SEARCH

SEARCH BY CITATION

Keywords:

  • Mammary gland side population cells;
  • Murine and human;
  • Microarray gene profiling

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

Similar to the bone marrow, the mammary gland contains a distinct population of Hoechst-effluxing side population cells, mammary gland side population cells (MG-SPs). To better characterize MG-SPs, their microarray gene profiles were compared to the remaining cells, which retain Hoechst dye (mammary gland non-side population cells [MG-NSPs]). For analysis, Gene Ontology (GO) that describes genes in terms of biological processes and Ontology Traverser (OT) that performs enrichment analysis were used. OT showed that MG-SP-specific genes were enriched in the GO categories of cell cycle regulation and checkpoints, multidrug-resistant transporters, organogenesis, and vasculogenesis. The MG-NSP-upregulated genes were enriched in the GO category of cellular organization and biogenesis, which includes basal epithelial markers, p63, smooth muscle actin, myosin, α6 integrin, cytokeratin (CK) 14, and luminal markers CK8 and CD24. Additional studies showed that a higher percentage of MG-SPs exist in the G1 phase of the cell cycle compared with the MG-NSPs. G1 cell cycle block of MG-SPs may be explained by higher expression of cell cycle-negative regulatory genes such as transforming growth factor-β2, insulin-like growth factor binding protein-5, P18INK4C, and wingless-5a (Wnt-5a). Accordingly, a smaller percentage of MG-SPs expressed nuclear β-catenin, possibly as a consequence of the higher expression of Wnt-5a. In conclusion, microarray gene profiling suggests that MG-SPs are a lineage-deficient mammary gland subpopulation expressing key genes involved in cell cycle regulation, development, and angiogenesis.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

Goodell et al. reported almost a decade ago that a population of bone marrow cells was capable of an elevated rate of Hoechst 33342 dye efflux and that flow cytometry (FACS) analysis performed at two different wavelengths permitted the isolation of these cells, termed side population cells (SPs) [1]. Bone marrow SPs were enriched at least 1,000-fold in hematopoietic stem cell (HSC) activity following transplantation into irradiated hosts. The hematopoietic SP phenotype was shown to be sensitive to the multiple drug resistance transporter protein (MDRTP) inhibitor verapamil [1], and more recently, to depend upon the presence of the adenosine-binding cassette (ABC) family member Abcg2 or breast cancer resistance protein 1 (Bcrp1) [2]. Subsequently, SP populations have been identified in many other normal tissues including skeletal muscle, pancreas, lung, retina, liver, testis, heart, epidermis, mammary gland, and embryonic tissue, suggesting that the SP phenotype might represent a universal property of stem cell populations [213]. Furthermore, a number of cancer cell lines and primary tumor cells contain an SP population, leading to the suggestion that these cells might constitute a population of drug-resistant “cancer stem cells” [1416]. Recently, purified side population of U373 glioma and MCF7 breast cancer cell lines, as well as a xenograft prostate tumor (LAPC-9), were shown to be more tumorigenic than the corresponding non-side population cells [17].

Similar to the bone marrow, the mammary gland in mice and human contains a distinct population of Hoechst-effluxing SPs known as mammary gland SPs (MG-SPs). MG-SPs constitute about 0.25%–5% of the epithelial cells [reviewed in ref. 18]. The remaining mammary gland cells, which retain the Hoechst dye, will be referred to as mammary gland non-side population cells (MG-NSPs). In mice, MG-SPs were enriched for the expression of a hematopoietic stem cell marker, Sca-1, and for label-retaining cells (LRCs) compared with MG-NSPs [19]. In addition, transplantation of a limited number of MG-SPs gave rise to outgrowths with normal alveolar and ductal structures [19]. However, enrichment in outgrowth potential could not be demonstrated, possibly due to the toxicity of the Hoechst dye [19]. A similar study by Alvi et al. showed that MG-SPs initially expressed lower levels of the luminal markers cytokeratin (CK) 19, an epithelial membrane antigen (EMA), and a myoepithelial marker, CK14, compared with the MG-NSPs [20]. However, upon exposure to differentiating conditions in vitro and in vivo, both SPs and NSPs gave rise to clones and outgrowths that expressed both luminal (CK18, CK19) and myoepithelial (CK14) markers, respectively. Similar to the previous studies, enrichment in outgrowth potential could not be established for the MG-SPs [20].

