Transforming growth factor-β regulates the sphere-initiating stem cell-like feature in breast cancer through miRNA-181 and ATM.

Recent studies indicate that a subset of cancer cells possessing stem cell properties, referred to as cancer-initiating or cancer stem cells (CSCs), have crucial roles in tumor initiation, metastasis and resistance to anticancer therapies. Transforming growth factor (TGF)-β and their family members have been implicated in both normal (embryonic and somatic) stem cells and CSCs. In this study, we observed that exposure to TGF-β increased the population of breast cancer (BC) cells that can form mammospheres in suspension, a feature endowed by stem cells. This was mediated by the micro (mi)RNA family miR-181, which was upregulated by TGF-β at the post-transcriptional level. Levels of the miR-181 family members were elevated in mammospheres grown in undifferentiating conditions, compared with cells grown in two-dimensional conditions. Ataxia telangiectasia mutated (ATM), a target gene of miR-181, exhibited reduced expression in mammospheres and upon TGF-β treatment. Overexpression of miR-181a/b, or depletion of ATM or its substrate CHK2, was sufficient to induce sphere formation in BC cells. Finally, knockdown of ATM enhanced in vivo tumorigenesis of the MDA361 BC cells. Our results elucidate a novel mechanism through which the TGF-β pathway regulates the CSC property by interfering with the tumor suppressor ATM, providing insights into the cellular and environmental factors regulating CSCs, which may guide future studies on therapeutic strategies targeting these cells.

Introduction

The TGF-β ligands are multitasking cytokines that play important roles in embryonic development, cell proliferation, motility and apoptosis, extracellular matrix production and modulation of immune function (Massague, 2008). These ligands signal through the heteromeric complex of transmembrane serine/threonine kinases, the type I and type II receptors (TβRI and TβRII), and activate both the Smad family of transcription factors and non-Smad signaling pathways (Derynck and Zhang, 2003). TGF-β plays a dual role in cancer: it limits proliferation in epithelial cells and early-stage cancer cells, whereas in late-stage cancer, it accelerates cancer progression and metastasis (Dumont and Arteaga, 2003; Roberts and Wakefield, 2003). In the cancer niche, TGF-β can be produced and secreted into the extracellular environment by both cancer cells and host cells, such as lymphocytes, macrophages and dendritic cells. In cancer patients, high levels of TGF-β at tumor sites correlate with high histological grade, risk of metastasis, poor response to chemotherapy, and poor patient prognosis (Dumont and Arteaga, 2003).

TGF-β signaling has been implicated in CSCs, or cancer-initiating cells, which are defined as a subset of cancer cells possessing stem cell properties. CSCs are considered the “seeds” of cancer for their crucial roles in tumor initiation, metastasis and resistance to anticancer therapies. They resemble embryonic stem cells (ESCs) and somatic (adult) stem cells by their abilities to self-renew and to undergo multilineage differentiation. Characterization of CSCs has been demonstrated in leukemia and solid tumors of the breast, lung, colon, prostate, pancreas, brain and head and neck [reviewed in (Ailles and Weissman, 2007; Visvader and Lindeman, 2008)]. In most of these studies, CSCs are prospectively isolated by immunosorting based on the expression of various stemness or multilineage related surface markers (Charafe-Jauffret ; Dontu, 2008; Visvader and Lindeman, 2008). Gene expression profiling suggests that the TGF-β pathway is active in CD44+ BC cells that are enriched for breast cancer stem cells (BCSCs), where its inhibition induces a more epithelial phenotype (Shipitsin ). In addition, the epithelial-mesenchymal transition (EMT) induced by TGF-β treatment or expression of Snail or Twist increases the number of transformed cells with BCSC properties (Mani et al., 2008).

A functional enrichment strategy relying on the characteristics of mammary stem cells (MSCs) to escape anoikis and grow into mammospheres in anchorage-independent conditions has been successfully used to obtain highly enriched and functional MSCs from both normal and cancerous breast tissue, as well as from BC cell lines (Dontu et al., 2003). Each of these mammospheres is clonally originated from one MSC, typically contains ~300 cells undergone various levels of differentiation, and averagely maintains one sphere-initiating MSC (for self-renewal) and some progenitor cells that can differentiate into both epithelial and myoepithelial lineages in Matrigel (Dontu et al., 2003). The sphere-forming efficiency (SFE) is thereby used to assess the number of MSCs in the bulk of normal or cancer cells, and ranges 0.1%~0.7% in normal mammary epithelial cells, whereas 1%~3% in BC cell lines (Charafe-Jauffret et al., 2009; Dontu et al., 2003). Genes that are differentially expressed by sphere cells highlight pathways implicated in maintaining the stem cell status (Charafe-Jauffret ). In this study, we examined the function of TGF-β in regulating the BC population with the sphere-forming CSC feature, and identified a novel miRNA-mediated mechanism that targets the ATM tumor suppressive pathway.

