The cDNA coding for a human codon-optimized variant of IBD-SBP-YAP1C (Supplementary Figure S1) was synthesized by Integrated DNA Technologies and provided in the pUCIDT (Kan) cloning vector (pMF1986). A mammalian expression plasmid encoding IBD-SBP-YAP1C (pMF1987) was generated by transferring the EcoRI/NotI-restricted fragment of pMF1986 into the EcoRI/NotI-restricted backbone fragment of pMF1839 (Walton et al., 2017). Mammalian expression vectors encoding mitochondrial (pMF1991), peroxisomal (pMF1992), or cytosolic (pMF2029) variants of IBD-SBP-YAP1C were generated by amplifying the IBD-SBP-YAP1C template via PCR (forward oligo: 5′-ggg​gga​tcc​cat​ggc​atc​aat​gca​gaa​gct​g-3′; reverse oligos: 5′-ccg​ggg​gcg​gcc​gct​cag​ttc​ata​tgt​t-3′ (for pMF1991 and pMF2029) and 5′-ccg​ggg​gcg​gcc​gct​caa​agc​tta​ctt​ttg​ttc​ata​tgt​tta​ttc​aat​gca-3′ (for pMF1992)) and subcloning the BamHI/NotI-restricted PCR products into the BamHI/NotI-restricted backbone fragments of pKillerRed-dMito (Evrogen) (for pMF1991) or pEGFP-N1 (Clontech) (for pMF1992 and pMF2029). Plasmid sequences were validated by DNA sequencing (LGC Genomics). The plasmids encoding EGFP-HsPEX11B (pTW110) (Fransen et al., 2001), c-roGFP2 (pMF1707) (Ivashchenko et al., 2011), po-roGFP2 (pMF1706) (Ivashchenko et al., 2011), or mt-roGFP2 (pMF1762) (Ivashchenko et al., 2011) have been described elsewhere. The plasmid encoding EGFP-HSPB1 was kindly provided by Prof. Dr. Ludo Van Den Bosch (KU Leuven, Belgium).

Cell culture was essentially performed as previously described (Lismont et al., 2019a). Briefly, all cells were cultured at 37°C in a humidified 5% CO₂ incubator in minimum essential medium Eagle α (Lonza, BE12-169F) supplemented with 10% (v/v) fetal bovine serum (Biowest, S181B), 2 mM UltraGlutamine I (Lonza, BE17-605E/U1), and 0.2% (v/v) MycoZap (Lonza, VZA-2012). Transfections were performed by using the Neon Transfection System (Thermo Fisher Scientific; 1,150 V, 20-ms pulse width, two pulses).

The Flp-In T-REx 293 cell lines expressing DD-DAO in peroxisomes or the cytosol have been detailed elsewhere (Lismont et al., 2019a). To generate po- or c-DD-DAO Flp-In T-REx 293 cell lines constitutively expressing c-IBD-SBP-YAP1C, po-IBD-SBP-YAP1C, or mt-IBD-SBP-YAP1C, the corresponding Flp-In T-REx 293 cells were transfected with pMF2029, pMF1992, or pMF1991, respectively. Starting from 2 days later, the cells were routinely cultured in a medium supplemented with 1) 10 μg/ml blasticidin (InvivoGen; ant-bl) and 100 μg/ml hygromycin B Gold (InvivoGen, ant-hg) to maintain the properties of the Flp-In T-REx 293 cell lines stably expressing peroxisomal or cytosolic DD-DAO, and 2) 200 μg/ml of G418 (Acros Organics, BP673-5) to select for cells carrying the neomycin resistance cassette (with a minimum period of 3 weeks). To modulate the expression levels of DD-DAO in these cells, they were incubated for 3 or 4 days in the absence or presence of 1 μg/ml doxycycline (DOX) (Sigma, D9891) and 500 nM Shield1 (Clontech, 632,189). Treatments were always followed by a 24-h chase period (no DOX, no Shield1) in order to remove the pool of residual cytosolic po-DD-DAO (Lismont et al., 2019a), unless specified otherwise.

Fluorescence microscopy was carried out as described previously (Ramazani et al., 2021). The following excitation filters (Ex), dichromatic mirrors (Dm), and emission filters (Em) were chosen to match the fluorescent probe specifications: DAPI (Ex: BP360-370; Dm: 400 nm cut-off; Em: BA420-460); EGFP (Ex: BP470-495; Dm: 505 nm cut-off; Em: BA510-550); and Texas Red (Ex: BP545-580; Dm: 600 nm cut-off; Em: BA610IF). The cellSens software (Olympus Belgium) was used for image analysis. Samples for immunofluorescence microscopy were fixed, counterstained with 0.5 μg/ml DAPI (Sigma, D-9542) in Dulbecco’s phosphate-buffered saline (DPBS) for 1 min, and processed as described (Passmore et al., 2020).

