Introduction

Proximity labeling technologies provide the biological and chemical sciences with various applications to study nucleic acids, lipids and proteins[1](#ref-CR1 “Qin, W., Cho, K. F., Cavanagh, P. E. & Ting, A. Y. Deciphering molecular interactions by proximity labeling. Nat. Meth. 1–11 https://doi.org/10.1038/s41592-020-01010-5

(2021).“),[2](#ref-CR2 “Kang, M.-G. & Rhee, H.-W. Molecular spatiomics by proximity labeling. Acc. Chem. Res. https://doi.org/10.1021/acs.accounts.2c00061

(2022).“),3,[4](https://www.nature.com/articles/s41467-025-65405-8#ref-CR4 “Becker, A. P. et al. Lipid- and protein-directed photosensitizer proximity labeling captures the cholesterol interactome. bioRxiv 2024.08.20.608660 https://doi.org/10.1101/2024.08.20.608660

(2024).“). Conceptually, proximity labeling uses enzymes fused to proteins of interest or directed to specific locations to determine their molecular environments by purifying proximity labeled biomolecules. Combining this approach with mass spectrometry-based proteomics allows unbiased and systematic determination of transient protein interactions as well as (sub)proteomes of organelles or cellular microcompartments that are inaccessible to biochemical purification methods5,[6](https://www.nature.com/articles/s41467-025-65405-8#ref-CR6 “Liu, X., Salokas, K., Weldatsadik, R. G., Gawriyski, L. & Varjosalo, M. Combined proximity labeling and affinity purification−mass spectrometry workflow for mapping and visualizing protein interaction networks. Nat. Protocol. 1–30 https://doi.org/10.1038/s41596-020-0365-x

(2020).“).

The primary cilium is a solitary plasma membrane microdomain with important functions in developmental biology and tissue maintenance[7](https://www.nature.com/articles/s41467-025-65405-8#ref-CR7 “Mill, P., Christensen, S. T. & Pedersen, L. B. Primary cilia as dynamic and diverse signalling hubs in development and disease. Nat. Rev. Genet. 1–21 https://doi.org/10.1038/s41576-023-00587-9

(2023).“),8. It functions as a specialized signaling compartment that translates extracellular cues into cellular responses by intricate mechanisms, employing second messengers and dynamic protein transport to and from the primary cilium[9](https://www.nature.com/articles/s41467-025-65405-8#ref-CR9 “Hilgendorf, K. I., Myers, B. R. & Reiter, J. F. Emerging mechanistic understanding of cilia function in cellular signalling. Nat. Rev. Mol. Cell Biol. 1–19 https://doi.org/10.1038/s41580-023-00698-5

(2024).“),[10](https://www.nature.com/articles/s41467-025-65405-8#ref-CR10 “Moran, A. L., Louzao-Martinez, L., Norris, D. P., Peters, D. J. M. & Blacque, O. E. Transport and barrier mechanisms that regulate ciliary compartmentalization and ciliopathies. Nat. Rev. Nephrol. 1–18 https://doi.org/10.1038/s41581-023-00773-2

(2023).“). Defects in these processes have been implicated in syndromic disorders, termed ciliopathies, affecting several tissues and cell types[7](https://www.nature.com/articles/s41467-025-65405-8#ref-CR7 “Mill, P., Christensen, S. T. & Pedersen, L. B. Primary cilia as dynamic and diverse signalling hubs in development and disease. Nat. Rev. Genet. 1–21 https://doi.org/10.1038/s41576-023-00587-9

(2023).“). A current hypothesis posits that cell type-dependent differences in the composition of primary cilia account for the pleiotropy of these syndromic disorders, as cilia dysfunctions in specific signal transduction mechanisms may have cell type- and tissue-specific consequences. Due to its small size (~1:10,000th of the cell)11 and difficulty to isolate pure primary cilia by classic biochemical methods12, proximity labeling approaches have been utilized to determine primary cilia proteomes of disease models and to investigate basic cilia biology13,14,15,16, such as dissecting the molecular composition of primary cilia during active signaling[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“),18. However, our knowledge of the protein composition of primary cilia is still incomplete as it stems from very few model cell types that are amenable to the available technologies.