Results

Identification of the miR-181 family members as TGF-β target genes

MiRNAs are naturally-occurring non-coding small RNA molecules that play crucial functions in cells by base pairing to the 3′ untranslated region (UTR) of target mRNAs, resulting in mRNA degradation or translation inhibition. To explore the role of TGF-β in miRNA regulation, we performed miRNA array analysis in the BC cell line MDA231 and the non-transformed mammary epithelial cells MCF10A. The miRNAs were filtered for expression changes greater than 1.5-fold after 24 h of TGF-β treatment; 13 miRNAs in MDA231 and 11 miRNAs in MCF10A cells were identified (Fig. 1A). Predicted target genes of these miRNAs were analyzed using TargetScanHuman 5.1 (Supplemental Table 1). Genes that were common to both cell lines were selected and mapped to their corresponding gene networks by Ingenuity Pathways analysis (IPA 8.6) (Fig. S1 and Supplemental Table 2).

Fig. 1

Identification of the miR-181 family as TGF-β target genes. A. MiRNAs altered by TGF-β treatment. Total RNAs were prepared from MDA231 and MCF10A cells treated with TGF-β (2 ng/ml) or vehicle for 24 h in 3 independent experiments, and subjected to Agilent miRNA microarrays. The differentially expressed miRNAs were selected by statistical criteria of greater than 1.5-fold changes (log base 2 = ±0.58) and p<0.05 by Students T-test (indicated in log 2 value). FC: fold change (in log 2) of TGF-β-treated sample compared to control sample. B. MDA231 and MCF10A cells were treated with TGF-β (2 ng/ml) or vehicle for indicated time. Total RNA was isolated and subjected to quantitative RT-PCR for miR-181a and miR-181b. Data was normalized to the level of U6, and then compared to that in untreated cells, which was set as 1. Each data point represents the mean ± S.D. of 3 wells. C. BC cell lines BT474, MDA361 and MCF7 were treated with TGF-β at the indicated final concentrations or vehicle for 72 h. Total RNA was isolated and subjected to quantitative RT-PCR for miR-181a and miR-181b.

It was noticed that members of the miR-181 family were upregulated by TGF-β in both cell lines (Fig. 1A). This miRNA family includes the guiding strands miR-181a/b/c/d that share the same seed sequence, and the passenger strands miR-181a* and miR-181a-2* that are present at lower abundances. It has been reported that miR-181 family members play an important role in hepatic cancer stem cells by targeting hepatic transcriptional regulators of differentiation (Ji et al., 2009). Elevated levels of miR-181 are observed in the cancer of breast, prostate and pancreas (Volinia et al., 2006). We thereby focused on the miR-181 family in this study, for their potential function in BCSCs as the downstream effectors of TGF-β signaling. Quantitative RT-PCR confirmed that TGF-β induced the expression of miR-181a and miR-181b in both MDA231 and MCF10A cells (Fig. 1B). A dose-dependent induction was also observed in other three BC cell lines, BT474, MDA361 and MCF7 (Fig. 1C).

The context-dependent effect of TGF-β on the sphere-initiating CSC-like feature

An in vitro cultivation system that allows for propagation of human mammary epithelial cells and BC cells in an undifferentiated state, based on their ability to proliferate in suspension as nonadherent mammospheres, has been established and used in several recent reports (Cicalese ; Dontu ; Mani ; Pece ). Using this approach, we examined the sphere forming efficiency (SFE) in a panel of BC cell lines. Because MDA231 and MCF10A cells could not form mammospheres using this system, three BC cell lines that could, i.e., BT474, MDA361 and MCF7, were selected for further investigation. All three cell lines exhibited SFE of 1~3%, which is consistent with the reported CSC populations in primary human cancers (Fig. 2A). The spheres formed from these cell lines also exhibited the typical size reported in others’ studies, and expressed the stem cell markers Oct4 and Nanog (Fig. 2B). Immunofluorescence assay further indicated that the spheres but not parental cells grown in 2D contained a small number of cells expressing CK5, a marker of mammary stem cells (Fig. 2C). The majority of the sphere cells expressed CK8/18, markers of differentiated cells, suggesting the heterogeneity of spheres formed from BC cell lines (Fig. 2C). When cells were pretreated with TGF-β for 3 days, the SFE of BT474 and MDA361 cells was significantly induced in a dose-dependent manner, whereas in MCF7, TGF-β decreased the SFE (Fig. 2D). These results suggest a context-dependent mechanism in the regulation of CSCs by TGF-β. Since both BT474 and MDA361 cells naturally overexpress HER2, whereas MCF7 cells are ER+/PR+/HER2−, we examined if overexpression of HER2 affected the effect of TGF-β on MCF7 sphere formation. HER2-transduced MCF7 cells (MCF7/HER2) were generated, and the overexpression and phosphorylation/activation of HER2 were confirmed by Western blot. However, similar to the control MCF7 cells, the SFE of MCF7/HER2 cells was also decreased by TGF-β (Fig. S2), indicating that other factors in BT474 and MDA361 cells are responsible for the context-dependent function of TGF-β.

