The study protocol conforms to the Ethical Guidelines of the World Medical Association Declaration of Helsinki and laws and regulations in Poland. Before the project commenced, appropriate approval was obtained from the Bioethics Committee at the Poznan University of Medical Sciences, Poland (no. 926/16). All subjects qualified for this study provided signed informed consent for inclusion before participating.

The study involved 201 individuals with CKD, CVD, and healthy volunteers (HVs) without renal or cardiovascular issues. They were divided into five experimental groups, named CKD1-2, CKD5, CVD1, CVD2, and HVs, as presented in Table 1. According to NICE Clinical Guidelines, the CKD patients were included in two groups based on their eGFR values (NICE Clinical Guidelines, 2014). The first group, CKD1-2, encompassed patients at the initial stage of CKD with a mean eGFR of 69 mL/min/1.73 m². The CKD5 group included end-stage renal disease patients treated with hemodialysis with a mean eGFR of 9.1 mL/min/1.73 m². Two groups of CVD patients varying in the degree of CVD clinical manifestation (CVD1 and CVD2) but without kidney dysfunction and thus eGFR >90 mL/min/1.73 m² were also analyzed.

The CKD1-2 and CVD1 groups included patients with the initial stage of atherosclerosis without any previous cardiovascular events and vascular interventions. Still, hypertension, hyperlipidemia, and non-obstructive CAD (non-hemodynamically significant stenosis less than 30%) were confirmed in coronary angiography. However, both groups differed in kidney function.

Patients in the CVD2 and CKD5 groups had advanced atherosclerosis, confirmed by coronary angiography and clinically manifested as CAD, with a history of at least one acute coronary syndrome (ACS) and/or after the vascular intervention. Again, only renal function differentiated these groups.

Complete blood count, lipid measurements and hsCRP analyses were performed. All individuals underwent an examination such as electrocardiography, echocardiography, Doppler ultrasonography, and coronary angiography. The HVs group was employed only as a reference to published previous studies. Individuals with diabetes, active acute infection, and malignant tumors were excluded from the study. The characteristics of experimental groups are summarized in Supplementary Table S1.

An immunomagnetic technique based on anti-CD14 antibodies was utilized for the separation of monocytes from blood samples. Two mL of peripheral blood was collected and directly used to obtain monocytes using a positive selection with CD14 Whole Blood MicroBeads and MACS Column (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s protocol. Also, plasma samples were obtained, and the whole population of leukocytes was isolated using the RBC lysis solution procedure (Dagur and McCoy, 2015). The whole leukocytes, isolated monocytes, and unlabeled cells remaining after magnetic separation were examined by fluorescence microscopy and flow cytometry to assess the morphology, preparation quality, purity, and apoptosis/necrosis rate. CD14 surface expression was analyzed by fluorescence microscopy and mass spectrometry analysis. Cells were labeled with CD14 FITZ-conjugated antibodies (Invitrogen, Waltham, MA, United States) and analyzed with an Accuri C6 flow cytometer (Becton Dickinson (BD), Franklin Lakes, NJ, United States) and confocal microscopy Leica TCS SP5 (Leica Microsystems, Wetzlar, Germany). The rates of apoptotic and necrotic cells were evaluated with dual staining with CellEvent Casp3/7-FITC and propidium iodide (Thermo Fisher Scientific, Waltham, MA, United States) as previously described (Tracz et al., 2021). Cells and plasma were aliquoted, frozen, and stored in a vapor phase of liquid N₂ until analysis. One hundred monocytes’ samples have been selected for non-targeted LC-MS/MS analysis. All samples were utilized for validation and various targeted analysis.

