The Wnt–β-catenin pathway represses let-7 microRNA expression through transactivation of Lin28 to augment breast cancer stem cell expansion

Many pathways are implicated in the regulation of stem cell self-renewal and differentiation. These pathways include Hedgehog, Notch and Wnt (Liu et al., 2010; Merchant and Matsui, 2010; Takebe et al., 2011; Wend et al., 2010), within which the Wnt–β-catenin pathway is crucial in both normal mammary development and tumorigenesis (Roarty and Rosen, 2010). β-catenin plays a central role as a transcriptional activator in the Wnt–β-catenin pathway (Logan and Nusse, 2004). In the absence of upstream signalling stimulation, cytoplasmic β-catenin is phosphorylated by glycogen synthase kinase-3β (GSK-3β) in a complex that includes Axin and the tumour suppressor adenomatous polyposis coli (APC) and is targeted for ubiquitin-mediated proteasomal degradation. Stimulation by Wnt ligands leads to the inhibition of phosphorylation and degradation of β-catenin, which subsequently enters the nucleus and binds to the LEF1/TCF family of transcription factors. In turn, the β-catenin–LEF1/TCF complex activates the expression of a cohort of target genes that impact cellular functions (Amit et al., 2002; Liu et al., 2002). As the Wnt–β-catenin pathway can regulate mammary stem cell self-renewal and differentiation, the aberrant activation of this pathway in human breast cancer may indicate a key role for it in cancer stem cells. Despite the fact that the Wnt–β-catenin pathway is a key regulator of breast cancer stem cells, the downstream transcriptional cascade in this process remains largely unknown.

MicroRNAs (miRNAs) play critical roles in many biological processes, including cancer processes, by directly inhibiting the expression of target mRNAs through a variety of molecular mechanisms (Bartel, 2009; Ventura and Jacks, 2009). Also, miRNAs undergo aberrant regulation during carcinogenesis and can act as either oncogenes or tumour suppressors (Lu et al., 2005). Let-7 is an important miRNA family consisting of 12 members with expression that is frequently downregulated in a number of human cancers (Calin et al., 2004; Johnson et al., 2005; Kumar et al., 2008). Moreover, let-7 is also the key regulator of breast cancer stem cell self-renewal and differentiation (Yu et al., 2007). A large body of evidence indicates that many miRNAs can regulate the Wnt–β-catenin pathway through the targeting of Wnt components (Huang et al., 2010); however, little is known regarding whether the Wnt–β-catenin pathway regulates the expression of miRNAs, especially in cancer stem cells.

Recent studies on induced pluripotent stem cells (iPS cells) support the concept that cancer stem cells arise through a reprogramming-like mechanism (Krizhanovsky and Lowe, 2009). By enforcing the expression of so-called reprogramming factors (OCT4, KLF4, SOX2, and Lin28 or c-MYC), differentiated somatic cells can be converted to iPS cells, which can differentiate into any tissue type. Several of these reprogramming factors are upregulated in human tumours and are considered to be putative oncogenes (Krizhanovsky and Lowe, 2009). Thus, iPS cells and cancer stem cells may share similar mechanisms that control the stem cell properties. Lin28, one of these reprogramming-like factors, is highly abundant in embryonic stem cells and developing cells, but in later stages, its expression declines (Moss and Tang, 2003). Lin28 may be overexpressed in some tumours (Viswanathan and Daley, 2010). Moreover, Lin28 has been found to be positively correlated with the percentage of ALDH1⁺ tumour cells in breast cancer, suggesting that Lin28 plays an important role in the regulation of breast cancer stem cells (Yang et al., 2010). Lin28 and Lin28B function as negative regulators of let-7 biogenesis through their ability to directly interact with the loop region of let-7 hairpins. Lin28/Lin28B accelerates the turnover of let-7 precursors and prevents both Drosha- and Dicer-mediated let-7 processing (Heo et al., 2008; Newman et al., 2008; Piskounova et al., 2008; Rybak et al., 2008). Conversely, let-7 targets Lin28 and downregulates its expression (Yang et al., 2010). These results suggest that a regulatory feedback loop exists between let-7 and Lin28 (Yang et al., 2010) that is crucial in modulating the self-renewal and differentiation of breast cancer stem cells.

In this study, we have demonstrated that the activation of the Wnt–β-catenin pathway suppresses mature let-7 miRNAs but not the primary transcripts, which suggests a post-transcriptional regulation of let-7 expression. We further identified Lin28 as a novel direct downstream target of the Wnt–β-catenin pathway. Accordingly, loss of function of Lin28 impairs Wnt–β-catenin-pathway-mediated let-7 inhibition and breast cancer stem cell expansion. Moreover, enforced expression of let-7 blocks Wnt–β-catenin-pathway-stimulated breast cancer stem cell expansion. Thus, the Wnt–β-catenin pathway, Lin28 and let-7 miRNAs, three of the most crucial stem cell regulators, connect in one signal cascade.