Cells were grown to 60–80% confluency, trypsinized, collected in cell culture medium, pelleted (150 x g, 5 min), and washed once with DPBS without calcium and magnesium (BioWest, L0615). Cell density was determined by Bürker chamber counting and adjusted to 10⁶ cells/ml. After being subjected to different treatments (for specifications, see Results section), N-ethylmaleimide (NEM) (TCI, E0136) dissolved in methanol (Fisher Scientific, M/4062/17) was added to a final concentration of 10 mM. Next, the cells were pelleted (150 x g, 5 min), resuspended in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% (v/v) Triton X-100, 10% (v/v) glycerol) containing 10 mM NEM and a protease inhibitor mix (Sigma-Aldrich, P2714)) at a density of 10⁷ cells/ml, and lysed on ice for 10 min. Thereafter, a cleared lysate was produced by double centrifugation (20,000 x g, 10 min), each time discarding the pellet.

The cleared lysates were mixed with 300 µl of (prewashed) high capacity streptavidin agarose beads (Thermo Scientific, 20359) and incubated at 4°C on a rotation mixer to enrich IBD-SBP-YAP1C complexes. After 2 h, the bead suspensions were transferred to Micro-Spin columns (Thermo Scientific, 89879) and consecutively washed five times with lysis buffer, five times with wash buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% (v/v) Triton X-100), and five times with 50 mM Tris-HCl pH 8.0. Finally, the disulfide-bonded interaction partners of IBD-SBP-YAP1C were eluted by incubating the columns three times for 15 min with 200 µl of elution buffer (10 mM DTT in 50 mM Tris-HCl pH 8.0). The eluates were pooled and subsequently processed for proteomics analysis. At each step, small aliquots were saved for immunoblot analysis. To analyze comparable amounts of IBD-SBP-YAP1C-containing protein complexes, the ratio of cleared lysate to streptavidin agarose beads was chosen such that the affinity matrix was slightly oversaturated, as determined by detection of residual IBD-SBP-YAP1C in the non-bound fraction.

Eluates were incubated for 30 min at 37°C to allow the DTT to fully reduce all proteins. Thereafter, the samples were alkylated (37°C, 30 min) with 25 mM iodoacetamide (Sigma, I-6125). Excess iodoacetamide was quenched with 25 mM DTT (AppliChem, A2948) (37°C, 30 min). Subsequently, the proteins were precipitated as described (Wessel and Flügge, 1984) and digested overnight with modified trypsin (Pierce, 90057) in the presence of 50 mM ammonium bicarbonate (Sigma-Aldrich, A6141), 5% acetonitrile (Applied Biosystems, 400315), and 0.01% ProteaseMAX surfactant (Promega, V2072). Trypsin was inactivated by addition of 0.5% (v/v) trifluoroacetic acid (Applied Biosystems, 400028). The resulting peptides were desalted with C18 ZipTip pipette tips (Merck Millipore, ZTC18S960) and loaded onto an Ultimate 3000 UPLC system (Dionex, Thermo Fisher Scientific) equipped with an Acclaim PepMap100 pre-column (C18; particle size: 3 μm; pore size 100 Å; diameter: 75 μm; length: 20 mm; Thermo Fisher Scientific) and a C18 PepMap analytical column (particle size: 2 μm; pore size: 100 Å; diameter: 50 μm; length: 150 mm; Thermo Fisher Scientific) using a 40 min linear gradient (300 nl/min) coupled to a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific) operated in data-dependent acquisition mode. After the initial pilot experiment, the mass spectrometry (MS) method was adapted, essentially doubling the maximum injection time for MS/MS. Peptides were identified by Mascot (Matrix Science) using Uniprot Homo sapiens as a database (# entries: 194619). S-carbamidomethylation (C), N-ethylmaleimide (C), and oxidation (M) were included as variable modifications. Two missed cleavages were allowed, peptide tolerance was set at 10 ppm and 20 mmu for MS and MS/MS, respectively.

Progenesis QI software (Nonlinear Dynamics) was used for the relative quantification of proteins based on peptides validated by the Proteome Discoverer 2.2 Percolator node. Only exclusive peptides having a peptide spectral match (PSM) with a posterior error probability (PEP) smaller than 0.001 (10^–3) in at least one of the conditions of the experiment were taken into account for quantification (Käll et al., 2008). Proteins that could not be unambiguously identified or were identified as keratins, extracellular proteins, or proteins that—according to the Human Protein Atlas database (http://www.proteinatlas.org)—are not expressed in HEK-293 cells, were manually removed. Proteins enriched at least 2.5-fold upon H₂O₂ exposure were retained as H₂O₂ targets. For every experiment, data derived from one biological replicate are shown.

The pre-immune serum was collected from a rabbit before immunization with the 21 kDa subunit of rat palmitoyl-CoA oxidase (Baumgart et al., 1996); the rabbit polyclonal antisera against EGFP (Fransen et al., 2001), PEX13 (Fransen et al., 2001), PRDX1 (Goemaere and Knoops, 2012), PRDX3 (Goemaere and Knoops, 2012), or PRDX5 (Goemaere and Knoops, 2012) have been described elsewhere; and the rabbit polyclonal antisera against TUBA (Santa Cruz Biotechnology, sc-5546) and the goat anti-rabbit secondary antibodies, conjugated to Texas Red (Calbiochem, 401355) or alkaline phosphatase (Sigma, A3687), were commercially obtained.

