The unique molecular mechanism of diabetic nephropathy: a bioinformatics analysis of over 250 microarray datasets

Abstract

Background/Aims

Diabetic nephropathy (DN) is one of the main causes of end-stage kidney disease worldwide. Emerging studies have suggested that its pathogenesis is distinct from nondiabetic renal diseases in many aspects. However, it still lacks a comprehensive understanding of the unique molecular mechanism of DN.

Methods

A total of 255 Affymetrix U133 microarray datasets (Affymetrix, Santa Calra, CA, USA) of human glomerular and tubulointerstitial tissues were collected. The 22 215 Affymetrix identifiers shared by the Human Genome U133 Plus 2.0 and U133A Array were extracted to facilitate dataset pooling. Next, a linear model was constructed and the empirical Bayes method was used to select the differentially expressed genes (DEGs) of each kidney disease. Based on these DEG sets, the unique DEGs of DN were identified and further analyzed using gene ontology and pathway enrichment analysis. Finally, the protein–protein interaction networks (PINs) were constructed and hub genes were selected to further refine the results.

Results

A total of 129 and 1251 unique DEGs were identified in the diabetic glomerulus (upregulated n = 83 and downregulated n = 203) and the diabetic tubulointerstitium (upregulated n = 399 and downregulated n = 874), respectively. Enrichment analysis revealed that the DEGs in the diabetic glomerulus were significantly associated with the extracellular matrix, cell growth, regulation of blood coagulation, cholesterol homeostasis, intrinsic apoptotic signaling pathway and renal filtration cell differentiation. In the diabetic tubulointerstitium, the significantly enriched biological processes and pathways included metabolism, the advanced glycation end products–receptor for advanced glycation end products signaling pathway in diabetic complications, the epidermal growth factor receptor (EGFR) signaling pathway, the FoxO signaling pathway, autophagy and ferroptosis. By constructing PINs, several nodes, such as AGR2, CSNK2A1, EGFR and HSPD1, were identified as hub genes, which might play key roles in regulating the development of DN.

Conclusions

Our study not only reveals the unique molecular mechanism of DN but also provides a valuable resource for biomarker and therapeutic target discovery. Some of our findings are promising and should be explored in future work.

INTRODUCTION

Owing to the increasing prevalence of diabetes mellitus, diabetic nephropathy (DN), which is one of its major microvascular complications, has become a worldwide public health problem [1]. DN is associated with more rapid progression, higher risk of cardiovascular disease and higher mortality rates compared with nondiabetic renal diseases (NDRDs). However, only a few antidiabetic drugs, including renin–angiotensin system (RAS) inhibitors and sodium–glucose co-transporter-2 (SGLT2) inhibitors, have been demonstrated to be beneficial to survival. The onset of microalbuminuria (MA) is regarded as an early marker of kidney injury in DN. However, emerging evidence has shown that MA is not an ideal diagnostic tool. As suggested by a recent biopsy study, comorbidity of DN and NDRD is observed in nearly 50% of patients with DN [2]. Therefore a comprehensive understanding of the unique molecular mechanism of DN is urgently needed to improve the diagnosis and treatment.

Numerous animal studies have suggested that the pathogenesis of DN is distinct from NDRDs in many aspects. These studies constructed our knowledge of the molecular mechanism of DN by steadily laying ‘stone on stone’. However, there is still a lack of large-scale, parallel and nonbiased comparisons regarding the molecular differences between DN and NDRD.

For the past 20 years, high-throughput technologies and bioinformatics have emerged as powerful tools to identify the molecular signatures of diseases. For example, using gene set enrichment analysis, a recent study identified 20 hub genes, including albumin (ALB), epidermal growth factor (EGF) and collagen type IV alpha 1 chain (COL4A1), in diabetic tubulointerstitial tissues [3]. Another RNA sequencing study found a set of unique genes for early DN by comparing its transcriptomic profile with advanced DN [4]. However, these studies did not include other chronic kidney diseases (CKDs) and only contained several datasets. To this end, we aim to provide integrative insight into the unique molecular mechanism of DN using a bioinformatics analysis of >250 microarray datasets.

MATERIALS AND METHODS

Dataset collection

The overall study design is presented in Figure 1. First, human microarray datasets including DN, NDRDs [hypertensive nephropathy (HN) and three types of glomerulonephritis (GN) including membranous nephropathy (MN), focal segmental glomerular sclerosis (FSGS) and immunoglobulin A nephropathy (IgAN)] and healthy controls were collected from gene expression omnibus (https://www.ncbi.nlm.nih.gov/geo/) in June 2019 according to the search criteria described in the supplemental documentation. To minimize the platform variation, all datasets were generated by either the Affymetrix U133 Plus 2.0 or U133 Array (Affymetrix, Santa Clara, CA, USA) , which both belong to the Affymetrix Human Genome U133 platform. The summary and sample description of each dataset were carefully evaluated by two investigators (L.T. Z. and Y.S.Z.) before being included.

FIGURE 1:

Overall study design and bioinformatics analysis workflow. The methods used in corresponding processes were presented in bold type in the figure. Quality control: QC stats in the simplify package and RNA degradation analysis in the affy package; data preprocessing: the RMA method in the affy package; DEG selection: the empirical Bayes method in the limma package and RMA: robust multiarray average.

Microarray dataset processing and DEG identification

The bioinformatics analysis was performed using R (version 3.6.0; R Foundation for Statistical Computing, Vienna, Austria) according to the systematic workflow described in Figure 1 [5]. Briefly, the quality of each dataset was examined using the general quality control stats in the simplify package and the RNA degradation analysis in the affy package. Then the 22 215 shared Affymetrix identifiers of the Human Genome U133 platform were extracted and datasets of the same category were pooled to expand the sample size. In particular, the healthy controls were pooled from all studies regardless of kidney diseases. The relative log expression graph was used to evaluate the consistency among datasets and those with significant bias were discarded. Then the robust multiarray average method in the affy package was used to preprocess the original data. Lastly, the empirical Bayes method in the limma package was used to select differentially expressed genes (DEGs). Statistically significant DEGs were defined as those with a P-value <0.05 after adjustment by the Benjamini–Hochberg method and fold changes >1.5.

