Abstract

Background

   Neoadjuvant chemotherapy (NAC) is an effective therapeutic regimen for
   patients with breast cancer. However, some individuals cannot benefit
   from NAC because of drug resistance. To date, valid strategies about
   enhancing sensitivity of breast cancer to NAC are still scarce. miRNAs
   have been reported to proverbially be involved in the onset and
   development of malignancies including drug resistance.

Methods

   [33]GSE73736 was downloaded from the GEO database. Student’s t-test was
   conducted to acquire differentially expressed-miRNAs (DE-miRNAs).
   Potential target genes of DE-miRNAs were predicted by miRTarBase. Gene
   Ontology annotation and Kyoto Encyclopedia of Genes and Genomes pathway
   enrichment analyses for these target genes were performed by database
   for annotation, visualization, and integrated discovery.
   Protein–protein interaction network was constructed by STRING database
   and visualized through Cytoscape software. The hub target gene-miRNA
   network was also established by Cytoscape software. Next, the
   expression of potential functional miRNAs in breast cancer cell lines
   and tissues was determined. Finally, the roles of miR-3617-3p,
   miR-3136-3p, and miR-520b in modulating breast cancer chemoresistance
   were further examined.

Results

   A total of 123 DE-miRNAs were identified, including 60 upregulated
   miRNAs and 63 downregulated miRNAs in the chemoresistant breast cancer
   group when compared with the chemosensitive group. Six hundred and
   seventeen and 1,146 potential target genes for the top 10 most
   upregulated and downregulated miRNAs were predicted, respectively.
   Enrichment analyses revealed that these target genes were enriched in
   some cancer-associated or chemo-resistance-associated pathways, such as
   MAPK signaling pathway, wnt signaling pathway, and p53 signaling
   pathway. MAPK1 and PRDM10 were identified as hub genes in the
   protein–protein interaction network. The top 25 hub genes were
   potentially regulated by 16 DE-miRNAs, among which miR-3617-3p and
   miR-3136-3p were commonly upregulated, whereas miR-520b was
   downregulated in two chemoresistant breast cancer cells compared with
   chemosensitive cell. By analyzing TCGA data, we found that expression
   of miR-3136-3p and miR-520b was increased and decreased in breast
   cancer tissues, respectively. Moreover, functional experiments
   demonstrated that miR-3136-3p and miR-3617-3p could reduce
   chemosensitivity of breast cancer, whereas miR-520b could reverse
   chemoresistance.

Conclusion

   The present study, based on bioinformatics analysis and experimental
   validation, brings to light novel mechanisms of breast cancer NAC
   resistance.

   Keywords: breast cancer, chemoresistance, neoadjuvant chemotherapy
   miRNA, bioinformatics analysis, enrichment analysis

Introduction

   Breast cancer is one of the most common malignancies among women, and
   its morbidity is growing yearly in most countries including America. In
   2017, according to the statistics provided by the American Cancer
   Society, more than 250,000 new cases of breast cancer and ~40,000
   breast cancer deaths are expected to occur.[34]^1 Surgery,
   radiotherapy, and chemotherapy are still three major traditional
   treatments for breast cancer. In the past decades, neoadjuvant
   chemotherapy (NAC) has been widely applied for improving disease free
   and overall survival and enabling more limited breast-conserving
   surgery in patients with breast cancer,[35]^2 and has substantially
   enhanced the efficacy of treatment for breast cancer. However, the
   benefit to patients with breast cancer from NAC is extremely limited in
   some individuals due to drug resistance.[36]^3 It is considerably
   meaningful to search and develop an approach which can reverse the drug
   resistance or increase the chemosensitivity of NAC in breast cancer.