Similar to mice mammary glands, a majority of human breast SPs initially do not express EMA or common acute lymphocytic leukemia antigen (CALLA), markers of human luminal and myoepithelial cells, respectively [21]. An in vitro growth assay showed that MG-SPs possessed a significantly higher cloning efficiency and therefore had a higher proliferative potential [21]. In addition, a single SP could give rise to CK18 and CK14 double-positive colonies as well as mixed CK18 and CK14 single-positive colonies, whereas NSPs gave rise to mainly CK14-positive colonies, suggesting that MG-SPs represent a putative stem cell population in the mammary gland [21]. Clarke et al. also showed that the human breast SPs are mainly negative for CALLA and MUC1 (also known as EMA) [22]. Breast SPs expressed higher levels of p21 cip1 (a cyclin-dependent kinase inhibitor), which is also expressed in hematopoietic stem cells, and musashi-1, a known marker of neural and intestinal stem cells [22]. These investigators showed that the MG-SPs had a greater capacity than MG-NSPs to produce differentiated structures in three-dimensional Matrigel cultures [22]. Finally, Dontu et al. have developed a nonadherent in vitro culture system that allows for enrichment of undifferentiated human breast epithelial cells in suspension, termed mammospheres [23]. These investigators showed that mammospheres were enriched for SPs, and compared with the freshly dissociated tissue, the mammospheres had a higher potential to give rise to mixed colonies containing myoepithelial, ductal, and alveolar cells [23].

Although these studies have suggested that MG-SPs may be considered candidate stem cells, evidence that they are enriched for functional stem cell activity is still only correlative [18]. Furthermore, it is apparent that even in the hematopoietic system, SPs may represent a heterogeneous cell population that contain cells with different outgrowth and progenitor characteristics [24]. With these caveats in mind, and to better characterize the MG-SP population, we have performed comprehensive gene expression profiling studies of these cells using Affymetrix microarrays (Santa Clara, CA, http://www.affymetrix.com). These studies have revealed that MG-SPs differentially express a number of key genes involved in cell cycle control, organogenesis, and angiogenesis and do not express a number of mammary gland differentiation markers. Furthermore, MG-SPs express a number of cell cycle checkpoints and multidrug transporter genes. These gene profiling studies provide an important addition to the stem cell databases and will be useful for future comparisons to other normal mammary gland stem and progenitor cells, as well as an increasing number of cancer models.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

Primary Mammary Epithelial Cell Culture and Hoechst Staining

The methods for the isolation of mammary glands, preparation of primary mammary epithelial cell cultures, as well as Hoechst staining, have been described previously [19]. For the microarray experiments, primary mammary epithelial cells (PMECs) were prepared from 15 to 20 virgin C57/BL6 mice 4–6 weeks of age. PMECs were cultured for 3 days to facilitate the preparation of sufficient numbers of single cells required for flow cytometry analysis and gene profiling. Analysis and sorting were performed on a triple laser MoFlo (Cytomation, Fort Collins, CO, http://www.cytomation.com). The Hoechst dye was excited at 350 nm, and its fluorescence was measured at two wavelengths, 450/20 band pass (Hoechst Blue) and 675 Edge Filler long pass (Hoechst Red).

Microarray Studies

Total RNA was isolated from the sorted MG-SPs and MG-NSPs using an RNA purification kit (PicoPure RNA Isolation Kit; Arcturus, Mountain View, CA, http://www.arctur.com), followed by mRNA amplification using a T7 global amplification method (Two-Cycle Target Labeling kit; Affymetrix), DNA fragmentation, biotinylation, and hybridization onto Affymetrix 430 2.0 array chips. Normalization and model-based expression analyses were performed using dChip (dChip software; Harvard School of Public Health, Boston, http://www.dchip.org) [25]. The scanned arrays passed the quality control standards such as visual inspection and eukaryotic control performance standards suggested by the manufacturer (Affymetrix). The cell intensity file (*.CEL) files were then used to perform further analysis in dChip [25] as outlined in the experimental design section.

Cell Cycle Analysis

Cell cycle analysis was performed on the sorted MG-SPs, MG-NSPs, and unsorted PMECs according to the protocol described [26]. Briefly, cells were washed in phosphate-buffered saline (PBS) and resuspended in 0.5 ml of PCBS solution (0.1 M phosphate-citrate, 0.15 M NaCl, 5 mM sodium EDTA, 0.5% bovine serum albumin, pH 4.8) containing 0.02% saponin (catalog no. S4521; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) and 10 μg/ml 7-aminoactinomycin D (7-AAD; catalog no. A9400; Sigma-Aldrich) followed by incubation for 20 min at room temperature. Cells were then washed one time in PBS and resuspended in 0.5 ml of PCBS containing 10μg/ml 7-AAD and 1 μg/ml pyronin Y (PY) (catalog no. P9172; Sigma-Aldrich). The mixture was kept at 4°C and analyzed the following day using a FACScan (Becton, Dickinson and Company, Franklin Lakes, NJ, http://www.bd.com).

Immunofluorescence

MG-SPs and MG-NSPs were sorted and plated on collagen-coated coverslips overnight to allow for cell adherence. Cells were then fixed in BD Cytofix/Cytoperm (catalog no. 554722; BD Biosciences, San Diego, http://www.bdbiosciences.com) for 15 min followed by washing and blocking using BD Perm/Wash buffer (catalog no. 554723; BD Biosciences) according to the manufacturer's guidelines. Immunofluorescence was carried out using nonphosphorylated active β-catenin (dephosphorylated on Ser37; catalog no. 05-01; Upstate Biotechnology, Lake Placid, NY, http://www.upstate.com) following the standard protocols recommended by the manufacturer. The antibody was used at a dilution of 1:200 at 4°C overnight. Immunofluorescence staining was also performed on cells cytospun onto slides immediately following a sort using another antibody that also recognizes nonphosphorylated active β-catenin (dephosphorylated on Ser37 or Thr41; catalog no. 05-665; Upstate).