Studies have shown that miRNAs play pivotal roles in controlling the maintenance and function of stem cells. We sought to identify the downstream target miRNAs of the Wnt–β-catenin pathway by screening the ZR-75-30 breast cancer cell line using a miRNA microarray system. Specifically, miRNA expression profiles were examined in high-Wnt and low-Wnt states with stable overexpression of β-catenin and GFP. All miRNAs showing a 1.5-fold or greater upregulation or downregulation in the high-Wnt state were chosen for further analysis (supplementary material Table S1). Of these miRNAs, we found that seven members of the let-7 miRNA family were downregulated by ∼2.5-fold in β-catenin stable overexpression cells, and quantitative real-time PCR (qPCR) results confirmed the changes (Fig. 1A). To verify the expression changes in more breast cancer cells and more Wnt stimulations, MDA-MB-231 and T-47D cells were also used. Consistent with the results in β-catenin activated ZR-75-30 cells, the expression levels of let-7a, let-7f and let-7g in these cells were dramatically decreased when the cells were treated with LiCl, a GSK3β inhibitor, as well as by stably overexpressing of HA–Wnt1 (Fig. 1B; supplementary material Fig. S1A). Conversely, knockdown of β-catenin expression resulted in the induction of let-7a and let-7f in both ZR-75-30 and MDA-MB-231 cells (Fig. 1C; supplementary material Fig. S1B). These results suggest that let-7 miRNAs are downstream target miRNAs of the Wnt–β-catenin pathway in breast cancer cells.

To investigate the mechanism through which the Wnt–β-catenin pathway regulates let-7 miRNAs, we used qPCR analysis to examine further the abundance of their primary transcripts in ZR-75-30 cells. Unexpectedly, we found that the expression of the let-7a/let-7f/let-7d and let-7g primary transcripts was not repressed in β-catenin stable overexpression cells (Fig. 1D). Consistently, these primary transcripts were also not induced in β-catenin knockdown cells (Fig. 1E). These findings indicate that the Wnt–β-catenin pathway utilises alternative mechanisms to downregulate multiple let-7 family members through a post-transcriptional pathway.

Lin28 and Lin28B RNA binding proteins have been demonstrated to bind to the stem loops of let-7 miRNAs and to inhibit their biogenesis by blocking Drosha- and Dicer-mediated cleavage and accelerating the turnover of let-7 precursors (Heo et al., 2008; Newman et al., 2008; Piskounova et al., 2008; Rybak et al., 2008). Therefore, we speculated that the Wnt–β-catenin pathway may inhibit let-7 family members through the transactivation of Lin28 or Lin28B. To verify this hypothesis, we used semi-quantitative RT-PCR to examine Lin28 and Lin28B expression in different breast cancer cell lines upon LiCl stimulation. Lin28B expression was undetected in three of four cell lines tested, and no changes were found in the ZR-75-30 cell line, but Lin28 expression was dramatically induced in all cell lines treated with LiCl (Fig. 2A, left). Similar results were achieved by qPCR analysis (Fig. 2A, right). Conversely, knockdown of β-catenin expression resulted in greatly reduced Lin28 expression in ZR-75-30 cells (Fig. 2B). We previously demonstrated that Pygo2, a newly found β-catenin interaction protein, is overexpressed in breast cancer stem cells and augments Wnt–β-catenin pathway activity (Chen et al., 2010). We examined whether Pygo2 knockdown also affects Lin28 expression. As shown in Fig. 2C, depletion of Pygo2 in ZR-75-30 cells caused a remarkable decrease in Lin28 expression. Next, we investigated whether the let-7 family members responded to Lin28 in the breast cancer cells in our experiments. As expected, when Lin28 mRNA and protein expression were effectively knocked down (supplementary material Fig. S2A), the levels of mature let-7a and let-7f transcripts were greatly increased (supplementary material Fig. S2B), whereas overexpression of Lin28 resulted in reduced mature let-7a and let-7f levels (supplementary material Fig. S2C). In contrast, the primary transcript levels of let-7a/let-7f/let-7d and let-7g were not changed in either experiment (supplementary material Fig. S2D,E). To further verify that both β-catenin and Lin28 could inhibit the biological function of let-7s as microRNAs, we performed a luciferase assay using a let-7 sensor containing a constitutively expressed luciferase reporter bearing two copies of sequences complementary to let-7a in the downstream 3′ UTR. As shown in Fig. 2D, endogenous let-7 miRNAs in the MDA-MB-231 cells greatly inhibited let-7 sensor activity, indicating that these miRNAs indeed target the 3′ UTR of luciferase and inhibit its expression. Enforced expression of β-catenin and Lin28 both remarkably rescued the sensor activity almost to the control vector levels, which suggests that the biological function of let-7 is inhibited. Finally, we treated the MDA-MB-231, ZR-75-30 and T-47D cells with recombinant Wnt1 to see weather the Wnt ligand also triggers the same activity. As expected, β-catenin expression was upregulated in all cell lines treated with recombinant Wnt1. Moreover, Lin28 was upregulated whereas let-7a/let-7f was downregulated in these cells (Fig. 2E). The regulatory consistency of let-7 suggests that the Wnt–β-catenin pathway and Lin28 may share the same signalling cascade and supports the hypothesis that the Wnt–β-catenin pathway inhibits let-7 miRNAs by activating Lin28 expression in breast cancer cells.

