Proteomic analysis of oxidative modifications in human atherosclerotic plaques and control arteries

Abstract

Cardiovascular disease is a major cause of worldwide mortality. A prime underlying cause is atherosclerotic plaque formation at specific arterial locations. Considerable evidence links plaques with chronic inflammation and leukocyte activation, with this potentially resulting in enhanced oxidant-mediated damage to arterial proteins. In this study, human carotid artery plaques and control tissue (superior thyroid artery) were analyzed by liquid chromatography-mass spectrometry (LC-MS/MS) for the presence of post-translational modifications (PTMs) generated by redox reactions on proteins. Of 7150 proteins detected across all samples, 2724 carried at least one PTM, with these including mono- and di-oxidation at methionine, tryptophan and histidine, conversion of cysteine to cysteic acid, chlorination and nitration of tyrosine and tryptophan, as well as carbamylation of lysine and N-terminal amines. Higher PTM levels were detected on proteins from plaques compared to control artery tissue, but similar modification types were detected. Elevated concentrations of PTMs were detected in plaques classified as unstable compared to stable, consistent with PTMs being associated with plaque instability. The levels of several PTMs correlate with the concentration of the neutrophil-derived heme protein myeloperoxidase, implicating this enzyme and its oxidants in protein damage, though direct causality has not been established except for chlorination. Large numbers of PTMs were detected on key metabolic proteins including those involved in glycolysis/gluconeogenesis, the pentose phosphate pathway, the tricarboxylic acid (TCA) cycle and oxidative phosphorylation. Particularly high levels were detected on glycolytic enzymes, including glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and Complex V subunits consistent with perturbations to cellular metabolism in atherosclerotic plaques.

Graphical abstract

Highlights

• •Unbiased in-depth proteomics of human atherosclerotic plaques for oxidative post-translational modifications (PTMs).
• •7150 proteins detected across all samples, with 2724 proteins carrying at least one PTM.
• •PTMs were more abundant on plaque than control artery proteins and associated with plaque instability.
• •PTMs correlate with the abundance of the neutrophil-derived enzyme myeloperoxidase.
• •Mapping to specific proteins shows marked damage to glycolytic enzymes including GAPDH.

1. Introduction

Cardiovascular disease (CVD) is a major cause of death worldwide, and results in significant healthcare, social and economic costs. The underlying cause is atherosclerosis, which is characterized by dysfunction of the endothelial cells lining the artery wall, the accumulation of fats and cholesterol at specific locations within the arterial tree, and formation of atherosclerotic plaques. Development of the plaques results in both an inward and outward remodeling (thickening) of the artery wall, and an increase in arterial stiffness [1,2]. Plaque enlargement during disease progression results in increasing luminal narrowing and possible occlusion, impaired blood flow and downstream ischemia. In addition to progressive narrowing and impaired perfusion of the end-organ, some plaques may undergo erosion, rupture or thrombus formation, causing acute vessel occlusion or peripheral embolization. These plaques are considered unstable or symptomatic due to the acute onset of clinical symptoms such as ischemic stroke (carotid and cerebrovascular arteries), myocardial infarction (coronary arteries), and acute lower limb ischemia (femoral and iliac arteries). Increasing age, high neutrophil numbers, and multiple genetic, environmental and lifestyle factors, contribute to the disease, with these including elevated levels of total cholesterol (particularly in low-density lipoprotein (LDL) and very-low-density lipoproteins), biological sex, hypertension, obesity, diabetes, smoking and low levels of physical activity (reviewed [[1], [2], [3]]).

Multiple biochemical processes can contribute to disease initiation and development. These include changes in the metabolic program and dysfunction of endothelial cells, decreased levels of nitric oxide, proliferation and migration of phenotypically-changed smooth muscle cells, accumulation of lipids within macrophage and other cells within the artery wall, apoptosis and necrosis of lipid-laden cells, damage to the extracellular matrix (ECM) that provides structural strength to the fibrous cap covering the lipid core, elevated uptake and decreased removal of lipids from the artery wall, enhanced numbers and activity of inflammatory cells, and an elevated level of oxidative stress [[2], [3], [4], [5], [6]]. Therapeutic reduction of lipid levels (e.g. by statins) and antithrombotics can improve outcomes, however considerable residual risk remains, resulting in an increased focus on other factors, and particularly modulation of inflammation [2,5,6]. The CANTOS (targeting interleukin-1β [7,8]) and LoDoCo/COLCOT (oral colchicine to target inflammation [9]) trials have established that chronic low-grade inflammation plays a key role, and that modulation of this can be of therapeutic benefit [8].

Considerable epidemiological evidence also supports a major role for high neutrophil numbers, their activation and the neutrophil-lymphocyte ratio in both CVD and all-cause mortality (e.g. Refs. [3,10,11]). Whilst the precise mechanisms through which inflammation drives atherosclerosis are unresolved, considerable evidence supports the hypothesis that neutrophil-mediated oxidative stress is a significant factor. Neutrophil activation results in the assembly of an NADPH oxidase complex (NOX2) at the plasma membrane, and release of the highly abundant heme enzyme myeloperoxidase (MPO) into the phagolysosome as well as extracellularly (e.g. via formation of neutrophil extracellular traps [12,13]). Both NOX2 and MPO generate high levels of oxidants (superoxide radicals, O₂^.-, and subsequently H₂O₂ from NOX2; hypochlorous acid (HOCl) and related species from MPO). High concentrations of MPO are strongly associated with arterial wall and low-density lipoprotein (LDL) modification, tissue and ECM damage and poor patient prognosis [[14], [15], [16], [17], [18], [19]].

Multiple studies have provided evidence for the presence of oxidized materials (post-translational modifications, PTMs) within human atherosclerotic plaques (reviewed [[20], [21], [22], [23]]). Increasing evidence indicates that protein modification is a significant factor due to the high abundance of proteins (∼60-70% by dry mass [24]) and their key role in arterial wall, plaque, tissue and cell structure. Previous studies have reported the presence of protein PTMs in plaques, but much of this data has arisen from studies using antibodies specific to particular modifications (e.g. Refs. [18,25]) thereby providing biased data with regards to the species examined, or have only examined specific proteins in, or extracted from plaques (e.g. LDL or high-density lipoproteins [[26], [27], [28]]). Thus, there is an incomplete picture, and limited unbiased data on the types, levels and sites of PTMs in human plaques and particularly on cellular proteins. We hypothesized that use of liquid chromatography coupled to mass spectrometry (LC-MS/MS), with in-depth sample fractionation, would allow these knowledge gaps to be addressed.

