Constructing mRNA, miRNA, circRNA and lncRNA Regulatory Network by Analysis of Microarray data in Breast Cancer

Background. Luminal tumors are the utmost frequent subtype of breast cancer (BC). Despite luminal BC has relatively good prognosis, in a subset of patients, disease relapse occursto endocrine therapy ;hence, there is a critical need to identify new strategies to promote the early detection and more effective therapies. Noncoding RNAs including microRNAs, long noncoding RNAs, and circular RNAs can interact with and modulate each other via diverse molecular mechanisms and make a complicated regulatory network.ncRNAs participate in diverse biological processes and disorders such as breast tumors. Therefore, understanding their regulatory mechanisms allow to develop new �eld of research and therapeutic options for BC patients. Methods. In this study, BC-specic RNA expression pro�les including mRNAs, miRNAs, lncRNAs, and circRNAs were retrievedfrom Gene Expression Omnibus microarray datasets, and differentially expressed(DE) items were obtained. Disease ontology, functional and pathway enrichment analyses were executed. The protein-protein interaction network was constructed, and hub mRNAs were extracted.The prognostic value of hub mRNAs in patients of BC were performed. Subsequently, a ceRNA network was established.


Introduction
Female breast cancer remains the most common neoplasm throughout the world and is the fth leading cause of cancer mortality in women (1).Estrogen receptor alpha is expressed in about 70% of breast cancers which are assigned as luminal A or luminal B subtypes BC (2).Luminal cancers have a better prognosis than other types of BC and are sensitive to endocrine therapy, however many patients eventually develop resistance (3).This situation is attributed to a complex regulatory network of breast cancer.Thus, the discovery and development of new molecular biomarkers and therapeutic targets is imperative in medical research and could potentially improve cancer diagnosis and treatment.During the last decades, advances in high-throughput technologies have made possible the development of new strategies in unraveling puzzles of the complex genetic nature of cancer.Therefore, studies have eventually shifted our conception of noncoding RNAs from worthless transcriptional products to major players that regulate many biological processes (4)(5)(6).
As novel non-coding RNAs (ncRNAs), circRNAs in contrast to linear RNAs are spliced backward from the 5′ to 3′ segment, with covalently closed rings lacking the 5′ cap and 3′ poly(A) tail.Compared to linear RNAs, this structure can be protected, against degradation by the exonucleolytic activity of RNase so their transcripts remain highly conserved and stable (7,8).Recent reports have revealed that aberrant expression of circRNAs is usually associated with diverse pathological conditions, such as neurodegenerative diseases (9), cardiovascular diseases (10), and cancers (11,12) with diagnostic and prognostic value.Although the exact mechanisms of most circRNAs remain unclear, several studies suggest circRNAs are responsible for multiple biological processes via different mechanisms such as gene transcription modulation and interference with RNA splicing (13,14), gene expression regulation at the RNA level through suppressing miRNAs as competing for endogenous RNA (ceRNA) (14), serving as protein scaffolds (15) and translation by encoding functional peptides (16).Previous studies in recent years have demonstrated that the circRNA-miRNA-mRNA regulatory network act as key regulators in breast cancer growth and invasion (17,18).For example, circABCB10 was signi cantly upregulated in BC tissue samples.CircABCB10 substantially promotes BC expansion and dissemination via the sponging of miR-1271(18).However, the functions of circRNAs in BC development and progression are still poorly understood.In addition to circRNAs, long non-coding RNAs as ncRNAs can reduce the miRNA suppression of targeting mRNA and indirectly regulate the target gene expression level.Plenty of evidence indicates that lncRNAs abnormally express in many cancer types and might be a prominent contributor to tumor progression (19)(20)(21)(22).In the current study, we established a breast cancer ceRNA network for both lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA based on bioinformatic prediction methods.In the beginning, we collected signi cant differentially expressed mRNAs, miRNAs, lncRNAs, and circRNAs according to the analysis of GEO database information.Then, we ltered circRNAs and lncRNAs that could bind to miRNAs and miRNA target genes to design a circRNA/lncRNA/miRNA/mRNA network.Besides, we reconstructed a proteinprotein interaction (PPI) network and performed functional and pathway enrichment analyses.Figure 1 shows the overall owchart of our in silico assessments.
Screening of differentially expressed items.All raw expression data les were subjected to background correction and quantile normalization using RMA (Robust Multi-Array Analysis) ) from affy Bioconductor package.(25,26).
The "limma" R package (27) was used to identify differentially expressed items between luminal tumor and normal samples.Differentially expressed genes (DEG) were collected with thresholds of adj.P-value < 0.01 and |log 2-fold-change| ≥ 2. The threshold values for differentially expressed miRNAs (DEmiRs) were set to |logFC|>1 and adj.Pvalue < 0.05.Based on the cut-off value, adjusted P-values < 0.05 and |log fold change (FC)|> 1.2 were used to select differentially expressed circRNAs (DEcRNAs).Differentially expressed lncRNAs(DELs) were selected according to (|logFC|) > 0.5 and the adjusted P-value < 0.01 cutoff criteria between normal and luminal tumor samples.Then, volcano plots were performed using R software ggplot2 package to visualise the results of differential expression analysis.