To determine whether Lin28 is a direct Wnt–β-catenin pathway target gene, we examined the genomic sequence within a 10-kb window centred on the transcriptional start site (TSS) of Lin28. The genomic sequence extending from 1.0 kb upstream to 114 bp (ATG site) downstream of the TSS contains seven consensus LEF/TCF-binding sites of CTTTG or GAAAC, with four sites showing conservation between human and mouse (Fig. 3A, conserved sites indicated as black dots). This fragment was cloned into a PGL3-basic promoter-less luciferase reporter cassette. The fragment was sufficient to drive the luciferase reporter activity in a dose-dependent manner when stimulated with LiCl, β-catenin or Wnt-1 in 293T cells (Fig. 3B). Equivalent Wnt-dependent reporter activity was conferred by a truncated fragment without the last three sites, which indicates that the conserved sites are responsible for the luciferase activity. Although mutation of the site 2 or 3 binding sequence had no effect, deletion or mutation of site 1 or 4 resulted in significantly compromised luciferase activity induced with LiCl, albeit the decrease for the site 1 mutation was minor (Fig. 3C). Similar results were achieved in MDA-MB-231 breast cancer cells when Wnt–β-catenin activity was stimulated with recombinant Wnt1 (Fig. 3D). In a chromatin immunoprecipitation (ChIP) assay, both LEF1 and β-catenin were found to occupy the endogenous Lin28 promoter at site 4 but not at site 1 in both MDA-MB-231 and T-47D cells (Fig. 3E), suggesting that only site 4 functions as a true endogenous LEF1/β-catenin-binding site on the Lin28 promoter. Taken together, our findings establish that the Wnt–β-catenin pathway activates Lin28 expression by direct binding to the Lin28 promoter, which provides compelling evidence that Lin28 is a direct target gene of the Wnt–β-catenin pathway.

To examine whether Lin28 is necessary for the Wnt–β-catenin-pathway-mediated repression of let-7 family members, we used two different shRNAs to inhibit Lin28 expression in MDA-MB-231 cells in the high Wnt state. Both of the shRNAs were observed to knock down the Lin28 expression significantly to levels comparable to those of negative control cells (Fig. 4A, left). Knockdown of Lin28 completely reversed the β-catenin-mediated repression of mature let-7a and let-7f, suggesting the essential role of Lin28 in this process (Fig. 4A, right). Conversely, overexpression of Lin28 almost completely compromised the β-catenin knockdown-induced let-7a and let-7f expression (Fig. 4B). As HRAS and HMGA2 are known let-7 targets, we examined the relationship between these proteins and the Wnt–β-catenin pathway. The results from different experiments all indicated that the Wnt–β-catenin pathway enhances HRAS and HMGA2 protein expression and that this activity depends on Lin28 activation and let-7 repression (Fig. 4A–C).

The Wnt–β-catenin pathway is known to enhance the proliferation of many cancer types including breast cancer (MacDonald et al., 2009; Matsuda et al., 2009). As expected, enforced expression of β-catenin resulted in a significantly increased growth rate in MDA-MB-231 cells. The simultaneous knockdown of Lin28 was found to almost completely compromise β-catenin activity (Fig. 4D). When plated at a clonal density, a significant decrease in the number and size of colonies was observed for Lin28-depleted cells (Fig. 4E). These results suggest that the transactivation of Lin28 is essential for the Wnt–β-catenin-pathway-mediated let-7 inhibition and cell proliferation. Moreover, overexpression of let-7a in Wnt-activated MDA-MB-231 cells was also observed to completely inhibit β-catenin-activated cell growth and colony formation (Fig. 4C–E), which underscores the importance of let-7 miRNAs as Wnt–β-catenin pathway downstream targets in regulating cell proliferation. Collectively, the above results verify the hypothesis that the repression of let-7 through the activation of Lin28 is necessary for Wnt–β-catenin-pathway-induced cell proliferation.