While proximity labeling technologies are an active area of research[19](https://www.nature.com/articles/s41467-025-65405-8#ref-CR19 “Milione, R. R., Schell, B.-B., Douglas, C. J. & Seath, C. P. Creative approaches using proximity labeling to gain new biological insights. Trends in Biochemical Sciences https://doi.org/10.1016/j.tibs.2023.12.005

(2023).“), the most frequently used proximity labeling methods are based on two enzymatic activities that use different chemistries to label nearby proteins: 1) promiscuous biotin ligases, such as BioID, or 2) peroxidases, such as ascorbate peroxidase (APEX). BioID-based technologies are simple to use and only require biotin and ATP as substrates. In cases where the biological system requires a constant supply of biotin, BioID is continuously active and consumes biotin, resulting in persistent labeling —a major challenge for time-resolved studies and in vivo application20. A recently developed light-activatable variant, LOV-turbo21, is a remarkable improvement, however, comes at the cost of a more complex experimental setup. APEX-based approaches require two substrates: a tyramide, which is oxidized to produce a phenoxyl radical that reacts with nearby targets, and hydrogen peroxide (H2O2), which is reduced to water22. H2O2 supplementation provides control over the enzymatic activity, yet, endogenous cellular peroxidases can also use H2O2 to oxidize tyramides[2](https://www.nature.com/articles/s41467-025-65405-8#ref-CR2 “Kang, M.-G. & Rhee, H.-W. Molecular spatiomics by proximity labeling. Acc. Chem. Res. https://doi.org/10.1021/acs.accounts.2c00061

(2022).“),23. To account for such potential non-specific labeling, experimental designs include complex, time- and resource-consuming specificity controls, such as mislocalized APEX transgenes or genetic ablation of the target structure24. Most critically, APEX requires H2O2 to be supplied in high concentrations (mM), which induces oxidative damage in virtually all biological contexts posing a significant challenge for in vivo studies25,26,27.

Here, we show that many commonly used cell culture models are incompatible with previous APEX2-based proximity labeling methods. Undesired background often exceeds APEX2-mediated proximity biotinylation due to endogenous peroxidase activities when potentially toxic H2O2 is added externally. In the work presented here, we could overcome these limitations of APEX-based proximity labeling by employing the enzyme D-amino acid oxidase (DAAO) from Rhodotorula gracilis28 to locally generate H2O2. Thereby, APEX2-mediated biotinylation is rendered dependent on an enzyme cascade, yielding a more versatile in situ APEX activation (iAPEX) system, which 1) expands the applicability to additional biological systems, 2) reduces toxicity by avoiding addition of exogenous H2O2, and 3) increases specificity of APEX labeling to circumvent complex genetic controls. Using this methodology, we could successfully determine the proteomes of primary cilia of cell types hitherto inaccessible to conventional APEX proximity labeling. In addition, we show that iAPEX is a versatile method enabling effective and organelle-specific protein labeling with superb spatial resolution on lipid droplets (LDs) and mitochondria, which provides the potential to probe dynamic protein interactions at membrane-contact sites with sub-organelle resolution. Finally, we provide proof-of-concept experiments for the in vivo application of iAPEX in Xenopus laevis.

Results

D-amino acid oxidase can activate ascorbate peroxidase

Quantitative proteomics on subcellular microdomains is technically challenging. Since proteomic information on primary cilia is limited to few cell types13,18, we aimed to determine cilia proteomes of cell lines commonly used to study primary cilia by employing cilia-APEX2, an experimental setup we have successfully applied to study the ciliary proteome in a quantitative and time-resolved manner in kidney epithelial cells[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“). As APEX2-based proximity labeling is widespread, we envisioned an easy transfer of the methodology to other cell types. Yet, performing APEX2 labeling reactions using hydrogen peroxide (H2O2) resulted in various degrees of background biotinylation within cell types of interest, such as C2C12 myoblasts, 3T3-L1 pre-adipocytes, and NIH/3T3 fibroblasts (Fig. 1a). While in IMCD3 cells biotinylation was specific to the expression and localization of the APEX2 enzyme (visualized by GFP fluorescence of NPHP31–200-GFP-APEX2), biotinylation was observed throughout the cell bodies of C2C12, 3T3-L1 and NIH/3T3 cells, with visibly higher overall signal than in cilia-APEX2 expressing IMCD3 cells (Fig. 1a). After generating an NIH/3T3 cell line that stably expresses cilia-APEX2, we performed APEX2 labeling reactions by H2O2 addition and investigated the amounts of biotinylation by SDS-PAGE and western blotting (Fig. 1b). Surprisingly, biotinylation in NIH/3T3 cells was independent of the presence of the cilia-APEX2 enzyme and greatly surpassed the amounts observed in the well-established IMCD3 cell line (Fig. 1b, lanes 4 vs. 6), which indicated excessive background biotinylation by endogenous peroxidases.