Then we defined the unique DEGs of DN as those dysregulated in DN but not in the other types of kidney diseases. Mathematically, it can be written as uniDN = DN − (IgAN∪MN∪FSGS∪HN), where uniDN denotes the unque DEG set of DN and ∪ denotes the union operation. Let GN = (IgAN∪MN∪FSGS) and the final equation becomes uniDN = DN − (GN∪HN)’, from which the unique DEGs of DN were identified.

GO and pathway enrichment analysis

The enrichment analysis was performed using the cluster Profiler package (version 3.12.0) in R [6]. Specifically, gene ontology (GO) annotation (http://geneontology.org/), the Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway (http://www.genome.jp/kegg) and the Reactome Pathway (http://www.reactome.org) query were realized via Annotation Hub, the KEGG Pathway Module and ReactomePA, respectively. The ggplot2 and enrich plot packages were used for visualization. A hypergeometric test was applied for enrichment analysis. P-values <0.05 after adjustment by the Benjamini–Hochberg method were regarded as statistically significant.

Construction of the protein–protein interaction network (PIN), recognition of hub genes and cluster analysis

The PIN analysis was performed based on three databases:the Biological General Repository for Interaction Datasets (BioGRID;http://thebiogrid.org/), the Database of Interacting Proteins (DIP; http://dip.doe-mbi.ucla.edu/) and the Human Protein Reference Database (HPRD; http://www.hprd.org/). Cytoscape software (version 3.6.0) was utilized to visualize PINs and analyze the characteristics of each node. The degree method in the CytoHubba plug-in was used to identify hub genes, defined as the number of edges incident to each node [7].

RESULTS

Characteristics of datasets

As presented in Table 1, a total of 255 Affymetrix U133 microarray datasets including biopsy-proven CKDs (DN tubulointerstitium n = 11, DN glomeruli n = 7; HN tubulointerstitium n = 21, HN glomeruli n = 15, IgAN tubulointerstitium n = 25, IgAN glomeruli n = 43; MN tubulointerstitium n = 18, MN glomeruli n = 18; FSGS tubulointerstitium n = 12, FSGS glomeruli n = 23) and healthy controls (normal tubulointerstitium n = 22, normal glomeruli n = 40) were collected. Healthy controls were either patients after tumor nephrectomy or pretransplant living donors. The resources of the datasets include GSE37463 [8], GSE47185 [9], GSE35489 [10], GSE21785 [11], GSE20602 [12], GSE50469 [13] and GSE69438 [14]. The baseline clinical characteristics of each included cohort can be found in the supplemental documentation.

Table 1.

Datasets included in the study after QC

Sample names	Cases, n	Controls, n	Resources	Platforms
DN glomeruli	7	18	GSE37463 [8], GSE47185 [9]	Affymetrix U133 Plus 2.0
DN tubulointerstitium	11	22	GSE35489 [10], GSE47185 [9]	Affymetrix U133 A
HN glomeruli	15	22	GSE47185 [9], GSE37463 [8], GSE21785 [11], GSE20602 [12]	Affymetrix U133 A
HN tubulointerstitium	21	22	GSE47185 [9], GSE37463 [8]	Affymetrix U133 A
IgAN glomeruli	43	22	GSE50469 [13], GSE37463 [8], GSE21785 [11], GSE20602 [12]	Affymetrix U133 A
IgAN tubulointerstitium	25	22	GSE35489 [10], GSE47185 [9]	Affymetrix U133 A
MN glomeruli	18	22	GSE47185 [9], GSE37463 [8], GSE21785 [11], GSE20602 [12]	Affymetrix U133 A
MN tubulointerstitium	18	22	GSE35489 [10], GSE47185 [9], GSE69438 [14]	Affymetrix U133 A
FSGS glomeruli	23	40	GSE37463 [8], GSE47185 [9], GSE20602 [12], GSE21785 [11]	Affymetrix U133 A Affymetrix U133 Plus 2.0
FSGS tubulointerstitium	12	22	GSE35489 [10], GSE47185 [9]	Affymetrix U133 A

Unique DEGs of DN

First, we constructed the gene expression patterns of all CKDs (Figure 2). As shown in Figure 3, we identified 417 DEGs (upregulated n = 303, downregulated n = 114) in the diabetic glomerulus, among which 129 DEGs were unique to DN (30.9%; upregulated n = 83, downregulated n = 46). In comparison, many more transcripts (upregulated n = 1291, downregulated n = 926) were dysregulated in the glomerular compartment of GN (Figure 2).

FIGURE 2:

Venn plots for identification of the unique DEGs in DN. Different colors represent different primary diseases. Overlapping areas represent shared DEGs. (A) Upregulated genes in glomerular compartments in various types of CKD; (B) downregulated genes in glomerular compartments in various types of CKD; (C) upregulated genes in tubulointerstitial compartments in various types of CKD and (D) downregulated genes in tubulointerstitial compartments in various types of CKD.

FIGURE 3:

Volcano plots showing the unique DEGs of the diabetic (A) glomerulus and (B) tubulointerstitium.

In the diabetic tubulointerstitium, 1251 DEGs were identified (Figure 2; upregulated n = 399, downregulated n = 874). Among these, 851 DEGs were DN-specific (68.0%; upregulated n = 205, downregulated n = 646) (Figure 3). Complete lists of the DEGs were presented in the Supplementary data, Table S1.

GO enrichment analysis of unique DEGs

The glomerular DEG set of DN was significantly enriched in 58 GO terms [Supplementary data, Table S2; molecular function (MF) n = 6, cellular component (CC) n = 4, biological process (BP) n = 48]. The 30 most enriched terms and involved genes were presented in Figure 4 and Table 2, respectively. Notably, the extracellular matrix, a CC term, was identified as the most enriched term. Further pathway analysis indicated that the involved DEGs were most related to the Wnt signaling pathway, the formation of fibrin clot (clotting cascade) and diseases associated with glycosaminoglycan metabolism (Supplementary data, Table S4). The enriched BP and MF terms include kidney epithelium development (BP), regulation of cell growth (BP), cholesterol homeostasis (BP) and cell adhesion molecule binding (MF).

FIGURE 4:

Significantly enriched GO terms of DEGs in the diabetic (A) glomerulus and (B) tubulointerstitium.

Table 2.

