Three same-sample expression profiles of TFs, genes and lncRNAs were downloaded from The Cancer Genome Atlas (TCGA; https://tcga-data.nci.nih.gov/; https://portal.gdc.cancer.gov/exploration?filters=%7B''op''%3A''and''%2C''content''%3A%5B%7B''op''%3A''in''%2C''content''%3A%7B''field''%3A''cases.primary_site''%2C''value''%3A%5B''Breast''%5D%7D%7D%5D%7D). The matched clinical information, including ER status, PR status, HER2 status and survival data were also obtained. All BRCA samples were divided into two sets: ER⁺/PR⁺, HER2⁻ BRCA and TNBC, based on clinical information. In total, the ER⁺/PR⁺, HER2⁻ BRCA dataset included 422 cancer samples and 22 matched normal samples, while the TNBC dataset contained 112 cancer samples and 3 matched normal samples.

To obtain more accurate information for analysis, the data of BRCA-associated lncRNAs and genes were extracted. Information regarding cancer-related lncRNAs was obtained from lnc2cancer 2.0, which is an updated database that provides comprehensive experimentally supported associations between lncRNAs and human cancer (20). Cancer-related genes were obtained from DisGeNET, which is the largest publicly available database of genes and variants associated with a variety of human diseases (21). A t-test was performed to identify differentially expressed cancer-related genes and lncRNAs in ER⁺/PR⁺, HER2⁻ BRCA and TNBC. P<0.05 was considered to indicate a statistically significant difference. A total of 6,626 and 555 ER⁺/PR⁺, HER2⁻ BRCA-specific genes and lncRNAs, and 5,576 and 326 TNBC-specific genes and lncRNAs were respectively obtained. These BRCA subtype-specific genes and lncRNAs were employed for subsequent analysis.

TF-gene interaction data were obtained from TRANSFAC (22). ER⁺/PR⁺, HER2⁻ BRCA- and TNBC-specific TF-gene interactions were extracted based on the aforementioned conditions for the retrieval of BRCA subtype-specific genes and co-expression data. Pearson correlation coefficient (PCC) values were calculated for all cancer-associated TF-gene interactions. Then, ER⁺/PR⁺, HER2⁻ BRCA and TNBC-specific TF-gene interactions were considered, providing the absolute PCC values were >0.3. Following filtering, 494 and 476 ER⁺/PR⁺, HER2⁻ BRCA- and TNBC-specific TF-gene interactions were respectively identified and employed for subsequent analysis.

An integrated and computational approach was developed to identify lncTDTs associated with patients with ER⁺/PR⁺, HER2⁻ BRCA and TNBC. For each BRCA subtype-specific lncRNA, the BRCA samples were sorted based on lncRNA expression; 25% of the samples exhibiting the highest and lowest lncRNA expression levels were selected. Additionally, interactions between TFs and genes were considered, providing the interactions were affected by a certain lncRNA, based on the independent analysis of each lncRNA-TF-gene triplet. The PPC values between each TF and gene were respectively calculated in the top and bottom 25% samples for a given lncRNA. A TF-gene interaction was considered to be affected by an lncRNA providing the absolute difference of the PCC values between the top and bottom 25% samples was >0.4. This lncRNA-TF-gene interaction was considered as a candidate lncTDT. For a particular lncRNA, all BRCA subtype-specific TF-gene interactions were calculated. Furthermore, 1,000 random alterations of the sample labels of expression profiles was performed to compare the absolute difference of PCC values with changes in the absolute difference of these values to determine significance. P<0.05 was selected as the threshold value to obtain significant lncTDTs associated with ER⁺/PR⁺, HER2⁻ BRCA and TNBC.

In order to further investigate the mechanism of lncTDTs, the lncTDTs were separated into six profiles based on their mode of regulation: i) Weaken inhibition, TFs can suppress the expression of a gene, while an lncRNA could weaken this inhibitory action; ii) strengthen inhibition, TFs inhibit the expression of a gene, while an lncRNA could promote this inhibitory action; iii) reverse inhibition, TFs induce the expression of gene, while an lncRNA can invert these activating effects favoring inhibition; iv) reverse activation, TFs inhibit the expression of a gene, while an lncRNA can invert these inhibitory effects favoring activation; v) weaken activation, TFs induce the expression of a gene, but the effects of the TF are weakened by an lncRNA; and vi) strengthen activation, TFs can induce the expression of a gene, which is promoted by an lncRNA.

