A chemoproteomic platform to quantitatively map targets of lipid-derived electrophiles

Abstract

Cells produce electrophilic products with the potential to modify and affect the function of proteins. Chemoproteomic methods have provided a means to qualitatively inventory proteins targeted by endogenous electrophiles; however, ascertaining the potency and specificity of these reactions to identify the most sensitive sites in the proteome to electrophilic modification requires more quantitative methods. Here, we describe a competitive activity-based profiling method for quantifying the reactivity of electrophilic compounds against 1000+ cysteines in parallel in the human proteome. Using this approach, we identify a select set of proteins that constitute “hot spots” for modification by various lipid-derived electrophiles, including the oxidative stress product 4-hydroxynonenal (HNE). We show that one of these proteins, ZAK kinase, is labeled by HNE on a conserved, active site-proximal cysteine, resulting in enzyme inhibition to create a negative feedback mechanism that can suppress the activation of JNK pathways by oxidative stress.

The functional diversity of mammalian proteomes is enriched by the post-translational modification (PTM) of proteins¹. Many PTMs are enzyme-catalyzed, but others reflect the direct (non-enzymatic) oxidative or electrophilic modification of nucleophilic residues, such as cysteine, by reactive small-molecules that are often the products of oxidative stress^2,3. Chronic inflammation or hypoxia, for instance, induces the peroxidation of polyunsaturated lipids to generate a broad range of electrophilic products². These lipid-derived electrophiles (LDEs) can modify DNA and proteins to promote cytotoxicity and have been implicated in the pathogenesis of many diseases⁴. A growing body of studies also suggests that, at lower and more physiological concentrations, LDEs serve as endogenous messengers that modulate the response of signaling pathways^5,6. 4-Hydroxynonenal (HNE), for instance, is a major product generated when free radicals initiate the non-enzymatic fragmentation of lipids in biological membranes^2,7. The levels of HNE and HNE-protein adducts are elevated in cells and tissues exposed to oxidative stress, and HNE can regulate redox-responsive signaling pathways by still poorly understood mechanisms^2,8,9. 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2) is another LDE produced by a set of enzymes that metabolize arachidonic acid¹⁰. 15d-PGJ2 exhibits anti-inflammatory and cytoprotective properties and has therefore been designated as a pro-resolving signal¹⁰. A third example is the LDE 2-trans-hexadecenal (2-HD), which is a product of sphingolipid metabolism and has recently been shown to function as a protein-modifying cofactor that promotes mitochondrial pathways for apoptosis¹¹.

Understanding the protein targets of LDEs is critical for elucidating their cellular functions and mechanisms of action. Here, chemoproteomic methods have proven particularly useful for inventorying a large number of proteins that react with LDEs in cells and tissues^2,12–17; however, quantifying the potency and specificity of these reactions to identify the most sensitive sites in the proteome to electrophilic modification has proven challenging. Here, we describe a competitive activity-based protein profiling (ABPP) method for quantifying the reactivity of electrophilic compounds against 1000+ cysteines in parallel in the human proteome. Using this approach, we identify select sets of proteins that are preferentially modified by HNE and 15d-PGJ2. We show that one of these proteins, ZAK kinase, is labeled by HNE on a conserved, active site-proximal cysteine residue, which inhibits the enzyme and suppresses the activation of JNK pathways by oxidative stress in cancer cells.

Results

Quantitative proteomic profiling of LDE-cysteine reactions

Among the 20 protein-coding amino acids, cysteine is unique owing to its high nucleophilicity, which renders its sensitivity to modification by endogenous and exogenous electrophiles and oxidants³. Cysteine reactions with electrophilic metabolites have been characterized for purified proteins^18,19, and, on a global scale in cells and tissues using mass spectrometry-based chemoproteomic^2,12–17 and imaging methods²⁰. These studies, along with analytical, quantum mechanical, and kinetic work^21,22, have, for the most part, confirmed the preferential reactivity that Michael acceptor electrophiles like HNE show for cysteine over other nucleophilic amino acids (e.g., lysine, histidine) in proteomes. We were interested in building on these past findings to determine whether individual cysteines in the proteome display differences in their reactivity with LDEs, and, if so, whether potential hot-spots for electrophile modification might constitute key nodes in signaling pathways of redox sensing and response.

We previously described a chemoproteomic method termed isoTOP-ABPP (isotopic Tandem Orthogonal Proteolysis-ABPP) and its use to quantify the intrinsic reactivity of cysteine residues in cell and tissue proteomes²³. Here, we envisioned that isoTOP-ABPP could be advanced to discover and quantify reactions between cysteines and electrophilic small-molecules in proteomes. In this ‘competitive’ version of isoTOP-ABPP (Fig. 1a, b), a proteome is treated with an electrophile (experimental sample) or DMSO (control sample). Both proteomic samples are then labeled with an alkynylated iodoacetamide (IA) probe (Fig. 1b) and conjugated by copper-catalyzed azide-alkyne cycloaddition (CuAAC or click²⁴) chemistry to light and heavy azide-biotin tags, respectively, each containing a Tobacco Etch Virus (TEV) cleavage sequence. The light and heavy samples are then mixed and subjected to our described isoTOP-ABPP protocol for peptide enrichment, cleavage, and identification, where IA-modified cysteines are identified and their extent of labeling quantified based on MS2 and MS1 profiles, respectively²³. Electrophile-sensitive cysteines are quantified by measuring the MS1 chromatographic peak ratios (R values) for heavy (DMSO-treated) over light (electrophile-treated) samples, with higher R values reflecting greater sensitivity to the electrophile. In this format, competitive isoTOP-ABPP can assay electrophiles against 1000+ cysteines in parallel directly in native proteomes without requiring any chemical modification to the electrophiles themselves.

