Characterisation of photo-crosslinking organoiridium catalysts. Chemical reactions that explore PPIs have been focused on photo-crosslinking by metal complexes (Ru(II), Pd(II), and Ir(III) complex) owing to its ability to make covalent connections between interacting proteins through zero-length oxidative coupling regardless of the labelling radius.29–31 Among these metal complexes, the Ir(III) complex can act as a photo-crosslinking catalyst in living cells without any catalysis-initiating additives, such as ammonium persulfate (Figs. 1a-b).37,38 Moreover, our Ir(III) photocatalyst exhibits the capability to facilitate multiple photo-crosslinking reactions through catalytic reaction cycle. In contrast, the photoaffinity labelling (PAL) methods rely on a single photo-conversion reaction that generates carbene or nitrene species from alkyl diazirine or arylazide probes, respectively. (Fig. 1b and Supplementary Fig. 1b). With these advantages in mind, we developed two different HaloTag ligand-conjugated Ir(III) complexes (Ir-HTL and sIr-HTL) as photo-crosslinking reagents by functionalizing the HaloTag ligand for specific targeting of intracellular POIs and subsequently used them for spatially resolved interactome mapping (Figs. 1a–b and Supplementary Figs. 2–9). Ir-HTL and sIr-HTL were rationally designed to control triplet excited state lifetime by changing the distance between the ancillary ligand bipyridine (bpy) and the carbonyl group that acts as the excited state quencher (Supplementary Scheme 1).38,39 As expected, time-correlated single photon counting in an aqueous solution demonstrated the triplet excited lifetime of Ir-HTL (704 ns) to be approximately two times longer than that of sIr-HTL (363 ns) because of the large distance (7.430 Å) between the carbonyl group and the bpy part of Ir-HTL (Fig. 1c).39,40 Results showed the emission intensity of Ir-HTL to be higher than that of sIr-HTL and the ultraviolet visible absorption spectra of both reagents to be similar (Fig. 1d). The addition of purified HaloTag protein led to an increase in emission intensity and excited state lifetime of all Ir(III) complexes owing to the hydrophobicity of the bound protein (Fig. 1e and Supplementary Fig. 10).41 The emission intensity of all Ir(III) complexes was almost saturated at the equivalent point of HaloTag protein (Fig. 1e and Supplementary Fig. 10), indicating a 1:1 binding ratio be-tween the protein and the Ir(III) complex. In addition, similar binding kinetics showing a 1:1 binding ratio were observed for binding between the HaloTag protein and Ir-HTL as well as sIr-HTL (Fig. 1f). Two Ir(III) complexes had the same optical properties except for their triplet excited state lifetime. Therefore, we expected that the longer triplet excited state lifetime of Ir-HTL would result in superior electron transfer capability to enhance the photo-crosslinking reaction.
Dihydrorhodamine 123 was used for detecting type I reactive oxygen species (ROS) produced by electron transfer to the oxygen, which is one of the biomolecules. Ir-HTL showed significantly higher type I ROS generation ability than sIr-HTL owing to its long excited-state lifetime (Supplementary Fig. 11). Next, the enhancement of the photo-crosslinking reaction was scrutinized in HEK293T cells. We initially verified the intracellular binding of Ir-HTL to the HaloTag protein by a competition assay with biotin-HTL (Supplementary Fig. 12). In the competition assay, the signal of biotin binding to the HaloTag protein decreased depending on the concentration of pre-incubated Ir-HTL. This reduction in the biotin binding signal indicates covalent binding of Ir-HTL to the HaloTag protein. Subsequently, two model constructs (GBP-NLS-HaloTag and LAMIN A/C-HaloTag) were used to understand the correlation between the triplet excited lifetime of the Ir(III) complexes and photo-crosslinking ability. Photo-irradiation successfully revealed the presence of bands of proteins photo-crosslinked by Ir(III) complexes (Fig. 1g and Supplementary Fig. 13). The efficiency of the crosslinking reaction was determined in terms of the correlation value (R2) indicating similarity between the two spectra (with or without light) that was obtained by a quantitative analysis using the line-scan spectra to high crosslinking efficiency; detailed calculations are represented in the Supplementary Materials). Ir-HTL was seen to have higher crosslinking efficiency than that of sIr-HTL for the complexes being located on the nucleus as well as on the nuclear membrane (Fig. 1i and Supplementary Fig. 13d). This result verifies that photo-crosslinking capability is substantially correlated with the triplet excited state lifetime of the photo-crosslinking catalyst.