Fig. 1: Background biotinylation in various cell types limits APEX-based proximity labeling applications.

a Immunofluorescence micrographs of IMCD3, C2C12, 3T3-L1, NIH/3T3, and IMCD3 cells stably expressing NPHP31–200-GFP-APEX2 (cilia-APEX2). Cilium formation was induced by 24 h growth factor deprivation. Cells were left untreated (–) or subjected to APEX2 proximity labeling (+) by incubation with 500 µM biotin tyramide (BT) for 30 min followed by 1 mM hydrogen peroxide (H2O2) for 1 min (3T3-L1) or 3 min (remaining cells). After fixation, primary cilia were visualized by anti-ARL13B antibody staining, cilia-APEX2 by GFP fluorescence, and biotin by fluorescently labeled streptavidin. Same imaging parameters for all images. b Wild-type (WT), cilia-APEX2-expressing IMCD3 and NIH/3T3 cells were lysed before (–) or after (+) APEX2 proximity labeling, followed by SDS-PAGE and western blot analysis (n = 4 independent experiments). Biotin detection as in (a), equal protein loading confirmed by total protein stain. c Diagram of cilia-iAPEX expression cassette with NPHP31–200-GFP-APEX2 (cilia-APEX2) and NPHP31–200-FLAG-DAAO (cilia-DAAO) transgenes in head-to-head orientation. A vector containing this cassette allows stable genomic integration via Flp-In recombinase and low-level expression of cilia-APEX2 and cilia-DAAO from truncated cytomegalovirus promoter (pCMVΔ6) and EF1α promoter lacking the TATA box (pEF1αΔ), respectively107. Enzymes are fused to N-terminal cilia-targeting sequences (amino acids 1-200 of murine Nephrocystin-3 (mNphp3)) and tagged with enhanced GFP (EGFP) for APEX2 and FLAG for DAAO. d Schematic: primary cilium harboring the in situ APEX activation (iAPEX) proximity labeling enzymes. APEX2 and DAAO are genetically targeted to primary cilia using constructs displayed in (c). A D-amino acid (D-AA) serves as DAAO substrate for in situ hydrogen peroxide (H2O2) production through oxidative deamination. Locally produced H2O2 and biotin tyramide (BT) are APEX2 substrates for proximity biotinylation, overcoming the need for external H2O2 addition (red X). e Immunofluorescence micrographs showing APEX2 proximity labeling in primary cilia of IMCD3 cells stably expressing the cilia-targeted iAPEX enzyme cascade (n = 5 independent experiments). APEX2 proximity labeling: incubation with biotin tyramide (BT) for 30 min and H2O2 for 3 min. DAAO-facilitated proximity labeling: D-alanine (D-Ala, 10 mM) added during BT incubation for 30 min. Cilia-DAAO is detected by anti-FLAG antibodies, all others as in (a). All scale bars = 5 µm.

To overcome non-specific proximity labeling and avoid external addition of H2O2, we expressed D-amino acid oxidase (DAAO) from Rhodotorula gracilis that oxidizes D-amino acids, the rare enantiomers of the predominant L-amino acids, to produce H2O2 intracellularly28,29,30,31. We specifically decided for this well characterized and established DAAO enzyme as it has been used successfully in several in vivo applications in human and rodent tissue30,32. To specify and restrict the subcellular localization of H2O2 production, we targeted DAAO to primary cilia by fusing it to the first 200 amino acids of the ciliary protein NPHP3, which we term cilia-DAAO (Fig. 1c). We hypothesized that locally produced H2O2 would be immediately used by nearby APEX2 to oxidize biotin tyramide (BT) for proximity labeling (Fig. 1d). To confirm the functionality of this enzymatic cascade in primary cilia, we generated an IMCD3 cell line stably expressing both cilia-APEX2 and cilia-DAAO, which localize specifically to primary cilia (Fig. 1e). In this cell line, proximity biotinylation in primary cilia could be achieved in the presence of biotin tyramide either by addition of H2O2 (to activate APEX2 directly) or by providing the DAAO substrate D-alanine (D-Ala) (Fig. 1e). Although small molecules can diffuse freely between the cilium and the cytoplasm[33](https://www.nature.com/articles/s41467-025-65405-8#ref-CR33 “Delling, M., Decaen, P. G., Doerner, J. F., Febvay, S. & Clapham, D. E. Primary cilia are specialized calcium signalling organelles. Nature https://doi.org/10.1038/nature12833