Electrophoretic mobility shift assays (EMSAs) were performed as described previously (Lismont et al., 2019a). A slightly modified protocol was used for samples taken during the proteomics sample preparation. Specifically, for the “input” and “non-bound” samples, 50 µl of cleared cell lysates and bead supernatants were respectively mixed with 2X SDS-PAGE sample buffer without reducing agent and heated to 65°C for 10 min. For the “bound” and “bound after elution” samples, 10 µl of bead volume was mixed with non-reducing 2X SDS-PAGE sample buffer and heated to 100°C for 10 min.

Subcellular fractionations of rat liver (Anthonio et al., 2009) and HEK-293 cells (Lismont et al., 2019c) were carried out as described elsewhere. The animal and human cell-related studies were reviewed and approved by the local UZ/KU Leuven ethics Committee (approval numbers: P092/2018, S63097, and S62366). Functional enrichment analysis was performed using g:GOST (https://biit.cs.ut.ee/gprofiler/gost) within the g:Profiler tool (Raudvere et al., 2019). Heat maps were generated with Graphpad Prism version 9.0.0 for Windows (GraphPad Software). The iCysMod (http://icysmod.omicsbio.info/index.php) (Wang et al., 2021) and TF2DNA (http://fiserlab.org/tf2dna_db//index.html) (Pujato et al., 2014) databases were used as resources to search for known protein cysteine oxidation sites and transcription factors in the human dataset, respectively.

To confirm the correct localization of the different IBD-SBP-YAP1C fusion proteins in the G418-enriched po-DD-DAO Flp-In T-REx 293 cell lines, we co-expressed compartment-specific fluorescent marker proteins (Figure 1). From this experiment, it is clear that the majority of the cells express IBD-SBP-YAP1C. However, it is also evident that, at least in some cells, varying portions of po-IBD-SBP-YAP1C and, to a certain extent, also mt-IBD-SBP-YAP1C still reside in the cytosol. Given that 1) the functionality of peroxisomal and mitochondrial targeting signals fused to a heterologous protein strongly depends on the protein context (Yogev and Pines, 2011; Kunze, 2018), 2) the priority of protein import into mitochondria and peroxisomes is governed by competition for binding to limiting amounts of import receptor (Weidberg and Amon, 2018; Rosenthal et al., 2020), and 3) expression of the YAP1C fusion proteins is driven by the cytomegalovirus promoter, one of the strongest naturally occurring promoters (Even et al., 2016), this observation may not be that surprising. Although this may complicate the interpretation of downstream results, this knowledge also allows us to correctly anticipate this shortcoming.

To validate the in cellulo trapping strategy for sulfenylated proteins, we first investigated IBD-SBP-YAP1C complex formation in different subcellular compartments upon exposure of cells to exogenous or peroxisome-derived H₂O₂. An outline of the experimental workflow is depicted in Figure 2A and detailed in the Materials and Methods section. Importantly, given that inhibition of catalase activity with 3-amino-1,2,4-triazole (3-AT) increases the responsiveness of Flp-In T-REx 293 cells to peroxisome-derived H₂O₂ (Lismont et al., 2019a), we routinely added this inhibitor to the assay medium. In a first experiment, Flp-In T-REx 293 cells expressing c-IBD-SBP-YAP1C were exposed or not to 1 mM H₂O₂ for 10 min and subsequently processed as detailed in the legend to Figure 2A. IgG blot overlay analysis of the input fractions (I), the non-bound fractions (NB), the streptavidin-bound fractions (B), and the streptavidin-bound fractions after elution with a reducing agent (BE) confirmed that exposure of the cells to external H₂O₂ triggered the formation of many c-IBD-SBP-YAP1C-containing higher molecular weight complexes that can be enriched on streptavidin beads and are sensitive to the reducing agent dithiothreitol (DTT) (Figure 2B). The latter feature is important to allow selective elution of target proteins from the affinity matrix. Next, a similar experiment was performed with Flp-In T-REx 293 cells expressing po-DD-DAO and a compartment-specific variant of IBD-SBP-YAP1C and in which peroxisomal H₂O₂ production was induced or not by supplementing the assay medium (DPBS) with 10 mM D- or L-Ala, respectively. Multiple IBD-SBP-YAP1C-containing protein complexes could be detected in all conditions in which peroxisomal H₂O₂ was produced (Figure 2C). Interestingly, a direct comparison of the staining patterns of the streptavidin-enriched fractions clearly shows that the interaction profiles of IBD-SBP-YAP1C with sulfenylated proteins differed considerably depending on the source of H₂O₂ as well as on the subcellular location of the YAP1C fusion protein (Figure 2D).