Fig. 2

The context-dependent effect of TGF-β on the sphere-initiating CSC-like feature. A. BC cell lines BT474, MDA361 and MCF7 contained a small cell population that could initiate spheres when cultured in suspension. Single cells were plated in ultralow attachment plates as described in Materials and Methods, so that cells with stem cell properties were allowed to grow as non-adherent spheroids (mammospheres). Images of the mammospheres were captured on day 7. Numbers of the mammospheres (diameter ≥ 70 μm) were counted, and the sphere forming efficiency (SFE) was calculated based on the numbers of cells that were initially seeded. Each data represents the mean ± S.D. of 3 wells. Bars equal 25 μm. B. Protein extracts were prepared from the parental (P) cells growing in 2D, 7-day sphere (S) cells, and cells that were dissociated from 7-day spheres and subsequently cultured in 2D differentiating condition (SD). Western blot was performed using indicated antibodies. GAPDH was used as a loading control. C. IFA images of MDA361 spheres and 2D cultured cells stained with antibodies against CK5 (green) and CK8/18 (red). DAPI: nuclei staining. The bar equals 25 μm. D. Cells grown in 2D were treated with TGF-β at the indicated concentrations for 72 h, before harvested for sphere formation assay in the absence of TGF-β. Spheres were counted on day 7 and SFE was calculated.

MiR-181 regulates sphere formation

The reported function of miR-181 in hepatic cancer stem cells urged us to examine its role in regulating BCSCs. Levels of the miR-181 family members were elevated in sphere cells by 2~4-fold, compared to the parental BT474, MDA361 and MCF7 cells grown in 2D; their expression levels were reduced when sphere cells were dissociated and re-plated to grow in 2D (Fig. 3A). This is unlikely to be an effect of the medium used in mammosphere culture, as the same medium and culture condition did not induce miR-181 expression in the MDA231 BC cells, which could not form mammospheres (Fig. S3). A hairpin inhibitor of miR-181a that efficiently inhibited miR-181a and partially inhibited miR-181b reduced both basal and TGF-β-induced SFE in BT474 and MDA361 cells, but had no effect in MCF7 cells (Fig. 3B). When a plasmid carrying the miR-181a/b gene cluster in chromosome 1 was transfected into the cells to overexpress miR-181, increased SFE was observed in both transfected BT474 and MDA361 cells, but not in MCF7 cells (Fig. 3C). Notably, the overexpression levels of miR-181a/b in plasmid transfected cells were comparable to those in TGF-β-treated cells (Fig. 1C), and were sufficient to significantly induce sphere formation in both BT474 and MDA361 cells (Fig. 3C, top panels). Therefore, although the induction of miR-181 by TGF-β was modest (1.5–2.5 folds), it seems to be sufficient to exert an effect on sphere formation. These data also suggest that TGF-β induces sphere formation through upregulating miR-181, which induces this stem cell phenotype in a context-dependent manner that requires certain factor(s) or functional link(s) present in BT474 and MDA361, but not MCF7 cells. We also compared levels of several previously reported cancer-related miRNAs in sphere cells and 2D-cultured parental cells. MiR-21 was also elevated in the spheres of BT474 and MDA361 cells (Fig. S4). However, a miR-21 hairpin inhibitor did not affect SFE as potently as the miR-181a inhibitor, when transfected alone or in combination with the miR-181a inhibitor (Fig. 3D).

