Abstract

   Breast cancer (BC) is the most common malignancy in female, but the
   role of androgen receptor (AR) in triple-negative breast cancer (TNBC)
   is still unclear. This study aimed to exam the performance of
   innovative biomarkers for AR positive TNBC in diagnosis and therapies.
   Four datasets ([32]GSE42568, [33]GSE45827, [34]GSE54002 and
   [35]GSE76124) were analyzed by bioinformatic methods and the
   differential expression genes (DEGs) between the AR positive TNBC
   tissues and normal tissues were firstly identified by limma package and
   Venn diagrams. Next, Gene Ontologies (GO) and Kyoto Encyclopedia of
   Genes and Genomes (KEGG) analysis were performed to explore the
   relationship between these DEGs. Then, the Protein-protein interaction
   (PPI) network was constructed. CytoHubba and bioinformatic approaches
   including Molecular Complex Detection (MCODE), Gene Expression
   Profiling Interactive Analysis (GEPIA), the Kaplan–Meier (KM) plotter
   and The Human Pro-tein Atlas (THPA) were used to identify the hub
   genes. Lastly, a miRNA-hub-gene regulatory axis was constructed by use
   of Target Scan database and ENCORI database. As a result, a total of
   390 common DEGs were identified, including 250 up-regulated and 140
   down-regulated. GO and KEGG enrichment analysis showed that the
   up-regulated DEGs were mostly enriched in the cell division, mitotic
   nuclear division, nucleosome, midbody, protein heterodimerization
   activity, cadherin binding involved in cell−cell adhesion, systemic
   lupus erythematosus and alcoholism, while the down-regulated DEGs were
   mainly enriched in carbohydrate metabolic process, extracellular space,
   extracellular region, zinc ion binding and microRNAs in cancer. Then,
   13 hub genes (CCNB2, FOXM1, HMMR, MAD2L1, RRM2, TPX2, TYMS, CEP55,
   AURKA, CCNB1, CDK1, TOP2A, PBK) were selected. The survival analysis
   revealed that only CCNB1 was associated with significantly poor
   survival (P <0.05) in TNBC patients. Finally, we found that
   hsa-miR-3163 took part in the regulation of CCNB1 and constructed a
   potential hsa-miR-3163-CCNB1 regulatory axis. The results of current
   study suggest that CCNB1 and hsa-miR-3163 may serve as highly potential
   prognostic markers and therapeutic targets for AR positive TNBC. Our
   findings may make contributions to the diagnosis and therapies of AR
   positive TNBC.

Introduction

   According to the latest cancer statistic, breast cancer is the most
   common malignant tumor in females, being the most important cause of
   cancer death in many countries [[36]1–[37]3]. TNBC is a special subtype
   of breast cancer defined as short of expression of the estrogen
   receptor (ER), progesterone receptor (PR) and human epidermal growth
   factor receptor 2 (HER2) [[38]4]. Furthermore, TNBC has been subdivided
   into four categories: luminal androgen receptor (LAR), mesenchymal
   (MES), basal-like immunosuppressed (BLIS), and basal-like immune
   activated (BLIA) [[39]5]. In addition, TNBC accounted for approx.
   10–15% of the breast cancer with high incidence of metastatic
   occurrence and poor overall survival [[40]4]. Despite the several
   standard chemotherapies and immunotherapies, TNBC was still lack of
   appropriate targets for treatment.

   AR is a transcription factor widely involved in the development of
   breast, and its importance is self-evident [[41]6]. AR was expressed in
   approx. 10–43% of TNBC, being involved in the invasion and metastasis
   of malignant tumor [[42]7, [43]8]. Nevertheless, the function of AR in
   TNBC was still unclear. In some studies, the expression of AR was
   associated with favorable outcomes, but in others, the opposite results
   had been achieved [[44]9–[45]14]. Therefore, it is urgent to understand
   the molecular mechanisms of AR positive TNBC progression to identify
   the potential biomarkers and therapeutic targets.

