Background
Cancer stem cell theory proposes that cancers may arise from malignant transformation of normal stem/progenitor cells. Alternatively, cellular plasticity and/or the tumor microenvironment may permit mature/differentiated cells to acquire a stem/progenitor phenotype [
1‐
6]. Tumorigenic stem/progenitor cells have been documented in hematologic malignancies as well as in solid tumors, although correct terminology for these cells (cancer stem cells versus tumor initiating cells) is still a matter of debate [
7‐
9]. Several studies implicate a subset of human breast cancer cells with an enhanced ability to form tumors in immunocompromised mice [
10,
11]. This subpopulation of cells also demonstrated the capacity for self-renewal and generation of heterogeneous progeny. At present, two distinct cell types have been described as CSCs for breast cancer. Cancer cells that display the cell surface marker profile of CD44
+/CD24
-/Lineage
- were the first described tumorigenic progenitor cell types for breast cancer [
10]. NOD/SCID mice implanted with as few as 200 CD44+/CD24- cells form tumors. In addition, disseminated cancer cells in bone marrow with CD44+/CD24- phenotype have been identified in patients, although the prognostic relevance of this is as yet unclear [
12‐
14]. Signaling pathways implicated in self-renewal and survival of normal organ-specific stem cells and embryonic stem cells, such as Hedgehog, Notch and Wnt/β-catenin, may be involved in maintaining "stemness" of CD44+/CD24-/lineage- cells [
15‐
18]. The gene expression pattern of CD44+/CD24- CSCs is more similar to normal CD44+/CD24- breast epithelial cells than to CD44-/CD24+ cells isolated from tumors [
19]. Recent studies have demonstrated enrichment of CSC gene expression signature in breast cancers that are classified as Claudin-low subtype [
20]. We demonstrated that breast cancer cells with CD44+/CD24- phenotype express elevated levels of invasion-associated genes and are invasive but this phenotype is not a requisite for homing and growth at sites of metastasis [
21]. In subsequent studies, normal and cancerous breast epithelial cells expressing higher levels of aldehyde dehydrogenase 1 (ALDH1) were described as normal and tumorigenic stem/progenitor cells [
22]. Functional assays revealed ALDEFLUOR-positive cells (aldefluor staining provides indirect estimation of all ALDHs in cells) to be highly tumorigenic relative to ALDEFLUOR-negative cells. Moreover, the most tumorigenic phenotype identified was ALDEFLUOR+/CD44+/CD24- cells [
22]. Additional refinement of the breast cancer stem cell phenotype has been described recently [
23].
During our analysis of breast cancer cell lines for subpopulations with the CD44+/CD24- phenotype, we observed that almost all cell lines with a CD44+/CD24- subpopulation were basal breast cancer cells that had undergone epithelial to mesenchymal transition (EMT) [
21]. Others have also reported enrichment of cells with the CD44+/CD24- phenotype in basal-like breast tumors [
14]. EMT is a developmental process during which epithelial cells acquire a fibroblastoid and invasive phenotype, down-regulate epithelial-specific proteins, and induce various mesenchymal proteins [
24]. There are specific changes in the gene expression profile during EMT. These include expression of vimentin and loss of E-cadherin expression; the change in expression of both of these markers has been associated with poor prognosis in breast cancer [
25‐
27]. Activation of oncogenic and receptor tyrosine kinase pathways such as Ras and Src, or signaling through transforming growth factor beta (TGFβ), hepatocyte growth factor (HGF) and platelet derived growth factor (PDGF) can trigger EMT [
24,
28]. These signaling pathways induce the expression of the SNAIL family of transcription repressors, which reduce E-cadherin expression [
29]. SNAIL family members involved in initiation and/or maintenance of EMT include SNAIL-1 [
30,
31], Snai2/SLUG [
32,
33], E12/E47 [
34], ZEB-1/ZFHX1A [
35], ZEB-2/ZFXH1B/Smad-Interacting Protein (SIP1) and TWIST [
36].
