Abstract

   MicroRNAs (miRNAs) play a great contribution to the development of
   diabetic nephropathy (DN). The aim of this study was to explore
   potential miRNAs-genes regulatory network and biomarkers for the
   pathogenesis of DN using bioinformatics methods.

   Gene expression profiling data related to DN ([37]GSE1009) was obtained
   from the Gene Expression Omnibus (GEO) database, and then
   differentially expressed genes (DEGs) between DN patients and normal
   individuals were screened using GEO2R, followed by a series of
   bioinformatics analyses, including identifying key genes, conducting
   pathway enrichment analysis, predicting and identifying key miRNAs, and
   establishing regulatory relationships between key miRNAs and their
   target genes.

   A total of 600 DEGs associated with DN were identified. An additional 7
   key DEGs, including 6 downregulated genes, such as vascular endothelial
   growth factor α (VEGFA) and COL4A5, and 1 upregulated gene (CCL19),
   were identified in another dataset ([38]GSE30528) from glomeruli
   samples. Pathway analysis showed that the down- and upregulated DEGs
   were enriched in 14 and 6 pathways, respectively, with 7 key genes
   mainly involved in extracellular matrix–receptor interaction, PI3K/Akt
   signaling, focal adhesion, and Rap1 signaling. The relationships
   between miRNAs and target genes were constructed, showing that miR-29
   targeted COL4A and VEGFA, miR-200 targeted VEGFA, miR-25 targeted
   ITGAV, and miR-27 targeted EGFR.

   MiR-29 and miR-200 may play important roles in DN. VEGFA and COL4A5
   were targeted by miR-29 and VEGFA by miR-200, which may mediate
   multiple signaling pathways leading to the pathogenesis and development
   of DN.

   Keywords: bioinformatics, diabetic nephropathy, differentially
   expressed genes, microRNAs

1. Indroduction

   Diabetic nephropathy (DN) is one of the most common and severe
   microvascular complications of diabetes. Approximately, 30% to 40% of
   all patients with diabetes develop DN.^[[39]1] DN has become the main
   cause of end-stage renal disease in developed countries.^[[40]2] As the
   lack of effective treatment for DN, expensive medical cost status is
   persisting across the globe. From 1990 to 2012, the number of deaths
   caused by DN increased by 94%.^[[41]3] Therefore, investigating the
   pathogenic mechanisms of DN and early screening of patients with DN are
   important to prevent the occurrence and development of DN.

   The role of miRNAs in the physiological mechanisms and therapeutic
   interventions of DN have received increasing attention.^[[42]4] The
   accumulating data and new discoveries are expected to generate new
   insight into the complex pathogenesis of DN.^[[43]5,[44]6] Regulating
   miRNAs expression could become a very attractive measure for preventing
   DN. Although a number of miRNAs have recently been identified related
   to DN,^[[45]7] more efforts are needed to discover new targets for DN
   treatment and to validate these targets through in vitro and in vivo
   experiments and large-scale population studies.

   The recent rise of bioinformatics has shed new light on studies of
   diabetic kidney disease. Researchers have used microarray data from
   mouse DN models in the Gene Expression Omnibus (GEO) database to
   analyze molecular mechanisms and genetic factors related to DN.^[[46]8]
   Previous bioinformatics analyses using human DN-related gene chip data
   ([47]GSE1009) from the GEO database found that the CXCR2, CASR, CSF2,
   VEGFA, and CD2AP genes played important roles in maintaining kidney
   functions in DN and that miR-1 targeting VEGFA and miR-25 targeting
   CD2AP might affect the occurrence and development of DN.^[[48]9] Cui et
   al^[[49]10] identified several important genes associated with DN using
   [50]GSE30122 chip data, including CD44, FN1, CCL5, and CXCR4; these
   genes were mainly enriched in the extracellular matrix (ECM)-receptor
   and cytokine–cytokine receptor interaction pathways. At the same time,
   3 potential miRNA (miR-17-5p, miR-20a, and miR-106a) were discovered.
   However, the regulatory network involving these miRNA-DEGs and the
   related signaling pathways remain unknown.

