Proteomic Analysis of Human Serum from Patients with Chronic Kidney Disease

Chronic kidney disease (CKD) is an important public health problem in the world. The aim of our research was to identify novel potential serum biomarkers of renal injury. ELISA assay showed that cytokines and chemokines IL-1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12 (p70), IL-13, IL-15, IL-17, Eotaxin, FGFb, G-CSF, GM-CSF, IP-10, MCP-1, MIP-1α, MIP-1β, PDGF-1bb, RANTES, TNF-α and VEGF were significantly higher (R > 0.6, p value < 0.05) in the serum of patients with CKD compared to healthy subjects, and they were positively correlated with well-established markers (urea and creatinine). The multiple reaction monitoring (MRM) quantification method revealed that levels of HSP90B2, AAT, IGSF22, CUL5, PKCE, APOA4, APOE, APOA1, CCDC171, CCDC43, VIL1, Antigen KI-67, NKRF, APPBP2, CAPRI and most complement system proteins were increased in serum of CKD patients compared to the healthy group. Among complement system proteins, the C8G subunit was significantly decreased three-fold in patients with CKD. However, only AAT and HSP90B2 were positively correlated with well-established markers and, therefore, could be proposed as potential biomarkers for CKD.


Protein Identification
Collected protein spots were destained using ferricyanide-thiosulfate as previously described [17] with some modifications. Briefly, gel pieces were washed with Milli-Q water and destained using a 30 mM potassium ferricyanide and 100 mM sodium thiosulfate solution for 5 min. Then, the gel pieces were rinsed three times with Milli-Q water, dehydrated using acetonitrile, and digested overnight (0.1% trypsin (MS grad, Promega, Madison, WI, USA) at 37 • C). Peptides were extracted using 0.2% trifluoroacetic acid (TFA) followed by incubation in 100% acetonitrile (ACN) for 30 min. Collected peptides were dried at 45 • C for 2 h, dissolved in 5 µL of 0.2% TFA and analyzed by a MALDI-TOF mass spectrometer Ultraflex (Bruker Daltonics GmbH, Bremen, Germany). The spectra were recorded in positive reflector mode from 700 to 3500 m/z. Peptide mass fingerprinting (PMF) was performed using the MASCOT software for searching matches in Swiss-Prot and NCBI databases.

LC-MRM/MS Analysis of Serum Digests
Peptides (10 µL) were separated in a reversed phase analytical column (100 × 2.1 mm i.d., Titan C18, 1.9 µm particle size (Supelco, Bellefonte, PA, USA)) with an Agilent 1290 Infinity UHPLC system coupled to a QTRAP 6500 (AB Sciex, Darmstadt, Germany) mass spectrometer. Proteins were separated using 400 µL/min flow rate and a gradient from 5%-95% mobile phase B, temperature 40 • C and a 25 min total run time. Mobile phase A consisted of 95% 0.1% v/v formic acid in water and 5% ACN, and mobile phase B consisted of 95% ACN and 5% 0.1% formic acid in water. The linear gradients were as follows (time: % B): 0.3 min: 5% B; 17 min: 40% B; 18 min: 95% B; 21.5 min: 95% B; 23 min: 5% B; 25 min: 95% B. All acquisition methods used the following parameters: 5200 V capillary voltage; source type Turbo Spray Ion Drive with temperature 500 • C; curtain gas 35 psi; declustering potential 51 V; collision energy was automatically optimized for each transition; flow rate 0.4 mL/min. Mass spectrometric data were analyzed using MultiQuant 3.0.2 software (AB Sciex, Darmstadt, Germany). Skyline 3.6.0 [19] was used to generate precursor/fragment ion pairs, so-called MRM transitions, in silico [20]. The following options were selected: the peptide length was set to 8-25 amino acids, no post-translational modification (PTM) and one missed cleavage was allowed. In addition, two or three charge states of peptides were chosen for further MRM experiments. At least two peptides were chosen for identification of the target protein (Supplementary Table S1). The MRM method included at least three MRM transitions per peptide to select the best transition. Data analysis was done, and the areas for all the transitions were calculated using the Analyst 1.6.2 and MultiQuant 3.0.2 software (AB Sciex, Darmstadt, Germany). Used peptides with unique sequences and scheduled MRM transitions are given in Table 2.