Figure 1.

Competitive isoTOP-ABPP for quantitative mapping of cysteine-lipid-derived electrophile (LDE) reactions in proteomes. (a) Competitive isoTOP-ABPP involves treatment of proteomes with DMSO or LDE, proteome labeling with an IA-alkyne (IA) probe, CuAAC-based incorporation of isotopically-labeled, TEV protease-cleavable biotin tags, enrichment with streptavidin, and sequential on-bead protease digestions to afford probe-labeled peptides for MS analysis. (b) Structures of the IA-probe and the competitive blockade of IA-cysteine reactions by an LDE. (c) Structures of three LDEs, HNE, 15d-PGJ2 and 2-HD, used in competitive isoTOP-ABPP experiments with their sites of reactivity marked with asterisks.

We applied competitive isoTOP-ABPP to quantitatively profile the proteome reactivity of three LDEs -- HNE, 15d-PGJ2, and 2-HD -- each of which possesses an α, β–unsaturated carbonyl that can react with nucleophilic cysteines via Michael addition (Fig. 1c). Competitive isoTOP-ABPP experiments were performed in quadruplicate using the soluble proteome of the human breast cancer cell line MDA-MB-231. Proteomes were treated with 100 μM of HNE, 15d-PGJ2, or 2-HD for 60 min followed by the IA-probe (100 μM, 60 min). A total of ~1000 cysteine reactivities were quantified across the aggregate data set, with at least 750 cysteine reactivities quantified for each LDE (Fig. 2a) more than 550 of which were quantified for all the three LDEs (Supplementary Fig. 1 and Supplementary Table 1). Most of the cysteine reactivities (> 98%) were unaffected or only marginally affected by LDE treatment (R < 5); however, a select subgroup showed marked reductions in their IA-probe reactivities (R > 5) following exposure to one or more LDEs (Fig. 2a). HNE and 15d-PGJ2 both targeted several cysteines in the proteome, the majority of which showed preferential reactivity with one of the two LDEs, while 2-HD exhibited no detectable high-sensitivity (R > 5) targets (Fig. 2a). We also analyzed LDEs in a second human cancer cell proteome (Ramos cells), which increased our proteome coverage (> 900 cysteines reactivities quantified for all three LDEs) and revealed good overall correlation of reactivity profiles compared to MDA-MB-231 proteomes (Supplementary Fig. 2).

Figure 2. Quantitative profiling of LDE-cysteine reactions in proteomes.

(a) Distribution of competitive isoTOP-ABPP ratios (R values) quantified from reactions with the human MDA-MB-231 proteome treated with 100 μM HNE (left, 5 technical replicates from 4 biological replicates), 15d-PGJ2 (middle, 6 technical replicates from 5 biological replicates), or 2-HD (right, 4 technical replicates from 3 biological replicates). A cut-off of five-fold or greater blockade of IA-probe labeling (R values > 5) is shown by a dashed line to mark cysteines that exhibit high sensitivity to LDEs, and proteins with cysteines showing the strongest competitive reactivity with LDEs are labeled by names. (b) Heat map of cysteines with R values > 5 illustrating examples of cysteines that display selectivity for reacting with one of the three tested LDEs. (c) Representative MS1 profiles for peptides containing cysteines that show selective competition with HNE (left) or 15d-PGJ2 (right). (d) Representative MS1 profiles for multiple cysteine-containing peptides from the same protein, only one of which shows sensitivity to LDE competition.

The quantitative ranking of cysteines based on the magnitude and selectivity of their inhibition illuminated “hot spots” for LDE reactivity in the proteome (Fig. 2a, b and Supplementary Table 2 and Supplementary Fig. 2). Examples included Cys22 of ZAK and Cys848 of MMS19, which were strongly and selectively blocked by HNE and 15d-PGJ2, respectively (Fig. 2c). Competitive isoTOP-ABPP also identified several proteins that possess multiple IA-reactive cysteines, only one of which proved sensitive to competitive blockade by an LDE (Fig. 2d). These data demonstrate that the quantified R values reflect measurements of individual LDE-cysteine reactions rather than general changes in protein abundance potentially caused by LDE exposure. We also found that LDE-sensitive cysteines spanned a broad range of intrinsic reactivity (Supplementary Fig. 3), as determined previously by measuring their extents and rates of IA labeling²³, suggesting that their modification by LDEs depend not only on cysteine nucleophilicity, but also on molecular recognition of the LDEs.