   In the present study, the differentially expressed genes (DEGs) between
   DN glomerular tissue and normal glomerular tissue were screened, and
   the results were combined with the data from the [51]GSE30122
   microarray data about glomerular tissue samples. Notably, the
   regulatory relationship between miRNAs and genes were constructed and
   validated using 5 target gene prediction databases. In addition, the
   regulatory network of miRNAs and target genes were connected with their
   corresponding signaling pathways to reveal potential regulatory
   mechanisms of DN pathogenesis and to promote the discovery of novel DN
   markers and therapeutic targets.

2. Methods

2.1. Acquisition of microarray data

   The gene expression profiles associated with DN were obtained from the
   NCBI GEO database ([52]https://www.ncbi.nlm.nih.gov/geo/). The
   following search terms were used: “diabetic nephropathy or diabetic
   kidney disease,” “Homo sapiens,” and “Expression profiling by array.”
   After careful screening, the microarray data ([53]GSE1009) submitted by
   Baelde et al^[[54]11] were selected for subsequent analysis. The
   experimental platform [55]GPL8300 used in the above study was the
   Affymetrix HG_U95Av2 Array (Affymetrix GeneChip U95 Version 2 Array).
   The dataset included a total of 6 samples: 3 DN glomerular tissue
   samples and 3 normal glomerular tissue samples. Another set of
   microarray data related to diabetic kidney disease (DKD) ([56]GSE30122)
   contained 69 samples,^[[57]12] including 9 DKD glomerular tissue
   samples, 26 normal control glomerular tissue samples, 10 DKD renal
   tubular tissue samples, and 24 normal control renal tubular tissue
   samples, of which 35 glomerular tissue samples were selected for
   analysis. Ethical approval was not necessary in this study because the
   gene expression profiles were downloaded from GEO, a public database,
   and we do not conduct new experiments in patients or animals.

2.2. Screening of DEGs

   The online analysis tool GEO2R^[[58]13] was used to preprocess and
   analyze the microarray data and screen for DEGs associated with DN. The
   screening conditions were P value <.01 and |logFC| ≥1.5.

2.3. Identification of key DEGs

   To analyze the interactions among the encoded proteins, the Search Tool
   for the Retrieval of Interacting Genes (STRING10.5)^[[59]14] database
   was used to construct a protein–protein interaction network. The
   CytoHubba plugin^[[60]15] was installed in the visualization software
   Cytoscape,^[[61]16] and the hubba values of each DEG were calculated
   according to the 12 topological algorithms by the CytoHubba plugin. The
   data were downloaded and sorted, and the top 15 hub nodes were
   selected. The common hub nodes calculated by ≥5 topological algorithms
   were eventually designated the hub genes. The intersections between the
   hub genes and the DEGs identified in the [62]GSE30122 microarray data
   were used to determine the key genes for DN.

2.4. Gene ontology functional enrichment and Kyoto Encyclopedia of Genes and
Genomes pathway analysis

   The gene ontology (GO) function and Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathway enrichment analyses were performed using the
   Database for Annotation, Visualization, and Integration Discovery
   (DAVID 6.8).^[[63]17] The P value <.05 and gene count ≥2 were set as
   the cutoff criteria.

2.5. Prediction of key miRNAs

   Web-based Gene Set Analysis Toolkit (WebGestalt)^[[64]18] online
   datebase was used to predict miRNAs that targeted the DEGs associated
   with DN. The differentially expressed miRNAs were selected according to
   the screening criterion of P value <.01. The results were exported into
   the Cytoscape software for analysis. The miRNA–protein interaction
   networks were constructed through network topology characteristics. The
   node degree was calculated using the Network analysis plugin, and
   miRNAs with a node degree >12 were selected as the key miRNAs.

2.6. Exploration of regulatory relationships between the key miRNAs and their
target genes

   The regulatory relationships between the key miRNAs and their target
   genes of DN were established based on Cytoscape and verified through 5
   miRNA target gene prediction databases, including TargetScan
   ([65]http://www.targetscan.org/vert_71/), starBase
   ([66]http://starbase.sysu.edu.cn/), miRTarBase
   ([67]http://mirtarbase.mbc.nctu.edu.tw/php/), microRNA.org
   ([68]http://34.236.212.39/microrna/home.do), and miRDB
   ([69]http://Mirdb.org/). The potential regulatory network for the
   pathogenesis and development of DN were established by combining these
   findings with the KEGG pathway analysis results.

