MiR-7 reduces the BCSC subset by inhibiting XIST to modulate the miR-92b/Slug/ESA axis and inhibit tumor growth

Background Breast cancer stem cells (BCSCs) are typically seed cells of breast tumor that initiate and maintain tumor growth. MiR-7, as a cancer inhibitor, decreases the BCSC subset and inhibits tumor progression through mechanisms that remain unknown. Methods We examined miR-7 expression in breast cancer and developed a BCSC-driven xenograft mouse model, to evaluate the effects of miR-7 overexpression on the decrease of the BCSC subset in vitro and in vivo. In addition, we determined how miR-7 decreased the BCSC subset by using the ALDEFLUOR, lentivirus infection, dual-luciferase reporter, and chromatin immunoprecipitation-PCR assays. Results MiR-7 was expressed at low levels in breast cancer tissues compared with normal tissues, and overexpression of miR-7 directly inhibited lncRNA XIST, which mediates the transcriptional silencing of genes on the X chromosome, and reduced epithelium-specific antigen (ESA) expression by increasing miR-92b and inhibiting slug. Moreover, miR-7 suppressed CD44 and ESA by directly inhibiting the NF-κB subunit RELA and slug in breast cancer cell lines and in BCSC-driven xenografts, which confirmed the antitumor activity in mice injected with miR-7 agomir or stably infected with lenti-miR-7. Conclusions The findings from this study uncover the molecular mechanisms by which miR-7 inhibits XIST, modulates the miR-92b/Slug/ESA axis, and decreases the RELA and CD44 expression, resulting in a reduced BCSC subset and breast cancer growth inhibition. These findings suggest a potentially targeted treatment approach to breast cancer.


Introduction
Breast cancer is often treated successfully with surgery and conventional therapies; however, approximately 25% of patients eventually develop treatment resistance and recurrence of metastatic disease, which is a major cause of mortality in women worldwide [1]. Studies have demonstrated that the cancer stem cells (CSCs) are typically seed cells of tumors that initiate and maintain tumor growth, and traditional chemotherapy is usually able to shrink the tumor mass, but often fails to completely eradicate CSCs [2][3][4].
Accumulating evidence suggests that noncoding microRNA (miR-7) is a breast cancer suppressor because it inhibits Kruppel-like factor-4 (KLF-4), which is one of the oncogenes that enhance breast cancer growth [5][6][7]. High expression of miR-7 significantly reduced both the capacity of self-renewal and invasion of breast cancer stem-like cells (BCSCs). Interestingly, the long noncoding RNA (lncRNA), terminal differentiation-induced lncRNA (TINCR), is required for normal epidermal differentiation by a posttranscriptional mechanism [8]. Recently, a report showed that TINCR modulated KLF4 expression by competing with miR-7, to mediate its oncogenic function [9,10]. MiR-7 can downregulate SET domain bifurcated 1 (SETDB1), an oncogene that is overexpressed in breast cancer, resulting in the inhibition of tumor progression and metastasis in mice, accompanied by a decrease in the BCSC subset in tumor tissues, as in our previous report [6]. The exact molecular mechanism, however, has yet to be elucidated. Therefore, the present study aimed to further investigate the mechanisms of miR-7-mediated reduction of the BCSC subset and elucidate its molecular linkage in breast cancer cells and in breast cancer xenografts in mice.
We report that lncRNA XIST (X-inactive-specific transcript), which is involved in genomic imprinting and the transcriptional silencing of genes on the X chromosome and is highly expressed in tumor tissues [11][12][13][14][15], is an epigenetic regulator targeted by miR-7 that directly bound and inhibited XIST expression in breast cancer by using the RNA in vivo precipitation (RIP) method, which underlies the attenuation of properties of BCSCs and the decrease of the BCSC subset. MiR-7 also modulated the miR-92b/Slug/ESA (epithelium-specific antigen) axis and reduced CD44 and Slug expression, resulting in a decrease in cancer cell stemness and the BCSC subset in breast cancer. Furthermore, miR-7 overexpression could suppress the tumorigenicity of CD44 + CD24 − ESA + BCSCs and reduce the BCSC subset in breast cancer animal model. Our study characterizes the regulation of XIST by miR-7 and provides evidence that overexpression of miR-7 or downregulation of XIST underlies the attenuation of properties of BCSCs and the decrease of the BCSC subset, which may be a potential strategy for breast cancer therapy.