Enriched GO terms and involved genes

ID	Description	Genes
Diabetic glomerular compartment
GO: 0031012	Extracellular matrix	MMP7, LUM, SLPI, SFRP1, CXCL12, SERPINE2, ANXA2, SERPING1, GPC4, DCN, MXRA7, ANXA4, F3, CCN2, FGF1, NDNF
GO: 0062023	Collagen-containing extracellular matrix	LUM, SLPI, SFRP1, CXCL12, SERPINE2, ANXA2, SERPING1, GPC4, DCN, MXRA7, ANXA4, F3, CCN2
GO: 0072073	Kidney epithelium development	PROM1, SFRP1, TFAP2B, CD24, SMAD5, EPHA7, KLHL3, FGF1, NPHS2
GO: 0002237	Response to molecule of bacterial origin	SLPI, VIM, CD24, ABCA1, HMGB2, DCN, LIAS, THBD, CEBPB, GCH1, NFKBIA, ADM
GO: 0098552	Side of membrane	VTCN1, SERPINE2, ABCA1, GPC4, LDLR, IL1RL1, F3, HLA-DRB4, CD83, NPHS2, RGS2
GO: 0009897	External side of plasma membrane	VTCN1, SERPINE2, ABCA1, GPC4, LDLR, IL1RL1, F3, CD83
GO: 0032496	Response to lipopolysaccharide	SLPI, VIM, ABCA1, HMGB2, DCN, LIAS, THBD, CEBPB, GCH1, NFKBIA, ADM
GO: 0001655	Urogenital system development	PROM1, SFRP1, TFAP2B, CD24, SMARCC1, DCN, SMAD5, EPHA7, KLHL3, FGF1, NPHS2
GO: 0001822	Kidney development	PROM1, SFRP1, TFAP2B, CD24, DCN, SMAD5, EPHA7, KLHL3, FGF1, NPHS2
GO: 0072001	Renal system development	PROM1, SFRP1, TFAP2B, CD24, DCN, SMAD5, EPHA7, KLHL3, FGF1, NPHS2
GO: 0032103	Positive regulation of response to external stimulus	CXCL12, STX3, NUPR1, CCL4, LDLR, THBD, IL1RL1, F3, CEBPB, NFKBIA
GO: 0045861	Negative regulation of proteolysis	SLPI, WFDC2, SERPINE2, TFAP2B, SERPING1, SMARCC1, CSNK2A1, CAMLG, PLAUR, KLF4
GO: 0016049	Cell growth	SFRP1, CXCL12, SERPINE2, SORBS2, CSNK2A1, PAPPA2, TAF9B, EPHA7, IFRD1, SOCS2, SGK1, RGS2
GO: 0001558	Regulation of cell growth	SFRP1, CXCL12, SERPINE2, CSNK2A1, PAPPA2, TAF9B, EPHA7, IFRD1, SOCS2, SGK1, RGS2
GO: 0001101	Response to acid chemical	SFRP1, BCHE, UFL1, ABCA1, FABP3, LAMTOR3, LDLR, KLF4, CEBPB, CCN2
GO: 0072009	Nephron epithelium development	PROM1, TFAP2B, CD24, KLHL3, FGF1, NPHS2
GO: 0032365	Intracellular lipid transport	ANXA2, ABCA1, FABP3, LDLR
GO: 0045823	Positive regulation of heart contraction	GCH1, CCN2, ADM, RGS2
GO: 0030193	Regulation of blood coagulation	SERPINE2, SERPING1, PLAUR, THBD, F3
GO: 1900046	Regulation of hemostasis	SERPINE2, SERPING1, PLAUR, THBD, F3
GO : 0061318	Renal filtration cell differentiation	PROM1, CD24, NPHS2
GO: 0072112	Glomerular visceral epithelial cell differentiation	PROM1, CD24, NPHS2
GO: 0050818	Regulation of coagulation	SERPINE2, SERPING1, PLAUR, THBD, F3
GO: 0072311	Glomerular epithelial cell differentiation	PROM1, CD24, NPHS2
GO: 0030195	Negative regulation of blood coagulation	SERPINE2, SERPING1, PLAUR, THBD
GO: 1900047	Negative regulation of hemostasis	SERPINE2, SERPING1, PLAUR, THBD
GO : 0032373	Positive regulation of sterol transport	ANXA2, ABCA1, NFKBIA
GO: 0032376	Positive regulation of cholesterol Transport	ANXA2, ABCA1, NFKBIA
GO: 0072010	Glomerular epithelium development	PROM1, CD24, NPHS2
GO: 0052547	Regulation of peptidase activity	SLPI, WFDC2, TFAP2B, CSNK2A1, EPHA7, PLAUR, KLF4, F3, PCOLCE2, CCN2
Diabetic tubulointerstitial compartment
GO: 0044282	Small molecule catabolic process	ENTPD1, IDUA, HYAL1, TST, IVD, GLS, ENPP4, ACOX1, MTRR, MCCC2, HADH, ENO1, ENTPD5, AHCY, HNMT, ECI2, PFKFB2, EHHADH, PEX7, ALDH7A1, ETFDH, DBT, ACADSB, CRYM, ACADM, PNP, IMPA1, ACADL, HAGH, KYNU, DECR2, MMUT, FUT6, KHK, RIDA, BCKDHB, MCCC1, GLUD2, AUH, GOT1, ALDH8A1, CRYL1, FAH, CTH, ALDH3A2, DERA, QDPR, ACAA1, HGD, UPB1, ALDH1L1, ALDH1B1, GK, KMO, NUDT3, ALDH6A1, QPRT, SORD, HADHA, ABAT, ACOX2, PRODH2, NPL, DPYS, FABP1
GO: 0016054	Organic acid catabolic process	IDUA, HYAL1, TST, IVD, GLS, ACOX1, MTRR, MCCC2, HADH, AHCY, HNMT, ECI2, EHHADH, PEX7, ALDH7A1, ETFDH, DBT, ACADSB, CRYM, ACADM, ACADL, KYNU, DECR2, MMUT, RIDA, BCKDHB, MCCC1, GLUD2, AUH, GOT1, ALDH8A1, CRYL1, FAH, CTH, ALDH3A2, QDPR, ACAA1, HGD, ALDH1L1, KMO, ALDH6A1, QPRT, SORD, HADHA, ABAT, ACOX2, PRODH2, NPL, FABP1