   In current study, we collected microarray data of [46]GSE42568,
   [47]GSE45827, [48]GSE54002 and [49]GSE76124 datasets from GEO. Limma
   package was used to identify the DEGs between the AR positive TNBC
   tissue and normal tissue. GO and KEGG pathway analyses were performed
   to clarify the potential functions and pathways of DEGs. The PPI
   network was created by the Search Tool for the Retrieval of Interacting
   Genes (STRING) database and visualized by Cystoscape software. Followed
   by the core module was constructed by MCODE and the hub genes were
   selected by the topological analysis methods. The expression levels of
   hub genes between TNBC tissues and normal tissues were validated by
   GEPIA (P <0.05). The online database Kaplan-Meier plotter was applied
   to the hub genes for prognostic evaluation. Subsequently, the CCNB1 was
   identified as the key gene in AR positive TNBC. Target miRNAs of CCNB1
   were predicted by from the intersection of ENCORI and TargetScan, and
   the correlation between CCNB1 and the miRNAs were verified. Finally, a
   regulatory miRNA of CCNB1 was identified, has-miR-3163, which was
   related to the molecular mechanisms of the AR positive TNBC. The result
   may further provide a new strategy for the diagnosis and treatment of
   AR positive TNBC.

Materials and methods

Microarray data

   Four gene expression microarray datasets ([50]GSE42568, [51]GSE45827,
   [52]GSE54002 and [53]GSE76124) were downloaded from the GEO
   ([54]https://www.ncbi.nlm.nih.gov/geo) database. All the datasets were
   processed on the [55]GPL570 platform (Affymetrix Human Genome U133 Plus
   2.0 Array). The normal breast tissue samples were selected from
   [56]GSE42568, [57]GSE45827 and [58]GSE54002, and the AR positive TNBC
   samples were selected from [59]GSE76124. [60]GSE42568, [61]GSE45827 and
   [62]GSE54002 contained 17, 11, and 16 normal breast tissue samples,
   respectively. [63]GSE76124 contained 37 AR positive TNBC samples.

Data preprocessing

   The regression calculation was performed to evaluate the datasets by
   FITPLM function in the AFFYPLM package. Followed by, the quality of
   data was evaluated by the weight map, residual symbol map, relative
   logarithmic expression map and RNA degradation map. Next, these
   datasets were read by RMA method, and the missing values were added by
   KNN method [[64]15]. Finally, the probe ID was replaced with the
   corresponding gene symbol by the [65]GPL570 platform annotation file.

Identification of DEGs

   The DEGs between AR positive TNBC samples ([66]GSE76124) and normal
   breast tissue samples ([67]GSE42568, [68]GSE45827 and [69]GSE54002)
   were analyzed by limma package, respectively. The genes that met cutoff
   criteria of |log2fold change (FC)|> 1 and p-value < 0.05 were
   considered as DEGs. Subsequently, the overlapping DEGs were screened
   out via drawing Venn diagrams for subsequent function analysis.

GO and KEGG pathway enrichment analysis for DEGs

   GO and KEGG analysis included functional annotation, enrichment
   analysis and pathways analysis for genes. GO analysis contained 3
   categories, biological processes (BPs), cellular components (CCs) and
   molecular functions (MFs) [[70]16, [71]17]. Online database DAVID
   ([72]https://david.ncifcrf.gov/, version 6.8) was employed to perform
   GO and KEGG pathway analysis for the up-regulated and down-regulated
   DEGs, respectively (p < 0.05) [[73]18].

PPI network construction and analysis

   The PPI network was constructed for DEGs by the STRING database
   ([74]https://string-db.org/, version 11.0) with a combined interaction
   score > 0.4, displayed by Cytoscape (version 3.8.2) [[75]19, [76]20].
   Based on the network, the cytoHubba was used to search for the hub
   genes [[77]21]. Subsequently, the MCODE app in Cytoscape was adopted to
   check these genes (degree cut-off = 2, node score cut-off = 0.2,
   k-score = 2, and max Depth = 100).

