A genetically encoded photoproximity labeling approach for mapping protein territories

Significance

A photoproximity labeling approach is described that allows capturing of protein interactomes using brief exposure to visible light. The technology—Light-induced Interactome Tagging (LITag)—involves genetically fusing an engineered flavoprotein to a protein of interest. Excitation of the flavin mononucleotide cofactor leads to covalent labeling of nearby proteins with exogenously added probes. The small size of the flavoprotein (~12 kDa) allows the application to many different protein systems, while its intrinsic fluorescence permits correlated cellular imaging and proteomics analyses. This capability is showcased through characterizing the poly(ADP-ribose) polymerase-1 neighborhood during DNA damage and defining the protein clients of the major vault protein, applications that highlight the power of the method for studying dynamic cellular processes and large macromolecular assemblies, respectively.

Abstract

Studying dynamic biological processes requires approaches compatible with the lifetimes of the biochemical transactions under investigation, which can be very short. We describe a genetically encoded system that allows protein neighborhoods to be mapped using visible light. Our approach involves fusing an engineered flavoprotein to a protein of interest. Brief excitation of the fusion protein leads to the labeling of nearby proteins with cell-permeable probes. Mechanistic studies reveal different labeling pathways are operational depending on the nature of the exogenous probe that is employed. When combined with quantitative proteomics, this photoproximity labeling system generates “snapshots” of protein territories with high temporal and spatial resolution. The intrinsic fluorescence of the fusion domain permits correlated imaging and proteomics analyses, a capability that is exploited in several contexts, including defining the protein clients of the major vault protein. The technology should be broadly useful in the biomedical area.

Protein–protein interactions (PPIs) have a pivotal role in regulating cellular function, and their dysregulation is implicated in various diseases, such as cancer and neurodegeneration (1). Accordingly, modulating PPIs is recognized as a promising therapeutic strategy (2). Methods for the characterization of PPIs are fundamental in molecular biology, and developing new tools to study protein networks is a vibrant field of research. In the last decade, proximity labeling (PL) techniques have emerged as a powerful approach for cataloging PPIs in living cells (3). PL utilizes engineered enzymes genetically fused to a protein of interest (POI). The PL enzyme activates an inert small-molecule substrate, generating a short-lived reactive intermediate that diffuses away from the enzyme active site to tag neighboring biomolecules covalently. The existing PL enzymes can be divided into two families—biotin ligases and peroxidases. Biotin ligases, such as BioID (4) and TurboID (5), utilize ATP and biotin to catalyze the formation of reactive biotin-5′-AMP, which labels lysine residues on proximal proteins (6). Peroxidases, e.g., APEX2 (7), catalyze the H₂O₂-mediated oxidation of biotin-phenol, generating a phenoxy radical which reacts with electron-rich residues, predominantly tyrosine. The biotinylation step is followed by streptavidin pulldown, allowing for protein characterization and quantification by various mass spectrometry (MS) techniques.

Compared to traditional methods for PPI characterization, such as affinity purification coupled with mass spectrometry (AP-MS), PL offers several advantages. First, the protein networks are probed in living cells and not in lysates; thus, the chances for nonphysiological artifacts are reduced. Second, PL can capture transient and weak PPIs, which are often missed in AP-MS. Furthermore, it enables the characterization of insoluble protein complexes (8) and is useful for interrogating dynamic processes, such as GPCR signaling (9, 10). Despite these evident advantages, existing PL methods are not without limitations. For example, some biotin ligases suffer from poor temporal resolution, as they require labeling for several hours; the engineering of TurboID has partially solved this problem by reducing the time needed for labeling down to 10 min (5). The peroxidase APEX2 requires much shorter incubation with H₂O₂ (30 to 60 s) to induce biotinylation; however, some applications cannot be accessed by APEX2 labeling due to the toxicity of H₂O₂.

The recent introduction of µMap, a photoproximity labeling (PPL) method, offers an attractive alternative to existing PL strategies (11, 12). This approach relies on the delivery of a chemical photocatalyst to a specific cellular site, e.g., through conjugation to an antibody. Selective targeting of the photocatalyst allows for the localized activation of the PL probe upon light irradiation, thereby tagging proximal proteins. The PPL approach has been shown to be efficient for studying cell surface biology and cell–cell interactions (13–15), as well as chromatin-associated PPIs in isolated nuclei (16). However, the reliance of PPL strategies on peptide- or protein-based delivery modalities generally impedes application within living cells, although cell-permeable photocatalysts can still be used intracellularly in cases where prior conjugation to a targeting protein is not required (17, 18). In addition, unlike with genetically encoded PL systems, supplementation of an exogenous photocatalyst could result in some off-target localization, increasing the PL background.

