Identification and Verification of Potential Biomarkers in Renal Ischemia-Reperfusion Injury by Integrated Bioinformatic Analysis

Background Renal ischemia-reperfusion injury (RIRI) plays an important role in the poor prognosis of patients with renal transplants. However, the potential targets and mechanism of IRI are still unclear. Method Differential gene expression (DEG) analysis and weighted correlation network analysis (WGCNA) were performed on the GSE27274 dataset. Pathway enrichment analysis on the DEGs was performed. To identify the hub DEGs, we constructed a protein-protein interaction (PPI) network. Finally, the hub genes were verified, and candidate drugs were screened from the DsigDB database. Results A hundred DEGs and four hub genes (Atf3, Psmb6, Psmb8, and Psmb10) were screened out. Pathway enrichment analysis revealed that 100 DEGs were mainly enriched in apoptosis and the TNF signaling pathway. The four hub genes were verified in animal models and another dataset (GSE148420). Thereafter, a PPI network was used to identify the four hub genes (Atf3, Psmb6, Psmb8, and Psmb10). Finally, eight candidate drugs were identified as potential drugs. Conclusion Three hub genes (Psmb6, Psmb8, and Psmb10) were associated with RIRI and could be potential novel biomarkers for RIRI.


Introduction
Several individuals suffer from end-stage kidney disease every year. According to Lv and Zhang [1], 2.618 million individuals underwent renal replacement therapy in 2010. In this population of individuals, 2.05 million individuals underwent dialysis, whereas the remaining patients underwent kidney transplantation. Kidney transplantation is the preferred treatment for patients with end-stage renal disease, providing a better prognosis than dialysis. During kidney transplantation, graft injury begins at the moment the graft is separated from the donor. After a brief warm ischemic period, the grafts are placed in a hypothermal preserving solution for a long cold ischemia time (CIT) [2]. Renal ischemia-reperfusion injury (RIRI) is a process during which tissues and organs undergo an initial interruption of blood flow and subsequent restoration of blood flow. In this process, tissues and organs are subjected to oxidative stress and inflammatory reaction. Early IRI induces later loss by chronic hypoxia, reduced kidney mass, graft vascular injury, and subsequent fibrosis [3]. IRI is an important factor affecting the early functional recovery of a transplanted kidney, usually manifested as acute tubular necrosis and delayed graft function (DGF) [4]. DGF is defined as the requirement for dialysis within 7 days of transplant, and it is distinguished from primary nonfunction (PNF) by the eventual recovery of renal function [5]. Prolonged CIT is an independent risk factor for DGF [6]. DGF is the most common complication after kidney transplantation and a risk factor affecting the survival rate of transplanted kidneys and patients. Clinically, IRI has been a challenge for several years, and it still remains one today. IRI is a multifactorial and intricate process involving the activation of cell death programs, endothelial dysfunction, transcriptional reprogramming, and activation of the innate and adaptive immune systems [7]. Therefore, exploration of the key molecular mechanisms and targets is needed to address the RIRI challenge.
With the help of bioinformatics, we gain new insights into the development of disease. Weighted gene coexpression network analysis (WGCNA) is a system biology method for describing the correlation patterns among genes across microarray samples [8]. WGCNA can identify modules of highly correlated genes, and these methods have been widely used in the fields of bioinformatics to identify candidate biomarkers. Gong et al. [9] performed WGCNA and constructed a protein-protein interaction (PPI) network to identify 10 DEGs in human anaplastic thyroid cancer. Zeng et al. [10] performed WGCNA and combined experimental verification to identify two hub genes in nonalcoholic fatty liver disease.
Some bioinformatic studies on RIRI have been reported. Guo et al. [11] found 10 hub genes and Hif-1α signaling pathways associated with RIRI. Additionally, Zhu et al. [12] found three genes and two miRNAs that may be potential targets for RIRI. By WGCNA, Lin et al. [13] found that Rplp1 and Lgals1 were associated with the development of acute kidney injury. Building on previous studies, our study further verified the accuracy of our results and found potential drugs to treat RIRI.
Above all, we focused on DEGs at both 6 hours and 24 hours after reperfusion and performed WGCNA to find genes linked with the development of IRI. Thereafter, we performed enrichment analysis for DEGs to find potential mechanisms. Finally, we performed a variation of hub genes and found potential drugs. This study sought to offer a novel insight into the pathogenesis of IRI and found potential biomarkers and drugs for treating IRI.