To corroborate the sulfenome mining strategy, we carried out an exploratory experiment in which cells expressing c-, mt-, or po-IBD-SBP-YAP1C were treated or not with 1 mM H₂O₂ for 10 min. After documenting IBD-SBP-YAP1C complex formation (Supplementary Figure S2), the DTT eluates were processed for LC-MS/MS analysis. After validation of the identified peptides, probable contaminating proteins (e.g., keratins, IgGs, and extracellular proteins) and proteins that could not be unambiguously identified were manually removed. A set of 48 proteins that were enriched 2.5-fold or more in at least one of the H₂O₂-treated conditions could be listed (Supplementary Table S1).

Gene ontology analysis of the hits revealed significant enrichment for proteins primarily implicated in cell redox homeostasis and cellular oxidant detoxification (Supplementary Figure S3). These include GSR, PRDX1, PRDX2, PRDX3, PRDX4, PRDX5, PRDX6, TXN, and TXNRD2. Interestingly, while 19 out of 48 proteins were enriched in all H₂O₂-treated samples, each IBD-SBP-YAP1C fusion protein also retained unique interactors (Figure 3). This finding agrees with our previous results showing that differentially-localized IBD-SBP-YAP1C proteins form different complexes upon treatment of po-DD-DAO Flp-In T-REx 293 cells with D-Ala (Figure 2D).

To gain more insight into the binding selectivity of each IBD-SBP-YAP1C, we calculated the percentage distribution of each interactor with each of the YAP1C-fusion proteins (Supplementary Table S2). In line with expectations, this analysis revealed that 1) po-IBD-SBP-YAP1C predominantly interacts with ACOX1, a bona fide peroxisomal matrix protein (Yifrach et al., 2018), and proteins that show a partial peroxisomal localization (e.g., LDHB (Schueren et al., 2014) and MDH1 (Hofhuis et al., 2016)), 2) mt-IBD-SBP-YAP1C mainly interacts with genuine mitochondrial proteins (e.g., TXNRD2, ME2, ECI1, HSD17B10, and PRDX3 (Rath et al., 2021)), 3) c-IBD-SBP-YAP1C preferentially interacts with proteins that are predominantly located in the cytosol (e.g., DNPEP, HBA2, UCHL3, TUBB1, and CDK4 (Thul et al., 2017)), and 4) interactors located simultaneously in peroxisomes, mitochondria, and the cytosol (e.g., HSPA9 (Jo et al., 2020) and PRDX5 (Knoops et al., 2011)) are trapped by all YAP1C fusion proteins. However, diverse interactors known to be located in the cytosol and/or nucleus (e.g., BOLA2B, CSTB, HPRT1, PRDX1, PRDX2, PRDX6, PRMT5, SKP1, and WDR77 (Thul et al., 2017)) also bound to po- and mt-IBD-SBP-YAP1C, at least to a certain extent. This phenomenon can most likely be explained by the fact that a small but significant portion (rough estimation: 5–25%) of these IBD-SBP-YAP1C-fusion proteins still resides in the cytosol (Figure 1). Similar reasoning can be applied with regard to the observation that a small portion of ACOX1 interacts with c-IBD-SBP-YAP1C. Indeed, it can be expected that the c-IBD-SBP-YAP1C-interacting portion of ACOX1 represents the protein pool that has not yet been imported into peroxisomes. Here, it is also important to note that the percentage distribution of some interactors displayed an unexpected behavior. For example, some cytosolically located target proteins preferentially interacted with po-IBD-SBP-YAP1C (e.g., MARCKSL1, RPL13, and SET). A comprehensive explanation is currently lacking, but it cannot be excluded that portions of these proteins are partially associated with peroxisomes. However, this remains to be further investigated.

Finally, this experiment clearly shows that some interactors (e.g., PRDX1, PRDX2, and TXN) are already partially trapped by their respective IBD-SBP-YAP1Cs even in the absence of H₂O₂ treatment (Supplementary Table S1). On one hand, this may point to background contamination. However, a more likely explanation is that these interactors are extremely sensitive to sulfenic acid formation, a hypothesis supported by their recognized roles in localized, rapid, specific, and reversible redox-regulated signaling events (Hanschmann et al., 2013).

A follow-up experiment was performed with Flp-In T-REx 293 cells expressing po-DD-DAO and compartment-specific variants of IBD-SBP-YAP1C and in which peroxisomal H₂O₂ production was induced by supplementing the assay medium with 10 mM D-Ala. However, this time we included different time points (0, 2, 5, 15, 30, and 60 min) (Supplementary Figure S4) and adapted the LC-MS/MS method providing a higher sensitivity in MS/MS. After validation of the MS results, a total of 444 unique proteins that were 2.5-fold or more enriched in at least one time point upon peroxisomal H₂O₂ production were retained. The number and overlap of interactors identified with po-, c-, or mt-IBD-SBP-YAP1C at each time point are visualized in Figure 4. From this overview, it is clear that the number of individual and common interactors of different IBD-SBP-YAP1Cs changed in function of time. One may argue that these differences are due to variations in experimental handling. However, given that peroxisomal H₂O₂ generation resulted in different but consistent response profiles for distinct interactor classes, this assumption is unlikely to hold. A specific example is shown for c-IBP-SBP-YAP1C, which traps multiple peroxiredoxins (PRDXs) and 14-3-3 proteins upon such treatment (Figures 5A,B). More examples can be found in Supplementary Table S3, Supplementary Table S4 and Supplementary Table S5, which respectively provide raw abundance heat maps of proteins trapped at different times by c-, po-, and mt-IBD-SBP-YAP1C in response to peroxisome-derived H₂O₂.