Pellets of monocytes were suspended in 1% sodium deoxycholate, 50 mM ammonium bicarbonate, and 8 mM dithiothreitol and vortexed at 1,500 rpm for 10 min at 60°C. Next, the samples were homogenized as described (Tracz et al., 2021). After centrifugation the supernatants were used for protein concentration assay (2-D Quant Kit, GE Healthcare, Uppsala, Sweden). Eight micrograms of protein mixture were reduced, alkylated and digested (Tracz et al., 2021). After that, sodium deoxycholate was precipitated with 1 μL of 10% trifluoroacetic acid. The supernatant containing 1 μg of the digested proteins was in random order analyzed by nano-LC-MS/MS using a 185-min gradient method with a Dionex UltiMate3000 RSLCnano System coupled with Q-Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) in one batch as described (Luczak et al., 2016a). Following LC-MS/MS analysis, the raw files were analyzed by Proteome Discoverer, v.2.2.0.388 (Thermo Fisher Scientific, Waltham, MA, United States). Qualitative analysis was performed as previously demonstrated (Luczak et al., 2016a) to scrutinize if any batch effect occurred. The technical/biological variability of samples was estimated by scatter plot and calculated using the Pearson correlation coefficients of the signal intensities and the total ion current. The identification of proteins was performed using the SEQUEST engine (embedded in the Discoverer v.2.2.0.388) against the UniProt reviewed human proteome set (released 20.04.2021) using tolerance levels of 10 ppm for MS, and 0.08 Da for MS/MS, and two missed cleavages were allowed. A percolator algorithm and target-decoy strategy were used to validate peptide identification. Peptide spectral matches (PSM) filtering was performed at a false discovery rate (FDR) of 1% (q-value <0.01). For protein quantitation, data were filtered to exclude reverse and contaminant proteins (keratins, trypsin). Only proteins detected in all samples were quantitatively and statistically analyzed (no missing values). Only proteins identified with a minimum of two peptides at >99% confidence level were accepted for quantitation.

Targeted analyses were performed using Multiple Reaction Monitoring (MRM) approach. This part of the study was conducted on plasma samples derived partially from another cohort of donors than non-targeted analysis to increase statistical power. MPO, sTNFα, SOD3, HMOX1, PON1, GPX3, NOS2, CAT and IFNG were quantitatively analyzed. A list of peptides and transitions (Supplementary Table S2) for these proteins was generated in silico by the open software Skyline 20.1.0.31.17 (Pino et al., 2020) and PeptideAtlas repository (Deutsch, 2010) as described previously (Tracz et al., 2021). The following filters were applied: precursor charge: 2-3, ion charge: 1, ion type: y, b. Peptides containing cysteine and methionine residues were excluded to prevent oxidation. To avoid missed cleavages, arginine and lysine were not allowed in the middle of the sequence, and sequences with lysine and arginine followed by proline were not considered. Two or three of the best transitions with ranks 1–3 for each peptide were monitored. Two isotope-labeled peptides (25 pg) were spiked into samples and used as internal standards for normalization purposes (Supplementary Table S2).

Proteins were prepared and analyzed using UPLC–ESI QqQ mass spectrometer (LCMS-8060, Shimadzu, Kyoto, Japan) as described previously (Tracz et al., 2021). An optimization experiment was applied to check RT, dwell time, and stability of heavy peptide standards. The dwell time was set to 10 ms for each transition based on the peak width observed in this experiment. Two transitions for each heavy standard were examined. Each MRM transition was evaluated for possible interferences by manually checking its neighborhood ion in MS mode. Skyline software was used for MRM peak integration, abundance, standard deviation, coefficient of variation, and fold change calculations. The sum of peptide intensities was used as the protein intensity value. The ratio of peptide to global standard was utilized as a normalization method.

The concentration of plasma homocysteine (Hcy) was measured with two methods: HPLC and MRM. The relative abundance of Hcy was also determined in GC-MS analysis, as described below.

In HPLC-based quantification, Hcy was assayed by conversion to Hcy-thiolactone and post-column derivatization and fluorescence detection as described (Jakubowski, 2002). The concentration of Hcy was calculated from a standard curve for DL-Hcy internal standards (0.5–100 µM) (Sigma-Aldrich, Poland) prepared identically as samples. Data were analyzed with Agilent ChemStation software.