GO: 0046395	Carboxylic acid catabolic process	IDUA, HYAL1, TST, IVD, GLS, ACOX1, MTRR, MCCC2, HADH, AHCY, HNMT, ECI2, EHHADH, PEX7, ALDH7A1, ETFDH, DBT, ACADSB, CRYM, ACADM, ACADL, KYNU, DECR2, MMUT, RIDA, BCKDHB, MCCC1, GLUD2, AUH, GOT1, ALDH8A1, CRYL1, FAH, CTH, ALDH3A2, QDPR, ACAA1, HGD, ALDH1L1, KMO, ALDH6A1, QPRT, SORD, HADHA, ABAT, ACOX2, PRODH2, NPL, FABP1
GO: 0050662	Coenzyme binding	QSOX1, CYBB, MICAL2, IVD, IDH3G, ACOX1, MTRR, HADH, IDH2, AHCY, ECI2, GRHPR, PNPLA3, CAT, ETFDH, ACADSB, IDH3B, CRYM, MAOB, ACADM, ACADL, KYNU, AIFM1, P4HA2, CRYZL1, MARC2, MCCC1, GPHN, GOT1, CRYL1, HPGD, NOX4, CTH, IDH3A, DDC, AOX1, ALDH1B1, NQO2, KMO, ALDH6A1, SORD, GPD1, HADHA, ABAT, ACOT7, ACOX2, DHTKD1
GO: 0048037	Cofactor binding	LCN2, QSOX1, CYBB, PTGES, MICAL2, IVD, IDH3G, FDX1, ACLY, ACOX1, ACO2, MTRR, HADH, IDH2, AHCY, ECI2, GRHPR, PNPLA3, CAT, ETFDH, ACADSB, IDH3B, CRYM, MAOB, CYP11B2, ACADM, CYB5A, CYP20A1, ACADL, KYNU, AIFM1, P4HA2, MMUT, CRYZL1, ACO1, MARC2, MCCC1, GPHN, GOT1, CRYL1, HPGD, NOX4, CTH, SDHC, SDHB, IDH3A, DDC, AOX1, ALDH1B1, NQO2, KMO, ALDH6A1, GSTM4, SORD, GPD1, HADHA, ABAT, ACOT7, ACOX2, CUBN, DHTKD1, AOC1
GO: 0006520	Cellular amino acid metabolic process	TST, IVD, GLS, MTRR, MCCC2, AHCY, HNMT, NARS2, PSMA3, PSMD1, PSMF1, ALDH7A1, DBT, PSMD12, ACADSB, CRYM, PEPD, ADI1, EPRS, KYNU, MMUT, PSMD11, RIDA, BCKDHB, MCCC1, GLUD2, AUH, PSMC2, GOT1, NR1H4, ALDH8A1, NOX4, FAH, CTH, GLYAT, ASS1, DDC, BPHL, QDPR, HGD, UPB1, KMO, ALDH6A1, GCLM, DPEP1, ABAT, FOLH1B, BHMT2, PRODH2, PSAT1, DPYS, SLC7A7
GO: 0009063	Cellular amino acid catabolic process	TST, IVD, GLS, MTRR, MCCC2, AHCY, HNMT, ALDH7A1, DBT, ACADSB, CRYM, KYNU, RIDA, BCKDHB, MCCC1, GLUD2, AUH, GOT1, ALDH8A1, FAH, CTH, QDPR, HGD, KMO, ALDH6A1, ABAT, PRODH2
GO: 0010038	Response to metal ion	ITPR3, CAV1, CALR, FABP4, CYBB, B2M, MEF2C, LOXL2, HNRNPD, PTGES, ABCC2, HOMER1, NEDD4L, CREB1, PRKAA2, SLC30A5, SLC11A2, BNIP3, CAT, MT1E, MAOB, MT1X, CYP11B2, CYB5A, MT1H, ALAD, MT1G, MT1F, LGMN, ACO1, GPHN, GOT1, EGFR, GGH, ASS1, QDPR, PRNP, CLIC4, DPEP1, CA2, SORD, CPNE3, ABAT, AOC1
GO: 1901605	Alpha-amino acid metabolic process	IVD, GLS, MTRR, MCCC2, AHCY, HNMT, ALDH7A1, CRYM, ADI1, KYNU, MMUT, RIDA, MCCC1, GLUD2, AUH, GOT1, NR1H4, ALDH8A1, NOX4, FAH, CTH, GLYAT, ASS1, QDPR, HGD, KMO, ALDH6A1, GCLM, DPEP1, BHMT2, PRODH2, PSAT1
GO: 0005777	Peroxisome	ACSL4, PEX19, ACOX1, ACSL3, ISOC1, IDH2, ECI2, GRHPR, PLAAT3, EHHADH, PEX7, CAT, ACSL1, CRYM, DECR2, MARC2, RIDA, ALDH3A2, EPHX2, ACAA1, HSDL2, ACOX2, AOC1, FABP1
GO: 0042579	Microbody	ACSL4, PEX19, ACOX1, ACSL3, ISOC1, IDH2, ECI2, GRHPR, PLAAT3, EHHADH, PEX7, CAT, ACSL1, CRYM, DECR2, MARC2, RIDA, ALDH3A2, EPHX2, ACAA1, HSDL2, ACOX2, AOC1, FABP1
GO: 0005759	Mitochondrial matrix	ERBB4, ISCA1, TST, GRSF1, IVD, IDH3G, GLS, PDP1, FDX1, HSPD1, ACO2, MCCC2, HADH, IDH2, CREB1, NARS2, ACSS3, MRPL42, GRPEL1, ALDH7A1, ETFDH, DBT, MRPL22, MAIP1, ACADSB, IDH3B, VDAC1, ACADM, ACADL, LRPPRC, HAGH, MMUT, RIDA, BCKDHB, MCCC1, MRPS18B, AUH, FOXO3, GLYAT, IDH3A, DLAT, ALDH1B1, ALDH6A1, LACTB2, NR3C1, ACSF2, HADHA, ABAT, DHTKD1
GO: 1901606	Alpha-amino acid catabolic process	IVD, GLS, MTRR, MCCC2, AHCY, HNMT, ALDH7A1, CRYM, KYNU, RIDA, MCCC1, GLUD2, AUH, GOT1, ALDH8A1, FAH, QDPR, HGD, KMO, ALDH6A1, PRODH2
GO: 0006732	Coenzyme metabolic process	NNMT, PFKFB3, ACSL4, NUP214, FOLR2, CD38, NUP50, PDP1, ACLY, ACSL3, MTRR, MCCC2, IDH2, ENO1, ENTPD5, PFKFB2, MAT2A, PRKAA2, RFK, ZBTB7A, ACSL1, STAT3, PNP, CYB5A, KYNU, MOCS2, MCCC1, GPHN, AMD1, GLYAT, DERA, DLAT, ALDH1L1, KMO, ACOT13, ACSF2, QPRT, GPD1, ACOT7, AASDHPPT, BHMT2, DHTKD1, INSR