Manual inspection of the 75 different po-IBD-SBP-YAP1C interactors (Supplementary Table S4) surprisingly revealed the presence of only a few proteins that are frequently (e.g., CAT and HSD17B4) (Yifrach et al., 2018) or sporadically (e.g., HSPA9) (Jo et al., 2020) detected in peroxisomes. This observation can potentially be explained in different ways. For example, given that the peroxisomal H₂O₂ sensor po-roGFP-ORP1 is already almost fully oxidized in Flp-In T-REx 293 cells under basal conditions (Lismont et al., 2019a), it may well be that no redox-sensitive protein thiol groups were left as targets for oxidation by newly formed H₂O₂. On the other hand, it cannot be ruled out that the peroxisomal proteins trapped by po-IBD-SBP-YAP1C represent the cytosolic protein pools that have not yet been imported into peroxisomes. To gain more insight into this problem, we compared the response kinetics of CAT and HSD17B4 to peroxisome-derived H₂O₂ in cells expressing po- or c-IBD-SBP-YAP1C (Supplementary Figure S5). From this figure, it is clear that 1) depending on the interactor and time point, the amount of protein trapped by po-IBD-SBP-YAP1C is up to 1.5-fold higher (e.g., CAT, 30 min) or up to 200-fold lower (e.g., HSD17B4, 15 min) than the amount trapped by c-IBD-SBP-YAP1C, and 2) the sulfenylation profiles of po-IBD-SBP-YAP1C (left panels) and c-IBD-SBP-YAP1C (middle panels) interactors can exhibit a bimodal behavior, thereby reflecting a heterogeneous character of protein thiol oxidation (see Discussion). The observation that the capturing ratios of the peroxisomal targets by po- and c-IBD-SBP-YAP1C vary between different time points (e.g., compare CAT at 5 and 30 min), demonstrates that at least a portion of these complexes was formed inside peroxisomes. However, it is also clear that, at least for HSD17B4, complex formation inside peroxisomes is a rather negligible phenomenon compared to complex formation in the cytosol.

Unlike external H₂O₂ addition, peroxisomal H₂O₂ production did not result in po-IBD-SBP-YAP1C-mediated trapping of ACOX1 or MDH1. Nonetheless, such treatment did result in the trapping of various cytosolic, cytoskeletal, and plasma membrane-associated proteins (Supplementary Table S4), many of which are known to be sulfenylated on at least one cysteine residue (Wang et al., 2021). Importantly, given that 1) we and others have previously demonstrated that peroxisome-derived H₂O₂ can efficiently permeate across the peroxisomal membrane (Mueller et al., 2002; Lismont et al., 2019a; Laporte et al., 2020), 2) a small portion of po-IBD-SBP-YAP1C is mislocalized to the cytosol (Figure 1), and 3) the fraction of interactor bound to po-IBD-SBP-YAP1C at the peak of protein sulfenylation is on average less than 10% of the amount of interactor bound to c-IBD-SBP-YAP1C, it is safe to conclude that the majority of cytosolic (e.g., PSMA7), cytoskeletal (e.g., POF1B), and plasma membrane-associated (e.g., ANXA2) interactors were trapped by the residual cytosolic fraction of po-IBD-SBP-YAP1C (Supplementary Figure S6). However, for some cytosolic interactors (e.g., PRDX1, PRDX2, SKP1, and TXN), the amount of po-IBD-SBP-YAP1C-bound protein was much higher than what one would expect from the estimated percentage of mislocalized po-IBD-SBP-YAP1C (Supplementary Figure S7). Although it was not the scope of this work to dig into the subcellular localization of all these and other targets of peroxisome-derived H₂O₂, we decided to explore this intriguing observation in more detail for PRDX1. Upon immunoblot analysis of subcellular fractions derived from HEK-293 cells or rat liver, a small but significant portion of PRDX1 appears to be associated with peroxisomes (Figure 6), thereby strengthening our working hypothesis.

A careful examination of the 53 different mt-IBD-SBP-YAP1C interactors (Supplementary Table S5) revealed the presence of nine proteins (ATP5F1A, ATP5F1B, GSR, HSP90AA1, HSPD1, PC, PCCB, PRDX3, and VDAC1) with a bona fide (partial) mitochondrial localization (Rath et al., 2021), thereby providing direct molecular evidence for the previously established peroxisome-mitochondria redox connection (Lismont et al., 2015). As observed for po-IBD-SBP-YAP1C (Supplementary Table S4; Supplementary Figure S6), mt-IBD-SBP-YAP1C also captured some cytosolic (e.g., S100A14), cytoskeletal (e.g., POF1B), and plasma membrane-associated (e.g., ANXA2) interactors (Supplementary Figure S8). This observation strengthens our prior interpretation (Supplementary Table S2) that also a small portion of mt-IBD-SBP-YAP1C is not yet imported into mitochondria. For comparison, we also included PRDX3, a mitochondrial member of the PRDX family (Supplementary Figure S8).