We, therefore, envisioned a PL system that is: i) fully genetically encoded and ii) triggered by visible light. Such an approach will benefit from the high spatiotemporal resolution and the benign nature of light activation while preserving the recognized advantages of genetically encoded systems, primarily the ability to perform the labeling with exquisite specificity within living cells. Here, we report such a method, termed “LITag” (Light-induced Interactome Tagging) (Fig. 1A). We demonstrate that this tool is broadly applicable and allows snapshots of protein neighborhoods to be reliably captured following irradiation times as short as a few seconds.

Fig. 1.

A genetically encoded approach for photoproximity labeling. (A) LOV*, an engineered LOV domain (PDB ID 7QF5), is fused to the POI. The FMN photocatalyst generates short-lived reactive species upon blue light irradiation, thereby tagging the POI interactome. (B) Light-dependent protein biotinylation in HEK293T cells expressing a LOV* fusion. Top—map of the DNA construct used. Bottom—western blot showing protein biotinylation with different irradiation times. Cells were incubated with BP (500 µM) for 30 min before blue light irradiation. (C) Local illumination of HEK293T cells expressing the LOV*-H2B fusion in the presence of BP (500 µM) results in spatially resolved photoproximity labeling. LOV* (green) was visualized by its intrinsic fluorescence. Biotinylation (orange) was visualized by staining with Neutravidin-Rhodamine Red-X. (Scale bar, 500 µm.) (D) Left—LOV* is targeted to various locations based on its fusion partners and can be visualized by its intrinsic fluorescence. A short pulse of blue light is sufficient to produce localized protein biotinylation. Right—fluorescence imaging of LOV* localization and biotinylation. Live-cell LITag labeling was performed by incubating the cells with BP (BA for nucleolus labeling) and irradiating for 1 to 3 s. LOV* (green) was visualized by its intrinsic fluorescence (anti-FLAG antibody staining was used to visualize LOV* on the plasma membrane). Biotinylation (red) was visualized by staining with Neutravidin-Rhodamine Red-X. Hoechst 33342 (blue) is a nuclear marker. (Scale bars, 10 µm.)

Results

A Genetically Encoded Approach for Fast Photoproximity Labeling.

When considering potential candidates for use in a genetically encoded PPL system, we were drawn to the LOV (Light-Oxygen-Voltage) domains, a family of small flavin mononucleotide- (FMN-) binding proteins that have found widespread use in the optogenetics area (19–23). Flavin derivatives are known to participate in single electron transfer (SET) reactions upon blue-light excitation into a triplet excited state (24–26). Indeed, antibodies conjugated to free flavin analogs have recently been used in cell surface PPL applications using phenol-based probes (15). We were especially interested in engineered versions of LOV domains originally developed as singlet oxygen (¹O₂) sensitizers for electron microscopy imaging applications and, more recently, for labeling nucleic acids (27–30). Importantly, transient absorption measurements indicate that the lifetime of the FMN triplet state (t_T) can be up to 30-fold longer in the context of the engineered LOV domains compared to the free cofactor (31), a feature that we anticipated would engender SET chemistry with exogenously added probes with suitable oxidation potentials.

To test this idea, we fused a series of engineered LOV domains to histone H2B and expressed these in HEK293T cells. The cells were then treated with a biotin–phenol (BP) probe and irradiated with blue light (SI Appendix, Fig. S1 A and B). Immunoblotting using a dye-labeled streptavidin revealed robust biotinylation of cellular proteins using a previously described mutant of an Arabidopsis thaliana phototropin LOV domain (termed SOPP in ref. 31, herein referred to as LOV*). Remarkably, labeling of cellular proteins could be detected with as little as 1 s irradiation when employing a suitable blue-light–emitting apparatus (Fig. 1B and SI Appendix, Fig. S1 C–F). Such short irradiation pulses did not induce significant toxicity in cells expressing LOV* fusions (SI Appendix, Fig. S1G). In addition, the light-triggered PL was accomplished in a spatially controlled manner, as demonstrated by employing local illumination with a photomask (Fig. 1C). Thus, the LOV*-based PPL approach (hereafter LITag) enables light-triggered interactome tagging with excellent spatiotemporal precision.