Gene Ontology (GO) and Ontology Traverser

Ontology Traverser (OT) is a web-based tool for performing enrichment analysis of microarray gene lists. The Traverser works by mapping gene lists onto the Gene Ontology (GO) data structure and calculating enrichment statistics for each GO node. The enrichment statistics are reported for GO nodes found in any path leading to GO annotations for the probes in the microarray gene list. All GO levels are considered in the analysis, and the entire connectedness of the GO data structure is being used (a directed acyclic graph). Detailed information regarding OT and the formulas used to calculate enrichment statistics have been published.

Quantitative Polymerase Chain Reaction

RNA was extracted from MG-SPs and MG-NSPs using an RNA purification kit (PicoPure RNA Isolation Kit; Arcturus). One microgram of RNA per sample was treated with DNase I as recommended by the manufacturer, followed by generation of cDNA using Superscript II (Invitrogen, Carlsbad, CA, http://www.invitrogen.com). One-tenth of the final reaction volume was used in quantitative polymerase chain reactions (Q-PCRs) containing 1 mM MgCl2, platinum Taq (Invitrogen), and 1 × SYBR green (Molecular Probes Inc., Eugene, OR, http://probes.invitrogen.com). PCR was performed using the ABI PRISM 7500 thermocycler (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com) with the following cycles: 94°C for 30 seconds, 55°C for 30 seconds, and 72°C for 30 seconds for 40 cycles. Relative quantitation of gene expression was calculated and normalized to glyceraldehyde-3-phosphate dehydrogenase. Q-PCR primers are listed in supplemental online Table 3.

Results and Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

Experimental Design

Gene expression profiles were obtained by hybridizing amplified RNA from four replicate MG-SP and MG-NSP samples to Affymetrix 430 2.0 microarray chips. To isolate MG-SP and MG-NSPs, mammary gland cells were stained using the Hoechst dye 33342 and fluorescence displayed at two wavelength emissions, blue and red. The MG-SP and MG-NSP regions are indicated by trapezoids on the left (R1) and right (R2), respectively (Fig. 1). We and others have previously shown that the R1 region is composed of side population cells since verapamil markedly inhibits their appearance [19, 20]. The cells in each region (R1 and R2) were sorted and their RNA was isolated and amplified by two rounds of in vitro amplification and applied to Affymetrix chips. Hybridization, scanning, and production of raw data files were performed according to the Affymetrix standard protocols. Normalization and model-based expression measurements were performed with dChip [25]. Selected quality assessment parameters are shown in supplemental online Table 1. One of the replicate sets was eliminated from further analysis due to the high percentage of probe sets called an array outlier after model-based expression analysis by dChip. The genes were filtered to eliminate those with very low expression values in most samples. From 45,000 probe sets, 16,744 probe sets were retained and used for further analysis. For normalization purposes and to allow for data comparison with other centers, universal mouse reference RNA (Stratagene, La Jolla, CA, http://www.stratagene.com) was hybridized to two chips. All other arrays were normalized to one of the universal mouse reference RNA arrays expressing average intensity closest to the median intensity of all chips. For comparison, a paired t test was performed with a false discovery rate (FDR) by permutation of class labels set at <10% to assure a low chance of false positive. Figure 2 shows the number of genes that demonstrated a statistically significant difference in expression values based on the indicated criteria for comparison. In these comparisons, MG-SPs were called the experimental set and MG-NSPs were called the baseline set. Therefore, negative values denote downregulated expression, and positive values denote upregulated expression. The list of genes in each comparison group, as shown in Figure 2, is provided in supplemental online Table 2A–2D. Q-PCR by SyberGreen showed similar expression patterns for 14 of 21 selected genes (70%; supplemental online Table 3).

Gene Ontology Grouping of MG-SP- and MG-NSP-Specific Genes

To examine the significant biological processes represented by MG-SP and MG-NSP-specific genes, we used GO annotations (http://www.geneontology.org) that describe genes in terms of their associated biological processes and OT that performs enrichment analysis of microarray gene lists. Gene lists that were differentially expressed by MG-SPs and MG-NSPs (771 genes at 1.5-fold difference, p value ≤ .05%, FDR <3.4%) were used, and a list of groups of GO biological processes with a p value enrichment of ≤ .05% were generated. The significant GO annotations and genes in each category are available in supplemental online Table 4A and 4B and demonstrated as a pie chart (Fig. 3).

MG-SP Gene Profiles

Cell Cycle Regulation and Cell Cycle Checkpoints.