Cell culture
Human breast cancer SK-BR-3, MCF-7, and MDA-MB-231 cells were obtained from the Cellular Institute in Shanghai, China. The LD cell line was established by our lab from a human breast cancer postsurgery sample that was cultured and characterized as shown in Additional file 1: Figure S1. SK-BR-3 and MCF-7 cells were maintained in 1640 medium. MDA-MB-231 and LD cells were cultured in DMEM.

Human breast cancer samples
Human breast cancer postsurgery samples were obtained from the Department of General Surgery of Zhongda Hospital at Southeast University in China. The investigation was approved by an ethics committee at Southeast University School of Medicine, and informed consent for the use of the postsurgery samples was obtained from the donors who were breast cancer patients. The clinical data of 12 breast cancer patients and specimens are shown in Additional file 4: Supplementary Table S1.

RT-qPCR
Reverse transcription quantitative real-time PCR (RT-qPCR) analyses were used and performed on an ABI Ste-pOnePlus Real-Time PCR System. Total cellular RNA was isolated from each sample by using a Qiagen RNeasy Kit (Qiagen, Valencia, CA). One microgram of total RNA per sample was subjected to cDNA synthesis using the Superscript III Reverse Transcriptase (Invitrogen). The mRNA levels of the genes of interest were expressed as the ratio of each gene of interest to β-actin mRNA per sample. cDNAs were amplified by PCR with primers as listed in Additional file 4: Supplementary Table S2 [16].

Western blotting
Cells and tumor tissue homogenates were lysed in protein extraction buffer (Novagen, Madison, WI, USA) according to the manufacturer's protocol. Briefly, proteins (10 μg/lane) were electrotransferred onto a polyvinylidene difluoride membrane, after which the membrane was blocked with 4% dry milk in Tris-buffered saline with Tween-20 for 1 h and then incubated with a rabbit antibody specific to human Slug (Affinity, Cincinnati, OH, USA), ESA (Proteintech, Chicago, USA), NF-κB subunit RELA (Santa Cruz Biotechnology, CA, USA), or CD44 (Proteintech, Chicago, USA) overnight at 4°C. The membrane was rinsed for 5 min with an antibody wash solution 3 times before adding goat anti-rabbit fluorescently labeled secondary antibody (LI-COR, Lincoln, USA) overnight at 4°C. Immunoreactive bands were detected by Odyssey scanning instrument (LI-COR Odyssey Imaging System, USA) [17].

Plasmid construction and dual-luciferase reporter assays
We used the miRcode algorithm (release 6.2, http:// www.mircode.org/) to search for miR-RNA targets of XIST, RELA, and Slug. Wild-type/mutated XIST, RELA, and Slug were generated by PCR from human genomic DNA. For the luciferase reporter assay, the PmirGLO Dual-Luciferase miRNA Target Expression Vector was used [6,16]. The oligonucleotide sequences (wild-type) were used are shown in Additional file 4: Supplementary  Table S3. Luciferase reporter assays were performed using the Dual-Luciferase Reporter Assay System (Promega) [21,22].

Infection with lentivirus encoding miR-7 vector
Lentivirus encoding miR-7 or empty vector was used to infect MDA-MB-231 or LD cells as previously described [6]. The clones with stable miR-7 expression were selected by GFP expression.

Chromatin immunoprecipitation (ChIP)-PCR assay
The ChIP-PCR assay was performed with Simple ChIP Assay Kits (Beyotime Institute of Biotechnology, Hangzhou, ZJ, China, Cell Signaling Technology). Briefly, cells were treated with 37% formaldehyde for 12 min and subjected to ultrasonication on ice. The DNA-protein complexes were immunoprecipitated with anti-RELA or anti-Slug antibody (Abcam, Cambridge, UK). The bound DNA fragments were then isolated by ChIP reactions and subjected to PCR using primers specific to the promoters of CD44 or (See figure on previous page.) Fig. 1 Detection of miR-7 and BCSC-related molecular expression. a Expression of miR-7, XIST, miR-92b, RELA, CD44, Slug, and ESA in breast cancer postsurgery samples analyzed by RT-qPCR. b-h Relative expression levels of miR-7 and XIST, miR-7 and RELA, miR-7 and Slug, miR-92b and XIST, miR-92b and Slug, RELA and CD44, and ESA and Slug in that order. i Relative expression levels of RELA, CD44, Slug, and ESA analyzed by Western blotting. All the data represent the mean ± S.D. (n = 12). Blue points represent adjacent noncancerous tissues; red points represent tumor tissues. N noncancerous tissues, T tumor tissues  ESA. The primers used in the ChIP assay are listed in Additional file 4: Supplementary Table S5 [24,25].

RIP assay
RNA pulldown assays were performed using a Target RNA Purification Kit (Shanghai Zeheng Biotech, China). All of the following experiments were performed as described previously [26] with small modifications. The probes used are listed in Supplementary Table S6. Cells were crosslinked by formaldehyde, equilibrated in glycine buffer and scraped with lysis buffer. Cell samples were sonicated and then centrifuged. The supernatants were transferred to a 2-ml tube, and 50 μl was saved as the input sample. Cell lysates were incubated with the biotintagged-specific probe or control probe for 3 h. The lysate supernatant was incubated with streptavidin beads for 1 h with rotation. The beads-sample mixture was washed. Subsequently, 10% of the beads-sample mixture was subjected to TRIzol extraction. Purified mRNA and miRNAs were detected by qRT-PCR assay using an All-in-One miRNA qRT-PCR Detection Kit (Gene Copoeia, USA).

In vivo experiment
Female nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice at 6 weeks of age and weighing 18 ± 1 g were ordered from Beijing Weitong Lihua Experimental Animal Technology Co., Ltd., China. All mice were raised in an SPF level animal facility at the Experimental Animal Center, School of Medicine, Southeast University.
The animal experiments followed the guidelines of the Animal Research Ethics Board of Southeast University. BCSCs (2 × 10 5 ) were injected into the right inguinal mammary fat pads in each mouse. Twenty-one days after injection, tumors of about 3-5 mm were found in the mice. At this point, twelve tumor-bearing mice were randomly divided into the following three groups of equal numbers (four per group): (1) the BCSC group (phosphate-buffered saline (PBS) locally injected into the BCSC-driven tumor site); (2) the drug group received the following: 30 mg/kg, adriamycin (Ad, purchased from Dalian Meilun Biotech Co., Ltd. of China, twice a week for a total of 3 weeks) and 10 mg/kg cyclophosphamide (Cy, purchased from Dalian Meilun Biotech Co., Ltd. of China, once a week for 3 weeks in 0.1 ml PBS); both were locally injected into the BCSC-formed tumor site; and (3) the group treated with miR-7 agomir, which was synthesized by the Rioribo Company and performed as described previously, whereby 2 nmol miR-7 agomir in 0.1 ml saline buffer was locally injected into the tumor site once every 3 days for a total of 7 times. Tumor development was assessed for up to 42 days. The tumor tissues were subjected to Western blotting, RT-qPCR, and immunohistochemistry assays. All experiments were repeated twice [16].

Statistics
Values of interest are presented as the mean plus or minus standard deviation. Statistical comparisons were performed using Student's t test. A P value < 0.05 was considered statistically significant.

MiR-7 and BCSC-related molecular expression in breast cancer
To identify miR-7 and BCSC-related molecular expression levels in breast cancer, we collected 12 postsurgery samples from breast cancer patients and used them in RT-qPCR. The results showed that mostly miR-7 expression was significantly lower in breast cancer tissues than in the adjacent noncancerous tissues (p = 0.0318). Based on these results, we further investigated whether the expression of XIST, miR-92b, RELA, CD44, slug, and ESA was related to low miR-7 expression. In contrast to miR-7 expression, the relative expression levels of XIST, RELA, CD44, slug, and ESA were significantly higher in breast cancer tissues than in adjacent noncancerous tissues (p = 0.0486, p = 0.038, p = 0.0297, p = 0.0420, and p = 0.0434, respectively) (Fig. 1a).
Interestingly, the lower miR-7 was, the higher XIST, RELA, CD44, slug, and ESA were, but the relative expression levels of XIST, RELA, CD44, slug, and ESA changed concurrently (Fig. 1c-e, g, h). Moreover, the expression levels of RELA, CD44, Slug, and ESA were further confirmed by Western blotting analyses (Fig. 1i). In addition, we found that miR-92b expression was similar to that of miR-7 and was decreased in breast cancer tissues compared with the adjacent noncancerous tissues (p = 0.0477).