Besides BP, light-dependent labeling was also observed using a biotin-aniline (BA) probe, but not with aryl-azide or aryl-diazirine warheads (SI Appendix, Fig. S1H), which is consistent with an oxidative SET pathway for probe activation. Compared to BP, proximity labeling with BA is more efficient. However, since some LOV*-independent activation of BA upon light irradiation was observed (SI Appendix, Fig. S1H), we chose BP as the preferred probe unless noted otherwise. LITag produces a biotinylation signal comparable to that of APEX2 (SI Appendix, Fig. S1I). However, the light-triggered PL enables tuning the labeling levels easily by altering the irradiation time; APEX2, being activated by H₂O₂ addition, is less amenable to fine control of the labeling duration. During these studies, we observed that the pretreatment of cells with 1 mM H₂O₂ for as little as 60 s led to increased LITag labeling efficiency (SI Appendix, Fig. S1J). This effect was not observed using purified proteins in vitro (SI Appendix, Fig. S1K), suggesting that the impact of peroxide in cells is indirect, potentially involving the amelioration of a redox-sensitive cellular quencher of the labeling reaction. With respect to this, we found that the in vitro reaction was less efficient in the presence of thiols such as glutathione (SI Appendix, Fig. S1L). Alternatively, peroxide treatment may affect the cellular probe concentration by modifying the plasma membrane permeability (32). Either way, all PPL experiments described herein were performed in the absence of added peroxide unless otherwise stated.

Next, we explored the scope of LITag. The system was found to be compatible with a broad range of cell types (SI Appendix, Fig. S2A) and could be performed within a number of different cellular organelles and on the cell surface by using appropriate protein fusions (Fig. 1D and SI Appendix, Fig. S2 B and C). These studies highlight the potential of LITag for performing correlated cell imaging and proximity labeling. Indeed, we observed excellent spatial overlap between the intrinsic LOV* domain fluorescence for each fusion and the corresponding biotin immunofluorescence signal following BP labeling. We noted that the nucleolus could not be efficiently labeled using BP, despite clear targeting of the LITag fusion to this region (SI Appendix, Fig. S2B and C). However, this issue could be overcome by switching to the BA probe (Fig. 1D and SI Appendix, Fig. S2D). The discrepancy in nucleolar labeling efficiency between BP and BA might reflect differences in the cellular localization of the probes. Future studies, for example testing other LITag-active probes, will help clarify this issue.

We sought to investigate the molecular mechanism of protein labeling by LITag. We envisioned two possible reaction pathways, both arising from the triplet excited state of the FMN cofactor (Fig. 2A). First, the highly oxidizing triplet-excited FMN (E_1/2^red = 1.5 V vs. SCE, ref. 25) can react with BP or BA through SET, resulting in a radical species that labels electron-rich amino acids, such as tyrosine (SI Appendix, Fig. S3A). Alternatively, it can produce singlet oxygen through energy transfer (31). Singlet oxygen further reacts with histidine residues, forming an endoperoxide intermediate that can be trapped by nucleophiles (33), yielding labeled oxohistidine (SI Appendix, Fig. S3B). To test these hypotheses, recombinant LOV* was employed in an in vitro labeling of phosphorylase B and bovine serum albumin using BP or BA. The samples were subjected to trypsin digestion and liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. We identified multiple tyrosine residues modified with BP and BA (Fig. 2B and SI Appendix, Fig. S3C), consistent with the proposed SET pathway. We also identified oxohistidine modified with BA (Fig. 2C and SI Appendix, Fig. S3 B and C) supporting the singlet oxygen-dependent labeling pathway. BP modification of oxohistidine was not observed. Lastly, peptides containing either tyrosine or histidine as a potentially active residue were subjected to PPL with BP or BA (Fig. 2D and SI Appendix, Fig. S3D). Consistent with the phosphorylase B and bovine serum albumin labeling reactions, BA could comparably label both peptides, while BP showed robust tyrosine modification and only faint labeling of histidine. Overall, our data suggest that both the SET-based and the singlet oxygen-based mechanisms are operational, depending on the nature of the employed probe.

Fig. 2.

Mechanism of FMN-mediated PPL. (A) The photoexcited FMN is a strong oxidant that can participate in SET reactions with probe molecules, resulting in the labeling of tyrosine residues. Alternatively, it can produce singlet oxygen, which further reacts with histidine residues to form an endoperoxide intermediate that can be trapped by aniline, giving rise to modified oxohistidine. While an electrophilic aromatic substitution at the ortho carbon atom is depicted, the aniline nitrogen atom could also react as a nucleophile with the electrophilic endoperoxide intermediate; our MS data cannot discern between these two possibilities. ISC, intersystem crossing. (B and C) MS/MS spectra of representative modified peptides from rabbit phosphorylase B. In vitro LITag labeling was achieved by 5 s irradiation in the presence of LOV* (10 µM), phosphorylase B (1 mg/mL), and probe (BP or BA, 500 µM), dissolved in PBS, pH 7.4. Tyrosine modification is observed with either BP or BA (B), while histidine modification is observed only with BA (C). Additional MS/MS spectra are in SI Appendix, Fig. S8. (D) Biotinylation of 15-mer peptides containing a single active residue (Y or H) is evaluated by western blot. Peptides (1 mg/mL) were dissolved in PBS, pH 7.4, in the presence of BP or BA (500 µM) and FMN (10 µM) and irradiated for 5 s. Silver staining was used as a loading control, indicating equal loading of each of the peptides with the two different probes. Since replacing tyrosine with histidine significantly changes the staining propensity, silver staining was not useful in comparing the amount of loading among the two peptides. Therefore, dot blotting of the peptide solutions, followed by ponceau staining, was used as an additional loading control.

