Androgen receptor/let-7a signaling regulates breast tumor-initiating cells

Androgen receptor (AR) is an important transcriptional factor, which is frequently expressed in invasive breast cancer and correlates patients’ prognosis. Our previous results indicate AR activation may increase let-7a expression in breast cancer cells, while let-7, a tumor suppressor, is reported to inhibit breast tumor-initiating cells (T-IC). The study aims to explore the effects of AR/let-7a signaling on breast T-IC and its regulatory mechanism. The results revealed that the expression of AR was significantly associated with let-7a and CD44+/24-/low especially in estrogen receptor positive (ER+) breast cancer tissues. The expression of AR and let-7a indicated better outcome, while patients with CD44+/24-/low phenotype had worse prognosis. AR activation induced by dihydrotestosterone (DHT) prevented cells proliferation, migration, invasion and self-renewal capacities in ER+ breast cancer cells, via transcriptional up-regulation of let-7a. In addition, AR could inhibit tumorigenesis and metastasis of ER+ breast cancer cells in the serial xenotranplanted animal models. Our data suggested that AR/let-7a signaling could inhibit the biological behavior of tumor-initiating cells (T-IC) in ER+AR+ breast cancers, which might become a new therapeutic target.


INTRODUCTION
Androgen receptor (AR) is frequently expressed in invasive breast cancers and may be a diagnostic and therapeutic target for breast cancer [1][2][3]. It is recently indicated tumor-initiating cells (T-IC), a small subset in cancer that has the capacities of self-renewal, mutipotent differentiation and tumorigenicity, may be affected by AR activation in prostate cancer, contributing to the progression of the malignant tumors [4][5][6][7]. AR is considered as a significantly prognostic factor in triplenegative breast cancer, which is enriched in T-IC [8].
Breast T-IC can be enriched by sorting for CD44 + /CD24 -/ low cells [7], by selecting for side-population cells that efflux Hoechst dyes [9], or by isolating spherical clusters of self-renewing cells ("mammospheres") from suspension cultures [10]. However, little information is known about the correlations between AR and breast T-IC capabilities in breast cancers.
As a critical transcriptional factor, AR regulates the transcription of various target genes, including MicroRNAs (miRNAs) [11]. MicroRNAs are endogenous, small, non-coding RNAs that interfere with protein expression by inducing cleavage of their specific target

Research Paper
transcripts or by inhibiting their translation, through complementing the 3' untranslated regions in the target transcripts [12]. Let-7 is the earliest discovered miRNAs and encodes an evolutionarily conserved family of 13 homologous miRNAs frequently deleted in variety of human cancers [13]. Let-7a, one of the members of let-7 family, is usually selected as the representative [14,15]. In addition, lack of let-7a is necessary for tumorigenicity and self-renewal of breast T-IC, while let-7a overexpression not only prevents cell proliferative and self-renewal capacities, but also converts higher metastatic T-IC into lower malignant cells [16].
We previously found AR activation induced a robust increase of let-7 family members in breast cancer cell lines, including let-7a [11]. We hypothesize that there are some correlations among AR, let-7a and BT-IC, and they maybe co-regulate the progression of breast cancer. The present study aims to examine the expression of AR, let-7a and CD44 + /CD24 -/low in invasive breast cancer tissue specimens. And the effects of AR/let-7a signalling on the stem-like behavior of breast cancer cells are also evaluated in vitro and in vivo. We may try to elucidate the mechanism of AR regulating let-7a.

AR decreased cell proliferation and invasion by upregulating let-7a in ER+AR+ breast cancer cells
We found more tumor cells with CD44 + /CD24 -/low phenotype could be detected in ER-positive and ARpositive breast cancers (Table 2). Otherwise, few studies focused the effects of AR on let-7a expression and BT-IC properties in this subtype. Therefore, ER+AR+ breast cancer cell lines, MCF7 and T47D were selected to perform the following analysis. Results revealed that compared with control groups, let-7a expression increased had a 1.5-fold increase in MCF7 and an 1.3-fold increase in T47D cells treated with DHT, while inhibiting let-7a activity by let-7a ASO in DHT-treated cells, let-7a was significantly decreased, similar to control groups. We also analyzed the expression of STAT3 and IL-6, known target proteins of let-7a. The level of pSTAT3 and IL-6 secretion decreased when cells exposed to DHT, while no difference was found in STAT3 expression. But when let-7a expression was re-suppressed, the level of pSTAT3 and IL-6 secretion were restored at a similar level of controls ( Figure 2A).
DHT treatment decreased cells proliferation, while re-lowering let-7a recovered cells growth activity, an effect similar to that illustrated without DHT treatment ( Figure 2B). In addition, the cell cycle examination confirmed the results of MTT assay. When cells treated with DHT, large number of cells was prevented in the G1 phase, G1-S arrest was obvious and proliferation index was decreased. However, let-7a re-down regulation restored the cells number and cells proliferation index in the S phase, an effect similiar to that showed in the control groups ( Figure 2C). DHT treatment also decreased cells healing ability, while re-knockdown of let-7a elevated cells healing ability ( Figure 2D). Same results could be seen in the cells invasion. The number of T47D cells that migrated through the insert pores was 162 ± 9.54/field in DHT treated group, which was different significantly with the control group (349 ± 9.64/field) and the let-7a blocked group (357 ± 12.77/field), suggesting that DHT treatment inhibited the invasion of T47D cells, but reknockdown of let-7a abolished the inhibited invasion ( Figure 2E).

AR decreased characteristics of self renewal of breast cancer cells
In breast cancer tissues, AR and let-7a expression were correlated with the phenotype of CD44 + /CD24 -/low . Quantitative RT-PCR result showed that the up-regulation of let-7a expression induced by AR was more obvious in MCF7 than in T47D cells ( Figure 2A). So MCF7 cell line was selected to analyze its capacity of self-renewing when AR was activated or AR expression was inhibited. BT-IC could be enriched by means of isolating spherical clusters of mammospheres from suspension cultures [10]. First, MCF7 cells treated with DHT were cultured in suspension to generate mammospheres. After 12 days, 1.60% ±0.36 of tumor cells formed mammospheres in DHT treated cells as compared with 3.07% ± 0.25 in the control group (P = 0.001, Figure 3A). AR expression was distinctly decreased in MCF7 cells expressing shRNA against AR by lentivirus, especially in 277-ihRNA2 transfected group ( Figure 3B). Therefore, we selected MCF7-277-ihRNA2 as AR knock-down group.
Mammospheres formation results showed that 3.73% ± 0.32 of tumor cells formed mammospheres in MCF7-277-ihRNA2 group, compared with 2.03% ± 0.31 of those in MCF7-277-iluc group (P = 0.003, Figure 3C). CD44 expression was only detected in MCF7-277-ihRNA2 cells ( Figure 3D), suggesting that these cells had the properties of tumor-initiating. Cells gathered from mammospheres were analyzed by flow cytometry assay, and results showed that tumor cells with CD44 + /CD24 -/low phenotype increased significantly from mammospheres than those in control groups ( Figure 3E), suggesting that breast tumor-initiating cells could be enriched by mammospheres cultivation .

AR knocking down promoted tumorigenesis and metastasis capacity
In order to analyze the effect of AR on BT-IC in vivo, we established a subcutaneous xenotransplanted tumor model in NOD/SCID mice with MCF7 cells in a gradually decreased cell density. There were significant differences of tumor volume between MCF7 expressing control and shRNA against AR groups. And tumors grew faster in MCF7-277-ihRNA2 group than in the control group ( Figure 4A-4B). When we decreased the injected cell density, all mice receiving MCF7 cells expressing shRNA against AR still had tumorigenic growth while the tumor formation rate decreased significantly in control group ( Figure 4C). In addition, liver metastasis was found in two of ten mice receiving MCF7-277-ihRNA2 cells with a cell density of 2 × 10 6 , but no metastasis was found in MCF7-277-iluc group. And metastatic cancer cells had high proliferation ability indicated by Ki67 staining ( Figure 4D).

AR transcriptionally regulated let-7a expression by activating let-7a promoter activity
We then analyzed the mechanism of let-7a upregulation by AR activation. DHT treatment induced AR imported into nucleus, indicated by western blot results in both MCF7 and T47D cell lines ( Figure 5A and 5B). We inferred that androgen/AR signaling might transcriptionally mediate the regulation of let-7a in ER+AR+ breast cancer cells, just like the mechanism in ER-AR+ breast cancer cells [11]. As CHIP results shown ( Figure 5C), cells treated with DHT induced an AR enriching on let-7a promoter region than those without DHT stimulation, which was confirmed by the following luciferase analysis. We constructed luciferase expression vectors of let-7a promoter based on androgen response element. The luciferase activity was higher in both AR over-expression MCF7 and T47D cells ( Figure 5D), suggesting that AR could directly activate the let-7a promoter activity. These results suggested that the mechanism of androgen-AR signaling regulating let-7a was also reasonable in ER+/AR+ breast cancer cells.

DISCUSSION
AR is a key transcription factor and mainly functions in nucleus, when being activated by androgens. Studies suggest AR play an important role in breast cancer development [18], in part via the role of let-7 up-regulation induced by AR activation [11]. Cancers may arise from rare self-renewing tumor-initiating cells (T-IC). In prostate cancer, AR expression is previously described to be associated with T-IC, and AR signal can influence T-IC [4]. Let-7 is the earliest discovered miRNAs [13], which may inhibit multiple BT-IC stem cell-like properties  by silencing more than one target [16]. Let-7a is one of the members of let-7 family and usually selected as the representative [14,15]. At present, few are known on the relationship among AR, let-7a and the self-renewing BT-IC in breast cancer, and also their interactions.
In breast cancer tissue, we identified that AR expression was positively correlated with let-7a expression, but negatively with CD44 + /CD24 -/low phenotype of cancer cells. We suppose that AR, let-7a and BT-IC may affect the progression of breast cancer together. AR expression in primary tumors is considered as a better prognostic indicator in a majority of previous studies [2,3,19]. T-ICs are characteristic by selfrenewal, rapid proliferation, differentiated tumor cells and drug resistance, which may be responsible for the recurrence and metastasis of breast cancer [7,10,20]. The present study also indicated that patients with CD44 + / CD24 -/low phenotype tumors had worse outcome. Let-7a expression is decreased in breast cancer, and the loss of let-7 expression is correlated with the development of poorly differentiated and aggressive cancers, which is considered to function as tumor suppressors [16,[21][22][23]. Overexpression of let-7a inhibits the ability of BT-ICs and reduces cells proliferation [24]. However, one report states the oncogenic activity of let-7a [25]. Therefore, more studies should be carried out to analyze the role of let-7a in breast cancer.
AR activation can inhibit estrogen-stimulated proliferation [26], and AR-induced cell apoptosis is also detected in ER+ breast cancer cells [27]. We found that androgen-induced AR activating signal affects cells behavior both in ER+AR+ breast cancer cells, and also in ER-AR+ breast cancer cells [11]. It is reported that AR/ ER ratio can influence cells response to androgen, and increased ratio of AR/ER inhibits E2-induced proliferation [28], which further illustrate that AR may function in breast cancer dependent on ER expression. And some co-activators have an important role in the phenotype of cells with different AR/ER ratios. From the present and our previous study, we concluded that the inhibition of some biological behavior of breast cancer cells induced by AR signaling might be via let-7a up-regulation in part [11]. However, in prostate cancer, let-7 suppresses AR expression and activity, which in turn leads to decreased cell proliferation and tumor growth [29,30]. Therefore, we suspects that AR and let-7a may have complicated associations depending on the different context. And more studies should be carried out to analyze whether let-7a will down-regulate AR expression and affect androgen-AR signaling in breast cancer cells.
The positive association between the CD44 + / CD24 -/low presence and AR expression was more obvious in ER+ breast cancers, which was consistent with our in vitro results. AR activation could decrease the mammospheres formation, while AR knockdown increase in ER+AR+ breast cancer cells when they were cultured in suspension. The flow cytometry assay results confirmed that decreased AR expression significantly increased the percentage of MCF7 cells with CD44 + /CD24 -/low phenotype whether in attachment or in suspension ( Figure 5). And CD44 expression could be detected in AR knockdown cancer cells, but not in control group, indicated by western blot results. There are several publications that suggest AR expression correlates significantly with cancer stem cells mainly in prostate cancers [31,32], but few is known about their correlations in breast cancers.. The important T-IC features are efficient xenograft formation and metastases initiation. In Al-Hajj's study [7], as few as 1000 cells with CD44 + /CD24 -/low phenotype can give rise to tumors in all cases, while up to 2 × 10 4 CD44 + /24 + cells fail to form tumors, which suggests that cancer cells with CD44 + /CD24 -/low phenotype have the tumorigenic capability upon serial xenotransplantation in vivo. We subcutaneously injected MCF7 cells of different AR expression into NOD/SCID mice with a gradually decreased cell density. AR knockdown could promote tumor formation and growth in the mice receiving cancer cells even with a decreased cell number, suggesting an enhanced tumorigenic capability. In addition, two mice developed into liver metastasis in AR knockdown group, but no in controls. No lung metastasis was observed in all mice. It is suspected that the heterogeneity of breast cancer cells contributes to the difference, and more studies should be performed to analyze this phenomenon.
AR is a key transcription factor and androgens induce AR translocate into the nucleus, where mainly function. Several recent studies state that AR can  directly regulate the transcription of miRNAs [33,34] , which is consistent to our study. Our group indicated that AR imported into the nucleus and up-regulated let-7a expression by binding to the androgen response elements of let-7a promoter region in breast cancer cells, upon androgen activation [11,35]. We speculated that the mechanism of AR affecting breast T-IC might be partially via transcriptional regulation of let-7a, since let-7a overexpression inhibits T-IC capabilities in ER+ breast cancers [24]. However, some additional transcription factors or co-regulating proteins may affect breast T-IC in ER-breast cancers.
In conclusion, our findings indicated that DHT-induced AR activation played a critical role in breast cancer, which was not only correlated with let-7a expression, but also related to BT-IC. DHT treatment induced AR translocated into the nucleus and transcriptionally up-regulated let-7a expression directly, and then decreased cells proliferation, migration, invasion, and self-renewal capacities. In addition, AR knockdown promoted tumor-formation and metastasis capacity, even when cancer cells were xenotransplanted into the mice with a decreased cell density. Patients with enhanced AR or let-7a expression predicted a better prognosis, while the tumor cells with CD44 + /CD24 -/low phenotype predicted a worse prognosis index. A better understanding of the effect and mechanism of AR-let-7a signalling and breast T-IC will lead to new therapeutic strategies.

Clinical specimens
One hundred and sixty-five invasive breast cancer tissues were selected, which were diagnosed between January 2005 and May 2005. These patients were registered in the Department of Breast Cancer Pathology and Research Laboratory, Tianjin Medical University Cancer Instituted and Hospital, Tianjin, China. All of these selected patients signed informed consent before the examination of the specimens, and the research was approved by the Institutional Ethic Committee of Tianjin Medical University Cancer Instituted and Hospital. Patients with bilateral primary breast cancers and those who received preoperative treatment were excluded, and their clinicopathological data were available. Participants in the cohort were women ranging in age from 32 to 78 years (median 51 years).

In situ hybridization(ISH) and evaluation of the staining
The MicroRNA ISH Detection Kit III (MK1031, BOSTER, China) and miRNAlet7a probes (38468-05, Exiqon, Denmark) were used for ISH in the present study. All procedures were carried out according to the manufacturer's instructions. Briefly, formalinfixed paraffin-embedded breast cancer tissue were cut into 4-µm sections and deparaffinized. Tissue permeabilization was used by proteinase. After prehybridization, sections were incubated with the hybridization mix and hybridized for overnight at 45 °C. For signal detection, anti-digoxigenin-alkaline phosphatase were used as the primary antibody, and the slides were then subsequently incubated with 5-bromo-4-chloroindo-3-yl-phosphate/nitro blue tetrazolium (BCIP/NBT) solution Counterstaining was performed by Nuclear Fast Red. The stained sections were then scored for expression of let7a miRNA under the microscopy. For the negative control, the hybridization solution was replaced with the pre-hybridization solution.

Cells culture
MCF-7 and T47D human breast cancer cell lines were obtained from American Type Culture Collection (ATCC, USA), and cultured in DMEM medium (Gibco, USA), which contained 10% fetal bovine serum (FBS, Gibco, USA) and 1% penicillin/streptomycin (Life Technologies, USA), and cultured in a incubator with 5% CO 2 at 37ºC.

Lentiviral vector production
Oligos encoding shRNA specific for AR were ligated into pSUPER.reto.puro, and the fragment containing the H1 promoter and hairpin sequences was subcloned into the lentiviral 277 vectors. The shRNA target sequences were as follows: AR-ihRNA1: CGACTACTACAACTTTCCA; AR-ihRNA2: AATGTTATGAAGCAGGGAT.
The plasmids were used to produce lentivirus in 293T competent cells with the packaging plasmids including pMD2.BSBG, pMDLg/pRRE and pRSV-REV.

Real-time reverse transcription PCR (real-time RT-PCR)
Total RNA was extracted from cells with TRIzol (Invitrogen, USA). RT-PCR system (TaKaRa, Japan) was used to reverse RNA, and 1 ul cDNA sample was quantified using primers with SYBR Green PCR Master mix (TaKaRa, Japan) by real-time PCR. Let-7a specific primers were used as previously reported [11]. All analyses were performed in triplicate.

IL-6 enzyme-linked immunosorbent assay (ELISA)
According to the manufacturer's protocol, culture supernatants of cells were harvested and concentrations of IL-6 proteins were evaluated by ELISA kits (R&D Systems, China). Briefly, protein was pipetted into a microplate containing IL-6 antibody and incubated for 2 h; IL-6 antibody with enzyme-linked was added and incubated for 1 h; added a substrate solution and incubated for 15 min. When positive control wells emerged blue color, stopping reagent was added to stop the reaction. Then samples were detected at 450 nm absorbance.
Flow cytometry analysis 5 × 10 5 cells were collected and fixed with methanol. Then cells were stained by propidium iodide (Sigma-Aldrich); finally all samples were analyzed by flow cytometer. Proliferation index (PI) was calculated as follows: PI = (S+G2/M)/G1. Cells were incubated with PE-labeled anti-CD24 and FITC-labeled anti-CD44 (Miltenyi Biotec, Germany) and then analyzed by flow cytometer for cell surface staining.

MTT, transwell and scratch assay
Cell proliferation and viability were measured by MTT assay. The procedure was carried out as previously reported [11]. For transwell assay, 24-well transwell chamer plates were used. 1 × 10 5 cells suspended with serum-free medium were seeded into the upper chamber (8 μm pore size, Corning, USA), which was coated with Matrigel (BD, USA) diluted by serum-free medium. Complete medium was added into the bottom chamber. The invaded cells were stained with Gimesa staining solution after 24 h and counted. For scratch assays, cells (5 × 10 5 ) were plated into cell culture dishes and cultured in serum free DMEM. The monolayer was wounded by a yellow pipette tip, and then removed the cellular debris with PBS. The exact initial wound areas were taken photographs every 2 days.

Tumor implantation
Indicated numbers of MCF7 cells were injected into the mammary fat pads of NOD/SCID mice (6-8 weeks, female). Before injection, all mice were treated with oral estrogen for two weeks. Within the following 30 days, all mice were examined for tumor formation by palpation. The volume calculated as follows: Volume (mm 3 ) = L.W 2 /2, and tumor size was measured every 3 days.

Chromatin immunoprecipitation
Cells were fixed using 1% formaldehyde for cross-linking the DNA and protein. ChIP was performed according to the ChIP Assay Kit's protocols (17-371, Upstate, USA). Briefly, sonicated nucleoprotein was incubated with 10µg of AR antibody (ab74272, Abcam, USA). 1µg polymerase antibody was replaced as positive control and 1µg normal mouse IgG as negative control. Sixty microliters of protein G-Sepharose beads was added for every reaction. The Sepharose immunocomplex slurries were incubated at 65ºC overnight to reverse crosslinks and were then RNase treated, proteinase K treated, deproteinized, and precipitated as described previously [37]. Immunoprecipitated and input DNA fractions were analyzed by PCR. In order to investigate the binding of AR and androgen response element (ARE) regions of the let-7a promoter, we first analyzed the potential ARE in Let-7a promoter regions by PROMO soft, and then designed the specific primers for the 251bp DNA ARE fragment in let-7a promoter region. They were as follows: let-7a-ARE-F: TTTTACATTGGGCATAGCCG; let-7a-ARE-R: TAGGCATTTGGAAGTTGGAC. The PCR products were detected by 1% agarose gel electrophoresis and analyzed with PCR Gel Image Analysis System.

Luciferase assay
Androgen response element of the 3200bp sequences upstream let-7a 5' end region was cloned into the luciferase reporter expression vector (pEZX-PG04, Genecopoeia, USA) AR transcript 1A CDS region was cloned into the pReceiver-M12 vectors (EX-M12, GeneCopoeia, USA). MCF7 and T47D cells were seeded into the 6-well plates at a density of (2-4) × 10 5 cells/ well. Luciferase let-7a promoter expression and AR overexpressing plasmids were transiently co-transfected into the cells using Lipofectamine 2000 (Invitrogen). Then the luciferase activities were evaluated by a Secrete-Pair Dual Luminescence Assay Kit (Genecopoeia), following the manufactures' instructions at the time of 24-48 hours after transfection. Luciferase activities were detected with a luminometer, and then normalized by luciferase activity in luciferase reporter.

Statistical analysis
Statistical analysis was performed by SPSS 19.0 statistical software. The data were reported as the Means ± SD. Statistical analysis was carried out using one-way ANOVA in comparison with he corresponding controls. Differences between AR, let-7a, CD44 + /CD24 -/ low and clinicopathological characteristics were analyzed via the Chi-square test. The association among AR, let-7a and CD44 + /CD24 -/low was tested using Spearman rank correlation with Pearson's test. Based on the Kaplan-Meier method, disease-free survival (DFS) and overall survival (OS) curves were generated. Differences between the curves were evaluated by the log-rank test. Probability values of < 0.05 were regarded as statistically significant.