GO: 0006790	Sulfur compound metabolic process	ACPP, ACSL4, CHST15, IDUA, BGN, MICAL2, HYAL1, TST, PDP1, ABCC2, ACLY, CIAO1, ACSL3, ARSB, ST3GAL1, MTRR, MCCC2, SULT1C2, AHCY, PAX8, MAT2A, ACSL1, ADI1, HAGH, KYNU, MMUT, MCCC1, NOX4, CTH, AMD1, GLYAT, DLAT, ACOT13, GNS, GSTM4, GCLM, ACSF2, DPEP1, ACOT7, BHMT2, SLC35D1
GO: 0002283	Neutrophil activation involved in immune response	LTF, C3, LCN2, ACPP, CD93, CFD, QSOX1, RNASET2, CYBB, B2M, LAIR1, RAB31, FCER1G, FCN1, MNDA, DNASE1L3, ATP6V1D, PGRMC1, ENPP4, RAB5C, ACLY, ARSB, CTSA, ILF2, CPPED1, DEGS1, PSMD1, CAT, IGF2R, CTSB, PSMD12, HMOX2, PNP, RAB14, ALAD, PSMD11, NRAS, PSMC2, METTL7A, RAB7A, GGH, LAMP2, DERA, MAGT1, ACAA1, SLC2A5, CMTM6, IQGAP2, GNS, CPNE3, AOC1
GO: 0002446	Neutrophil mediated immunity	LTF, C3, LCN2, ACPP, CD93, CFD, QSOX1, RNASET2, CYBB, B2M, LAIR1, RAB31, FCER1G, FCN1, MNDA, DNASE1L3, ATP6V1D, PGRMC1, ENPP4, RAB5C, ACLY, ARSB, CTSA, ILF2, CPPED1, DEGS1, PSMD1, CAT, IGF2R, CTSB, PSMD12, HMOX2, PNP, RAB14, ALAD, PSMD11, NRAS, PSMC2, METTL7A, RAB7A, GGH, LAMP2, DERA, MAGT1, ACAA1, SLC2A5, CMTM6, IQGAP2, GNS, CPNE3, AOC1
GO: 0042119	Neutrophil activation	LTF, C3, LCN2, ACPP, CD93, CFD, QSOX1, RNASET2, CYBB, B2M, LAIR1, RAB31, FCER1G, FCN1, MNDA, DNASE1L3, ATP6V1D, PGRMC1, ENPP4, RAB5C, ACLY, ARSB, CTSA, ILF2, CPPED1, DEGS1, PSMD1, CAT, IGF2R, CTSB, PSMD12, HMOX2, PNP, RAB14, ALAD, PSMD11, NRAS, PSMC2, METTL7A, RAB7A, GGH, LAMP2, DERA, MAGT1, ACAA1, SLC2A5, CMTM6, IQGAP2, GNS, CPNE3, AOC1
GO: 0043312	Neutrophil degranulation	LTF, C3, LCN2, ACPP, CD93, CFD, QSOX1, RNASET2, CYBB, B2M, LAIR1, RAB31, FCER1G, FCN1, MNDA, ATP6V1D, PGRMC1, ENPP4, RAB5C, ACLY, ARSB, CTSA, ILF2, CPPED1, DEGS1, PSMD1, CAT, IGF2R, CTSB, PSMD12, HMOX2, PNP, RAB14, ALAD, PSMD11, NRAS, PSMC2, METTL7A, RAB7A, GGH, LAMP2, DERA, MAGT1, ACAA1, SLC2A5, CMTM6, IQGAP2, GNS, CPNE3, AOC1
GO: 0051287	Nicotinamide adenine dinucleotide (NAD) binding	IDH3G, HADH, IDH2, AHCY, GRHPR, IDH3B, CRYL1, HPGD, IDH3A, AOX1, ALDH1B1, SORD, GPD1, HADHA
GO: 0050660	Flavin adenine dinucleotide binding	QSOX1, CYBB, MICAL2, IVD, ACOX1, MTRR, ETFDH, ACADSB, MAOB, ACADM, AIFM1, NOX4, AOX1, NQO2, KMO, ACOX2
GO: 0006631	Fatty acid metabolic process	C3, ACSL4, CAV1, CD74, PTGES, RGN, IVD, PDP1, ACOX1, ACSL3, HADH, ECI2, PRKAA2, EHHADH, DEGS1, PNPLA3, PEX7, ETFDH, ACSL1, ACADSB, ACADM, ACADL, DECR2, MMUT, LYPLA1, CRYL1, HPGD, CES2, ALDH3A2, EPHX2, ACAA1, MSMO1, GSTM4, ACSF2, HADHA, ACOT7, ACOX2, QKI, FABP1
GO: 0043202	Lysosomal lumen	AGRN, RNASET2, CD74, IDUA, BGN, SGSH, HYAL1, CTSL, ARSB, CTSA, CTSB, TPP1, LGMN, GPC3, LAMP2, GNS, CUBN, CTSV
GO: 0016999	Antibiotic metabolic process	CYBB, IDH3G, ABCC2, ADH5, ACLY, ST3GAL1, ACO2, IDH2, GPX3, CAT, STAT3, IDH3B, MAOB, ACO1, EGFR, SDHC, SDHB, IDH3A, DLAT, ALDH1B1, DPEP1, NPL
GO: 0044438	Microbody part	ACSL4, PEX19, ACOX1, ACSL3, ECI2, GRHPR, PLAAT3, EHHADH, PEX7, CAT, ACSL1, CRYM, DECR2, ALDH3A2, EPHX2, ACAA1, ACOX2, FABP1
GO: 0044439	Peroxisomal part	ACSL4, PEX19, ACOX1, ACSL3, ECI2, GRHPR, PLAAT3, EHHADH, PEX7, CAT, ACSL1, CRYM, DECR2, ALDH3A2, EPHX2, ACAA1, ACOX2, FABP1
GO: 0045177	Apical part of cell	ANXA1, ITPR3, OXTR, PLAT, ABCC2, HOMER1, USH1C, SORBS2, SLC30A5, SLC11A2, PDZK1, SLC3A2, LRP2, ATP6V1C1, LGMN, ERBB3, STX3, NOX4, ABCC6, EZR, EGFR, SLC2A5, VAMP3, CLCN5, TUBG1, CLIC4, SLC22A4, SLC17A1, DPEP1, CA2, PRKCI, KL, SLC7A9, THY1, PODXL, CUBN, MTDH, CTSV, FABP1
GO: 0031983	Vesicle lumen	LTF, C3, LCN2, CFD, QSOX1, CALR, RNASET2, B2M, TOR4A, HGF, FCN1, MNDA, VEGFC, SERPINF2, ACLY, ARSB, CTSA, ILF2, CPPED1, PSMD1, CAT, PSMD12, PNP, ALAD, PSMD11, VEGFA, PSMC2, EGFR, GGH, DERA, ACAA1, KNG1, GNS, APOE, APOH, PLG, AOC1
GO: 0072350	Tricarboxylic acid metabolic process	IDH3G, ACLY, ACO2, IDH2, IDH3B, ACO1, SDHC, SDHB, IDH3A, ASS1, DLAT
GO: 0005775	Vacuolar lumen	C3, AGRN, RNASET2, CD74, IDUA, BGN, MNDA, SGSH, HYAL1, CTSL, ACLY, ARSB, CTSA, CPPED1, PSMD1, CTSB, TPP1, LGMN, GPC3, GGH, LAMP2, GNS, CUBN, CTSV