Materials and Methods
2.1. Data Information. Gene expression profiles of GSE27274 (GPL6101) and GSE148420 (GPL14746) were obtained from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) [14]. GSE27274 contains six sham group samples and 18 kidney IRI samples (six samples at 6, 24, and 120 hours after reperfusion). However, in our analysis, we used six samples at 6 and 24 hours after reperfusion. GSE148420 contains four normal samples and eight kidney IRI samples.

Screening Differentially Expressed
Genes. All data were log-transformed, and expression matrices were normalized. The R package of "Limma" was used to identify DEGs between IRI samples and sham group samples. DEGs with jlog 2FoldChangej > 1 and false discovery rate ðFDRÞ < 0:05 were selected.

Function Enrichment Analysis of DEGs.
To further investigate the molecular mechanisms of potential differentially expressed IRGs, we used the David database (https:// david.ncifcrf.gov) [15] to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. GO terms and KEGG pathway with P value < 0.05 were considered significantly enriched.

Identification of IRI-Associated Modules by WGCNA and
Matching Common Genes. In this study, we constructed a coexpression network for normalized gene expression data of the GSE27274 dataset using the "WGCNA" package. The soft threshold was set to 20, whereas scale-free R 2 was 0.8. The minimum number of genes in modules was set as 30 genes. The R package of "VennDiagram" was used to identify the intersection of the differential genes obtained from the GEO database and significant module genes.
2.5. PPI Network Analysis for the Identification of Hub Genes and ROC Curve Analysis in Another Dataset. To further identify hub genes, we established a PPI network on the Search Tool for the Retrieval on Interacting Genes (STRING) (https://string-db.org/) [16], which was used to predict associations among proteins. Additionally, we set the minimum required interaction score at 0.9. Cytoscape plug-in, Cyto-Hubba, was used to identify hub genes. The diagnostic accuracy of hub genes was tested in another dataset.
2.6. Animal Model. Eighteen male Wistar rats weighing 200-300 g were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd (Beijing, China). Wistar rats were subjected to ischemia/reperfusion, as described in our previous study [17]. Rats were allocated to three groups (n = 18) as follows: (1) the six hours after reperfusion group (n = 6), (2) the 24 hours after reperfusion group (n = 6), and (3) the sham group (n = 6). Rats in the 6 and 24 hours after reperfusion groups were anesthetized. Thereafter, their right kidneys were removed, and their left renal pedicles were isolated from connective tissues and clamped for 48 minutes using a vascular clamp. For those in the sham group, they underwent the same procedure but without vessel clipping. The rats were sacrificed after 24 hours, and blood and kidneys were collected for further analysis.

Creatinine Analysis and Hematoxylin and Eosin (H&E)
Staining. An automated biochemical analyzer BS-200 (Mindray, Shenzhen, China) was used to measure the creatinine levels of the collected blood samples. The kidney specimens were embedded in paraffin, sectioned, and stained with H&E. Tubular damage was scored, as described previously [18].
2.8. Western Blot Analysis. Total proteins from frozen tissues were extracted using lysis buffer (P0013J, Biyuntian, China). The protein concentration of samples was measured (BCA Protein Assay Kit (Biyuntian, China)), and equal amounts of protein were loaded for electrophoresis and subsequent transfer to polyvinylidene fluoride membranes (Millipore-Sigma, Burlington, MA, USA). Using 5% BSA for 1 hour, membranes were blocked, and the primary antibody was incubated overnight at 4°C. On the second day, a secondary antibody was incubated at room temperature for 1 hour.

Prediction of Candidate Drugs for Key Genes.
Approved drugs associated with RIRI were obtained from the DsigDB database [19], which was obtained from Enrichr (https:// maayanlab.cloud/Enrichr/). Candidate drugs with adjusted P < 0:01 were screened.
2.11. Statistical Analysis. DEGs were analyzed using the R package of "Limma" and were defined with jlog 2 FoldChangej > 1 and false discovery rate ðFDRÞ < 0:05. Differences between the two groups were identified using the t-test. P < 0:05 was considered statistically significant. Statistical analysis was performed using GraphPad Prism 9 software (GraphPad Software, San Diego, CA, USA). Figure 1 depicts the workflow of the study.  Table 1). (Figure 3(a), Table 2) and KEGG ( Figure 3(b), Table 3) enrichment analyses of these DEGs showed aging, cell surface, and heparin binding as the most common biological terms for biological process, cellular component, and molecular function. Moreover, pertussis was the most enriched term in the KEGG pathway analysis.

