miR-432 is a novel biomarker for PCNSL and is associated with cell adhesion: An integrated bioinformatics analysis

Background: Primary central nervous system lymphoma (PCNSL), a rare form of the non-Hodgkin's lymphoma (NHL), usually has a poor prognosis, and molecular pathogenesis of PCNSL has not been fully elucidated. Here, potential miRNA biomarkers were investigated in patients with PCNSL using an integrated bioinformatics analysis. Methods: Expression profile arrays (GSE122011, GSE139031, and GSE25297) were obtained from the Gene Expression Omnibus (GEO). Free-scale miRNA co-expression networks were constructed with 27 PCNSL patients from GSE122011 by the weighted gene co-expression network analysis (WGCNA) in order to identify candidate biomarkers. Subsequently, miRNA-miRNA networks were visualized with the Cytoscape. Expression of candidate miRNAs was assessed in serum samples from GSE139031, including 42 PCNSL patients and 77 non-cancer individuals, and the sensitivity and the specificity were assessed by the receiver operating characteristic (ROC) curve. From GSE25297, differentially expressed genes (DEGs) from the PCNSL tissues (n = 7) and the normal lymph nodes (n = 7) were compared, target genes of candidate miRNAs were downloaded from TargetScan database, and target genes that were also down-regulated in GSE25297 were used to construct the protein-protein interaction (PPI) networks and for the gene ontology (GO) analysis. Results: miRNAs were clustered into two groups with 8 modules in 27 patients with PCNSL. One group consisted of the yellow and the turquoise modules, and the second group consisted of the other six modules. In the miRNA-miRNA network, the highest nodes were observed between miR-432 and miR-330-3p, which were from the yellow and the turquoise modules, and only miR-432 was closely associated with both the yellow (0.977, P = 2.88E -18 ) and the turquoise modules (0.525, P = 0.005). Additionally, patients with PCNSL had higher serum miR-432 expression compared with that in the non-cancer controls in

Conclusion: Up-regulated miR-432 expression is a novel biomarker for patients with PCNSL and may be associated with cell adhesion. Background Primary central nervous system lymphoma (PCNSL) is a rare subtype of the non-Hodgkin's lymphoma 1 and usually has a poor prognosis. Thus, accurate and timely diagnosis is necessary 2 . Until now, stereotactic biopsy has been the gold standard for PCNSL diagnosis 3 . Recently, discovery of molecular biomarkers from the cerebrospinal fluid (CSF) 4 and the serum 5 have become an area of extensive research. Among the novel biomarkers, microRNAs (miRNAs) have been considered in PCNSL diagnosis because the primary central nervous system diffuse large B-cell lymphoma (PCNS-DLBCL) has a distinct miRNA signature compared to that in the germinal center (GC)-DLBCL and the non-GC-DLBCL 6 . Despite this finding, there is a limited clinical application of miRNA. Hence, there is a necessity of further research into the discovery of diagnostic miRNAs using a variety of methods. In the present study, a combined analysis using multiple databases was performed, and key miRNAs associated with PCNSL were investigated with the goal of providing a theoretical basis for the development of diagnostic markers.

Materials And Methods
Download and analysis of expression profiles from GEO.
Expression profiles of GSE122011, GSE139031, and GSE25297 were downloaded from the GEO database (https://www.ncbi.nlm.nih.gov/gds). MiRNA expression profiles from 27 patients with PCNSL were extracted from GSE12201. Serum miRNA expression profiles from 42 patients with PCNSL and 77 non-cancer controls were extracted from GSE139031. From GSE25297, differentially expressed genes in seven PCNSL and seven normal lymph node sets were screened using the R software (FDR < 0.01 and log 2 fold change (FC, tumor-control) > 2 or < -2).
WGCNA and ROC analysis.
MiRNA co-expression networks from 27 patients with PCNSL from GSE12201 were constructed using WGCNA analysis with the WGCNA package in R. Modules of the highly correlated miRNAs, module membership (MM), and a module heatmap were created as per a previously reported method 7 . Sensitivity and specificity of each candidate miRNA in GSE139031 were evaluated by ROC analysis as per a previously reported method 8 .
PPI and GO analysis.
MiR-432 target genes were downloaded from the TargetScan (http://www.targetscan.org/vert_72/) 9 , and target genes shared with the down-regulated genes in GSE25297 were subjected to a protein-protein interaction analysis with the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) (https://stringdb.org/) 10 . The GO analysis was performed for genes with the STRING interactions, including biological process (BP), molecular function (MF), and cellular component (CC).
Statistical methods.
IBM SPSS Statistics 21.0,GraphPad Prism version 5 and R software with corresponding packages were used to perform the statistical analysis. Relative miR-432 expression was compared with the independent t-tests, and P < 0.05 was considered statistically significant.