Recent studies have shown that some members of the SNAIL family confer an EMT phenotype to breast epithelial cells, which correlates with cells changing phenotype from CD44-/CD24+ to CD44+/CD24- [
37,
38]. However, it is not known whether all genes that induce EMT confer a CD44+/CD24- phenotype to breast epithelial cells or if all breast epithelial subtypes are equally susceptible to such EMT-mediated phenotypic change.
In this study we have utilized the basal cell phenotype MCF-10A breast epithelial cell line [
39] to study the association between CD44+/CD24- and the EMT phenotype. Although the majority of MCF-10A cells are CD44-/CD24+ or CD44+/CD24+, a fraction of these cells are CD44+/CD24-. Gene expression analysis of CD44+/CD24- cells compared to CD44-/CD24+ cells revealed increased expression of 32 EMT associated genes including SLUG, ZEB-1, ZEB-2, Hedgehog signaling associated gene Gli-2, and the metastasis-associated gene SATB-1. Transgenic overexpression studies showed that only SLUG had the capacity to alter the phenotype of CD44-/CD24+ MCF-10A cells to induce a subpopulation of CD44+/CD24- cells. However, transgenic overexpression of SLUG in the luminal type breast cancer cell line, MCF-7, generated cells with a CD44+/CD24+ phenotype, suggesting that basal cell types but not luminal cell types are susceptible to EMT associated acquisition of CD44+/CD24- phenotype. Additionally, only specific EMT associated genes induced a CD44+/CD24- phenotype in MCF-10A cells. For example, overexpression of the NF-κB subunit of p65 upregulates expression of ZEB-1 and ZEB-2 genes [
40], and this resulted an increase in the percent of CD44+/CD24+ MCF-10A cells but not the percent of CD44+/CD24- cells.
Discussion
In this report, we describe gene expression patterns in the CD44+/CD24- and CD44-/CD24+ subpopulations of MCF-10A cells. Unlike previous studies using similar subpopulations of epithelial cells from normal or cancerous breast, our study identified a large number of genes that are differentially expressed between cells with these two phenotypes. This could be partly due to MCF-10A cells being more homogenous, as well as not having to amplify RNA before microarray analysis. Based on the claudin expression pattern, CD44+/CD24- cells that were used in this study appear to more closely resemble stem cells [
20]. Several of the genes upregulated 10-fold or higher in CD44+/CD24- cells were cell surface molecules, which can potentially be used to further fractionate CD44+/CD24- cells into distinct subpopulations and assayed for stem/progenitor status. These molecules/markers include EPHA4, EPHA5, ST-2, platelet growth factor receptor-like-1, LRP-1/CD91, and Toll-like receptor 4.
Previous studies linking EMT to the CD44+/CD24- phenotype identified only a few genes that may be involved in generating CD44+/CD24- cells. These genes included TWIST1, TWIST2, SNAIL, SLUG/SNAI2, FOXC2, and ZFHX1B/ZEB-2 [
38]. TWIST2, SLUG, and ZEB-2 were also identified in our study to be elevated in CD44+/CD24- cells (Table
1). We observed enhanced expression in CD44+/CD24- cells of an additional 29 genes that have been associated with EMT. Pathway analysis suggested that expression and/or activity of 27 of these genes are linked (Figure
3A). However, not all genes involved in EMT are likely to have the ability to induce the CD44+/CD24- phenotype. We had previously demonstrated that MCF-10A cells overexpressing constitutively active p65 undergo EMT, which is accompanied by upregulation of ZEB-1 and ZEB-2 [
40]. However, MCF-10A cells overexpressing p65 displayed a CD44+/CD24+ phenotype, suggesting that ZEB-1 and ZEB-2-mediated EMT did not involve generation of CD44+/CD24- cells. Nonetheless, p65 overexpressing CD44+/CD24+ cells may have stem cell like properties similar to MCF-7-SLUG cells. In this respect, it has recently been demonstrated that SRC-mediated transformation of MCF-10A cells involves the NF-κB-Lin28B-let7-IL-6 axis and this signaling axis generates cancer cells with stem cell like properties [
55]. Thus, it is likely that cells with the CD44+/CD24- and CD44+/CD24+ phenotype have stem cell like properties at least based on the mammosphere assay. Clinical prognostic studies using CD44 and CD24 as markers may have to consider both cell types. Since MCF-10A cells that have acquired an EMT phenotype without co-expressing an oncogene such as SRC and Ras or the cytokine IL-6 are not tumorigenic [
55], we did not characterize MCF-10A-SLUG or MCF-10A-p65 cells in xenograft models for CSC behavior.
Although SLUG was thought to function in a redundant manner with SNAIL, several recent studies suggest unique functions for SLUG. For the following reasons SLUG appears to be much more relevant for generating breast cancer cells with a CSC phenotype than SNAIL: 1) SLUG is downstream of Notch1 and Jagged 1 and is essential for Notch1:Jagged 1 mediated EMT, tumor growth, and metastasis [
56]; 2) Jagged 1 expression in primary breast cancer correlates with SLUG but not SNAIL expression [
56]; 3) SLUG expression correlates with poor breast cancer prognosis [
57]; 4) SLUG but not SNAIL or TWIST is expressed in ES cells and is part of the ES cell signature activated in several cancer cells [
58,
59]; 5) SLUG contributes to the function of the stem cell factor c-kit signaling pathways [
60]; c-kit is overexpressed in basal breast cancers, which are also predominantly of CD44+/CD24- phenotype [
21,
61]; 6) E-cadherin, whose experimental depletion in mammary epithelial cells leads to generation of CSC-like cells [
62], represses SLUG but not SNAIL or TWIST expression [
56]; 7) SLUG is a potent repressor of the tumor suppressor p53 and deregulation of p53 activity is often observed in breast cancers [
63]; 8) SLUG-/- embryonic fibroblasts show reduced expression of several genes linked to self-renewal and chromatin remodeling [
64]; 9) SLUG confers resistance to radiation, a characteristic of CD44+/CD24- cells [
65,
66]; 10) SLUG but not SNAIL expression, is linked to ductal development in the breast including tubule maintenance or growth within invasive ductal carcinoma [
67]; and 11) SLUG is highly expressed in basal type breast cancers, which also tend to express higher levels of stem cell-associated genes [
20,
51,
68]. These observations along with the results of this study highlight critical role of SLUG in breast cancer.
Gli-2, a hedgehog pathway mediator linked to EMT [
69], failed to alter the CD44 and CD24 phenotype of MCF-10A cells. This is somewhat surprising considering previous reports showing elevated expression of Gli-2 in primary mammary stem/progenitor cells and its association with increased mammosphere-forming ability [
16]. Overall our results raise two important issues: 1) not all EMT associated genes can induce the CD44+/CD24- phenotype when overexpressed; and 2) not all breast epithelial subtypes (luminal versus basal) acquire a subpopulation of cells with the CD44+/CD24- phenotype upon overexpression of EMT-associated genes. In this respect, our results agree with the observation that primary breast cancers expressing EMT markers are predominantly of basal-like phenotype [
47,
51].
The ability of only a few EMT-associated genes to induce the EMT phenotype and upregulate the expression of genes linked to the CD44+/CD24- phenotype suggests that genes with this dual capacity have stronger transcriptional activity, and in particular target chromatin modifying genes. Several chromatin modifying genes including HMGA2, EPC2, CBX6, SMARCA1, and SMARCA3 showed increased expression in CD44+/CD24- cells (Additional file
2, Table S2). Future studies focusing on these genes may help to identify drugs that target the CD44+/CD24- subpopulation of cancer cells, as has been demonstrated recently using human mammary epithelial cells depleted of E-cadherin or overexpressing TWIST [
62].
Competing interests
HN, SB and Indiana University have submitted a patent application for ANTXR1, which is expressed at higher levels in CD44+/CD24- cells, as a marker and target for breast cancer stem cells.
Authors' contributions
PBN performed most of the experiments described in the manuscript including generating cell lines and RT-PCR. HA did mammosphere assays and RT-PCR. CB and EFS designed and performed several flow cytometry assays. PF did RT-PCR studies, whereas SB did H&E analysis of 3 D cultures and writing the manuscript. RG provided primary tumor samples, which were analyzed as part of this study and participated in the design of the study. HN was responsible for designing experiments, flow cytometry, and writing the manuscript. All authors have read and approved the final manuscript.