In the MRM-based approach, Hcy was quantified by screening specified precursor to fragment ion transitions. Two transitions were analyzed for 136.05 m/z precursor ion: 90.15 and 56.20 m/z. Plasma samples and standards (0.5–100 µM) (8 μL) were diluted with 8 μL of water, and then 16 μL of 500 mM DTT was added. The mixture was vortexed for 10 min in RT, and afterward, 80 μL of 0.1% formic acid was added, and samples were filtrated through a centrifugal Millipore 10-kDa filter at 14,000 g for 25 min. The filtrates were analyzed by UPLC-ESI-QqQ mass spectrometer (LCMS-8060, Shimadzu, Kyoto, Japan). Five μL of samples were injected into a C18 Kinetex column, 100 × 2.1 mm (i.d), 100A pore size, 2.6 μM particle size (Phenomenex, Torrance, California, United States) with 0.1% formic acid (as phase A) and 0.1% formic acid in acetonitrile (B). The analyses were performed with a constant flow rate of 0.5 mL/min at 40°C with a linear gradient of B from 5% to 90% over 8 min. MRM analyses were performed with a dwell time of 100 ms. Data were analyzed using LabSolutions version 5.86 (Shimadzu, Kyoto, Japan). The concentration of Hcy was calculated from a standard curve for DL-Hcy standards.

The concentration of selected proteins was measured using a commercially available ELISA kit for INFγ (EHIFNG; Thermo Fisher Scientific, Waltham, United States), sTNFα (ELH-TNFα; RayBiotech Life, Atlanta, United States), and iNOS/NOS2 (E-EL-H0753; Elabscience, Houston, United States). ELISA approach was also utilized to quantify protein carbonyls (STA-310; Cell Biolabs, San Diego, United States) and oxLDL (EH0943; FineTest, Hubei, China). All assays were prepared according to the manufacturer’s instructions and measured using an Infinite M200 PRO multimode reader (TECAN, Männedorf, Switzerland).

Plasma AOPPs were determined using spectrophotometric detection according to (Witko-Sarsat et al., 1996) with minor modifications. Two hundred μL chloramine T-solution (0–200 µM) or 200 μL of plasma diluted 1:5 with PBS were mixed with 10 µL 1.16 M potassium iodide and 20 µL acetic acid. The absorbance at 340 nm was measured immediately using an Infinite M200 PRO multimode reader (TECAN, Männedorf, Switzerland).

The concentration of reduced (GSH) and oxidized (GSSG) glutathione, as well as the ratio between GSSG/GSH, was determined in monocytes’ samples using OxiSelect™ assay (STA-312; Cell Biolabs, San Diego, United States) according to the manufacturer’s instructions. An infinite M200 PRO multimode reader (TECAN, Männedorf, Switzerland) was utilized for recording the absorbance. Level of glutathione components, intermediates in the glutathione synthesis, and molecules involved in glutathione components: L-cysteine, L-glutamic acid, L-glycine, γ-L-glutamyl-L-cysteine, L-cysteinyl-L-glycine, 5-oxo-L-proline, and L-ornithine, were determined using GC-MS analysis. GPX3 was quantified using the MRM approach as described above.

Colorimetric assay Oxystat (BI-5007; Biomedica, Vienna, Austria) was used to characterize the oxidative status in analyzed samples. The quantitative determination of total circulating peroxides, including organic peroxides and hydroperoxides, was performed using 10 µL plasma samples according to the manufacturer’s instructions and measured with Infinite M200 PRO reader (TECAN, Männedorf, Switzerland).

Plasma samples were derivatized and analyzed by gas chromatography-mass spectrometry (GC/MS) using TRACE 1310 GC system connected to TSQ8000 triple quad (Thermo Fisher Scientific, Bremen, Germany) as described previously (Marczak et al., 2021). The identified compounds were quantified by applying extracted ion chromatograms (EIC) based on selected m/z values and retention times corresponding to specific compounds. The unique masses were chosen based on NIST library spectra and EIC peaks were integrated to measure their areas. The total ion current was measured by integrating all TIC peaks and used for compounds normalization.