Here, we provide a detailed proteomic analysis of PTMs present on cellular proteins in advanced, symptomatic, human carotid artery plaques and non-atherosclerotic control artery (superior thyroid) samples. The data are consistent with the presence of many different types of modifications, with these being present on a wide range of proteins in both diseased and control tissue. Higher levels are detected in plaques compared to control tissue, and in vulnerable (rupture-prone) compared to stable plaques, and the levels of some of these species correlate with MPO levels. Analysis of cellular proteins indicates significant levels of PTMs on key enzymes of the glycolytic and pentose phosphate pathways, and more limited numbers on mitochondrial electron transport chain proteins.

2. Materials and methods

2.1. Human artery samples

Carotid plaques were collected from 76 consecutive patients recently diagnosed with cerebral embolic ischemia (stroke, transient ischaemic attack) or retinal ischemia (complete vision loss, amaurosis fugax) and processed as previously described [29,30]. Plaques were removed in toto during carotid endarterectomy. For 8 of the patients included in the study, the superior thyroid artery was also surgically dissected, ligated, and removed as part of the operational procedure, with this material then used as non-atherosclerotic control tissue as described previously [30]. The number of control samples obtained within the study period was limited, as collection of the superior thyroid artery was only possible when clinically indicated and did not substantially increase duration of surgery, since its removal is not otherwise justifiable. Plaques and artery sections (76 + 8, 84 total) were subsequently washed and stored at −80 °C in a deoxygenated preservative buffer made up using Chelex-treated MilliQ water (PBS, with 20 mM EDTA, 100 μM 2,6-di-tert-butyl-p-cresol, 100 mM methionine and 1 mM sodium azide) to minimize post-collection oxidation and modification.

2.2. Ethics

This study was conducted according to the Helsinki declaration, under the approval of The Danish National Committee on Health Research Ethics (journal number H-20002776). Patient data and biological materials (atherosclerotic plaque and thyroid superior artery sections) were collected with written, verbal and informed consent from the patients.

2.3. Sample preparation

Protein extraction, clean-up, and digestion was carried out as described previously [29,30] using an automated Opentrons OT2 liquid handling platform. Briefly, the protein component of atherosclerotic plaques and artery tissue was extracted and solubilized using 80% trifluoroacetic acid using the SPEED protocol [31], then neutralized with Tris base. Proteins were reduced, alkylated, and captured on magnetic beads in the presence of acetonitrile (ACN), followed by washing with acetonitrile/ethanol and on-bead digestion with Lys-C and trypsin, separated from the magnetic bead, acidified with trifluoroacetic acid (TFA, 0.5% final concentration) and stored at −80 °C until further analysis.

2.4. LC-MS analysis

Peptide samples were separated using a Dionex Ultimate RSLCnano system (Thermo Scientific) equipped with an Aurora C18 column (15 cm × 75 μm, 1.6 μm particle size, IonOpticks), maintained at 45 °C and coupled to an upgraded timsTOF Pro (Bruker) mass spectrometer. Peptides were eluted using a 28-min binary gradient (mobile phase A: 0.1% v/v formic acid (FA) in water; mobile phase B: 0.1% v/v FA in ACN) with data acquisition in DIA-PASEF mode [32]. The workflow is summarized in Fig. 1.

Fig. 1.

Overview of experimental workflow for LC-MS/MS analysis of control tissue and human carotid artery atherosclerotic plaques. Human carotid artery atherosclerotic plaques (n = 76) and control (superior thyroid artery tissue, n = 8) were obtained from carotid artery endarterectomy operations. Proteins were extracted using the SPEED protocol [31], then reduced, alkylated, captured on magnetic beads and subjected to on-bead digestion with Lys-C and trypsin. For a subset of the plaque samples (n = 6) the peptides were subjected to high pH reverse-phase liquid chromatography fractionation to give 36 fractions for each plaque. These fractions were then subject to LC-MS/MS analysis using data-dependent acquisition (DDA) with parallel accumulation-serial fragmentation (PASEF) to generate an in-depth spectral library from the 216 (36 x 6) DDA runs. These samples, plus peptide solutions from 70 additional plaque samples and 8 control tissue samples, were subject to data-independent acquisition (DIA)-PASEF, with the resulting MS data searched using a spectral library generated from the DDA data in combination with pseudo-DDA spectra from diaTracer [33] converted DIA data. Precursors were summarized to peptidoforms (peptides, either native or carrying ≥1 modification), PTMs (a specific modification on a specific site) and protein levels. Modified peptidoforms (and thereby PTMs) were normalized to total parent protein intensity prior to further data analysis.

2.5. Off-line fractionation and spectral library generation

Six of the plaques were subjected to an in-depth fractionation protocol and used for spectral library generation. For these plaques, an additional delipidation step was performed with methyl tert-butyl ether after protein digestion followed by acidification and sample desalting, prior to storage at −80 °C until further processing. These samples were each separated into 36 fractions by high-pH reverse phase chromatography with a Kinetex EVO C18 column using a 50 min binary gradient (Buffer A: 20 mM ammonium formate, pH 10.5 in H₂O; Buffer B: 20 mM ammonium formate, pH 10.5 in 80% v/v CH₃CN). The gradient was ramped from an initial composition of 95% Buffer A: 5% Buffer B, to 38% B in 35 min, then to 60% B in 7 min and finally 95% B in 2 min and kept at 95% B for 6 min, The flow rate was maintained 40 μL min⁻¹. Eluates were collected every minute from 0 to 48 min, with eluates from the 0-12 min period pooled, resulting in a total of 36 fractions per sample. Fractionated peptides were then dried down by vacuum centrifugation before being resuspended in 0.1% TFA and separated using a Dionex Ultimate RSLCnano system (Thermo Scientific) equipped with an Aurora C18 column coupled to a timsTOF Pro (Bruker) mass spectrometer with data acquisition in DDA-PASEF mode. All 216 DDA runs (6 samples x 36 fractions) and all DIA-PASEF runs were searched with MSFragger (v. 4.2) implemented in the FragPipe software suite (v. 23.0) [34]. For DIA-PASEF files, the files were initially converted to DDA-like files using the diaTracer module [33] but were otherwise handled as DDA-PASEF files. The search included the following variable modifications: oxidation of Met, Pro and Trp (+15.9949 Da), dioxidation of Met, His and Trp (+31.9898 Da), trioxidation of Cys (+47.9847 Da), carbamidomethylation of Cys (+57.0215 Da), carbamylation of N-terminal amines and Lys residues (+43.0058 Da), chlorination of Tyr and Trp (+33.961 Da), nitration of Tyr and Trp (+44.985 Da) and oxidation of His to 2-oxoHis (+13.9793 Da). Peptide-spectrum matches were re-scored and validated using MSBooster (v. 1.3.9 [35]) and Percolator (v. 3.7.1 [36]) and filtered to obtain a false-discovery rate of 1%. The easyPQP python package (v. 0.1.52) was used to generate a spectral library from the search results.