LITag with Quantitative Proteomics for Mapping Protein Territories.

Encouraged by the high temporal resolution and the broad scope of LITag, we next integrated it into a quantitative proteomics-MS workflow (Fig. 3A). We initially focused on the mitochondria, an organelle that has been extensively studied using existing PL approaches (5, 34) and that consequently provides an excellent opportunity to test the fidelity of our PPL approach. Efficient targeting of the LOV* domain to the mitochondrial matrix was achieved using a HEK293T cell line stably expressing a fusion with a mitochondrial targeting sequence (MTS) derived from COX4 (Fig. 3B). These cells were treated with BP for 30 min before being irradiated for different times. Immunoblotting revealed labeling after a few seconds of exposure to blue light (Fig. 3C), while immunofluorescence imaging of the irradiated cells indicated that this labeling was restricted to the mitochondria (Fig. 3B). For quantitative proteomics-MS analysis, we integrated LITag (with 3 s irradiation) into a stable isotope labeling by amino acids in cell culture (SILAC) workflow (SI Appendix, Fig. S4 A and B). This led to the identification of over 160 proteins that met our enrichment criteria (>1.5-fold change, FDR < 0.05) (Fig. 3D and SI Appendix, Fig. S4 C and D and Dataset S1). Gratifyingly, all of the enriched proteins are known to be mitochondrially localized based on human protein atlas annotation, as well as previous proteomics studies (Fig. 3E) (35). We also observed remarkable specificity with regard to submitochondrial localization. Since the radical intermediate is membrane-impermeant (34), only proteins in the matrix or the inner mitochondrial membrane (IMM) are expected to be labeled. Indeed, our list of hits does not include any proteins residing in the mitochondrial outer compartments—the intermembrane space (IMS) or the outer mitochondrial membrane (OMM) (Fig. 3E and Dataset S1). The submitochondrial localization of one hit, ARL2, is uncertain (36); the LITag data suggest that it occupies the matrix or the IMM. In complementary studies, we showed that the MTS fusion LITag system could also be used to enrich mitochondrial RNA, in this case employing an RT-qPCR readout (Fig. 3F).

Fig. 3.

LITag labeling of the mitochondria. (A) Workflow for LITag coupled with SILAC-based proteomics. HEK293T cells, stably expressing the MTS-LOV* fusion in the mitochondrial matrix, were cultured in either “heavy” or “light” media and incubated with BP. In the forward experiment, LITag was induced in the “heavy” cells by 3 s irradiation, while “light” cells were kept in the dark as a negative control. Biotinylated proteins were pulled down, digested, and subjected to LC-MS/MS analysis. (B) Top—MTS-LOV* is targeted to the mitochondria. Bottom—spatially restricted biotinylation of the mitochondria. LOV* (green) is visualized by its fluorescence. Mitochondria (orange) are labeled with MitoTracker Deep Red FM. Biotinylation (red) is visualized by Neutravidin-Rhodamine Red-X. Hoechst 33342 (blue) is a nuclear marker. (Scale bars, 10 µm.) (C) Western blot showing protein biotinylation with different irradiation times. (D) Analysis of SILAC data from the mitochondrial matrix LITag experiment. The Top histogram shows the Log₂(FC) distribution of the mitochondrial proteins in the data set. Red bars represent significant hits (>1.5-fold change, FDR < 0.05). The Bottom histogram shows the Log₂(FC) distribution of the non-mitochondrial proteins. FC, fold change. (E) Analysis of specificity. Top—1,600 proteins are detected by proteomics, most of which (over 80%) are not mitochondrial. The labeling specificity is excellent, as only mitochondrial proteins are enriched. Bottom—labeling is restricted to the mitochondrial matrix and the IMM, with no enrichment of proteins residing in the outer compartments. (F) LITag labeling of mitochondrial mRNAs. Biotinylated RNA molecules were pulled down following proximity labeling, and the enrichment of two mitochondrial mRNAs, MT-CO2 and MT-ND1, was evaluated by RT-qPCR. No enrichment was observed for the negative controls, XIST and GAPDH. Error bars = SD, n = 4. Data were analyzed using a one-sample t test, ***P < 0.001.