Results
Eight modules were constructed for miRNAs in PCNSL.
Co-expressed miRNAs in PCNSL were clustered into eight modules. Among them, the yellow and the turquoise modules constituted one subgroup, and the green, the black, the brown, the red, the blue, and the pink modules constituted another subgroup (Fig. 1A).
Module clustering indicated that the yellow module had the highest correlation with the turquoise module (Fig. 1B). Additionally, higher miRNA expression overall was observed in the yellow and the turquoise modules (Fig. 1C). miR-432 is closely associated with the yellow and the turquoise modules.
A miRNA-miRNA network was constructed for the top 20 miRNAs in the turquoise and the yellow modules to identify key miRNAs in PCNSL ( Fig. 2A and Fig. 2B). A higher number of nodes was observed between miR-330-3p and miR-432 compared to those in other miRNAs in these two modules. Meanwhile, MM values of all miRNAs from these two modules were calculated, and only miR-432 was significantly associated with both modules (Table 1).
These results suggested that miR-432 may be a key link between the yellow and the turquoise modules. To identify whether abnormal miR-432 was also expressed in the peripheral circulation, serum miR-432 was determined in patients with PCNSL and non-cancer samples. The results confirmed that serum miR-432 was higher in patients with PCNSL compared with 7 that in the non-cancer control ( Fig. 3A; P < 0.0001). The ROC curve for PCNSL generated with serum miR-432 has an AUC of 0.77 with a 95% confidence interval (CI) of 0.6923 to 0.8550 (P < 0.0001), which suggests that miR-432 may be a potential discriminator between PCNSL and non-cancer profiles.
miR-432 may be involved in the regulation of cell adhesion in PCNSL.
To explore the potential biological function of miR-432 in PCNSL, down-regulated miR-432 target genes were screened. The miR-432 target gene set was downloaded from the TargetScan database, and genes down-regulated in PCNSL were analyzed in GSE25297 and RASGRF1 (Fig. 4C). The GO analysis of these seven genes was performed using the "Analysis" function in STRING, and BP, MF, and CC and have been shown in Tables 2-4.
Furthermore, we observed that four target genes, CTNND2, GLDN, NRCAM, and PTPRD were involved in regulating cell-cell adhesion.

Discussion
In this study, up-regulated miR-432 was found in the tumor tissues and in the serum of patients with PCNSL and was explored as a potential diagnostic biomarker for PCNSL.
Compared with the normal lymph nodes, 15 miR-432 target genes were down-regulated in PCNSL and linked to a cell-cell adhesion. These results indicated that serum miR-432 may be a potential marker for identifying PCNSL, and miR-432 may be associated with the regulation of cell adhesion.
Although stereotactic biopsy is the gold standard for PCNSL diagnosis, non-invasive diagnostic methods are under investigation 11 , and liquid biopsy has been presented as a promising method 2 . MiR-30c in the CSF could serve as a biomarker to distinguish between PCNSL and SCNSL 12 . MiR-151a-5p and miR-151b exhibited the most prominent differences in the blood of patients with PCNSL 13 . Additionally, it is necessary to develop novel, more effective miRNA biomarkers because there are limited molecular biomarkers with a widespread clinical applicability. Recently, WGCNA was used for the development of disease-related diagnostic and prognostic markers by creating relationships between gene modules and clinic traits 14 . In this study, the WGCNA approach was used to construct the miRNA co-expression network from 27 patients with PCNSL. All miRNAs were divided into two groups with eight modules. The yellow and the turquoise modules consisted of a subgroup with a high correlation. In order to find the key miRNAs shared between these two modules, interaction networks of the top 20 miRNAs were constructed. In parallel, MM values of all miRNAs in these two modules were calculated. Finally, miRNA-432 was identified as a potential key miRNA due to the higher node number and close association between these two modules. In fact, an abnormal expression of miR-432 has been observed in multiple tumor tissues and linked to the drug resistance in tumors and the malignant phenotype. For example, miR-432-3p was highly expressed in the ESCC tumors compared with that in the corresponding non-cancerous esophageal mucosa, and upregulated miR-432 could decrease sensitivity of the cancer cells to chemotherapeutic drugs 15 . MiR-432 has also been reported to play an important role in tumor progression; it can suppress tumor growth, but also promote tumor progression by regulating cell proliferation and apoptosis 16

Conclusions
In brief, serum miR-432 was up-regulated in patients with PCNSL and may be associated with the regulation of cell-cell adhesion. Thus, miR-432 may be a target for both diagnosis and treatment.

Ethics approval and consent to participate
Not applicable.

Consent for publication
Not applicable.

Availability of data and material
All data generated during this study are included in this published article.