2.6. Data analysis and statistics

DIA-PASEF data were processed in DIA-NN [32] using a library constructed from the DDA-PASEF acquisitions from the off-line fractionated samples (see above). Further data analysis and visualization were performed in R. Precursors carrying modifications were only considered for downstream analysis with DIA-NN-determined post-translational modification (PTM) site confidence levels of > 90% to provide high confidence site assignments while retaining sufficient coverage for this discovery-scale PTM analysis. Precursor intensities were log2-transformed and normalized across samples by median-centering prior to summarization of precursor signal to peptidoform (peptides, either native or carrying one or more modifications) and protein level. For modified peptidoforms, intensities were normalized within each sample to the corresponding parent protein abundance (i.e. by subtracting log2 total protein intensity). PTM-level (site and modification specific) intensities were calculated from these normalized values. Each type of modification on each protein was further summed by exponentiating normalized log2 PTM intensities back to a linear scale, summing modified sites of each modification type (e.g. multiple Met monooxidations) on the given protein, then log2-transforming these summed intensities. All abundance (intensity) based analyses reported here were performed on normalized data. Sample clustering was performed using the Monte Carlo Reference-based Consensus Clustering (M3C) package [37] for its k-means algorithm and clustering stability based on the top 10% most variable proteins. The optimal number of clusters was determined by identifying the peak Relative Cluster Stability Index (RCSI), an empirical p-value below 0.05 indicating significance relative to null distributions generated through Monte Carlo simulations, and the lowest entropy value calculated by the algorithm. Using a modified MSqRobHurdle [38] workflow, statistical inference with linear modeling was carried out, with differential detection based on precursor count only considered when modeling of differential abundance based on intensity failed. p-values were adjusted separately for differential abundance and detection, with Benjamini-Hochberg [39] adjusted p-values < 0.05 considered significant. Cellular proteins were defined as proteins neither in the matrisome (as defined in Ref. [40]) nor plasma-derived proteins (defined as ‘secretome’ or ‘erythroid cell’ species in the human Protein Atlas: https://www.proteinatlas.org/).

2.7. Data availability

In accordance with Danish legislation relating to the General Data Protection Regulation (GDPR), raw mass spectrometry data can only be shared under a data processing agreement between the data controller and processor. Such data sharing agreements are available and can be completed by contacting the corresponding author.

3. Results

Fig. 1 presents an overview of the workflow used to detect PTMs in human carotid plaques and control tissue. The parent proteomes of the 76 atherosclerotic plaques (from 76 patients) and 8 non-atherosclerotic sections from superior thyroid arteries have been reported previously [29,30], but these samples have not been examined previously for PTMs. Analysis of these samples resulted in the identification of 7150 proteins using a conservative 1% false-discovery rate (FDR) cut-off. The peptides detected were then searched for PTMs including oxidation/hydroxylation at multiple residues, dioxidation/dihydroxylation at multiple residues, trioxidation at Cys, chlorination and nitration at Tyr and Trp residues, and carbamylation at Lys and the N-terminus (see Materials and Methods).

Of the 7150 proteins detected across all the artery samples, 2724 were detected carrying at least one PTM. After filtering out precursors carrying PTMs with less than 90% site localization confidence, 3608 PTMs on 1154 proteins were quantified. These data were then subjected to cluster analysis using unsupervised consensus clustering (Monte Carlo Reference-based Consensus Clustering package [37]), with this resulting in an optimized, and statistically significant, output of three clusters (Fig. 2A). Higher or lower numbers of clusters were not statistically significant when compared to the null hypothesis. There was clear segregation of the non-atherosclerotic control tissue samples (cluster type C) from two clusters derived from the plaques (cluster types A and B) when the clusters were assessed and visualized by multidimensional scaling (MDS) (Fig. 2B). These clusters correlate strongly with the morphological phenotypes of the plaques determined independently from the proteomic analyses (Fig. 2C). Thus, the type A cluster consisted primarily of soft and intermediate/mixed plaques with ulceration and bleeding, whereas the type B cluster was mainly derived from plaques categorized as hard and intermediate/mixed plaques without ulceration or bleeding. Cluster type A contained a higher number of plaques derived from males, whereas cluster B had equal numbers of plaques from female and male subjects.

Fig. 2.

Sample clusters based on proteomic profiles correspond well with clinical parameters for control tissue and plaques. (A) Relative cluster stability index (RCSI) values for clusters number (K) 2-8 obtained by Monte Carlo reference-based consensus clustering. Cluster numbers with a significant Monte-Carlo p-value of < 0.05 are represented by red dots, while non-significant cluster numbers are colored black. (B) Multidimensional scaling (MDS) plot of distances between protein abundance profiles for the control artery and plaque proteomes. Each sample is plotted on a two-dimensional scatterplot of the first two components of a MDS analysis of the top 200 most abundant proteins, with each shape representing an individual control artery (square) or plaque (triangle). Shapes are colored based on their cluster, with open/closed symbols indicating the biological sex of the donor (open = female, closed = male). (C) Tile plot overview of all samples, including consensus cluster, tissue type, clinical parameters (plaque consistency, ulceration, and bleeding), and the biological sex of the donor (cf. Data in Refs. [29,41]).

The detection of modifications was assessed based on distinct peptidoforms and proteins. For peptidoforms, the percentage was calculated as the number of distinct modified peptidoforms divided by the total number of distinct peptidoforms (native or modified). Proteins were classified as modified if at least one of its detected peptidoforms carried a PTM, and the percentage was calculated as the number of proteins with at least one modified peptidoform divided by the total number of detected proteins. Type A (soft and intermediate plaques) contained the highest percentage of unique modified peptidoform count (4.7%) and proteins (13.7%) followed by type B (hard and intermediate) plaques, and finally the superior thyroid artery tissue (cluster type C; Fig. 3A and B).

Fig. 3.