   MiRNAs are a group of small, nature-occurred, noncoding RNAs with
   ~18–24 nucleotides in length. MiRNAs regulate gene expression, mainly
   by binding to the 32 untranslated region of target mRNA at
   posttranscriptional level, thereby causing direct degradation of mRNA
   or inhibition of protein translation.[37]^4 Through this approach,
   miRNAs are involved in a wide range of biological processes (BPs), such
   as proliferation, differentiation, apoptosis, autophagy, migration, and
   invasion.[38]^5 Many studies have demonstrated that miRNAs play key
   roles in the occurrence and progression of malignancies including
   breast cancer.[39]^6 Recently, miRNAs have been reported to be
   associated with chemoresistance of breast cancer.[40]^7^–[41]^9
   However, studies regarding the impact of miRNAs in regulating breast
   cancer NAC resistance are still rare. In this study, we successfully
   identified several chemoresistance-associated miRNAs (miR-3617-3p,
   miR-3136-3p, and miR-520b) in breast cancer via a series of processes,
   including identification of differentially expressed-miRNAs
   (DE-miRNAs), functional and pathway enrichment analysis, construction
   and analysis of the protein–protein interaction (PPI) network and
   miRNA-target gene network, and experimental validation in cell lines.
   The three miRNAs, miR-3617-3p, miR-3136-3p, and miR-520b, may be
   important novel NAC response predictors, prognosticators, or
   therapeutic targets in breast cancer in the future.

Materials and methods

MiRNA microarray

   The microarray data set [42]GSE73736 based on [43]GPL20986 MRA-1001B1
   Human miRNA array miRbase release 19 was downloaded from the National
   Center for Biotechnology Information Gene Expression Omnibus (GEO)
   database ([44]http://www.ncbi.nlm.nih.gov/geo). This data set contained
   20 cases of breast cancer tissues. According to Mille-Payne response,
   these 20 cases included 10 cases of NAC-resistant breast cancer tissues
   ([45]GSM1901882–[46]GSM1901891) which were MP1, and 10 cases of
   NAC-sensitive breast cancer tissues ([47]GSM1901892–[48]GSM1901901)
   which were MP5. The MP1 group represents that breast cancer in this
   group is totally unresponsive to NAC, whereas the MP5 group indicates
   that breast cancer in this group disappears after administration of
   NAC.

Screening for DE-miRNAs

   Data were first normalized by the normalize BetweenArray function from
   R package “LIMMA” from the Bioconduc-tor project.[49]^10 After
   normalization, the Student’s t-test was introduced to compare
   difference between the NAC-resistant group and the NAC-sensitive group.
   P-value <0.05 and |fold change (FC)|>2 were set as the thresholds for
   screening DE-miRNAs. This screening method was defined as “Previous
   analysis.” In addition, the online GEO2R analytic tool provided by the
   GEO database was further validated by our analytic results of “Previous
   analysis.”

Prediction of target genes of DE-miRNAs

   The miRTarBase ([50]http://mirtarbase.mbc.nctu.edu.tw/php/index.php),
   an experimentally validated microRNA-target interactions database, is
   widely used for predicting target genes of miRNA.[51]^11 In this study,
   miRTarBase was applied to predict targets of the top 10 most
   upregulated and down-regulated DE-miRNAs.

Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes
(KEGG) pathway enrichment analysis

   GO annotation and KEGG pathway enrichment analyses for the predicted
   target genes of the selected 20 DE miRNAs were conducted by using the
   online website, the database for annotation, visualization, and
   integrated discovery (DAVID, [52]http://david-d.ncifcrf.gov/).[53]^12

Construction and analysis of PPI network and miRNA-target network

   To assess the interactive relationship among these predicted target
   genes of the top10 most upregulated and downregulated DE-miRNAs, the
   STRING database ([54]http://string-db.org) was introduced to construct
   the PPI network.[55]^13 In search of hub genes in the two networks,
   Cytoscape software (version 3.6.1) was used to analyze the degree of
   connectivity.[56]^14 According to degree, we screened out the top 25
   hub genes, after which miRNA-hub genes networks were established by
   Cytoscape software.