Survival analysis and validation of the hub genes

   Online database KM plotter([78]http://kmplot.com/analysis/) was adopted
   to evaluate the prognostic value of the identified hub genes [[79]22].
   In our research, the parameters were set as follows: (1) the negative
   expression of ER, PR and HER-2; (2) selected only JetSet best probe
   set; (3) excluded biased arrays; (4) survival: RFS. The mRNA expression
   levels of the hub genes between normal breast samples and BC samples
   were verified by the GEPIA([80]http://gepia.cancer-pku.cn/)
   platform(|log2FC| cut-off>1 and p-value cut-off< 0.01) [[81]23]. THAP
   database([82]https://www.proteinatlas.org/) was applied to validate the
   expression levels of CCNB1 in normal breast tissues and breast tumor
   tissues [[83]24].

Identification of candidate miRNAs

   ENCORI and TargetScan databases were employed to predict the target
   miRNAs of CCNB1. In our study, the parameters of ENCORI were set as
   follows: (1) CLIP data: high stringency (≥3); (2) Degradome data: with
   or without data; (3) Pan-cancer: 1 cancer type. TargetScan
   ([84]http://www.targetscan.org/vert_72/; version 7.2) was a widely used
   database to predict miRNAs [[85]25]. The candidate miRNAs of CCNB1 were
   selected through the intersection of ENCORI and TargetScan.

Identification of the miRNA-mRNA regulation axis

   ENCORI was used to verify the correlation between CCNB1 and the
   candidate miRNAs. It was also used to compare the expression levels of
   candidate miRNAs between BC samples and normal tissue samples. The
   overexpression of CCNB1 was associated with unfavorable prognosis in AR
   positive TNBC. Accordingly, we hypothesized that the miRNAs regulating
   CCNB1 was associated with a favorable prognosis. Subsequently, the KM
   plotter was adopted to assess the prognosis of the candidate miRNAs
   (the dataset: TCGA, molecular subtype: TNBC). Eventually, the
   miRNA-mRNA regulation axis was displayed by Cytoscape.

Drug sensitivity analysis

   The NCI-60 database, containing data from 60 cancer cell lines, was
   analyzed by CELLMINER website
   ([86]https://discover.nci.nih.gov/cellminer/). The expression status of
   target genes and z-score for cell sensitivity data (GI50) was
   downloaded from the website and assessed through Pearson correlation
   analysis to determine the correlation between target gene expression
   and drug sensitivity.

Results

Data quality assessment

   All the points were distributed uniformly in the weight and residual
   symbol maps. In the relative logarithmic expression map, all samples
   were close to the zero point without outliers. In the RNA degradation
   diagram, the 5′-terminal was lower than the 3′-terminal and the slope
   was suitable [[87]15] ([88]S1 Fig). The results indicated that these
   data were suitable for further analysis.

Identification of DEGs

   Our study included 44 normal breast tissues and 37 AR positive TNBC
   tissues, the normal breast tissue datasets ([89]GSE42568, [90]GSE45827
   and [91]GSE54002) and AR positive TNBC tissues dataset ([92]GSE76124)
   were analyzed by R software to obtain DEGs, respectively. [93]GSE76124
   vs [94]GSE42568, [95]GSE76124 vs [96]GSE45827 and [97]GSE76124 vs
   [98]GSE54002 were analyzed, resulting in 2896 (1771 up-regulated and
   1125 down-regulated), 3360 (2091 up-regulated and 1269 down-regulated)
   and 3158 (1166 up-regulated and 1992 down-regulated) DEGs, respectively
   ([99]Fig 1). Subsequently, the overlapping DEGs were identified via the
   Venn diagrams. Compared with normal breast tissues, 390 common DEGs
   were discovered in AR positive TNBC tissues, of which 250 were
   up-regulated (P < 0.05, log2FC > 1) and 140 were down-regulated (P <
   0.05, log2FC <‐1) ([100]Fig 2).

Fig 1. Volcano map of DEGs, red color represents of the up-regulated and
down-regulated genes.

   [101]Fig 1
   [102]Open in a new tab

   (a) [103]GSE76124 VS [104]GSE42568, (b) [105]GSE76124 VS [106]GSE45827,
   (c) [107]GSE76124 VS [108]GSE54002.

Fig 2. Venn diagram of DEGs in [109]GSE76124 VS [110]GSE42568, [111]GSE76124
VS [112]GSE45827 and [113]GSE76124 VS [114]GSE54002.

   [115]Fig 2
   [116]Open in a new tab

   (a) the overlapped 250 up-regulated DEGs (P < 0.05, log2FC > 1), (b)
   the overlapped 140 down-regulated (P < 0.05, log2FC <‐1).

GO and KEGG pathway enrichment analysis

   The GO and KEGG pathway analysis were performed to explore the
   biological functions for the 390 overlapping DEGs by means of DAVID
   software. The top five GO terms in the BP, CC and MF categories are
   shown in [117]Fig 3. The upregulated DEGs were mostly enriched in cell
   division and mitotic nuclear division among the BP categories, the
   nucleosome among the CC categories, and protein heterodimerization
   activity and cadherin binding involved in cell−cell adhesion among the
   MF categories. The downregulated DEGs were mainly associated with
   carbohydrate metabolic process among the BP categories, extracellular
   space and extracellular region among the CC categories, and zinc ion
   binding among the MF categories. In the KEGG pathway analysis, the
   upregulated DEGs were mostly enriched in the systemic lupus
   erythematosus and alcoholism, and the downregulated DEGs were mostly
   enriched in MicroRNAs in cancer ([118]Fig 3).

Fig 3. GO and KEGG pathway enrichment analysis for DEGs.

   [119]Fig 3
   [120]Open in a new tab

   (a) up-regulated DEGs enrichment in GO, (b) down-regulated DEGs
   enrichment in GO, (c) up-regulated DEGs enrichment in KEGG pathway, (d)
   down-regulated DEGs enrichment in KEGG pathway.

PPI network construction and analysis

   The STRING database was adopted to construct the PPI network for the
   390 overlapping DEGs, displayed in Cytoscape. The result revealed that
   the PPI network contained 287 nodes and 3420 edges ([121]Fig 4a). The
   most significant module was identified by the MCODE based on the PPI
   network. The result showed that the major module contained 70 central
   nodes and 2271 edges, and all of which were up-regulated genes. The
   topological analysis methods in cytoHubba were adopted to select the
   hub genes. As a result, CCNB2, FOXM1, HMMR, MAD2L1, RRM2, TPX2, TYMS,
   CEP55, AURKA, CCNB1, CDK1, TOP2A, PBK were identified as hub genes
   which were in the intersection of at least five methods. In addition,
   all the 13 hub genes were contained in the most meaningful module
   ([122]Fig 4b).

Fig 4. The PPI network of DEGs and the highest score clustering module of the
PPI network.

   [123]Fig 4
   [124]Open in a new tab

   (a) the PPI network of DEGs, red color indicates up-regulated DEGs,
   blue color represents downregulated DEGs and yellow color indicates the
   highest score clustering module generated by MCODE. (b) the highest
   score clustering module of the PPI network was magnified and green
   color indicates the hub genes of the PPI network.

Expression and survival analysis for hub genes

   The expression levels of the 13 hub genes were examined by GEPIA
   database (|log2FC| cut-off value = 1 and p value cut-off value = 0.01).
   [125]Fig 5 indicated that the expression levels of the 13 hub genes
   were significantly higher in breast cancer tissues than in normal
   breast tissues (P < 0.05). Subsequently, the KM plotter was adopted to
   the survival analysis for mentioned 12 hub genes (CDK1 data was not
   found), resulting that only the CCNB1 was associated with statistical
   poor survival (P<0.05) in TNBC ([126]Fig 6). Eventually, THAP database
   was employed to explore the relative expression of CCNB1 in BC.
   Immunohistochemistry (IHC) in THPA database verified that CCNB1 was
   upregulated in breast cancer tumor tissues ([127]Fig 7).

Fig 5. The expression of 13 hub genes.

   [128]Fig 5
   [129]Open in a new tab

   All the 13 hub genes highly expressed in breast cancer samples
   contrasted to normal samples (P < 0.01). Red color represents tumor
   tissues and grey color represents normal tissues.

Fig 6. Kaplan–Meier plotter for the hub genes.

   [130]Fig 6
   [131]Open in a new tab

   The Kaplan–Meier test p-value < 0.05: CCNB1. The Kaplan–Meier test
   p-value > 0.05: AURKA, CCNB2, CEP55, FOXM1, PBK, HMMR, RRM2, TOP2A,
   MAD2L1, TPX2, TYMS.

Fig 7. THPA website analysis of CCNB1 protein in BC.

   [132]Fig 7
   [133]Open in a new tab

   CCNB1 protein expression in BC specimens and non-cancerous breast
   tissues via THPA website analysis. Three representative images of BC
   ([134]https://www.proteinatlas.org/ENSG00000134057-CCNB1/pathology/brea
   st+cancer#img) and non-cancerous breast tissues were presented
   ([135]https://www.proteinatlas.org/ENSG00000134057-CCNB1/tissue/breast#
   img). N, normal; T, tumor; BC, breast cancer.

Identification of candidate miRNAs

   MiRNAs are small regulatory molecules (a short non-coding RNAs) which
   own vital biological functions [[136]26]. A total of 42 miRNAs
   regulating CCNB1 were predicted by the ENCORI database. 698 miRNAs
   regulating CCNB1 were predicted by the TargetScan database, including 6
   conserved sites and 692 poorly conserved sites. Eventually, 24
   overlapping miRNAs were selected via the Venn diagrams (Figs [137]8 and
   [138]9).

Fig 8. Validation of the overlapped miRNAs in the two databases (ENCORI and
TargetScan) via Draw Venn diagram.

   Fig 8
   [139]Open in a new tab

Fig 9. Correlation between miRNA-CCNB1 pairs identified by ENCORI database.

   [140]Fig 9
   [141]Open in a new tab

MiRNAs expression analysis and survival analysis

   The expression levels of target genes are negatively regulated by
   miRNAs binding to specific target sites via base pairing interactions
   to induce mRNA silence and degradation [[142]27, [143]28]. The miRNA is
   negatively correlated with the mRNA [[144]29]. Therefore, the
   expression of the candidate miRNAs regulating CCNB1 should be lower in
   BC, related with poor prognosis.

   ENCORI was used to verify the correlation between the CCNB1 and the
   candidate miRNAs. Among the 24 miRNA-CCNB1 pairs, 7 negatively
   correlated pairs were considered statistical (P < 0.05). The KM plotter
   were applied to survival analysis for the 7 candidate miRNAs. [145]Fig
   10 shows that the low expression of hsa-miR-181C-5p and hsa-miR-3163
   are associated with the poor prognosis of TNBC. Finally, the ENCORI
   pan-cancer analysis was performed to compare the expression levels of
   hsa-miR-181C-5p and hsa-miR-3163 between breast cancer and normal
   samples. The result showed that only hsa-miR-3163 was significantly
   down-regulated in BC samples ([146]Fig 11).

Fig 10. Kaplan–Meier plotter for miRNAs negatively correlated with CCNB1
expression in TNBC patients.

   [147]Fig 10
   [148]Open in a new tab

   TNBC patients with low expression of hsa-miR-181C-5p (d) and
   hsa-miR-3163 (f) had a poor prognosis (P < 0.05).

Fig 11. The expression levels of miRNAs.

   [149]Fig 11
   [150]Open in a new tab

   Hsa-miR-181C-5p. (a) and hsa-miR-3163 (b) between breast cancer and
   normal samples. The miRNA-CCNB1 regulation axis (c).

MiRNA-CCNB1 regulation axis

   The hsa-miR-3163 was selected to construct the miRNA-mRNA regulatory
   axis by bioinformatics analysis, visualized by Cytoscape software
   ([151]Fig 11c).

Drug sensitivity analysis to CCNB1

   The influence of target genes on drug sensitivity was assessed by using
   CellMiner database, which could facilitate better precision treatment.
   Drug sensitivity was measured by z-score, and the higher the scores
   implied that cells were more sensitive to the drug treatment ([152]Fig
   12). Notably, CCNB1 was associated with cell resistance to the
   treatment of denileukin difttitox (ontak). Furthermore, CCNB1was
   related to the increased sensitivity of cells to thioguanine and
   allopurinol.

