Stem Cell Reviews

, Volume 3, Issue 2, pp 124–136

Prospective Isolation and Functional Analysis of Stem and Differentiated Cells from the Mouse Mammary Gland


  • Joseph Regan
    • The Breakthrough Breast Cancer Research CentreInstitute of Cancer Research
    • The Breakthrough Breast Cancer Research CentreInstitute of Cancer Research

DOI: 10.1007/s12015-007-0017-3

Cite this article as:
Regan, J. & Smalley, M. Stem Cell Rev (2007) 3: 124. doi:10.1007/s12015-007-0017-3


Prospective isolation and in vitro and in vivo analysis of primary mouse mammary epithelial cells has been used to separate cell subpopulations and identify stem, progenitor and differentiated cell compartments. Progress has been made from cell separation strategies based on a single marker of the luminal epithelial or myoepithelial compartments to use of markers that allow simultaneous isolation of non-epithelial, basal/myoepithelial and luminal epithelial cells. Transplant analysis has shown that mammary stem cells are found in the basal/myoepithelial compartment, whereas in vitro colony progenitors are found in the luminal compartment. A basal population enriched for stem cell activity can be purified from the myoepithelial cells and the most recent data shows that the luminal population can now be prospectively split into estrogen receptor positive and estrogen receptor negative cells. Future work aims to molecularly characterise these populations to identify new drug targets, which can be used to specifically kill breast cancer stem cells.


Stem cellsCancerMammary glandSide populationCD24

Stem Cells in Normal Biology and Cancer

Recent data has provided strong evidence supporting the cancer stem cell hypothesis [14]. This hypothesis has two related components. The first of these concerns the cellular origin of tumours and the question of whether they arise from tissue stem cells [2]. The second is the suggestion that tumours are driven by cellular components that exhibit the properties of stem cells [510].

Adult stem cells are defined by their capacity to self-renew and to generate daughter cells that can differentiate down several cell lineages to form all the cell types found in the mature tissue [1115]. The differentiation capacity of these tissue-specific stem cells is largely restricted to cell types within a particular organ, a characteristic that distinguishes them from embryonic stem cells. It is this ability to self renew that allows a stem cell population to persist throughout the lifetime of an organism, providing new cells for tissue genesis, maintenance and regeneration. This long life also means that stem cells have the potential to accumulate the multiple mutations that are required for carcinogenesis, making them excellent candidates for the cells of origin of cancer. Furthermore, normal stem cells and cancer cells share several important properties. These include (1) the faculty for self-renewal, (2) the ability to differentiate, (3) the activation of cytoprotective mechanisms, i.e. increased membrane transporter activity, overexpression of anti-apoptotic proteins, and active telomerase expression, in addition to (4) the ability to migrate [16]. Even properties such as anchorage independence, a hallmark of transformed cells, have been described as a property of normal tissue stem cells [1719].

During normal tissue homeostasis, stem cell self-renewal divisions are asymmetric divisions in which a stem cell is able to replicate itself as well as generate a daughter transit-amplifying or progenitor cell that undergoes differentiation into the functional cell lineages found in the tissue [20]. In some situations, however, such as during normal development and tumorigenesis, stem cells may undergo symmetric divisions in which they produce two identical stem cell progeny, thus allowing for stem cell expansion [21]. It is hypothesised that deregulation of this normally highly regulated self-renewal process may be a key early event in carcinogenesis [16].

The development of animal models permitting the direct assessment of stem cell properties in tumour subpopulations has provided strong experimental support for the second component of the cancer stem cell hypothesis—that tumours contain and are driven by cellular elements displaying stem cell properties [3, 59]. These animal models have enabled the in vivo characterisation of prospectively identified subpopulations of tumour cells and shown them to display the defining stem cell properties of self-renewal and differentiation [5]. It is thought tumour growth is driven by stem cell self-renewal, whereas the aberrant differentiation of their progeny contributes to tumour phenotypic heterogeneity.