Fig. 3

MiR-181 regulates sphere formation. A. Total RNA was isolated from parental cells, sphere cells and sphere-to-differentiation cells as described in Fig. 2B, and subjected to qRT-PCR for miR-181a, miR-181a*, miR-181b and miR-181d. Data was normalized to the level of U6, and then compared to that in parental cells, which was set as 1. B. Cells grown in 2D were transiently transfected with miR-181a hairpin inhibitor or a control reagent. TGF-β (3.5 ng/ml) or vehicle was added at 6 h after transfection. After 72 h, cells were harvested; half were used for RNA extraction and qRT-PCR of miR-181a/b (bottom), and half for sphere formation assay (top). SFE was assessed on day 7. C. Cells grown in 2D were transiently transfected with a miR-181a/b expression plasmid or the vector control. At 72 h after transfection, cells were analyzed for the levels of miR-181a/b by qRT-PCR (bottom) and for SFE (top). * p<0.005. D. BT474 cells grown in 2D were transiently transfected with the hairpin inhibitors of miR-181a or miR-21, either alone or in combination at equal amount. At 72 h after transfection, cells were analyzed for the levels of miR-181a and miR-21 by qRT-PCR (bottom) and for SFE (top).

TGF-β induces miR-181a/b at the post-transcriptional level

Two distinct mechanisms have been reported in the regulation of miRNAs by TGF-β. The TGF-β downstream effectors Smads are reported to bind to and activate the promoter of miR-155 (Kong ). Whereas in the regulation of miR-21, Smad2/3 bind to the primary transcript of miR-21 through interacting with the Drosha miRNA processing complex, which facilitates miR-21 maturation (Davis ). To investigate which mechanism is involved in the regulation of miR-181, we examined levels of the primary (pri-) miR-181a-1 and the precursor (pre-) miR-181a-1 in TGF-β-treated cells by qRT-PCR. Although TGF-β induced the mature forms of miR-181 (Fig. 1C), it decreased their primary and precursor forms in all four cell lines tested (Fig. 4A), suggesting the regulation occurs at the level of miRNA maturation. It has been reported that MDA231 cells do not undergo Smad4 translocation into the nucleus in response to TGF-β stimulation (Ren ). In these cells, decreased pri- and pre-miR-181a-1 levels and increased mature miR-181 levels were still observed (Fig. 1B & 4A), consistent with the reported observation that the Smad2/3-Drosha interaction is Smad4-independent (Davis ). In RNA immunoprecipitation (RIP)-coupled RT-PCR, pri-miR-181a-1, but not the mature miR-181a, was detected in the precipitates of Smad2/3 and Drosha, but not IgG, in a TGF-β-inducible manner (Fig. 4B), suggesting that similar to the regulation of miR-21, TGF-β induces binding of Smad2/3 to the primary transcripts of miR-181 and regulates their maturation. Smad4 knockdown using siRNA did not interfere, but instead increased, miR-181 levels and SFE (Fig. 4C), suggesting that lower levels of Smad4 may contribute to a switch of Smad2/3 function from Smad4-mediated transcriptional regulation to Drosha-mediated miRNA maturation.

Fig. 4

TGF-β induces miR-181a/b at the post-transcriptional level. A. Quantitative RT-PCR of the primary (pri-) miR-181a-1 and the precursor (pre-) miR-181a-1 in cells treated with TGF-β at the indicated concentrations for 72 h. Data was normalized to GAPDH in each sample and compared to that of untreated cells, which was set as 1. B. RIP was carried out in BT474 cells treated with TGF-β (3.5 ng/ml) or vehicle for 24 h. Antibodies against Smad2/3, Drosha or IgG (as a negative control) were used in the immunoprecipitation as described in the Materials and Methods. Quantitative RT-PCR of pri-miR-181a-1 and the mature miR-181a were performed, using the pre-RIP RNA sample without reverse transcription reaction (−RT) as a negative control. Data was compared to the control RIP sample with IgG. C. BT474 cells grown in 2D were transiently transfected with siRNA targeting Smad4 or control siRNA. After 72 h, cells were analyzed for the levels of miR-181a/a*/b/d by qRT-PCR (left) and for SFE (right). * p<0.005.

ATM, a miR-181 target, suppresses sphere formation through phosphorylating CHK2

One of the predicted miR-181 target gene is ATM, a serine/threonine kinase that functions as a tumor suppressor. In the function annotations of the predicted target genes of TGF-β-regulated miRNAs (Fig. S1), ATM appeared in 11 out of the 12 top ranked functional networks (Supplemental Table 2). To determine if ATM is a real target of miR-181, we first interrogated the 3′UTR sequence of human ATM gene transcript using TargetScanHuman 5.1 (www.targetscan.org) and miRDB (www.mirdb.org). Two potential miR-181a/b/c/d targeting sites, at the positions 123 and 3525 in the ATM 3′UTR, were identified (Fig. 5A). We then cloned these two putative miR-181 binding regions, either together or individually, downstream to a luciferase reporter gene in psiCHECK vector, and transfected 293 cells with these constructs or a control vector containing a scrambled sequence. Co-transfection with the plasmid expressing miR-181a/b efficiently suppressed expression of the luciferase reporter followed by either putative miR-181 binding site, but not the scrambled sequence; the reporter construct containing both miR-181 binding sites showed the strongest inhibition by miR-181a/b (Fig. 5B). Consistently, TGF-β also suppressed these ATM 3′UTR reporters containing one or two miR-181 binding sites (Fig. 5B). Suppression of the ATM 3′UTR reporter by overexpressed miR-181a/b and TGF-β treatment was also observed in transfected BT474, MDA361 and MCF7 cells (Fig. 5C).