Prediction Of Lncrnas
In addition to selected differentially expressed lncRNAs based on the GSE45827, the interactions between lncRNA and DEmiRNAs were evaluated by DIANA TOOLs databases (https://diana.e-ce.uth.gr/lncbasev3).

Prediction Of Mirna Target Genes
The DEmiRNA target genes predictions were retrieved from an online analysis tool called MiRWalk 2.0 (http://mirwalk.umm.uniheidelberg.de),Since the results of MiRWalk 2.0 were derived from different miRNA-target prediction programs (DIANA-microTv4, DIANA-microT-CDS, miRanda, mirBridge, miRDB4, miRmap, miRNAMap, PicTar2, PITA, RNA22v2, RNAhybrid, and Targetscan), their projections were very reliable(28).If there was no interaction between differentially expressed genes (DEGs) obtained from the GEO database and predicted genes in the database, they could be removed after functional enrichment analysis and network centrality analysis.

Construction Of The Ppi Network
To construct a PPI network among DEGs identi ed, we used the Search Tool for the Retrieval of Interacting Genes (STRING) database (version 11) (29).The STRING database provides protein-protein interaction based on coexpression experiments, gene fusion, text mining, co-occurrence, and, neighborhood regions as the forecasting methods.DEGs with a joined score > 0.9 were selected to create a network using Cytoscape (30).Then, the network was processed for module analysis, using Molecular Complex Detection (MCODE) Plugin in Cytoscape software with default parameters as follows: degree cutoff ≥ 2, node score cutoff ≥ 2, K-core ≥ 2, and maximum depth = 100 (31,32).Cytoscape (v3.8.3) was used to visualize and compute the main characteristics of the PPI network to explore the hub genes, such as average clustering coe cient distribution, closeness centrality, average neighborhood connectivity, node degree distribution, shortest path length distribution, and topological coe cients.

Functional Enrichment Of Degs
To further examine the possible function of mRNAs in the ceRNA network, disease ontology (DO), functional enrichment of gene ontology (GO) including the cellular component (CC), molecular function(MF), and biological process (BP), Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analyses were conducted using cluster Pro ler, and DOSE in R (20).statistical signi cance was represented by the adjusted P-value (Q-value) < 0.05.Also, assessing miRNAs regulatory roles of DEmiRs was performed with DIANA-miRPath (https://www.microrna.gr/miRPath),an online software.

Survival Analysis Of Candidate Hub Genes
The clinical outcomes of hub mRNAs between BC samples and normal tissues were evaluated using the GEPIA (Gene Expression Pro ling Interactive Analysis) webserver.It develop new possibility for data exploration of gene functions based on records that have been arranged from TCGA ana GTEx projects (33).