The therapeutic implications of a cancer stem cell driving solid tumour growth are profound. Current therapies treat all of the phenotypically different cancer cells that constitute a tumour as though they have unlimited proliferative potential and can acquire the ability to metastasise. It is known, however, that small numbers of disseminated cancer cells detected at sites distant from primary tumours never manifest metastatic disease [22]. It is possible that this is the result of a highly efficient immune response that destroys disseminated cancer cells before they can form a tumour of detectable size. An alternative explanation, favouring the cancer stem cell hypothesis, is that only the dissemination of rare cancer stem cells can lead to metastatic disease and that most cancer cells lack the ability to form a new tumour. If this is the case, new therapies must seek to identify and destroy this cancer stem cell population [1].

An understanding of normal breast stem cells and of breast cancer stem cells would improve our understanding of the origins of breast cancer and enable novel therapies targeted at stem cells to be developed. Until recently, however, analysis of the properties and regulation of mammary stem cells has been limited by a lack of methods for prospective isolation of cellular subpopulations. Using the mouse mammary gland as a model system, we have established methods for the separation and characterisation of the epithelial cell subpopulations within the mammary gland, including candidate stem cell populations and their differentiated daughters. The remainder of this review will focus on the mouse mammary gland as a model system and our previous and ongoing work using this system.

The Mouse as a Model System for the Prospective Isolation of Mammary Epithelial Stem Cells

The mouse mammary gland is the ideal system for improving our understanding of the cellular biology of the breast and clarifying the link between normal and tumour stem cells. Cleared fat-pad transplantation (Fig. 1) has been a standard technique in mammary-gland biology since the pioneering work of De Ome in 1959 [23]. The ability to transplant mammary cells into their normal microenvironment and assess their growth, differentiation, and tumorigenic capabilities is one of the most singular features of mammary gland studies [24]. This attribute is based on the low prepubertal growth rate of mammary epithelium and the accessibility of the mammary fat pad. There are five pairs of mammary fat pads in the mouse. Within the fourth (abdominal) mammary fat pad of the 3-week old mouse the mammary epithelium is still concentrated in the nipple area having not yet grown into the mammary fat pad beyond the lymph node. Using this as an anatomical fixed point, the fat pad can be cut away or ‘cleared’ from the nipple to the lymph node. The bulk of the fat pad is then left free of epithelium and ready to receive cells. The endogenous epithelium must be cleared as otherwise the endogenous tree would overgrow the transplanted cells before they are able to generate their own outgrowth. Transplantation of numbers in excess of 5 × 104 unfractionated primary mammary cells results in success rates greater than 90% and generates a mammary tree, containing both luminal and myoepithelial cells, that resembles the normal mammary tree in all ways except that it is not connected to the nipple. It can even respond to pregnancy by generating alveolar structures identical to those seen in the unmanipulated gland. Immortalized mouse mammary epithelial cell lines do not, to our knowledge, generate a normal mammary tree when transplanted in this way—a normal structure is only generated when freshly harvested primary cells, or those that have been cultured for a short period of time are used.
Fig. 1

Cleared fat pad transplantation. Schematic representation of the procedure together with an example of a carmine-stained transplanted outgrowth. a The endogenous epithelium has not grown past the lymph node in the 4th mammary fat pad of the 21-day old female mouse. b The portion of the fat pad from the nipple to the lymph node, which contains the endogenous epithelium, is removed, leaving an epithelium-free fat pad (‘clearing’). c Test cells are injected into the epithelium-free fat pad. df Fat pads are wholemounted and carmine stained after varying periods, usually 6–8 weeks after transplant. If the injected cells had a stem cell activity, an epithelial outgrowth very similar to that observed in the intact gland is seen, except that it is not connected to a nipple (d). The outgrowth will even respond to pregnancy by undergoing alveolar development (e). If non-marked cells are being used for transplantation, to be sure that outgrowths are derived from the transplanted tissue and not from either a failure of the clearing procedure or ingrowth of epithelium from the 5th fat pad, then outgrowth must be seen to have a central point of origin with ducts extending in all directions (f; bar = 5 mm)

It is predicted that an epithelial cell subpopulation that is enriched for candidate stem cells should produce more successful transplants at limiting cell numbers than a subpopulation that has not been enriched. Transplantation of cells at limiting dilution (<2 × 104) has thus become a key assay in mouse mammary stem-cell biology [25, 26].