Fig. 5

ATM, a miR-181 target, suppresses sphere formation through phosphorylating CHK2. A. The miR-181a/b/c/d targeting sites in the 3′UTR of ATM mRNA predicted by TargetScanHuman 5.1 (www.targetscan.org) and miRDB (www.mirdb.org). B. The psiCHECK luciferase reporters containing the miR-181 targeting site at 123, the site at 3525 or both sites, or a control reporter containing a scrambled sequence, were used to transfect 293 cells, together with a miR-181a/b expressing plasmid or vector. For the TGF-β treatment experiments, cells were transfected with each reporter; TGF-β (3.5 ng/ml) or vehicle was added at 24 h after transfection. Luciferase activity was analyzed at 48 h post transfection, and compared to the cells transfected with the control plasmids. C. The luciferase reporter containing both miR-181 targeting sites in ATM 3′UTR was transfected into BT474, MDA361 or MCF7 cells, together with the miR-181a/b expressing plasmid or vector. For the TGF-β treatment experiments, cells were transfected with the reporter only; TGF-β (3.5 ng/ml) or vehicle was added at 24 h after transfection. Luciferase activity was analyzed at 48 h post transfection, and compared to the cells transfected with the control plasmids. D. Protein extracts of parental (P) and sphere (S) cells were prepared and subjected to Western blot analysis. GAPDH was used as a loading control. E. BT474 cells grown in 2D were transfected with a miR-181a hairpin inhibitor (or control reagent) or a miR-181a/b expressing plasmid (or vector). TGF-β or vehicle was added at 6 h after transfection. After 72 h, cells were harvested; half were used for RNA extraction and qRT-PCR of miR-181a (bottom), and half for protein extraction and Western blot analysis (top). F. Total RNA isolated from cells treated with TGF-β at the indicated concentrations for 72 h was analyzed for ATM mRNA level by qRT-PCR. Data was normalized to GAPDH in each sample and compared to that of untreated cells. G. Cells grown in 2D were treated with TGF-β or vehicle for 72 h before lysed and subjected to Western blot analysis. H. Cells grown in 2D were transiently transfected with siRNA targeting ATM, CHEK2, BRCA1, p53 or control siRNA. At 72 h after transfection, cells were analyzed for SFE as described in Fig. 3B. * p<0.005. I. Western blot of the indicated proteins in cells transfected with ATM siRNA or control siRNA.

The levels of ATM significantly reduced in the spheres formed by all three BC cell lines (Fig. 5D), which was consistent with the elevated miR-181 levels in the spheres (Fig. 3A). The miR-181a inhibitor increased basal ATM expression and abolished the suppressive effect of TGF-β on ATM, whereas miR-181a/b overexpression decreased ATM protein level (Fig. 5E). Treatment with TGF-β, which induced miR-181, reduced the mRNA levels of ATM in BT474 and MCF7, but not significantly in MDA361 (Fig. 5F). Nonetheless, at the protein level, ATM was significantly suppressed by TGF-β in all three cell lines (Fig. 5G). Similar to the pattern of cell line specific regulation by TGF-β, knockdown of ATM by siRNA significantly increased SFE in BT474 and MDA361, but not MCF7 cells (Fig. 5H). These results indicate that the distinct effect of TGF-β and miR-181 in different cell lines is due to a context-dependent function of ATM.

ATM is an important cell cycle checkpoint kinase that phosphorylates a wide variety of substrates, including p53, BRCA1 and CHK2 (Kastan and Lim, 2000). To further identify which ATM downstream effector is involved in the regulation of sphere formation, we individually knocked down the expression of p53, BRCA1 and CHK2 using siRNAs. Knockdown of CHK2, but not the other two genes, induced SFE in all three cell lines tested (Fig. 5H). Reduced CHK2 phosphorylation, as a result of the reduced ATM levels, was observed in the spheres formed by all three cell lines, compared to the cells grown under regular culture conditions (Fig. 5D). This suggests that the ATM effector CHK2 functions as a suppressor of sphere formation. Western blot further indicated that treatment with TGF-β, overexpression of miR-181a/b, or transfection of ATM siRNA all reduced CHK2 phosphorylation at Thr68, a reported ATM phosphorylation site (Kastan and Lim, 2000), in BT474 and MDA361, but not MCF7 cells (Fig. 5E, 5G & 5I). These results further suggest that in BT474 and MDA361 cells, ATM negatively regulates sphere formation via activating CHK2, which may be controlled by another kinase in MCF7 cells.

