Integrated bioinformatic analysis identifies COL4A3, COL4A4, and KCNJ1 as key biomarkers in Wilms tumor

Abstract

Wilms tumor (WT) is one of the most common pediatric solid tumors, affecting 1 in 10,000 children, worldwide. A subset of WT patients has poor prognosis, which is associated with a high risk of advanced and/or recurrent disease. Therefore, candidate markers are urgently needed for the diagnosis and effective treatment of WT. We evaluated three mRNA microarray datasets to identify the differences between normal kidney tissue and WT tissue. Gene expression profiling revealed 130 differentially expressed genes (DEGs). Enrichment analysis and gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed for the DEGs. Subsequently, we established a protein-protein interaction (PPI) network to reveal the associations among the DEGs and selected 10 hub genes, all of which were downregulated in WT. The expression of COL4A3, COL4A4, KCNJ1, MME, and SLC12A1 in WT tissues was significantly lower than that in normal renal tissues. Survival analyses using the Kaplan-Meier method showed that patients with WT and low expression of COL4A3, COL4A4, and KCNJ1 exhibited remarkably poor overall survival. The correlations among COL4A3, COL4A4, and KCNJ1 in WT were analyzed using cBioPortal; COL4A3, COL4A4, and KCNJ1 were positively correlated with each other. Thus, these genes were considered clinically significant and might therefore play important roles in carcinogenesis and the development of WT.

Introduction

Wilms tumor (WT) is a type of kidney cancer that develops during childhood; it is usually diagnosed in children between 3 and 4 years of age [1]. WT accounts for more than 90% of renal tumors in children and is the most common solid renal malignant tumor worldwide, with 1 in every 10,000 children being affected by the disease [2,3]. It is the fourth leading cause of malignant tumors in children, accounting for 5% of all cancers and 95% of all kidney cancers in children [4]. WT is prevalent on a global scale, although there are significant differences in the associated morbidity and prognosis [5]. The treatment regimen for WT usually includes surgery, chemotherapy, and radiotherapy [6] and the long-term survival outcomes for pediatric WT patients have gradually improved over the last several decades. However, specific subgroups of patients, such as those with relapse and anaplastic histology, have poor event-free survival and are at risk of developing significant late effects in response to therapy [7,8].

The exact molecular mechanisms underlying the development of WT are still unclear; thus, it is necessary to investigate the potential biomarkers of WT as well as its biologic perturbations to prevent the occurrence of tumors and to develop effective therapeutic measures. In this respect, continued investigation of WT cell proliferation and WT recurrence, and carcinogenesis is necessary. Microarrays and RNA-sequencing have been widely utilized to explore the differentially expressed genes (DEGs) and gene expression profiling data are available for download from the Gene Expression Omnibus (GEO) database.

In this study, we downloaded and analyzed three gene expression profile datasets to identify DEGs between normal renal tissue and WT tissue and to investigate the molecular functions and clinical significance of these genes in WT. In total, ten hub genes and 130 DEGs that may serve as potential biomarkers of WT were identified. These results provide new insights into the pathogenesis of disease. Importantly, some of the newly identified genes may serve as biomarkers for the diagnosis and prognosis of WT.

Materials and methods

Microarray data

GEO is a publicly-available genomics database that contains gene chips, microarrays, and high-throughput gene expression data (http://www.ncbi.nlm.nih.gov/geo) [9]. We downloaded three gene expression datasets (GSE19249, GSE6280, and GSE66405) from GEO [10-12], containing data corresponding to normal kidney samples and WT samples (8 vs. 8, 6 vs. 2, and 4 vs. 28, respectively) (Table 1).

Table 1.

Statistics of the three microarray datasets

Dataset ID	Platforms	Country	NK	WT	Total
GSE19249	GPL570	USA	8	8	16
GSE6280	GPL96	Netherlands	6	2	8
GSE66405	GPL17077	Germany	4	28	32

DEG identification

GEO2R (http://www.ncbi.nlm.nih.gov/geo/geo2r/), which allows interaction analysis, is a tool that can compare at least two datasets in GEO series to identify DEGs according to experimental conditions. Genes with values above the cutoff (P<0.05 and |log FC (fold change)| ≥ 1) were regarded as significant DEGs between normal kidney samples and WT samples. DEGs with the same change in the dataset were combined using the web tool Venn diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/).