Modifications vary between the control artery tissue and plaque types significantly, but the pattern of PTMs remains relatively consistent. Percentage values were calculated from counts of distinct native (gray bars) and modified (red bars) (A) peptidoforms and (B) proteins quantified in the dataset. Peptidoforms were classified as modified if they carried at least one PTM, while proteins were classified as modified if at least one of its peptidoforms carried a PTM. Statistical significance was calculated by one-way ANOVA with post hoc Tukey HSD test and statistical significance was accepted for adjusted p-value < 0.05. (C) Abundance expressed by summed parent-protein-normalized PTM intensities and (D) Relative PTM abundance, expressed as the fraction of total PTM intensity represented by each modification type within each sample. Data is provided for each type of modification in each sample of control artery (type C) and plaques of types A and B.

We then assessed the overall extent and types of modifications (for structures and identities see Suppl. Fig. S1) across the samples and the clusters (i.e. both tissue and plaque type). Although the PTM abundance (summed PTM intensity for each modification type) and detection (peptidoform count) varied between the samples (Fig. 3C, Suppl. Fig. S2A) as expected from donor variability, the pattern of PTMs (relative abundance or detection) within each cluster showed a degree of consistency across tissue and plaque type (Fig. 3D, Suppl. Fig. S2B). Several PTMs exhibited differential patterns between the control and plaque groups, and between the two plaque clusters (A vs. B), consistent with the number, abundance and types of PTMs being dependent not only on the presence of disease, but also the nature of the plaque (Fig. 4, Suppl. Fig. S3). Amongst the 12 types of PTMs searched for and detected on these peptides, 9 (75%) showed statistically increased abundance in either or both of the plaque types when compared to the control samples (Fig. 4). Significantly higher levels of cysteic acid (RSO₃H from Cys), Met sulfoxide (i.e. monooxidation of Met), dioxidation of Trp, and chlorination of Tyr were detected in both plaque types compared to control tissue (Fig. 4). Significantly higher abundance of 2-oxoHis, carbamylation of Lys residues, and Met sulfone (i.e. dioxidation of Met [42]) were detected in type A plaques compared to control tissue (Fig. 4). Type B plaques contained higher levels of nitrated Trp and Tyr compared to control tissue (Fig. 4). Control arteries contained higher levels of N-terminal carbamylation and chlorination of Trp compared to both plaque types and type B plaques, respectively (Fig. 4). Type A plaques contained higher abundance of 2-oxoHis; carbamylation of Lys; and mono- and dioxidation of Met (i.e. Met sulfoxide and sulfone) compared to type B plaques, which in turn contained higher levels of nitration of Trp compared to type A plaques (Fig. 4).

Fig. 4.

PTM levels vary across tissue and plaque type. Plots of PTM abundances (summed parent-protein normalized intensities of each modification type, log2) for control artery (type C, green color) samples and plaques of type A (red color) and type B (blue color). Statistical significance between the clusters was assessed using linear modeling, followed by pairwise comparisons using estimated marginal means (emmeans). Statistical significance was accepted at a Tukey adjusted p-value < 0.05 (* indicates p < 0.05, **p < 0.01, and ***p < 0.001).

Differences in specific PTM and protein abundance (based on intensity) and detection (based on precursor count) were examined across the clusters. 853 PTMs exhibited significant differential abundance between control and plaque tissue, with 702 PTMs more abundant in plaques and 151 PTMs being more abundant in the control artery samples. In addition, 96 PTMs were found to be differentially detected (i.e. present solely, or predominantly, in one group and not the others), with 75 PTMs detected more frequently in the plaques and 21 PTMs more frequently in controls.

Of the PTMs identified, 733 differed significantly in abundance between the two plaque types, with 631 PTMs found to be more abundant in type A (soft/unstable) plaques and 102 more abundant in type B (hard/stable) plaques. A higher absolute number of PTMs were detected on cellular proteins when compared to ECM species (as defined in Ref. [40]), or plasma-derived proteins (as defined in the human Protein Atlas: https://www.proteinatlas.org/), though ECM species were relatively more modified when taking the more modest numbers of ECM species into account.

Of the PTMs detected on cellular proteins 507 were found to be differentially abundant between the control and plaque samples (459 more abundant in plaques; 48 more abundant in controls; Fig. 5A), with a further 27 cellular PTMs being differentially detected (11 in plaques, 16 in controls). 346 cellular protein PTMs were differentially abundant between plaque types (305 more abundant in type A; 41 more abundant in type B: Fig. 5A).

Fig. 5.

PTM numbers and modified protein abundance vary between tissue plaque types. Volcano plots providing a comparison of cellular (A) PTM and (B) modified protein abundance between plaque and control tissue, and between plaque types A and B. (C) Volcano plots showing the comparison of cellular PTM abundances for each type of modification between plaque and control artery tissue (plaque vs control, top of the panel) and plaque types (type A vs type B, bottom of panel). Differential abundance analyses were performed on parent-protein-normalized intensity values. Each dot represents a cellular PTM with a log2 fold change between the samples compared (x axis) plotted against the FDR-adjusted p-value (reversed log 10; y axis), with red dots indicating PTMs on proteins of metabolic pathways. Light gray dots below the horizontal dashed line indicate non-significant changes, while dots above this line indicate significant changes (FDR-adjusted p-value < 0.05).

The majority of proteins detected as carrying PTMs carried only a single, or low numbers of modifications of the same type. Examining the differential abundance of modified cellular proteins therefore yielded similar results as for the PTM-carrying peptidoforms (Fig. 5B). The most abundant PTMs were mono- and di-oxidation of Met residues, with these being detected at higher levels in plaques compared to control arteries, and in type A compared to type B plaques (Fig. 5C).

In order to examine whether the presence of PTMs is associated with the abundance (concentration) of the parent protein, plots were generated of protein abundance (log2 intensity-based absolute quantification (iBAQ) values) versus rank order with the presence of the various PTMs annotated (Fig. 6). For the mono-oxidations (Fig. 6A) there was a clear association between high abundance and PTM detection, with this PTM being less frequently detected on low concentration proteins. In contrast, significant levels of cysteic acid formation (the sulfonic acid form of Cys), nitration, chlorination and in particular carbamylation were detected on both high and low abundance species (Fig. 6D-G) consistent with some PTMs being generated with a degree of selectivity, and at specific sites, on low abundance proteins.

Fig. 6.