Construction Of The Cerna Network
Based on the potential interaction between mRNAs, miRNAs, circRNA, and lncRNAs, the ceRNA network was visualized by Cytoscape software.

Results
Microarray data preprocessing.Following quality assessment and removing inaccurate expression data, 691 DEGs Subsequently, 20 circRNAs were selected based on adjusted P-value, consisting of 10 down-regulated and 10 upregulated circular RNAs.All data are presented in the Supplementary le (Table 1_4).
Identication of circRNA-miRNA interactions.Evidence suggests that some circular RNAs containing speci c miRNA binding sites can reduce miRNA suppression to target mRNA, thus playing an important role in tumorigenesis.We predicted miRNAs that could trap by circRNAs to construct the circRNA-miRNA regulatory network using three online databases, CSCD, CircBank, and CircInteractome.Latter, we selected circRNAs that could intersect with more than two DEmiRNAs.Therefore, 24 circRNA-miRNA pairs consisting of 7 DEcircRNAs (3upregulated and 4 downregulated) and 14 DEmiRNAs were maintained.The fundamental structure of the 7 circRNAs including MRE (microRNA response element), RBP (RNA binding protein), and ORF (open reading frame) elements are exhibited in Figure .3.The basic characteristics of the 7 circRNAs are listed in Table 1.DEmiRNAs, they would be eliminated.Then, lncRNAs that could intersect with more than two DEmiRNAs were selected.Using this technique, 37 lncRNA-miRNA interaction pairs including 9 DElncRNAs and 8 DEmiRNAs were identi ed.The basic information of the 9 lncRNAs is listed in Table 2.

Ppi Network Construction And Module Analysis
To predict the interaction relevance, the STRING database was applied.As a result, 629 nodes and 4252 protein pairs with a combined weight score of > 0.4 were found in the network (Supplementary Table 5).All nodes with a combined score of > 0.9 were imported into Cytoscape software for visualization.After clustering analysis with MCODE, six modules with a score of > 5 were detected(Supplementary Table 6).After centrality analysis, the top 50 nodes with the degree, closeness, and betweenness indices values, higher than the mean value of the whole network, were considered as key nodes.Figure 4 shows the PPI network of the selected DEGs.

Functional Annotation
Enrichment analysis renders an approach for facilitating the biological interpretation of outcomes from highthroughput datasets (34).To obtain the biological functions of the 50 top DEGs, DO functional, GO analyses, and KEGG pathway enrichment was conducted using the cluster Pro ler package of R software.The functional characterization of miRNAs was exploited using DIANA-miRPath v3 (Supplementary Table 7).Using disease ontology analysis of 50 top DEGs, revealed that 23 of them were associated with breast carcinoma, hereditary breast, and ovarian cancer, and breast disease (P < 0.05).Gene ontology enrichment in BP terms revealed that these DEGs were mainly involved in response to steroid hormone, regulation of protein serine/threonine kinase activity, response to ketonea antibiotic, response to oxygen levels, Fc receptor signaling pathway, positive regulation of T cell activation, positive regulation of leukocyte cell-cell adhesion, regulation of DNA metabolic process, leukocyte cell-cell adhesion, cell junction assembly, and positive regulation of cell-cell adhesion.For CC, DEGs were mainly enriched in focal adhesion, cell-substrate adherens junction, and cell-substrate junction.For MF, DEGs were mainly enriched in RNA polymerase II core promoter sequence-speci c DNA binding, protease binding, and integrin binding (P < 0.05).In the KEGG pathway analysis, the top enriched pathways included bacterial invasion of epithelial cells, focal adhesion, proteoglycans in cancer, Rap1 signaling pathway, regulation of actin cytoskeleton, PI3K-Akt signaling pathway, tight junction, and ECM-receptor interaction (P < 0.05).The results are demonstrated in Fig. 5A-C.In addition, the cluster Pro ler package of R software and DIANA MirPath web server was used to conduct GO analysis and KEGG pathway analysis for 14 miRNAs, respectively.For BP, 14 DEmiRs were mainly enriched in the regulation of angiogenesis, regulation of endothelial cell migration, and regulation of epithelial cell migration (P < 0.05).Moreover, the top ve KEGG analysis pathways included Proteoglycans in cancer, Adherens junction, Cell cycle, FoxO signaling pathway, and Hippo signaling pathway (P < 0.05) ( Fig. 5D).
Prediction of miRNA target genes.MiRwalk database was searched to determaine the interaction between 14 DEmiRs and 691DEGs.If the predicted genes from the database were not considered as hub nodes, they would be removed from our list.Using this approach, 821 miRNA-mRNA interaction pairs were obtained between 14 miRNA and 691 differentially expressed genes.Also, 43 interaction pairs were identi ed between 14 miRNAs and 12 hub mRNAs.