We also identified a subset of 429 different c-IBD-SBP-YAP1C interactors (Supplementary Table S3). KEGG pathway analysis revealed enrichment for proteins implicated, among others, in ribosome biology, carbon metabolism, biosynthesis of amino acids, and proteasome functioning (the -log₁₀ (p_adj) values are 5.162 × 10^–27, 3.512 × 10^–8, 2.293 × 10^–6, and 1.325 × 10^–6, respectively). Surprisingly, despite the fact that it is well known that H₂O₂ can, directly or indirectly, oxidatively modify different classes of proteins involved in signal transduction (e.g., kinases, phosphatases, proteases, antioxidant enzymes, transcription factors, etc.), no transcription factors were found to interact with c-IBD-SBP-YAP1C. One potential explanation for this finding is that members belonging to this group of proteins are oxidatively modified via redox relay, and not through direct oxidation of redox-sensitive cysteines to sulfenic acid (see Discussion). Another interesting observation is that with the exception of proteins involved in the maintenance of cellular redox homeostasis (e.g., GSR, PRDX1, PRDX2, PRDX3, PRDX4, PRDX6, TXN, and ERP44), virtually all other interactors display a sulfenylation peak at 5 or 15 min (Figure 5C). In addition, proteins belonging to the same protein family share in general a common pattern (e.g., compare the peak time values in Figure 5 and Supplementary Figure S9). Protein families that, besides the PRDXs (Figure 5A) and 14-3-3 (Figure 5B), are abundantly modified by peroxisome-derived H₂O₂ include constituents of the cytoskeleton (Supplementary Figures S9A,B), annexins (Supplementary Figure S9C), protein chaperones (Supplementary Figure S9D), S100 proteins (Supplementary Figure S9E), and negative regulators of endopeptidase activity (Supplementary Figure S9F). Finally, it is worth noting that some c-IBD-SBP-YAP1C interactors (e.g., ANXA2, DSG1, DSP, FLG2, GAPDH, and JUP) already appear to be considerably sulfenylated under basal conditions (Supplementary Table S3). Once again, this observation strongly supports a role for these proteins in redox-regulated housekeeping signaling pathways. However, another potential explanation is that the high basal sulfenylation state of some of these proteins is a confounding effect of the hyperoxic in vitro cell culture environment (typically ∼18% O₂), which differs from the in vivo situation (∼1–6% O₂) (Stuart et al., 2018).

In a laboratory setting (patho)physiological scenarios of how H₂O₂ can drive cellular signaling events are often mimicked by the external addition of this oxidant to cultured cells. In order to assess how H₂O₂ levels and sulfenylation responses in our DD-DAO-based approach compare to this often-used strategy, we treated cells with different concentrations of H₂O₂, ranging between 10 µM and 1 mM, for 10 min (Supplementary Figure S10). Using cells expressing c-IBD-SBP-YAP1C, we identified 326 proteins that were 2.5-fold or more enriched in at least one of the treated conditions (Supplementary Table S6). Note that, due to the increased sensitivity settings in this LC-MS/MS run, the number of hits greatly exceeded the number of targets identified in the initial validation experiment (Supplementary Table S1).

The number and overlap of interactors identified upon treatment of the cells with different H₂O₂ concentrations are visualized in Figure 7A. From these data, it can be deduced that treatment with 100 µM H₂O₂ yields the highest number of sulfenylated proteins. This can be explained by the fact that, at this concentration of H₂O₂, 60% of all targets reach their sulfenylation peak (Figure 7B), including the antioxidant defense enzymes (Supplementary Table S6) that did not even reach their peak within 1 h of peroxisomal H₂O₂ production (Supplementary Table S3). Interestingly, PRDX5 is the only PRDX that was exclusively sulfenylated by externally added H₂O₂ (Supplementary Table S6). This may be explained because 1) PRDX5 has the lowest expression level of all PRDXs in HEK-293 cells (Geiger et al., 2012) and 2) its preferred substrates are lipid peroxides instead of hydrogen peroxide (Knoops et al., 2011). Given that PRDX5 is already more than 2.5-fold enriched from 10 µM H₂O₂ onwards, peroxisomal H₂O₂ production by DD-DAO is far less stringent than the external addition of 10 µM H₂O₂. Taking into consideration that data in the literature suggest a 390- to 650-fold concentration difference between the extra- and intracellular H₂O₂ levels (Huang and Sikes, 2014; Lyublinskaya and Antunes, 2019), peroxisomal H₂O₂ production most likely results in intracellular concentrations of less than 15–26 nM. This claim is supported by the observation that within the time frame of the experiment (60 min), the equilibrium between peroxisomal H₂O₂ production by DD-DAO and the cell’s antioxidant defense mechanisms has not yet been reached (as demonstrated by the fact that antioxidant enzymes did not reach their sulfenylation peak). Here, it is also worth mentioning that there is relatively little overlap between the protein targets of peroxisome-derived and externally added H₂O₂ (Figure 8A). When comparing the sulfenylation profiles upon addition of external H₂O₂ (concentration curve) or peroxisomal H₂O₂ production (time curve), it is clear that depending on the H₂O₂ source, different proteins show distinct responses. A set of examples is shown in Figures 8B–J. Altogether, these findings highlight that data obtained with external H₂O₂ cannot simply be extrapolated to internally produced H₂O₂, which reflects a more physiological condition.