Knockdown of ATM enhances the in vivo tumorigenesis of BC cells

To further examine the role of ATM in tumorigenesis, we constructed MDA361 cells stably expressing doxycycline (Dox)-inducible ATM shRNA. Treatment with Dox efficiently decreased the protein level of ATM (Fig. 6A) and induced sphere formation (Fig. 6B). When 5 × 105 cells were injected into the mammary fat pads of immunocompromised mice, 3 out of 5 mice treated with Dox formed tumors within 3 weeks, whereas no tumor was developed in the control group (−Dox; n=5) (Fig. 6C).

Fig. 6

Knockdown of ATM enhances the in vivo tumorigenesis of BC cells. A. MDA361/tetO-shRNA(ATM) cells were treated in the absence or presence of Dox (1 μg/mL) for 48 h and analyzed for ATM expression by Western blot. B. The same cells were treated in the absence or presence of Dox for 48 h before analyzed for SFE as described. * p<0.005. C. NSG mice were injected in the no. 4 mammary fat pad with 5 × 105 of MDA361/tetO-shRNA(ATM) cells, and divided into 2 groups (5 mice per group) for treatment with Dox or control. Tumor formation was indicated by Tumors/Injections at 3 weeks after injection.

Discussion

Similar to embryonic and somatic stem cells, the self-renewal and differentiation of CSCs are simultaneously regulated by intrinsic (cancer cell-endowed) and extrinsic (microenvironmental) factors. Here we reported that TGF-β, a cytokine whose level is often elevated in the tumor microenvironment and associated with advanced breast cancers, stimulated the signature phenotype of CSCs to proliferate in suspension as nonadherent mammospheres. This regulation of CSCs by a microenvironmental factor is dependent on certain intrinsic pathways within cancer cells, such as the signaling axis of ATM and CHK2. As a result, CSCs that carry different genetic or epigenetic alterations may respond differently to the same cues in the cancer niche. Other factors in the tumor microenvironment that regulate CSCs, and how CSCs, in turn, modify the cancer niche and regulate their neighbor cells are yet to be identified.

Members of the TGF-β family have been implicated in the development of various organs and the maintenance of ESC pluripotency (Topczewska ; Watabe and Miyazono, 2009). Nodal and activin have been reported to maintain pluripotency of human ESCs by controlling the expression of Nanog, a critical transcriptional factor for the “stemness” status, through binding of Smad2/3 to Nanog promoter (Vallier et al., 2009). In our study, TGF-β treatment induced Nanog expression in all three BC cell lines tested (Fig. 5G), whereas the sphere-forming CSC property was only induced in BT474 and MDA361, but not MCF7 cells (Fig. 2D). This suggests that increased expression of Nanog is not sufficient to induce the sphere-forming phenotype of CSCs. Instead, knockdown of CHK2 consistently induced sphere formation in all BC cell lines (Fig. 5H). Although the role of the miR-181/ATM/CHK2 axis in the regulation of embryonic and somatic stem cells needs to be further investigated, it is likely that in cancer, both induction of Nanog and suppression of CHK2 function through the mechanism identified herein mediate the regulation of CSCs by TGF-β.

It has been reported that levels of the miR-181 family members are elevated in EpCAM-positive hepatic cancer stem cells and in embryonic livers (Ji et al., 2009). In another study, significant upregulation of miR-181b and miR-181d is observed in the livers of mice during early carcinogenesis (Wang et al., 2010). Expression of the tissue inhibitor of metalloprotease 3 (TIMP3), another validated target of miR-181, is markedly suppressed in these livers. TGF-β is found to induce miR-181b through a Smad4-dependent mechanism in hepatic cells, as knockdown of Smad4 by siRNA interferes with miR-181b expression in these cells (Wang et al., 2010). In contrast, our data indicated that in BC cells, Smad4 knockdown instead increased miR-181 expression and SFE (Fig. 4C). The entire miR-181 family is encoded by three genomic loci in chromosomes 1, 9 and 19, and the transcription of these loci is controlled by different promoter regions without sequence homology. Our data herein suggested that at least in BC cells, TGF-β upregulates the entire miR-181 family at the post-transcriptional level through the Smad4-independent functions of Smad2/3, such as their interaction with Drosha. This upregulation simultaneously increases both the guiding strands and the passenger strands. Since the passenger strands usually undergo rapid degradation and exist at much lower basal levels compared to the guiding strands, their fold induction by TGF-β treatment seemed to be more significant (miR-181a* and miR-181a-2*, Fig. 1A). However, it is also possible that TGF-β has a specific effect on the stability of miR-181a* and miR-181a-2*, resulting in further increases of these passenger strands. Argonaute proteins, the effector molecules in miRNA-mediated RNA interference, are involved in multiple miRNA-related functions, including the incorporation of miRNA into the RNA-Induced Silencing Complex (RISC), cleavage of the target mRNA, miRNA maturation, and removal of the passenger strand from RISC after maturation (Diederichs and Haber, 2007). Whether TGF-β regulates miRNA maturation and the fate of the passenger strands through affecting the function of Argonaute proteins is an interesting direction to further investigate.