Determining the potency of HNE-cysteine reactions

We next focused on identifying the most sensitive sites for LDE reactivity in the proteome. The MDA-MB-231 cell proteome was treated with varying concentrations of HNE (5, 10, 50, 100 and 500 μM) for 60 min and then the IA-labeling profile of each reaction was quantitatively compared to a DMSO-control sample by isoTOP-ABPP. (Supplementary Table 3). By combining the R values at all 5 HNE concentrations, we could extrapolate IC₅₀ values for HNE-blockade of IA probe-labeling for ~700 of the ~1100 total cysteines detected in the study (Fig. 3a and Supplementary Table 3). This analysis revealed that the vast majority of cysteines were modified by HNE with low potency (IC₅₀ values > 100 μM), but a select few cysteines, including C22 of ZAK, C41 of EEF2, C24 of FN3KRP, and C1001 of RTN4, exhibited much higher sensitivities with IC₅₀ values ranging from 6 to 23 μM (Fig. 3a, b). We next confirmed that these hypersensitive cysteines were also inhibited by HNE in situ by treating MDA-MB-231 cells with 50 or 100 μM HNE and then preparing proteomes for analysis by competitive isoTOP-ABPP (Fig. 3c, d and Supplementary Table 4). Another set of cysteines showed reductions in IA-probe labeling in situ, but not in vitro (Fig. 3d, black diamonds). This finding suggests that certain proteins may preferentially react with HNE in living cells, although we cannot exclude, at this point, that the reductions in IA-probe labeling observed for these proteins reflect a decrease in their overall abundance in HNE-treated cells.

Figure 3.

Determining the potency of HNE-cysteine reactions in proteomes and in cells. (a) Box-and-whisker (maximum 1.5 IQR) plots showing the distribution of R values for ~1100 cysteines quantified from competitive isoTOP-ABPP experiments with the MDA-MB-231 proteome treated with 5, 10, 50, 100, and 500 μM HNE. The IA-labeling of cysteines from ZAK, EEF2, RTN4, and FN3KRP exhibit exceptional sensitivity to HNE competition compared to the rest of the cysteines in the proteome. The data reported here is representative of an experiment run in duplicates. (b) Representative MS1 profiles for HNE-sensitive cysteines in ZAK and RTN4 showing concentration-dependent blockade of IA-labeling by HNE. (c) Distribution of R values quantified from competitive isoTOP-ABPP experiments with proteomes from MDA-MB-231 cells treated in situ with DMSO or HNE (100 μM, 60 min), confirming that cysteines in ZAK, EEF2, RTN4 and FN3KRP are also highly sensitive to HNE competition in living cells (2 biological replicates). (d) Comparison of R values obtained from in vitro versus in situ competitive isoTOP-ABPP experiments. Red and black diamonds mark cysteines that show similar or different in vitro versus in situ R values, respectively.

Functional analysis of HNE modification of ZAK kinase

ZAK kinase (also known as MLK7 or MLTK) functions as a mitogen-activated kinase kinase kinase (MAP3K) that can activate all three major MAPK (ERK, JNK, and p38) pathways in mammalian cells^25–27, with some preference for JNK²⁶, and is involved in response pathways to stressors such as UV radiation²⁸ and chemotherapeutic agents²⁷. Sequence and structure comparisons allowed us to map the HNE-sensitive cysteine in ZAK (C22) to a location proximal to the glycine-rich ATP-binding loop (“P-loop”) (Fig. 4a). Interestingly, among all 20 human MAP3Ks, ZAK is the only member that possesses a cysteine at this position (Fig. 4a), and this cysteine is highly conserved across ZAK orthologues in vertebrates (Supplementary Fig. 4). This information, combined with the high sensitivity displayed by C22 for HNE (Fig. 3a) motivated us to further characterize this interaction and its impact on ZAK activity.

Figure 4.

Functional characterization of HNE modification of ZAK kinase. (a) Crystal structure of human MAP3K9 (left, PDB: 3DTC) and multiple sequence alignment of ZAK with other 19 human MAP3Ks showing the HNE-sensitive cysteine C22 of ZAK is located next to the kinase’s ATP binding loop (“P-loop”; note that C22 corresponds to I150 in MAP3K9) and is unique to ZAK relative to other MAP3K enzymes. (b) Selective IA-labeling of wild-type (WT), but not C22A-ZAK, and concentration-dependent competition of IA-labeling of WT-ZAK by HNE as measured by gel-based ABPP using an IA-rhodamine probe. ZAK were expressed as FLAG-tagged proteins in HEK293T cells by stable transfection and immunoprecipitated prior to IA-probe labelling and analysis. (c) An HNE-alkyne probe (HNEyne¹²) selectively labels WT-, but not C22A-ZAK in proteomes and in living cells as determined by gel-based ABPP. For b and c, the experiments were repeated three times with consistent results. (d) Catalytic activity of immunoprecipitated WT-, but not C22A-ZAK is inhibited by HNE as measured using a Myelin Basic Protein (MBP) substrate assay. A K45M-ZAK mutant, in which a conserved active-site lysine was mutated, showed no detectable activity and thus served as a catalytically dead control enzyme. All three ZAK variants (WT, C22A, and K45M were expressed at similar levels in transfected HEK293T cells; Supplementary Fig. 4). Data are presented as mean values ± s.e.m.; N = 3 experiments/group. **, P < 0.01, ***, P < 0.001, t-test, two sided. (e) SILAC-ABPP of acylphosphate-ATP probe reactivity for ZAK-transfected HEK293T cell proteomes pre-treated with DMSO (heavy cell proteome) or HNE (light cell proteome; 100 μM, 30 min) showing that the probe interaction with ZAK, but not other proteins, is substantially impaired by HNE treatment.