In comparison, many more enriched terms were identified in the tubulointerstitial DEG set of DN (Supplementary data, Table S3; N = 467, MF n = 32, CC n = 65, BP n = 370). Dysregulation of a wide range of small molecule catabolic processes, including organic acid, carboxylic acid, amino acid and fatty acid, was suggested to be involved in DN (Figure 4). Moreover, neutrophil-mediated immunity, oxidative stress, aging and platelet activation might also participate in the development of DN (Figure 4). We also performed GO enrichment analysis in upregulated and downregulated DEG sets separately. The results can be found in the Supplementary data, Table S4.

Pathway enrichment analysis

Two databases (KEGG pathway, Reactome) were used for pathway enrichment analysis. After analyzing the whole DEG set, there is no evidence of an enriched term within the glomerular compartment, and the same holds true for the upregulated and downregulated DEG sets.

In contrast, 75 terms were found enriched in the DEG set of the tubulointerstitial compartment (Supplementary data, Table S5; KEGG pathway n = 27, Reactome pathway n = 48). Similar to the results of the GO enrichment analysis, pathway enrichment analysis revealed a wide range of metabolism disorders. Notably, valine, leucine and isoleucine degradation was identified as the term with the highest enrichment (q value = 3.86E-11). Other enriched terms include the peroxisome proliferator-activated receptor (PPAR) signaling pathway, epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor resistance, FoxO signaling pathway, neutrophil degranulation and autophagy (Figures 5 and 6). Moreover, the upregulated genes were mainly related to hemostasis, signaling by receptor tyrosine kinases, platelet activation and extracellular matrix organization (Figure 5 and Supplementary data, Table S5). The downregulated genes, not surprisingly, were mainly associated with metabolism disorders, FoxO signaling pathway, PPAR signaling pathway, etc., most of which can also be found in the list of the whole set enrichment terms (Figure 5 and Supplementary data, Table S5). The enrichment map for the KEGG enrichment analysis of the diabetic tubulointerstitium further revealed the inner relationship between different enriched pathways (Figure 6). Most metabolism-related pathways have shared genes and constitute a pathological network. It is also noteworthy that ferroptosis might also contribute to renal tubular cell death as well as autophagy.

FIGURE 5:

Significantly enriched pathways for (A) dysregulated DEGs, (B) upregulated DEGs and (C) downregulated DEGs in the diabetic tubulointerstitium.

FIGURE 6:

Enrichment map for the KEGG enrichment analysis of the diabetic tubulointerstitium. Overlapping gene sets tend to cluster together by edges.

PIN reconstruction and hub gene identification

To further reveal the relationship of the DEGs, PINs with physical interaction were reconstructed (Figure 7). Each edge represents a physical interaction established by a given method, such as mass spectrometry or two-hybrid assay. While calculating topological scores, the edges between two nodes were regarded as one. Compared with those of the glomerular compartment, tubulointerstitial DEGs formed a much more complex network (glomerular PIN: 36 nodes, 30 edges; tubulointerstitial PIN: 597 nodes, 1870 edges). As presented in Tables 3 and 4, nodes with the highest degrees were recognized as hub genes for each network. Hub genes of the glomerular PIN include AGR2, VIM, CSNK2A1, ANXA2 and TERF1, whereas 10 genes, including EGFR, LMNA and HSPD, were identified as hub genes in the tubulointerstitial PIN.

FIGURE 7:

PIN reconstruction of the diabetic (A) glomerulus and (B) tubulointerstitium. Red, upregulated nodes; green, downregulated nodes.

Table 3.

Hub genes identified in the glomerular PIN

Rank	Gene symbol	Full name	Description	Degrees
1	AGR2	Anterior gradient 2, protein disulfide isomerase family member	This gene encodes a member of the disulfide isomerase family of ER proteins that catalyze protein folding and thiol-disulfide interchange reactions. The encoded protein has an N-terminal ER-signal sequence, a catalytically active thioredoxin domain and a C-terminal ER-retention sequence. This protein plays a role in cell migration, cellular transformation and metastasis and is as a p53 inhibitor. As an ER-localized molecular chaperone, it plays a role in the folding, trafficking and assembly of cysteine-rich transmembrane receptors and the cysteine-rich intestinal gylcoprotein mucin. This gene has been implicated in inflammatory bowel disease and cancer progression	10
2	VIM	Vimentin	This gene encodes a type III intermediate filament protein. Intermediate filaments, along with microtubules and actin microfilaments, make up the cytoskeleton. The encoded protein is responsible for maintaining cell shape and integrity of the cytoplasm and stabilizing cytoskeletal interactions. This protein is involved in neuritogenesis and cholesterol transport and functions as an organizer of a number of other critical proteins involved in cell attachment, migration and signaling. Bacterial and viral pathogens have been shown to attach to this protein on the host cell surface. Mutations in this gene are associated with congenital cataracts in human patients	4
3	CSNK2A1	Casein kinase 2 alpha 1	Casein kinase II is a serine/threonine-protein kinase that phosphorylates acidic proteins such as casein. It is involved in various cellular processes, including cell cycle control, apoptosis and circadian rhythm	4
4	ANXA2	Annexin A2	This gene encodes a member of the annexin family. Members of this calcium-dependent phospholipid-binding protein family play a role in the regulation of cellular growth and in signal transduction pathways. This protein functions as an autocrine factor, which heightens osteoclast formation and bone resorption	4
5	TERF1	Telomeric repeat binding factor 1	This gene encodes a telomere- specific protein that is a component of the telomere nucleoprotein complex. This protein is present at telomeres throughout the cell cycle and functions as an inhibitor of telomerase, acting in cis to limit the elongation of individual chromosome ends	3