To probe the direct photo-crosslinking reaction, we performed experiments in vitro as well as in living cells using the FKBP-rapamycin binding protein (FRB)-FK506 binding protein (FKBP) model system (Supplementary Fig. 14).42 The FRB and FKBP are capable of facing each other in the presence of rapamycin, which leads to direct contact between them (Supplementary Figs. 14a–c).43 Photo-irradiation revealed the significant presence of a new photo-crosslinking band below 100 kDa that was not detectable without the addition of rapamycin (Supplementary Fig. 14d; red asterisk mark). This newly generated band corresponds to the FKBP12-EGFP (41 kDa)–HaloTag-FRB (52 kDa) crosslinked protein complex. In addition, the crosslinking signal was not detected with the FKBP12-EGFP complex in the absence of HaloTag construct, which indicates that the crosslinking reaction between interacting proteins occurs only at the direct contact point of Ir-HTL. Additional crosslinking analysis performed for living cells using a similar model system (FRB-EGFP; 40 kDa and FKBP25-V5-HaloTag; 63 kDa) revealed the presence of a newly produced crosslinking band upon treatment with rapamycin. The new crosslinking band was detected by binding of α-GFP antibody to FRB and α-V5 antibody to FKBP25 in a manner similar to that of the analysis of the in vitro process described previously (Supplementary Fig. 15, a red asterisk mark). Taken together, these results confirm that the direct photo-crosslinking reaction can precisely detect the intracellular interactome of the POI.
Spatially resolved localisation and photo-catalytically recyclable protein crosslinking in living cells. PPIs were analysed by using photo-crosslinking inside the nucleus considering the challenging nature of chemically targeting the undruggable classes of proteins, such as transcription factors and regulatory proteins, present within it.44 In addition, the presence of various subcellular compartments (e.g., nucleoplasm, nucleolus, and nuclear bodies) within the nucleus prompted us to evaluate the spatial resolution for PPIs using this photocatalyst. We selected PTBP1, POU2F1 (an undruggable target), and PSMA2 as three POIs, respectively consisting of the spliceosome, heterochromatin condensate, and 26S proteasome in the nucleus (Fig. 2a). Specific relocation of the photocatalyst to the desired location in the nucleus was done using a GBP nanobody system. GBP is a representative nanobody, an antigen-binding fragment from an antibody, which can target GFP (Fig. 2b). Using the nanobody system with POI-EGFP and GBP-HaloTag-NLS, we, therefore, devised the new proteomic mapping method, POINT (Fig. 2c, and Supplementary Table 1). This method involved the covalent labelling of GBP-HaloTag-NLS by Ir-HTL treatment subsequent to the expression of GBP-HaloTag-NLS and POI-EGFP (Fig. 2c: i). This enabled the precise localisation of the Ir-HTL connected to GBP-HaloTag-NLS depending on the location of the POI that is tagged with GFP (Fig. 2c: ii). The localisation of Ir-HTL in a specific region thus allowed for the interactome to be captured by photo-crosslinking (Fig. 2c: iii). In this workflow, we considered the experimental condition of only GBP-HaloTag being expressed in cells without GFP-POI as the reference control for implementing a strict cut-off in the proteomic analysis.