A cohort of 201 human blood samples derived from 5 experimental groups was analyzed. Plasma and cells were randomized before sample preparation and then prior to analysis. Non-targeted proteomic analyses of monocytes were performed on 20–25 biological replicates for each experimental group and two technical repetitions were applied. Targeted analyses using the MRM approach were accomplished using plasma samples from 20 to 25 samples for each group, derived partially from another cohort of individuals than non-targeted analysis. Two technical repetitions were performed. Homocysteine was determined using the HPLC method for 30 biological repetitions (from each group) in duplicate. For the ELISA analysis, 10–15 samples from each group were compared with 2 technical repetitions. AOPPs and peroxide concentration were measured in 25 samples from each group prepared in triplicate. For glutathione-related analysis, 15–25 samples from each group were analyzed with 3 technical replicates. GC analysis was performed on 30 biological repetitions from each experimental group.

Statistical analyses were performed using Perseus 1.6.13 (Cox and Mann, 2008), Statistica 12.0 (StatSoft, Inc., Kraków, Poland), or MetaboAnalyst software (Xia et al., 2009). The chi-square test was used for categorical variables. Data distribution was assessed using a Shapiro−Wilk test and Leven’s test to evaluate the equality of variances. Data were log2 transformed, and the isolation Forest algorithm was used for outlier detection. The data were statistically analyzed using one-way ANOVA or Kruskal−Wallis, followed by Bonferroni or Dunn’s post hoc tests. Multiple testing was performed with or without the HVs group. A Mann - Whitney U-test or Student’s unpaired t-test was utilized to compare 2 groups. Statistical significance was accepted as p < 0.05 with the Benjamini−Hochberg (B-H) correction, and FDR was set to 5%. For results derived from non-targeted LC-MS/MS proteomics, the fold changes in the level were assessed. According to our previous study and others, fold change 1.4 was considered significant (Levin, 2011; Tracz et al., 2021). Therefore, a protein was found to be differentially expressed if the difference between at least two groups was statistically significant (p < 0.05) and the fold change was ±1.4. Only differentially expressed proteins (DEPs) identified with a minimum of two unique peptides at >99% confidence level were accepted.

The linear correlation was analyzed using the Spearman or Pearson correlation coefficient. Multivariate analyses were carried out by unsupervised principal component analysis (PCA) and hierarchical clustering. For hierarchical clustering and heat map visualization, auto-scaling was additionally performed. Data derived from relative measurements presented as box plots were logarithmically transformed and automatically scaled. Results of concentration measurements were presented without transformation and scaling. The black dots on box plots correspond to the normalized (log10 and auto-scaled) protein abundance. The notch indicates the 95% confidence interval around the median of each group. The mean level is shown with a yellow diamond.

Bioinformatic analysis was conducted using Ingenuity Pathway Analysis software (IPA; Ingenuity Systems, Redwood City, United States) as previously (Tracz et al., 2021). Pathview software was utilized for data integration and pathway visualization (Luo and Brouwer, 2013; Luo et al., 2017). Network analysis and visualization of proteomics data were also performed with Cytoscape String App (Doncheva et al., 2018). All identified proteins were annotated according to their Gene Ontology in the cellular compartment, pathway, and biofunction using UniProtKB list. All DEPs were subjected to the enrichment analysis to determine the top canonical pathways, biological functions, and upstream regulators associated with the observed differences in protein profiles using the right-tailed Fisher’s exact test with B-H multiple corrections. In IPA software z-score algorithm was also used to predict the direction of change for a given function or pathway. A negative z-score indicated inhibition and positively predicted activation. Only functions with z-scores ≥2 (for activation) and ≤ −2 (for inhibition) were considered significant. On this basis, graphical summaries of performed functional analysis were automatically generated with the machine learning method in IPA software. As a result, networks demonstrating the most significant pathways, biological functions, diseases, upstream regulators, and predicted relationships, were obtained.