(2013).“), the functionality of the cascade required DAAO to be localized to cilia, as a DAAO enzyme localized to the cytosol (cyto-DAAO) did not activate cilia-APEX2 (Supplementary Fig. 1a). This suggests that H2O2 produced by DAAO in the cytoplasm does not diffuse into primary cilia, probably due to rapid detoxification. Interestingly, in cilia with strong biotin signals we observed a reduction in the cilia-DAAO signal (Supplementary Fig. 1b,c). As cilia-DAAO is detected via the FLAG epitope, we interpret this anti-correlation as a potential biotinylation of the tyrosine residue within the FLAG epitope, which may mask antibody binding.

D-amino acids are inert in most biological systems but show biological activity in rare instances, such as D-serine as a putative gliotransmitter34,35,36. Therefore, we tested different amino acids and derivatives as potential DAAO substrates for APEX-based proximity labeling in the cilia-APEX2 and cilia-DAAO expressing IMCD3 cell line. Except for D-valine, all D-amino acids tested induced biotin tyramide-dependent biotinylation in primary cilia in IMCD3 cells, which confirmed suitability and stereo-specificity of these substrates for DAAO-dependent APEX labeling (Supplementary Fig. 1d). Taken together, our experiments confirmed the functionality of the DAAO-APEX enzymatic cascade, which we term “in situ APEX activation” (“iAPEX”) proximity labeling.

Local hydrogen peroxide production minimizes toxicity

As the local restriction of APEX2 activation within primary cilia may limit its application for whole-cilium proteomics, we assessed the sub-ciliary localization of biotinylated proteins by ultrastructure expansion microscopy (U-ExM)37. U-ExM confirmed that both cilia-APEX2 and cilia-DAAO were confined to the membrane of the primary cilium (Fig. 2a and Supplementary Fig. 2a) with varying degrees of co-localization. However, after activation of the iAPEX cascade, biotinylation was not restricted to the membrane and could be detected throughout the entire cilium, indistinguishable from the activation by external addition of H2O2 (Fig. 2b and Supplementary Fig. 2b), indicating that iAPEX labeling generates sufficient phenoxyl radicals to probe the entire cilium.

Fig. 2: In situ D-amino acid oxidase-mediated hydrogen peroxide production enables APEX2 proximity labeling.

a, b Ultrastructure expansion microscopy (U-ExM) confocal images showing cilia-targeted iAPEX localization and proximity labeling. Cells were fixed, cross-linked, and embedded in a water-expandable gel. a Expanded primary cilium from an RPE-1 cell line stably expressing cilia-iAPEX reveals membrane localization of APEX2 (cyan) and DAAO (yellow), probed with antibodies against GFP and ALFA tag, respectively. Acetylated tubulin (acTub) is shown in magenta. b U-ExM micrographs of IMCD3 cells stably expressing cilia-iAPEX demonstrates biotinylation within the entire cilium. Cells were either untreated (–), labeled with biotin tyramide (BT) and H2O2, or 10 mM D-methionine (D-Met). Biotin visualization as in Fig. 1 (yellow). Scale bars = 5 µm (adjusted to expansion factors = 4 (a) and 4.3 (b)). c**–e** Quantification of absolute ciliary biotin signals in micrographs obtained from proximity labeling experiments performed in IMCD3 cilia-iAPEX cell line shown as violin plots. Dotted and dashed lines represent quartiles and medians, respectively. c Type and concentration of D-amino acid affect biotinylation. Indicated concentrations of D-Ala or D-Met were incubated for 30 min. n = 40 cilia per condition. d DAAO shows stereoselectivity for D-amino acids and allows labeling with low concentrations of D-Met. n = 77 cilia from two experiments. e Shorter substrate incubation leads to comparable biotinylation as H2O2-induced labeling. Incubation with BT and 10 mM D-Met for indicated times. n = 20 cilia from two experiments. Where indicated, 1 mM H2O2 was incubated for 3 min. Data analysis: one-sided Kruskal-Wallis test followed by Dunn’s multiple comparison. Numbers indicate p values. f**–h** D-Met-activated cilia-DAAO generates minute amounts of H2O2. O2 consumption rates (OCR) measured by Seahorse metabolic flux analysis with an average of 30.000 cells (coefficient of variation <5.0%). f OCRs of cilia-iAPEX IMCD3 cells determined after treatment with or without oligomycin (blocking cellular respiration), followed by D-Met or L-Met addition to activate DAAO. g OCRs of wild-type (WT) and h cilia-iAPEX IMCD3 cells were recorded and normalized to OCR after oligomycin treatment before addition of indicated amino acids (100%). Insert shows zoom. Lines depict means, error bars standard deviations (n = 3). Source data are provided as a Source Data file.