The Prognostic Value Of Hub Mrnas
The clinical outcomes of key genes were evaluated through survival analysis.GEPIA webserver was applied to determine whether the expression of hub genes were related to the breast cancer patients' overall survival (OS) and disease-free survival (DFS).This analysis indicated that for FOXO3, lower expression is closely related to improved OS (Logrank p = 0.0091; Fig. 6A).Besides, overexpression of RHOA is slightly associated with better DFS in BC patients (Logrank p = 0.015; Fig. 6B).Moreover, higher expression of KIT is associated with slightly higher survival rate in BC patients and this effect was altered over time (Logrank p = 0.016; Fig. 6C).All the other hub genes did not signi cantly in uence the prognosis of patients with breast cancer.

Intermodulation Of Ncrnas And Their Targeted Gene
Non-coding RNA as main actors in gene regulatory networks, directly or indirectly regulate the function of mRNAs and the disruption of these regulatory networks which are reported in various types of cancer (35)(36)(37).Here, we showed the intermodulation of the distinct regulatory mechanism among various ncRNA and their targeted genes in breast tumors.
Based on the present study, circ_0007535, circ_0014130, and circ_0007001 circRNAs, which could intersect with four miRNAs were selected.circ-0007535 and circ_0014130 were down-regulated, while circ_0007001 was up-regulated in cancer tissues versus normal tissues.In our predicted circRNA-miRNA network, miR-183, miR-188, miR-7, and miR-630 were recognized as target miRNAs for circ_0007535.There is uncertainty about the action mechanism of miR-183 in cancers.miR-183 has been reported to work as an oncogene or tumor suppressor in a wide range of human cancers(38).Several studies suggested that miR-183 acts as an oncogenic factor in human breast cancer thereby promotes cell proliferation and inhibits apoptosis in BC.For example, Yan Cheng et al found that miR-183 was upregulated in breast tumors, enhances cell proliferation migration, invasion, and inhibits cell apoptosis by targeting PDCD4 (39).Besides, the miR-183/-96/-182 cluster is highly expressed in most breast cancers and promotes cell proliferation and migration (40).What's more, overexpression of mir-183 and mir-494 serves a critical role in BC metastasis and increases the growth and spread of cancer cells in human BC cell lines (41).In our results, miR-183 was up-regulated in breast cancer tissues and based on mirwalk prediction, it can regulate the expression of FOXO3, RHOA, RAC1, CAV1, IGF1, NCOA3, CCNB1, and ITGB1.Along with circ_0007535, we found three possible DEcircRNAs, circ_0002727, circ_0044927, and circ_0089153 can intercommunicate with miR-183 and thereby regulate the expression of above 12 DEGs.Multiple studies show that miR-188 as a potent tumor suppressor, induces apoptosis, and inhibits proliferation in breast cancer cells (42).In our results, there was down-regulation of miR-188 in breast cancer tissues.We identi ed four probable crosstalk between these miRNAs and circRNAs, including circ_0007535-miR188-FOXO3/RHOA/IGF1, and circ_0002727-miR188-FOXO3/RHOA/IGF1.In the current study, it was found that circ_0014130 may adsorb corresponding miRNA through engaging with miRNA binding sites.It's MREs were detected via online tools including miR-1207, miR-200b, and miR-200c.As a target miRNAs, miR-1207-5p act as an oncogene and stimulates BC cell growth and proliferation by targeting STAT6 (43).In triple-negative breast cancer (TNBC), antagomir-1207-5p/chemotherapy combination increases sensitivity to treatment (44,45).Instead, miR-1207-5p acts as an antioncogene by impairing cell spreading, migration, and invasion in gastric cancer (46) and lung cancer (47).However, in our results, miR-1207 was down-regulated in luminal cancer tissues versus normal tissues, and further study required to explain the action mechanism of miR-1207 in luminal tumors.In this paper,we predicted miR-1207-5p as a target for circ_0014130 and hsa_circ_0007001 which may present a promising strategy for BC regulation.In addition, miR-200b-3p and miR-200c were associated with the ER status of BC cells and were down-regulated in the basal TNBC compared to ER+/epithelial cancer cell lines (48).They can repress a program of mesenchymal genes to keep an epithelial state and inhibit metastasis in breast cancer (48-50).Consequently, triplenegative breast cancer cells could bene t from the restoration of these microRNAs.Based on our assessment, there are links between circ_0007001 and miR-200b-3p in breast tumors, thus, inhibition of circ_0007001 by RNA interference (RNAi) may be an interesting target for TNBC treatment.
In this study, also a second regulatory network, interactions between differentially expressed lncRNAs, miRNAs, and mRNAs were determined.In our in silico assessments, among 9 selected DElncRNAs, ve including NORAD, MALAT1, TUG1, ZFAS1, OIP5-AS1 could intersect with more than three DEmiRNAs like miR-183, miR-182, miR-  Consequently, triple-negative breast cancer cells could bene t from the restoration of these microRNAs.Based on our assessment, there are links between circ_0007001 and miR-200b-3p in breast tumors, thus, inhibition of circ_0007001 by RNA interference (RNAi) may be an interesting target for TNBC treatment.

