no competing financial interest.