As a further test of the SILAC-LITag proteomics workflow, we generated a stable U2OS cell line expressing a doxycycline-inducible LOV* fused to poly(ADP-ribose) polymerase 1 (PARP1), an essential regulator of various DNA repair pathways that catalyzes the polymerization of ADP-ribose units, attaching poly(ADP-ribose) (PAR) chains to itself and other target proteins (Fig. 4A) (37). The LOV* fluorescence was exploited in a live-cell imaging experiment to track the recruitment of the fusion protein to DNA damage induced by laser micro-irradiation. Indeed, PARP1-LOV* is recruited rapidly to the line of DNA damage, as expected for a functional PARP1 protein (Fig. 4B and Movie S1) (38). Next, the cells were treated with BP for 30 min in the presence of the damage-inducing agent H₂O₂. Immunoblotting revealed significant protein tagging after only 1 s of blue light irradiation (Fig. 4C). Employing the SILAC-LITag workflow, we identified 99 hits (>1.5-fold change, FDR < 0.05) (Fig. 4D and SI Appendix, Fig. S5 A and B and Dataset S2). Most of these enriched proteins are known to be PAR binders, PARylation targets, or both (Fig. 4E). We observed a large number of RNA-binding proteins (Fig. 4D), consistent with the extensive involvement of RNA-processing factors in the DNA damage response (39). For example, the top hits on the list, FUS and EWSR1, are RNA-binding proteins that participate in DNA repair and localize to DNA breaks through PAR binding (SI Appendix, Fig. S5C) (40, 41). We were particularly intrigued by the high enrichment of HMGB1, a multifunctional chromatin-associated protein implicated in various DNA repair pathways (42). HMGB1 accumulates at sites of DNA damage (43); however, the mechanism for this process is not yet understood. We wondered if HMGB1 recruitment is dependent upon the synthesis of PAR chains at damage sites. Indeed, laser microirradiation experiments in U2OS cells expressing an eGFP-tagged HMGB1 show that treatment with PARP1 inhibitor significantly reduces HMGB1 accumulation, strongly implicating PARylation in HMGB1 recruitment at the site of DNA damage (Fig. 4F and SI Appendix, Fig. S5D). Collectively, these PARP1 data indicate that the LITag approach can be used to faithfully capture a protein neighborhood during a dynamic cellular process such as the DNA damage response.

Fig. 4.

Investigation of PARP1 neighborhood following DNA damage. (A) Treatment with genotoxic agents like H₂O₂ induces DNA lesions such as single-strand (ss) breaks, to which PARP1 is rapidly recruited, resulting in PARP1 activation and PAR synthesis. PAR signaling recruits various DNA repair factors, which are captured by LITag. (B) Live-cell imaging showing the recruitment of the PARP1-LOV* fusion to the site of DNA damage induced by laser microirradiation. Fluorescence images were taken at different time points following damage induction. (Scale bar, 5 µm.) (C) Western blot showing the light-dependent protein biotinylation in U2OS cells expressing PARP1-LOV*. (D) Volcano plot displaying the SILAC-LITag data. 99 proteins were enriched (orange and red), among which 54 are RNA-binding proteins (red). (E) Out of 99 significant hits, 86 are known to be PAR binders, PARylation targets, or both. The enriched proteins are grouped according to these features and sorted by their log₂ (FC), as represented by bar height. (F) Laser microirradiation and live-cell imaging of eGFP-tagged HMGB1 in U2OS cells. Left—fluorescence images of the cells following DNA damage. Talazoparib (250 nM) was used as a PARP1 inhibitor. (Scale bar, 5 µm.) Right—quantification of HMGB1 accumulation at the site of DNA damage. The plus sign denotes an image taken right after microirradiation. Error bars = SD, n = 21.