We first expressed FLAG-tagged versions of wild type (WT) and a C22A mutant of ZAK by stable transfection in HEK293T cells and found that WT-ZAK showed much stronger IA-rhodamine probe labeling as measured by gel-based ABPP. The IA-rhodamine labeling of WT-ZAK was blocked by pre-treatment with HNE over a concentration range that closely matched the HNE-sensitivity profile observed for C22 in competitive isoTOP-ABPP experiments (Fig. 4b, compare to Fig. 3a). We next used an alkyne-functionalized HNE probe (HNEyne)¹² to verify direct labeling of WT-, but not the C22A-ZAK mutant in vitro and in living cells (Fig. 4c). HNE also inhibited WT-, but not C22A-ZAK activity in a concentration-dependent manner, as measured with an in vitro Myelin Basic Protein (MBP) substrate assay²⁹ (Fig. 4d). We note that C22A-ZAK exhibited reduced basal activity compared to WT-ZAK, but the residual activity of C22-ZAK, which was still much greater than a catalytically dead K45M-ZAK mutant, was insensitive to HNE (Fig. 4d and Supplementary Fig. 5). Considering that C22 is predicted to reside adjacent to the ATP-binding loop of ZAK, we postulated that the HNE-induced loss of kinase activity might be due to blockade of ATP-binding. We tested this hypothesis by performing a competitive SILAC (Stable Isotope Labeling by Amino acids in Cell culture)-ABPP³⁰ experiment using an acylphosphate-ATP probe³¹, which revealed that probe-labeling of ZAK, but not other proteins, was substantially reduced in cell proteomes treated with HNE (Fig. 4e).

HNE modification of ZAK suppresses JNK activation in cells

We next evaluated the activation state of MAPK signaling pathways in HEK-293T cells stably expressing WT-ZAK, C22A-ZAK, or K45M-ZAK. WT-ZAK-expressing cells, but not C22A- or K45M-ZAK-expressing cells showed increased JNK and, to a lesser extent, p38 and ERK pathway activation compared to mock-transfected cells, as measured by anti-phosphoprotein blotting (Fig. 5a, b and Supplementary Fig. 6). We next treated cells with H₂O₂ (1 mM, 30 min) to induce oxidative stress, a process that is known to activate the JNK pathway³². H₂O₂-treatment stimulated JNK activity in both WT- and C22A-ZAK-transfected cells (but not K45M-ZAK-transfected cells), with WT-ZAK cells showing the greater level of activation (Fig. 5b,c). Strikingly, however, pre-treatment with HNE (100 μM, 30 min) produced opposing effects in WT- and C22A-ZAK cells, blocking H₂O₂-dependent JNK activation in the former cell model, while hyper-activating JNK activity in the latter. We interpret these findings to indicate the existence of both ZAK-dependent and ZAK-independent pathways for HNE modulation of JNK activation (Fig. 5d). By modifying C22 on ZAK, HNE blocks the contribution that this kinase makes to activation of the JNK pathway. HNE is also known to itself promote oxidative stress³³ that likely adds to the effects of H₂O₂ and, in the context of an HNE-resistant C22A-ZAK mutant, would serve to further augment activation of the JNK pathway. In this model, the HNE-ZAK interaction acts as a negative-feedback loop that tempers activation of the JNK pathway under high and/or persistent levels of oxidative stress (Fig. 5d). We further tested this idea by evaluating the effects of HNE alone on JNK pathway activity in ZAK-expressing cells. A dramatic concentration-dependent activation of JNK was observed in C22A-ZAK cells, but not in WT-ZAK cells, which showed higher basal JNK activation that was mostly unaffected by HNE (Fig 5e,f). While we were initially surprised that HNE treatment did not appear to block the basal JNK activation caused by WT-ZAK, we should note that HNE also activated JNK in mock-transfected cells to a level that matched the basal JNK activity observed in WT-ZAK-transfected cells. Thus, the residual JNK activation observed in WT-ZAK-transfected cells may reflect ZAK-independent pathways of JNK activation by HNE (Fig. 5e,f). That JNK activation was much higher in C22A-ZAK cells compared to the other cell models indicates this HNE-insensitive form of ZAK, which still retains some catalytic activity (see Fig. 4d), combines with ZAK-independent, HNE-stimulated pathways to further enhance JNK activation.

Figure 5.

HNE modification of ZAK suppresses JNK pathway activation in cells. (a) WT-ZAK-transfected HEK-293T cells show higher basal JNK activation compared to mock-, C22A-ZAK-, or K45M-ZAK-transfected. (b,c) Western blots (b) and normalized phosphorylated JNK levels (c) showing that H₂O₂ treatment (1 mM, 30 min) increases JNK activation in WT- and C22A-ZAK cells and this increase is blocked or amplified in WT- and C22A-ZAK cells, respectively by pre-treatment with HNE (100 μM, 30 min). (d) A model diagramming ZAK-dependent and ZAK-independent pathways for HNE modulation of JNK activation. Dashed line designates the potential for oxidative stress to generate HNE and initiate a negative feedback loop to limit JNK activation. (e,f) Western blots (e) and normalized phosphorylated JNK levels (f) showing dramatic, concentration-dependent activation of JNK by HNE (50 or 100 μM, 60 min) in C22A-ZAK cells, but not in WT-ZAK cells. Note that mock- and K45M-ZAK transfected cells also show modest, but significant elevations in JNK activity following HNE treatment, which is consistent with previous studies indicating that HNE can activate JNK by multiple pathways⁴⁰. For a, c and f, data are presented as mean values ± s.e.m.; N >= 4 experiments/group (biological replicates). *, P < 0.05; **, P < 0.01; ^##, P < 0.01; t-test, two-sided.

