Identification and integrative analysis of microRNAs in myelodysplastic syndromes based on microRNAs expression profile

Myelodysplastic syndromes (MDS) are a group of malignant hematological disorders characterized by the abnormal development of hematopoietic stem cells and increased risk of acute myelogenous leukemia. Although the pathogenesis of MDS has not been fully understood, various alterations of microRNAs (miRNAs) have been reported in MDS. This study aimed to explore the molecular mechanisms of MDS by integrative bioinformatics analysis of miRNAs expression profile. The GSE81372 expression profile dataset was downloaded from Gene Expression Omnibus database. The differentially expressed miRNAs (DEMs) between MDS and normal controls were identified and targets of miRNAs were predicted. Subsequently, gene ontology (GO) functional and pathway enrichment analyses of target genes were performed. Finally, pathway relation network and miRNA–GO regulatory network were constructed and analyzed. A total of six upregulated and 35 downregulated DEMs were identified. The results showed that target genes of DEMs mainly participated in the process of signal transduction, blood coagulation, apoptotic process, cell proliferation, transmembrane transport, and angiogenesis. The significantly enriched pathways included MAPK signaling pathway, PI3K‐Akt signaling pathway, TGF‐beta signaling pathway, Hippo signaling pathway, and P53 signaling pathway. Moreover, miR‐195‐5p, miR‐4505, miR‐22‐3p, and miR‐148a‐3p were selected as hub miRNAs in miRNA–GO regulatory network and their aberrant expression might be closely associated with MDS pathogenesis. Our discovery provides a registry of miRNAs and pathways that are disrupted in MDS, which has the potential to be used in clinic for diagnosis and target therapy of MDS in future.

contributors in the pathophysiology of MDS. 4,5 However, to date, the precise molecular mechanisms behind the occurrence and development of MDS are still not fully elucidated, and current effective treatment and diagnosis methods of MDS are limited because of the heterogeneity and complexity of this disease. Therefore, it is important to clarify the molecular pathogenesis of MDS and to explore effective targeted therapies for clinical use.
MicroRNAs (miRNAs) are short noncoding RNA molecules that repress expression of genes by inhibiting translation or inducing degradation of target mRNA at the posttranscriptional level. 6 miRNAs regulate numerous biological processes such as proliferation, apoptosis, and differentiation. In addition, miRNAs can be either oncogenes or tumor suppressors in the pathogenesis of different cancers, which may be due to the fact that over half of miRNA genes are located in cancer-related genomic regions. 7 In recent years, increasing studies have shown that miRNAs are implicated in normal hematopoiesis, and dysregulation of miRNA has been found in hematological malignancies including MDS. [8][9][10] It is relatively little known about miRNAs in pathogenesis, prognosis, and therapy of MDS. In the present study, we applied an integrative bioinformatics approach to analyze the miRNAs expression profiles of MDS. We aimed to provide a systematic perspective toward understanding molecular mechanisms and exploring new therapeutic targets for MDS.

| Microarray data
The microarray expression profile GSE81372 was downloaded from Gene Expression Omnibus database, which was based on the GPL16384 Affymetrix Multispecies miRNA-3 Array platform. The dataset contained the miRNAs expression profiles of bone marrow CD34+ cells from 12 MDS patients and six normal controls, which was deposited by Xu et al. 11

| Differential expression analysis
The raw data were firstly preprocessed using the Affy package, then the probe-level data in CEL files were converted into expression value matrix. Data preprocessing was performed using robust multiarray average algorithm, including background correction, quartile data normalization, and probe summarization. 12 The significance analysis of microarray method was used to identify the differentially expressed miRNAs (DEMs) between MDS patients and normal controls. 13 Only miRNAs with p < .05 and jfold-changej >2 were considered as DEMs.
Hierarchical cluster analysis was performed and cluster dendrogram was constructed to assess the characterizations of screened DEMs. 14

| Target genes prediction of miRNAs
Interactions between miRNA and mRNA were predicted based on the TargetScan and miRanda databases. 15 The intersections recognized by two main algorithms were considered as candidate target genes of DEMs.

| Functional enrichment analysis
Gene ontology (GO) analysis was applied to explore the main functions of target genes of DEMs identified in this study. Specifically, two-side Fisher's exact test was used to classify GO category and the false discovery rate (FDR) was calculated using Benjamini-Hochberg method to correct the p value. FDR < 0.05 was set as the threshold value to select significant GO categories. Besides, the enrichment score was calculated to access the enrichment level for each GO category.