Both high and low abundant proteins are modified. Proteins are plotted by intensity-based absolute quantification (iBAQ) values (log2) against rank order. Native proteins are indicated with gray dots, while red dots indicate proteins containing (A) single oxidations (m/z +16 species) of Met or Trp, (B) dioxidations (m/z +32 species) of Met or Trp, (C) 2-oxoHis residues (oxidation of His, m/z +14 species), (D) cysteic acid residues (tri-oxygenated, sulfonic acid, m/z +48 species), (E) nitrations (m/z +45 species) of Tyr or Trp, (F) chlorinations (m/z +34 species) of Tyr or Trp, or (G) carbamylations (m/z +43 species) of Lys or N-terminal amines.

Considerable numbers of differentially abundant PTMs were detected on cellular metabolic proteins. Consequently, PTMs present on proteins involved in glycolysis/gluconeogenesis, the tricarboxylic acid (TCA) cycle, the pentose phosphate pathway (PPP), and mitochondrial oxidative phosphorylation were examined in depth (Fig. 7A–C). Significantly more PTM sites were detected on enzymes involved in glycolysis/gluconeogenesis, the TCA cycle and the PPP in both plaque types when compared to control artery samples (Fig. 7A). The abundance of these PTMs was, in general, also significantly higher in plaques compared to control (except for PTMs on TCA cycle proteins in type B plaques); and in Type A compared to type B plaques, with the exception of PTMs on oxidative phosphorylation proteins (Fig. 7B). The greatest number of PTMs, and particularly Met oxidations, were observed on proteins involved in glycolysis for both control and plaque tissue (Fig. 7C). Met oxidation (both mono- and di-) was prevalent across all the metabolic pathways examined; multiple other PTMs were also observed on proteins involved in glycolysis albeit at different levels (Fig. 7C and D). Comparison of the abundance (by intensity) and detection (by precursor count) of the modified proteins indicated that the differences were often Met oxidations, with these detected on multiple proteins (Fig. 7, Fig. 8, Fig. 9). This was true for all four pathways, and particularly glycolysis/gluconeogenesis.

Fig. 7.

Multiple PTMs are present on proteins involved in glycolysis/gluconeogenesis, the tricarboxylic acid (TCA) cycle, the pentose phosphate pathway (PPP) and mitochondrial oxidative phosphorylation. (A) Total detection (count) and (B) abundance (parent-protein-normalized intensity, log2) of PTMs summed across all proteins involved in the different metabolic pathways for control artery (type C, green color) and plaques of type A (red color) and type B (blue color). Statistical significance between the clusters was assessed using negative binomial generalized linear modeling (for (A) count data) or linear modeling (for (B) abundance data), followed by pairwise comparisons using estimated marginal means (emmeans). Statistical significance was accepted at a Tukey adjusted p-value < 0.05. * Indicates p < 0.05, **p < 0.01, and ***p < 0.001. (C) Tile plot showing number of PTM sites, grouped by modification type, across the four pathways for control tissue as well as type A and type B plaques. (D) Mapping of modifications to individual proteins for the glycolytic/gluconeogenesis pathways for the plaques (type A and B combined), with the number of different types of PTMs detected on each protein coded to the box width and coloration.

Fig. 8.

Metabolic proteins are differentially modified in tissue types. Tile plots showing differential abundance (fold change) and detection (odds ratio) between plaque and control tissue for modified proteins of: (A) glycolysis/gluconeogenesis, (B) TCA cycle, (C) pentose phosphate pathway, and (D) oxidative phosphorylation. Each tile represents a modified protein with a log2 fold change (purple to red scale) or natural log odds ratio (green to orange scale) between the tissues (more or less abundant/detected in plaque compared to control tissue) indicated. Light gray tiles indicate non-detected modified proteins. Differential abundance analysis was performed on parent-protein-normalized values. Significance was accepted at an FDR-adjusted p-value of < 0.05. * Indicates p < 0.05, **p < 0.01, and ***p < 0.001.

Fig. 9.

Metabolic proteins are differentially modified in plaque types. Tile plots showing differential abundance (fold change) and detection (odds ratio) between type A and type B plaques for modified proteins of: (A) glycolysis/gluconeogenesis, (B) TCA cycle, (C) pentose phosphate pathway, and (D) oxidative phosphorylation. Each tile represents a modified protein with a log2 fold change (purple to red scale) or natural log odds ratio (green to orange scale) between the plaque types (more or less abundant/detected in type A compared to type B) indicated. Light gray tiles indicate non-detected modified proteins. Differential abundance analysis was performed on parent-protein-normalized values. Significance was accepted at an FDR-adjusted p-value of < 0.05. * Indicates p < 0.05, **p < 0.01, and ***p < 0.001.

The protein carrying the highest number of different PTM types was glyceraldehyde-3-phosphate dehydrogenase (GAPDH), with six different types observed (Met oxidation, Met and Trp dioxidation, cysteic acid formation, 2-oxoHis, Lys carbamylation; Fig. 8, Fig. 9). High fold-changes between plaque and control in the extent of oxidation were also detected on other proteins including citrate synthase (CS, more abundant in plaque; Fig. 8), fructose-1,6-bisphosphatase 1 (FBP1, observed more in plaque; Fig. 8) and lactate dehydrogenase A (LDHA, observed more in control; Fig. 8). The modifications on GAPDH were markedly more abundant in both plaque types (A and B) than control arteries (Fig. 8, Fig. 10A, top part) and were also over-represented in type A versus type B plaques (Fig. 9, Fig. 10A, lower part). These PTMs include the active site Cys152 residue of GAPDH with this detected as the hyperoxidized (irreversible) cysteic acid species, although the extent of this modification was not significantly different between the plaque and control samples, or between the plaque clusters. These modifications are mapped to the protein structure in Fig. 10B and color-coded by PTM type.

Fig. 10.

PTMs detected on GAPDH differ between control tissue and plaques, and between plaque types. (A) Differential abundance (log2 fold change) of each PTM site in the sequence of GAPDH for plaque versus control (top section) and type A and B plaques (bottom plot). Each lollipop represents a PTM detected along the GAPDH amino acid sequence (x-axis) with a log2 fold change (y-axis) between the tissues (top; more or less abundant/detected in plaque compared to control) or plaque types (bottom; more or less abundant/detected in type A compared to type B plaques) compared. Differential abundance analysis was performed on PTM intensities normalized to total GAPDH abundance. Significance was accepted at an FDR-adjusted p-value < 0.05 and denoted with closed/open lollipops. (B) Human GAPDH monomer crystal structure (PDB: 4WNC). The NAD⁺ molecule that binds in proximity to the catalytic site is shown in orange and modified amino acids are colored according to type of modification, with the catalytic Cys152 residue also labelled. Visualization was achieved with the PyMOL Molecular Graphics System (version 3.1.13 Schrödinger, LLC).