Examination of co-localisation of proteins in confocal images of HEK293T cells for spatially resolved proteomic analyses of the GFP-GBP system validated the specific localisation of Ir-HTL by this system (Fig. 2d). Ir-HTL is largely localised on the mitochondria without expression of GFP-GBP system because of localisation tropism for iridium complexes (Supplementary Fig. 16)45,46, and expression of GBP-HaloTag-NLS makes Ir-HTL localised on all across the nucleus with driving force by ligand recognition of Ir-HTL to HaloTag protein (Fig. 2d). Each nuclear compartment displayed characteristic structures, and the signals of Ir-HTL showed a notable overlap with expression of POU2F1-EGFP, PTBP1-EGFP, or PSMA2-EGFP. Line-scan spectra from the signals of EGFP and Ir-HTL were recorded for precise inspection of targeting ability. These results indicate that the GFP-GBP system is capable of specific targeting of Ir-HTL to the respective POI. The photo-crosslinking on the four specific locations was further validated by immunostaining with α-V5 (for GBP-HaloTag-NLS; Fig. 2e) and α-GFP (for POI-EGFP; Supplementary Fig. 17) antibodies. Each pattern of photo-crosslinking bands was associated with distinctive spectra depending on the location of Ir-HTL. For the detailed profiling of distinctive crosslinking signals, each line-scan spectrum obtained as a result of crosslinking of the respective POI was cross-checked with their correlation value (Supplementary Fig. 18).
With the above spatially resolved photo-crosslinking of Ir-HTL, we sought to verify the catalytic recyclability of this photo-crosslinking reaction to enable detailed protein-protein interaction network analysis. For this, we compared our POINT method with Spotlight method which doesn’t have reaction recyclability (vide supra; Fig. 1b and Supplementary Fig. 1b).42 Through the catalysed reaction, the photo-crosslinking intensity was increasingly enhanced in time-dependent manner and almost saturated in 10 min irradiation (Supplementary Fig. 19). With the recyclability of POINT method, the crosslinking efficiency of it was remarkably higher (16%) than that of Spotlight (4%) under the condition of GBP-HaloTag-NLS and PTBP1-GFP co-expression although crosslinking efficiency is almost close each other in 2 min irradiation. Based on this information, we set 10 min irradiation as a photo-crosslinking condition for following proteomic analysis. In addition to recyclability, cell toxicity evaluation showed suitability by maintaining the viability around 100% before and after 10 min irradiation (Supplementary Fig. 20). It verifies that our iridium photocatalyst can be one of the photocatalysts that enable to reflect biological context like photocatalyst of µMap-based TargetID has no toxicity followed by successful target identification of drug conjugate under physiological condition.23 Therefore, the crosslinking results taken together provide plausibility to enable drawing landscape of intact intracellular PPIs with enhanced photo-crosslinking and no effect on cell viability.
Proteomic identification of subnuclear interactome by POINT. To establish the reliability of the POINT method, we first analysed the protein interaction network of PTBP1, which is known to be a key regulator of mRNA splicing and an RNA chaperone. We optimised the sampling process for the mass spectrometry analysis of the interactome of PTBP1 as the first target POI in this study (POI-#1). POI-EGFP and GBP-HaloTag-NLS were co-expressed in HEK293T cells followed by in vitro biotinylation of AviTag fused with GBP-HaloTag-NLS (Fig. 3a; top). Photo-crosslinked proteins constituting the in vitro biotinylated bait were selectively acquired by streptavidin (SA) magnetic bead enrichment owing to its extraordinarily high affinity to biotin (Kd ≈ 10 − 14 mol/L) compared to that of other conventional epitope tag antibodies (e.g., anti-Flag M2 antibody, Kd ≈ 6.5 − 9 mol/L). Instead of conventional treatment with 8 M urea for washing of physically adsorbed proteins followed by denaturation, we used sodium dodecyl sulphate (SDS) to prevent the critical loss of proteins that occurred due to their excessive dissociation from the SA beads. Biotinylated proteins did not detach from the SA beads when washed with SDS, and non-specific binding proteins could thus be efficiently removed (Supplementary Fig. 21).