Constructed PPI Network for Hub Gene Identification.
To construct a PPI network, we uploaded 100 DEGs to the STRING database (version: 11.5; https://cn.string-db.org/) (Figure 4(a)). Thereafter, we analyzed the top six hub genes using the Cytoscape plug-in. Based on five algorithms of the

BioMed Research International
Cytohubba plug-in (Figure 4(b)), four genes (Psmb6, Psmb8, Psmb10, and Atf3) were identified as key genes of RIRI. These genes were selected for further validation in the RIRI model. Figure 5(a), H&E staining of rat kidneys helped in finding the RIRI group (6 and 24 hours after reperfusion groups), which had significant proximal tubular injury. Concurrently, serum creatinine and BUN levels were significantly higher in the RIRI group than in the sham group ( Figure 5(b)). Additionally, the expression of kidney injury molecule-1 (Kim-1) was positively correlated with the severity of RIRI. Kim-1 was more highly expressed in the RIRI group than in the sham group (Figures 5(c) and 5(d)). Altogether, these results indicated that the RIRI model was successfully established.

External Validation and Verification of Hub Genes.
We performed ROC curve analysis to predict the values of four hub genes. The AUC value for Psmb6, Psmb8, Psmb10, and Atf3 was 0.719, 1, 1, and 0.906, respectively, in the GSE148420 dataset (Figures 6(a)-6(d)). Consistent with our analysis of GSE27274, these four hub genes were highly expressed in the RIRI group (6 and 24 hours after reperfusion groups). However, due to lack of statistical significance, they were placed in Figures S1(a)-1(h). Additionally, we performed western blot analysis and found that Psmb6, Psmb8, and Psmb10 were highly expressed in the RIRI group (Figures 6(f)-6(k)), whereas Atf3 was highly expressed in the 6 hours after reperfusion group (Figures 6(e) and 6(l)).
3.6. Identification of Candidate Drugs Associated with Renal IRI. Eight candidate drugs (Table 4) with adjusted P value < 0.01 were identified in the DSigDB database. Psmb6 and Psmb8 were the most common drug targets, indicating these two genes might play a key role in the development of RIRI.

Discussion
IRI is an inevitable procedure in kidney transplantation, and the severity of kidney injury has a close correlation with graft function. Despite several studies on IRI, efforts to improve long-term kidney transplant outcomes remain challenging. Only a better understanding of RIRI pathogenesis and identification of possible targets can improve the progress of grafts.
A total of 100 DEGs were identified, and pathway enrichment analyses were performed. From the GO enrichment analysis, aging, positive regulation of apoptotic process, response to ethanol, response to drug, response to xenobiotic stimulus, and cellular response to lipopolysaccharides (LPS) were the top biological processes. Cell surface, extracellular space, and lysosomal membrane were the top cellular components. Heparin binding, threonine-type endopeptidase activity, transcriptional activator activity, and RNA polymerase II transcription regulatory region sequencespecific binding were the top molecular functions. As a special form of cell death, apoptosis, characterized by cell shrinkage, DNA fragmentation, and activation of caspases, was first proposed by Kerr et al. [20]. During RIRI, tubular epithelial cells are mostly severely damaged, and kidney injury can lead to a sharp decline in renal function. LPS is the major cell wall component of Gram-negative bacteria [21]. In the classical way, activation of TLR4 by LPS requires the interaction between LPS-binding proteins, followed by a series of reactions. LPS-TLR4 binding activates downstream signaling of Table 1: Overlap of DEGs originating from 6 hours and 24 hours of reperfusion with genes from turquoise modules.

A3GALT2
CLDN5 GSN  NR1I2  SLC25A11   ACOX1  CLIC1  HNF4A  NT5E  SLC34A2  ADAMTS1  CLU  HPN  PCOLCE  SNRPA  ADHFE1  CRY1  ICAM1  PEX11A  SORD  ALDH9A1  CTSD  IER2  PGPEP1  STX8  ANXA1  CTSS  IFITM3  PLSCR1  SULT1A1  ANXA2  CXCL10  IRF1  PMM1  TAX1BP3  ANXA5  CXCL11  JUNB  PSMB10  TIMP1  ARPC1B  DMGDH  KIF22  PSMB6  TINAG  ASL  EIF4EBP1  KLF6  PSMB8  TMBIM1  ATF3  EMP3  LAMC2  PVR  TMEM37  ATP6V0E2  EPN1  LCN2  RASD1  TNFRSF12A  BTG2  FBL LGALS3  KEGG enrichment analysis was mainly enriched in the TNF signaling pathway, proteasome, complement, and coagulation cascades. The inflammatory cytokine, TNF-α, has been established to play an important role in IRI. Several studies [22,23] have indicated that TNF-α mRNA levels were raised significantly in the RIRI group, compared with the sham group. Furthermore, Nagata et al. [24] reported that RIRI can be ameliorated by the anti-TNF-α agent, infliximab. The complement system, an important component of the innate immunity system, is an important participant in the immune-inflammatory reaction, which comprises >40 blood-circulating, membrane-associated,   [25,26] have confirmed that complement is the key mediating factor of chronic renal tubulointerstitial fibrosis after IRI. Activation of the complement system involves three activation pathways: the classical pathway, alternative pathway, and lectin pathway. During RIRI, strong inflammatory complement substances with biological activity are produced by the activation of complement; they include anaphylatoxins (C3a and C5a) and the cytolytic membrane attack complex (MAC) [27]. Additionally, we constructed a PPI network and identified four hub genes: Atf3, Psmb6, Psmb8, and Psmb10. According to our western blot analysis results, the proteins of Psmb6, Psmb8, and Psmb10 were consistently highly expressed in the 6-hour and 24-hour reperfusion groups, whereas Atf3 was only highly expressed in the 6 hours after reperfusion group. Our rt-PCR results indicated a higher tendency in the RIRI group compared to the sham group. However, this tendency was not statistically significant. The lack of statistical significance was probably due to the small sample size and reduced mRNA level changes of these genes compared to protein level changes.
Psmb6, Psmb8, and Psmb10 are all β-subunits of the 20S proteasome core components. The proteasome complex 26S [28] consists of the 20S catalytic core and 19S regulatory particle. The 20S core catalyzes the chamber, which contains β-subunits (β1, β2, and β5), and the 19S regulatory particle binds the polyubiquitinated substrates. Proteasome 20S has been reported to be a probable biomarker of various diseases, such as renal cell carcinoma [29], psoriasis [30], and myocardial ischemia-reperfusion injury [31]. Currently, a study [32] on hypoxia-induced pulmonary vascular in psmb6 has been reported. The expression of psmb6 increased after chronic hypoxia. During RIRI, the production of a large amount of reactive oxygen species (ROS) promoted the activation of Nrf2, which in turn plays a critical role in the protection against ROS-mediated injury.

BioMed Research International
The Nrf2 agonist can alleviate tunicamycin-induced endoplasmic reticulum (ER) stress and increase the expression of Psmb6 [33]. These findings were consistent with our results. Basler et al. [34] found that psmb8 inhibitor ameliorated the pathological symptoms of dextran sulfate sodiuminduced colitis by reducing inflammation. Li et al. [35] found that the deletion of psmb10 attenuates ang II-induced atrial inflammation and oxidative stress. Hypoxia, inflammation, and oxidative stress, as widely known, are closely related to IRI [24,36]. However, these three proteasome subunits have not been reported to be associated with IRI. Our results demonstrated for the first time that the expression levels of these proteins were closely related to the development of RIRI.
Based on the DSigDB database, we found some drugs that were related to these targets by screening for drugs with adjusted P < 0:01. These drugs are listed in Table 4. Clopamide [37] is an oral diuretic used to treat some diseases, such as cardiac failure, nephrosis, chronic kidney failure, and cirrhosis. Azacyclonol [38], a metabolite of terfenadine, is a central nervous system inhibitor. Among these drugs, carfilzomib and borterzomib [39,40] have been reported to be associated with IRI. Both drugs are proteasome inhibitors, and their potentially therapeutic targets are Psmb6 and Psmb8. Wu et al. [39] found that carfilzomib could reverse BNIP3L degeneration and restore mitophagy to alleviate brain IRI. However, borterzomib has different effects on the outcome of IRI-indifferent organs. Liu et al. [40] found that borterzomib alleviated myocardial IRI by activating the Nrf2/HO-1 signaling pathway. Nevertheless, borterzomib exacerbates RIRI, despite decreased secretion of infiltrating T cells and proinflammatory factors [41].
This study had some limitations. First, these findings were obtained from in vivo experiments but not from in vitro experiments and clinical samples. Second, these proteins had higher expression levels in the IRI group. However, their role in RIRI is unknown. Hence, further studies on the role of hub genes in RIRI should be conducted.

Conclusions
By bioinformatic analysis, we found some novel genes associated with the development of RIRI, and these hub genes are potential therapeutic targets for RIRI. Drugs based on target prediction may improve RIRI outcomes.

Data Availability
All data were acquired from public databases, including the GEO database.

Conflicts of Interest
The authors declare no conflicts of interest.