Statistical Analysis
Statistical analysis of the multiple reaction monitoring (MRM) data and ELISA data were performed in the R environment [21]. Statistically significant differences between groups of patients and healthy individuals were accepted as p < 0.05, assessed by the Wilcoxon rank sum test with Benjamini-Hochberg adjustment for MRM data and p < 0.01 for ELISA data. Correlations between the concentrations of serum proteins, cytokines, urea and creatinine were analyzed using the R Hmisc package (based on Spearman's rank correlation coefficient).

Results
Two-dimensional fluorescence difference gel electrophoresis (2D-DIGE) is a convenient method to identify differences in protein profiles of the samples compared; however, it imposes a number of requirements on an analyte. In particular, successful separation and visualization of protein spots in gel depend upon concentrations of the proteins analyzed. The serum proteome has a dynamic range of more than ten orders of magnitude; thus, there is an excess of major proteins of albumin (more than 60% of the total amount of proteins), alpha-, beta-and gamma-globulin fractions (over 30%), with the rest being less than 10% of the total number of serum proteins [4,22,23]. Prior to 2D gel-electrophoresis, samples were depleted with the use of a commercial "ProteoMiner" kit (Bio-Rad) in order to remove major proteins. As can be seen in Figure 1, removal of most of the major fraction facilitated better blood serum protein separation. Protein spots with different expression levels between two patients and two healthy subjects were cut off and identified; these data are presented in Supplementary Table S2. Some proteins were presented by several spots on 2D gel, probably being isoforms of the same protein.

Results
Two-dimensional fluorescence difference gel electrophoresis (2D-DIGE) is a convenient method to identify differences in protein profiles of the samples compared; however, it imposes a number of requirements on an analyte. In particular, successful separation and visualization of protein spots in gel depend upon concentrations of the proteins analyzed. The serum proteome has a dynamic range of more than ten orders of magnitude; thus, there is an excess of major proteins of albumin (more than 60% of the total amount of proteins), alpha-, beta-and gamma-globulin fractions (over 30%), with the rest being less than 10% of the total number of serum proteins [4,22,23]. Prior to 2D gelelectrophoresis, samples were depleted with the use of a commercial "ProteoMiner" kit (Bio-Rad) in order to remove major proteins. As can be seen in Figure 1, removal of most of the major fraction facilitated better blood serum protein separation. Protein spots with different expression levels between two patients and two healthy subjects were cut off and identified; these data are presented in Supplementary Table S2. Some proteins were presented by several spots on 2D gel, probably being isoforms of the same protein. Fifty-six unique, differentially expressed proteins between CKD patients and healthy subjects were identified with MALDI mass spectrometry, and we surmised there might be potential diagnostic markers of early CKD stages among these proteins. To test this hypothesis, 20 proteins with differential expression were quantified in the serum of CKD patients (n = 26) and healthy volunteers (n = 10) using MRM (multiple reaction monitoring). At present, MRM is an advanced method of mass spectrometry and allows simultaneous quantitation of numerous protein  Fifty-six unique, differentially expressed proteins between CKD patients and healthy subjects were identified with MALDI mass spectrometry, and we surmised there might be potential diagnostic markers of early CKD stages among these proteins. To test this hypothesis, 20 proteins with differential expression were quantified in the serum of CKD patients (n = 26) and healthy volunteers (n = 10) using MRM (multiple reaction monitoring). At present, MRM is an advanced method of mass spectrometry and allows simultaneous quantitation of numerous protein concentrations by the signal intensity of daughter ions, being fragments of known parent peptides [24]. For this purpose, we analyzed amino acid sequences of chosen proteins and selected relevant peptides to quantify proteins of interest using Skyline software (Table 2).
Two apolipoproteins, APOA1 and APOA4, were included in the list of proteins selected for MRM, with APOE added as a component that plays an important role in lipid metabolism and is associated with impaired hemodialysis and renal transplant functioning. Moreover, 18 complement system proteins were added into the expanded list, as they were directly involved in inflammatory reactions ( Table 2).
The quantification results for proteins with differential expression in 2D-DIGE and those associated with the complement system are shown in Figures 2 and 3.  In general, concentrations of 26 proteins enlisted in Table 2 were significantly different in patients with CKD and healthy subjects (Figures 2 and 3, Supplementary Tables S3 and S4). IGSF22, HSP90B2, AAT, C4BPA, C3, C1R and C9 concentrations increased more than four times in the group of CKD patients as compared to the control. Serum CFH, CUL5, PKCE, APOA4, APOE, CCDC171, CCDC43, VIL1, antigen KI-67, NKRF, APPBP2, CAPRI, C1QC, C1S, C4, C5, C8A, C8B and MBL2 concentrations were two to three times elevated in patients with CKD. Concentrations of the remaining proteins such as CCT4, PLG, APOA1, LMCD1, CKAP2L, PLB1, PPEF2, Gal-3, MASP2, C6 and FCN3 were not significantly different between the group of CKD patients and healthy subjects. Only the concentration of C8G was significantly decreased (three-fold) in patients with CKD.