Although previous studies have demonstrated that the Wnt–β-catenin pathway, Lin28 and let-7 miRNAs all play central roles in stem cell self-renewal and differentiation, their possible involvement in the same signalling cascade that regulates cancer stem or cancer-initiating cells has not been addressed. Such cells in breast cancer can be enriched from established cancer cell lines using either mammosphere cultures or FACS sorting for a CD44⁺ CD24⁻ population (Al-Hajj et al., 2003; Ponti et al., 2005). Using western blotting and qPCR, we detected higher levels of β-catenin and Lin28 proteins but lower levels of let-7 members in mammospheres of the MDA-MB-231, MCF7 and T-47D cell lines in comparison to their corresponding adherent cultures (Fig. 5A). Knockdown of β-catenin expression in MDA-MB-231 cells resulted in decreased Lin28 expression and increased let-7a transcript levels in the mammospheres (Fig. 5B). Consequently, β-catenin-depleted cells formed fewer, smaller mammospheres (Fig. 5C). The expression trend of let-7 and Lin28 regulated by β-catenin in mammospheres suggests a functional relevance of these factors in breast cancer stem cells. Given our result that the Wnt–β-catenin pathway/Lin28/let-7 cascade promotes breast cancer cell proliferation, we reasoned that this phenomenon is caused by stem cell expansion. To verify this hypothesis, we infected MDA-MB-231 cells with β-catenin-expressing lentiviruses and examined mammosphere formation. The enforced expression of β-catenin in MDA-MB-231 cells resulted in significantly more and larger mammospheres. Simultaneous knockdown of Lin28 or overexpression of let-7a was found to compromise the β-catenin-enhanced mammosphere formation almost completely (Fig. 5D). We also examined whether the Wnt–β-catenin pathway/Lin28/let-7 cascade affects the size of the CD44⁺ CD24⁻ population in breast cancer cells. T-47D cells were used for this analysis because ∼90% of MDA-MB-231 cells are of this type, making it difficult to score any potential increase. As shown in Fig. 5E, the enforced overexpression of β-catenin in T-47D cells by lentiviral infection yielded a greatly increased CD44⁺ CD24⁻ population, but the overexpressed β-catenin was not able to induce an increase in the CD44⁺ CD24⁻ population in the Lin28-depleted or let-7a-overexpressed cells. Collectively, these results demonstrate that let-7 repression through the activation of Lin28 is necessary for Wnt–β-catenin-pathway-induced expansion of the breast cancer stem cell population.

To investigate whether the Wnt–β-catenin pathway is associated with Lin28 expression in breast cancer patients, tissue microarrays from 82 patients with breast cancer who had undergone mammary gland resection were examined by immunostaining with β-catenin and Lin28 antibodies. The overall expression levels of both the β-catenin and Lin28 proteins were significantly higher in breast cancers than in adjacent tissues as indicated by the representative samples in the tissue microarrays (Fig. 6A, left) and the summary of all the samples (supplementary material Table S2). Moreover, correlation analyses revealed strong correlations between β-catenin expression and Lin28 levels (supplementary material Table S2; Fig. 6A, right). To analyse further the relationship between the Wnt–β-catenin pathway, Lin28 and let-7, 16 pairs of fresh tumour samples along with their adjacent tissues were used to detect the expression of β-catenin, Lin28 and let-7a by western blotting and qPCR. As shown in supplementary material Fig. S3, the β-catenin and Lin28 expression levels in the cancers were higher than in the adjacent tissues, which is consistent with the tissue microarray results. Conversely, the let-7a expression levels in the cancers were dramatically lower than those of the adjacent tissues. Moreover, Lin28 expression showed a positive relationship, whereas let-7a expression was found to be inversely related to β-catenin. It has been well documented for the mouse mammary tumor virus (MMTV)-wnt-1 mouse that the stem cell-enriched Lin⁻CD29^hi CD24⁺ population in premalignant mammary tissue is greatly expanded (Shackleton et al., 2006), indicating the critical role for the Wnt–β-catenin pathway in the expansion of breast cancer stem cells. We therefore measured the expression levels of β-catenin, Lin28 and let-7a in this model using immunostaining, western blotting and qPCR. As expected, β-catenin expression was greatly increased in premalignant mammary tissues compared with tissues from wild-type littermates. Lin28 and let-7a showed coordinated increased and decreased expression levels. Moreover, the let-7 targets HRAS and HMGA2 also showed increased protein expression levels in premalignant mammary tissues of the MMTV-wnt-1 mouse (Fig. 6B). To confirm the regulating cascade in vivo in a second setting, we examined an additional model with a high Wnt state. The model used, the Apc^min/+ mouse, is a mouse intestine adenoma model bearing a truncated Apc allele. Results similar to those of the MMTV-wnt-1 mice were achieved and further confirm the relationship between the Wnt–β-catenin pathway, Lin28 and let-7a (Fig. 6C). Finally, the in vivo functional relevance of the Wnt–β-catenin pathway/Lin28/let-7 cascade was examined. We intraductally injected cholesterol-, OMe-conjugated Lin28 siRNA into the premalignant mammary tissue of MMTV-wnt-1 mice through and examined whether Lin28 knockdown could rescue the let-7 repression and the stem cell phenotype driven by the Wnt–β-catenin pathway. As illustrated in Fig. 6D, premalignant mammary tissues exhibited decreased let-7a and let-7f levels and an expanded Lin⁻ CD29^hi CD24⁺ population when compared with tissues from wild-type littermates; however, knockdown of Lin28 dramatically recovered the let-7 expression and the size of the stem cell population when compared with the wild-type littermates. Similarly, delivery of a cholesterol-conjugated let-7a agomir into the premalignant mammary tissue of MMTV-wnt-1 mice was also observed to dramatically rescue the stem cell phenotype (Fig. 6E). Taken together, these results provide solid evidence of the existence of a Wnt–β-catenin pathway/Lin28/let-7 cascade in vivo.

Aberrant activation of the Wnt–β-catenin pathway has been found in a wide range of cancers, especially in cancers derived from intestine, skin, mammary gland and haematopoietic cells. Moreover, the Wnt–β-catenin pathway may preferentially influence stem/progenitor cell expansion in these cancers (Wend et al., 2010). Although many downstream target genes for both normal development and tumorigenesis have been identified in different cellular contexts, the genes that mediate the Wnt–β-catenin pathway activity in maintaining the stem cell properties are still not very clear. c-Myc, a reprogramming factor and a well-known Wnt target, was recently demonstrated to be an important stem cell regulator in normal and cancerous cells (Kim et al., 2010; Smith et al., 2011). However, no study has established a convincing relationship between the Wnt–β-catenin pathway and c-Myc in stem cells. In contrast, recent studies on both iPS and ESCs suggest that the role of the Wnt–β-catenin pathway in reprogramming and maintaining stem cells could be independent of c-Myc induction (Marson et al., 2008; Ying et al., 2008). In breast cancer, current studies indicate that Hedgehog, Notch and the Wnt–β-catenin pathway, as well as Bmi1 (Liu et al., 2006) and Lin28 (Yang et al., 2010) transcription factors are major regulators that affect cancer stem cell properties. However, the downstream molecular events of the Wnt–β-catenin pathway responsible for this remain largely unknown. In this study, we identified Lin28 as a novel downstream target of the Wnt–β-catenin pathway and found that Lin28 is necessary for Wnt–β-catenin-pathway-mediated breast cancer stem cell expansion, which highlights a molecular mechanism for the Wnt–β-catenin pathway in regulating the stem cell properties of cancer cells. More recently, it was demonstrated that the Wnt–β-catenin pathway acts as a new reprogramming factor to promote somatic cells to pluripotency, although the underlying molecular rationale is unclear (Ying et al., 2008). Thus, our studies have shed light on the connection between the Wnt–β-catenin pathway and the reprogramming process.

In regulating stem cell functions, miRNAs are proposed to be important factors because expression levels of certain miRNAs in embryonic stem cells are different from those of differentiated embryoid bodies (Suh et al., 2004). Consistently, tumours analysed by miRNA profiling have shown significantly different miRNA profiles of stem cells compared with differentiated cells from the same sample (Papagiannakopoulos and Kosik, 2008; Sun et al., 2010). In breast cancer, the inhibition of let-7, miR-30 and miR-200 and the induction of miR-181 and miR-495 are required for the maintenance of stem cell properties (Hwang-Verslues et al., 2011; Iliopoulos et al., 2010; Shimono et al., 2009; Wang et al., 2011; Yu et al., 2010; Yu et al., 2007), suggesting miRNAs are important in regulating breast cancer stem cells. As the Wnt–β-catenin pathway is a dominant regulator of breast cancer stem cells, we speculate that this pathway may affect the stem cell properties through modulations in miRNA expression. The relationship between the Wnt–β-catenin pathway and miRNAs has been verified in different cellular contexts, although it remains obscure in stem cells, especially cancer stem cells. In contrast to the rare downstream target miRNAs, many upstream modulated miRNAs of the Wnt–β-catenin pathway have been identified, of which miR-135a/b (Nagel et al., 2008) and miR-315 (Silver et al., 2007) upregulate, whereas miR-200a (Korpal et al., 2008; Saydam et al., 2009), miR-21 (Hashimi et al., 2009), miR-203 (Thatcher et al., 2008) and miR-8 (Kennell et al., 2008) downregulate β-catenin transcriptional activity through targeting the Wnt–β-catenin pathway components. Among the few Wnt-regulated miRNAs, the Wnt–β-catenin pathway regulates miR-15/16 maturation, rather than its transcription. However, the underlying mechanism is unknown (Martello et al., 2007). Another study showed that miR-122a expression is downregulated in APC-driven gastrointestinal cancers. The mechanisms by which the Wnt–β-catenin pathway regulates miR-122a and 122a inhibition are unknown (Wang et al., 2009). Additionally, miR-375 is downregulated by β-catenin. However, the function of miR-375 and the transcriptional mechanism through which miR-375 is regulated by the Wnt–β-catenin pathway are not clear and require further investigation (Ladeiro et al., 2008). In an attempt to identify novel downstream target miRNAs of the Wnt–β-catenin pathway, we performed screening experiments in breast cancer cells. In our screening, 15 miRNAs were upregulated, whereas 48 miRNAs were downregulated upon stimulation of the Wnt–β-catenin pathway, which is consistent with the notion that an overall downregulation of miRNAs occurs in many cancers, compared with their normal tissue counterparts (Hammond, 2006). It is noted that two known downregulated miRNAs, miR-15 and miR-375, are present in our list, suggestive of a successful screening strategy for our experiment. The obvious downregulation of let-7 members is remarkable, because let-7 miRNAs have been demonstrated to play an important role in breast cancer stem cells. Both qPCR and reporter assays of several cell lines verified that let-7 miRNAs are novel downstream target miRNAs of the Wnt–β-catenin pathway. Furthermore, we found that the Wnt–β-catenin pathway regulates let-7 miRNAs through a post-transcriptional pathway by directly transactivating Lin28 expression. To the best of our knowledge, this report is the first to find that the Wnt–β-catenin pathway inhibits let-7 miRNAs through Lin28 transactivation.

Recent advances in the characterisation of stem cells in mammary epithelium and breast cancer cells have opened the door to understanding the signalling and transcriptional programming underlying both the development of normal mammary stem cells and the proliferation/differentiation of their malignant counterparts (Al-Hajj et al., 2003; Ponti et al., 2005; Shackleton et al., 2006; Stingl et al., 2006). The Wnt–β-catenin pathway, Lin28 and let-7 are defined as crucial mammary/breast cancer stem cell regulators. However, the relationship between them is unknown. In this study, we have functionally established a connection between the Wnt–β-catenin pathway and Lin28/let-7 that is essential for Wnt–β-catenin-pathway-mediated breast cancer stem cell expansion. This model is strongly supported by: (1) the converging effects of the Wnt–β-catenin pathway, Lin28 and let-7 on breast cancer stem cell properties; (2) the dependence of the Wnt–β-catenin pathway in let-7 inhibition and breast cancer stem cell expansion on Lin28 activation; (3) the dependence of the Wnt–β-catenin pathway in breast cancer stem cell expansion on let-7 inhibition; and (4) the expression pattern and functional relevance of β-catenin, Lin28 and let-7a in cancer samples and tumour models. As such, our study presents the first mechanistic characterisation of the Wnt–β-catenin pathway function in breast cancer stem cells and points to Lin28 and/or let-7 as important downstream targets in these cells in vitro and in vivo. As the Wnt–β-catenin pathway is aberrantly activated in many cancer types, future work should address whether the Wnt–β-catenin pathway/Lin28/let-7 cascade is also involved in the regulation of other cancers. Moreover, re-expression of let-7 miRNAs might be a new therapeutic option in treating cancers with an aberrantly activated Wnt–β-catenin pathway.

HEK 293T (human embryonic kidney 293T) and human breast cancer cell lines ZR-75-30 (human breast ductal carcinoma), MDA-MB-231 (human breast adenocarcinoma), MCF7 (human breast adenocarcinoma) and T-47D (human breast ductal carcinoma) were obtained from the American Type Culture Collection and grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal calf serum (Gibco), penicillin and streptomycin. To induce the Wnt–β-catenin pathway, cells were grown in the presence of 20 mM LiCl (Sigma) for 24–36 hours or 100 ng/ml Wnt1 protein (Abcam, cat. no. ab84080) for 48 hours. All cell lines were grown at 37°C with 5% carbon dioxide.

The following antibodies were used: anti-Lin28 (Abcam, cat. no. ab46020), anti-β-catenin (Cell Signaling Technology, cat. no. 9562), anti-β-actin (Sigma-Aldrich, cat. no. A1978), anti-HRAS (Proteintech Group, cat. no. 18295-1-AP), anti-HMGA2 (R&D Systems, cat. no. AF3184), anti-LEF-1 (Santa Cruz Biotechnology, cat. no. sc-8591), anti-HA (Sigma-Aldrich, cat. no. H9658). The human cell surface markers utilised were CD24-PE (BD Biosciences) and CD44-PE-Cy5 (eBioscience). Cholesterol-conjugated mmu-let-7a agomir and negative control agomir as well as cholesterol-, 2′-O-methyl (OMe)-conjugated mouse lin28 siRNA and negative control siRNA for in vivo delivery were obtained from Ribobio Co. (Guangzhou, China).

Total RNAs were isolated using TRIzol reagent (Invitrogen). cDNAs were prepared from these RNAs using a ReverTra Ace qPCR RT kit from Toyobo. Quantitative PCR was performed using a Rotor gene 6000 Sequence Detection System with the Thunderbird SYBR qPCR Mix (Toyobo). Mature let-7a, let-7f and let-7g miRNAs were quantified using a predesigned miRCURY LNA™ Universal RT microRNA PCR Assay (Exiqon). A eukaryotic 18S rRNA or U6 snRNA endogenous control was used as an internal standard, and the results were calculated using the ΔΔCT (where CT is threshold cycle) method. Primer sequences are provided in supplementary material Table S3.

Stable transfected cells were plated in 96-well dishes at 1000 cells/well. Cell growth rates were monitored using a Cell Counting Kit-8 (CCK-8, Dojindo Molecular Technologies) and an MTT Assay Kit (Promega).

Stable transfected cells were plated in six-well dishes at 500 cells/well. After 10 days, the cultures were washed twice with PBS, incubated with methanol for 20 minutes, stained with Crystal Violet for 30 minutes, and washed with tap water. The colonies were counted under a low magnification microscope and a group of more than 10 cells was defined as a colony.

Confluent cells were trypsinised into single-cell suspensions. These culture cell suspensions and the tissue digested cell suspensions were washed with fluorescence-activated cell sorting (FACS) buffer (2% foetal bovine serum in PBS), counted, and stained with fluorophore-conjugated antibodies. A total of 10⁶ cells in 100 µl of FACS buffer were incubated with antibodies for 30 minutes at 4°C. Unbound antibodies were washed off, and the cells were sorted using a Beckman EPICS XL instrument.

Cells (10⁴ cells/ml) were cultured in ultra-low attachment plates in serum-free DMEM/F12 (Invitrogen) supplemented with B-27 (1∶50; Invitrogen), 20 ng/ml epidermal growth factor (EGF; BD Biosciences), 20 ng/ml basic fibroblast growth factor (bFGF; BD Biosciences) and 4 µg/ml insulin (Sigma), and fed every 3 days.

Total RNAs were harvested using TRIzol (Invitrogen) and an RNeasy mini kit (Qiagen) according to the manufacturer's instructions. The samples were labelled using a miRCURY™ Hy3™/Hy5™ Power labelling kit and hybridised on a miRCURY™ LNA Array (v.14.0). Scanning was performed on an Axon GenePix 4000B microarray scanner. GenePix pro V6.0 was used to read the raw intensity of the images. Signals that were more than twice the background were removed, and datasets were median-centred before calculating fold change values.

A chromatin immunoprecipitation (ChIP) assay was performed following the Upstate Biotechnology protocol. Briefly, T-47D cells were fixed with 1% paraformaldehyde at room temperature for 10 minutes, washed, and lysed with SDS lysis buffer (50 mM Tris-HCl, 1% SDS, 10 mM EDTA, and protease inhibitors). The lysates were sonicated to reduce DNA lengths to between 300 and 600 bp. The soluble fraction was diluted, precleared with salmon sperm DNA–protein-A–agarose, divided into two tubes and incubated with specific antibodies or control IgG. The immune complexes were then precipitated with protein A/G beads and eluted with elution buffer (0.1 M NaHCO₃ and 1% SDS). The eluted samples were reverse cross-linked and treated with proteinase K. DNA was purified by phenol–chloroform extraction and dissolved in distilled water. Real-time PCR quantification of the ChIP samples was performed in triplicate using the Thunderbird SYBR qPCR Mix (Toyobo). Primer sequences for the Lin28 promoter are provided in supplementary material Table S3.

Oligonucleotides encoding let-7a1 pre-miRNA or shRNA targeting Lin28, β-catenin, Pygo2 or lacZ (control) were synthesised by Invitrogen and cloned under the control of the U6 promoter in the lentiviral vector lentilox pLL3.7. For overexpression of Lin28, β-catenin S33Y, HA-Wnt1 or GFP (control), the cDNAs were cloned under the control of the EF1α promoter in the lentiviral vector pLV-CS2.0. The generation of lentivirus vectors was performed by co-transfecting pLL3.7 or pLV-CS2.0 carrying the expression cassette with helper plasmids pVSV-G and pHR into 293T cells using Lipofectamine 2000 (Invitrogen). The viral supernatant was collected 48 hours after transfection, and viral titres were determined by transducing HeLa cells at serial dilutions and analysing the GFP expression using flow cytometry. Cells at 50–70% confluency were infected with viral supernatants containing 10 µg/ml Polybrene for 24 hours, after which fresh medium was added to the infected cells, which were later selected with puromycin or G418. The oligonucleotide sequences are provided in supplementary material Table S4.

To evaluate the miRNA function of let-7, a pMIR-REPORTTM luciferase reporter vector (Ambion) with two copies of sequences complementary to let-7a cloned into its 3′ UTR (let-7 sensor) was used. The reporter vector plasmid was transfected into MDA-MB-231 using Lipofectamine 2000 according to the manufacturer's instructions. To correct for the transfection efficiency, a luciferase reporter vector without the let-7 target was transfected in parallel. The luciferase activity was assayed using a luciferase assay kit (Promega). let-7 miRNA function was expressed as a percentage reduction in the luciferase activity of cells transfected with the reporter vector containing the let-7 target sequences compared with cells transfected with the vector without the let-7 target.

To generate the luciferase reporter vectors, the Lin28 promoter fragment was amplified from human genomic DNA and cloned into the KpnI–HindIII site of the firefly luciferase plasmid pGL3-basic-IRES. For reporter assays, HEK 293T cells in 24-well plates were transfected at 50–60% confluency using the calcium phosphate method. Both Lin28 promoter constructs (50 ng) and cytomegalovirus (CMV)–β-galactosidase (25 ng) reporter plasmids were co-transfected with the β-catenin S33Y or Wnt-1 expression plasmids or in the presence of LiCl (10 to 20 mM) for 24 hours. The total amount of plasmid DNA transfected was made equivalent by adding empty vectors. Cells were harvested after 24 hours and processed for luciferase and β-galactosidase assays, and the data were normalised to the β-galactosidase levels.