3. Results

3.1. Identification of DEGs

   Overall, 600 DEGs were identified when DN glomerular tissue samples
   versus normal glomerular tissue samples from [70]GSE1009, of which 257
   were downregulated and 343 were upregulated. The same method was used
   to analyze and identify 301 DEGs between DKD and normal glomerular
   tissues in the [71]GSE30122 microarray data, with 253 downregulated
   genes and 48 upregulated genes.

3.2. Screening of key genes of DN

   The 15 downregulated hub genes (eg, VEGFA, EGFR, and ITGA3) and 15
   upregulated hub genes (eg, AGT, CXCR2, and CCL19) were selected (Table
   [72]1). By intercrossing with the glomerular tissue-related DEGs in
   [73]GSE30122, the 7 key genes for DN were identified, including 6
   downregulated genes (VEGFA, ITGA3, ITGB5, COL4A3, COL4A5, and CBLB) and
   1 upregulated gene (CCL19) (Fig. [74]1).

Table 1.

   Hub genes of diabetic nephropathy.

   3.2.
   [75]Open in a new tab

Figure 1.

   Figure 1
   [76]Open in a new tab

   Differentially expressed genes (DEGs) common to [77]GSE1009 and
   [78]GSE30122. (A) Venn diagram of the up- and downregulated DEGs in
   [79]GSE1009 and [80]GSE30122. (B) Venn diagram of the hub genes
   identified in GSD1009 and the DEGs in glomerular tissues identified in
   [81]GSE30122.

3.3. GO enrichment analysis of DEGs

   GO functional enrichment analysis was performed on the DEGs associated
   with DN from the aspects of molecular function (MF), cellular component
   (CC), and biological process (BP).

   The GO analysis results for the downregulated DEGs showed that most of
   these genes had MF such as protein-binding, integrin-binding and
   calcium ion-binding. These DEGs were mainly concentrated in the
   extracellular exosome, cell surface, actin cycoskeleton and plasama
   membrane, as revealed by the GO cellular component analysis. GO
   biological process analysis showed these genes were mainly involved in
   positive regulation of cell migration, lung development, glomerulus
   development, and glomerular basement membrane development (Fig.
   [82]2A).

Figure 2.

   Figure 2
   [83]Open in a new tab

   (A) GO analysis of the downregulated DEGs in DN (top 10). (B) GO
   analysis of the upregulated DEGs in DN (top 10). BP = biological
   process, CC = cellular component, DEGs = differentially expressed
   genes, DN = diabetic nephropathy, GO = gene ontology, MF = molecular
   function.

   The GO analysis results for the upregulated DEGs showed that these DEGs
   mainly had MF such as growth factor and cytokine activity,
   protease-binding, and calcium ion-binding. These DEGs were mainly
   concentrated in the extracellular region, integral component plasma
   membrane, cell junction, and axon, as revealed by the GO cellular
   component analysis. GO biological process analysis showed these genes
   were mainly involved in positive regulation of cell death and
   proliferation, protein autophosphorylation, and GTPase activity (Fig.
   [84]2B).

3.4. KEGG pathway analysis of DEGs

   KEGG pathway analysis of DEGs associated with DN revealed that the
   downregulated DEGs were involved in a total of 14 pathways (P < .05),
   including ECM–receptor interaction, cancer, focal adhesion, complement
   and coagulation cascades, PI3K/Akt and Rap1 signaling pathways. KEGG
   pathway analysis of the upregulated DEGs revealed 6 pathways (P < .05),
   including neuroactive ligand–receptor interaction, cytokine–cytokine
   receptor interaction, renin secretion, and PI3K/Akt signaling pathways
   (Table [85]2). Among these pathways, the 7 key genes were mainly
   involved in the PI3K/Akt, focal adhesion, ECM–receptor interaction, and
   Rap1 signaling pathways (Table [86]3).

Table 2.

   KEGG analysis of the differentially expressed genes.

   3.4.
   [87]Open in a new tab

Table 3.

   Key genes-related signaling pathways.

   3.4.
   [88]Open in a new tab

3.5. Prediction of miRNAs targeting DEGs

   A total of 9 groups of key miRNAs that targeted the downregulated DEGs
   were predicted: miR-506, miR-23a/b, the miR-200 family
   (miR-200b/c/429), miR-124a, the miR-181 family (miR-181a/b/c/d), the
   miR-25/32/92/363/367 cluster, the miR-17 family (miR-17-5p, miR-20a/b,
   miR-106a/b, and miR-519d), miR-27a/b, and the miR-29 family
   (miR-29a/b/c) (Fig. [89]3A). Only 2 groups of key miRNAs targeted the
   upregulated DEGs were found (the miR-17 family and the miR-200 family)
   (Fig. [90]3B).