To identify the minimum concentrations of D-amino acids required for efficient labeling, we titrated the DAAO substrates D-alanine (D-Ala) and D-methionine (D-Met) and assessed cilia-iAPEX-catalyzed biotinylation efficiency by immunofluorescence microscopy (Fig. 2c and Supplementary Fig. 2c). Quantitation of biotinylation in primary cilia showed a concentration-dependent increase in biotinylation for both D-Ala and D-Met (Fig. 2c). High concentrations of both D-amino acids led to stronger biotinylation in cilia compared to 3 min labeling with H2O2, although D-Ala did not reach the same levels as D-Met. For D-Met saturating signals were achieved at 4 mM (Fig. 2c), while notable biotin signals could be observed at concentrations as low as 125 µM when DAAO-catalyzed proximity biotinylation was performed for 30 min (Fig. 2d). We therefore focused on the use of D-Met as DAAO substrate for our applications. As biotin tyramide exhibits moderate cell permeability[2](https://www.nature.com/articles/s41467-025-65405-8#ref-CR2 “Kang, M.-G. & Rhee, H.-W. Molecular spatiomics by proximity labeling. Acc. Chem. Res. https://doi.org/10.1021/acs.accounts.2c00061

(2022).“), we aimed to increase the temporal resolution of biotinylation by pre-incubating cells with biotin tyramide before D-Met addition. A time course of the labeling reaction revealed that 5 min incubation with 10 mM D-Met after 30 min biotin tyramide pre-incubation was comparable to short activation (3 min) with 1 mM H2O2 (Fig. 2e). While we incubated for 2 min with H2O2 in a former study[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“), 5 min labeling is still compatible with our previous temporal resolution. Further experiments revealed that 5 min pre-incubation with biotin tyramide was sufficient, as longer pre-incubation times did not increase the labeling efficiency (Supplementary Fig. 2d). Interestingly, pre-incubation with D-Met to initiate local H2O2 production prior to biotin tyramide addition decreased biotinylation in a time-dependent manner (Supplementary Fig. 2d), indicating that prolonged H2O2 production might interfere with APEX2 function. Thus, the overall temporal resolution that can be achieved in IMCD3 cells is 5 min, which compares favorably to recently developed proximity labeling methods21,38.

To assess the consequence of local H2O2 production and biotinylation on the function of primary cilia, we investigated intraflagellar transport in IMCD3 cells transiently transfected with IFT38-GFP-APEX2 and cilia-localized DAAO by live-cell microscopy (Supplementary Movie 1) and could not observe any obvious difference after 30 min H2O2 production (Supplementary Movie 2) or iAPEX labeling (Supplementary Movie 3).