4. Discussion

In this study, > 3600 PTMs were identified on proteins present in human symptomatic carotid atherosclerotic plaques and non-atherosclerotic control artery samples. This is the first detailed, unbiased, and in-depth analysis of the extent and types of oxidative PTMs present in human artery samples. Multiple PTMs have been detected, together with their sites, on >2700 proteins. Particularly high levels are present on proteins from plaques compared to (non-overtly diseased) control arteries, and from plaques classified as soft/unstable compared to hard/stable.

These analyses complement previous studies on the native proteomes of different plaque types [29,30,41] and the presence of protein fragments (the ‘degradome’ [41]), arising from protease activity or oxidation [43]. In line with previous data [29,30,41], unbiased and unsupervised computational clustering of the PTM data resulted in three distinct clusters, corresponding to species derived from control arteries (type C) and two plaque types (Fig. 2). These clusters closely align with the parent protein data [30] and gross plaque morphology, with the latter assessed independently and in a blinded manner. The majority of type A plaques contained ulceration and bleeding and were classified as being soft or mixed/intermediate type (Fig. 2), features associated with an unstable morphology [[44], [45], [46]]. In contrast, the type B plaque cluster showed features consistent with a stable morphology, and of a hard or mixed/intermediate type, without intraplaque hemorrhage or ulceration (Fig. 2).

The number of PTM-containing peptidoforms (Fig. 3A) and proteins (Fig. 3B) was significantly higher in plaques compared to non-atherosclerotic tissue. Higher numbers were also detected from type A (soft/unstable) compared to type B plaques (hard/stable). Although the PTM levels (both by abundance (intensity) and detection (count)) were different between the tissue samples (Fig. 3C, Suppl. Fig. 2A), the pattern of PTMs was similar between the control and diseased samples, and between the different plaque clusters (Fig. 3D, Suppl. Fig. S2B). All of the identified PTMs are well-established species (Suppl. Fig. S1 [43]) and have been detected previously on isolated proteins [43]. Some of these have been detected previously on apolipoproteins B100 and A-I isolated from human plaques (e.g. Refs. [26,47]). These data confirm, and significantly extend, previous reports on the presence of chlorinated and nitrated species in human plaques determined by antibody approaches [18,25,[48], [49], [50]], though these previous studies did not provide data on the identity of the proteins involved (only data on co-localization), nor the sites of the modifications within the protein sequences. The significant number of PTMs detected on proteins from control arteries may appear surprising, but it should be noted that these originate from donors of advanced age undergoing carotid endarterectomy operations, and atherosclerosis has both local and systemic features. Thus, the control arteries are also likely to have been exposed to chronic low-grade inflammation, with this contributing to their PTM load. Patients with symptomatic carotid stenosis also frequently harbor atherosclerotic plaques in other vascular territories. Consequently, the thyroid artery samples cannot be regarded as a truly “healthy” vessel, as they are harvested in close proximity to advanced plaques in patients with generalized atherosclerotic disease. Future studies should assess samples from young healthy subjects, though such tissue is difficult to obtain.

The overall proteome of plaques (and plaque subtypes) differs from that of normal, healthy, non-atherosclerotic artery wall (e.g. Refs. [30,[51], [52], [53]]). A change in the absolute abundance of a PTM may therefore simply reflect a change in parent protein abundance, rather than a change in the degree to which a protein is modified. This makes it essential to account for differences in parent protein abundance when comparing PTMs in tissue and plaque types. We therefore expressed the abundance of the modified peptidoforms and PTMs relative to parent protein abundance, by normalizing the intensities of the modified peptidoforms PTMs to the intensity of its parent protein (both native and modified forms). This reduces the confounding effect of protein abundance differences between tissue and plaque types, and the results reported here are therefore not due to changes in parent protein abundance.

About 40% of PTMs detected in the control and plaque clusters were Met monooxidations (Met sulfoxide) with this being the most abundant PTM (Fig. 3D). Significant levels of dioxygenated Met (Met sulfone [42]) were also detected. The abundance of these, and other PTMs, varied between the control tissue and plaques, and between plaque types (Fig. 4). Data on the (approximate) abundance of each type of modification, and in each sample type, is provided by the log2 intensity data (vertical axes) in Fig. 4 and the detection counts in Suppl. Fig. 3. In addition to the oxidized Met species, type A plaques also contained a significantly higher abundance of all the other PTMs, with the exception of N-terminal carbamylation (more abundant in control), Trp monooxidation and chlorination, and nitrations. Type B plaques contained a higher abundance of cysteic acid (from Cys/cystine), chlorinated Tyr, nitrated Trp and Tyr, and dioxidation of Trp. Only Trp monooxidation did not differ between plaque or tissue types. Type A plaques contained higher levels of oxohistidine, carbamylated Lys, and Met mono- and dioxidation compared to type B, with only Trp nitration found to be more abundant in type B plaques. However, both nitrations and chlorinations displayed modest detection frequency (Suppl. Fig. S2 and S3) and therefore a lower robustness of their quantification. The higher levels of most oxidative PTMs in plaques, and particularly unstable (type A) plaques, is consistent with a heightened inflammatory state, and supports the hypothesis that increased inflammation and oxidants play a central role in atherosclerosis progression [6].

High levels of PTMs, as indicated by both abundance (intensity) and detection (count), were found on cellular proteins (Fig. 5), with the majority of the differentially-abundant PTMs being more abundant in plaques compared to control tissue (Fig. 5A left panel; Fig. 5C top panel), and type A compared to type B plaques (Fig. 5A right panel; Fig. 5C bottom panel). Most modified proteins were detected with only a single PTM of any particular type, resulting in similar data at both the PTM and protein level (Fig. 5A and B). This suggests that some oxidative modifications are ubiquitous and driven to some extent by target abundance, with the iBAQ data (Fig. 6) supporting this assertion. Thus, mono-oxygenations of Met and Trp (Fig. 6A, red symbols; native proteins indicated by gray symbols) are clustered, to a significant extent, on high abundance proteins. This is consistent with a limited selectivity of the oxidants that generate them (reviewed [43]). However, other PTMs show a greater degree of selectivity with regards to the proteins on which they are present, with a number of low abundance proteins detected with cysteic acid (from Cys; Fig. 6D, red versus gray symbols), nitrations (Fig. 6E), chlorinations (Fig. 6F) and carbamylations (Fig. 6G), suggesting that the oxidants that generate these specific PTMs are either more selective, or are generated in close proximity to the target.