Discussion
Non-coding RNA regulatory networks as bridges between DNA and protein play signi cant roles in speci c biological processes.Although nearly 40 years have elapsed since circular RNAs were rst discovered, they were initially neglected as functionless byproducts of errors in the normal splicing process(63, 64).Nonetheless, in the past few years, continuous achievements in sequencing technology and new bioinformatics algorithms leading to the enigma of circRNAs have gradually been unveiled that these molecules serve a signi cant role in pathogenesis and progression in a broad spectrum of human diseases, in particular in cancer(65-68).Furthermore, circRNAs show valuable opportunity as diagnostic and prognostic biomarkers, due to high levels of evolutionary conservation and abundance, stability, and tissue-speci city (69, 70).For instance, the plasma level of circ_0001785 in breast cancer patients compared to healthy donors is upregulated and acts as a diagnostic biomarker for BC detection (71).Based on the evidence, circRNAs have great therapeutic potentials.Thus, the application of RNA interference (RNAi) for oncogenic circRNAs and induction of tumor-suppressive circRNAs expression for tumor suppressive circRNAs in cancer cells or tissue could have anticancer effects (72)(73)(74).Additionally, lncRNAs would serve as promising biomarkers for clinical diagnosis and prognosis (75-78).Research has indicated that circular RNAs share similar features and biological functions with lncRNAs.For example, they regulate mRNA expression by adsorbing miRNAs like miRNA sponges (3,79), serve as protein scaffolds (15,80), remained stable in extreme conditions over a relatively long time(81, 82), and are associated with a variety of diseases(83, 84).Numerous studies that simultaneously investigate these regulatory networks are required to further discover the complexity of the networking between different types of non-coding RNAs to nd new opportunities for the therapeutic management of breast cancer.
In this study, based on the potential interaction between mRNAs, miRNAs, circRNAs, and lncRNAs, we constructed a ceRNA regulatory network.First, breast cancer-speci c RNAs, including mRNAs, miRNAs, lncRNAs, and circRNAs were ltered and differentially expressed items were obtained.Next, we identi ed seven circRNAs could interact with more than two of DEmiRNAs (hsa_circ_0007535, hsa_circ_0002727, hsa_circ_0005240, hsa_circ_0014130 ,hsa_circ_0044927, hsa_circ_0007001, hsa_circ_0089153).As reported previously, some circRNAs can regulate the expression of the target genes by competitive binding with miRNA response elements to inhibit their expression(85).
Thus, we predicted MREs in seven circRNAs mentioned through CSCD, CircInteractome, and CircBank online tools.
As part of this study, to better explain the action mechanism of nRNAs in gene expression regulation by targeting miRNA, we constructed a PPI network.Following centrality analysis, we determined the top 50 nodes based on degree, closeness, and betweenness indices.It was necessary to prune the fty identi ed genes to make the study more reliable and to achieve this goal, enrichment analyses were performed.The functional enrichment and pathway analyses elucidated these genes were related to many signi cant tumor-associated pathways.Additionally, we obtained 23 breast tumor-associated genes depend on disease ontology.In the nal analysis, 12 hub genes were picked out based on centrality indices, modularity analysis, DO, and functional enrichment analyses including FOXO3, RHOA, EZH2, KIT, HSP90B1, NCOA3, RAC1, IGF1, CAV1, CXCR4, CCNB1, and ITGB1.