Although the basic functions of primary cilia still appeared intact in the timeframe of iAPEX labeling, H2O2 in primary cilia is likely to cause oxidative damage locally. To assess potential toxicity of the iAPEX system for the entire cell, we determined the H2O2 production by cilia-DAAO employing oxygen (O2) consumption measurements as DAAO activity converts O2 to equimolar amounts of H2O239. When blocking cellular respiration with oligomycin the cilia-iAPEX IMCD3 cell line consumed approximately 1 fmol/(min×cell) O2 at steady state (Fig. 2f). After addition of D-Met the O2 consumption increased to about 1.2 fmol/(min×cell) (Fig. 2f). This rise in O2 consumption was D-amino-acid- and cilia-DAAO-dependent, as it was not observed in the parental cell line (Fig. 2g). Although D-Ala led to a comparable maximum O2 consumption rate as D-Met, our kinetic analysis indicated a 30 min delay to reach this maximum (Fig. 2h), which agrees with the observed differences in iAPEX-catalyzed biotinylation (see Fig. 2c). Assuming an average cell volume of 4.000 fL, the increase in O2 consumption of 0.2 fmol/(min×cell) would result in sub-µM H2O2 concentrations within 1 sec of DAAO activity if the cell completely lacked mechanisms to detoxify H2O2. However, as physiological redox signaling requires an existing, potent antioxidant system40, our data indicate that the amount of H2O2 produced by cilia-DAAO is within physiological H2O2 concentrations and can therefore be considered non-toxic for most cell types.

Cilia-iAPEX locally restricts proximity biotinylation and prevents off-target biotinylation

To test whether the iAPEX labeling cascade overcomes the high background observed in select cell lines (see Fig. 1a, b), we introduced cilia-iAPEX into NIH/3T3 cells using the Flp-In system, and isolated a clonal cell line, in which both enzymes localized to primary cilia (Fig. 3a). Most importantly, iAPEX labeling was specific to primary cilia in this cell line and indicated that spatially restricted H2O2 production by cilia-DAAO prevented non-specific biotinylation of other cellular structures, as observed after direct activation of cellular peroxidases (Fig. 3a). SDS-PAGE and western blot analysis of whole-cell lysates from cilia-iAPEX expressing NIH/3T3 cells confirmed high background biotinylation when using H2O2 as a substrate, while DAAO activation resulted in markedly reduced but specific biotinylation (Fig. 3b, lanes 11 vs. 12). Interestingly, even in whole-cell lysates from the established IMCD3 cell line we observed overall stronger signals when using H2O2 compared to D-Met-based proximity labeling (Fig. 3b, lanes 8 vs. 9), despite weaker biotin signals in primary cilia (see Fig. 2c). To gain a deeper understanding of the events occurring during APEX labeling, we established a live-cell imaging setup to visualize the subcellular localization of peroxidase activity by the oxidation of the peroxidase substrate Amplex UltraRed (AmUR) to a fluorescent resorufin product41,42. After loading cells that stably express cilia-iAPEX with AmUR, we noticed a burst in peroxidase activity throughout the entire cell shortly after H2O2 addition, which ceased over time when only the APEX activity in the primary cilium remained (Fig. 3c and Supplementary Movie 4). In contrast, local production of H2O2 by cilia-DAAO prevented non-specific peroxidase activity, as we observed resorufin signals exclusively in primary cilia for prolonged labeling times (Fig. 3d and Supplementary Movie 5). These results indicate that other cellular peroxidases are capable of oxidizing substrates, such as biotin tyramide to biotinylate nearby proteins by proximity labeling when H2O2 is added to the cells. We further hypothesize that the observed initial burst in peroxidase activity may cause significant non-specific biotinylation and thereby high background when studying proteomic environments of targets that are expressed at low levels, such as for primary cilia proteomics.

Fig. 3: iAPEX enables cilia-specific biotinylation in NIH/3T3 bypassing high cellular background.

a Immunofluorescence micrographs of NIH/3T3 WT and cilia-iAPEX cells with IMCD3 cilia-iAPEX cells as control (n = 5 independent experiments). Cells were left untreated, labeled using BT and H2O2, or BT and D-Met, as indicated. b Western blot analysis of WT and cilia-iAPEX-expressing IMCD3 and NIH/3T3 cells. Cells were lysed before or after BT and H2O2, or BT and D-Met treatment (n = 4 independent experiments). Asterisk marks cross-reactive band of the anti-GFP antibody. Please note that the GAPDH loading control was not analyzed on the same gel, but from identical samples analyzed in parallel. c, d Live-cell confocal imaging micrographs were captured to observe peroxidase-dependent Amplex UltraRed (AmUR) oxidation to resorufin in IMCD3 cells stably expressing cilia-APEX2. c Cells were treated with 50 µM AmUR together with 1 mM H2O2 where indicated. Resorufin and GFP fluorescence of cilia-APEX2 were monitored at 4.8-second intervals over a total duration of 264 s (see also Supplementary Movie 4). d AmUR oxidation reveals exclusive cilia-APEX2 activity when DAAO-dependent H2O2 production was performed after addition of 50 µM AmUR and 10 mM D-Met (see also Supplementary Movie 5). Two cilia are shown per condition. Scale bars = 5 µm in all panels.