One of the significant findings in this analysis was that MG-SP-specific genes were mapped to the enriched GO categories of cell cycle, cell cycle regulation, and cell cycle checkpoints (Fig. 3A). The genes in these categories included those encoding G2/M cell cycle proteins such as cyclin A2; cyclin B1; cyclin B2; Ki67; ubiquitin-conjugated enzyme e2c; kinesin family members 11, 22, and 2c; kinetochore-associated protein 2; and cdc 20 (supplemental online Table 4A). Cell cycle analysis of the MG-SPs and MG-NSPs showed that approximately 83% of the MG-SPs existed in G1 phase of the cell cycle compared with 61% of the MG-NSPs and 43% of the unsorted PMECs (Table 1; Fig. 4). The mechanism responsible for the accumulation of a higher percentage of MG-SPs in G1 phase of the cell cycle requires further studies. However, we speculate that the MG-SP G1 phase cell cycle block may be explained by the increased expression of negative cell cycle regulatory proteins such as p18INK4C (an early G1 phase cell cycle inhibitor); checkpoint kinase Chk1; aurora kinases A and B; transducer of ERBB2, 2 (TOB-2); insulin-like growth factor binding protein-5 (IGFBP-5); and ligands such as transforming growth factor β-2 (TGF-β2) and Wnt-5a, a noncanonical Wnt ligand.

TGF-β2 and IGFBP-5 expression levels were also increased in the LRCs in the skin bulge, an epidermal stem cell niche microenvironment, and have been proposed to mediate the quiescent state of these cells [27]. Cyclin D-type binding protein 1 has been shown to bind to the D-type cyclins and to inhibit the kinase activity of cyclin D1 (a G1 phase cyclin), resulting in the suppression of E2F-mediated transcriptional activity [28]. These results suggest that the self-renewal and engraftment of the MG-SPs may be similarly affected by the negative regulation imposed by these proteins.

Wnt-5a is a Wnt ligand that activates the noncanonical Wnt signaling pathway in many biological systems, including Xenopus embryos and mammalian cells [29]. There is increasing evidence that the noncanonical Wnt signaling may antagonize the canonical Wnt-β-catenin pathway by several proposed mechanisms, including the degradation of β-catenin by Siah2 [29]; activation of Nemo-like kinase, which phosphorylates T cell factor/lymphoid enhancer factor transcription factors and interferes with β-catenin DNA binding [30]; and competition for disheveled or other molecules shared by the two pathways [31]. Since Wnt-5a transcripts were upregulated in MG-SPs (supplemental online Table 3), we hypothesized that Wnt-5a may inactivate the canonical Wnt signaling pathway by inhibiting β-catenin nuclear localization. To examine Wnt signaling activity, the expression of active β-catenin, a central player in the canonical Wnt pathway, was studied by flow cytometry and immunofluorescence using an antibody to nonphosphorylated β-catenin (active β-catenin). Flow cytometry revealed that about 25% of the MG-SPs and 65% of the MG-NSPs expressed intracellular nonphosphorylated β-catenin (data not shown). However, since intracellular flow did not allow discrimination between the nuclear versus cytoplasmic active β-catenin expression, we performed nuclear localization studies by immunofluorescence. Immunofluorescence staining of the MG-SPs and MG-NSPs cultured overnight on collagen-coated coverslips showed that β-catenin was localized to the nuclear heterochromatin regions (Fig. 5A). Wnt hyperplastic glands also showed a similar nuclear staining pattern, as was seen in a subpopulation of MG-NSPs (Fig. 5B). Nuclear β-catenin was expressed in approximately 40% of the MG-NSPs, 11% of the MG-SPs, and 1% of the unsorted cells (Fig. 5B). A direct cytospin of cells onto slides also showed that a significantly higher percentage of MG-NSPs expressed active nonphosphorylated β-catenin (supplemental online Fig. 5A, 5B). Therefore, overnight culture did not appear to affect the behavior of these cells. These data suggest that the canonical Wnt signaling pathway is inactive in the majority MG-SPs. In the mammary gland, Wnt-5a has been shown to reduce growth and motility of the epithelial cells, and Wnt-3a, a canonical Wnt signal activator, has been shown to increase the growth of the mammary gland cells, including MG-SPs [3234]. It has also been proposed that both the canonical Wnt and Hedgehog (Hh) pathways are inactive in stem cells and in the stem cell niche microenvironment until an inflammatory signal triggers these cells to exit the niche and become transit amplifying cells [35]. Similarly, the Wnt canonical pathway has been reported to be silent in the skin-bulge niche microenvironment [36]. Similar to the skin bulge, we speculate that the block of MG-SPs in the G1 phase of the cell cycle may be due to an inactive canonical Wnt signaling. Whether Wnt-5a contributes to the inactivation of canonical Wnt signaling and the G1 phase block of the MG-SPs requires further investigation.