Cell culture

   The human breast cancer cell line, BCap37, was purchased from the cell
   bank of the Chinese Scientific Academy. Two multidrug-resistant breast
   cancer cells, Bads-200 and Bats-72, were previously selected by our lab
   via treatment of Bcap37 for dose-dense and dose-escalation
   administration with paclitaxel.[57]^15 The three breast cancer cell
   lines, Bcap37, Bads-200, and Bats-72, were maintained in RMPI-1640
   medium (Gibco, Life Technologies, Carlsbad, CA, USA) with 10% FBS
   Biological Industries, Cormwell, CT, USA) and 1%
   penicillin-streptomycin at 37°C and 5% CO[2].

RNA extraction and quantitative-PCR (q-PCR)

   Total RNA was extracted from Bcap37, Bads-200, and Bats-72 cell lines
   by using RNAiso plus Reagent (TaKaRa, Kusatsu, Japan). MiRNAs reverse
   transcription primers and q-PCR primers were purchased from Ribobio Co.
   Ltd (Guangzhou, China). Total RNA was first reversely transcribed into
   complementary DNA through the PrimeScript RT Reagent Kit (TaKaRa,
   RR0037A). Then, SYBR Premix Ex Taq (TaKaRa, RR420A) was employed to
   conduct q-PCR. All experiments were performed in triplicate. The U6
   small nuclear RNA was used as the internal control.

Expression data from the cancer genome atlas (TCGA)

   The miRNA-sequencing level three data in patients of breast invasive
   carcinoma was downloaded from TCGA database
   ([58]https://genome-cancer.ucsc.edu/). The expression levels of
   potential miRNAs were extracted from the downloaded miRNA information.
   Statistical analysis between two groups was performed using Student’s
   t-test. Differences were considered significant with a P-value <0.05.

Drug sensitivity test

   The roles of miR-3617-3p, miR-3136-3p, and miR-520b in modulating
   breast cancer chemoresistance were determined using drug sensitivity
   test, as reported previously.[59]^9 In total, 2,500 cells of Bads-200
   and Bats-72 were seeded in 96-well plates and were cultured for 12
   hours, after which cells were transfected with miRNA mimics or negative
   control miRNA (Ribobio Co. Ltd) at a final concentration of 50 nM using
   Lipofectamine™ 3000 (Invitrogen, Carlsbad, CA, USA). After 24 hours,
   the cells were treated with paclitaxel for 48 hours. Then, MTT solution
   was added and incubated for 4 hours. Subsequently, the medium was
   replaced with 0.15 mL dulbecco’s modified eagle media. Finally, the
   absorbance in each well was detected at 570 nm through a microplate
   reader (Bio-Rad, Sunnyvale, CA, USA).

Results

Identification of DE-miRNAs and their targets in chemoresistant breast cancer

   A microarray data set [60]GSE73736 was downloaded from the GEO
   database. This data set included 10 cases of chemoresistant breast
   cancer tissues ([61]GSM1901882–[62]GSM1901891) and 10 cases of
   chemosensitive tissues ([63]GSM1901892– [64]GSM1901901). The data were
   first normalized, and the results are presented in [65]Figure 1. Then,
   the data were further processed by unpaired t-test (P<0.05,
   |log[2]FC|>1). In total, 123 DE-miRNAs were successfully screened out,
   as shown in [66]Figure 2A. Among the 123 DE-miRNAs, 60 miRNAs were
   significantly upregulated and 63 miRNAs were down-regulated. The top 10
   most upregulated and downregulated miRNAs are listed in [67]Tables 1
   and [68]2, respectively.

Figure 1.

   Figure 1
   [69]Open in a new tab

   Normalization of [70]GSE73736. (A) Data before standardization; (B)
   data after standardization.

Figure 2.

   [71]Figure 2
   [72]Open in a new tab

   Differentially expressed miRNAs (DE miRNAs) in ten cases of
   NAC-resistant (DR) breast cancer tissues and 10 cases of NAC-sensitive
   (DS) breast cancer tissues. (A) Heatmap of DE-miRNAs; (B) information
   of breast cancer tissues.

   Abbreviation: DR, drug resistant; DS, drug sensitive; NAC, neoadjuvant
   chemotherapy.

Table 1.

   Top 10 most upregulated differentially expressed miRNAs between DR
   group and DS group of human breast cancer
   MiRNA Average expression in DR group Average expression in DS group FC
   log[2]FC P-value
     __________________________________________________________________

   hsa-miR-1911-3p 11.0600 1.7120 6.460280 2.691597 0.0196
   hsa-miR-3130-5p 5.0940 0.9988 5.100120 2.350531 0.0260
   hsa-miR-3136-3p 3.4370 0.6950 4.945324 2.306065 0.0201
   hsa-miR-4697-5p 2.8520 0.6169 4.623116 2.208865 0.0281
   hsa-miR-4277 3.2360 0.7020 4.609687 2.204669 0.0018
   hsa-miR-4776-3p 6.6250 1.4660 4.519100 2.176035 0.0336
   hsa-miR-3619-3p 6.8330 1.5420 4.431258 2.147716 0.0454
   hsa-miR-5008-3p 6.8090 1.5380 4.427178 2.146387 0.0128
   hsa-miR-4423-3p 4.8600 1.1050 4.398190 2.136910 0.0061
   hsa-miR-3681-3p 14.0400 3.6100 3.889197 1.959472 0.0114
   [73]Open in a new tab

   Abbreviations: DR, drug resistant; DS, drug sensitive; FC, fold change.

Table 2.

   Top 10 most downregulated differentially expressed miRNAs between DR
   group and DS group of human breast cancer
   MiRNA Average expression in DR group Average expression in DS group FC
   log[2]FC P-value
     __________________________________________________________________

   hsa-miR-1253 0.9842 6.6550 0.147889 –2.75742 0.0442
   hsa-miR-584-5p 1.6060 9.7650 0.164465 –2.60415 0.0290
   hsa-miR-520b 0.7801 4.5860 0.170105 –2.55551 0.0215
   hsa-miR-1245b-5p 0.6776 3.4550 0.196122 –2.35018 0.0336
   hsa-miR-711 0.5090 2.5880 0.196677 –2.34610 0.0407
   hsa-miR-92a-1-5p 1.0600 5.2480 0.201982 –2.30770 0.0267
   hsa-miR-663b 0.9303 4.4860 0.207379 –2.26966 0.0184
   hsa-miR-617 1.9690 9.3850 0.209803 –2.25289 0.0447
   hsa-miR-922 0.6251 2.8600 0.218566 –2.19386 0.0208
   hsa-miR-586 1.4700 6.6050 0.222559 –2.16774 0.0188
   [74]Open in a new tab

   Abbreviations: DR, drug resistant; DS, drug sensitive; FC, fold change.

   In order to determine the reliability of our screening method, we
   further analyzed DE-miRNAs using the online analytic tool, GEO2R,
   provided by GEO database. Finally, 118 significant DE-miRNAs (P<0.05),
   including 59 upregulated miRNAs and 59 downregulated miRNAs, were
   obtained. Through getting intersection of the two sets of DE-miRNAs, we
   discovered that a lot of miRNAs commonly appeared in both two sets, as
   shown in [75]Figure S1.

   Next, miRTarbase, an experimentally validated miRNA-target interactions
   database, was employed to predict the target genes of the top 10 most
   upregulated and downregulated DE-miRNAs. Finally, 1,763 potential
   target genes were generated, among which 617 were for the upregulated
   miRNAs and 1,146 for the downregulated miRNAs.

Functional and pathway enrichment analyses

   To further explore the systematic features and biological functions of
   the aforementioned potential target genes, GO functional annotation and
   KEGG pathway enrichment analyses were performed via the online website,
   DAVID. Three GO categories, including BP, cellular component (CC), and
   molecular function (MF), were selected in the functional annotation.

   The top 10 GO terms of the target genes of the top 10 most upregulated
   DE-miRNAs are shown in [76]Figures 3A–C, including regulation of
   transcription and regulation of anti-apoptosis in the BP category,
   nuclear lumen and intracel lular organelle lumen in the CC category,
   and DNA binding and zinc ion binding in the MF category. KEGG pathway
   analysis ([77]Figure 3D) for the target genes of the top 10 most
   upregulated DE-miRNAs indicated that these target genes were enriched
   in some types of cancer, including glioma, melanoma, chronic myeloid
   leukemia, and non-small cell lung cancer. In addition, some of these
   potential target genes were also related to other pathways, such as the
   insulin signaling pathway and focal adhesion.

Figure 3.

   [78]Figure 3
   [79]Open in a new tab

   GO annotation and KEGG pathway enrichment analysis for the target genes
   of the top 10 most upregulated miRNAs. (A) Enriched BP of these target
   genes; (B) enriched CC of these target genes; (C) enriched MF of these
   target genes; (D) enriched KEGG pathways of these target genes.

   Abbreviations: GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and
   Genomes; BP, biological process; CC, cellular component; MF, molecular
   function.

   Subsequently, we conducted GO functional and KEGG pathway analyses for
   the target genes of the top 10 most downregulated DE-miRNAs. These
   target genes were involved in a variety of GO terms, such as regulation
   of transcription, negative regulation of apoptosis, and negative
   regulation of programmed cell death in the BP category ([80]Figure 4A),
   intracellular organelle lumen and membrane-enclosed lumen in the CC
   category ([81]Figure 4B), and ion binding and nucleotide binding in the
   MF category ([82]Figure 4C). The enriched KEGG pathways ([83]Figure 4D)
   for these target genes included the MAPK signaling pathway, wnt
   signaling pathway, and p53 signaling pathway. Of note, the three
   signaling pathways have been demonstrated to be associated with
   chemoresistance in the course of cancer
   treatment.[84]^9^,[85]^16^,[86]^17 Besides, 34 target genes (TRAF1,
   FGF7, XIAP, FOXO1, BCL2L1, PTEN, IGF1R, RAC1, FAS, RUNX1, TRAF6, MYC,
   FZD9, EGFR, CTBP1, MSH3, TGFBR2, CYCS, TP53, ITGA2, CDK6, CDK4, STAT3,
   RAD51, FZD6, MAPK1, CCDC6, NRAS, CDKN1A, LAMA4, CCND1, CRKL, LAMC1, and
   CRK) were involved in pathways in cancer.

Figure 4.

   [87]Figure 4
   [88]Open in a new tab

   GO annotation and KEGG pathway enrichment analysis for the target genes
   of the top 10 most downregulated miRNAs. (A) Enriched BP of these
   target genes; (B) enriched CC of these target genes; (C) enriched MF of
   these target genes; (D) enriched KEGG pathways of these target genes.

   Abbreviations: GO, Gene Ontology; KEGG, Kyoto Encyclopedia of Genes and
   Genomes; BP, biological process; CC, cellular component; MF, molecular
   function; ER,.

Construction and analysis of PPI network and miRNA-target network

   It has been widely acknowledged that proteins rarely act alone, but
   interact with each other. In view of this perspective, the STRING
   database was applied to construct the PPI network of these potential
   target genes of the top 10 most downregulated and upregulated miRNAs.
   After analyzing the data from STRING using Cytoscape software, we
   screened out the top 25 hub nodes according to degree ([89]Table 3).
   For downregulated miRNAs, the top 10 hub genes were MAPK1, CALM3, ESR1,
   CCND1, PPP2CA, CDKN1A, CYCS, CREB1, CDC27, and IGF1R, among which MAPK1
   showed the highest node degree (degree=57). For the upregulated miRNAs,
   the top 10 hub genes were PRDM10, PHLPP2, PIK3CB, H2AFX, CALM3, POLR2F,
   HSPA5, DICER1, ACTG1, and BTRC. Among these hub genes, PRDM10 showed
   the highest node degree, which was 67. For better visualization of
   interactions of these hub genes, additionally, we constructed networks
   based on the screened top 25 hub genes, as presented in [90]Figure 5.

Table 3.