Among these hub-genes, FOXO3 shows considerable impacts on OS between the BC and normal tissue groups.Moreover, high expression of RHOA is slightly associated with a higher rate of disease-free survival in BC patients.
FOXO3, a member of the Forkhead box O (FoxO) transcription factor family, is important mediator of many different bioprocess like cell cycle progression, apoptosis, survival, and DNA damage.FOXO3, also known as FOXO3a, play a role a tumor suppressor in variety of human cancers and has a protective role in ER + breast tumors.Many evidences have revealed that FOXO3a serves as a prognostic biomarker in multiple cancers, especially in Luminal-like BC.While tamoxifen therapy is is still a good choice for ER + breast cancer patients, in many cases, acquired resistance to antihormone therapy is inevitable (93,94).In tamoxifen-resistant BC, upregulation of growth factor signaling pathways, activation of phosphatidylinositol 3-kinase (PI3-K)/Akt, including PTEN down-regulation, make the growth of resistant cells easier.FOXO3a works downstream of PI3K/AKT pathway and is inactivated by overactivation of the PI3K/Akt pathway in the most of breast tumors (95,96).Meanwhile, dysregulation of the PI3K/AKT signaling pathway is one of the most frequent oncogenic aberrations in TNBC (94).Noteworthy, FOXO3a may be an attractive therapeutic target, especially cells like TNBC and also in tamoxifen resistance breast tumors.Several miRNA binding sites in 3′UTRs of FOXO3a can greatly increas the level of regulation (94,97) ).RhoA and Rac1 have been described as oncogenes due to their overexpression within malignant tumors such as breast cancer, and their associations to metastasis, migration, invasion, cell cycle progression, drug resistance, and diverse clinical outcomes (101,102).Hence, targeting RHOA1and RAC1 represents a right approach for suppressing cancer cell proliferation (103,104).Our results showed that RHOA and RAC1were up-regulated in breast cancer tissues.
Parallelly, microarray analysis revealed that circ_0089153, circ_0007001, circ_0044927, NORAD, MALAT1, TUG1, ZFAS, OPI5-AS1 were up-regulated in breast cancer tissues, and might serve as oncogenic ncRNAs by competitive binding with miRNAs(miR-342, miR-101, miR-200b, and miR-200c) to inhibit their expression and therefore increase the expression of RHOA and RAC1 genes.In addition, miR-1207, which is down-regulated in breast cancer tissues, is a negative regulator of RAC1, thus overexpression of miR-1207 may lead to signi cant tumour regression.Therefore might be anticipated that application of precise RNA interference (RNAi) to oncogenic circRNAs and lncRNAs or targeted delivery of miRNAs to reduce the mRNA levels of oncogenes could develop an effective means for inhibiting tumor growth and aggression.

