Background
Breast cancer (BC) is a heterogeneous and malignant neoplasm derived from breast tissue, and accounts for about 16% of all cancers and 22.9% of invasive cancers in women [
1]. The most common cause of the BC-related mortality is metastasis [
2,
3]. Currently, BC diagnosis and prognosis is evaluated on the basis of disease stage, histological grade, and the expression level of hormone receptors [
4]. However, clinical and pathological symptoms have limited predictive value in detecting early BC, and the clinical outcomes are highly variable on account of its heterogeneity. In addition, the underlying molecular mechanisms of BC still remain unclear. Therefore, it is vital to identify potential molecular diagnostic markers and/or therapeutic targets to combat BC, especially the invasive form.
Long non-coding RNAs (lncRNAs) is a class of ncRNA over 200 nucleotides long [
5], and are reportedly involved in a number of cellular processes, for example, transcriptional and post-transcriptional regulation [
5,
6]. Due to their strong tissue specificity, lncRNAs are potentially effective early diagnostic biomarkers of various cancers [
7]. Identification of a BC specific lncRNA biomarker may therefore be of clinical significance for the diagnosis and prognosis of BC. Several lncRNAs have been reported to be associated with BC initiation and progression [
8,
9], and although some have been found to predict clinical outcomes for BC, the results are inconsistent due to limited tissue samples. Furthermore, studies without large sample size are also not able to determine with statistical power whether these lncRNAs are associated with survival or other clinical factors. The Cancer Genome Atlas (TCGA) is an open-access and large-scale database which can provide multidimensional molecular profiles for a large number tumor samples. To increase the statistical reliability of our studies, we identified BC specific lncRNAs using data obtained from TCGA database.
The competing endogenous RNA (ceRNA) hypothesis presented by Salmena et al. [
10] was proposed as a novel regulatory mechanism between ncRNA and coding messenger RNA. LncRNAs contain miRNA-response elements (MREs) which function as ceRNAs, and play a key role in various pathological processes like tumorigenesis [
11]. Zhang et al. [
12] have proved the biological role of lncRNA related-ceRNAs in glioblastomas. A recent study has demonstrated that lncRNA NUTF2P3-001 acts as a ceRNA to communicate with KRAS by competitively binding to hsa-mir-3923, and the up-regulation of NUTF2P3-001 reverses the suppressive effect of hsa-mir-3923 on KRAS, leading to the proliferation and invasion of pancreatic cancer [
13]. In addition, the aberrant expression of 7-lncRNA signature (called LncRisk-7) led to differential gene expression via a dysregulated lncRNA-associated ceRNA network, contributing to pancreatic ductal adenocarcinoma progression [
14]. Collectively, these findings show that dysregulation of important lncRNAs in the ceRNA network also disrupt the miRNA-mediated lncRNA/mRNA ceRNA interactions and therefore contribute to cancer initiation and progression [
15,
16]. Nevertheless, very little information is available on BC ceRNAs.
In our work, RNA sequencing data of 1109 BC samples and 113 adjacent non-tumor breast tissues samples were retrieved from the TCGA database. To the best of our knowledge, this is the first study to use large scale sequencing database (TCGA) and ceRNA network to identify BC-specific lncRNAs. This new approach of predicting cancer specific lncRNA and ceRNA networks can elucidate the lncRNA-mediated ceRNA regulatory mechanisms in the development and prognosis of BC, and identify novel lncRNAs as potential diagnostic biomarkers or therapeutic targets.
Discussion
BC is a common malignant gynecological cancer, and is one of the main causes for the cancer-related deaths in women [
29]. The lack of specific diagnostic and prognostic biomarkers may contribute to the current low survival rate among BC patients. To improve clinical outcomes therefore, it is essential to explore the exact regulatory mechanisms of BC initiation and progression, and to identify the potential BC-related prognostic signatures that predict those outcomes. Growing experimental evidence indicates that lncRNAs play important roles in many biological processes, and ceRNA activity is closely related to the development of cancers [
30,
31].