Having shown that LITag is a viable approach for PPL analyses, we next applied the strategy to characterize the interactome of the major vault protein (MVP), the primary component of the 13 MDa vault particle (44). MVP has been implicated in diverse cellular processes and is often associated with apoptosis resistance (45). However, the precise molecular mechanisms by which MVP contributes to cellular physiology remain poorly understood. In principle, characterization of the vault particle interactome by LITag could help bridge this knowledge gap. With this in mind, we fused LOV* to either the N or C terminus of human MVP and expressed the constructs in HeLa cells (Fig. 5 A and B). Differential centrifugation revealed that the N-terminal-tagged MVP (LOV*-MVP) is efficiently incorporated into vault particles, while the C-terminal-tagged MVP is not (Fig. 5C). Cells expressing LOV*-MVP were incubated with biotin-containing probes and irradiated for 3 s. Robust protein tagging was observed in the presence of BA, while the use of the BP probe afforded weaker labeling in this case (SI Appendix, Fig. S6A); therefore, we chose BA for this LITag proteomics experiment. This led to the identification of 78 proteins as potential MVP interaction partners (>1.5-fold change, FDR < 0.05) (Fig. 5D and SI Appendix, Fig. S6 B and C and Dataset S3). PARP4, a known component of the vault particle (46), was highly enriched, second only to MVP itself, further validating the incorporation of LOV*-MVP into vault particles. Many of the enriched proteins are linked to nuclear transport (Fig. 5E); this is consistent with previous literature suggesting nuclear trafficking as a major process employing the vault particle (47, 48). Interestingly, several of the hits are chaperones or co-chaperones (Fig. 5E and Dataset S3), implying a potential role for MVP in regulating protein folding. In this regard, the interaction of MVP with an HSP70 co-chaperone is implicated in apoptosis resistance (49). We asked whether these proteins interact with the whole vault particle or rather with the nascent MVP monomers. To answer this question, the native vault nanoparticles were isolated from soluble HeLa lysates using sucrose gradient centrifugation (46). Western blot analysis of the sucrose gradient fractions showed that BAG2, a nucleotide-exchange factor for the 70 kDa heat-shock proteins, cosediments with the vault particle proteins MVP and PARP4 (Fig. 5F). This result, which was further confirmed in a non-small cell lung cancer cell line (A549, Fig. 5G), supports the idea that MVP interacts with the folding regulators in the context of an intact vault particle. It remains to be seen whether these chaperones are actively engaged in folding client proteins while associated with the vault particle, versus being transported to where they are needed in the cell. Regardless, the MVP example indicates that the LITag approach is compatible with a large molecular particle, and, as such, may have utility in the study of analogous systems such as the assembly of viral capsids in infected cells.

Fig. 5.

Studying the interactome of the major vault protein (MVP). (A) Right—the structure of the vault particle (PDB ID 4V60). A vault consists of two identical half-vaults, which align at their waists to form a barrel-like structure; each half-vault comprises 39 MVP monomers. A single monomer within the upper half-vault is highlighted in blue. Left—the small LOV* domain is fused to the N terminus of MVP. (B) Live-cell imaging of LOV*-MVP expression in HeLa cells. LOV* (green) is visualized by its fluorescence. Hoechst 33342 (blue) is a nuclear marker. (Scale bar, 5 µm.) (C) Differential centrifugation separates whole vault particles from soluble cytosolic proteins. Top—workflow: following cell lysis, the chromatin fraction is removed by centrifugation, providing a clear soluble fraction (WCL). The latter is subjected to ultracentrifugation; individual proteins and small complexes remain in the supernatant (Sup), while the vault particles pellet with the microsomal fraction (Pell). Bottom—western blot showing the incorporation of the exogenously expressed MVP constructs in the vault particle. The C terminus-tagged MVP remains largely in the supernatant, suggesting that it is not incorporated in the vault particles. On the other hand, the N terminus-tagged MVP (LOV*-MVP) is mainly in the pellet, similarly to the endogenous MVP, implying its efficient incorporation in the vault. (D) Volcano plot displaying the SILAC-LITag data. HeLa cells, transfected with exogenous MVP missing the LOV* domain, were used as the negative control. Seventy-eight proteins were enriched (blue and red)—the vault proteins MVP and PARP4 are the top hits; next is BAG2, a co-chaperone of the HSP70 proteins. (E) Enriched proteins were categorized based on their molecular function. Many hits are involved in nuclear transport and protein folding regulation. (F and G) Sucrose gradient centrifugation was performed to isolate the native vault particles from soluble cell lysate. Western blot analysis shows that BAG2 co-sediments with the vault particles. Two different cell lines were used: F—HeLa; G—A549.

Discussion

We have demonstrated that an engineered LOV domain can be deployed as a genetically encoded photoproximity labeling tool. We show that protein tagging with exogenous probes can operate through redox-based or energy transfer-based mechanisms, depending on the nature of the probe. We further illustrate how the combination of LITag with quantitative proteomics generates spatially defined “snapshots” of protein neighborhoods. This capability is showcased through the characterization of PARP1 neighborhood during DNA damage and of the protein clients of MVP, applications highlighting the power of the method for studying dynamic cellular processes and large macromolecular assemblies, respectively. A unique feature of LITag is its outstanding temporal resolution (labeling in a few seconds), which could be important for studying dynamic biological processes that operate under short timescales. In addition, minimizing the PL duration is imperative for preventing artifactual results caused by non-physiological changes during the protein labeling process, such as structural destabilization (50). Finally, the intrinsic fluorescence of the LOV* fusion domain permits correlated imaging and proteomics analyses, a unique capability compared to existing genetically encoded PL approaches.