Chlorination of Tyr is specific to, and a hallmark of MPO-derived hypochlorous acid, (HOCl) or chloramines (RNHCl) [19]). The high reactivity of these oxidants [54,55] is inconsistent with a selectivity for specific Tyr residues, and it is therefore likely that these chlorination arises from site-specific HOCl/RNHCl generation, possibly via MPO binding to particular proteins, as reported for ECM species and lipoproteins [22,23,56].

The formation of cysteic acid on specific proteins may arise from differences in reactivity between parent Cys residues. This PTM is generated by multiple oxidants, with the rate of oxidation driven partly by the thiol pK_a, with the thiolate anion being more reactive than the neutral form [57,58]). The absence, or slow repair of intermediates (e.g. sulfenic and sulfinic acids) on the pathway to cysteic acid may also be a significant factor in the selective formation of this species [57].

All of the PTMs detected here have been previously associated with MPO-derived oxidants [19,59,60], though this does not indicate causality, except for the chlorinations which are specific (in mammalian cells) to MPO-derived oxidants. Previous studies have reported that MPO is elevated in plaques relative to control tissue (e.g. Refs. [16,18,61]) and in soft/unstable (type A) versus hard/stable (type B) plaques [29]. As data on the MPO levels present in these tissue samples was available, correlation analysis was carried out between MPO levels and the total global PTM load (Fig. 11A), for the sum of carbamylation, chlorination and nitration (Fig. 11B), and carbamylation alone (Fig. 11C). In each case positive and significant correlations were detected for the plaques (p 0.038, p 0.0002 and p 0.0001 respectively) but not control artery samples. These data are consistent with, but do not prove, that MPO is involved in the generation of these PTMs; a number of these species have also been linked previously to enhanced atherosclerosis [62,63]. Together these data support a major role for MPO in plaque PTM formation, though a role for other oxidants cannot be excluded, as nitrations, oxidations and carbamylation can, for example, also be induced by peroxynitrite (ONOO⁻/ONOOH, and derived species such as peroxynitrosocarbonate) [64], metal ions [[65], [66], [67]] and isocyanic acid/isocyanate (HOCN/⁻OCN, from urea or SCN⁻ degradation [68,69]), respectively.

Fig. 11.

Plaque PTM burden correlates with MPO levels. Association between abundance (parent-protein-normalized intensities, log2) of (A) global PTM load (B) carbamylations, chlorinations and nitrations; and (C) carbamylations and MPO concentration (intensity, log2) was assessed by linear regression separately for control (green) and plaques (purple) with sex included as a covariate.

As a significant number of PTMs (both types, and sites modified; Fig. 7, Fig. 8, Fig. 9) are present on key metabolic enzymes, these were mapped to specific metabolic pathways and enzymes. Proteins involved in glycolysis were modified to a major extent, and these PTMs were more abundant than for other metabolic pathways (Fig. 7, Fig. 8, Fig. 9). Proteins of the TCA cycle, PPP and mitochondrial oxidative phosphorylation complexes were however also affected. In each case the PTM loads (precursor detection (count) and/or abundance (intensity)) were greater in the plaques than controls (Fig. 7A and B). A clear over-abundance of Met oxidation was detected on proteins from all four pathways, though to varying extents. Of particular interest is the high PTM abundance on the rate-determining enzymes of glycolysis: hexokinase (HK), phosphofructokinase (PFKM), pyruvate kinase (PKM) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Fig. 7D) [70,71]. Whether, and how, the individual PTMs modulate enzyme activity, and the total pathway flux, is unclear, though some of the detected PTMs are associated with enzyme inactivation in vitro (e.g. Met 44 and Cys 152 of GAPDH [[71], [72], [73]]). The role of specific PTMs is difficult to assess, due to both their number, and the abundance of sites affected, with the overall effect on cell metabolism likely to arise from the cumulative effects of all the PTMs. In vitro studies have shed light on the effects of some specific oxidants and generated PTMs (e.g. on glycolytic and PPP enzymes; reviewed [74]).

The high abundance of PTMs on glycolytic enzymes is consistent with the established dependence of arterial endothelial [[75], [76], [77]] and smooth muscle cells [78,79] on glycolysis for energy requirements, and a high abundance of these proteins (cf. data for cancer cells [80]). Exposure of endothelial cells to inflammatory conditions, as occurs during atherosclerosis, can enhance this glycolytic dependence [81] and strong links have been reported between abnormal glycolytic flux and disease development [4]. Furthermore, an association has been reported between altered metabolism and plaque instability, with high-risk plaques exhibiting increased glycolysis, elevated amino acid utilization and a decreased level of fatty acid oxidation when compared to stable plaques [82].

Enhanced glycolysis is associated with decreased O₂ availability and partial hypoxia, processes that can drive a switch from oxidative phosphorylation to glycolysis (either aerobic or anaerobic), with an upregulation of proteins involved in glycolysis and the PPP [83], and also oxidant formation [84]. Enhanced glycolytic activity also occurs in human coronary artery smooth muscle cells exposed to hypoxia (Jørgensen, submitted), with the KLF4-PFKFB3 pathway closely associated with phenotypic switching of these cells from a quiescent contractile phenotype towards the migratory, proliferative and synthetic phenotype present in plaques [78,85]. Metabolite analysis has shown reduced glucose and elevated lactate levels, together with increased expression of the genes of several glycolysis and PPP enzymes, in high-risk compared to low-risk plaques [82]; despite this metabolic reprogramming, no differences in ATP levels have been reported in such plaques [82]. Metabolic adaptation, with changes in polarization have also been reported for plaque macrophages, with a switching from oxidative phosphorylation to glycolysis, and increased generation of pro-inflammatory cytokines [4]. These previous observations are, in general, consistent with the current data.

In contrast to the abundant changes observed for glycolytic enzymes, lower levels were observed on enzymes unique to the PPP, with only PGM (phosphoglucomutase; 9 modifications), GPI (glucose-6-phosphate isomerase; 8 modifications), TKT (transketolase; 8 modifications), G6PD (glucose-6-phosphate dehydrogenase; 4 modifications) and TALDO1 (transaldolase 1; 4 modifications) showing significant PTM levels. A number of dual pathway (glycolysis and PPP) enzymes (e.g. fructose-1,6-bisphosphatase, FBP; phosphofructokinase, PFK) were also modified to a significant extent (Fig. 8A–C; Fig. 9A, C, Suppl. Fig. 4A, C). Despite the lower levels of PTMs detected on PPP proteins, these were still significantly elevated in plaques compared to the control arteries, and in type A plaques compared to type B plaques (Fig. 8C and Fig. 9C, Suppl. Fig. 4C), potentially as a result of increased cellular demand for NADPH for cellular antioxidant defenses and repair.