In recent years, some studies have investigated the ceRNAs in BC. For instance, Chen et al. [
32] analyzed the BC ceRNA network on the basis of common miRNAs as well as co-expression, but did not consider miRNA expression. Another study also established a BC specific ceRNA network to investigate its underlying molecular mechanisms based on the PCC of miRNA–mRNA pairs [
33]. However, both studies focused on the roles of mRNAs rather than that of lncRNAs in the BC ceRNA networks. In 2018, Zhou et al. [
34] constructed four BC-related ceRNA networks by combining the miRNA targets and the expression data of lncRNAs, miRNAs and mRNA, but they did not take into account the relationship between survival and lncRNAs, nor construct the prognostic signature. In our study, in addition to constructing the ceRNA networks by combining lncRNA, miRNA, and mRNA expression data, we also investigated the association of lncRNA and OS in BC patients. Furthermore, based on the theory of ceRNA network, we established the 4-lncRNAs prognostic signature. With the goal of identifying lncRNAs significantly associated with OS, we established an lncRNA–miRNA–mRNA ceRNA network using the information obtained from the TCGA database. Univariate regression analysis on the DElncRNAs of the ceRNA network identified 7 lncRNAs—ADAMTS9-AS1, AC061992.1, LINC00536, HOTAIR, AL391421.1, TLR8-AS1 and LINC00491—that were associated with OS. Multivariate analysis showed significant prognostic value of 4 of those lncRNAs (ADAMTS9-AS1, LINC00536, AL391421.1 and LINC00491) in the OS of BC patients. A cumulative risk score of the 4 lncRNAs was calculated, which indicated that this 4-lncRNA signature independently predicted OS in BC patients. To the best of our knowledge, this is the first report integrating a ceRNA network with TCGA data to build an lncRNA-related risk score, and evaluate the OS of BC patients. Our study will help improve the understanding of lncRNA-mediated ceRNA regulatory mechanisms in BC and identify novel lncRNAs as therapeutic targets.
In the current study, among this 4-lncRNA signature, ADAMTS9-AS1 was demonstrated to play important roles in the progression and prognosis of cancer. ADAMTS9-AS1 is an antisense lncRNA, and growing evidence has implicated that a large amount of antisense lncRNAs play crucial roles in the cancer [
35,
36]. Li et al. [
37] reported that ADAMTS9-AS1 could predict the survival status of patients with esophageal squamous cell carcinoma. Another study has also indicated a prognostic role of ADAMTS9-AS1 in patients with colon adenocarcinoma [
38]. In addition, ADAMTS9-AS1 has been demonstrated to be a risk lncRNA in ovarian cancer, which is involved in the progression of ovarian cancer [
39]. In our study, we noticed that ADAMTS9-AS1 with low-expression could compete with up-regulated miRNAs (hsa-mir-182, and hsa-mir-21), to regulate the expression of the target genes such as CHL1, SPRY2, and TCEAL7 involved in the ceRNA network. Previous studies have shown high expression of hsa-mir-182 in MCF-7 breast cancer cells [
40,
41]. In addition, the high-expression of hsa-mir-21 was reported to be correlated to the metastasis and poor prognosis of BC patients [
42]. The remaining three lncRNAs of the ceRNA network (LINC00536/AL391421.1/LINC00491) were up-regulated and competed with the decreased hsa-mir-204 and hsa-mir-144 levels. Down-regulation of has-mir-204 has been suggested to enhance cell proliferation and invasion in gastric cancer [
43], and low-expression of has-mir-204 is related to the poor prognosis of acute myeloid leukemia patients [
44]. In addition, decreased expression of has-mir-144 is strongly correlated with the progression of colorectal cancer [
45]. No study so far has reported any association of LINC00536, AL391421.1 or LINC00491 with cancer. This is the first study to show aberrant expression of ADAMTS9-AS1, LINC00536, AL391421.1 and LINC00491 in BC, and indicates a potential prognostic role of this 4-lncRNA signature in BC. In addition, the bioinformatics based investigation of lncRNAs will be helpful in future experimental studies.
Although the findings of our study have important clinical implications, the limitations must also be noted. First, a longer follow-up duration is required to verify our results, and second, the findings based on the TCGA database will need to be verified using other experimental methods. In addition, the biological roles of ADAMTS9-AS1, LINC00536, AL391421.1, and LINC00491 in BC also need to be further investigated.