Mass analysis of the crosslinked interactome of PTBP1 by POINT showed a total of 47 enriched proteins, including known interactors, such as RAVER1, MATR3, and SFPQ (Fig. 3b; left, Supplementary Fig. 22a, and Supplementary Data 1). Among the enriched proteins, RAVER1 was ranked the highest in terms of its interaction among the proteins in the interactome of PTBP1, which is in agreement with the well-known ability of the PRI3 motif of RAVER1 to bind to the RRM2 motif of PTBP1 and to act as a PTBP1 corepressor for splicing.47,48 MATR3 is an RNA and DNA binding protein located in the nuclear matrix and its GILGPPP motif binds to the RRM2 motif of PTBP1 for alternative splicing.49 Newly formed long interspersed nuclear elements utilize the interaction between MATR3 and PTBP1 to repress splicing for the purpose of RNA insulation.50 SFPQ, a known PTBP1 binding partner, is an essential factor for pre-mRNA splicing.51 These findings, taken together with our results, collectively support the capability and specificity of our method for the identification of interactomes.
Furthermore, to demonstrate the significance of the interactome analysis performed using POINT, we compared our results showing the enriched proteins to those obtained by TurboID, a popular proximity labelling enzyme used for interactome analysis by the biotinylation of lysine residues (Fig. 3a; bottom, and Supplementary Data 2).13 In this analysis, we also utilized the PTBP1-EGFP construct with GBP-TurboID-NLS which led to the identification of more enriched interactors in the PTBP1-GFP:GBP-TurboID-NLS sample compared to those in the GBP-TurboID NLS sample (Fig. 3b; right). Our POINT method revealed 29 interactors that were not detected by TurboID (Fig. 3c; blue box). We attribute this result to the ability of POINT, which is free from labeling radius, to target the tyrosine residue via direct oxidative coupling to make di-tyrosine; whereas, the labeling modality of TurboID is based on targeting lysine residues via biotinylation (Fig. 3a). Furthermore, dynamic lysine acetylation of proteins frequently occurs in splicing- and translation-associated proteins.47 The lysine modality of TurboID does not easily identify with precision the landscape of the splicing interaction network of PTBP1 when the lysine of the interactome is acetylated. This is the reason that POINT can enrich various interactomes in the context of RNA splicing processes by targeting tyrosine residues in the interactome. Consequently, FUBP1, SET, and DAZAP1, which are associated with dynamic lysine acetylation, were only detected by POINT.52,53 Moreover, we also detected unknown interactors of PTBP1, such as HCFC1, SET, SIRT1, ANP32B, and SUPT5H, involved in histone modifications, the correlation of which with the regulation of alternative splicing has been highlighted in previous studies, despite its pathways and other associated factors not being fully characterized yet.54–56 These enriched proteins imply the existence of moonlighting functions of splicing regulation which are required to be confirmed by further studies. Comparative analysis with POINT and TurboID methods collectively indicates the potential of POINT for establishing a detailed PPI network, which can be attributed to the high precision resulting from direct-contact labelling reaction and its modality that differs from that of the conventional TurboID method. The following network clustering for the proteins identified by POINT categorizes them into three subgroups: mRNA splicing, cellular response to stress, and nucleic acid binding; this corresponds well to the functionality of PTBP1 (Fig. 3d). Among the three categories, mRNA splicing had the highest fidelity in gene ontology (GO) annotation analysis in terms of biological process, which verified the reliability of the identification of the PTBP1 splicing network (Fig. 3e).