Cilia-iAPEX proteomics increases specificity and sensitivity of cilia protein identification

To directly compare the iAPEX- with the APEX2-based proximity labeling method as a discovery tool in proteomics applications, we performed iAPEX (DAAO-dependent) and APEX2 (H2O2-induced) proximity labeling in cilia-iAPEX IMCD3 cells using desthiobiotin tyramide (DTBT) as a substrate, as this allowed competitive elution of APEX-biotinylated proteins after isolation by streptavidin chromatography (Fig. 4a). Abundant non-specific biotinylation in APEX2-based proximity labeling setups requires controls to precisely assess the background43,44. To this end, for cilia proteomics we previously expressed cilia-APEX2 in Cep164*-/-* cells that lack primary cilia[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“). Triplicates of the iAPEX-labeled samples and duplicates of the controls were analyzed by SDS-PAGE and western blotting. Our analyses confirmed reduced biotinylation by iAPEX compared to APEX2 labeling (Supplementary Fig. 3a, lanes 6-7 vs. 8-10), while several cilia components, represented by IFT88 and IFT57, were isolated more efficiently after iAPEX labeling (Fig. 4b, lanes 13-15 vs. 11,12). This indicated a higher sensitivity of cilia-iAPEX compared to previous setups, while the Cep164*-/-* controls confirmed specificity of isolation, as no ciliary proteins were isolated in the absence of cilia (Fig. 4b, lanes 9,10). To quantitatively assess the performance of cilia-iAPEX vs. cilia-APEX2, isolated proteins were digested by trypsin, labeled with tandem-mass-tags (TMT) and analyzed by liquid chromatography and multistage mass spectrometry (LC-MS3)5,[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“),45,46 (Fig. 4a). We quantified the relative abundances of 5982 identified proteins within the individual samples (see Supplementary Data 1). Replicates showed high similarity across conditions, as evidenced by pairwise multiscatter plots and high Pearson correlation coefficients (Supplementary Fig. 3b). When assessing candidate ciliary proteins by statistical analysis of relative enrichments between cilia-iAPEX and control samples, we applied stringent TMT enrichment ratios of 23 which resulted in 175 high confidence candidate cilia proteins (Supplementary Fig. 3c). Surprisingly, within the same experiment the same TMT enrichment ratio cutoff between cilia-APEX2 and control samples identified 799 putative cilia proteins (Supplementary Fig. 3d). A direct comparison showed that the enrichment of known cilia proteins was similar in both approaches, however, cilia-iAPEX proteomics separated known cilia proteins much better from non-cilia proteins (Fig. 4c). Gene Ontology (GO) term enrichment analyses confirmed higher specificity of the iAPEX setup, as evidenced by the absence of non-ciliary processes and the lower p values of ciliary categories (Supplementary Fig. 3e,f). Hierarchical clustering of the relative protein abundances within the experiment confirmed high reproducibility of the replicate samples (Supplementary Fig. 4a). Proteins enriched in both iAPEX and APEX2-labeled samples formed three clusters highly enriched in known cilia proteins (Fig. 4d, Supplementary Fig. 4b and Supplementary Data 1), which covered 38% of the cilia-APEX2 proteome that was based on three independent datasets[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“). Interestingly, the proteins in these cilia clusters contributed to the lower correlation with control samples, underscoring their specificity and higher abundance in the iAPEX samples (see Supplementary Fig. 3b, orange). A GO term enrichment analysis revealed high statistical significance of components associated with cilia and related microtubule-based structures (Fig. 4e), while our previous cilia-APEX2 proteome contained many non-ciliary categories[17](https://www.nature.com/articles/s41467-025-65405-8#ref-CR17 “May, E. A. et al. Time-resolved proteomics profiling of the ciliary Hedgehog response. J. Cell Biol. 220, https://doi.org/10.1083/jcb.202007207