Figure 3.

   Figure 3
   [91]Open in a new tab

   (A) MiRNA–protein interaction network between targeting miRNAs and the
   downregulated DEGs. The V-shaped nodes represent the key miRNAs (degree
   >12). The color gradient represents the degree (top 10 miRNAs). The
   round nodes represent the downregulated genes related to the key miRNAs
   or other miRNAs. (B) MiRNA–protein interaction network between
   targeting miRNAs and the upregulated DEGs. The triangle nodes represent
   the key miRNAs (degree >12). The color gradient represents the degree
   (top 10 miRNAs). The round nodes represent the upregulated genes
   related to the key miRNAs or other miRNAs. DEGs = differentially
   expressed genes, miRNAs = microRNAs.

3.6. Exploration of relationships between the key miRNAs and their target
genes

   The regulatory relationships between the miRNAs and their target genes
   were established using Cytoscape, which showed that the single gene was
   regulated by multiple miRNAs and that the single miRNA could target
   multiple genes (Fig. [92]4). Subsequently, miR-29 targeting VEGFA and
   COL4A5 and miR-200 targeting VEGFA were verified in all 5 databases. A
   total of 4 databases showed that miR-29 targeted COL4A3, the miR-17
   family targeted VEGFA, and miR-181 targeted CBLB. Only 3 databases
   showed that miR-17 family targeted COL4A3, miR-23 targeted COL4A5, and
   miR-506 targeted ITGA3, and only 2 databases showed that miR-181
   targeted ITGA3 (Table [93]4). Integrating with the result of KEGG
   pathway analysis, it was indicated that these key miRNA-target genes
   were mainly involved in the ECM–receptor interaction, PI3K/Akt
   signaling pathways, and focal adhesion to influence the pathogenesis
   and development of DN (Fig. [94]5).

Figure 4.

   Figure 4
   [95]Open in a new tab

   Global network diagram of the interactions between the key miRNAs and
   their target genes. The diamond nodes represent the 9 key miRNAs. The
   hexagonal nodes represent the 5 hub genes (ITGB5 and CCL19 did not
   correspond to the miRNAs). The large round nodes represent genes
   associated with the 9 key miRNAs, and the small round nodes represent
   the other upregulated genes and miRNAs. miRNAs = microRNAs.

Table 4.

   Verification of potential miRNAs-target genes regulatory relationships
   using 5 databases.

   3.6.
   [96]Open in a new tab

Figure 5.

   Figure 5
   [97]Open in a new tab

   Schematic diagram of potential miRNAs-genes regulatory network in DN.
   DN = diabetic nephropathy, miRNAs = microRNAs.

4. Discussion

   The recent rise of bioinformatics has shed new light on studies of
   multiple diseases including autoimmune disorders, neoplastic disease,
   and DN among others. Especially, these disease-related miRNAs were
   successfully identified. For example, the differentially expressed
   miRNAs including miR-371a-5p and let-7a-5p may be potential diagnosis
   biomarkers to illuminate pathogenic mechanisms in systemic lupus
   erythematosus^[[98]19–[99]21] and lupus nephritis.^[[100]22,[101]23] In
   addition, many miRNAs were predicted using the WebGestalt
   toolkit^[[102]24] then validated to be important in acute ischemic
   stroke.^[[103]25–[104]27] Bioinformatics analysis was also performed to
   explore the underlying functions, pathways, and networks of the
   differentially expressed miRNAs and targeted genes in various types of
   cancer such as hepatocellular carcinoma,^[[105]28] gastric
   carcinoma,^[[106]29] breast cancer,^[[107]30] prostate
   cancer,^[[108]31] and so on. Many miRNAs identified as potential
   biomarkers based on bioinformatics analysis were further confirmed by
   reverse transcription quantitative PCR and other functional
   experiments.