The transplanting of marked candidate stem cells mixed with unmarked stem cell depleted cells is a refinement to this process and can be used to assess the contribution of the marked cells to the outgrowths generated [2729]. A possible confounding factor in interpreting the results of this assay and one caveat to its use in testing stem-cell-like behaviour is that the epithelium-free fat pad might be a stem-cell niche that promotes daughter-cell reversion to a more stem-cell-like phenotype.

It is generally accepted that the Terminal End Buds of the immature gland contain stem cell activity [30]. However, there is also a role for stem cells in the adult mammary gland. The epithelial ductal network of the mature virgin animal is relatively simple and quiescent, although there are cyclical changes such as the appearance and regression of secondary branches and alveolar buds that occur with each estrous cycle. These structures are the precursors of the fully differentiated, milk-secreting alveoli that will appear during pregnancy. The extent to which these changes occur depends on animal strain. The burst of proliferation that occurs at pregnancy sees the full elaboration of the mammary epithelial structures, with the formation of secondary and tertiary ductal branches and the maturing of alveolar buds into fully differentiated secretory alveoli. Ultimately, the fully differentiated lactating gland is packed with epithelial alveolar tissue, with minimal adipose tissue remaining. Weaning causes these structures to regress via an apoptotic mechanism until the mammary epithelium once again resembles the virgin state [31]. This cycle of epithelial proliferation, differentiation and apoptosis occurs with every pregnancy. Similar processes to these occur during human pregnancy and the human menstrual cycle, although there are differences in the extent of de novo proliferation. Only stem cells have the replicative potential needed to maintain this process.

Adult mammary epithelial stem cells also have a role in the replacement of cells that are shed from the epithelium into the lumen during routine cell turnover as is seen, for example, in the recovery of epithelial cells from milk during lactation [32]. This might be an important source of cell loss in the resting gland in vivo. Without the replacement of cells as they are shed into the lumen of the alveolar and ductal systems the epithelial tree would not be able to maintain its integrity. So, two potential roles for stem cells in the mammary epithelium exist but whether this means that there are two (or more) stem cell types in a possible stem cell hierarchy or that one stem cell type has the potential to perform several tasks depending on the cues it is given, remains to be seen [33]. Furthermore, the relationship between the two main structural subunits (ducts and lobules/alveoli) and two main cell types (myoepithelial and luminal) of the mammary gland and stem cells remains unclear.

Clonal Analysis of Mammary Epithelium

For many years, the identity of this/these stem cell(s) has remained elusive. As shown by analysis of mouse mammary tumour virus (MMTV) retroviral integration sites, a complete mammary gland can develop from the progeny of a single cell, but the identity of this cell was unclear [25]. The location of undifferentiated cap cells at the tips of the end buds led to the suggestion that they were either stem cells [34, 35] or myoepithelial precursors [36]. Morphological studies have also suggested that undifferentiated small light cells distributed within mammary ducts may be stem cells [37]. However, the lack of appropriate cellular markers for the isolation of live cells has prevented the acquisition of functional data to confirm these assignments.

We argued that to fully understand stem cell identity, the mechanisms behind their potential roles and their relationships with the differentiated cell types of the gland, it was necessary to be able to study separately the behavior of the individual cell types within the gland. Clonal growth and analysis is a powerful method of dissecting cell subpopulations within a heterogeneous tissue such as the mammary gland. However, before being analysed for cell-type specific features, it must first be possible to separate and then grow such cells as clones. We developed systems for the separation and cloning of mouse mammary luminal epithelial and myoepithelial cells [3840] based on those previously described for human breast [41] and rat mammary gland [42]. As well as these systems providing a new methodological approach to the study of mouse mammary epithelial cells and their cell-cell interactions, they also provided information on the phenotypic stability of mouse epithelial cells in culture and enabled their comparison with corresponding rat and human cells. In addition, they provided the baseline for studying the effects of various growth and differentiation promoting factors on the two mouse mammary epithelial cell populations at the clonal level.

