Regulatory Network of MicroRNAs, Target Genes, Transcription Factors and Host Genes in Endometrial Cancer

EC is the fourth most common malignancy in women in the developed world after breast, colorectal and lung cancer. Worldwide, similar to 74,000 women die from endometrial cancer each year (Bell, 2014). Molecular targeted therapies are now a focus of attention. They may provide useful future strategies for control of endometrial malignancies in developing countries and across the world (Thanapprapasr et al., 2013). However, the molecular events involved in the development and progression of EC remain unclear (Li et al., 2013). In this study, we mainly focus on the relationships between genes and miRNAs in EC in order to discover the regulation mechanism. TFs and miRNAs are prominent regulators for gene expression (Hobert, 2008). TFs are proteins that can activate or repress transcription by binding to the upstream regions of genes (Tran et al., 2010). MiRNAs comprise a family of small, non-coding RNAs, which regulate gene expression at the posttranscriptional level (Torres et al., 2013). In EC, miRNAs are associated with regulation of gene expression, epigenetic dysfunction and carcinogenesis, thus, miRNAs are likely to have key roles in diagnosis, prognostic prediction, and therapy in EC (Banno et al., 2013).


Introduction
EC is the fourth most common malignancy in women in the developed world after breast, colorectal and lung cancer. Worldwide, similar to 74,000 women die from endometrial cancer each year (Bell, 2014). Molecular targeted therapies are now a focus of attention. They may provide useful future strategies for control of endometrial malignancies in developing countries and across the world (Thanapprapasr et al., 2013). However, the molecular events involved in the development and progression of EC remain unclear . In this study, we mainly focus on the relationships between genes and miRNAs in EC in order to discover the regulation mechanism.
TFs and miRNAs are prominent regulators for gene expression (Hobert, 2008). TFs are proteins that can activate or repress transcription by binding to the upstream regions of genes (Tran et al., 2010).
MiRNAs comprise a family of small, non-coding RNAs, which regulate gene expression at the posttranscriptional level (Torres et al., 2013). In EC, miRNAs are associated with regulation of gene expression, epigenetic dysfunction and carcinogenesis, thus, miRNAs are likely to have key roles in diagnosis, prognostic prediction, and therapy in EC (Banno et al., 2013).
Lu-Chen Xue 1,3 , Zhi-Wen Xu 2,3 *, Kun-Hao Wang 2,3 , Ning Wang 2,3 , Xiao-Xu Zhang 1 , Shang Wang 2,3 The relations between miRNAs and target genes are studied. MiRNAs regulate a variety of cellular processes via the regulation of multiple target genes. They control the cellular expression machinery and have a profound impact on the regulation of signaling pathways (Lv et al., 2012).
Host genes are genes where certain miRNAs locate. It is indicated that miRNAs are transcribed in parallel with their host transcripts and two different transcription classes of miRNAs, exonic and intronic, were identified (Rodriguez et al., 2004). Intronic miRNA and its host gene have a close relation (Baskerville et al., 2005).
Numerous experimental data suggest that the differentially expressed genes and miRNAs contribute to the pathogenesis of EC, while related genes and miRNAs also play a part. However, the data can be found in scattered form, which makes it difficult to analyze the mechanism of EC systematically. In this study, genes and miRNAs are used to construct three networks including differentially expressed network, related network and global network. Each network is combined with three relationships of miRNAs targeting at genes, TFs regulating miRNAs and miRNAs locating on their host genes. Then contrast tables about upstream and downstream of genes and miRNAs are extracted to find the similarities and figure out differences. Crucial genes and miRNAs in the

Data collection and processing of genes and miRNAs
The comprehensive data of EC were manually collected from databases and literatures. Symbols of genes were unified officially by checking them in NCBI database (http://www.ncbi.nlm.nih.gov/gene).
Related genes were collected from GeneCards database (http://www.genecards.org/) and pertinent literatures. In GeneCards database, the first 100 relative genes were picked out. Furthermore, some TFs predicted by Pmatch method (Chekmenev et al., 2005) were also included in related genes. 1000nt promoter region sequences of differentially expressed genes' targets were downloaded from UCSC database (http://genome.ucsc.edu/) (Fujita et al., 2011). Pmatch method that combines pattern matching and weight matrix approaches was used to identify transcription factor binding sites (TFBSs) in 1000nt promoter region sequences and mapped TFBSs onto promoter region of targets. Matrix library of Pmatch is also sets of known TF-binding sites collected in TRANSFAC, so it provides the possibility to search for a large variety of different TF binding sites. The vertebrate matrix was used and be restricted with high quality criterion. The dataset of the related genes was considered as set U 5 .
Differentially expressed miRNAs were collected from mir2Disease (http://www.mir2disease.org/) (Jiang et al., 2009) and pertinent literatures. Related miRNAs were extracted from pertinent literatures. The datasets of differentially expressed miRNAs and related miRNAs were separately considered as set U 6 and set U 7 .