   The bioinformatics technology has also provided new perspectives for
   researchers to study the potential molecular mechanisms and regulatory
   targets of DN. In the present study, a total of 600 DEGs associated
   with DN were screened. VEGFA, ITGA3, ITGB5, COL4A3, COL4A5, CBLB, and
   CCL19 were identified as hub genes based on a network topology. The key
   miRNAs related to DN were also predicted, including miR-200b/c,
   miR-29a/b/c, miR-25, miR-27, miR-23, miR-181, miR-17, miR-506, and
   miR-124a. In the present study, we made an appropriate significance
   level cutoff (the screening criterion of P value <.01 and network node
   degree >12) for more detailed investigation of the data and narrowed
   down the screening range of key miRNAs based on WebGestalt software
   tool^[[109]18,[110]32] and previous screening cutoff
   point.^[[111]9,[112]24,[113]33] Strikingly, this study suggested the
   potential regulatory network of the miRNAs-genes and signaling
   pathways. For example, miR-29 targeted VEGFA and COL4A5, and miR-200
   targeted VEGFA, which mediated ECM–receptor interaction and PI3K/Akt
   signaling pathways as well as other signaling pathways to initiate the
   pathogenesis of DN.

   VEGFA is the most important vascular endothelial growth factor and
   plays an important role in angiogenesis, tumor growth, and ischemic
   diseases. The up- and downregulation of the podocyte VEGFA levels
   during renal development led to glomerular disease.^[[114]34] This
   study identified that VEGFA was a key gene that was downregulated in
   DN, consistent with previous studies. Sivaskandarajah et al^[[115]35]
   found that selective knockout of VEGFA in glomerular podocytes of
   diabetic mice resulted in a significant increase in proteinuria with
   severe glomerular scar formation and increased apoptosis. However,
   Veron et al^[[116]36] found that overexpression of the VEGFA isoform
   (VEGF164) in the podocytes of adult transgenic mice resulted in
   proteinuria, glomerular hyperplasia, glomerular basement membrane
   thickening, and podocyte disappearance, which might be related to
   different types of diabetes models or the length of studies by
   different research institutes.

   This study showed that VEGFA was regulated by multiple miRNAs,
   including miR-29 and miR-200. Previous studies have confirmed that
   VEGFA is a target of miR-29^[[117]37,[118]38] and that miR-200
   suppresses VEGFA expression.^[[119]39,[120]40] The loss of miR-29a/b is
   thought to be closely related to progressive kidney damage in diabetes.
   MiR-29a and miR-29b promote resistance to renal fibrosis and prevent
   podocyte apoptosis.^[[121]41,[122]42] MiR-29c can induce apoptosis and
   increase the accumulation of extracellular matrix proteins, whereas
   knockdown of miR-29c prevented high glucose-induced
   apoptosis.^[[123]43] Several studies have shown that miR-200 family
   members can regulate the epithelial-to-mesenchymal transition (EMT) and
   its reversal process, and that the overexpression of miR-200 family
   members can prevent the TGF-β-induced EMT.^[[124]44,[125]45] The EMT
   plays an important role in renal fibrosis, and renal interstitial
   fibrosis is a common feature of the progression of various kidney
   diseases developing into end-stage renal failure,^[[126]46] suggesting
   that miR-200 can affect the development of DN.

   The KEGG pathway analysis showed that VEGFA was mainly enriched in the
   PI3K/Akt and focal adhesion pathways. Currently, studies in vivo and in
   vitro have indicated that the PI3K/Akt signaling pathway is involved in
   DN tubular cell apoptosis, mesangial cell hyperplasia, and podocyte
   dysfunction. High glucose induces increased Akt expression and
   accelerates the transdifferentiation of renal tubular epithelial cells,
   leading to renal fibrosis.^[[127]47–[128]49] Previous studies reported
   that miR-200 ultimately promoted glomerular hypertrophy and renal
   fibrosis through the PI3K/Akt signaling pathway.^[[129]50,[130]51]
   These studies indicated that the mechanism of miR-200 inducing DN by
   targeting VEGFA through the PI3K/Akt signaling pathway was feasible.
   Thus, miR-200 may be a prospective target for the prevention or
   treatment of renal fibrosis in diabetic patients. This mechanism needs
   to be further verified in vivo and in vitro experiments as well as
   large-scale clinical trials in the future.