Strikingly, whereas upon peroxisomal H₂O₂ production the sulfenylation peak rapidly decreases to baseline levels for most proteins, treatment with external H₂O₂ most often results in a slower and rather modest decrease (Figures 8B–J, compare the blue and red profiles). Given that we have previously shown that po-DD-DAO-mediated H₂O₂ production results in steadily increasing levels of cytosolic H₂O₂ and disulfide bond formation over time (Lismont et al., 2019a), a potential explanation is that the fast decrease in sulfenylation observed under this condition represents a combined effect of 1) a thioredoxin (TXN)-mediated reduction of the disulfide bond between IBD-SBP-YAP1C and its interactors (Izawa et al., 1999), and 2) a continuous depletion of freely available redox-sensitive cysteine thiols (e.g., due to overoxidation or disulfide bond formation with other proteins). Importantly, our observation that under conditions of oxidative stress transient disulfide bond formation between HSPB1 and c-IBD-SBP-YAP1C (Figure 9A) precedes previously reported (Zavialov et al., 1998) disulfide-mediated changes in the oligomeric state of HSPB1 (Figure 9C), is in line with this view. On the other hand, upon external H₂O₂ addition, the modest decrease in sulfenylation after peaking may be explained by reduced availability of TXN and other thiol-disulfide reductases as a consequence of their overoxidation, a phenomenon supported by the observation that also these enzymes themselves are less sulfenylated at these H₂O₂ concentrations (Figures 8H–J). Finally, the observation that proteins can be released from c-IBD-SBP-YAP1C underscores the transient nature of these disulfide bonds, even in an oxidizing environment.

The findings presented thus far clearly demonstrate that the YAP1C-based sulfenome mining approach is a very powerful and efficient tool to identify targets of peroxisome-derived H₂O₂. As mentioned above and described elsewhere (Lismont et al., 2019a), we used Flp-In T-REx 293-derived cell lines in which the expression levels and stability of po-DD-DAO can be strictly controlled to overcome possible interfering effects of H₂O₂ produced by newly synthesized DD-DAO that has not yet been imported into peroxisomes. To document the validity of this assumption at the proteome level, we also performed a sulfenome mining experiment with Flp-In T-REx 293 cells expressing c-DD-DAO and c-IBD-SBP-YAP1C (Supplementary Figure S11). Importantly, even in a condition where initially 100% of DD-DAO was located in the cytosol, only 8 out of the 226 identified targets were enriched in the chase condition (Figure 10A; Supplementary Table S7). In addition, to gain more insight into target proteins that may be indirectly retained on the affinity matrix through electrostatic interaction with truly sulfenylated proteins (Liebthal et al., 2020), Flp-In T-REx 293 cells expressing c-IBD-SBP-YAP1C were treated with 1 mM H₂O₂ for 10 min and, after enrichment of the corresponding c-IBD-SBP-YAP1C complexes and completing the normal washing procedure, the streptavidin column was first three times eluted with high salt (1 M NaCl) and subsequently three times with DTT (this experiment was done in parallel with the experiment shown in Supplementary Figure S10). LC-MS/MS analysis of the eluates revealed the presence of 294 distinct interactors, of which only 9 predominantly eluted in a NaCl-dependent manner (Figure 10B; Supplementary Table S8). These findings confirm that the vast majority of targets identified are bona fide sulfenylated proteins. Finally, one can argue that one biological replicate per condition may not suffice to draw reliable conclusions. However, the aim of this study was not to generate a full inventory of the responsiveness of redox-sensitive proteins to peroxisome-derived H₂O₂, but rather to provide an insight into the dynamics and localization of the major H₂O₂ targets. Therefore, we have adopted extremely stringent validation criteria for peptide identification (PEP of peptide spectral match (PSM) < 10^–3 in at least one of the conditions of the experiment), resulting in proteins that are very confidently identified but that are only the tip of the iceberg in terms of protein abundance. In addition, in the following section, we discuss additional observations that support the reliability of our data.