| Pathway enrichment analysis
Kyoto Encyclopedia of Genes and Genomes (KEGG) knowledge database was used for the classification of correlative gene sets into their respective pathways. 16 The Fisher's exact test was used to calculate the significance p value and Benjamini-Hochberg procedure was used to calculate FDR. FDR < 0.05 was used as the cutoff criteria to identify the significant pathways. The enrichment score was also calculated to access the enrichment level for each pathway.

| Pathway network analysis
The network of the significantly enriched pathways was built according to the interaction between pathways among the KEGG database. Each pathway in the network was measured by counting the number of upstream and downstream pathways, which were shown as in-degree or out-degree, respectively. 17

| miRNA-GO network analysis
The relation between significant GO items and target genes can be got via GO enrichment analysis, then we can get the relevance between miRNAs and significant GO items based on the fact that miRNA can combine with 3 0 -UTR of target mRNA. The adjacent relation matrix of GO items and miRNAs was built, and we evaluated the degrees of miRNAs and GO items in the network using the methods of graph theory. 18 Those who had the highest degree were the core miRNAs or GO items.

| DEMs identification
After data processing, a total of 41 miRNAs, six upregulated and 35 downregulated, were identified to be differentially expressed in MDS patients compared with normal controls ( Table 1). The hierarchical cluster analysis showed that the 12 MDS samples distributed in MDS cluster and six normal samples in control cluster, and no overlap was found between them ( Figure 1). This observation showed that separation of the miRNAs expression profiles between MDS and normal controls.

| Targets prediction of miRNAs
In present study, targets of these DEMs were identified based on sequence complementarities and free energy of the predicted RNA duplex using TargetScan and miRanda. In total, 71 767 target mRNAs were obtained by miRanda, and 26 301 target mRNAs were obtained by TargetScan; a total of 5641 target mRNAs overlapped between two datasets (Table S1).

| Functional enrichment analysis
In order to functionally annotate the target genes, we performed  (Table S2).  (Table S3).  Table 2.

| miRNA-GO network analysis
In order to clarify the regulatory status of miRNAs and GO items, a miRNA-GO network was built according to interactions between miRNAs and GO items ( Figure 5). The network provided us with the key drivers of MDS, including miR-195-5p, miR-4505, miR-22-3p, and miR-148a-3p. It was also noticed that blood coagulation was the core GO items with the highest degree in the network. The top 10 significant DEMs identified by miRNA-GO network were shown in   that these phosphorylation levels are positively correlated with the rate of intramedullary apoptosis. 24,25 In accord with previous studies, Wnt activation may also contribute to the pathogenesis of MDS. Gene expression profiling of hematopoietic cells supports a role for Wnt pathway activation in MDS, AML, and therapy-related myeloid neoplasms. 26 Moreover, Wnt activation in HSCs has been directly implicated in self-renewal of leukemia stem cells, and is associated with a poorer outcome in AML patients. 27 In addition, we also found VEGF signaling pathway was significantly enriched in MDS, suggesting that angiogenesis was extensively implicated in carcinogenesis. The imbalance of oncogene and cancer suppressor was a crucial mechanism of MDS progression, such as the mutations of TP53 gene have an unfavorable prognosis in MDS. 28 Taken together, these pathways and genes involved in tumor cell proliferation, apoptosis, self-renewal and angiogenesis might be taken as diagnostic biomarkers and potential therapeutic targets for MDS.
Pathway analysis identified that focal adhesion, adhesion junction and calcium signaling pathway were enriched. Emerging evidence point to bone marrow microenvironment (BMME) abnormalities as central participants in the progression of MDS pathogenesis whereby, inhibition of Notch signaling pathway by miR-195-5p-DLL1 axis contributes to the excess apoptosis in low-grade MDS. 11 Macrophages overexpressing miR-148a-3p increased their ROS production through the PTEN/AKT pathway, likely to defend against bacterial invasion.
Moreover, miR-148a-3p also enhanced M1 macrophage polarization and pro-inflammatory responses through PTEN/AKT-mediated upregulation of NF-κB signaling. 32 However, there were several limitations of the present study.
First, our study was limited to the small amount of data. Therefore, a meta-analysis including larger sample sizes may be performed in future. Second, the results were not verified by biological experiments. Thus, further experimental studies are still needed to confirm the findings of this study.

| CONCLUSIONS
In summary, our results provide a comprehensive bioinformatics analysis of miRNAs and pathways which may be involved in the carcinogenesis of MDS. Our findings may be helpful for understanding the complex mechanisms underlying MDS and guiding the development of targeted therapies for patients with MDS.

ACKNOWLEDGMENTS
This study was supported by the Science and Technology Development Project of Luoyang City (grant no. 1503007A-4).