MiR-7 regulates the expression of BCSC-related molecules in breast cancer cells
RT-qPCR showed that miR-7 expression in clinical samples was negatively correlated with the expression of XIST, RELA, CD44, slug, and ESA. We further evaluated whether these expression levels would be changed in   Figure S1), while the SK-BR-3 cell line exhibits ER − / PR − /HER2 + . Interestingly, FCM analysis showed that CD44 expression was concurrently decreased in the four cell lines transiently transfected with miR-7 mimic, but not with miR-7 mimic control; CD44 expression was simultaneously rescued in MDA-MB-231 (Fig. 2a), MCF-7 (Fig. 2d), SK-BR-3 (Fig. 2g), and LD cells (Fig. 2j) after transfection with a miR-7 inhibitor. Similarly, the expression levels of XIST, slug, and ESA were synchronously reduced but miR-92b expression was increased in the miR-7 mimic transiently transfected cells (shown in Additional file 2: Figures S2 a to d).
Since NF-κB subunit RELA function may be implicated in CD44 expression [27], we further analyzed RELA and CD44 expression by RT-qPCR and Western blotting. Consistent with the XIST, Slug, and ESA expression patterns, CD44 and RELA expression levels were also reduced in the miR-7 mimic-transfected cells (Additional file 2: Figures S2 e to g,  Fig. 2b, e, h, and k). Furthermore, CD44 expression was concomitantly decreased in RELA-downregulated cells (Additional file 3: Figure S3). It is thus evident that the synthetic miR-7 mimic acts as a regulator to modulate the expression of BCSC-related molecular genes in breast cancer cell lines.

MiR-7 inhibits CD44 expression by targeting RELA
To explore the mechanism of miR-7-mediated regulation of RELA and CD44 expression, a dual-luciferase reporter assay was first conducted in MDA-MB-231 cells. By TargetScan and miRcode algorithm prediction, we found that RELA was one of the miR-7 candidate genes and the RELA 3′-UTR contained two miR-7 binding sites (Fig. 3a). The following reporter assay was performed as previously described [6,16]. The results showed the miR-7 mimic reduced the relative luciferase activity of the wild-type vector, which contained the putative miR-7 binding sites in the RELA 3′-UTR, but not the relative luciferase activity of the mutant A or B vector or the A plus B vector (Fig. 3b). Next, we tested whether RELA had enhanced occupancy in the promoter region of its target gene CD44 and conducted a ChIP assay. Based on the JASPAR database prediction, we found that there were seven putative RELA-binding sites in the CD44 promoter (Fig. 3c). ChIP-PCR results indicated that RELA was directly bound to the − 1234 to − 1243, − 1654 to − 1663, and − 2073 to − 2082 regions in the CD44 promoter in MDA-MB-231 cells (Fig. 3d). To further confirm these findings, we investigated whether silencing RELA could decrease CD44 expression in MDA-MB-231, MCF-7, and SK-BR-3 cells. As shown in Fig. 3e-j, the CD44 transcriptional and translational expression levels were significantly decreased after transfection with siRELA recombinants in comparison with the control cells.