Shortly after this work was published as a preprint, two related studies appeared, describing a light-dependent PL method using the FMN-binding variant, miniSOG (51, 52). Along with the similarities between the approaches, LITag stands out for its superior temporal resolution, as it accomplishes labeling in a few seconds compared to several minutes. Furthermore, while the other papers describe an exclusive singlet oxygen-dependent labeling pathway for oxohistidine modification, we reveal an orthogonal redox-mediated labeling pathway that results in the modification of tyrosine. This mechanistic understanding provides means to orient protein tagging towards various residues through the choice of the exogenous probe employed (phenol vs. aniline) and can be important for designing new probes for PPL.

We view the LITag approach as complementary to existing PL and PPL approaches, albeit one that combines the advantages of a genetically encoded system with the spatial–temporal control afforded by an optical trigger. For example, spatially resolved illumination can be particularly advantageous for studying asymmetric systems with non-uniformly distributed PPIs. While the current study involved a repurposed LOV domain mutant, we imagine that further optimization of the system may be possible through directed evolution approaches, perhaps in conjunction with the use of additional cell-permeable probes with suitably altered redox properties. Among the more exciting prospects that our approach offers for the future is the application of the PPL strategy in an optically transparent model organism. Future efforts from our group will be directed along these lines.

Materials and Methods

This section describes only the materials and methods relevant to the primary LITag workflow, i.e., PPL in living cells and subsequent processing by western blot, immunofluorescence or quantitative proteomics. A full description of the materials and methods used in this study is in the SI Appendix.

Cell culture.

HEK293T, U2OS, HeLa, A549, and HCT116 were cultured in Dulbecco's Modified Eagle Medium (DMEM, Thermo Fisher 11995-005), supplemented with 10% v/v FBS (Atlanta Biologicals S12450H), 100 U/mL penicillin, 100 µg/mL streptomycin (ThermoFisher 15140-122), and 2 mM L-Glutamine (Thermo Fisher 25030-081). Cells were maintained in an incubator at 37 °C with 5% CO₂. For HEK293T, plates were pre-coated with 0.01 % poly-L-lysine solution (Sigma P8920).

LITag labeling.

To perform LITag experiments, cells were washed twice with PBS (137 mM NaCl, 12 mM Phosphate, 2.7 mM KCl, pH 7.4) or Live Cell Imaging Solution (ThermoFisher A14291DJ). A solution of biotin-phenol (BP, ApexBio A8011, 500 µM, 500× dilution from a 250 mM solution in DMSO) or biotin–aniline (BA, Iris Biotech LS-3970, 250 µM, 1,000× dilution from a 250 mM solution in DMSO) in Live Cell Imaging Solution was added, and the cells were incubated for 30 min at 37 °C under 5% CO₂. Following probe incubation, cells were irradiated using two Kessil PR160L lamps (440 nm, 450 mW/cm²) for the indicated time. Subsequently, cells were washed twice with PBS, and further processing was performed.

Western Blot.

Following LITag labeling, 1×SDS sample loading buffer (100 mM Tris-Cl, pH 6.8, 3% SDS, 15% Glycerol, 2.25% β-mercaptoethanol, 0.015% bromophenol blue) was added to lyse the cells. Samples were boiled at 98 °C for 30 mins and were subsequently loaded on a 10% bis-tris polyacrylamide gel. Proteins were transferred to a nitrocellulose membrane which was then blocked with 5% w/v nonfat dry milk in TBST (20 mM Tris-Cl, pH 7.6, 150 mM NaCl, 0.1% v/v Tween-20) for 30 mins at room temperature. Next, membranes were incubated with the indicated primary antibodies (diluted in TBST containing 1% BSA) at room temperature for 2 h or at 4 °C overnight. After washing three times with TBST, the appropriate LI-COR IRDye secondary antibodies (1:10,000 dilution in TBST) were applied for 1 h at room temperature. For the detection of the biotin signal produced by LITag, membranes were incubated with an IRDye® 800CW-labeled streptavidin (LI-COR 926-32230, 1:10,000 dilution in TBST) for 1 h at room temperature. Following incubation with the dye-labeled secondary antibodies, membranes were washed three times with TBST and imaged with LI-COR Odyssey Infrared Imaging System.

Immunofluorescence.