Three networks construction
Three networks of EC were constructed, which are differentially expressed network, related network and global network. Set U 1 , U 2 and U 3 were combined to get the global network. The global network includes all the regulatory relations of miRNAs targeting at targets, TFs regulating miRNAs and host genes including miRNAs. It is a reference to get the differentially expressed network and related network.
The relations between differentially expressed genes and differentially expressed miRNAs that included in the global network were selected to constitute the differentially expressed network. If a pair of gene and miRNA extracted from U 4 and U 6 can be found in the global network, the elements and the relation can be included in the differentially expressed network. The related network was constructed by using the same method with elements in set U 5 and U 7 finally, the software Cytoscape was used to show the network graphically. It helps to analyze the regulatory pathways in EC.

Differentially expressed genes in EC
It is generally accepted that the alterations in the expression of critical genes take major responsibility for the evolution of cancer. CDH1, SDHB, MLH3 and MSH2 can be found in the NCBI SNP database. KRAS, CTNNB1 and ERBB2 can be found in KEGG database. PTEN, TP53, SLC2A1, MSH6, MSH3 and RBL2 come from the Cancer GeneticsWeb database. In pertinent literatures, mutations in PIK3CA, KRAS and PTEN were reported (Machay et al., 2014). In a research of 243 endometrial tumors, PTEN was the most frequently mutated gene (108 tumors, 44%) followed by PIK3CA (97 tumors, 40%) (Cheung, 2012). Moreover, FGFR2, CTNNB1, TP53, KRAS and AKT1 mutations in samples were revealed (Cheung, 2012). Mutations in ESCO1, CHTF18 and MRE11A were uncovered in EC, and the cooccurrence of mutations in ESCO1 and CHTF18 was statistically significant (Price et al., 2013). Mutations in MSH6, TGFBR2 and BAX were found frequently in EC (Kawaguchi et al., 2009). Twist homolog 1 (TWIST1) is a transcription factor and serves as a powerful oncogene, which has been observed to be up-regulated expression and associated with metastasis in several cancers including EC (Li et al., 2014). COPZ2 in EC was frequently down-regulated (Tsuruta et al., 2011). FOXO1, a crucial regulator of progesterone-dependent differentiation of the normal human endometrium, could be repressed expression by several miRNAs (Xu et al., 2013). CCND1 was up-regulated in EC targeted by the down-regulated potential tumor suppressor miR-503 (Xu et al., 2013). DICER1 was higher in normal endometrium than in EC .