   Hub genes identified in the PPI networks
   Downregulated miRNAs
     __________________________________________________________________

   Upregulated miRNAs
     __________________________________________________________________

   Gene symbol Degree Gene symbol Degree
     __________________________________________________________________

   MAPK1 57 PRDM10 67
   CALM3 41 PHLPP2 64
   ESR1 41 PIK3CB 37
   CCND1 39 H2AFX 35
   PPP2CA 36 CALM3 34
   CDKN1A 35 POLR2F 30
   CYCS 33 HSPA5 29
   CREB1 31 DICER1 29
   CDC27 30 ACTG1 28
   IGF1R 30 BTRC 25
   DNAJC10 30 MDM2 25
   PRKACB 26 CDKN1A 25
   FYN 26 HNRNPU 22
   CUL3 25 HNRNPA1 22
   FOXO1 25 ATM 22
   IL8 25 FZR1 21
   PARP1 24 FUS 20
   TCEB1 24 IGF1R 20
   HSPA5 24 BCL2L1 19
   UBE2V2 22 ARF1 19
   RPS6KB1 22 ACTRT3 18
   UBE2V1 21 NR3C1 17
   SEC61A1 21 MYH9 17
   TGFBR2 20 SNRPE 17
   WWP1 20 POLQ 16
   [91]Open in a new tab

   Abbreviation: PPI, protein–protein interaction.

Figure 5.

   [92]Figure 5
   [93]Open in a new tab

   The top 25 hub genes. (A) For downregulated miRNAs; (B) for upregulated
   miRNAs.

   Next, we further constructed the miRNA-hub gene network using Cytoscape
   software. As shown in [94]Figure 6A, the hub target genes of
   upregulated miRNAs could be potentially regulated by the top 10 most
   upregulated miRNAs. Of note, miR-4766-3p and miR-4697-5p could regulate
   the most hub target genes (n=4), suggesting that the two miRNAs may
   possess a significant role in manipulating drug resistance of human
   breast cancer. The results presented in [95]Figure 6B demonstrate that
   the top 25 hub genes from target genes of the top 10 most downregulated
   miRNAs could be potentially modulated by six miRNAs, namely
   miR-1245b-5p, miR-1253, miR-922, miR-520b, miR-584-5p, and
   miR-92a-1-5p. Among these miRNAs, miR-520b was found to potentially
   target the most hub genes (n=11). Additionally, miR-1253 and miR-584-5p
   could regulate 10 and six hub genes, respectively. The three miRNAs may
   act as crucial negative regulators in chemoresistance of breast cancer.

Figure 6.

   [96]Figure 6
   [97]Open in a new tab

   MiRNA–hub gene networks. (A) For upregulated miRNAs; (B) for
   downregulated miRNAs.

   Altogether, the construction and analysis of the PPI network could
   provide us with some information in searching for key genes (such as
   MAPK1 and PRDM10) which play important roles in chemoresistance of
   breast cancer. The construction of the miRNA-hub gene network is
   considerably helpful for us to catch sight of the most potentially
   functional miRNAs.

Expression and roles of potential miRNAs in breast cancer

   Two chemoresistant cell lines, Bads-200 and Bats-72, originated from
   chemosensitive Bcap37, were chosen to determine the expression of the
   16 potential miRNAs mentioned above in chemoresistant breast cancer. As
   shown in [98]Figure 7A, miR-3617-3p, miR-3136-3p, and miR-3130-5p were
   significantly upregulated, whereas miR-584-5p and miR-520b were
   downregulated in Bads-200 cell when compared to Bcap37. Expression
   levels of miR-3617-3p, miR-3136-3p, miR-4277, and miR-92a-1-5p in
   Bats-72 cell were obviously much higher than the Bcap37 cell, whereas
   miR-520b was significantly downregulated in the Bats-72 cell
   ([99]Figure 7B). In conclusion, compared to the Bcap37 cell, the
   expression values of miR-3617-3p and miR-3136-3p were commonly
   increased, whereas miR-520b was decreased in Bads-200 and Bats-72 cell
   lines. Next, we further analyzed miR-3617-3p, miR-3136-3p, and miR-520b
   expression in breast cancer tissues by using the TCGA database, and
   eventually discovered that miR-3136-3p and miR-520b were, respectively,
   upregulated and downregulated in breast cancer tissues compared to
   adjacent normal breast tissues ([100]Figure S2). However, miRNA
   miR-3617-3p was not found in the downloaded miRNA data from TCGA.

Figure 7.

   [101]Figure 7
   [102]Open in a new tab

   Expression and roles of potential miRNAs in breast cancer cell lines.
   (A) Comparison of miRNA expression between chemoresistant cell Bads-200
   and chemosensitive cell Bcap37; (B) comparison of miRNA expression
   between chemoresistant cell Bats-72 and chemosensitive cell Bcap37; (C)
   overexpression of miR-3136-3p and miR-520b significantly reduced and
   increased the sensitivity of Bads-200 cell line to paclitaxel (1,400
   nM), respectively; (D) overexpression of miR-3617-3p and miR-3136-3p
   significantly reduced the sensitivity of Bats-72 cell line to
   paclitaxel (130 nM), whereas overexpression of miR-520b could reverse
   chemoresistance.

   Notes: *P<0.05. Error bars represent SD for n=3.

   Abbreviation: ns, no significance.

   Next, we evaluated the roles of miR-3617-3p, miR-3136-3p, and miR-520b
   in modulation of breast cancer chemoresistance. [103]Figures 7C and D
   demonstrated that miR-520b could effectively overcome paclitaxel
   resistance, while miR-3136-3p could significantly promote the
   development and progression of paclitaxel resistance in both Bads-200
   and Bats-72 cells. For miR-3617-3p, we found that it could only reduce
   the chemosensitivity of the Bats-72 cell.

   In a few words, all these findings suggested that miR-3617-3p,
   miR-3136-3p, and miR-520b may be three key regulators of
   chemoresistance in breast cancer.

Discussion

   It has been widely acknowledged that miRNAs, a class of small
   endogenous noncoding RNAs, are frequently dysregulated in breast
   cancer. The abnormal expression of miRNAs is closely associated with
   the occurrence and development of breast cancer, such as
   proliferation,[104]^18 apoptosis,[105]^19 angiogenesis,[106]^20 and
   metastasis.[107]^21 Recently, our group has found that miR-20a and
   miR-27b were involved in chemoresistance of breast
   cancer.[108]^7^,[109]^9 The roles of miRNAs in chemoresistance of
   breast cancer was also reported by other groups.[110]^8 However,
   studies regarding systematic analysis of miRNAs’ roles in breast cancer
   NAC resistance are still rare.

   In this study, we first screened out 123 DE-miRNAs in chemoresistant
   breast cancer tissues compared to chemo-sensitive breast cancer
   tissues. Among these DE-miRNAs, miR-1253 and miR-1911-3p were shown to
   be the most downregulated and upregulated in the chemoresistance group.
   Previous investigations indicated that miR-1253 acted as a tumor
   suppressor in non-small cell lung cancer.[111]^22 MiR-1911-3p has been
   reported to have a correlation with neuroglioma.[112]^23 In addition to
   the two miRNAs, most identified DE-miRNAs have been confirmed to be
   associated with tumorigenesis.

   KEGG pathway enrichment analysis revealed that the potential target
   genes of the top 10 most upregulated and downregulated DE-miRNAs were
   enriched in some cancer-associated pathways, such as glioma, melanoma,
   chronic myeloid leukemia, and non-small cell lung cancer, and some
   chemoresistance-associated pathways, such as the MAPK signaling
   pathway, wnt signaling pathway, and p53 signaling pathway.

   In the PPI networks, MAPK1 and PRDM10 were the hub genes with the
   highest connectivity degree of 57 and 67, respectively. MAPK1 has been
   demonstrated to be involved in chemoresistance of breast cancer in our
   previous study.[113]^9 PRDM10, a member of PRDI-BF1 and RIZ homology
   domain containing (PRDM), has been lately discovered to be linked with
   the onset and development of several malignancies, such as
   nasopharygeal carcinoma[114]^24 and undifferentiated pleomorphic
   sarcoma.[115]^25 However, currently, studies about the impact of PRDM10
   on chemoresistance are still absent. Therefore, the function and
   mechanism of PRDM10 in chemoresistance of breast cancer need to be
   further investigated.

   Through constructing hub gene-miRNA networks, we found that the top 25
   hub genes could be potentially modulated by 16 DE-miRNAs (from the top
   10 most upregulated and downregulated DE-miRNAs). Among these 16
   DE-miRNAs, miR-520b was the miRNA which could potentially regulate the
   most hub genes (n=11). Cui et al[116]^26 suggested that overexpression
   of miR-520b could increase the sensitivity of breast cancer cell lines
   to complement-dependent cytotoxicity. Additionally, miR-520b has also
   been reported to increase chemosensitivity of hepatocellular carcinoma
   cells to doxorubicin.[117]^27 All these findings support that miR-520b
   may be a key negative modulator in breast cancer NAC resistance.

   Then, the expression values of these miRNAs in Bads-200, Bats-72, and
   Bcap37 were determined by q-PCR. When compared with Bcap37,
   miR-3617-3p, and miR-3136-3p were markedly upregulated and miR-520b was
   significantly downregulated in both Bads-200 and Bats-72 cells. In view
   of this, the three miRNAs may be the most important regulators in
   breast cancer NAC resistance.

   Next, we further examined the role of miR-3617-3p, miR-3136-3p, and
   miR-520b in regulating breast cancer chemoresistance. Paclitaxel was
   chosen as the first tested drug as it is a chemotherapy medication
   employed to treat multiple types of cancer. In breast cancer,
   particularly, paclitaxel is considered as one of the first-line
   chemotherapeutic agents and is also widely used as an NAC agent
   clinically.[118]^28 Our results suggested that all the three miRNAs,
   especially miR-3136-3p and miR-520b, are key regulators in breast
   cancer chemoresistance.

   Besides, miR-3130-5p and miR-584-5p were upregulated and downregulated
   in the Bads-200 cell. MiR-4277 and miR-92a-1-5p were upregulated in the
   Bats-72 cell. These miRNAs may also play important roles in modulating
   breast cancer NAC resistance. Despite some positive results being
   obtained in our current study, many more experiments need to be
   launched to confirm our preliminary analytic data in the future.

Conclusion

   To sum up, our present study provides a comprehensive analysis of
   DE-miRNAs and their potential target genes which may be involved in
   breast cancer NAC resistance. All findings have demonstrated that
   miR-3617-3p, miR-3136-3p, and miR-520b may be three key functional
   miRNAs in regulating breast cancer NAC resistance, and they may serve
   as NAC response predictors, prognosticators, and therapeutic targets in
   the future.

Supplementary materials

   Figure S1

   The intersection of our previous analytic DE-miRNAs and GEO2R analytic
   DE-miRNAs. (A) For downregulated DE-miRNAs; (B) for upregulated
   DE-miRNAs.

   Abbreviation: DE-miRNAs, differentially expressed miRNAs.
   [119]cmar-10-4747s1.tif^ (163KB, tif)
   Figure S2

   Abnormal expression of miR-3136-3p and miR-520b in breast cancer based
   on TCGA database. (A) The miR-3136-3p expression in tumor tissues was
   compared with normal tissues; (B) the miR-3136-3p expression in tumor
   tissues was compared with matched adjacent normal tissues; (C) the
   miR-520b expression in tumor tissues was compared with normal tissues;
   and (D) the miR-520b expression in tumor tissues was compared with
   matched adjacent normal tissues. ***P<0.001; ****P<0.0001. Error bars
   represent SD.

   Abbreviation: TCGA, the cancer genome atlas.
   [120]cmar-10-4747s2.tif^ (209.7KB, tif)

Acknowledgments