ER: endoplasmic reticulum. Descriptions of the genes were abstracted from the RefSeq database.

Table 4.

Hub genes identified in the tubulointerstitial PIN

Rank	Gene symbol	Full name	Description	Degrees
1	EGFR	Epidermal growth factor receptor	The protein encoded by this gene is a transmembrane glycoprotein that is a member of the protein kinase superfamily. This protein is a receptor for members of the epidermal growth factor family. EGFR is a cell surface protein that binds to epidermal growth factor. Binding of the protein to a ligand induces receptor dimerization and tyrosine autophosphorylation and leads to cell proliferation	99
2	LMNA	Lamin A/C	The nuclear lamina consists of a two-dimensional matrix of proteins located next to the inner nuclear membrane. The lamin family of proteins make up the matrix and are highly conserved in evolution. During mitosis, the lamina matrix is reversibly disassembled as the lamin proteins are phosphorylated. Lamin proteins are thought to be involved in nuclear stability, chromatin structure and gene expression. Vertebrate lamins consist of two types, A and B. Alternative splicing results in multiple transcript variants. Mutations in this gene lead to several diseases: Emery–Dreifuss muscular dystrophy, familial partial lipodystrophy, limb–girdle muscular dystrophy, dilated cardiomyopathy, Charcot–Marie–Tooth disease and Hutchinson–Gilford progeria syndrome	48
3	NRAS	NRAS proto-oncogene, GTPase	This is an N-ras oncogene encoding a membrane protein that shuttles between the Golgi apparatus and the plasma membrane. This shuttling is regulated through palmitoylation and depalmitoylation by the ZDHHC9-GOLGA7 complex. The encoded protein, which has intrinsic GTPase activity, is activated by a guanine nucleotide-exchange factor and inactivated by a GTPase activating protein. Mutations in this gene have been associated with somatic rectal cancer, follicular thyroid cancer, autoimmune lymphoproliferative syndrome, Noonan syndrome and juvenile myelomonocytic leukemia	47
4	HSPD1	Heat shock protein family D (Hsp60) member 1	This gene encodes a member of the chaperonin family. The encoded mitochondrial protein may function as a signaling molecule in the innate immune system. This protein is essential for the folding and assembly of newly imported proteins in the mitochondria. This gene is adjacent to a related family member and the region between the two genes functions as a bidirectional promoter. Several pseudogenes have been associated with this gene. Two transcript variants encoding the same protein have been identified for this gene. Mutations associated with this gene cause autosomal recessive spastic paraplegia 13	46
5	HEXIM1	HEXIM P-TEFb complex subunit 1	Expression of this gene is induced by hexamethylene-bis-acetamide in vascular smooth muscle cells	43
6	TGOLN2	Trans-golgi network protein 2	This gene encodes a type I integral membrane protein that is localized to the trans-Golgi network, a major sorting station for secretory and membrane proteins. The encoded protein cycles between early endosomes and the trans-Golgi network and may play a role in exocytic vesicle formation	42
7	MAPK6	Mitogen-activated protein kinase 6	The protein encoded by this gene is a member of the Ser/Thr protein kinase family and is most closely related to MAP kinases. MAP kinases also known as extracellular signal-regulated kinases, are activated through protein phosphorylation cascades and act as integration points for multiple biochemical signals. This kinase is localized in the nucleus and has been reported to be activated in fibroblasts upon treatment with serum or phorbol esters	40
8	RC3H2	Ring finger and C-x8-C-x5-C-x3-H (CCCH)-type domains 2	RC3H2 is a protein coding gene. GO annotations related to this gene include mRNA binding	36
9	CHD3	Chromodomain helicase DNA binding protein 3	This gene encodes a member of the CHD family of proteins which are characterized by the presence of chromo (chromatin organization modifier) domains and SNF2-related helicase/ATPase domains. This protein is one of the components of a histone deacetylase complex referred to as the Mi-2/nucleosome remodeling and deacetylase(NuRD) complex which participates in the remodeling of chromatin by deacetylating histones. Chromatin remodeling is essential for many processes including transcription. Autoantibodies against this protein are found in a subset of patients with dermatomyositis	35
10	SUZ12	SUZ12 polycomb repressive complex 2 subunit	This zinc finger gene has been identified at the breakpoints of a recurrent chromosomal translocation reported in endometrial stromal sarcoma. Recombination of these breakpoints results in the fusion of this gene and juxtaposed with another zinc finger protein 1 (JAZF1)	34

Descriptions of the genes were abstracted from the RefSeq database. As no summary was available for RC3H2 in Refseq, the GeneCards summary was used instead.

DISCUSSION

DN is the leading cause of end-stage kidney disease worldwide. Its pathogenesis is distinct from other CKDs such as GN and HN. In this study we reported a unique transcriptomic feature of DN based on bioinformatics analysis of >250 microarray datasets. To our knowledge, this is the largest comparative bioinformatics study on this topic.

Our study found more DEGs in the diabetic tubulointerstitium than in the glomerulus. Moreover, a larger proportion of DEGs were DN-specific in the tubulointerstitium (68.0% versus 30.9%). As a result, the number of unique DEGs identified in diabetic tubulointerstitium was 6.59 times as many as in the glomerulus, despite the fact that more tubulointerstitial samples were included in the analysis. In addition, the tubular DEG formed a much more complex PIN. These data provide compelling evidence for the presence of diabetic tubulopathy and support the tubulocentric view of DN.