Differentially expressed network of EC
There are 5 TFs (PTEN, TP53, TWIST1, CCND1 and AKT1), another 17 differentially expressed genes, 49 differentially expressed miRNAs and 56 host genes in differentially expressed network. It is composed of three relationships including TFs regulating miRNAs, miRNAs targeting at target genes and host genes including miRNAs. The single nodes that do not have regulatory relations with others are omitted in Figure 1 though they play a function too. The differentially expressed network contains many incorrect data linkages when EC emerges, so it can be used as a faulty map of EC. According to the faulty map, we can systematically get the knowledge of the regulation relations in EC.
Besides TFs, other pathways are significant too. CDH1 is targeted by hsa-miR-9(hsa-miR-9-1, hsa-miR-9-2 and hsa-miR-9-3). Besides CDH1, hsa-miR-9 also targets at FOXO1, a crucial regulator of progesterone-dependent differentiation of the normal human endometrium which could be repressed expression by miRNA (Xu et al., 2013). Expression of miR-9 could significantly reduce the abundance of FOXO1 (Myatt et al., 2010). It shows that hsa-miR-9 is an important miRNA involved in EC. MSH2 and MSH6 are targeted by hsa-miR-155 and hsa-miR-21 in common, which indicates that the two genes may have some similarities.
The functions of miRNAs that involved in differentially expressed network are worth concern. For example, hsa-miR-125b can be regulated by AKT1 and TP53 and can target at DICER1, ERBB2, AKT1 and TP53. It is reported that hsa-miR-125b contributes to cells invasion partly through directly down-regulating ERBB2 expression (Shang et al., 2012). Hsa-miR-200c is widely investigated in recent years. It is significantly increased in EC regulated by TP53 and TWIST1 (Cohn et al., 2010). CCND1 has a binding sequence of miR-503 within its 3 untranslated region, and was confirmed to be a direct target of miR-503 by the fluorescent reporter assays (Xu et al., 2013). Abnormal suppression of miR-503 leads to an increase in the CCND1 level, which may promote carcinogenesis and progression (Xu et al., 2013). The content changes of miRNAs help us to know how to adjust them in the pathologic network to control the occurrence of cancer. For example, hsa-miR-155 is validated to be overexpressed in the regulatory network and take effect in the pathogenesis (Choi et al., 2012). The up-regulation impacts the expression of MSH2, MSH6, MLH1, CCND1, CTNNB1 and KRAS. Correcting individual expression level not only can reduce the morbidity of EC, but also rectify the genes and miRNAs downstream to normal standard.
The relation of host genes and its miRNA presents some special features in differentially expressed network. The mutation of host genes may lead to the differentially expression of miRNAs that included in it. Every differentially expressed miRNA in Figure 1 has host gene. A host gene may include one more miRNAs while a miRNA may come from more than one host genes. Hsa-miR-9-1, hsa-miR-9-2 and hsa-miR-9-3 have their own host genes and hsa-miR-9-2 locates in two host genes, MIR9-2 and LINC00461. Hsa-let-7c has three host genes including LINC00478 while LINC00478 hosts three miRNAs including hsa-let-7c, hsa-miR-99a and hsa-miR-125b-2. The host genes are not differentially expressed in EC except COPZ2, which is the host gene of hsa-miR-152. Epigenetic silencing documented in miR-152 was consistent with its location at 17q21.32 in intron 1 of the COPZ2 gene (Tsuruta et al., 2011). It is noted that the frequencies of down-regulation of COPZ2 in EC cell lines and primary cases (<0.5-fold expression) were 100% (13 of 13) and 75.7% (53 of 70) (Tsuruta et al., 2011). The method of researching differentially expressed miRNAs' host genes helps to research the pathogenesis of cancers.
We can conclude that investigating the differentially expressed genes and miRNAs has great significance in cancer prevention and gene therapy. The differentially expressed network partially reveals the mechanism of EC. If the faulty regulations can be corrected, EC may be prevented and even cured in theory.

Related network of EC
Related network of EC is more complicated than the differentially expressed network because it contains more elements and pathways. Besides differentially expressed genes and miRNAs, the related genes and miRNAs are also involved in EC-related network. New added related genes not only regulate differentially expressed miRNAs but also regulate related miRNAs. New added related miRNAs can target at differentially expressed genes and related genes. For example, hsa-miR-152 is a differentially expressed miRNA, which targets at DNMT1 in differentially expressed network. In related network it targets at added related genes E2F3, MET and MTOR (Banno et al., 2013). The differentially expressed gene TWIST1 has an added miRNA hsa-miR-106b that targets at it in related network. It is suggested that hsa-miR-106b could suppress EC cell invasion by down-regulating TWIST1 expression (Dong et al., 2014). CDC25A, a new added related gene, could be targeted by 5 miRNAs (hsa-miR-21, hsa-miR-424, hsa-miR-449a, hsa-miR-503 and hsa-miR-125b) that are all differentially expressed miRNAs. Hsa-miR-449a may act as a tumor suppressor by targeting CDC25A in EC (Ye et al., 2014). The other 480 miRNAs may have the similar functions to CDC25A. The related network contributes to the understanding of EC additionally.

Global network
The differentially expressed network and the related network are both involved in the global network. The global network contains all the regulatory relations between all the genes and miRNAs. It is an experimentally validated biological network in human body. It can be a reference for the study of the differentially expressed network and related network, especially the first one.

Network of predicted TFs and differentially expressed miRNAs in EC
We got predicted TFs by using P-match method above. They are probably involved in transcription process of EC because they may cause the similar results with differentially expressed TFs in EC. Here the regulatory network of predicted TFs and differentially expressed miRNAs in EC is analyzed in Figure 2. Obviously, 4 predicted TFs including E2F1, E2F3, NFKB1 and ZEB1 are more significant than others in Figure 2. E2F1 could regulate 5 miRNAs and be targeted by 4 miRNAs. NFKB1 could regulate 8 miRNAs and be targeted by 3 miRNAs. ZEB1 regulates 5 miRNAs and is targeted by 4 miRNAs. Specifically, 8 pairs of TF and miRNA form FBLs, which are E2F1 and hsa-miR-106a, E2F1 and hsa-miR-20a, NFKB1 and hsa-miR-9-1, hsa-miR-9-2 and hsa-miR-9-3, ZEB1 and hsa-miR-429, hsa-miR-200b and hsa-miR-200c. In Figure 2, it can be seen that a TF can indirectly influence another TF through regulating some differentially expressed miRNAs. The network of predicted TFs and miRNAs indicates the possible transcription regulation in EC. It may further reveal transcription mechanism of EC.