The networks of lncTDTs for ER⁺/PR⁺, HER2⁻ BRCA and TNBC were constructed and the degree analysis was performed using Cytoscape 3.0 (http://www.cytoscape.org/).

For all TFs and genes associated with ER⁺/PR⁺, HER2⁻ BRCA and TNBC, functional enrichment analyses were performed with the Enrichr tool online web server using default parameters (23). GO terms (http://geneontology.org/; Accessed May 9, 2019; P<0.01) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways (https://www.kegg.jp/; release 90.1; P<0.05) associated with lncTDTs in ER⁺/PR⁺, HER2⁻ BRCA and TNBC were determined.

The regression coefficient was employed to verify whether each lncTDT in ER⁺/PR⁺, HER2⁻ BRCA and TNBC was associated with survival based on expression data. A multivariate cox regression model was generated for each lncTDT and standardized Cox regression coefficients were obtained. Then, a risk score was established for each sample based on the expression profile of an lncTDT weighted by the standardized Cox regression coefficient. Furthermore, the BRCA samples were divided to high- and low-risk groups based on the median risk score. Finally, Kaplan-Meier survival analysis of the two groups was performed followed by a log-rank test. All analyses were performed using R 3.3.2 software (https://www.r-project.org/).

An integrated and computational approach was performed to identify lncTDTs in TNBC and ER⁺/PR⁺, HER2⁻ BRCA based on expression profiles and experimentally verified interactions. A total of 2,645 and 1,417 lncTDTs were identified in TNBC and ER^+/PR⁺, HER2⁻ BRCA; two lncTDT networks were also constructed for TNBC and ER^+/PR⁺, HER2⁻ BRCA (Fig. 1A and B; Table SI). There were 325 nodes and 1,417 edges in the TNBC lncTDT network; there were 583 nodes and 2,645 edges were identified in the ER⁺/PR⁺, HER2⁻ BRCA lncTDT network. The nodes included 218 lncRNAs, 107 TF-gene interactions, 33 TFs and 83 genes in the TNBC lncTDT network (Fig. 1C). The nodes included 410 lncRNAs, 173 TF-gene interactions, 69 TFs and 142 genes in the ER⁺/PR⁺, HER2⁻ BRCA lncTDT network (Fig. 1C). The degrees of all nodes in the TNBC network (R²=0.61; Fig. 1D) and the ER⁺/PR⁺, HER2⁻ BRCA network (R²=0.85; Fig. 1E) exhibited scale-free distribution, indicating that the networks were biological regulatory networks. In addition, degree analysis of the lncRNAs for each lncRNA in the networks for TNBC and ER⁺/PR⁺, HER2⁻ BRCA. The results revealed that the lncRNA with the highest degree was RP11-82L18.2 in the TNBC lncTDT network. lncRNA RP11-82L18.2 was determined to mediate 18 TF-gene interactions (Fig. 1F). The lncRNA with the highest degree was RAMP2 antisense RNA 1 (RAMP2-AS1) in the ER⁺/PR⁺, HER2⁻ BRCA lncTDT network; lncRNA RAMP2-AS1 mediated 30 TF-gene interactions (Fig. 1G). These results indicated that lncRNAs serve essential and specific roles in the lncTDT networks in TNBC and ER⁺/PR⁺, HER2⁻ BRCA.