A number of cell cycle checkpoint genes, such as those encoding aurora kinases A and B, Gadd45g, Chk1, and cyclin D type binding protein 1, displayed significantly higher expression in the MG-SPs compared with MG-NSPs. These genes are normally upregulated in response to genotoxic agents to prevent cell cycle entry, allowing time for DNA repair. These data suggest that MG-SPs might have an active DNA damage response to ensure chromosomal repair and genomic stability. As mentioned previously, isolation of MG-SPs requires use of the DNA vital dye, Hoechst 33342. Hoechst 33342 is toxic, which may cause cell death or the induction of a DNA damage response. However, a DNA damage response is usually very rapid and involves post-translational processing, relocalization, and conformational change or stability of pre-existing proteins [37]. Therefore, it is unlikely that the short exposure to the Hoechst dye induced the upregulation of transcriptional machinery of DNA damage response genes. Furthermore, since the MG-SPs efficiently efflux the Hoechst dye compared with the MG-NSPs, it is likely that the higher expression of checkpoint genes is not a consequence of Hoechst dye–induced DNA damage. Rather, it is possible that MG-SPs possess an active DNA damage machinery to allow for efficient repair in case of a DNA damage insult. Interestingly, a recent study proposed 11 signature cancer stem cell genes, predicting therapy failure and poor prognosis in patients with solid as well as hematopoietic cancers [38]. Up-regulation of 8 of the 11 genes was associated with poor prognosis, and 4 of those 8 genes were also upregulated in the MG-SPs, including those encoding Ki67, cyclin B1, BUB1, and kinetochore-associated 2 (KNTC2). Three of the four genes (those encoding Ki67, cyclin B1, and KNTC2) were included in the Ontology Traverser-enriched GO category of cell cycle regulation (supplemental online Table 4A).

Multidrug-Resistant Transporters.

MG-SPs, compared with MG-NSPs, express increased levels of MDRTPs such as ATP-binding cassette, subfamily B (MDR/TAP), member 1A (mdr 1a); ATP-binding cassette, subfamily B (MDR/TAP), member 1B (mdr 1b); ATP-binding cassette, subfamily C (CFTR/MRP), member 4; ATP-binding cassette, subfamily C (CFTR/MRP), member 9; and ATP-binding cassette, subfamily G (WHITE), member 2 (Bcrp1 or ABCG2). The expression of these transporters by the MG-SPs suggested that the SP phenotype may be contributed by a number of ATP-binding cassette transporters in the mammary gland, not just Bcrp1 as seen in the hematopoietic stem cells. These results have been confirmed by the analysis of the SP population in Bcrp1/Mdr 1a/1b triple knockout mice [6]. These investigators reported that whereas Bcrp1 is almost exclusively responsible for the SP phenotype in the bone marrow, both Mdr 1a and Mdr 1b transporters contributed to this phenotype in the mammary gland. Additional studies have demonstrated that Bcrp1 is highly expressed in alveolar epithelial cells during late pregnancy and lactation but only expressed in a small number of cells, usually basal to the myoepithelial cells, in mammary ducts in virgin mice [6]. Interestingly, a novel estrogen receptor response element has been reported in the Bcrp1 promoter [39].

Cells that express MDRTPs, such as SPs, are protected from chemotoxic agents as well as from hypoxic conditions. MDRTPs are responsible for the efflux of agents such as mitoxantrone, topotecan, doxorubicin, daunorubicin, and methotrexate [14]. Likewise, cells that express Bcrp1 are protected from hypoxic conditions since Bcrp1 binds to and prevents the accumulation of heme (a porphyrin), which is detrimental to cell survival in a hypoxic environment [16]. These results, coupled with higher expression of DNA damage repair and checkpoint genes by the MG-SPs, suggest that several mechanisms may be in place to protect MG-SPs against cellular insult and to ensure the maintenance of chromosomal integrity specifically in these cells.

Angiogenesis.

Another enriched GO category for MG-SPs was angiogenesis and vascular development (Fig. 3A; supplemental online Table 4A). The genes included in this category were known endothelial-specific genes such as Tie-2 (a receptor for angiopoietin 1 and 2), enos (endothelial nitric-oxide synthase), angiopoietin-2, Flt-1 (also known as vascular endothelial growth factor receptor-1), and FLK-1 (also known as vascular endothelial growth factor receptor-2). Others that belong to this category but were not endothelial-specific genes were plexin D1, neuropillin 1, endoglin, EGF-like domain 7 (Egfl7), and cysteine rich protein 61 (CYR61). CYR61 and Egfl7 are known major inducers of angiogenesis. CYR61 is a member of the cysteine-rich CYR61 family of secreted matricellular proteins, which are key modulators of cell-matrix interactions [40]. The expression of matricellular proteins increases during development and growth, and in response to injury [41]. CYR61 binds to integrin receptors and promotes cell adhesion, migration, and proliferation [42]. Egfl7 is an endothelial cell-derived secreted factor expressed at high levels in the vasculature associated with tissue proliferation and is downregulated in most of the mature vessels in normal adult tissues [43]. A number of the endothelial genes expressed by the MG-SPs are proposed endothelial precursor markers, such as Tie-2, Flk-1, Flt-1, and CD31 [44]. We have examined the expression of CD31 (platelet endothelial cell adhesion molecule [PECAM-1]) by flow cytometry. As seen in Figure 6, a higher percentage of MG-SPs express CD31 compared with MG-NSPs (18.6% ± 7.8% vs. 2.17% ± 0.8%).