KEGG and GO enrichment analysis of DEGs

Gene ontology (GO) is considered one of the main bioinformatic methods to analyze the biologic processes associated with the DEGs [13]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis is used to explore biologic systems and high-level functions based on a wide range of molecular datasets using high-throughput experimental technology methods [14]. The network of interactions between molecules was visualized using Cytoscape (version 3.6.1), an open source software [15]. ClueGO (version 2.5.4) is a plug-in app that allows the creation and visualization of functionally grouped networks of terms/pathways [16]. The GO annotation and KEGG pathway enrichment analysis were performed using ClueGO. P<0.05 was considered significant.

Construction of the PPI network

To construct a protein-protein interaction (PPI) network with a confidence score above 0.4 as the cut-off criterion, the Search Tool for the Retrieval of Interacting Genes (STRING; http://string-db.org) online database was utilized [17]. Subsequently, Cytoscape was used to establish and visualize the PPI network for protein interactions.

Hub gene selection and analysis

The CytoHubba plugin in Cytoscape was used to explore maximal clique centrality (MCC) in each protein node. The top ten genes were identified as hub genes in our study. CytoHubba was used to analyze the network of hub genes and their co-expressed genes. Gene expression levels between normal kidney samples and WT samples were compared using the Oncomine database (http://www.oncomine.com) [18,19]. Survival analysis for the hub genes was performed using a Kaplan-Meier curve and P-values were calculated using the UCSC Xena Functional Genomics Explorer (https://xenabrowser.net/) [20], with data retrieved from the TARGET database (ocg.cancer.gov/programs/target). Correlations between gene expression levels were reported in cBioPortal (http://www.cbioportal.org) [21,22].

Statistical analysis

Student’s t-test for independent samples was performed to assess the significant differences between fetal kidney tissues and WT tissues. The Kaplan-Meier method was used to generate survival curves, and the log-rank test was used to assess significant differences in overall survival time between the high and low expression level groups. Correlations were separately evaluated using the pairwise correlation among genes according to the Pearson chi-square test. A value of P<0.05 was considered significant.

Results

Identification of DEGs in WT

DEGs (811 in GSE19249, 2514 in GSE6280, and 2,749 in GSE66405) were determined after standardizing the three gene expression profiles. We performed Venn diagram analysis to visualize the intersection of the DEG profiles (Figure 1). Ultimately, 130 genes were differentially expressed in the three groups; among these, 11 were upregulated and 119 were downregulated.

Figure 1.

The Venn diagram of the DEGs. In total, 130 DEGs were remarkably differentially expressed in the three groups. DEGs, differentially expressed genes.

KEGG and GO enrichment analysis of DEGs

The results (Table 2) of GO annotation analysis showed that the DEGs were enriched in terms associated with the following biologic processes (BP): “renal system development”, “kidney development”, and “urogenital system development”. DEGs were enriched in the following cellular component (CC) terms: “basolateral plasma membrane”, “cluster of actin-based cell projections”, and “brush border”. Additionally, DEGs were enriched in the following molecular function (MF)-associated term: “growth factor-receptor binding”. Furthermore, KEGG pathway analysis showed that the DEGs were mostly associated with “protein digestion and absorption”, “aldosterone-regulated sodium reabsorption”, and “collecting duct acid secretion” (Table 2).

Table 2.

Significantly enriched GO terms and KEGG pathways of DEGs

Category	Term	Description	Count in gene set	P-value
BP term	GO: 0072001	renal system development	18.00	3.56E-11
BP term	GO: 0001822	kidney development	17.00	1.54E-10
BP term	GO: 0001655	urogenital system development	18.00	4.14E-10
BP term	GO: 0072006	nephron development	12.00	1.73E-09
BP term	GO: 0098739	import across plasma membrane	11.00	6.00E-09
BP term	GO: 0015698	inorganic anion transport	12.00	7.18E-08
BP term	GO: 0006821	chloride transport	10.00	1.07E-07
CC term	GO: 0005903	brush border	11.00	1.42E-09
CC term	GO: 0031526	brush border membrane	8.00	1.31E-08
CC term	GO: 0098862	cluster of actin-based cell projections	11.00	7.75E-08
CC term	GO: 0016323	basolateral plasma membrane	12.00	5.82E-07
MF term	GO: 0070851	growth factor receptor binding	8.00	1.15E-05
KEGG pathway	hsa: 04974	Protein digestion and absorption	8.00	2.71E-06
KEGG pathway	hsa: 04960	Aldosterone-regulated sodium reabsorption	5.00	2.95E-05
KEGG pathway	hsa: 04966	Collecting duct acid secretion	4.00	1.35E-04
KEGG pathway	hsa: 04512	ECM-receptor interaction	6.00	1.55E-04
KEGG pathway	hsa: 04514	Cell adhesion molecules (CAMs)	7.00	5.49E-04

Construction of the PPI network and hub gene selection

We established a PPI network related to the DEGs using STRING tools and Cytoscape. In total, 169 edges and 92 nodes were present in the PPI network (Figure 2). In the PPI network, the top ten hub genes were recognized using CytoHubba (Table 3). The expression of potassium inwardly rectifying channel subfamily J member 1 (KCNJ1) was the most significantly altered. All hub genes were underexpressed in WT samples.