To assess the reproducibility of our sulfenome mining approach, we exploited the fact that small but significant amounts of po- and mt-IBD-SBP-YAP1C are still located in the cytosol (Figure 1, Supplementary Table S2). Specifically, we selected all predominantly cytosolic targets of peroxisome-derived H₂O₂ that were trapped in independent experiments by c-IBD-SBP-YAP1C (Supplementary Table S3), po-IBD-SBP-YAP1C (Supplementary Table S4), or mt-IBD-SBP-YAP1C (Supplementary Table S5). Strikingly, upon comparison, only 5 out of the 37 cytosolic interactors of po-IBD-SBP-YAP1C and 3 out of the 21 cytosolic interactors of mt-IBD-SBP-YAP1C were not present in the pool of c-IBD-SBP-YAP1C interactors (Figure 11A). This clearly demonstrates that the interactors identified are specific and not merely a random result of a single experiment. The consistency of our data is further strengthened by the fact that most of the po- and mt-IBD-SBP-YAP1C cytosolic interactors identified correspond to the more abundantly captured interaction partners of c-IBD-SBP-YAP1C (Figure 11B). Note that the different thresholds of sensitivity in the trapping of sulfenylated proteins in the cytosol by c-, po-, and mt-IBD-SBP-YAP1C can be explained by their relative concentrations in this subcellular compartment.

To validate our findings on a more conceptual level, we connected our data to previous observations that oxidative stress rewires cellular carbohydrate metabolism (Grant, 2008; Mullarky and Cantley, 2015). In this context, it is important to highlight that 1) NADPH, generated during the metabolism of glucose via the oxidative arm of the pentose phosphate pathway (Ox-PPP), acts as an important redox cofactor that fuels most cellular antioxidant systems, and 2) oxidative stress can inhibit glycolytic enzymes as a controlled response that redirects the metabolic flux from glycolysis to PPP (Grant, 2008). Both carbohydrate fluxes are directly connected via glucose-6-phosphate, a common intermediate that, upon isomerization to fructose-6-phosphate, continues through glycolysis and, upon oxidation to 6-phospho-D-glucono-1,5-lactone, enters the Ox-PPP (Figure 12). While the Ox-PPP enzyme activities appear to be maintained during oxidant exposure, such treatment has been shown to inhibit multiple glycolytic enzymes, including GAPDH and PKM, by directly oxidizing cysteine residues (Mullarky and Cantley, 2015). Oxidative inhibition of GAPDH and PKM have respectively been shown 1) to redirect glycolytic flux towards Ox-PPP (through an accumulation of metabolites upstream of GAPDH), and 2) to increase the biosynthesis of glycine and cysteine, two glutathione precursors (through the build-up of 2-phosphoglycerate and subsequent activation of the serine synthesis pathway), thereby conferring cells resistant to oxidative stress (Mullarky and Cantley, 2015). Interestingly, a careful analysis of our sulfenome mining data indicated that multiple glycolytic enzymes, including GAPDH and PKM, are also targets of peroxisome-derived H₂O₂ (see Figure 12).

Lastly, also the conserved sulfenylation profiles between targets belonging to the same protein family (Supplementary Figure S9) as well as the observation that a decrease in the sulfenylation peak of HSPB1 coincides with an increase in disulfide bond formation (Figure 9) support the idea that our findings are solid.

In our protocol, free and oxidatively modified cysteines are differentially alkylated (Figure 13), providing more information on which cysteines are oxidized. In short, after cell treatment, free thiols are irreversibly blocked with NEM, cells are lysed, and IBD-SBP-YAP1C-containing protein complexes are enriched on the streptavidin affinity matrix. Upon DTT elution, the originally sulfenylated cysteines captured by IBD-SBP-YAP1C are reduced and become available for iodoacetamide alkylation, thereby resulting in the formation of an S-carbamidomethyl cysteine. As LC-MS/MS analysis can distinguish between different cysteine modifications, it is possible to determine at what stage in the protocol the cysteines of the identified peptides were alkylated. Here, it is essential to point out that S-carbamidomethylation does not necessarily indicate that the cysteine was indeed sulfenylated, since S-carbamidomethylated cysteines can also result from pre-existing disulfide bridges or cysteines that are inaccessible to NEM. In addition, as LC-MS/MS analysis does not detect every peptide of a protein, the detection of redox-sensitive cysteine-containing peptides is no absolute requirement to categorize a hit as an authentically sulfenylated protein. In the 784 IBD-SBP-YAP1C-interactors that were identified during the time course of our experiments, we could detect 150 proteins containing at least one S-carbamidomethyl cysteine (Supplementary Table S9). As could be expected, the number of detected S-carbamidomethylated cysteines is determined by the sensitivity settings of the LC-MS/MS run as well as by the stringency of the treatment which intrinsically correlates with peptide abundance (e.g., treatment of cells with external H₂O₂ resulted in more S-carbamidomethyl cysteines than H₂O₂ production inside peroxisomes). A comparative analysis of our data with the list of proteins in the iCysMod database revealed that, of the 337 redox-sensitive cysteines identified, 91 cysteines have already been reported to be sulfenylated by others, 172 cysteines were already shown to undergo other oxidative thiol modifications, and 74 cysteines represent novel redox-regulated cysteines in humans (Supplementary Table S9).