TFs are known to activate or inhibit gene expression; however, this is a complex process in diseases, such as cancer. In addition, lncRNAs can influence the effects of TFs on their target genes; thus, these lncRNA-mediated TFs-gene interactions were defined as lncTDTs. The regulatory patterns of lncTDTs in TNBC and ER⁺/PR⁺, HER2⁻ BRCA are complex. The lncTDTs were divided into six patterns based on the regulatory action of lncRNAs and TFs (Fig. 2A). In total, there were, 1.27, 0.42, 18.77, 13.20, 24.49 and 41.85% lncTDTs involving lncRNAs that could weaken inhibition, strengthen inhibition, reverse inhibition, reverse activation, weaken activation and strengthen activation of TF-regulated gene expression, respectively for TNBC. In addition, there were 0.076, 0.87, 7.18, 4.23, 33.50 and 54.14 lncTDTs involving lncRNAs that could weaken inhibition, strengthen inhibition, reverse inhibition, reverse activation, weaken activation and strengthen activation of the effects of TF on gene expression for ER⁺/PR⁺, HER2⁻ BRCA (Fig. 2B). In TNBC and ER⁺/PR⁺, HER2⁻ BRCA, the most common regulatory pattern was strengthened activation, while the least common were weakened inhibition and strengthened inhibition. Additionally, it was determined that an lncRNA can exhibit various regulatory patterns; for example, a particular lncRNA could weaken, strengthen or reverse the effects of different TFs on gene expression. The cancer-associated lncRNA HOX transcript antisense RNA (HOTAIR), which is an important cancer-related lncRNA (24), could weaken activation and reverse inhibition of gene expression in TNBC (Fig. 2D). Furthermore, HOTAIR was proposed to reverse inhibition, and strengthen or weaken activation in ER⁺/PR⁺, HER2⁻ BRCA (Fig. 2E); only the most significant HOTAIR-related lncTDTs are shown. The most common patterns of HOTAIR in TNBC and ER⁺/PR⁺, HER2⁻ BRCA were weaken activation and strengthen activation, respectively. These results indicated the complexity of lncTDTs in TNBC and ER⁺/PR⁺, HER2⁻ BRCA.

TNBC and ER⁺/PR⁺, HER2⁻ BRCA are two important subtypes of BRCA. The mechanisms underlying the development and progression of these subtypes differ. Therefore, investigating the common characteristics of lncTDTs in TNBC and ER⁺/PR⁺, HER2⁻ BRCA may improve the understanding of this disease. In the present study, the lncRNAs, TFs, genes and TF-gene interactions in TNBC and ER⁺/PR⁺, HER2⁻ BRCA were analyzed. The majority of these molecules differed between TNBC and ER⁺/PR⁺, HER2⁻ BRCA; however, 142 lncRNAs, 18 TFs, 23 genes and 23 TF-gene interactions were reported in the two BRCA subtypes (Fig. 3A-D). In addition, five common lncTDTs were identified between TNBC and ER⁺/PR⁺, HER2⁻ BRCA (Fig. 3E). In the present study, 80% of the common lncTDTs exhibited the same regulation pattern (Fig. 3F); these lncTDTs included: NCK1-antisense RNA 1 (NCK1-AS1)/nuclear factor of activated T cells 2 (NFATC2)/CD3g molecule, NCK1-AS1/NFATC2/cytotoxic T-lymphocyte associated protein 4, small nucleolar RNA host gene 6 (SNHG6)/interferon regulatory factor 1 (IRF1)/interleukin 12B (IL12B) and SNHG6.1/IRF1/IL12B, which exhibited weakened activation (Fig. 3G). Of all the common lncTDTs, only PRC1 antisense RNA 1 (PRC1-AS1)/NK2 homeobox 5 (NKX2-5)/procollagen-lysine, 2-oxoglutarate 5-dioxygenase 1 (PLOD1) revealed a different regulatory pattern in TNBC and ER⁺/PR⁺, HER2⁻ BRCA. In addition, PRC1-AS1/NKX2-5/PLOD1 exhibited reversed inhibition and strengthened inhibition patterns (Fig. 3H). It was identified that lncRNA NCK1-AS1 serves an important role in the two BRCA subtypes; lncRNA NCK1-AS1 could promote proliferation and induce cell cycle progression in cervical cancer (25). Downregulation of lncRNA NCK1-AS1 was reported to increase chemosensitivity to cisplatin in cervical cancer (26). In the present study, lncRNA NCK1-AS1 was determined to regulate 12 and 14 TF-gene interactions in TNBC and ER⁺/PR⁺, HER2⁻ BRCA, respectively (Fig. 3I). The regulatory patterns in ER⁺/PR⁺, HER2⁻ BRCA are varied and complex. The present results indicated that the lncTDTs possessed common and specific characteristics in TNBC and ER⁺/PR⁺, HER2⁻ BRCA.