(2021).“), suggesting false-positive hits. Agreeingly, our cluster analysis also identified proteins that were enriched only in the H2O2-treated samples, in wild-type and Cep164*-/-* cells, that formed four clusters (Fig. 4f and Supplementary Fig. 4c). GO term enrichment analysis of these clusters identified a large fraction of proteins located in the endoplasmic reticulum (ER) (Fig. 4g), suggesting that ER resident peroxidases in IMCD3 cells can biotinylate nearby proteins in a H2O2-dependent manner. Such peroxidases may cause non-specific labeling and contribute to potential false-positive hits in conventional APEX2 labeling setups. To rule out that the proteins in the false-positive clusters were absent from the iAPEX labeled samples for technical reasons, due to an oxidation-dependent mass shift, we investigated the oxidized peptides in our dataset. When comparing peptide distributions of oxidized and non-oxidized peptides across all samples, we observed a highly similar distribution in the false-positive clusters and among peptides from IFT proteins (Supplementary Fig. 4d), supporting the notion that the proteins identified by conventional APEX2 labeling are likely false-positives. Taken together, iAPEX-based proteomics shows high sensitivity to analyze the proteome of subcellular microdomains and significantly reduces the number of false-positives by lowering background biotinylation activities.

Fig. 4: Quantitative primary cilia proteomics using iAPEX outperforms conventional APEX2-based system in IMCD3 cells.

a Schematic: cilia-iAPEX-based proximity labeling workflow for proteomic analysis of IMCD3 primary cilia. iAPEX labeling in cilia-iAPEX (cilia-APEX2 and cilia-DAAO) expressing cells with desthiobiotin tyramide (DTBT) and D-Met for 30 min. For APEX2 proximity labeling cilia-iAPEX expressing wild-type and cilia-APEX2 expressing Cep164*-/-* cells (control) were pre-incubated with DTBT for 30 min, followed by 3 min H2O2. After cell lysis, labeled proteins were isolated by streptavidin chromatography and competitively eluted with biotin. Input, Unbound and Eluate fraction analysis by SDS-PAGE and western blotting. For mass spectrometric analysis, eluted proteins were digested in-solution using trypsin, peptides labeled with tandem mass tags (TMTs) and fractionated offline via reverse-phase chromatography. Quantitative proteomics was performed using LC-MS³: peptides were selected (MS¹), fragmented for identification (MS²), and TMT reporter ions quantified (MS³). b Western blot analysis after proximity labeling from IMCD3 cells, as outlined in (a) (n = 2 independent experiments). IMCD3 Ift88*-/-* cell lysate served as antibody specificity and untreated control. SDS-PAGE and western blot analysis of Input and Eluate samples using indicated antibodies. Input 0.063%, Eluate 8.5%. c Volcano plot of statistical significance versus protein enrichment in cilia-APEX2 (left) and cilia-iAPEX (right) compared with control samples. Calculated p values (unpaired two-sided Student’s t test) were plotted against TMT ratios for 5982 proteins. Gray and red circles indicate identified and known cilia proteins, respectively. Representative subunits of kinesin-2 (Kif3a), IFT-A (Ift122), IFT-B (Ift88) and the BBSome (Bbs4) are highlighted. Dotted lines indicate TMT ratios of 23. d Selected clusters of two-way hierarchical cluster analysis of IMCD3 cilia-iAPEX proteome show known cilia proteins and highest scoring candidate cilia proteins. Legend shows relative protein abundances (in %). Full cluster analysis shown in Supplementary Fig. 4. e GO term enrichment analysis of protein clusters in (d) shows enrichment of ciliary categories. p values were calculated by one-sided Fisher’s exact test. f Selected clusters with proteins identified after H2O2-mediated cilia-APEX2 proximity labeling. g GO term enrichment analysis of protein clusters in (f) identified enrichment of non-ciliary categories in H2O2 treated samples. p values were calculated by one-sided Fisher’s exact test.

Determining the cilia-iAPEX proteome of NIH/3T3 cells

As iAPEX proximity labeling allowed the specific biotinylation of proteins in NIH/3T3 cells (see Fig. 3), we sought to gain proteomic information on the ill-defined proteome of NIH/3T3 primary cilia. By combining hierarchical clustering with the increased specificity of the iAPEX system, we envisioned that the cilia proteomes could be investigated without the need for genetic