Discussion
Modern diagnostics of CKD and monitoring of its course are based on the measurement of serum creatinine concentrations and subsequent calculation of the glomerular filtration rate [3,4]. The disadvantage of this method is its poor effectiveness in diagnosing early CKD. There are a number of studies where potential protein biomarkers of CKD were proposed and evaluated in the serum or plasma of patients [7,8,[11][12][13][14]. Peptide biomarkers can also provide insight into CKD diagnosis and Hierarchical clustering of all analytes was performed by using the Euclidian distance method. The blue and red colors represent negative and positive Spearman's rank correlation coefficients between the two analytes, respectively.

Discussion
Modern diagnostics of CKD and monitoring of its course are based on the measurement of serum creatinine concentrations and subsequent calculation of the glomerular filtration rate [3,4]. The disadvantage of this method is its poor effectiveness in diagnosing early CKD. There are a number of studies where potential protein biomarkers of CKD were proposed and evaluated in the serum or plasma of patients [7,8,[11][12][13][14]. Peptide biomarkers can also provide insight into CKD diagnosis and progression. In a recent study, 273 specific urinary peptides (CKD273 classifier) were identified as reliable predictors of CKD progression [26] and cardiovascular events in CKD [27]. In this study, we hypothesized that if some proteins showed significant quantitative changes in patient serum and correlated with well-established markers, then they would be potential biomarkers for CKD diagnosis. The first stage of our research involved the screening of serum proteins with differential expressions in patients with CKD and healthy individuals using the 2D-DIGE approach. Twenty-one proteins from the 2D dataset were selected as potential protein candidates for biomarkers of the disease (Table 2). AAT, APOA4 and VIL1 concentrations have previously been shown to increase in blood plasma as CKD progresses [28][29][30][31]. Our data confirm the increased blood serum concentrations of these proteins in patients; whereas only serum AAT values correlated with elevated levels of creatinine and cytokines. Our target proteomics results are compared with previously reported data from studies of different renal pathologies in Table 3.  AAT, which is supposed to have anti-inflammatory properties, belongs to the group of serine proteinases [27]. Exogenous AAT inhibited apoptosis and inflammation in renal reperfusion [28] and reduced the urine kidney injury molecule-1 (KIM-1) concentration [30]. Several studies have demonstrated increased blood plasma AAT concentrations in hypoxia [31] and acute myocardial infarction [48,49].
Based on the MRM results, APOE concentrations were significantly higher in the group of patients with CKD, whereas serum APOA1 levels showed no changes. However, in previously reported studies, plasma APOE and APOA1 levels differed for patients with and without CKD, apparently depending on the disease stage and therapy [35,50,51]. In particular, hemodialysis and pharmacotherapy might have an impact on the levels of these proteins.
High antigen KI-67 and CUL5 expression levels in tissues are associated with cancers [44,[52][53][54][55][56]. We observed increased serum concentrations of these proteins in patients with CKD in this study. Furthermore, we saw increased levels of other intracellular proteins such as NKRF, CCDC171, CAPRI, CCDC43, APPBP2, IGSF22 and PKCE, which might be associated with renal tissue necrosis and, thus, with disease progression. The literature on these proteins in CKD is limited and/or contradictory. IGSF22 is known to be similar to cytoskeletal proteins in its structure, with a one-nucleotide substitution in the IGSF22 gene associated with the development of renal carcinoma [41]. The information on PKCE is controversial; protein kinase activation in proximal nephron cells leads to impaired functioning of the mitochondria, oxidative stress, energy deficiency and cell death [57]. At the same time, there is evidence that PKCE has protective functions against ischemic injury in other tissues, particularly in the myocardium and neurons [58][59][60][61]. NKRF is a transcriptional repressor of NF-kappa-B responsive genes [62]. Increased expression of NKRF transcription factor was shown to upregulate IL-1-induced secretion of IL-8 [63]. In our study, increased serum levels of IL-1 and IL-8 were also demonstrated, suggesting that NKRF could synergize with IL-1 to induce IL-8 expression. It should be noted that serum HSP90B2 concentration positively correlated with levels of creatinine, urea and a number of cytokines. However, the data on changes of serum HSP90B2 in various pathologies are presently lacking. We have evaluated key complement system proteins known to be important components of innate immunity and to play important roles in body defense, inflammation, tissue regeneration and other physiological processes. We found that serum C1S, C1QC and C4 concentrations increased two to three times, and that of C1R was elevated four times in patients with CKD, as compared to the control subjects. Increased levels of these proteins might indicate the activation of the complement system under the classic pathway. It is noteworthy that concentrations of lectin pathway activators such as FCN3, Gal-3 and MASP2 did not show any significant difference between the groups, although the MBL2 concentration was twice as high in patients with CKD. MBL2 has been previously shown to be capable of binding with apoptotic and necrotic cells, thereby promoting the activation of phagocytosis of dying cells [64].
Serum concentrations of complement system inhibitors factor H and C4BP were elevated two and four times, respectively, in CKD patients. We found significantly elevated concentrations (two times) of late lytic cascade proteins such as C5, C8A and C8B, and both C3 and C9 were more than four times higher in CKD patients. These data are in consistent with previously reported results about higher plasma MAC (C5b-9) levels in patients with renal diseases [45,46].
It is remarkable that the serum concentration of the subunit C8G did not elevate following the increase of C8A and C8B subunits constituting a single C8 protein. Unlike C8A and C8B, the C8G subunit is a member of the lipocalin family. As Lovelace et al. have previously reported [65], C8g is not involved in the formation of a membrane attack complex, but it only enhances its activity [65]. Perhaps, the subunit C8G has a regulator function as a complement system inhibitor. Furthermore, it has been suggested that decrease in serum C8G might be specific for other chronic diseases.

Conclusions
Early diagnosis of renal dysfunction is essential to improve disease progress and survival of patients with CKD. In this research we quantified differentially expressed proteins and complement components using the MRM approach and measured serum concentrations of cytokines, chemokines and growth factors using the multiplex Luminex xMAP technology in patients with CKD and healthy subjects.
Patient serum levels of AAT, IGS22 and HSP90B2 had a greater than four-fold change. In addition, the data analysis showed a mild correlation between increased serum concentrations of AAT and HSP90B2 in patients with CKD and well-established markers of CKD such as creatinine and urea. Thus, we suggest that proteins AAT and HSP90B2 might be associated with kidney diseases and might be potential markers of CKD. Further investigations of these proteins as early biomarkers are needed to elucidate their clinical usefulness.

Conflicts of Interest:
The authors declare no competing interests.