Conclusion
The present study listed several ncRNAs, which could bind to miRNA to modulate breast cancer development.Here, regulatory networks were designed for both lncRNA-miRNA-mRNA and circRNA-miRNA-mRNA and highlighted the importance of cooperation between circRNAs with lncRNAs which could regulate the same miRNAs and could might target RHOA to control BC progress (Fig. 8).
The current study provides a new perspective into the possible molecular mechanisms among various RNA crosstalks underlying the carcinogenesis and progression in luminal breast tumors and especially endocrineresistant breast cancer patients.The aim of this study was to revive researchers' enthusiasm in studying the potential role of ncRNAs as novel prognostic biomarkers and to make new treatment decisions for breast cancer patients.Improving our understanding of the nature of ceRNA crosstalk makes new opportunities to establish ceRNA-based therapeutic purposes.However, these ndings need further experimental validation and con rmation in the future.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for
Not applicable.
Protein-protein interaction analysis of differentially expressed genes.Six modules with a score of >5.0 were depicted in the gure in unique shapes and colors.Edges represent interactions between two genes.The signi cance of protein nodes in the network is presented by a degree where small and large sizes denote low and high values, respectively.The border color indicates the fold change for nodes where upregulated nodes are de ned in red and downregulated nodes are de ned in green.

Figures Figure 1 Flowchart
Figures

Figure 3 Basic
Figure 3

Figure 5 Functional
Figure 5

Table 1
Basic characteristics of these 7 circRNAs.Based on the potential interaction between miRNAs and lncRNAs, the interactions between lncRNA and 14 DEmiRNAs that linked with DEcircRNAs, were detected by DIANA TOOLs databases (https://diana.e-ce.uth.gr/lncbasev3).To improve the reliability of data, two lters (breast tissue selection and high con dence level) were applied.If the predicted miRNAs in the database were not matched with 14 A total of 203 lncRNA probes were identi ed in GSE59247 datasets with adjusted P-value < 0.01.Next, 122 lncRNAs probe ID with |logFC| > 0.5 and adjusted P-value < 0.01 were detected among 11 normal and 130 breast tissue samples.