   This study also found that COL4A3 and COL4A5 were downregulated in DN
   glomeruli. COL4A3 was regulated by miR-29 and miR-17, and COL4A5 was
   regulated by miR-29 and miR-23. Collagen type IV (COL4) is the main
   collagen component of the ECM and plays an important role in the renal
   pathological changes of DN.^[[131]52] The accumulation of ECM is
   considered the key to glomerulosclerosis and renal failure in
   DN.^[[132]53] Zhao et al^[[133]54] revealed that miR-23b was
   downregulated in DN and promoted renal fibrosis and proteinuria via the
   PI3K/Akt signaling pathway.^[[134]55] COL4A3 and COL4A5, which are
   subtypes of type IV collagen, may be regulated by multiple miRNAs,
   including miR-29 and miR-23, and participate in the pathogenesis of DN
   through the ECM–receptor interaction and PI3K/Akt signaling pathways.

   The hub genes found in this study also include ITGA3, ITGB5, CBLB, and
   CCL19. ITGA3 encodes integrin α3, and ITGB5 encodes integrin β5, which
   have been found to be differentially expressed in the glomeruli and
   tubules of DN patients.^[[135]12] Integrin α3β1 is mainly expressed in
   podocytes and is thought to play a key role in podocytes attachment to
   the glomerular basement membrane (GBM)^[[136]56]; a decrease in
   integrins leads to the detachment of podocytes from the GBM, which
   causes proteinuria and promotes renal damage.^[[137]57,[138]58]
   However, Sawada et al^[[139]59] found that ITGA3 expression was
   upregulated in early DN podocytes. Therefore, the roles of ITGA3 and
   ITGB5 in DN require further investigation. Currently, chemokines and
   their receptors play a beneficial role in deepening understanding of
   the inflammatory mechanism of DN. CCL19, which is a member of the
   chemokine family, was identified as a key factor in this study and
   might be involved in kidney damage in diabetic conditions. CCL19
   expression was increased in DN patients. Many previous studies have
   confirmed the roles of CCL2, CXCL9, CXCL16, and other chemokines and
   their receptors in DN and have verified that high levels of chemokines
   mediate renal injury.^[[140]60–[141]62] At present, blocking RAAS is
   the main clinical strategy for treating the inflammatory process of DN.
   Newer strategies based on chemokine receptor antagonists or
   immunosuppressive therapy are still in the experimental stages.
   Moreover, there are lack of relevant animal experiments and clinical
   studies for confirming the effect of CCL19 in DN. Thus, CCL19 may be a
   new research target. Notably, miR-181 targeted both ITGA3 and CBLB,
   whereas ITGA3 was regulated by miR-506, which have not been previously
   reported in the field of DN and may be the starting point for future
   studies.

   Other key miRNAs identified in this study were miR-25, miR-27, and
   miR-124a. Previous studies have reported that miR-25 downregulation
   plays an important part in diabetic kidney injury, such as high
   glucose-induced podocyte apoptosis, persistent proteinuria, and
   secondary hypertension.^[[142]63–[143]65] In addition, high glucose
   stimulation has been shown to increase miR-27a expression, leading to
   the worsening of renal function and increasing tubulointerstitial
   fibrosis in DN patients.^[[144]66–[145]68]

   Previous studies based on bioinformatics analysis have mostly analyzed
   either gene expression profiling or noncoding RNA profiling, but not
   conducted between miRNAs, genes, and signaling pathway. In the present
   study, we made a perfect combination of miRNAs-genes-signaling pathway
   and identified several important miRNAs-genes regulatory network in DN
   such as miR-200-VEGFA-PI3K/Akt pathway. Furthermore, a few novel
   miRNA-target pairs (eg, miR-23 targeting COL4A5, miR-181 targeting
   ITGA3, miR-17 targeting VEGFA and COL4A3, and miR-506 targeting ITGA3)
   were found to be involved in diabetic kidney injury. These key genes
   and miRNAs have potential value for the diagnosis and treatment of DN
   and may provide new breakthroughs for further studies focusing on the
   new biomarkers and therapeutic targets of DN. However, we only
   identified candidate aberrantly DEGs and predicted some potential
   miRNAs-genes regulatory network in DN based on bioinformatics analysis.
   Our study is just at the tip of the iceberg of understanding for the
   prevention, diagnosis, treatment, and prognosis of DN. Further
   molecular biological experiments are necessary to confirm the function
   of the identified DEGs as well as the interaction between microRNAs and
   genes in DN. In future studies, we also need to analyze more clinical
   samples to validate these results.

Acknowledgments