In the study by Davis et al., the pre-miR-21 level is increased upon TGF-β treatment (Davis ), whereas in our study, both pri- and pre-miR-181a-1 levels decreased in TGF-β-treated cells (Fig. 4A). Following Drosha-mediated cleavage of the primary transcripts, the miRNA hairpin precursors are further processed by the Dicer RNase and/or Argonaute proteins to generate the mature miRNAs (Diederichs and Haber, 2007). Therefore, the miRNA precursor, as an intermediate during miRNA maturation, only transiently exists and undergoes rapid turnover. As such, the levels of miRNA precursors detected by PCR in TGF-β-treated cells are affected by their turnover time, which may be different for each miRNA regulated through the Smad2/3-Drosha mechanism. Our results further suggest that the level of Smad4 may determine the function of Smad2/3 by altering the equilibrium between Smad4-mediated transcriptional regulation (favored at a high Smad4 level) and Drosha-mediated miRNA maturation (favored at a low Smad4 level). It was recently reported that the tumor suppressor p53 interacts with the Drosha processing complex through the association with DEAD-box RNA helicase p68 (DDX5) and facilitates the processing of primary miRNAs to precursor miRNAs (Suzuki ). Since p53 can interact with Smad2 (Cordenonsi ), whether p53 plays a role in mediating the interaction between Smad2/3 and Drosha needs further investigation, and may reveal a functional link between the p53 and TGF-β pathways in regulating miRNA biogenesis.

A novel miR-181 target, ATM, was identified in this study. ATM is a key regulator of the DNA damage response through phosphorylating a variety of proteins involved in DNA repair, cell cycle regulation and apoptosis (Kastan and Lim, 2000). Consistent with this function, ATM deficient tumors have been shown to be more sensitive to DNA double-strand break (DSB)-inducing agents (Tribius et al., 2001). Small molecule inhibitors of ATM have also been shown to sensitize cancer cells to DNA-damaging drugs, and are proposed to be used as drug-sensitizing agents for anti-cancer chemotherapy. However, a number of studies suggest an opposite effect of ATM mutation/deletion, which correlates with resistance to DNA-damaging chemotherapy and poor patient survival (Austen et al., 2007; Haidar et al., 2000). Our results indicated that suppression of ATM or CHK2 could induce the sphere-forming CSC phenotype. Since CSCs have been implicated in resistance to chemotherapy, it is possible that mutation/deletion of ATM or CHK2, or their downregulation by factors such as TGF-β and miR-181, contribute to drug resistance through regulating the CSC population. Similarly, pharmaceutical inhibition of ATM is unlikely to have a beneficial effect due to its potential influence on CSCs. Further studies are required to evaluate the therapeutic value of TGF-β/miR-181 interventions for their effects on CSCs and drug resistance.