In the next step, differentially expressed proteins (DEPs; p < 0.05; fold change ≥1.4; ≥2 peptides) significantly altered amid experimental groups with atherosclerosis-related and non-related to CKD were identified (Supplementary Table S3). First, the groups with initial symptoms of atherosclerosis and healthy kidneys (CVD1) or initial stages (1 and 2) of CKD (CKD1-2) were compared. Only 74 DEPs were identified in the group of 2,383 proteins, representing approximately 3.1% of all quantified proteins (Figure 2C; Supplementary Table S3). Also, unsupervised principal component analysis (PCA) (Supplementary Figure S1 and Supplementary Figures S2A; left panel) showed that CKD1-2 and CVD1 clusters together. Graphical presentation of up- and downregulated DEPs in CKD1-2 and CVD1 comparison is demonstrated on volcano plot (Supplementary Figures S2A; right panel).

The identified 74 DEPs were then included in the bioinformatic surveys and functionally analyzed utilizing IPA software. Proteins were annotated according to their Gene Ontology regarding canonical pathways, diseases, and biofunction categories and subjected to enrichment analysis to determine the top categories. Functional analysis revealed that DEPs identified in the CKD1-2 vs. CVD1 comparison were mainly involved in Vasculogenesis, Exocytosis, Bleeding, Morbidity/Mortality, and Kidney Damage GO categories (Figure 2A). According to the calculated z-score, these biofunctions were activated in CKD1-2 compared to CVD1. On the other hand, Metabolism of Lipids, Concentration of Cholesterol, Paxillin Signaling, Coagulation, Hemostasis, and Cell Spreading categories were also overrepresented according to the Fisher test but were predicted to be inhibited in CKD1-2 as compared to CVD1. Serum Response Factor (SRF) was identified as a primary negative regulator responsible for observed changes. The graphical summary of these overrepresented categories and predicted relationships between them is presented in Figure 2A. A detailed list of annotations is presented in Supplementary Tables S4, S5.

In the next step, proteomic profiles of monocytes derived from patients with advanced atherosclerosis and different renal function were compared. Thus, CVD2 (healthy kidneys) and CKD5 (kidney failure) groups were analyzed. As a result, 679 DEPs were identified (p < 0.05; fold change ≥1.4; ≥2 peptides; Figure 2C; Supplementary Table S3). PCA revealed that CKD5 and CVD2 are in separate clusters (Supplementary Figure S1 and Supplementary Figures S2B). Among 679 DEPs, 539 were upregulated, and only 140 were downregulated in CKD5 compared to CVD2, as presented on the volcano plot (Supplementary Figures S2B).

Functional analysis of DEPs identified in this comparison revealed Integrin Signaling and Actin Cytoskeleton Signaling as the top-ranked canonical pathways inhibited in CKD5. Also, Coagulation Pathway and Acute Phase Response Signaling were inhibited in CKD5 as compared to CVD2 (Figure 2B; Supplementary Table S4). The canonical pathways most activated in CKD5 included: EIF2 Signaling, NRF2-mediated Oxidative Stress Response, and Glutathione Redox Reactions. Analysis of biological functions and disease categories revealed that Activation of Macrophages, Cell Survival, Binding of Monocytes, Chemotaxis, Recruitment of Leukocytes/Phagocytes, and Cell Movement were overrepresented and inhibited in CKD5 in comparison to CVD2. Among the biofunctions and disease categories activated in CKD5: Morbidity/Mortality, Damage of Kidney, Synthesis/Metabolism of Reactive Oxygen Species, Inflammatory Response and Inflammation of Endothelial Cells, Growth of Bacteria, and Bleeding categories should be stated (Figure 2B; Supplementary Table S5). Among others, TNFα, IFNG, and NFKBIA were identified as the primary upstream regulators with predicted inhibition functionally associated with many of the dysregulated biological functions observed in the study (Figure 2B). Some of these functional categories were revealed as dysregulated in the whole fraction of leukocytes as we presented in the previous study (Tracz et al., 2021). For instance, proteins involved in the Activation/Recruitment of Leukocytes and Integrin Signaling were dysregulated in an advanced stage of CKD. However, the overrepresentation of proteins involved in the synthesis and metabolism of ROS was revealed for the first time as more active in CKD5 than in CVD2. Moreover, this biofunction was predicted to be functionally related not only to Inflammatory Response but also to Integrin Signaling and EIF2 Signaling (Figure 2B). Furthermore, TNFα, IFNG, IL17A, IL6, NFKBIA, and TLR4 were predicted as the top upstream negative regulators of these changes. All of these compounds are closely related to inflammatory reactions and ROS production. Therefore, in the next step, we decided to get more functional insight into oxidative stress (OS) and antioxidant defense in CKD and CVD in relation to inflammation, focusing on the comparison of CKD5 and CVD2. First, we more closely examined the abundance of 109 DEPs derived from non-targeted proteomic analysis and assigned by Gene Ontology to the synthesis and metabolism of ROS and inflammation (Supplementary Table S3). Then, we complemented these results with more targeted techniques to 1) confirm selected results and 2) assess other specific proteins involved in ROS production, different mediators of OS, and finally, some oxidative stress-induced effects at the level of proteins and lipids. More detailed data derived from these measurements are presented in Supplementary Table S6. Some of these compounds were confirmed by more than one method.