Discussion

Long viewed as biomarkers of oxidative damage, LDEs have more recently gained attention as second messengers that can regulate diverse cellular processes^5,34. Considering that the signaling and pathophysiological functions of LDEs may differ across the endogenous concentration ranges found for these compounds, it is imperative to understand the potencies of LDE-protein interactions in biological systems. With this goal in mind, we created a competitive isoTOP-ABPP platform to quantitatively map LDE-reactivity across 1000+ cysteines in parallel, resulting in the discovery of discrete sites of hypersensitivity, or “hot spots”, for LDE modification in the human proteome.

Among the most LDE-sensitive cysteines, C22 of ZAK captured our interest due to the proposed role that this kinase plays in activating JNK, ERK, and p38 MAPK pathways in both cancer²⁵ and inflammation²⁷. To date, only a handful of studies have investigated ZAK function and its modes of regulation remain poorly understood. That HNE inhibits human ZAK by modifying an active site-proximal cysteine conserved among ZAK orthologues, but not other MAP3K enzymes, suggests ZAK acts as a special node in MAPK signaling pathways that confers sensitivity to lipid oxidation products. In this way, HNE modification of ZAK may limit the extent of JNK activation caused by oxidative stress, which could help certain cell types, such as tumor and immune cells, survive in the presence of high levels of reactive oxygen species. Further studies of ZAK function would benefit from the development of selective inhibitors for this enzyme. It is noteworthy, in this regard, that covalent inhibitors targeting active-site cysteine residues have recently been introduced for many kinases³⁵, leading us to speculate that the C22-HNE interaction discovered herein may offer a medicinal chemistry starting point for the development of ZAK inhibitors. Beyond ZAK, we also identified several other kinases in our competitive isoTOP-ABPP experiments that possess cysteines inhibited by HNE, albeit with lower potencies (Supplementary Table 5). Prominent among these was cysteine (C311) in AKT1/2/3, which is an active site-proximal residue implicated in substrate binding³⁶ and was inhibited by HNE with an IC₅₀ value of ~60 μM (Supplementary Table 3). These proteomic findings nicely confirm recent work showing that recombinant AKT2 is modified by HNE on C311¹⁸.

We believe that competitive isoTOP-ABPP offers several advantages over more conventional proteomic approaches for the discovery and characterization of protein-small molecule reactions in biological systems. First, quantitative inhibition values are measured in relative terms that are independent of absolute protein abundance. The method is therefore able to sift through signals that span a broad range of intensities to identify reactive sites that are more likely to bear functional consequence. Here, the site-specificity afforded by isoTOP-ABPP is important, since it permits the discovery of potent electrophile-cysteine reactions that may occur on proteins that display other unaffected and/or non-functional cysteines (Fig. 2d). Endogenous electrophiles also vary considerably in their structures and the stability of the protein adducts that they form. These features can complicate the direct detection of electrophile-protein interactions in proteomic studies. By competitively assessing electrophile-protein interactions using a structurally simple IA-probe that forms stable adducts with cysteines, isoTOP-ABPP affords a robust and uniform way to detect and quantify cysteine-electrophile adducts in proteomes by MS. The method should also prove applicable to profiling electrophile-protein reactions in primary cells and/or tissues²³, since isotopic tags are conveniently appended to these adducts by click chemistry.

There are, however, some potential shortcomings of competitive isoTOP-ABPP that should also be discussed. Because this method indirectly identifies sites of electrophile action by competitive displacement, it remains possible that, in certain instances, blockade of IA-probe labeling could be due to electrophile modification of a neighboring residue rather than the IA-labeled cysteine itself. We addressed this potential concern for ZAK by follow-up studies using a clickable HNE probe¹², which labeled WT- but not a C22A-ZAK mutant. We envision that a similar strategy could be used to confirm additional LDE-cysteine interactions discovered in competitive isoTOP-ABPP experiments. Competitive isoTOP-ABPP may also fail to detect certain electrophile-cysteine reactions that occur on very low abundance proteins or on non-proteotypic peptides incompatible with MS detection. Alternative protease digestion protocols should prove capable of identifying at least a subset of these electrophile-cysteine reactions. We should further note that, while the potency of LDE-cysteine reactions is likely important for proteins that are inhibited by LDEs in cells, this parameter may be less relevant for LDE-cysteine reactions that stimulate protein function^2,9. Indeed, only a modest level of LDE modification may be needed to activate some signaling pathways, such as, for instance, the HNE-mediated induction of the heat-shock response by modifying heat-shock proteins to release bound HSF1^2,9. We identified several lower-potency, IA-reactive cysteines in Hsp90 (Supplementary Fig. 7), and it is possible that one or more of these HNE-cysteine reactions contributes to disrupting Hsp90-HSF1 interactions. Finally, natural electrophiles are also known to target other amino acids beyond cysteine^14,37, and these reactions could be quantitatively profiled by adapting competitive isoTOP-ABPP to map other nucleophilic amino acids in proteomes. Here, the continued exploration of simple reactive groups that preferentially and stably modify specific amino acid side chains in proteins is warranted^38,39.

ONLINE METHODS

Preparation of human cancer cell line proteomes

MDA-MB-231 cells (from ATCC) were grown in L15 media supplemented with 10% fetal bovine serum at 37 °C in a CO₂-free incubator. For in vitro labelling experiments, cells were grown to 100% confluency, washed three times with PBS and scraped in cold PBS. Ramos cells (from ATCC) were grown in RPMI media supplemented with 10% fetal bovine serum at 37 °C in a 5% CO₂ incubator and cells were harvest at the density of 1x10⁶ cells/mL. MDA-MB-231 and Ramos cell pellets were isolated by centrifugation at 1400 x g for 3 min, and the cell pellets stored at −80 °C until further use. The harvested cell pellets were lysed by sonication in PBS buffer and fractionated by centrifugation (100,000 x g, 45 min) to yield soluble and membrane proteomes. The proteomes were prepared fresh from the frozen cell pallets prior to each experiment.