Low fold-changes in modifications levels were detected on mitochondrial oxidative phosphorylation complex proteins in plaques (Fig. 9D), although these were generally elevated when compared to control artery samples (Fig. 8D). The proteins affected by PTM formation (with complex site) were: NDUFAB1 and NDUFS (Complex I; both with single modifications), SDH (Complex II; single modification), CYC1 and UQCRC2 (Complex III; 2 and 3 modifications respectively), COX7 (Complex IV; 1 modification) and multiple ATP5F (27 modifications) and ATP6 (30 modifications) subunits of Complex V (ATP synthase). PPA1 (inorganic pyrophosphatase 1) was also detected with a single modification. Surprisingly, these PTMs were most abundant on Complex V subunits, with only limited numbers observed on Complexe I-IV proteins involved in electron transport. This may be associated with a decreased importance of oxidative phosphorylation (and therefore oxidant flux), and lower protein abundance in endothelial and smooth muscle cells, when compared to glycolytic proteins [[75], [76], [77], [78], [79]]. Further studies on this PTM distribution would seem warranted. Whilst a high protein abundance may help explain the high level of PTMs on particular proteins, other factors may also contribute, including a longer lifetime (slower turnover) that would favor PTM accumulation (cf. data for murine proteins [86]), as well as potential alterations in repair/removal capacity, with these also being susceptible to oxidative inactivation (e.g. proteasome [87,88] and lysosomal activity [89]).

This study has a number of strengths and weaknesses. Strengths include the number of human plaques examined, and their comparison with matched “healthy” (superior thyroid) artery tissue. The depth and breadth of the analyses, and the use of unbiased approaches, is also a significant strength. The proteome data have also been compared (in a blinded manner) with plaque gross morphology and clinical data including the presence of hemorrhage/erosion and bleeding, and ultrasound images, allowing associations to be drawn between PTMs and plaque instability.

Whilst a significant number of PTMs have been examined, these analyses are not exhaustive due to the computational power required to analyze large numbers of samples for multiple PTMs, with this increasing exponentially with both sample size and the number of PTMs analyzed for. Additional modifications are undoubtedly present. The data obtained are also not absolute (only relative) quantifications, due to the lack of authentic standards of the modified peptides required for calibration. A further weakness of the study is an incomplete analysis of Cys-derived products (e.g. sulfenic and sulfinic acids, disulfides [57]) as these are destroyed by the reduction and alkylation used in sample processing; these steps are however essential for good proteome coverage. A lack of comparable histology, and validation of the PTMs by other methods (e.g. immunofluorescence) is also a weakness, with these being incompatible with the (destructive) LC-MS/MS workflow, and tissue heterogeneity which precludes the easy use of neighboring tissue as a comparator. Studies on asymptomatic plaques, earlier stages of disease and additional controls (including other artery beds) would clearly be worthwhile, but obtaining such samples is challenging for ethical reasons. Use of post-mortem samples would be possible, but the extended time before sampling is permitted would raise concerns regarding artefactual PTM generation which are difficult to control for.

5. Conclusions

The analyses described here indicate that both control arteries and advanced human carotid artery plaques contain large numbers of different PTMs (including mono- and di-oxidations at Met, Trp, His and Tyr residues, cysteic acid formation from Cys, chlorination and nitration of Tyr and Trp, and carbamylations on Lys and N-termini) on a large number of proteins. Higher PTM levels are present on plaque proteins than those from control arteries, but the types and patterns of PTMs remained relatively constant. Higher PTM levels, in total and for specific species, are present in unstable (type A) compared to stable (type B) plaques consistent with a role in plaque instability. These data show a clear ‘fingerprint’ of plaque protein modifications and an association of these levels (at least in part) with MPO protein, its oxidants and inflammation. Some of these species may be suitable as disease biomarkers. Large numbers of PTMs were detected on key cell proteins including those involved in glycolysis/gluconeogenesis, the pentose phosphate pathway, the tricarboxylic acid cycle and oxidative phosphorylation. Particularly high PTM levels were associated with key glycolytic enzymes consistent with marked perturbations to cellular metabolism in atherosclerotic plaques.

CRediT authorship contribution statement

Karen C. Yang-Jensen: Writing – review & editing, Writing – original draft, Visualization, Validation, Methodology, Investigation, Formal analysis, Data curation. Lasse G. Lorentzen: Writing – review & editing, Visualization, Validation, Supervision, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation. Karin Yeung: Writing – review & editing, Validation, Resources, Methodology, Investigation, Formal analysis, Data curation. Camilo López-Alarcón: Writing – review & editing, Formal analysis. Jonas P. Eiberg: Writing – review & editing, Supervision, Resources, Project administration, Investigation, Funding acquisition, Data curation, Conceptualization. Michael J. Davies: Writing – review & editing, Writing – original draft, Supervision, Resources, Project administration, Funding acquisition, Data curation, Conceptualization.

Ethics declaration

Written informed consent to take part in the study and to publish the article has been obtained from all participants or their legal representatives. The privacy rights of participants have been observed.

This study included organ or tissue donors. This study includes human biological material and consent was obtained from donors, or their next of kin or legal representatives, for use in this study and for publication of the article. The samples used in this research were not sourced from executed prisoners or prisoners of conscience.

This study was performed in compliance with relevant laws, regulatory frameworks and guidelines where the research took place. This study was approved by the Danish National Committee on Health Research Ethics. (Approval No. H-20002776).

Funding

This work was supported by grants from the Novo Nordisk Foundation (NNF20SA0064214 to M.J.D. and NNF22OC0079176 to L.G. L.) and the University of Copenhagen-Novo Nordisk Foundation BRIDGE program (fellowship to L.G.L.).

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:MJD is a member of the Editorial Board of Redox Biology but was not involved in the review or decision making on this manuscript. MJD is joint Editor-in-Chief of the journal Redox Biochemistry and Chemistry, and a member of the Editorial Board of Free Radical Biology and Medicine, which are both sister journals of Redox Biology. The other authors do not have conflicts of interest with regard to the data reported in this manuscript.

Appendix A. Supplementary data

The following is the Supplementary data to this article: