RNA‐sequence‐based microRNA expression signature in breast cancer: tumor‐suppressive miR‐101‐5p regulates molecular pathogenesis

Aberrantly expressed microRNA (miRNA) are known to disrupt intracellular RNA networks in cancer cells. Exploring miRNA‐dependent molecular networks is a major challenge in cancer research. In this study, we performed RNA‐sequencing of breast cancer (BrCa) clinical specimens to identify tumor‐suppressive miRNA in BrCa. In total, 64 miRNA were identified as candidate tumor‐suppressive miRNA in BrCa cells. Analysis of our BrCa signature revealed that several miRNA duplexes (guide strand/passenger strand) derived from pre‐miRNA were downregulated in BrCa tissues (e.g. miR‐99a‐5p/‐3p, miR‐101‐5p/‐3p, miR‐126‐5p/‐3p, miR‐143‐5p/‐3p, and miR‐144‐5p/‐3p). Among these miRNA, we focused on miR‐101‐5p, the passenger strand of pre‐miR‐101, and investigated its tumor‐suppressive roles and oncogenic targets in BrCa cells. Low expression of miR‐101‐5p predicted poor prognosis in patients with BrCa (overall survival rate: P = 0.0316). Ectopic expression of miR‐101‐5p attenuated aggressive phenotypes, e.g. proliferation, migration, and invasion, in BrCa cells. Finally, we identified seven putative oncogenic genes (i.e. High Mobility Group Box 3, Epithelial splicing regulatory protein 1, GINS complex subunit 1 (GINS1), Tumor Protein D52, Serine/Arginine‐Rich Splicing Factor Kinase 1, Vang‐like protein 1, and Mago Homolog B) regulated by miR‐101‐5p in BrCa cells. The expression of these target genes was associated with the molecular pathogenesis of BrCa. Furthermore, we explored the oncogenic roles of GINS1, whose function had not been previously elucidated, in BrCa cells. Aberrant expression of GINS1 mRNA and protein was observed in BrCa clinical specimens, and high GINS1 expression significantly predicted poor prognosis in patients with BrCa (overall survival rate: P = 0.0126). Knockdown of GINS1 inhibited the malignant features of BrCa cells. Thus, identification of tumor‐suppressive miRNA and molecular networks controlled by these miRNA in BrCa cells may be an effective strategy for elucidation of the molecular pathogenesis of this disease.

Aberrantly expressed microRNA (miRNA) are known to disrupt intracellular RNA networks in cancer cells. Exploring miRNA-dependent molecular networks is a major challenge in cancer research. In this study, we performed RNA-sequencing of breast cancer (BrCa) clinical specimens to identify tumor-suppressive miRNA in BrCa. In total, 64 miRNA were identified as candidate tumor-suppressive miRNA in BrCa cells. Analysis of our BrCa signature revealed that several miRNA duplexes (guide strand/passenger strand) derived from pre-miRNA were downregulated in BrCa tissues (e.g. miR-99a-5p/-3p, miR-101-5p/-3p, miR-126-5p/-3p, miR-143-5p/-3p, and miR-144-5p/-3p). Among these miRNA, we focused on miR-101-5p, the passenger strand of pre-miR-101, and investigated its tumor-suppressive roles and oncogenic targets in BrCa cells. Low expression of miR-101-5p predicted poor prognosis in patients with BrCa (overall survival rate: P = 0.0316). Ectopic expression of miR-101-5p attenuated aggressive phenotypes, e.g. proliferation, migration, and invasion, in BrCa cells. Finally, we identified seven putative oncogenic genes (i.e. High Mobility Group Box 3, Epithelial splicing regulatory protein 1, GINS complex subunit 1 (GINS1), Tumor Protein D52, Serine/Arginine-Rich Splicing Factor Kinase 1, Vang-like protein 1, and Mago Homolog B) regulated by miR-101-5p in BrCa cells. The expression of these target genes was associated with the molecular pathogenesis of BrCa. Furthermore, we explored the oncogenic roles of GINS1, whose function had not been previously elucidated, in BrCa cells. Aberrant expression of GINS1 mRNA and protein was observed in BrCa clinical specimens, and high GINS1 expression significantly predicted poor prognosis in patients with BrCa (overall survival rate: P = 0.0126). Knockdown of GINS1 inhibited the malignant features of BrCa cells. Thus, identification of tumor-suppressive miRNA and molecular networks controlled by these miRNA in BrCa cells may be an effective strategy for elucidation of the molecular pathogenesis of this disease.

Introduction
Breast cancer (BrCa) is the most common malignancy among women, and~2 million cases are newly diagnosed each year, resulting in more than 620 000 deaths annually (Bray et al., 2018;Ferlay et al., 2015). In the general population,~12% of women will develop BrCa in their lifetime (Howlader et al., 2017). In contrast,~70% of women who inherit BRCA1 or BRCA2 mutations will develop BrCa by 80 years of age (Kuchenbaecker et al., 2017). A recent study reported that germline mutations in TP53 and PTEN also increase the risk of BrCa development (Economopoulou et al., 2015).
Based on gene expression signature analysis, BrCa can be classified into intrinsic molecular subtypes (Perou et al., 2000;Sotiriou et al., 2003). According to the 12th St Gallen International Breast Cancer Conference, BrCa can be classified into the following five subtypes, which can facilitate the selection of treatment strategies: luminal-A, luminal-B [human epidermal growth factor receptor 2 (HER2)-positive], luminal-B (HER2-negative), HER2-positive, and triple negative (Goldhirsch et al., 2011). These intrinsic molecular subtypes are related to the biological features of BrCa and are essential for treatment selection.
Many studies have indicated that noncoding RNAs derived from the human genome are functional and play pivotal roles in various cellular activities, e.g. cell proliferation, movement, and death (Gebert and MacRae, 2019;Treiber et al., 2019). Among noncoding RNAs, microRNA (miRNA) are short RNA molecules (19-22-nucleotide single-stranded RNA molecules) that play roles in regulating protein-coding and noncoding RNA expression in cells (Gebert and MacRae, 2019;Treiber et al., 2019). Importantly, a single miRNA regulates many RNA transcripts, and bioinformatics studies have shown that more than half of the RNA molecules transcribed from the genome are controlled by miRNA (Bartel, 2009). In cancer cells, intracellular RNA networks are disrupted due to the influence of abnormally expressed miRNA. These aberrantly expressed miRNA play critical roles in the malignant transformation of cancer cells.
To identify tumor-suppressive or oncogenic miRNA in cancers, miRNA expression signatures provide valuable information. RNA-sequencing technology is suitable for producing miRNA signatures. Recently, we reported the miRNA expression signature of triplenegative BrCa (TNBC), and 104 miRNA (56 upregulated miRNA and 48 downregulated miRNA) were found to be significantly dysregulated in TNBC tissues . TNBC is a subtype of BrCa in which estrogen receptor (ER), progesterone receptor, and HER2 are not expressed;~15-20% of BrCa cases are TNBC (Foulkes et al., 2010;Goldhirsch et al., 2011). TNBC is highly aggressive in nature, and metastases are frequently observed. Therefore, the prognosis of TNBC is worse than that of other subtypes of BrCa (Foulkes et al., 2010;Goldhirsch et al., 2011). Based on our TNBC signatures, we identified tumor-suppressive miR-204-5p and novel oncogenic genes regulated by this miRNA . Interestingly, several miR-204-5p target genes were found to be closely associated with BrCa pathogenesis . The discovery of oncogenic networks mediated by tumor-suppressive miRNA will contribute to the elucidation of the molecular mechanisms mediating the pathogenesis of BrCa.
Breast cancer is a heterogeneous cancer, and treatment strategies differ for each subgroup (Foulkes et al., 2010;Goldhirsch et al., 2011;Perou et al., 2000;Sotiriou et al., 2003). Thus, elucidation of the universal molecular pathways mediating BrCa will lead to the development of new treatment strategies for this disease. Accordingly, in this study, we created the RNA-sequencing-based miRNA expression signature of BrCa using clinical BrCa specimens, including ERpositive, HER2-positive, and TNBC specimens. In total, 64 miRNA were identified as candidate tumorsuppressive miRNA in BrCa cells. Analysis of our BrCa signature revealed that several miRNA duplexes (guide strand/passenger strand) derived from pre-miRNA were downregulated in BrCa tissues. Despite the general consensus that passenger strands derived from miRNA duplexes have no regulatory activity, our recent studies have revealed that some passenger strands actually function by targeting several genes (Mah et al., 2010;McCall et al., 2017).
Based on our current miRNA signature of BrCa, the expression levels of both strands of the miR-101 duplex (miR-101-5p: the passenger strand and miR-101-3p: the guide strand) were significantly reduced in cancer tissues, suggesting that these miRNA have tumor-suppressive functions. Many previous reports have demonstrated that miR-101-3p acts as a tumorsuppressive miRNA in various cancers . In contrast to miR-101-3p, the functional significance of miR-101-5p and RNA networks regulated by this miRNA in cancer cells is poorly understood. Accordingly, in this study, we showed that ectopic expression of miR-101-5p attenuated aggressive phenotypes, e.g. proliferation, migration, and invasion, in BrCa cells. Moreover, GINS complex subunit 1 (GINS1) was directly controlled by miR-101-5p in BrCa cells, and its expression contributed to BrCa oncogenesis.

Collection of clinical breast cancer specimens, breast epithelial specimens, and BrCa cell lines
To construct the miRNA expression signature of BrCa, 20 clinical tissue specimens (five specimens each for ER-positive BrCa, HER2-positive BrCa, TNBC, and normal breast epithelium) were collected following surgical resection at Gunma University Hospital.
To validate the expression levels of miRNA and target genes, 27 clinical specimens (18 BrCa specimens and nine normal breast epithelial tissues) were collected at Kagoshima University Hospital. Twenty-one paraffin blocks of BrCa specimens were used for immunostaining. The clinical features of these patients are shown in Table 1. Informed consent was obtained from all patients. This study was approved by the Bioethics Committee of Gunma University (approval nos 2016-023 and 2017-167) and Kagoshima University (approval no. 160038:28-65). The study methodologies conformed to the standards set by the Declaration of Helsinki.

Construction of the miRNA expression signature for BrCa
The miRNA expression signatures of 20 samples with BrCa (Table 1) were generated by small RNA sequencing using HiSeq 2000 (Illumina, San Diego, CA, USA). Small RNA sequencing and data mining were performed as previously described, and a false discovery rate (FDR) less than 0.05 was considered significant (Goto et al., 2017;Koshizuka et al., 2017;Toda et al., 2018;Yonemori et al., 2017).
2.3. RNA preparation and quantitative reverse transcription polymerase chain reaction (qRT-PCR) Total RNA including miRNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) in clinical specimens and ISOGEN reagent (NIPPON GENE, Tokyo, Japan) in BrCa cells. The qRT-PCR was performed as previously described Yamada et al., 2018a,b,c). TaqMan probes and primers used in this study are listed in Table S1.

Transfection of BrCa cells with miRNA, small interfering RNA (siRNA), and plasmid vectors
The miRNA, siRNA, and vectors were transfected into cancer cells as described in our previous reports using the reagents listed in Table S1 Yamada et al., 2018a,b,c).

Assays of cell proliferation, cell cycle, migration, and invasion
Cell proliferation, migration, and invasion were assessed as described previously Yamada et al., 2018a,b,c).

Evaluation of miR-101-5p binding sites by luciferase reporter assays
The 3 0 UTR of GINS1 and the 3 0 -UTR lacking the putative miR-101-5p binding sites were cloned into the psiCHECK-2 vector (C8021; Promega, Madison, WI, USA). Luciferase reporter assays were performed as previously described Yamada et al., 2018a,b,c). The cloned sequences are shown in Figs 4 and S1.

Clinical data analyses of BrCa
The clinical significance of miRNA and their target genes was investigated with The Cancer Genome Atlas (TCGA; https://tcga-data.nci.nih.gov/tcga/) in BrCa.

Western blotting and immunohistochemistry
Western blotting and immunohistochemistry were performed as described previously Yamada et al., 2018a,b,c). Primary antibodies are listed in Table S1.

Genes affected by GINS1 expression in BrCa cells
Gene expression levels and clinical information were downloaded from cBioPortal (http://www.cbioportal. org/) on 8 January 2019. The normalized mRNA expression levels of RNA-sequencing data were provided as Z-scores. Gene set enrichment analysis (GSEA) was performed based on mRNA sequence data from cBioPortal. A heatmap of gene expression was constructed using the BrCa RNA-sequence database. Overexpressed genes in BrCa tissues showing high GINS1 expression in TCGA were classified into known pathways using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database with the Enrichr program.

Statistical analysis
Mann-Whitney U tests were applied for comparisons between two groups. For multiple groups, one-way analysis of variance and Tukey tests for post-hoc analysis were applied. These analyses were performed with GRAPHPAD PRISM 7 (GraphPad Software, La Jolla, CA, USA) and JMP PRO 14 (SAS Institute Inc., Cary, NC, USA). For other analyses, EXPERT STATVIEW (version 5, SAS Institute, Inc.) was used.

Results
3.1. Creation of a miRNA expression signature for BrCa by small RNA sequencing RNA sequencing was performed to create the miRNA expression signature of BrCa. We sequenced 20 small RNA libraries (15 BrCa specimens and five normal breast epithelium specimens). The clinical features of the specimens used to create the miRNA signature are summarized in Table 1. We obtained between 10 112 255 and 15 495 422 total reads in this study. After filtering out noise (fragments that did not completely match the human genome sequence), between 4 781 591 and 13 003 597 miRNA reads were mapped on the human genome sequence (Table S2). Read sequences matching the human genome were categorized into small RNA according to their biological functions (Table S2). Finally, we constructed the miRNA expression signature of BrCa containing miRNA with markedly downregulated expression (Table 2; FDR < 0.05).
In total, 64 miRNA were significantly downregulated in BrCa tissues ( Table 2). Analysis of our BrCa signature revealed that 11 miRNA duplexes (guide strand/passenger strand) derived from pre-miRNA were downregulated in BrCa tissues (Table S3).
To verify the credibility of the BrCa signature, expression levels of miR-101-5p and miR-101-3p in clinical specimens (18 BrCa specimens and nine normal breast epithelial specimens) and two cell lines (MDA-MB-231 and MCF-7) were measured. Table 1 shows the information on the clinical specimens used for this study. The expression levels of the two miRNA, i.e. miR-101-5p (P = 0.0396) and miR-101-3p (P = 0.0047), were significantly reduced in BrCa tissues (Fig. 1A,B). Moreover, we confirmed that the expression levels were low in the two cell lines (Fig. 1A,B).
Next, we analyzed whether miRNA expression affected the prognosis of patients with BrCa by TCGA database analysis. Kaplan-Meier overall survival curves showed that low expression levels of miR-101-5p (P = 0.0316) and miR-101-3p (P = 0.0280) were associated with overall survival in patients with BrCa (Fig. 1C,D).

MicroR-101-5p was incorporated into the RNA-induced silencing complex (RISC) in BrCa cells
Next, we aimed to verify that miR-101-5p (passenger strand) had actual functions in BrCa cells. It is essential that miRNA are incorporated into the RISC to control target genes. Ago2 is a fundamental component of the RISC. Therefore, immunoprecipitation using anti-Ago2 antibodies was performed after transfection of miR-101-5p into MDA-MB-231 cells. The amount of miR-101-5p incorporated into the protein was measured by PCR. Levels of miR-101-5p in the immunoprecipitates were much higher than those in mock-, miR-control-or miR-101-3p-transfected cells (P < 0.0001; Fig. S4).

Candidate oncogenic targets regulated by miR-101-5p in BrCa cells
The TargetScan Human 7.1 database predicted that 2896 candidate genes had miR-101-5p binding sites in the 3 0 -UTR. We also investigated genes that were upregulated in clinical BrCa specimens by microarray analysis [Gene Expression Omnibus (GEO) accession number: GSE118539] and compiled a list of 1121 genes. Finally, 104 oncogenic targets regulated by miR-101-5p were identified in BrCa cells (Table 3). Our selection strategy for miR-101-5p targets is shown in Fig. S5.
Moreover, we confirmed that three genes (i.e. GINS1, TPD52, and SRPK1) were significantly downregulated by miR-101-5p transfection into both MDA-MB-231 and MCF-7 cells (Fig. 3). These three genes are essential for biological analysis of BrCa cells. We further analyzed the oncogenic functions of GINS1 in BrCa cells because this gene has not been described frequently in studies of cancer.
TargetScan database analysis showed that one putative miR-101-5p binding site was present in the 3 0 -UTR of GINS1 (Figs 4B and Fig. S1). Additionally, luciferase reporter assays showed that the luminescence intensity was markedly decreased by cotransfection with miR-101-5p and a vector carrying wild-type GINS1 3 0 -UTR. In contrast, the vector with a deleted miR-101-5p target site showed no change in luminescence intensity (Fig. 4C). These data indicated that GINS1 was directly regulated by miR-101-5p in BrCa cells.
We also investigated the direct regulation of TPD52 and SRPK1 by miR-101-5p in BrCa cells. The luminescence intensities were significantly reduced by cotransfection with miR-101-5p and vectors carrying wild-type TPD52 and SRPK1 3 0 -UTR, suggesting that these two genes were directly regulated by miR-101-5p (Fig. S6).   . Cell proliferation, migration, and invasion were significantly blocked by ectopic expression of miR-101-5p or miR-101-3p. Error bars are represented as mean AE SD. P-values were calculated using Bonferroni-adjusted Mann-Whitney U-test. *P < 0.01, **P < 0.0001.
A multivariate Cox proportional hazards model showed that high expression of GINS1 was an independent predictive factor for survival [hazard ratio (HR): 1.64, 95% confidence interval (CI): 1.12-2.41, P = 0.0102], as were well-known clinical prognostic factors such as N stage and M stage (Fig. 5C). Next, we investigated the expression levels of GINS1 in BrCa clinical specimens by immunostaining. GINS1 was strongly overexpressed in several cancer lesions compared with that in adjacent noncancerous lesions (Fig. 6). The clinical features of the specimens used to immunostaining are summarized in Table 1.

Effects of GINS1 silencing in BrCa cells
To validate the oncogenic functions of GINS1 in BrCa cells, we used knockdown assays with siRNA in two BrCa cell lines, MDA-MB-231 and MCF-7 (Fig. 5A). The two siRNA, siGINS1-1 and siGINS-2, used in this assay significantly suppressed GINS1/GINS1 expression in BrCa cells (Fig. 7A,B).
Functional assays showed that malignant phenotypes of BrCa cells (e.g. cell proliferation, migration, and invasive abilities) were significantly blocked by siGINS1 transfection in BrCa cells (Fig. 7C-E). Furthermore, cell cycle assays showed that G 0 /G 1 phase arrest was detected in siGINS1-transfected cells (Fig. S3).
Similar results were observed in another cell line, MDA-MB-157. Indeed, ectopic expression of miR-101-5p and knockdown of GINS1 significantly blocked cancer cell aggressive phenotypes in MDA-MB-157 cells (Fig. S8).

Genes affected by GINS1 expression in BrCa clinical specimens
Finally, we identified the differentially expressed genes that were affected by GINS1 in BrCa. Our strategy is shown in Fig. S9. GSEA for the differentially expressed genes in BrCa tissues showing high expression of GINS1 in TCGA identified 11 signaling pathways (Fig. S10). We categorized GINS1-regulated genes using KEGG pathways. In total, seven pathways were identified based on overexpressed genes in BrCa tissues showing high GINS1 expression in TCGA (Fig. 8A). In particular, genes involved in DNA replication pathways were identified by heatmap analysis (Fig. 8B).
Among these genes involved in DNA replication pathways, we investigated the clinical significance of the relationship between gene expression and prognosis of the patients with BrCa by TCGA database analysis. High expression of four genes (GINS1, MCM4, MCM6, and RFC3) was significantly associated with poor prognosis in patients with BrCa (Fig. 8C).

Discussion
Notably, a single miRNA regulates a wide range of different RNA transcripts (protein-coding and nonprotein-coding genes) in various normal and abnormal cells. Based on the unique nature of miRNA, novel RNA networks in human cancer cells can be identified from analysis of relevant miRNA. Currently available high-throughput technologies, e.g. oligo-microarrays, PCR-based arrays, and RNA-sequencing, have enabled the construction of miRNA expression signatures of BrCa (Ma et al., 2018), revealing the abnormal expression of many miRNA (Adhami et al., 2018;Gupta et al., 2019;Khordadmehr et al., 2019;Klinge, 2018;Kurozumi et al., 2017;Mehrgou and Akouchekian, 2017). One approach to identify the most important miRNA from a large number of candidate miRNA is to identify crossovers of miRNA that have been reported in multiple independent studies. Previous studies have shown that miR-139-5p, miR-195-3p, miR- 205-3p, and miR-99a-5p are frequently downregulated and function as tumor-suppressive miRNA in BrCa cells (Adhami et al., 2018;Gupta et al., 2019;Khordadmehr et al., 2019;Klinge, 2018;Kurozumi et al., 2017;Mehrgou and Akouchekian, 2017). These miRNA were included in the signature we created in this study.
Furthermore, a major advantage of this signature is that it contained multiple passenger strands of miRNA derived from miRNA duplexes, e.g. miR-99a-3p, miR-101-5p, miR-144-5p, and miR-145-3p. As a general theory of miRNA biogenesis, the guide strand of miRNA derived from the miRNA duplex is incorporated into the RISC and regulates gene expression (Bhayani et al., 2012;Mah et al., 2010;McCall et al., 2017). In contrast, the passenger strand is degraded and does not regulate genes in cells (Bhayani et al., 2012;Mah Fig. 6. Expression of GINS1 in clinical BrCa tissues. Immunohistochemistry staining of GINS1 in BrCa specimens. Overexpression of GINS1 was observed in cancer cells, whereas negative or low expression of GINS1 was observed in normal cells. Scale bars of 940 and 9400 represent 1 mm and 100 µm, respectively. McCall et al., 2017). However, our recent studies have shown that some passenger miRNA have tumor-suppressive functions in cancer cells (e.g. miR-144-5p, miR-145-3p, miR-150-3p, and miR-455-5p) (Arai et al., 2019;Misono et al., 2018;Misono et al., 2019;Uchida et al., 2019). These miRNA and their target oncogenic genes are closely associated with cancer pathogenesis (Arai et al., 2019;Misono et al., 2018;Misono et al., 2019;Uchida et al., 2019). In the future, we will attempt to clarify the new molecular networks of BrCa using passenger strands of miRNA as indicators.
We focused on miR-101-5p and explored new aspects of this miRNA in BrCa cells. Many studies . GUSB was used an internal control (*P < 0.0001). (B) GINS1 protein expression was evaluated by western blot analysis 72 h after transfection with si-GINS1-1 and si-GINS1-2 into BrCa cell lines. GAPDH was used as a loading control. (C) Cell proliferation was identified by XTT assays 72 h after transfection with siGINS1-1 and siGINS1-2 (*P < 0.0001). (D) Cell migration activity was determined using migration assays (*P < 0.001, **P < 0.0001). (E) Cell invasion was determined by Matrigel invasion assays (*P < 0.01, **P < 0.0001). Error bars are represented as mean AE SD. P-values were calculated using Bonferroni-adjusted Mann-Whitney U-test.
have shown that downregulation of miR-101-3p (the guide strand) occurs frequently in many cancers and that this miRNA acts as a tumor suppressor . Previous studies have clarified that miR-101-3p regulates various pivotal oncogenes and that downregulation of this miRNA affects cancer cell proliferation, metastasis, drug resistance, and angiogenesis via targeting of several oncogenic targets, e.g. EZH2, STMN1, VHL, SOX2, and DNMT3A . In BrCa, downregulation of miR-101-3p was detected in all subtypes of BrCa tissues, and miR-101-3p acted as a tumor suppressor (Liu et al., 2015;Liu et al., 2016;Ren et al., 2012;Zhang et al., 2015Zhang et al., , 2019Zhang et al., , 2015Zhang et al., , 2019. Compared with reports of miR-101-3p, few studies have reported the tumor-suppressive functions of miR-101-5p and its target molecules in cancer cells. More recently, downregulation of miR-101-5p was reported in non-small cell lung carcinoma tissues compared with that in normal tissues (Chen et al., 2019). Overexpression of miR-101-5p was shown to suppress the aggressive phenotypes of cancer cells (in vitro) and pulmonary metastasis (in vivo) by regulating CXCL6 (Chen et al., 2019). Our data also showed that miR-101-5p acted as an antitumor miRNA in BrCa cells. Notably, both strands of miRNA derived from the miR-101 duplex were found to have tumor-suppressive functions in cancer cells.
Next, we aimed to elucidate miR-101-5p-regulated oncogenes and oncogenic pathways in BrCa cells. Analysis of our miRNA target search revealed that seven genes (HMGB3, ESRP1, GINS1, TPD52, SRPK1, VANGL1, and MAGOHB) were closely associated with poor prognosis. Among these targets, three genes (GINS1, TPD52, and SRPK1) were strongly controlled by miR-101-5p in BrCa cells. Aberrant expression of TPD52 (encoding TPD52) has been reported in a wide range of cancers, including BrCa, and several tumorsuppressive miRNA have been reported to be involved in regulating the expression of these genes (Balleine et al., 2000;Byrne et al., 2014;Roslan  et al., 2014). SRPK1 (encoding serine-arginine protein kinase 1) is involved in the regulation of several mRNA processing pathways, and its overexpression has been reported in multiple cancers (Patel et al., 2019). High expression of SRPK1 is correlated with poor disease outcomes in patients with BrCa (Hayes et al., 2007;van Roosmalen et al., 2015). Knockdown of SRPK in BrCa cells inhibits metastasis to distant organs (Hayes et al., 2007;van Roosmalen et al., 2015). Further functional analyses of these genes will reveal the biological characteristics of BrCa. Starting from antitumor miRNA and using TCGA database analyses, we were able to identify effective prognostic markers and therapeutic targets for BrCa, indicating that our miRNA-based strategy was feasible.
In this study, we focused on GINS1 and showed that its aberrant expression was closely related to BrCa malignant phenotypes. Chromosomal DNA replication is a tightly controlled essential process in the eukaryotic cell cycle, and many proteins are involved in each step of DNA replication (Labib and Gambus, 2007;MacNeill, 2010;Seo and Kang, 2018;Sun et al., 2016). The GINS complex (SLD5, GINS1, GINS2, and GINS3) is involved in the minichromosome maintenance complex and Cdc45 with proteins in a replisome progression complex (Labib and Gambus, 2007;MacNeill, 2010;Seo and Kang, 2018;Sun et al., 2016). A previous study of GINS1 in BrCa cells showed that knockdown of GINS1 inhibited BrCa cell growth by delaying DNA replication (Nakahara et al., 2010). This result was consistent with our current data. Another study showed that high expression of GINS1 in cancer cells promoted cell proliferation, transplantation, and metastatic properties (Nagahama et al., 2010). Overexpression of PSF1 was reported non-small lung cancers, and its expression was useful as a prognostic marker (Kanzaki et al., 2016). These findings indicated that aberrantly expressed GINS1 was involved in cancer pathogenesis.
Anlotinib is a newly developed multitarget receptor tyrosine kinase inhibitor used for patients with treatment failure non-small cell lung cancer with metastases (Shen et al., 2018). Interestingly, GINS1 was identified as an anlotinib-mediated downstream gene, and knockdown of GINS1 markedly inhibited the proliferation of synovial sarcoma cells (Tang et al., 2019). Aberrant expression of cell cycle-regulated genes is a common molecular mechanism of cancer cell malignancies, and these genes are potential cancer therapeutic targets. Cyclin-dependent kinases, i.e. CDK4 and CDK6, are essential for transition from the G 0 /G phase to the S phase of the cell cycle. Recently, several CDK4/6 inhibitors (e.g. abemaciclib, palbociclib, and ribociclib) have been developed, and several clinical trials have demonstrated the therapeutic effects of these inhibitors on hormone receptor-positive/HERnegative BrCa (Iwata, 2018;Matutino et al., 2018;Spring et al., 2019). Clinical trials of CDK4/6 inhibitors are also progressing in other subtypes of BrCa (Iwata, 2018;Spring et al., 2019). Our current data showed that knockdown of GINS1 could markedly suppress malignant phenotypes in BrCa cells by affecting several cell cycle-and DNA replication-controlled genes. Controlling genes involved in DNA replication may represent a potential approach for cancer treatment. Thus, GINS1 could be a novel diagnostic and therapeutic target for patients with BrCa.

Conclusion
We produced a novel RNA-sequencing-based BrCa miRNA signature. Our signature revealed that several novel miRNA, including passenger strands of miRNA, were downregulated in BrCa tissues. The BrCa miRNA signature created in this study established a basis for exploring new molecular RNA networks in BrCa. This is the first report demonstrating that miR-101-5p (the passenger strand of the miR-101 duplex) acted as a tumor-suppressive miRNA in BrCa cells. Several oncogenic targets regulated by miR-101-5p were closely associated with BrCa pathogenesis and oncogenesis. Moreover, we demonstrated that GINS1, which we identified from analyses of genes regulated by miR-101-5p, may be a novel diagnostic and therapeutic target in BrCa. Our approach based on analysis of miRNA signatures could contribute to elucidation of the molecular pathogenesis of cancer. SM, and TI; resources, KM, TF, JH, YK, and SN; writing-original draft preparation, HT and NS; writing-review and editing, NS, SK, and SN; visualization, HT, YY, and NN; supervision, NS; funding acquisition, NS, SK, and SN.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. A partial sequence of the 3 0 untranslated region (3 0 -UTR) of the GINS1 gene. A putative binding site for miR-101-5p is shown in the 3 0 -UTR. Fig. S2. Sequences of miR-101-1 and miR-101-2 in the human genome. Stem-loop sequences of miR-101-1 and miR-101-2; red characters indicate mature miRNA. Fig. S3. Cell cycle assays (flow cytometry) in MDA-MB-231 cells with ectopic expression of miR-101-5p and siGINS1. Cell cycle phase distributions (G 0 /G 1 , S, and G2/M) are shown in the bar chart. By transfection of miR-101-5p and siGINS1, G 0 /G 1 phase arrest was detected in MDA-MB-231 cells. Fig. S4. Incorporation of miR-101-5p into the RISC in BrCa cells. Mature miRNA (miR-101-5p and miR-101-3p) were transfected into MAD-MB-231 cells, and incorporated miRNA was immunoprecipitated using anti-Ago2 antibodies. Incorporated miRNA was evaluated by qRT-PCR (*P < 0.0001). Expression of miR-21-5p was used for normalization. Error bars are represented as mean AE SD. P-values were calculated using Bonferroni-adjusted Mann-Whitney U-test. Fig. S5. The strategy for identification of miR-101-5p target oncogenes in BrCa cells. Fig. S6. Direct regulation of TPD52 and SRPK1 by miR-101-5p in BrCa cells. Dual luciferase reporter assays showed that luminescence activities were reduced by cotransfection with wild-type vectors (A: TPD52 and B: SRPK1) and miR-101-5p in MDA-MB-231 cells. Normalized data were calculated as Renilla/firefly luciferase activity ratios (*P < 0.001). Error bars are represented as mean AE SD. P-values were calculated using Bonferroni-adjusted Mann-Whitney U-test. Fig. S7. Inverse correlation between expression of miR-101-5p and GINS1 in BrCa patients (TCGA database analysis, n = 1006), as detected by Spearman's rank tests (P = 0.00103, r = -0.082). Fig. S8. Expression of GINS1 was significantly reduced by siGINS1 transfection into MDA-MB-157 cells (A). Functional assays, cell proliferation (B), migration (C), and invasion (D), in MDA-MB-157 cells with transfection of miR-101-5p and siGINS1. Cell proliferation, migration, and invasion assays were described in Materials and Methods (2.4 and 2.5). *P < 0.001, **P < 0.05. Error bars are represented as mean AE SD. P-values were calculated using Bonferroni-adjusted Mann-Whitney U-test.  Table S1. Reagents used in this study. Table S2. Annotation of reads aligned to small RNA. Table S3. Downregulated miRNA in BrCa compare with normal breast (guide/passenger strand).