In vitro LDE treatment

HNE was purchased from EMD biosciences, 15d-PGJ2 was purchased from Cayman Chemicals and 2-HD was purchased from Santa Cruz Biotechnology. Proteome samples were diluted to a 4 mg protein/mL solution in PBS. For each profiling experiment, one aliquot of the proteome sample (0.5 mL) was treated with 100 μM of LDE using 5 μL of a 10 mM stock and the other aliquot was treated with 5 μL of either ethanol (for HNE and 15d-PGJ2) or DMSO (for 2-HD) as control. Samples were prepared in at least four replicates for each proteome-LDE combination. For the concentration-dependent profiling experiments using HNE, aliquots of the proteomes (0.5 mL each) were treated with 5, 10, 50, 100 and 500 μM of HNE using 5 μL of 0.5, 1.0, 5, 10 and 50 mM of stock solution, respectively. After 60 min of treatment at room temperature, both the LDE-treated and control aliquots were passed through a NAP-5 Sephadex column (GE Healthcare) to remove any unreacted LDE. The volume of each aliquot was increased to 1 mL with the concentration at 2 mg/mL after this step.

In situ HNE treatment

After MDA-MB-231 cells were grown to 100% confluency, the media was removed and replaced with fresh serum-free media containing 100 μM HNE (20 μL of 50 mM stock in 10 mL media). A control flask of cells was treated with 10mL of serum-free media containing 20 μL of ethanol in parallel. The cells were incubated at 37 °C for 1 hour and harvested as detailed above to prepare HNE-treated and control proteomes, respectively. Samples were prepared in duplicate.

Protein labelling and click chemistry

Each of the control and LDE-treated proteome samples (~2 mg protein/mL in 1mL volume) was treated with 100 μM of IA-probe using 10 μL of a 10 mM stock in DMSO. The labeling reactions were incubated at room temperature for 1 hour. Click chemistry was performed by the addition of 100 μM of either the Heavy-TEV-Tag (for the control sample) or Light-TEV-Tag (for the LDE-treated sample) (20 μL of a 5 mM stock), 1 mM TCEP (fresh 50X stock in water), 100 μM ligand (17X stock in DMSO:t-Butanol 1:4) and 1 mM CuSO₄ (50X stock in water). Samples were allowed to react at room temperature for 1 hour. After the click chemistry step, the light and heavy-labelled samples were mixed together and centrifuged (5900 x g, 4 min, 4 °C) to pellet the precipitated proteins. The pellets were washed twice in cold MeOH, after which the pellet was solubilized in PBS containing 1.2% SDS via sonication and heating (5 min, 80 °C). Samples were subjected to streptavidin enrichment of probe-labelled proteins, sequential on-bead trypsin and TEV digestion, and liquid chromatography-mass spectrometry (LC-MS) analysis according to the published isoTOP-ABPP protocol²³.

MS Data Analysis

For each LC/LC-MS/MS run, IA-probe labelled peptides were identified by SEQUEST⁴¹ and DTASelect⁴², and the quantification of heavy/light ratios (isoTOP-ABPP ratios, R) was performed by an in-house software (CIMAGE) as previously described²³. The software was advanced to be able to detect and quantify cases where near complete LDE blockade of IA-probe labelling was achieved (e.g., very small or no light peak) and assign an empirical ratio cut-off of 15 to each of such cases. After ratios for unique peptides were calculated for each run, overlapping peptides with the same labelled cysteine (e.g., same local sequence around the labelled cysteines but different charge states, MudPIT segment numbers, or tryptic termini) were grouped together (as a “peptide entry”), and the median ratio from each group was recorded as the “R” ratio of this peptide entry in this run.

For the in vitro profiling with 3 different LDEs, each experiment (one proteome-LDE combination) contained at least 4 replicate runs. The R ratios of each peptide entry from multiple replicate runs were averaged and the mean value was reported as its final R ratio in this experiment. Cases were excluded from further analysis when large variability in R ratios across multiple replicate runs were observed, (s.d. > 2/3 of the mean) or R ratios were not quantified in two or more replicate runs for any of the six proteome-LDE experiments.

For the in vitro profiling with HNE competition at different concentrations, all runs were searched using SEQUEST and filtered with DTASelect as described above. Because the mass-spectrometer was configured for data-dependant fragmentation, peptides are not always identified in every run. In the case of probe-modified peptides that were sequenced in one, but not the other runs, a featured algorithm of CIMAGE was utilized to identify the corresponding peak pairs in the runs without the SEQUEST identification and obtain quantification as previously described.

For the in situ profiling with HNE at 50 and 100 μM, two replicate samples were prepared for each condition (4 runs in total). Peptide entries without quantified ratios from at least two runs (either two different concentrations or two replicates at the same HNE concentration) were removed from further analysis.

In all cases, the false-positive rate after quantification was found to be less than 1%. All of these values can be found in Supplementary Tables 1, 3 and 4.