The method used for the isolation of mouse mammary epithelial cells in this study has since been modified as described [43, 44]. After removal of the intra-mammary lymph nodes, the fourth mammary fat pads of 10–12 week-old virgin female mice are harvested. Fine mechanical mincing is followed by collagenase/trypsin digestion to liberate epithelial tissue fragments (organoids), non-epithelial fragments and non-epithelial single cells. After a series of washes, and incubations with red blood cell lysis buffer to remove erythrocytes, the resulting suspension is ‘pre-plated.’ The majority of the contaminating fibroblasts, which are single cells, attach to the tissue culture plastic, whereas the epithelial organoids do not. The epithelial and non-epithelial fragments, and contaminating single cells such as lymphocytes and some fibroblasts, are decanted, leaving the majority of the fibroblasts behind. The organoid preparations next undergo a second round of enzymatic digestion to process the organoids to single mammary cells. These cells are then stained with antibodies and sorted into cellular subpopulations for analysis.

Our 1998 study [39] used two cell-type specific rat monoclonal antibodies for flow sorting, namely 33A10 and JB6, originally characterised by Sonnenberg et al. [45]. 33A10 reacts with a milk fat globule membrane antigen on luminal epithelial cells. JB6 was originally reported as staining only basal cells in the outer layer of normal mouse parenchyma and not reacting with differentiated myoepithelial cells in Balb/c mice, although in induced tumours it stained all basal cells [45]. However, we reported that JB6 stained all basal/myoepithelial cells in the Parkes mouse strain used [39], similar to that of the CD10 marker, which had been used for both human and rat myoepithelial cell sorting [41, 42]. Using these antibodies, we were able to separate and clone both cell types as primary cultures and, consequently, separately study their growth and differentiative ability in vitro. At the time the study was carried out the technology and reagents available allowed only limited numbers of antibodies to be used on mouse cells and we were confined to two-way sorting for cells that either did or did not express a single marker.

We found that the key to efficient culture of mouse mammary epithelium at clonal density (approximately 1 cell plated per mm2 of culture surface area) was low oxygen levels (5% vol/vol or less), conditions which enabled many hundreds of mouse clones from flow-sorted luminal and myoepithelial cells to be analysed. Bulk culture of primary mouse mammary epithelial cells and clonal culture of human and rat mammary epithelial cells do not require low oxygen conditions. The study identified four morphologically distinct clone types in adult virgin mouse mammary parenchyma. Type A, B, and C clones were derived from 33A10-positive luminal epithelial cells, whereas Type D clones were derived from JB6-positive myoepithelial/basal cells. Both Type A and Type D clones had limited proliferative potential and appeared to represent differentiated rather than proliferative compartments. Type B clones consisted of uniformly compact and cuboidal cells and were the most similar to the phase contrast appearance of mouse mammary epithelial cells generally observed in bulk cultures [46]. In contrast, the Type C luminal clones were very heterogeneous in their cellular morphologies [39]. However, attempts to resolve the question of the pluripotentiality of different mouse luminal clone types by recloning them was unsuccessful, as cells isolated from all the primary clones consistently failed to give rise to second generation clones [38].

The study also demonstrated that the expression of cell type-specific cytoskeletal markers is deregulated in vitro, leading to promiscuous expression in luminal epithelial clones of basal/myoepithelial markers usually within 24–48 h of being placed in culture [38, 39]. Some basal derived clones also expressed luminal-type cytokeratin markers, although this was only seen 6 days after being placed in culture and only weakly. Promiscuous expression of cytokeratin markers was not seen in cloned human mammary epithelial cells in vitro [41] and was seen only to a limited extent in a subpopulation of cloned rat mammary luminal cells in vitro [42]. In vivo, cytokeratins are important in cell organization and tissue architecture [4749], with significant roles in the mediation of cell identity and cell polarity [50]. Furthermore, it has been shown that differences in expression of cytokeratins are characteristic of human [51] and rat [42] mammary epithelial cell types in vivo and also of luminal epithelial and myoepithelial cells of both species in vitro [41, 42]. The deregulation and promiscuous expression of cytokeratin markers observed in mouse mammary epithelial cells in vitro suggests, therefore, that these cells lose some aspect of their basic cell identity when placed in culture. This also suggests that the maintenance of a differentiated cellular identity in virgin mouse mammary epithelial cells is not cell-autonomous but requires constant extracellular signals, which are not provided for in separated cells grown on tissue culture plastic or glass. This plasticity of mouse mammary cells when placed in culture has led to our current practice of never transplanting or harvesting cells for gene expression analysis if they have had any time in culture. We only carry out such studies on freshly separated cell subpopulations.