Fig 12. Drug sensitivity analysis to CCNB1.

   [153]Fig 12
   [154]Open in a new tab

   The scatter plot indicates the correlation between target gene
   expression and drug sensitivity (the z-score of the CellMiner
   interface) for the Pearson correlation test using NCI-60 cell line
   data. 3 associations are shown, ordered by P-value.

Discussion

   Breast cancer is the most common malignant tumor in female. TNBC is a
   subtype of BC, lacks the expression of hormone receptors as well as
   human epidermal growth factor receptor 2 (HER2). Due to the absent of
   molecular targets, TNBC has no chance of endocrine treatment and HER2
   target therapy. Consequently, an increasing number of studies were
   performed to detect new diagnosis biomarkers and therapeutic targets of
   TNBC. The role of AR in TNBC remains poorly understood. Therefore, it
   is significant to explore the mechanisms of AR in TNBC.

   At present, the bioinformatic methods were seldom applied to study in
   AR positive TNBC. In this study, four datasets ([155]GSE42568,
   [156]GSE45827, [157]GSE54002 and [158]GSE76124 datasets) were analyzed
   by bioinformatic methods, including 44 normal breast tissues and 37 AR
   positive TNBC tissues. A total of 390 common DEGs (P < 0.05 and |
   log2FC |>1) were identified by use of limma package and Venn diagrams,
   of which 250 were up-regulated (log2FC > 1) and 140 were down-regulated
   (log2FC < ‐1). Subsequently, DAVID was used to GO and KEGG pathway
   enrichment analysis. In the GO enrichment analysis, the DEGs were
   mainly related to cell division, mitotic nuclear division, nucleosome,
   protein heterodimerization activity, carbohydrate metabolic process,
   cadherin binding involved in cell−cell adhesion, extracellular space,
   and zinc ion binding. In the KEGG pathway analysis, the DEGs were
   mainly related to MicroRNAs in cancer. Afterwards, the PPI network was
   established, comprehending 287 nodes and 3420 edges. 13 hub genes were
   selected based on the PPI network. Furthermore, GEPIA showed that the
   expression levels of the 13 hub genes were higher in breast tumor
   tissues than in normal tissues (P<0.01). The Kaplan–Meier plotter
   manifested that only CCNB1 was associated with significantly poorer
   survival (P<0.05) of TNBC patients. Eventually, IHC in THPA verified
   that CCNB1 was upregulated in breast cancer tumor tissues. Accordingly,
   CCNB1 may be a highly potential prognostic marker for AR positive TNBC.

   CCNB1 (Cyclin B1) is a member of the cyclin family, which plays an
   important role in the cell cycle at the G2/M (mitosis) transition
   [[159]30]. The overexpression of CCNB1 can cause uncontrolled cell
   growth and carcinogenesis by binding CDK1 [[160]31]. It has been
   reported that overexpression of CCNB1 was correlated with adverse
   pathological types and clinical outcomes [[161]32, [162]33], such as BC
   [[163]34–[164]36], colon adenocarcinoma [[165]31], non-small cell lung
   cancer [[166]37] and esophageal squamous cell cancer [[167]38]. A study
   showed that CCNB1 could be used as a prognostic factor of breast cancer
   death, with good repeatability [[168]39]. Furthermore, CCNB1 expression
   levels were significantly increased from benign to malignant breast
   tissues [[169]36]. Knockdown of CCNB1 could suppress cell proliferation
   in TNBC [[170]40]. Besides, a study in prostate cancer identified CCNB1
   as a bona fide AR target gene in prostate stromal cells. The negative
   regulation of CCNB1 by AR was mediated through switching between E2F1
   and E2F4 on the promoter of CCNB1 [[171]41]. Accordingly, the CCNB1 may
   be a highly potential biomarker and therapeutic target for AR positive
   TNBC.