Retroviral overexpression of FLAG-tagged ZAK proteins in HEK-293T cells

Full-length cDNA encoding human ZAK-β (BC001401) in pOTB7 was purchased from Open BioSystems and subcloned into pFLAG-CMV-6c (Sigma-Aldrich). ZAK-C22A and ZAK-K45M mutants were generated by QuikChange site-directed mutagenesis using the primer 5′-atttgatgacttgcagttttttgaaaacgccggtggaggaagttttg - 3′ and 5′-ggacaaggaggtggctgtaatgaagctcctcaaaatagag - 3′ and their complements. Wild-type and mutant ZAK were cloned into a modified pCLNCX retroviral vector. Retrovirus was prepared by taking 3.0 μg of each of pCLNCX and pCL-Ampho vectors and 18 μL of FuGENE HD reagent (Roche) to transfect 60% confluent HEK-293T cells. Medium was replaced after 1 day of transfection and on the next day virus-containing supernatant was collected, filter sterilized and stored at −70 °C. 1 mL of virus-containing medium was used to infect target cells in presence of 8 μg/mL of polybrene for 72 hours and infected cells were selected in medium containing 100 μg/mL of hygromycin. Surviving cells after the selection were expanded and cultured in regular DMEM medium with 10% FCS.

Immunoprecipitation of FLAG-tagged ZAK proteins

HEK-293T cells with stable expression of wild-type or mutant ZAK were grown to 100% confluency on a 10 cm plate. Cells were collected, washed with cold PBS (2 x 10 mL) and lysed in 1 mL of PBS supplemented with 1x cOmplete EDTA-free protease inhibitor cocktails by sonication. Cell lysates were fractionated by centrifugation (100,000 x g, 45 min.) and the soluble fraction was incubated with 50μL of Anti-FLAG M2 affinity gel (Sigma-Aldrich) at 4 °C for 3 hours. Beads were washed with 5 x 1 mL of cold PBS (10 min per incubation) and FLAG-ZAK were eluted by either 150 μg/mL of FLAG-peptide solution provided by manufacturer or by 4x gel loading buffer depending on the downstream applications.

In-gel fluorescence characterization of ZAK labelled by IA probe

FLAG-tagged wild-type and C22A mutant ZAK were immunoprecipitated from HEK-293T cells (1 x 10⁷). After washing with PBS, the beads were suspended in 100 μL of PBS buffer and labelled with 250 nM of IA-rhodamine (add 1 μL of 25 μM probe stock in DMSO). After 1 hour of labelling at 4 °C, 50 μL of 4x gel loading buffer was added and the beads were boiled for 5 min to elute the bound proteins. Gel samples were separated by SDS-PAGE (50 μL of sample/lane) and visualized in-gel using a Hitachi FMBio II flatbed laser-induced fluorescence scanner (MiraiBio, Alameda, CA). For testing HNE blockade on IA labelling of ZAK by gel, soluble lysate of HEK-293T overexpressing WT-ZAK were incubated with 10, 50 and 100 μM of HNE (add 2 μL of 5, 25 and 50 mM stock) for 30 min and then subjected to immunoprecipitation.

In-gel fluorescence characterization of ZAK labelled by HNEyne probe

50 μL of soluble lysate (1 mg/mL in PBS) of HEK-239T cells transfected with mock, ZAK-WT and ZAK-C22A was labelled with 10 μM of HNEyne¹² (Cayman Chemicals, 1 μL of 500 μM stock in ethanol) for 1 hour at room temperature. Cycloaddition was performed with 200 μM rhodamine-azide, 1 mM TCEP, 100 μM TBTA ligand and 1 mM CuSO₄. The reaction was allowed to proceed at room temperature for 1 hour before quenching with 20 μL of 4x SDS-PAGE loading buffer (reducing). Quenched reactions were separated by SDS-PAGE (40 μL of sample/lane) and visualized in-gel using a fluorescence scanner. For the in situ HNEyne labelling, WT- and C22A-ZAK transfected cells were grown in a 6-well plate to 100% confluency and switched into 1 mL of serum-free DMEM medium. Cells were labelled with 5 μM of HNEyne probe (1 μL of 5 mM stock) for 1 hour at 37°C. Cells were then harvested, washed with cold PBS and lysed in 200 μL of PBS with protease inhibitors. 50 μL of soluble lysates were subjected to the cycloaddition protocol as described above and probe labelling was monitored by in-gel fluorescence.

ZAK in vitro kinase activity assay

The kinase activity assay protocol was adapted from Yu et al²⁹. Kinase assay buffers, Myeilin Basic Protein (MBP) substrate, and ATP stock solution were purchased from SignalChem. Radio-labelled [^33P]-ATP was purchased from PerkinElmer. 10 mg of soluble lysate of HEK-293T cells transfected with each of wild-type, C22A and K45M of ZAK were immunoprecipitated and then eluted with 2 x 300 μL FLAG-peptide buffer. Each sample was concentrated to 100 μL using an Amicon centrifugal filter (30 kDa cutoff) and exchanged to the assay kinase buffer (5 mM MOPS, pH 7.2, 2.5 mM β-glycerol-phosphate, 5 mM MgCl₂, 1 mM EGTA, 0.4 mM EDTA, 0.05 mM DTT and 40 ng/μL BSA) to a final volume around 10 μL. For each ZAK construct, 4 reactions were set up and each reaction starts with mixing 10 μL of immunoprecipiated ZAK, 5 μL of MBP (1 mg/mL) and 5 μL of HNE (10 or 100 μM) or H₂O together. No-enzyme and no-substrate controls were prepared in parallel. The mixed samples were incubated on ice for 15 min and 5 μL of [^33P]-ATP assay cocktail (250 μM, 167 μCi/mL) was then added to initiate the kinase reaction. Each reaction mixture was incubated in 30°C for 15 min and the reaction was terminated by spotting 20 μL of the reaction mixture onto individual pre-cut strips of phosphocellulose P81 paper. The spotted P81 strips were air dried and then washed with 10 mL of 1% phosphoric acid for 3 x 10 min. ZAK activity was measured by counting the radioactivity on the P81 paper in the presence of scintillation fluid in a scintillation counter after subtracting the value obtained from the corresponding no-substrate control, and was normalized to that of ZAK-WT without HNE treatment. Experiments were performed in triplicates. 10 μL of each ZAK variant used in setting up the kinase reaction were run on a SDS-PAGE gel and immunoblotted with an anti-FLAG antibody to ensure that they are enriched at similar levels (Supplementary Fig. 4).