Figure 2.

The PPI network was established according to DEGs. Upregulated genes were red nodes, and downregulated genes were purple nodes. PPI, protein-protein interaction; DEGs, differentially expressed genes.

Table 3.

Top ten hub genes with higher MCC of connectivity

No.	Gene symbol	Full name	Alias	MCC
1	KCNJ1	potassium inwardly rectifying channel subfamily J member 1	KIR1.1, ROMK, ROMK1	46
2	SLC12A1	solute carrier family 12 member 1	BSC1, NKCC2	45
3	CASR	calcium sensing receptor	CAR, EIG8, FHH, FIH, GPRC2A, HHC, HHC1, HYPOC1, NSHPT, PCAR1, hCasR	40
4	CLCNKB	chloride voltage-gated channel Kb	CLCKB, ClC-K2, ClC-Kb	37
5	COL4A3	collagen type IV alpha 3 chain	ATS2, ATS3	37
6	COL4A4	collagen type IV alpha 4 chain	ATS2, BFH, CA44	36
7	AGT	angiotensinogen	ANHU, SERPINA8, hFLT1	36
8	ITGB3	integrin subunit beta 3	BDPLT16, BDPLT2, CD61, GP3A, GPIIIa, GT	34
9	ITGA6	integrin subunit alpha 6	CD49fB, VLA-6, ITGA6	32
10	MME	membrane metalloendopeptidase	CALLA, CD10, CMT2T, NEP, SCA43, SFE	27

Expression analysis of hub genes

A network of hub genes and their co-expressed genes was generated using CytoHubba (Figure 3). In the Cutcliffe Renal dataset [18], the expression levels of COL4A3, COL4A4, KCNJ1, MME, and SLC12A1 in WT patient samples were considerably lower than those in normal kidney samples (Figure 4). In the Yusenko Renal dataset [19], the gene expression levels in WT patient samples were significantly comparable to those in the normal kidney samples (Figure 5).

Figure 3.

The network of the hub genes and their co-expression genes. The co-expression genes were green and purple nodes, whereas the hub genes were the others. The highest MCC gene was red node, and the lowest MCC gene was yellow node. MCC, maximal clique centrality.

Figure 4.

Gene expression levels between normal kidney and WT samples. In the Cutcliffe Renal dataset, the expression levels of COL4A3 (D), COL4A4 (E), KCNJ1 (H), MME (I), and SLC12A1 (J) in patients with WT were significantly lower than those in individuals with normal kidneys. There are no significantly differential expression levels of AGT (A), CASR (B), CLCNKB (C), ITGA6 (F), and ITGB3 (G) between normal kidney and WT samples. WT, Wilms tumor.

Figure 5.

Gene expression levels between normal kidney and WT samples. In the Yusenko Renal dataset, the expression levels of COL4A3 (A), COL4A4 (B), KCNJ1 (C), MME (D), and SLC12A1 (E) in patients with WT were significantly lower than those in individuals with normal kidneys. WT, Wilms tumor.

Survival analysis

Using data from the TARGET database (https://ocg.cancer.gov/programs/target), the correlation between the overall survival rates of patients and expression of COL4A3, COL4A4, KCNJ1, MME, and SLC12A1 was analyzed. Kaplan-Meier curves were drawn using the UCSC online tool. Among a total of 132 WT patients, 66 patients were in the low-expression group and 66 patients were in the high-expression group. Low expression of COL4A3, COL4A4, and KCNJ1 in WT patients was associated with a remarkably poor overall survival (Figure 6). Therefore, the expression of COL4A3, COL4A4, and KCNJ1 is associated with clinical prognosis and may have a crucial role in the oncogenesis, development, and metastasis of WT.

Figure 6.

Survival analysis of hub genes. Patients with WT and low gene expression of COL4A3 (D), COL4A4 (E), and KCNJ1 (H) showed remarkably worse overall survival. There are no statistical significance between high and low gene expression of AGT (A), CASR (B), CLCNKB (C), ITGA6 (F), ITGB3 (G), MME (I), and SLC12A1 (J). WT, Wilms tumor.