   MiRNAs are small non-coding RNA molecules, widely involved in multiple
   processes of cancer progression including tumor genesis, development,
   drug resistance and metastasis [[172]42, [173]43]. MiRNAs negatively
   regulate target genes by inducing mRNA silencing and degradation. It
   can be detected on tissue samples and biological fluids including
   serum, plasma and urine [[174]44]. Therefore, miRNA has been regarded
   as a promising potential biomarker for cancer diagnosis and treatment.
   This research aimed to explore the potential miRNAs in AR positive TNBC
   as biomarkers and therapeutic targets. 24 miRNAs regulating CCNB1 were
   predicted by TargetScan database and ENCORI database, of which 13 were
   downregulated in breast cancer. And only 7 of the 13 miRNAs were
   considered statistically significant (P < 0.05). Afterwards, the KM
   plotter was adopted to assess the prognostic value of the 7 miRNAs in
   TNBC. Consequently, the hsa-miR-3163 was identified with good
   prognostic value in TNBC.

   Functionally, hsa-miR-3163 played as a tumor suppressor. A recent study
   in breast cancer indicated that MiR-513c and miR-3163 were
   downregulated in tumor tissues, which might serve as tumor suppressors
   [[175]45]. This is in line with the results of our research. Although
   there were few studies on the function of miR-3163 in breast cancer,
   miR-3163 were identified as a tumor suppressor in several kinds of
   malignant tumors. It has been reported that overexpression of
   hsa-miR-3163 may enhance the sensitivity of hepatocellular carcinoma
   cells to molecular targeted agents [[176]46]. In non-small cell lung
   cancer, hsa-miR-3163 could inhibit tumor cell growth [[177]47]. In
   retinoblastoma cancer stem cells, hsa-miR-3163 was associated with
   proliferation, apoptosis and multi-drug resistance [[178]48]. The
   expression of hsa-miR-3163 has been downregulated in colorectal cancer,
   whereas the colorectal cancer cell proliferation is suppressed by the
   overexpression [[179]49, [180]50]. In the future, further studies in
   vivo and in vitro should be conducted to identify the functional
   mechanism of miR-3163 in AR positive breast cancer.

   Taken together, in accord with previous studies, our research indicated
   that hsa-miR-3163 was related to inhibition of malignant tumors, such
   as breast cancer. As we know, there were no study investigating the
   role of hsa-miR-3163 in AR positive TNBC. According to our study,
   hsa-miR-3163 and CCNB1 may be highly potential biomarkers and
   therapeutic targets for AR positive TNBC. However, some limitations in
   our study should be noticed. Firstly, the common data sources from TCGA
   and GEO database were analyzed only by bioinformatic methods without in
   vivo and in vitro experiments. Secondly, our study only preliminarily
   revealed the expression levels of CCNB1 and his-miR-3163 in AR positive
   TNBC, but seldom involved the functional mechanisms and signaling
   pathways. Therefore, additional prospective studies are needed to
   verify the value of CCNB1 and his-miR-3163 in AR positive TNBC and the
   interaction of them should be further explored by experiment.

Conclusions

   In summary, through bioinformatics analysis, we suggest CCNB1 and
   hsa-miR-3163 as innovative potential biomarkers for AR positive TNBC.
   Our findings provide crucial information and direction for the
   elucidation of potential biomarkers and the therapeutic targets of AR
   positive TNBC. A series of future experiments will be conducted to
   verify these predictions.

Supporting information

   S1 Fig. Maps to evaluate the quality of the data.

   (a) weight map, (b) residual symbol map, (c) relative logarithmic
   expression map, (d) RNA degradation map.

   (TIF)
   [181]Click here for additional data file.^ (7.4MB, tif)

Data Availability

   The the Four gene expression microarray data sets presented in this
   study can be found in the GEO database
   ([182]https://www.ncbi.nlm.nih.gov/geo) at: GSE42568, GSE45827,
   GSE54002 and GSE76124. This information is also found in the
   "[183]Materials and methods" sections of this article.

Funding Statement

   The author(s) received no specific funding for this work.

References