Measurement of HNE blockade of ATP binding of ZAK by SILAC-ABPP

HEK-293T cells with stable expression of wild-type ZAK were passaged six times in DMEM medium minus l-lysine and l-arginine (Thermo) supplemented 10% dialyzed FBS (Gemini), 1% PSQ (1% vol/vol 10,000 units penicillin, 10 mg streptomycin, 29.2 mg l-glutamate solution) and 100 μg/mL [¹³C₆, ¹⁵N₄] l-arginine–HCl and [¹³C₆,¹⁵N₂] l-lysine–HCl (heavy) or l-arginine–HCl and l-lysine–HCl (light) (Sigma-Aldrich). Soluble proteomes of light and heavy ZAK-WT transfected HEK-293T cells (3 mL each at 9 mg/mL) were treated with 100 μM of HNE (6 μL of 50 mM stock) or DMSO for 30 min at room temperature, respectively. The proteomes were gel filtrated by PD-10 columns (GE Healthcare) to remove unreacted HNE and split into four aliquots (0.5 mL, 6 mg/mL). Each of the four aliquots of light and heavy proteomes was labelled with 20 μM of acyl-phosphate-ATP probe (ActivX Biosciences) for 10 min and then heavy and light proteomes mixed together to proceed with reduction/alkylation, streptavidin enrichment, trypsin digest according to a modified version of the vendor-provided “Xsite Kinase Analysis” protocol⁴³. The four trypsin-digested samples were analysed by LC-MS/MS and enriched peptides identified by SEQUEST and DTASelect. The amounts of probe-labeled ZAK and other proteins in the proteome with and without HNE treatment were quantified using the CIMAGE module that was developed for quantitative SILAC-ABPP⁴⁴. Quantitation data from the four MS runs were combined and peptides from keratin proteins and decoy proteins were removed. Cases were excluded from further analysis when proteins were not quantified in at least two out of the four runs and with at least three unique tryptic peptides, or quantified with large variability across multiple runs (s.d. > 2/3 of the mean). R ratios (DMSO-treated heavy versus HNE-treated light proteomes) of protein targets that passed the filter were plotted in Fig. 4e. As an internal control, untreated and unenriched proteomes were digested with trypsin and analysed by LC-MS/MS to quantify the relative levels of proteins in heavy and light amino acid-labeled cells. This control confirmed ZAK showed an R ratio of 0.62 (heavy/light), which was similar to the median ratio of 0.72 (heavy/light) quantified for all proteins. Quantified ratios from these SILAC-ABPP experiments are reported in Supplementary Table 6.

Western blotting of phospho-MAPKs in ZAK-transfected HEK-293T cells

Mouse and rabbit monoclonal antibodies against phospho-ERK 1/2 (Thr202/Tyr204), phosphor-SAPK/JNK (Thr183/Thr185), phosphor-p38 MAPK (Thr180/Thr182) and total ERK 1/2, SAPK/JNK and p38 MAPK were purchased from Cell Signaling Technology (catalogue #9106, 9255, 9216, 9212, 9252,4695). HEK-293T cells transfected with mock, WT-ZAK, C22A-ZAK and K45M-ZAK were seeded into a 12-well plate with 2.5x10⁵ cells per well. Cells were grown in regular DMEM medium with 10% FBS for 24 hours and starved in serum-free DMEM medium for another 24 hours. Cells were then treated at 37°C either with 100μM of HNE (2 μL of 50 mM stock) for 30 min followed by 1 mM of H₂O₂ for 30 min, or with 50 and 100 μM of HNE alone for 60 min. After the treatment, cells were harvested, washed with 2 x 1 mL of cold PBS, and then lysed by sonication in 100 μL of PBS buffer supplemented with 1x cOmplete protease inhibitors cocktail and 1 x PhosSTOP phosphatase inhibitors cocktail (Roche). 30 μg of soluble lysate of each sample was separated by SDS-PAGE, transferred to nitrocellulose membrane, blocked in 5% milk TBST and blotted against the primary antibodies (1:2000) listed above for 16 hours at 4°C. After washing in TBST (3 x 10 min), membranes were blotted with IRDye secondary antibodies (1:10,000) for 1 hour at room temperature and scanned by an Odyssey imaging system (LI-COR). Protein band intensities were quantified by ImageJ⁴⁵ and ratios of phosphor-MAPK over total MAPK were computed. Experiments were repeated in at least four replicates.