We speculate that MG-SPs may be an important component of the mammary gland stem cell niche, contributing to mammary gland development, remodeling, and neovascularization. Neovascularization involves the incorporation of endothelial precursor cells derived mainly from the bone marrow or blood. However, solid tissue, such as muscle-derived SPs, may also contain endothelial precursor cells. It has been shown that muscle-derived SPs expressed endothelial markers Tie-2 and CD31 and engrafted into endothelium during vascular regeneration in response to injury [45]. Furthermore, there is increasing evidence that tissue-resident endothelial cells provide direct developmental signals and participate in the differentiation and growth of many organs, including the neural system, liver, pancreas, and kidney, a role that is more than a sole provider of oxygen and nutrients [46]. Similarly, Shen et al. have reported that the coculture of endothelial cells with neural stem cells promoted the self-renewal of neural stem cells and inhibited their differentiation [47]. These endothelial expanded neural stem cell clones had a significantly higher capacity to give rise to the differentiated neuroepithelial cells [47]. In addition, a similar study has demonstrated that the coculture of neural stem cells with endothelial cells promotes the conversion of neural cells to endothelial cells by a mechanism that was independent of cell fusion [48]. In addition, a recent report describes a close association between HSCs and sinusoidal endothelium in spleen and bone marrow, as well as endosteum, and proposes that endothelium serves as a niche microenvironment for the HSCs [49].

The present study indicates that about 18% of the MG-SPs expressed the endothelial surface marker CD31, compared with approximately 2% of the MG-NSPs. Previous studies have shown that MG-SPs can give rise to differentiated epithelial structures in vitro and to mammary gland outgrowths in vivo [1922]. Whether this subpopulation of MG-SPs belongs to the endothelial and/or epithelial lineage requires a comprehensive analysis using other endothelial and/or epithelial markers and functional in vivo transplantation studies.

Organogenesis.

OT showed enrichment of genes involved in organogenesis, pattern formation, and development in the MG-SPs. These genes included transcription factors such as the homeobox A3 (Hoxa3), homeobox A5 (Hoxa5), Tbx18, Hesr-1 (a homeobox and notch target gene), Gli3 (a hedgehog transcription factor), Msx-1, and Msx-2. Hedgehog signaling and Notch pathways are known to play essential roles in the self-renewal of stem cells in neural, hematopoietic, intestinal, and mammary gland tissue [35, 50, 51].

Homeobox genes Msx-1 and Msx-2 play essential roles in mammary gland development [52, 53]. Msx-1 and Msx-2 double mutants fail to form a mammary bud indicating their essential role in mammary gland development [52, 53]. In Msx-2 mutant mice, mammary development is arrested at the mammary sprout stage. Msx-1 transcripts are localized to the mammary epithelium; however, Msx-2 transcripts are confined to the stroma in the periductal areas [54]. Msx-2 is expressed in fetal epithelium and mesenchymal cells and shifts after birth to the mammary mesenchymal cells adjacent to the mammary ducts [55]. It has also been shown that removal of epithelium from the gland reduced the expression of Msx-2 from the stroma, indicating a role for Msx-2 in epithelial-mesenchymal interactions and mammary development [54]. Therefore, these results suggest that Msx-1 and Msx-2 genes may regulate the development and fate determination of the MG-SPs. It is likely that MG-SPs in the periductal areas may mediate epithelial-stromal interactions and thus may regulate mammary gland development.

Homeobox A5 (Hoxa5) has been shown to be downregulated in a majority of human breast tumors by promoter methylation [56]. Hoxa5 has a consensus binding site in the promoter region of p53 and consequently regulates the expression of p53 [56]. Furthermore, overexpression of Hoxa5 in epithelial cancer cells expressing wild-type p53 causes apoptotic cell death [56]. Hoxa5 has also been shown to directly control the expression of progesterone by binding a consensus binding site in the promoter region of progesterone receptor [57]. Hoxa3 has been show to have a role in carotid artery development [58]. However, a role for Hoxa3 in the mammary gland has not been reported.