Cells were cultured in either a 35 mm glass-bottom petri dish (MatTek P35G-1.5-14-C) or a 24-well glass-bottom plate (Cellvis P24-1.5H-N). LITag labeling was performed at 30 to 40% confluency as described above. Subsequently, cells were fixed with 4% formaldehyde for 10 min, then permeabilized with 0.5% Triton in PBS for 15 min at room temperature. 3% BSA or 2.5% goat serum in PBS were used for blocking for at least 1 h at room temperature. For experiments that only need biotin staining, cells were incubated with NeutrAvidin™-Rhodamine Red™-X (2 µg/mL, ThermoFisher A6378) and Hoechst 33342 (10 µg/mL) for 1 h at RT, washed three times with PBS, and imaged using a confocal microscope. When staining with a primary antibody was also needed, the antibody was incubated with the cells overnight at 4 °C, followed by staining with a secondary antibody, NeutrAvidin™-Rhodamine Red™-X, and Hoechst 33342. Cells were washed three times with PBS and imaged using a confocal microscope. More information regarding the microscope specifications is in the SI Appendix.

SILAC-Based Proteomics.

Cells were cultured in “DMEM for SILAC” (ThermoFisher 88364) supplemented with 10% dialyzed FBS (ThermoFisher A3382001), 100 U/mL penicillin, 100 µg/mL streptomycin (ThermoFisher 15140-122), and the appropriate amino acids: “heavy” (H) media—L-lysine-¹³C₆,¹⁵N₂ dihydrochloride (Cambridge Isotope Laboratories (CIL) CNLM-291-H) and L-arginine-¹³C₆,¹⁵N₄ hydrochloride (CIL CNLM-539-H); “medium” (M) media—L-lysine-4,4,5,5-d₄ dihydrochloride (CIL DLM-2640) and L-arginine-¹³C₆ hydrochloride (CIL CLM-2265-H); “light” (L) media—L-lysine hydrochloride (Sigma L1262) and L-arginine hydrochloride (Sigma A3909). Cells were cultured in SILAC media for at least eight doublings to ensure complete isotopic labeling. For the LITag labeling step, cells were seeded in six-well plates; labeling was performed at 95 to 100% confluency. We used a total of 2 × 10⁷ cells per experimental replicate (1 × 10⁷ cells for each SILAC condition). After LITag labeling (as described above), the cells were washed twice with PBS, followed by an additional wash in PBS for 5 min to remove the excess of probe. The cells were incubated with lysis buffer (10 mM tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.5% sodium lauryl sarcosine, 0.1% sodium deoxycholate, 1× Halt protease inhibitor cocktail, ThermoFisher 78438; 150 µL/well) for 10 min at RT and were subsequently transferred to a 5 mL Eppendorf tube. The cell lysate was sonicated on ice using Fisherbrand 505 Sonic Dismembrator (3 mm microtip, 25% amplitude, 4 × 15 s), and the solution was clarified by centrifugation (10 min at 4 °C, 17,000 g). The protein concentration was determined by BCA (1 to 2 mg/mL), and samples were diluted with the appropriate volume of lysis buffer so that the concentration of the different L/M/H samples was even. SILAC samples were mixed 1:1 and diluted with one volume of binding buffer (0.25% NP-40 alternative in TBS, pH 7.6). Streptavidin magnetic beads (Streptavidin Mag Sepharose, Cytiva 28985799, 300 µL slurry), pre-equilibrated with binding buffer, were added, and the samples were rotated head-over-head for 2 h at RT. Subsequently, the beads were collected and washed with binding buffer (once), 1% SDS in PBS (twice), 1 M NaCl in PBS (twice), and 100 mM ammonium bicarbonate (three times). The beads were suspended in 300 µL of 5 mM DTT in 100 mM ammonium bicarbonate and mixed at 56 °C for 30 min. The suspension was cooled down to RT, and 9 µL of 500 mM iodoacetamide in water was added, followed by rotation at RT for 20 min in the dark. The beads were washed three times with 50 mM ammonium bicarbonate and then resuspended in 300 µL 50 mM ammonium bicarbonate containing 3 µg of trypsin gold (Promega V5280). Following rotation at 37 °C overnight, the supernatant was collected and dried using a SpeedVac vacuum concentrator. The samples were then subjected to LC-MS/MS analysis. More information regarding the MS operation and data processing is in the SI Appendix.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Movie S1.

The PARP1-LOV* fusion is recruited rapidly to the site for DNA damage. Live-cell imaging of the PARP1-LOV* green fluorescence, tracking the recruitment of the fusion to the line of DNA damage in real-time.

Data, Materials, and Software Availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD035265 (53). A python script for missing-data imputation of the proteomics datasets was deposited in Zenodo, DOI: 10.5281/zenodo.6883783 (54).