Fourteen-week-old virgin female MMTV-wnt-1 (FVB) and wild-type (FVB) mice were anesthetized. The keratin plugs were removed from the surface of the nipple, revealing the duct orifice. The mammary ducts were cannulated using a 1.0-cm, 34-gauge, blunt-ended needle attached to a 1-ml tuberculin syringe. Next, 5 nmol of cholesterol-conjugated microRNA agomir or cholesterol-, OMe-conjugated siRNA in 25 µl PBS was infused into the mammary gland once every 3 days for a total of 12 days (four times). Twenty days after the first injection, the mammary glands of the injected site were dissected from the female mice. The animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Xiamen University College of Medicine.

After mechanical dissociation with surgical scissors, the tissue was placed in culture medium 1640 supplemented with 5% bovine calf serum and containing 300 U/ml collagenase (Sigma) and 100 U/ml hyaluronidase (Sigma) and digested for 3 hours at 37°C. The resultant organoid suspension was vortexed and red blood cells lysed in NH₄Cl. A single-cell suspension was obtained by sequential dissociation of the fragments by gentle pipetting for 1–2 minutes in 0.25% trypsin and then for 2 minutes in 5 mg/ml dispase II plus 0.1 mg/ml DNase I (Sigma-Aldrich) followed by filtration through a 40-µm mesh. Cells were resuspended to 10⁶ cells in 100 µl of cold FACS buffer. Staining was performed using the following antibodies and reagents on ice for 30 minutes as follows: biotin-conjugated Lin (CD45, CD31, TER119, 50 µl/ml; EasySep, cat. no. 19757), PE-conjugated CD24 (10 µl/100 µl; StemCell Technologies, cat. no. 10814), FITC-conjugated CD29 (10 µl/100 µl; BioLegend, cat. no. 102205), PE-conjugated IgG isotype control (10 µl/100 µl), FITC-conjugated IgG isotype control (10 µl/100 µl). Unbound antibodies were washed off, propidium iodide (2 µg/ml final concentration) was added and the mixture incubated on ice for 20 minutes. The cells were sorted using a Beckman EPICS XL instrument. Forward scatter (FS) and side scatter (SS) were used to select most of the single cell populations. The PI-positive, non-viable cells were excluded from the final analysis. Lin⁺ cells were gated out of the final analysis. Single stained samples were used as compensation controls. Cells incubated in non-specific IgG isotype control (conjugated to PE or FITC) were used to set the gates that define Lin⁻, CD24⁺ or CD29^hi cells, respectively.

Tissue microarrays from 82 patients with breast cancer were purchased from Shanghai Outdo Biotech Co., Ltd. The human tissue preparation and analysis were approved by the Institutional Review Board of The First Affiliated Hospital of Xiamen University. The tissues were fixed in 10% formalin, embedded in paraffin, and sectioned. After dewaxing and rehydration, the sections were pretreated with peroxidase blocking buffer (Maxim, Fuzhou, China) for 20 minutes at room temperature. Antigen retrieval was performed by boiling in Tris-EDTA (pH 9.0) for 20 minutes. After treatment with a blocking buffer (5% normal goat serum in PBS) for 1 hour at room temperature, sections were incubated with anti-Lin28 antibody (1∶300, overnight at 4°C, Abcam) and anti-β-catenin antibody (1∶200, overnight at 4°C; Cell Signaling Technology) in the blocking buffer. Secondary antibody reagents were obtained from the SABC kit (Boster Biological Technology, Wuhan, China) or DAB kit (Maxim Biological Technology, Fuzhou, China).

The pictures were taken using a computerized imaging system (Leica Microsystems, Imaging Solutions Ltd, Cambridge, UK). Under high-power magnification (200×), photographs of three representative fields were captured using Leica QWin Plus v3 software; identical settings were used for each photograph. For the reading of each antibody staining, a uniform setting for all of the slides was applied. The mean integrated optical density (IOD) was used to represent the relative expression of indicated protein. IOD is a computer-assisted method which is commonly used to quantify both the area and the intensity of the positive staining in immunohistochemistry. We calculated the IOD of each photograph acquired from the tissue microarray sections by using Image-Pro Plus v6.2 software (Media Cybernetics Inc., Bethesda, MD). The typical positive staining area was located, with the help of a pathologist, in the segmentation panel, standard optical density was chosen in the intense calibration panel, and background was subtracted in the optical density calibration panel, which was chosen to exclude all the areas without epithelial or tumour tissues. We also subtracted the background in serial sections by using a non-specific IgG instead of primary antibody. The ‘mean IOD’ (IOD/tissue area) represents the expression level of indicated protein.

For the tissue microarrays, Fisher's exact test was used to compare qualitative variables; quantitative variables were analysed using t-tests and Spearman's rank correlation tests. The data are presented as the means ± s.d. All statistical tests were two-sided Student's t-tests, and a P<0.05 was considered to be statistically significant.