Five DEPs involved in glutathione synthesis and metabolism were identified in non-targeted LC-MS/MS analysis: glutathione reductase (GSR), glutathione peroxidase 1 (GPX1), glutathione synthetase (GSS), glutathione S-transferase P (GSTP1), and isocitrate dehydrogenase (IDH1). We supplemented these results with an analysis of monocytes’ concentration of oxidized (GSSG) and reduced (GSH) form of glutathione and an assessment of the GSSG/GSH ratio. We also tested the plasma glutathione peroxidase 3 (GPX3) level using a targeted MRM approach. Eventually, we utilized the GC-MS approach to assess the abundance of glutathione components and some of the intermediates involved in glutathione synthesis: L-cysteine, L-glutamic acid, L-glycine, 5-oxo-L-proline, and L-ornithine. These experimental results are graphically illustrated in Figure 3. The level of GSR, an essential enzyme that recycles oxidized glutathione (GSSG) back to the reduced form (GSH), was significantly decreased in CKD5 as compared to CVD2 (Figures 3A, I). Consequently, the concentration of monocytes’ oxidized form of glutathione (GSSG) was elevated in CKD5 group; however, the reduced form of glutathione (GSH) was not altered in this comparison (Figures 3F, G). We also noticed the downregulation in CKD5 of other antioxidant enzymes, cytosolic GPX1 and extracellular GPX3, catalyzing the reduction of hydrogen peroxides using glutathione as a reducing agent (Figures 3B, C). On the other hand, the level of GSTP1 and cytoplasmic IDH1 was elevated in CKD5 compared to CVD2 (Figures 3D, E). The primary role of GSTs is conjugating GSH to different electrophilic xenobiotics to detoxify them. IDH1 is responsible for NADPH production, a molecule required to regenerate GSH by GSR. The level of GSS was significantly elevated in CKD5 compared to HVs, CKD1-2, or CVD1, but not in CKD5 vs. CVD2 comparison (Figure 3I; Supplementary Table S3). And finally, the ratio of GSSG to GSH was higher in CKD5 than in CVD2 (Figure 3H). In contrast, low molecular compounds involved in glutathione synthesis: cysteine, glutamic acid, glycine, 5-oxo-L-proline, and ornithine, did not demonstrate any significant differences in abundance between CKD5 and CVD2 (Supplementary Table S6). However, the glutamate level was elevated in both groups compared to HVs (Figure 3I).

In the next step, we assessed other essential components of the antioxidant defense system: catalase (CAT), superoxide dismutases (SODs), peroxiredoxin 1 (PRDX1), and heme oxygenase 1 (HMOX1) (Figure 4A). We demonstrated the downregulation of CAT in CKD5, and found elevated accumulation of SOD1 in monocytes and SOD3 in plasma of CKD5 patients’ as compared to CVD2. The mitochondrial form of SOD (SOD2) was also identified in our study but did not differentiate between CKD5 and CVD2 groups (Supplementary Table S3). Other components of the antioxidative system: PRDX1 and HMOX1, were also upregulated in CKD5 compared to CVD2. On the other hand, thioredoxin family enzymes: TXN, TXNDC5, TXNDC12, and TXNDC17 were upregulated in both groups with advanced atherosclerosis (CKD5 and CVD2) as compared to HVs and CKD1-2 and CVD1. But no differences were revealed between CKD5 and CVD2 (Supplementary Table S3).