Hairy/enhancer of split-related (Hesr) proteins, such as Hesr-1/Hey1, are basic helix-loop-helix transcription factors implicated in cell fate decisions. Hesr-1 is a direct transcriptional target of the Notch signaling pathway. The Notch signal transduction pathway has also been implicated in the self-renewal of stem cells in hematopoietic, neural, and germ cells. Using Notch receptor activating ligands as well as blocking antibodies, Dontu et al. have shown a role for the Notch pathway in self-renewal and fate determination of mammary gland stem cells [51]. It has also been demonstrated that Hesr-1/Hey1 is expressed in the neural precursor cells and that expression of Hesr-1/Hey1 promotes the maintenance of neural precursor cells and inhibits neurogenesis [50]. Finally, the expression of Notch 1 and Notch 4 ligands were shown to be upregulated by the MG-SPs. Collectively, these data suggest that the Notch pathway is active in these cells and may regulate the maintenance of a progenitor state of the mammary cells as reported for mammospheres [51].

A role for Gli3 in mammary gland development has not been reported, except for the possible involvement of Gli3 in the development of mammary placode induction [59]. Other studies have implicated Gli3 in the correct patterning and differentiation of the neural progenitor cells [60], as well as a role in specification of epaxial muscle progenitor cells [61]. There is increasing evidence for the role of the Hh pathway in stem cell self-renewal [35]. Together, these data indicate that Gli3 may mediate MG-SPs self-renewal and/or fate determination.

MG-NSP-Specific Gene Profiles: Cellular Organization and Biogenesis

MG-NSPs express a number of genes belonging to the enriched GO category of cellular organization and biogenesis (Fig 3.; supplemental online Table 4B). These genes include basal-specific mammary epithelial cell markers such as p63, smooth muscle actin, myosin, α6 integrin (CD49f), and CK5/14, as well as the luminal epithelial markers CK8 and CD24. We also compared our data to the tumor database reported by Perou et al. [62]. This comparison was interesting since a number of human basal tumor markers, such as CK5, CK17, nebulette, enolase 3β muscle, annexin A8, laminin γ2, and calgranulin (a HER2 and/or basal human tumor marker), were differentially highly expressed by the MG-NSPs. MG-NSPs also demonstrated increased expression of CK6a, a putative mammary gland stem cell marker [63]. These data indicate that the majority of mammary gland lineage-specific markers are expressed by the MG-NSPs and that MG-SPs represent a lineage-deficient epithelial cell population. The low expression of mammary lineage markers by the MG-SPs has been reported previously in humans and mice by several groups, which helps validate the gene expression profiles [20, 21].

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

We have performed gene profiling on a subpopulation of mammary gland cells known as mammary gland side population cells. The stem cell potential of these cells has been demonstrated by several in vitro studies; however, direct in vivo evidence is still lacking. Therefore, the exact identity of the MG-SPs is still not known. These gene profiling studies have attempted to clarify SP identity by revealing unique genes, signaling pathways, and markers expressed in this population. In summary, our gene profiling using OT has allowed us to build several hypotheses that require further testing and functional validation. Specifically, MG-SPs lacked the expression of several lineage-specific mammary gland markers, such as myoepithelial and/or luminal markers. Therefore, MG-SPs appear to represent a heterogeneous stem/progenitor cell population lacking mammary gland cell differentiation markers. It is also possible that MG-SPs may serve as a component of the mammary gland stem cell niche, providing support to the mammary gland stem cells. MG-SPs displayed reduced expression of CK6, a putative mammary gland stem/progenitor cell marker. Thus, CK6 expression may be higher in the more committed progenitors. In addition, MG-SPs may have mechanisms in place to protect them from DNA damage, hypoxic conditions, and chemotoxic insults since they express DNA damage response genes as well as MDRTPs. Why in particular these cells as compared to other mammary epithelial cells require such protective mechanisms is not clear and requires further investigation. The higher expression of endothelial-specific genes, angiogenic factors, and genes involved in development by MG-SPs suggests that these cells may be an important component of the stem cell niche and as such may play an essential role in mammary gland development. Additionally, gene profiling showed that MG-SPs are in a state of readiness to enter the cell cycle, since they express a number of G2/M cell cycle regulatory genes and they exist mainly in the G1 phase of the cell cycle. The blockade of MG-SPs in G1 may be due to an inactive canonical Wnt signaling pathway. It is also intriguing to speculate that MG-SPs may be involved in tumorigenesis, since these cells express several of the MDRTPs, as well as 4 out of the 11 cancer stem cell signature genes. In contrast, MG-NSPs express a number of mammary gland luminal and myoepithelial markers, suggesting that MG-NSPs may include differentiated or more committed progenitor cells.

Table Table 1.. Cell cycle analysis of MG-SP, MG-NSP, and unsorted PMEC cells
Thumbnail image of
thumbnail image

Figure Figure 1.. A representative mammary gland side population cell (MG-SP) profile. Mammary gland cells were stained using the Hoechst dye 33342 and fluorescence displayed at two wavelength emissions, blue (FL7) and red (FL8). MG-SP and mammary gland non-side population cell (MG-NSP) regions are indicated by trapezoids on the left (R1) and right (R2), respectively. As indicated by the flow cytometry profile, MG-SP represent approximately 2% and MG-NSPs approximately 16.33% of the mammary gland cells. The remaining 81.67% are dead cells and cellular debris.