MiR-7 directly targets XIST and slug to decrease ESA but increases miR-92b expression
To explore the effects of miR-7 on XIST and the miR-92b/Slug/ESA axis, we found that XIST contained three predicted binding sites for miR-7 (Fig. 4a) and one for miR-92b (Fig. 4e) based on TargetScan and miRcode algorithm prediction. The results indicated that the miR-7 or miR-92b mimic significantly decreased the relative luciferase activity of the wild-type vector compared with the control (Fig. 4b-f), suggesting that the inhibition of XIST expression was regulated by miR-7 and miR-92b. Next, we found that there were no sites for miR-7 in the ESA 3′UTR, but the slug mRNA contained one, as shown in Fig. 4g. The result in Fig. 4h shows that miR-7 reduced the relative luciferase activity of the wild-type vector but not the mutant vectors, indicating the inhibition of slug expression. Additionally, we further used biotin-tagged XIST antisense oligonucleotides (XIST probe) and performed RIP assay to pull down the XIST complex by beads and then identified the pulldown of miR-7-5p from the XIST complex precipitate to confirm miR-7 targeting of XIST in MDA-MB-231 cells [26]. shown in Fig. 4j. Specific isolation of miR-7-5p from the XIST probe and control probe (input) in cells is shown in Fig. 4k. Figure 4l represents the RT-PCR analysis of miR-7-5p isolation efficiency in cells. These data strongly demonstrated that, in addition to the dualluciferase reporter assay, the RIP results further provided evidence that miR-7 could actually bind to XIST in MDA-MB-231cells. Furthermore, we conducted ChIP-PCR in MDA-MB-231 cells to determine whether slug would directly increase ESA transcription by binding to the ESA promoter region. Figure 5a shows two putative Slug-binding sites in the ESA promoter. ChIP-PCR results indicated that Slug was bound to PCR-amplified fragments at two DNA binding sites (Fig. 5b).
Interestingly, RT-qPCR results also showed that miR-92b expression was increased in Lenti-miR-7-infected MDA-MB-231 cells (Fig. 5c), which may be due to inhibition of XIST. To verify this, we silenced XIST by siRNA and found that knockdown of XIST enhanced miR-7 and miR-92b expression, as shown in Fig. 5d, e. More interestingly, there was one binding site for miR-92b in the Slug 3′UTR (Fig. 5f). Figure 5j shows that miR-92b decreased the relative luciferase activity of the wild-type Slug 3′UTR vector due to the inhibition of slug expression in MDA-MB-231 and LD cells. Moreover, we validated the RT-qPCR results by Western blotting and found that miR-7 and miR-92b respectively inhibited the expression of slug and ESA in MDA-MB-231 cells following transfection of miR-7 or miR-92b mimic (Fig. 5h, i). Overall, although Slug is a strong transcription factor and directly bound to regions in the ESA promoter to induce ESA expression, miR-7 and miR-92b could directly downregulate Slug expression, resulting in reduced ESA expression.

MiR-7 reduces BCSC-driven tumor growth by decreasing the BCSC subset
Furthermore, we evaluated the potential role of miR-7 agomir in suppressing BCSC growth in vivo. To this end, we established a BCSC-driven xenograft model by locally injecting of 2 × 10 5 BCSCs, which were isolated from MDA-MB-231 cells, into the right inguinal mammary fat pads of NOD/SCID mice. All mice generated tumor xenografts 17 days after injection. The tumor-bearing mice were treated with miR-7 agomir, Ad plus Cy, or PBS control 4 days later. Figure 6a contains photographs of breast cancer-bearing mice 21 days after treatment. The mouse weight changes and the dynamic tumor growth were monitored (Fig. 6b, c). As shown in Fig. 6b, we found that weight loss was approximately 5% on day 7 after initiation of miR-7 agomir treatment (from day 21 to day 28), and it mostly maintained mouse weight similar to that of control mice 21 days after treatment (from day 21 to day 42). This suggested that local injection of the miR-7 agomir caused no severe side effects. However, in response to the combined Ad plus Cy treatment, the mice showed serious weight loss, suggesting severe side effects in the drug-treated mice. Figure 6d shows photographs of the tumors 21 days after treatment. The quantification analysis of tumor sizes is shown in Fig. 6e, in which the tumor sizes were slightly smaller in mice treated with miR-7 agomir than mice treated with Ad plus Cy, but the difference was not statistically significant (p > 0.05). Furthermore, the tumor sizes in the two groups of mice were much smaller than those in the mice treated with PBS control (p < 0.0208 or p < 0.0161). The counts of CD44 + CD24 − BCSCs (Fig. 6f) and CD44 + CD24 − CD44 + BCSCs (Fig. 6h) were also significantly decreased in the tumor tissues of mice treated with the miR-7 agomir and in the Ad plus Cy group (Fig. 6g-i). Western blotting results showed that the expression levels of REAL, CD44, Slug, and ESA were concurrently reduced (Fig. 6j-k).