Consequently, we checked possible sources of ROS in monocytes, including NADPH oxidase complex, myeloperoxidase (MPO), and induced endothelial nitric oxide synthase (iNOS/NOS2).

First, we looked at inducible NOS and MPO (Figure 4A). The level of NOS2 was the highest in CKD5 compared to other groups. Surprisingly, an abundance of MPO was significantly decreased in CKD5 and revealed the lowest level in non-targeted MS analysis amid analyzed groups. Moreover, we verified the MPO accumulation with the targeted MRM method, and the same results were obtained. Correlation analysis was performed in the next step to detect possible relationships between these OS factors. A positive correlation was revealed only for MPO and CAT (ρ = 0.82) (Figure 4B), SOD1 and PRDX1 (0.84), and NOS2 with HMOX1 (0.68). All correlation coefficients and their statistics are presented in Supplementary Table S7.

Then, we pay attention to homocysteine (Hcy), another trigger of oxidative stress closely associated with cardiovascular risk, NOS2 activity, and NO production in endothelial cells. The abundance of this molecule was elevated exclusively in CKD5 compared to other groups (Figure 4A). In detail, its level was even 2.8 times higher, as shown by three independent methods (Figure 4A, Supplementary Table S6). However, other groups did not differentiate between each other. Consequently, we checked the abundance of PON1, an enzyme with an anti-atherogenic role closely associated with Hcy level. Abundance of PON1 was decreased in CKD5 in comparison to CVD2, however in both groups with advanced atherosclerosis, the level of PON1 was downregulated in comparison to HVs and patients with initial atherosclerosis (Figure 4A). The same picture was observed for PON1/HDL ratio (Figure 4A, Supplementary Table S6).

Finally, we tested the abundance of p22phox (CYBA), gp91-phox (CYBB), NCF1, NCF2, and NCF4 - membrane and cytoplasmic components of the NADPH-oxidase complex, and we did not observe any difference between HVs, CKD and CVD groups (Supplementary Figure S3).

In the next step, we focused on the final effects of OS on protein and lipid oxidative modifications. The abundance of advanced oxidation protein products (AOPPs) and protein carbonyls was elevated exclusively in CKD5 compared to other groups (Figures 5A, B). In detail, AOPP and protein carbonyl concentrations were 1.72 and 1.44 times higher in CKD5 compared to CVD2, respectively (Figure 5A).

A completely different effect of oxidative stress was revealed for LDL and HDL lipoproteins. The level of the oxidized form of LDL (oxLDL) was elevated in both CKD5 and CVD2, but no differences were shown between CKD5 and CVD2 (Figure 5C). The same picture was demonstrated for oxLDL/LDL and oxLDL/HDL ratios (Figures 5D, E). Moreover, oxLDL negatively correlated with the total level of cholesterol (ρ = −0.65), but only when CKD groups were analyzed. A positive correlation between total cholesterol and LDL was obtained for CKD and CVD groups in both combined and separated analyses (ρ = 0.88–0.91). OxLDL also correlated with Hcy level but only when CKD patients were analyzed (with the average for all methods ρ = 0.62).

Finally, we checked the overall level of systemic oxidative stress by measuring the total peroxide concentration in plasma (Figure 5F). The obtained results confirmed that oxidative stress is undoubtedly elevated in CKD5 and CVD2 compared to other experimental groups, but no difference between CKD5 and CVD2 was shown.