Download figure to PowerPoint

thumbnail image

Figure Figure 2.. Differentially expressed number of genes by the mammary gland side population cells (MG-SPs) and non-side population cells (MG-NSPs). Each bar represents the number of genes differentially expressed by MG-SPs and MG-NSPs based on a paired t test comparison plotted against the selected criteria, p value, fold change, and false discovery rate (FDR). a, 1.2-fold change, FDR 9.4%, p value ≤ .05; b, 1.5-fold, FDR 3.4%, p value ≤ .05; c, twofold change, FDR 0%, p value ≤ .05; d, twofold change, FDR 0%, p value ≤ .01.

Download figure to PowerPoint

thumbnail image

Figure Figure 3.. Ontology Traverser analysis of mammary gland side population cell (MG-SP) and non-side population cell (MG-NSP)-specific genes. Significant Gene Oncology categories represented by MG-SP-specific (A) and MG-NSP-specific (B) genes. The group of genes used for this analysis included 771 differentially expressed genes at a 1.5-fold difference in expression value, FDR = 3.4%, and a p value ≤ .05.

Download figure to PowerPoint

thumbnail image

Figure Figure 4.. Flow cytometry analysis of MG-SP and MG-NSP cell cycle status. MG-SPs and MG-NSPs were sorted and stained with PY and 7-AAD for the quantification of RNA and DNA, respectively, as described in Materials and Methods. A higher percentage of MG-SPs (83.6%) exist in G1 phase of the cell cycle compared with MG-NSPs (61%). Abbreviations: 7-AAD, 7-aminoactinomycin D; MG-NSP, mammary gland non-side population cell; MG-SP, mammary gland side population cell; PY, pyronin Y.

Download figure to PowerPoint

thumbnail image

Figure Figure 5.. Immunofluorescence staining for active β-catenin. (A): Representative MG-SP, MG-NSP, and Wnt hyperplasia tumor cells stained for active β-catenin. Most MG-NSP and Wnt hyperplasia tumor cells express a distinct nuclear active β-catenin staining in the heterochromatin regions. (B): Bar graph indicates the percentage of unsorted PMECs (1% ± 0.2%), MG-SPs (11.3% ± 6%), and MG-NSPs (41% ± 3.2%) that stained positive for the nuclear active β-catenin. A total of 651 unsorted PMECs, 354 MG-SPs, and 568 MG-NSPs were counted. Abbreviations: MG-NSP, mammary gland non-side population cell; MG-SP, mammary gland side population cell; PMEC, primary epithelial cell.

Download figure to PowerPoint

thumbnail image

Figure Figure 6.. Expression of CD31 in MG-SPs versus MG-NSPs. Primary epithelial cells were stained with Hoechst 3342 as described in Materials and Methods, followed by staining with the anti-CD31 antibody. MG-SPs and MG-NSPs were gated and analyzed for the expression of CD31. A significantly higher percentage of MG-SPs expresses CD31 compared with MG-NSPs (18.6% ± 7.8% vs. 2.17% ± 0.8%, respectively). Bar graph represents accumulated data from four independent experiments. Abbreviations: MG-NSP, mammary gland non-side population cell; MG-SP, mammary gland side population cell.

Download figure to PowerPoint

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

We thank Drs. Daniel Medina, Margaret A. Goodell, Yi Li, Kaylee Schwertfeger, Tracy Vargo Gogola, and Heather L. Lamarca for helpful comments. We thank Dr. Charles M. Perou for the comparison of microarray data to the reported human tumor database and for helpful comments. In addition, we thank Mike Cubbage, Chris Threeton, and Tatiana Goltsova for excellent flow cytometry assistance and Lisa While and Hilary N. Lewis at the Baylor Microarray Core Facility. This work was supported by grants F32 CA101560-03 (F.B.) and UO1 CA84243 (J.M.R.) from the National Cancer Institute.

Authors' Note

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

Recently, MG-SPs have been analyzed for the expression of endothelial specific adhesion molecule-1 (ESAM-1) by FACS. ESAM-1 is an adhesion molecule shown to be upregulated by long-term HSCs [64]. The majority of MG-SPs expressed ESAM compared to MG-NSPs (83.4 ± 6.4 vs. 4.2 ± 0.5).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results and Discussion
  6. Conclusion
  7. Acknowledgments
  8. Authors' Note
  9. Disclosures
  10. References
  11. Supporting Information
FilenameFormatSizeDescription
SC050375SuppFile1.pdf37KTable S1
SC050375SuppFile2.pdf112KTable S2a
SC050375SuppFile3.pdf632KTable S2b
SC050375SuppFile4.pdf282KTable S2c
SC050375SuppFile5.pdf144KTable S2d
SC050375SuppFile6.pdf33KTable S3
SC050375SuppFile7.pdf22KTable S4a
SC050375SuppFile8.pdf18KTable S4b
SC050375SuppFile9.jpg1383KFigure S5

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.