Table 2
Basic characteristics of the 9 lncRNAs which are dominated as important molecules in breast cancer regulatory network.
(52)miR149, miR-200b, miR,200c, miR-101, and miR-342.All of the above ve LncRNAs were upregulated and could function as oncogenic ncRNAs contributing to the progression and invasivasion of breast tumors through modulating downstream pathways.Therefore, it is imperative to explore the molecular pathways implicated in BC drug resistance or tumor metastasis to re ne therapeutic outcomes.Increased expression of lncRNA-NORAD in breast cancer tissues facilitates proliferation and invasion of breast cancer cells and is associated with poor prognosis.The study by Ke Zhou et al. showed that lncRNA NORAD was overexpressed in breast cancerous tissues and stimulated BC cell progression by regulating the TGF-β pathway(51).We showed that NORAD could promote a lncRNA associated with diverse types of cancer such as breast tumors.The expression and action mechanism of TUG1 in breast carcinoma is a quite arguable subject.The study by Teng Li. demonstrated that the TUG1 expression was enhanced in breast cancer tissues and highly invasive BC cell lines and was related to clinical and pathologic tumor characteristics including tumor size, distant metastasis, and TNM staging(52).Whereas other studies reported downregulation of TUG1 in triple-negative BC and that its expression increases cisplatin sensitivity (57)55) cells(53).Based on our results, TUG1 could regulate the expression of 12 hubs DEmRNAs including FOXO3, RHOA, EZH2, KIT, HSP90B1, NCOA3, RAC1, IGF1, CAV1, CXCR4, CCNB1, and ITGB1by trapping miR-183, miR-182, miR-7, and miR-101.In most studies, Metastasis Associated Lung Adenocarcinoma Transcript 1 (MALAT1) is commomly overexpressed and acts as a poor prognosticator lncRNAs in some types of cancers, including breast tumors(54,55).The action mechanism of MALAT1 on breast cancer cell proliferation is disputable, as reported that MALAT1 can act as metastasis promoter (56) and cell proliferation suppressor(57).The controversial evidence in the role of MALAT1 in tumorigenesis from distinct studies may be based on the speci c tumor subtypes or different cell types (58, 59).Likewise, MALAT1 might regulate the above mentioned hub DEGs by trapping miR-183, miR-342, miR-7,miR-200b, miR-200c, and miR-101.The study by Haibing Xiao et al., revealed that MALAT1 can modulate ZEB2 expression via trapping miR-200s as a competing endogenous RNA and serve as a possible therapeutic target in clear cell kidney carcinoma (60).Similarly, MALAT1 promotes melanoma expansion by inhibiting the expression and function of miR-183 and ITGB1 signal activation(61).Furthermore, the repressive function of miR-101 on the 7,   miR149, miR-200b, miR,200c, miR-101, and miR-342.All of the above ve LncRNAs were upregulated and could function as oncogenic ncRNAs contributing to the progression and invasivasion of breast tumors through modulating downstream pathways.Therefore, it is imperative to explore the molecular pathways implicated in BC This study listed many ncRNAs, which might bind to miRNAs to regulate breast cancer progression.All potential circRNA-miRNA-mRNA and lncRNA-miRNA-mRNA crosstalks in breast cancer are summarized in Table3.
(53) resistance or tumor metastasis to re ne therapeutic outcomes.Increased expression of lncRNA-NORAD in breast cancer tissues facilitates proliferation and invasion of breast cancer cells and is associated with poor prognosis.The study by Ke Zhou et al. showed that lncRNA NORAD was overexpressed in breast cancerous tissues and stimulated BC cell progression by regulating the TGF-β pathway(51).We showed that NORAD could promote the occurrence and development of breast tumors by adsorbing miR-183, miR-182, miR-7, miR149, miR,200c, miR-101, and miR-342 as a sponge to regulate the expression of 12 hubs DEmRNAs including FOXO3, RHOA, EZH2, KIT, HSP90B1, NCOA3, RAC1, IGF1, CAV1, CXCR4, CCNB1, and ITGB1.Another lncRNA, taurine upregulated 1 gene (TUG1) is a lncRNA associated with diverse types of cancer such as breast tumors.The expression and action mechanism of TUG1 in breast carcinoma is a quite arguable subject.The study by Teng Li. demonstrated that the TUG1 expression was enhanced in breast cancer tissues and highly invasive BC cell lines and was related to clinical and pathologic tumor characteristics including tumor size, distant metastasis, and TNM staging(52).Whereas other studies reported downregulation of TUG1 in triple-negative BC and that its expression increases cisplatin sensitivity in TNBC cells(53).Based on our results, TUG1 could regulate the expression of 12 hubs DEmRNAs ZEB2 expression via trapping miR-200s as a competing endogenous RNA and serve as a possible therapeutic target in clear cell kidney carcinoma (60).Similarly, MALAT1 promotes melanoma expansion by inhibiting the expression and function of miR-183 and ITGB1 signal activation(61).Furthermore, the repressive function of miR-101 on the autophagy and growth of colorectal cancer cells was abrogated by MALAT1(62).Consequently, NORAD, TUG1, and MALAT1 may be desirable biomarkers for breast cancer diagnosis and therapeutic targets in breast cancer management.It is noteworthy that these ncRNAs could interact with other 378 DEGs in the network and construct a complex regulatory RNA network.