The functional analysis of identified DEPs also revealed dysregulation of various processes related to inflammation, especially when CKD5 and CVD2 were compared (Figure 2B). Sixty-one DEPs were assigned by both, IPA and Cytoscape analyses to different inflammation-related categories, i.e., Inflammatory Response, Chemokine Signaling, Growth of Bacteria, and Inflammation of Endothelial Cells. According to z-score, activation of these categories was predicted in CKD5. Simultaneously, downregulation of TNFα and INFG as upstream regulators were predicted in CKD5 as compared to CVD2 (Figure 2B). An inhibition was estimated to Activation of Macrophages and Acute Phase Response Signaling. Therefore, we decided to scrutinize more closely the abundance of DEPs assigned by IPA and Cytoscape analyses to inflammation (Figure 6).

To confirm the prediction obtained in bioinformatic analysis, the abundance of sTNFα and INFG was verified by two independent methods, ELISA (TNF₁, INFG₁) and MRM (TNF₂, INFG₂). The obtained results confirmed the prediction and revealed that the concentration of sTNFα and INFG was much lower in CKD5 compared to CVD2 (Figures 6B, C). However, the abundance of sTNFα and INFG in CKD5 was comparable to HVs, and only the result for CVD2 was much higher.

Moreover, a similar picture was obtained for TACE (ADAM17) and inflammatory cytokine TGFB1 (Figure 6A; Supplementary Table S3). TACE is the enzyme responsible for the cleavage of membrane TNFα and releasing the soluble form of TNFα to the plasma. The highest abundance of TACE and TGFB1 was assessed for the CVD groups. However, 10 other involved in TNF signaling pathway proteins identified in the study were upregulated in CKD5 compared to CVD2 (Figure 7).

On the other hand, the abundance of 47 proteins related to inflammatory response was upregulated in CKD5 in comparison to CVD2, including markers of inflammation: CRP and CST3 (Figure 6A). Among the upregulated proteins, a lot of damage-associated molecular patterns (DAMPs) molecules were identified, among other HMGB1, S100 proteins, galectins, and HSPs. DAMPs are compounds triggering an innate immune response by interaction with pattern recognition receptors (PRRs). The latter ones were also identified in this study as upregulated, among them Toll-like receptors (TLR2 and TLR8, IKBKB, PIK3R1, IRF5), proteins involved in NOD-like and RIG-I-like (DDX3X, ISG15, TRIM25, TMEM173, IL18, MAVS, OAS3, PKN1, NAIP) and MAPK signaling pathways (MAPK1, MAPK10, MAP2K1, CASP, ARRB, MAX, JNK, MEK1, ERK, RSK2, IKK) as well as scavenger receptors (CD68, CD44) (Supplementary Table S3). To visualize these results, the protein-protein interaction network was generated by Cytoscape StringApp for proteins with a role in the inflammatory response. The retrieved network revealed potential interactions among almost all analyzed inflammation-related DEPs (Figure 8). Clustering these proteins with the K-means algorithm revealed three different clusters; however, most DEPs were included in the same cluster (marked in green) with TNF and INFG located in the center of the cluster.

Finally, we examined associations between all mentioned oxidative stress and inflammation-related factors and the progression of CKD. Correlation analysis demonstrated that level of Hcy, AOPPs, SOD1, SOD3, PRDX1, oxLDL, and ratios between oxLDL and LDL or HDL negatively correlated with a parameter of kidney function decline–eGFR (all with correlation coefficients of ρ ≤ −0.6) when only CKD patients (CKD1-2 and CKD5) were analyzed. Positive correlations (ρ ≥ 0.7) were revealed for GPX3, CAT, TNFα, and MPO (Supplementary Table S7). Most of these compounds also correlated with each other. AOPPs, Hcy, GPX3, and CAT correlated additionally with markers of inflammation: CRP and CST3, but also only when CKD groups were analyzed. Also, IL18 correlated with SOD1, SOD3, PRDX1, GSTP1, IDH1, MAPK1, ILF3, MIF and TXNDC5. When CVD-derived results were correlated, some of the identified inflammatory proteins revealed significance, for instance, CST3 and ILF3 (ρ = 0.67) but not with oxidative stress-related parameters.