Correlation analysis

The correlation among COL4A3, COL4A4, and KCNJ1 was analyzed using the cBioPortal online platform. In WT samples, COL4A3 expression was found to be positively correlated with that of COL4A4 (R = 0.8889, P<0.01) and KCNJ1 (R = 0.6164, P<0.01), and that of COL4A4 was found to be positively correlated with KCNJ1 expression (R = 0.5657, P<0.01) (Figure 7).

Figure 7.

Correlations among COL4A3, COL4A4, and KCNJ1 in WT. The correlations among COL4A3, COL4A4, and KCNJ1 in WT were analyzed, and they were positively correlated with every two. WT, Wilms tumor.

Discussion

WT, also known as nephroblastoma, is the most common form of kidney cancer in children. It is caused by poorly differentiated mesenchymal renal stem cells [23]. This disease accounts for 90% of the renal tumors and 7% of all cancers in children [24]. WT therapy usually involves multiple modes of treatment including surgery, chemotherapy, and radiotherapy. The long-term survival rates of pediatric WT patients have gradually improved in recent years. However, chronic health conditions secondary to treatment, including renal failure, infertility, cardiotoxicity, restrictive lung disease, and subsequent malignant tumor development, affect nearly a quarter of the WT survivors [25,26]. Therefore, candidate biomarkers are urgently needed for the diagnosis and effective treatment of this condition. Microarray technology provides an opportunity to explore gene variations in WT and is a useful approach for identifying new biomarkers in other cancers as well.

In this study, we identified the key biomarkers, COL4A3, COL4A4, and KCNJ1-all of which were downregulated in WT samples-through bioinformatic analysis. Our different approaches-through GEO, ONCOMINE, and TARGET database analyses-revealed that the expression of COL4A3, COL4A4 and KCNJ1 in these three databases was consistent. WT patients with low expression of COL4A3, COL4A4, and KCNJ1 showed remarkably poor overall survival. Additionally, COL4A3, COL4A4, and KCNJ1 positively correlated with each other in WT tissues. Therefore, COL4A3, COL4A4, and KCNJ1 are possible clinically relevant genes with important roles in the carcinogenesis and development of WT.

Svetlana et al. [27] identified that genes, including those coding for adhesion molecules-COL4A3 and CDH5-were downregulated in early non-small-cell lung cancer. Further, Nie et al. [28] suggested that COL4A3 overexpression is associated with the development of gastric cancer. The upregulation of COL4A3 was associated with other tumors, including WT, breast cancer, and ovarian cancer [29-31]. Wang et al. [32] reported that COL4A4 expression is reduced in patients with clear cell renal cell carcinoma (ccRCC) compared to that in paracancerous tissues, whereas patients with ccRCC and high expression of COL4A4 exhibit remarkably longer overall survival. Chattopadhyay et al. [33] suggested that genes (KRT4 and COL4A4) coding for proteins involved in the extracellular matrix tissue and cell communication pathways are significantly downregulated in esophageal cancer. Li et al. [34] reported that the expression of COL4A4 is markedly downregulated in esophageal squamous cell carcinoma. Guo et al. [35] reported that KCNJ1 has low expression and is associated with poor prognosis in ccRCC; moreover, it plays an important role in the growth and metastasis of this carcinoma. Valletti et al. [36] reported that low KCNJ1 expression in patients with ccRCC can be used in the preliminary diagnosis of the condition and prediction of the prognosis and therapeutic response.

In our study-in addition to COL4A3, COL4A4, and KCNJ1-we identified seven other hub genes associated with WT, including SLC12A1, CASR, CLCNKB, AGT, ITGB3, ITGA6, and MME. Dysregulation of these genes has also been reported in glioma, hyperparathyroidism, Bartter syndrome, and gastric, lung, bladder, and breast cancers; thus, these genes may represent valuable biomarkers for tumor diagnosis, treatment, and prognosis [37-43]. COL4A3, COL4A4, and KCNJ1 were downregulated in WT tissues and were associated with worse overall survival, suggesting that these genes might serve as biomarkers and therapeutic targets for WT. Nonetheless, further studies are required to investigate the function of these genes in WT.

Conclusion

The purpose of this study was to identify DEGs that might be involved in the carcinogenesis or progression of WT. In our study, newly identified DEGs and hub genes provided an opportunity to understand the mechanism underlying the carcinogenesis and development of nephroblastoma. COL4A3, COL4A4, and KCNJ1 may be the key genes in WT and could represent targets for the diagnosis and treatment of the disease. As we have adopted a bioinformatic approach, further biological experiments are required to better understand the role of these genes in WT.

Disclosure of conflict of interest

None.