Materials and Methods

Cell lines, plasmids and viruses

This information can be found in the Supplemental Materials

Sphere formation assay

Mammosphere culture was performed as described by Dontu et al. with slight modifications (Dontu et al., 2003). Single cells were plated in ultralow attachment plates (Corning; Corning, NY) at a density of 4,000 cells/mL in serum-free DMEM/F12 (Invitrogen; Carlsbad, CA) supplemented with 10 ng/mL basic fibroblast growth factor (bFGF; Sigma-Aldrich), 20 ng/mL epidermal growth factor (EGF; Invitrogen), 5 μg/mL insulin (Sigma-Aldrich), and 0.4% bovine serum albumin (BSA; Sigma-Aldrich). On day 7~9, numbers of the mammospheres (diameter ≥ 70 μm) were counted, and SFE calculated based on the numbers of initially seeded cells. To ensure that each mammosphere was clonally originated from a single cell, cells grown in the regular 2D conditions were labeled with PKH67 green fluorescent or PKH26 red fluorescent cell linkers (Sigma-Aldrich) following the manufacturer’s protocol, and mixed at a 1:1 ratio, prior to initial seeding for sphere forming culture. After 7~9 days, spheres were monitored under a Nikon Eclipse TE2000-S fluorescent microscope (Nikon; Melville, NY), and >95% of the spheres were labeled with a single dye (data not shown). Mammospheres of 7~9 days were collected by gentle centrifugation (320 g), washed with 1×phosphate buffered saline (PBS), and subjected to RNA or protein preparation (described below). Some mammospheres were enzymatically dissociated by incubation in trypsin-EDTA solution (Invitrogen) for 2 min at 37 °C. Single cell suspensions were then plated in tissue-culture coated plates to allow differentiation under the regular 2D culture conditions. To study the effect of TGF-β on mammosphere formation, cells were treated with TGF-β at the indicated dosages for 72 h before plating in undifferentiating sphere-culture conditions in the absence of TGF-β. When cells were transfected with plasmids, siRNAs or miRNA inhibitors, TGF-β was added at 6 h post transfection, and cells were treated for 3 days before sphere culture.

RNA extraction, reverse transcription (RT) and real time quantitative PCR (qPCR)

This information can be found in the Supplemental Materials.

RNA immunoprecipitation (RIP) assay

RIP was performed using a protocol modified from the chromatin immunoprecipitation (ChIP) assay described previously (Wang et al., 2005). In brief, cells were cross-linked for 10 min with 1% formaldehyde, lysed, and sonicated (7 watts, 10 sec ×6). The lysates were cleared and subjected to immunoprecipitation with Smad2/3 or Drosha antibodies, or normal rabbit IgG (as a control). Precipitated RNA was isolated using TRIzol (Invitrogen) and subjected to RT and qPCR, as described above, using primers to detect pri-miR-181a-1 and the mature miR-181a. Samples precipitated with IgG were used as controls.

Western blot analysis

Preparation of cell lysates and Western blot were carried out as described previously (Wang ). Primary antibodies included: Oct4, Nanog, GAPDH, Phospho(P)-HER2Y1248, HER2, ATM, P-CHK2T68, and CHK2 (Cell Signaling; Danvers, MA).

Immunofluorescence assay (IFA)

IFA was carried out as described previously (Wang ). Fluorescent images were captured using a Princeton Instruments cooled CCD digital camera from a Zeiss upright LSM 510 2-Photon confocal microscope with a 20×/0.6 objection. Immunostaining of the mammospheres was performed in suspension, after fixation in 4% paraformaldehyde for 2 h followed by 20 min in cold methanol, following the same protocol as above. Primary antibodies include Cytokeratin (CK) 5, CK8/18 (Abcam; Cambridge, MA), E-cadherin and N-cadherin (Santa Cruz Biotechnology; Santa Cruz, CA). The fluorescent antibodies are Alexa Fluor 488-goat-α-rabbit IgG and Alexa Fluor 594-goat-α-mouse IgG (Invitrogen).

Cell transfection and RNA interference (RNAi) studies

DNA transfection was performed using Lipofectamine 2000 (Invitrogen) following the manufacturer’s protocol, as described previously (Wang et al., 2006). The miRIDIAN miRNA hairpin inhibitors of miR-181a and miR-21 as well as the negative control were purchased from Dharmacon (Lafayette, CO). Silencer siRNAs against human Smad4, ATM, CHEK2, BRCA1 and p53 as well as the All Stars negative control siRNA were purchased from Qiagen. MiRNA inhibitors and siRNAs were transfected into cell lines using DharmaFECT Duo Transfection Reagent (Dharmacon) according to the manufacturer’s procedures. In 6-well plate format, a final concentration of 25 nM miRNA inhibitors or 100 nM siRNAs and 6 μL of DharmaFECT Duo Transfection Reagent mixed in 2 mL of serum-free medium were used for each transfection.

In vivo tumorigenesis

Six-week-old female NOD/SCID/IL2Rγ-null (NSG) mice were injected in the no. 4 mammary fat pad with 5 × 105 of MDA361/tetO-shRNA(ATM) cells, and divided into 2 groups (5 mice per group) for treatment with Dox or control. For the group with Dox treatment, cells were pretreated in vitro with Dox (1 μg/mL) for 2 days before injection, and animals were administered 1 mg/mL Dox in 5% sucrose through drinking water starting at 2 days before cell injection. Mice were monitored for tumor formation twice weekly. Experimental protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at City of Hope.

MiRNA microarray and Ingenuity Pathways analysis of predicted miRNA target genes

This information can be found in the Supplemental Materials.