According to the GO enrichment analysis, the unique DEGs of the glomerular compartment were related to the extracellular matrix, cell growth, regulation of blood coagulation, cholesterol homeostasis, intrinsic apoptotic signaling pathway and renal filtration cell differentiation. However, we did not find enriched terms in either the Reactome or KEGG pathway database using the cluster profile package in R, which might be due to the conservative algorithm design (fewer background genes in hypergeometric test) of this package. Consistent with the notion that glomerulosclerosis is a key pathological feature of DN, we observed that the extracellular matrix was the most significantly enriched GO term in the diabetic glomerulus. Further pathway analysis revealed that the related DEGs were enriched in the wnt signaling pathway, the formation of fibrin clot (clotting cascade) and diseases associated with glycosaminoglycan metabolism. Dysregulation of glomerular wnt/β-catenin signaling has been documented to increase the deposition of fibrous tissue in DN, despite the fact that its mechanism remains to be elucidated [15]. Interestingly, most of the enriched terms (extracellular matrix, cell growth, regulation of blood coagulation and apoptosis) fell into the broader BPs of wound healing. A recent bioinformatics analysis using urinary proteomics also showed that wound healing is a hallmark in both uncomplicated diabetes and incipient DN [16]. Experimental studies have shown that activation of coagulation, one of the early processes in wound healing, can facilitate immune cell infiltration, impair glomerular permselectivity and induce apoptosis of podocytes and endothelial cells. On the other hand, chronic injury induces progressive accumulation of extracellular matrix and improper tissue development, which contributes to the development of renal lesions in DN. Moreover, our analysis also observed a disturbance of cholesterol homeostasis pathways in the diabetic glomerulus. Interestingly, the altered expression of related DEGs, including ABCA1 and low-density lipoprotein receptor (LDLR), has been shown to promote cholesterol accumulation by affecting the efflux and influx of LDL, which aggravates glomerulopathy in DN [17, 18]. We also identified five hub genes of the glomerular PIN, including AGR2, VIM, CSNK2A1, ANXA2 and TERF1. Prior work has documented some of their roles in the development of DN. For example, Huang et al. [19] reported that CK2α (encoded by CSNK2As) can regulate the process of fibrosis in the diabetic kidney via the nuclear factor κB (NF-κB) pathway. In addition, a recent study reported a significant upregulation of AGR2, Tff2 and Gkn3 in the diabetic mouse model after administration of dapagliflozin [20]. However, their roles in the development of diabetic tubulopathy remain to be elucidated.

In the diabetic tubulointerstitium, the significantly enriched BPs and pathways included metabolism, the advanced glycation end products–receptor for advanced glycation end products signaling pathway in diabetic complications, the EGFR signaling pathway, the FoxO signaling pathway, autophagy and ferroptosis. In contrast to the diabetic glomerulus, a wide range of metabolism disorders emerged as the top pathological processes in diabetic tubulopathy. It is widely accepted that diabetes is associated with altered cellular metabolism. By analyzing diabetic kidney transcriptomic data from a cohort of southwestern American Indians, Sas et al. [21] reported dysregulation of glycolysis, fatty acid and amino acid metabolism pathways in the human kidney. Our analysis highlighted that these changes are characteristic in diabetic tubulointerstitium, and the PPAR signaling pathway might play important roles in regulating renal tubular metabolism (especially in fatty acid metabolism) in DN. We observed a significant dysregulation of downstream target genes of PPARs that participate in the maintenance of renal metabolic homeostases, such as glycerol kinase (GK), 3-hydroxy-3-methylglutaryl-coenzyme A synthase 2 (HMGCS2) and acyl-coenzyme A dehydrogenase long chain (ACADL). We have also observed a significant downregulation of the pathway for branched-chain amino acids (BCAAs; including valine, leucine and isoleucine) degradation. Urinary BCAA catabolites are decreased in diabetic subjects [22]. Elevations of plasm BCAAs are independently associated with an increased risk of type 2 diabetes and insulin resistance [23]. Whether urinary BCAAs could be used as an early biomarker of diabetic tubulopathy remains to be investigated.

Our study also showed ferroptosis might be involved in renal tubular cell death in DN. Ferroptosis is an iron-dependent form of cell death resulting from lipid-based reactive oxygen species, which has implications in tumors, nervous system diseases and kidney diseases such as polycystic kidney disease and acute kidney injury [24]. A recent study reported that inhibition of ferroptosis can reduce myocardial ischemia–reperfusion injury in diabetic models [25]. Considering there is still a lack of literature, it is noteworthy for future studies to investigate the role of ferroptosis in DN.

Ten genes with the highest degrees, including EGFR, LMNA and HSPD1, were identified as hub genes in the tubulointerstitial PIN. Interestingly, some of these genes, such as EGFR and HSPD1, have been gaining increasing research attention in recent years [26, 27]. Using global protein network analysis and subsequent functional validation, Aluksanasuwan et al. [27] reported HSPD1 was involved in diabetic tubular cell dysfunction probably via regulation of intracellular adenosine triphosphate (ATP) production. Here we only discussed several terms or genes according to their importance given by statistics tests. However, it is rather common that those with less statistical significance have more important roles in real circumstances. Therefore readers are encouraged to explore the supplementary materials for further information.

However, some limitations are worth noting. First, although >250 datasets were included, only transcriptomics data were analyzed in our study. Future work should therefore include proteomic and metabolomic data to provide a more comprehensive understanding of the molecular mechanism. Second, expression of mRNAs, rather than proteins, was used to reconstruct PINs, which might not accurately reflect the real situation. Third, to increase the specificity of unique DEG identification, the union but not the intersection operation was used to generate the DEG set of GN. Despite its logical reasonability, this operation might miss some potential genes that play important roles in DN. Last but not least, several resources of heterogeneity, including different genetic and clinical backgrounds of the study populations, might also introduce bias to our results.

In conclusion, our study not only reveals the unique molecular mechanism of DN, but also provides a valuable resource for biomarker and therapeutic target discovery. Some of our findings are promising and should be explored by future work.

SUPPLEMENTARY DATA

Supplementary data are available at ckj online.

FUNDING

Funding was provided by the National Natural Science Foundation of China (Youth Program 81900698). This project was supported by the Scientific Research Projects of Wuxi Municipal Health and Family Planning Commission (Q201761).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.