To further understand the functions and mechanisms of lncTDTs in TNBC and ER⁺/PR⁺, HER2⁻ BRCA, GO and KEGG pathway functional enrichment analyses were performed for genes and TFs associated with lncTDTs. The lncTDTs in TNBC or ER⁺/PR⁺, HER2⁻ BRCA were enriched for certain cancer-associated functions, including 'positive regulation of transcription, DNA-templated', 'negative regulation of transcription, DNA-templated', 'positive regulation of gene expression' and 'cellular response to cytokine stimulus' (Fig. 4A and B). Additionally, the lncTDTs in TNBC and ER⁺/PR⁺, HER2⁻ BRCA were both enriched in the PI3K/AKT signaling pathway which is a key pathway for cancer development and treatment (data not shown) and this pathway is key in the development and treatment of cancer (27-29); alterations in the PI3K/AKT signaling pathway also contribute to drug resistance in BRCA (30,31). In the present study, numerous genes in lncTDTs were enriched in this pathway for TNBC and ER⁺/PR⁺, HER2⁻ BRCA (Fig. 4C). Of note, lncTDTs were associated with the PI3K/AKT signaling pathway in the two BRCA subtypes; however, the genes differed. Furthermore, a common gene, CD19, was determined to be associated with the PI3K/AKT signaling pathway in TNBC and ER⁺/PR⁺, HER2⁻ BRCA. These results indicated that lncTDTs may serve essential roles in TNBC and ER⁺/PR⁺, HER2⁻ BRCA by affecting certain important biological processes and pathways.

A risk score-based analysis on the expression of the lncRNA, TF and gene of an lncTDT was performed for the prediction of survival to evaluate the potential of lncTDTs as prognostic biomarkers in the two BRCA subtypes. A total of 16.08 and 7.86% lncTDTs were significantly associated with survival in TNBC and ER⁺/PR⁺, HER2⁻ BRCA, respectively (Fig. 5A and B); few lncTDTs (0.28 and 1.36%) exhibited a highly significant association with survival in the two respective BRCA subtypes. In addition, the prognostic value of HOTAIR-mediated lncTDTs in the two BRCA subtypes was investigated. Certain HOTAIR-mediated lncTDTs in TNBC were associated with survival (data not shown). A total of seven HOTAIR-mediated lncTDTs in ER⁺/PR⁺, HER2⁻ BRCA were significantly related to survival (Fig. 5C), including HOTAIR/E2F transcription factor 1 (E2F1)/cyclin E1 (CCNE1), HOTAIR/CCAAT enhancer binding protein α (CEBPA)/apolipoprotein B, HOTAIR/CEBPA/carboxylesterase 1, HOTAIR/signal transducer and activator of transcription 5A/insulin like growth factor 1, HOTAIR/estrogen receptor 2/membrane metalloendopeptidase, HOTAIR/Kruppel like factor 11/hemoglobin subunit β and HOTAIR/early growth response 2/ATP binding cassette subfamily B member 1. Of these HOTAIR-mediated lncTDTs, HOTAIR/E2F1/CCNE1 was determined to be related to survival with the highest degree of significance (P=0.02); the lncTDTs that were associated with survival in TNBC and ER⁺/PR⁺, HER2⁻ BRCA with a high degree of significance are presented in Fig. 5D and E. The patients in the high-risk group exhibited shorter survival than those in the low-risk group. The results of the present study indicated that lncTDTs may collectively influence the survival of patients with TNBC or ER⁺/PR⁺, HER2⁻ BRCA; thus, lncTDTs could be considered as potential biomarkers for these subtypes of BRCA.