DOI
PTEN has abundant adjacent nodes in networks of EC. There are 6 miRNAs target at PTEN which regulates 1 miRNA in differentially expressed network. Hsa-miR-21 and PTEN form FBLs. 9 miRNAs target at PTEN and PTEN regulates 2 miRNAs in related network. In global network, 25 miRNAs target at PTEN which regulates 10 miRNAs. The upstream and downstream in three networks show a terraced variation. The miRNAs in the global network provide support for EC's study. The upstream of PTEN influences the downstream indirectly by affecting on PTEN. Furthermore, the upstream and downstream can be affected by their predecessors and successors.

Feature of pathways about differentially expressed miRNAs in EC
Similarly, adjacent nodes of each differentially expressed miRNA were extracted and compared in three networks. Upstream and downstream about each differentially expressed miRNA were listed in table. Take hsa-miR-21 for an example. Hsa-miR-21 is a well-studied oncomiR that is hyperexpressed in many cancers and may be a useful biomarker for cancer diagnosis (Banno et al., 2013).
Hsa-miR-21 is regulated by PTEN and targets at PTEN, ERBB2, TGFBR2, MSH2 and MSH6 in differentially expressed network. In related network, four more genes regulate hsa-miR-21. They are EGFR, NFKB1, REL and RELA. Hsa-miR-21 targets at 10 more genes (CDC25A, CDKN1A, E2F1, E2F2, EGFR, NCOA3, PDCD4, SERPINB5, TGIF1 and TIMP3) in related network. EGFR and hsa-miR-21 form FBLs in related network. E2F1 is a predicted TF in related network, which means hsa-miR-21 may influence more miRNAs by impacting on E2F1. In global network, hsa-miR-21 has relations with more TFs and targets. It has abundant adjacent nodes, which suggests the hsa-miR-21 is crucial in EC.

Feature of pathways about predicted TFs in EC
Then the predicted TFs which were proposed by using P-match method were focused on. Some predicted TFs interact with no miRNA in differentially expressed network and related network, which suggests that the TFs may have little influence. The predicted TFs that have upstream and downstream in differentially expressed network and related network were focused on. Take E2F1 for an example and analyze the pathway. E2F1 regulates 5 miRNAs in differentially expressed network and 7 miRNAs in related network. It is targeted by 4 miRNAs in differentially expressed network and 5 miRNAs in related network. E2F1 is not a differentially expressed TF in EC, but it is suspicious to cause similar changes with differentially expressed TFs so that the study of predicted TFs has the enlightenment significance. Hsa-miR-106a and E2F1, hsa-miR-20a and E2F1 form FBLs in differentially expressed network. Hsa-miR-106b and E2F1 form FBLs in related network additionally.

Discussion
Some special genes, miRNAs and the interactions in EC are brought to the forefront. For example, TWIST1, a helix-loop-helix transcription regulator, is known to induce epithelial-mesenchymal transition and promote tumor metastasis (Dong et al., 2014). Besides, TWIST1 also plays a part in cervical cancer (CC) which is another gynecologic cancer. It likewise regulates hsa-miR-214 and hsa-miR-214 targets at PTEN (Wang et al., 2014). There may be some similarities between EC and CC in molecular level. The predicted TFs in this study may enlighten the research of EC. To get further knowledge of genes and miRNAs in pathways relies on much more experiments. It will help much to discover the pathogenesis of EC based on the network in this study.
In conclusion, this study systematically concludes the relation of genes and miRNAs that differentially expressed in EC. Three regulatory networks which are differentially expressed network, related network and global network are constructed. The differentially expressed network is emphasized to analyze the regulation mechanism. Then upstream and downstream of differentially expressed genes, differentially expressed miRNAs and predicted TFs are extracted. Some remarkable genes and miRNAs are highlighted and their influences on other elements as well as the EC are discussed. Some pathways are proposed logically but remain to be validated by scientific experiments. This study contributes to the research of EC and provides a new vision for further research..