Discussion
Crucial concerns for the BCSC subset are nearly related with the chemoresistance and recurrence of breast cancer; we explored the mechanisms of decreasing the BCSC subset regulated by miR-7. Our findings in the present study revealed that miR-7 and miR-92b expression levels were markedly decreased in breast cancer surgery samples in contrast to the high expression of lncRNA XIST, RELA, CD44, slug, and ESA, suggesting a negative correlation between miR-7/miR-92b and the expression of BCSC stemnessrelated molecules in vivo. In Fig. 2 and Figure S2 a to d, we found that the expression levels of XIST, RELA, CD44, Slug, and ESA were remarkably decreased in the miR-7 mimic transiently transfected cells, especially in the case of XIST, which may regulate the stemness characteristics and tumorigenicity of some types of cancer cells, resulting in cancer progression  [28][29][30][31][32][33]. To verify these claims, we attempted to identify whether XIST expression was regulated by miR-7/miR-92b. The results indicated that miR-7/ miR-92b could significantly inhibit the reporter activity of XIST, whereas knockdown of XIST led to high miR-7/miR-92b expression, which suggested the reciprocal regulation between miR-7 and XIST in MDA-MB-231 and LD cells.
To further confirm that miR-7 directly targets XIST in breast cancer, we performed a RIP assay. The results in Fig. 4 g to j showed that miR-7 could directly bind to XIST in MDA-MB-231 cells, which fully demonstrated that miR-7 could inhibit XIST expression.
It is known that RELA is not only a member of the NF-κB family and participates in many biological processes but is also involved in the NF-κB signaling pathway, which correlates with BCSC proliferation activity [28]. Thus, we evaluated whether RELA could directly regulate CD44 expression. As expected, ChIP-PCR results showed that RELA had increased occupancy in the promoter region of CD44, and this was further confirmed by knockdown of RELA, which resulted in reduced CD44 expression. Next, we observed that miR-7 could directly regulate RELA expression, as confirmed by the luciferase reporter assay. Since CD44 is tightly associated with the BCSC phenotype [10,16,34], we hypothesized that miR-7 decreased the BCSC subset mainly by targeting RELA to inhibit CD44 expression. Our findings agreed with a recent report by Cantley et al., which reported a targeted relationship between miR-7 and the NF-κB subunit RELA and miR-7 inhibition of RELA expression could impact the proliferation, invasion, and apoptosis of non-small cell lung cancer cells. However, the report did not involve the CSC subset [35].
In the present study, we also observed low miR-92b and high XIST expression concurrently in breast cancer tissues and cell lines. These findings were in agreement with several studies on XIST and miR-92b expression in tumors [13][14][15][36][37][38].
In addition, we found that miR-7 negatively regulated XIST expression but positively regulated miR-92b expression and that high miR-92b expression enhanced the inhibition of transcription factor Slug activity, resulting in decreased ESA expression. The miR-7 mimic not only decreased ESA expression by indirectly increasing miR-92b expression to inhibit slug activity but also directly downregulated Slug to reduce ESA expression. These epigenetic regulatory effects finally led to a decrease in the BCSC subset percentage in breast cancer.
It is possible that XIST serves as a competing endogenous RNA to bind miR-7 to inhibit its function. Since while XIST was knocked down, the expression of miR-7 was clearly increased. Similarly, miR-92b exhibited an inverse relationship with XIST gene expression. Consequently, the results from the in vitro experiments demonstrated that the miR-7 mimic could exert multiple reciprocal regulatory actions to inhibit the expression of XIST, RELA, Slug, CD44, and ESA but increase miR-92b expression.
In vivo animal experiments demonstrated that the local treatment of the BCSC-driven xenograft with the miR-7 agomir had the same effects as applying the combined treatment of Ad plus Cy in a mouse model, which resulted in a reduction in tumor growth and improvement of tumor-bearing mouse survival as well as decrease in BCSC phenotypes (CD44, ESA). A combination of Ad plus Cy treatment of breast cancer is well known as one of the conventional therapies in the past few decades [39,40]. Our results suggested a potential value of miR-7 agomir as a local treatment reagent in the treatment of breast cancer.
(See figure on previous page.) Fig. 6 MiR-7 agomir reduces BCSC-driven xenograft growth. a Representative images show the BCSC xenograft growth in NOD/SCID mice 21 days after treatment with Ad+Cy, miR-7 agomir, or PBS. b Mouse weight changes. c Tumor dynamic growth plot. d Tumor sizes 21 days after treatment. e Quantification of tumor sizes. f-i FCM analysis of CD44 + CD24 − and CD44 + CD24 − ESA + BCSC percentages. j, k Western blotting analysis of CD44, RELA, Slug, and ESA expression levels in tumor tissues Fig. 7 MiR-7 decreases the BCSC subset by inhibiting XIST to modulate the miR-92b/Slug/ESA axis. Notes: XIST long noncoding RNA X-inactive-specific transcript, ESA epithelium-specific antigen. → Promoting; ⊥ inhibiting; reciprocal inhibiting