We further applied the POINT method to reveal the subnuclear interactomes of the heterochromatin condensate POU2F1 and 26S proteasomal subunit PSMA2 for transcription and proteolysis, respectively. We identified a total of 21 and 16 proteins as potential interactors of POU2F1, and PSMA2, respectively (Supplementary Figs. 22b–c and Supplementary Data 1). Results showing scatter plots, Pearson’s correlation value, and distribution of subcellular localisation for enriched interactomes are provided (Supplementary Figs. 23–24). However, the majority of the identified interactomes of POU2F1 and PSMA2 have not yet been experimentally verified. Therefore, we preferentially selected the potential interactomes based on the molecular function of each POI. POU2F1 (POI-#2) has cooperativity with the TATA box binding protein which is involved in transcription-coupled DNA double strand break repair (DSBR).57,58 PRPF19, which showed the highest enrichment score in the analysis of POU2F1 by POINT, functions as a DNA interstrand crosslink-inducing agent involved in DSBR.59 Interestingly, the SETD1A histone methyl transferase, a key interactor of PRPF19 involved in DSBR,60 was concurrently enriched. PSMA2 (POI-#3) is known to have cooperativity with the INO80 chromatin remodeling complex,61 and RUVBL2 in the interactome of PSMA2 enriched by POINT is involved in the formation of INO80 for positioning the nucleosome.62
With these considerations, we selected PRPF19 and RUVBL2 as potential interactors of POU2F1 and PSMA2, respectively. We further validated the interaction (PTBP1–RAVER1) and the newly identified subnuclear interactions (POU2F1-PRPF19, and PSMA2–RUVBL2) of 3 POIs by POINT through co-localisation and photo-crosslinking of interactors (Fig. 4). The protein crosslinking was cross-checked for each interactor, and the results showed clear signals of crosslinking for RAVER1, PRPF19, and RUVBL2 when the photocatalyst targeted PTBP1, POU2F1, and PSMA2, respectively (Figs. 4a–c, left). The imaging of signals recorded from POI-EGFP, the photocatalyst, and selected interactors showed substantial overlaps (Figs. 4a–c; co-localisation). Moreover, we performed high-resolution imaging with the Airyscan mode for a precise analysis of co-localisation (Supplementary Fig. 25). RAVER1, which is an interactor of PTPB1, showed a higher Pearson’s correlation value with PTBP1 compared to that with POU2F1 and PSMA2, thus indicating a substantial PPI between PTBP1 and RAVER1. To validate the crosslinking efficiency of each interactor, we confirmed the photo-crosslinking reaction for each interactor depending on the targeted subnuclear protein of the photocatalyst (Figs. 4d–f and Supplementary Fig. 26). In line with Pearson’s correlation values from co-localisation, the photo-crosslinking intensity by immunostaining of each identified interaction partner was the highest under the POINT reaction condition with Ir-HTL incubation (PTBP1–RAVER1; Fig. 4d, POU2F1–PRPF19; Fig. 4e, and PSMA2–RUVBL2; Fig. 4f).
Next, we tested possibility of artifacts in GFP–GBP system. For this experiment, we prepared and expressed PTBP1-HaloTag protein in HEK293 cells, and Ir-HTL was directly targeted to the PTBP1-HaloTag protein in the cells followed by initiation of POINT reaction with light. In this reaction, RAVER1 was still shown as the photo-crosslinked interacting partner of PTBP1-HaloTag under the POINT reaction condition (Supplementary Fig. 27). This result supported that GFP-GBP system can nicely capture the interactome of POI of POI-GFP construct. Collectively, these results demonstrate the capability of POINT to identify spatially resolved PPIs within subcellular compartments comprising subnuclear domains.
In brief, we established a proteomic method based on photo-crosslinking by an organoiridium catalyst for intracellular interactome mapping (POINT) having wide applicability and high precision. POINT shares a similarity with other photocatalytic proximity labelling methods such as MAURA21 and µMap22–24, in terms of their use of photoactivable heavy metal catalysts (e.g. Ru or Ir-based catalyst) for chemical modification of the spatial proteome. However, each method employs distinct modification mechanisms. MAURA involves histidine oxidation through singlet oxygen, and µMap utilizes C-H insertion via nitrene generation, while POINT achieves tyrosine-tyrosine crosslinking through phenoxyl radical generation. As a result, we anticipate that POINT can be potentially combined with the MAURA or µMap labelling methods to construct more comprehensive and detailed spatial protein maps in the future. This technology can be utilized as a promising proteomic tool to analyse protein-protein interactions and allow for the detection of an expanded range of proteins that have been previously difficult to access.