Three-dimensional Clonal Culture of Mouse Mammary Epithelium

At least two hypotheses could explain the extracellular signals responsible for reinforcing cellular identity in vivo. Firstly, it is possible that each cell type requires contact with, or paracrine signals from other cells of the same or different types, with, for example, luminal cells requiring signals from myoepithelial cells and vice versa. Cells cultured separately are deficient in signals from the other cell type and could begin to dedifferentiate as a result. Alternatively, it is possible that mouse cells are exceptionally sensitive to the presence or absence of extracellular matrix (ECM). A number of cell types, including mouse mammary epithelial cells [5255], rely on ECM for their morphological and functional differentiation. Despite not supporting proliferation [56], ECM derived from the Engelbreth–Holm–Swarm murine sarcoma (EHS matrix) will maintain ß-casein synthesis in single mouse mammary cells derived from midpregnant mice and suspended therein [57]. ECM has also been shown to be a requirement for the suppression of apoptosis in mammary epithelium [5860]. As well as providing a network of support for cells in vivo, ECM contains ligands that interact with cells via cell surface molecules such as the integrins [56, 6165] and cell surface ß-1,4-galactosyltransferase [6669]. Indeed, the disruption of ß1-integrin function in the mammary epithelium of transgenic mice has been shown to disrupt mammary epithelial proliferation during pregnancy [70, 71] and the use of blocking antibodies for the modulation of integrins has been demonstrated to restore some degree of functional normality in the early stages of malignant progression in breast cells [72, 73].

To further investigate the hypothesis that maintenance of the differentiated identity of mouse mammary epithelial cells requires ECM, we cultured separated virgin mouse mammary luminal epithelial and myoepithelial cells on EHS matrix at clonal density [40]. The morphological examination of the resulting clones along with their cytokeratin expression profiles were used as an index of whether the differentiated identity maintained in vivo was equally maintained by the isolated cells cultured on the EHS matrix. We found that ECM maintained a more differentiated phenotype in mouse mammary basal/myoepithelial cells, which formed flat clones and showed a marked reduction in the extent to which inappropriate antigens were expressed. The uniformity in phenotypes seen with the myoepithelium-derived clones contrasted greatly with those derived from purified luminal epithelial cells. These clones showed diverse morphologies and staining patterns, indicating a greater complexity in the factors determining their specific phenotypes. The growth of mouse mammary luminal epithelial cells on a thick gel layer resulted in the formation of spheroidal clones (“mammospheres”). Within these mammospheres the central mass of cells retained a more differentiated phenotype (cytokeratin 18-positive/19-positive) and did not express basal markers (cytokeratin 14-negative). These cells had no contact with ECM but were surrounded in all directions by other cells of luminal origin. However, the outer layer of the mammospheres retained the promiscuous cytokeratin expression, which stained with both luminal and basal markers (i.e., they were cytokeratin 18-positive/19-positive/14-positive). Interestingly, clones derived from luminal cells plated on thinner layers of EHS matrix formed, in about 10% of cases, two layered clones in which the layer of cells next to the matrix consisted of larger, cytokeratin 14-positive only clones, whereas the cells growing on top of these were more compact, cuboidal cytokeratin 18-positive cells.

This data suggested that sorted single luminal epithelial cells from the virgin mouse mammary parenchyme can generate both luminal and basal epithelial cells, at least in vitro, depending on the manner in which they interact with each other and the surrounding matrix. Under the culture conditions used, luminal epithelium-derived clones did not generate fully differentiated myoepithelial cells with both basal cytokeratins and α-isoform smooth muscle actin. However, it is entirely possible that the differentiative potential of the clones might be medium-dependent. The effects of different medium components on the different mouse cell types have yet to be established.

We proposed that homotypic cell–cell interactions reinforce and maintain the luminal epithelial phenotype and are responsible for the differences between the cytokeratin 18-positive/14-negative phenotype of the innermost cells in the mammospheres and the principally double-positive cells of the flat clones. Whether it is the topography or the extent of these interactions that is important in maintaining the luminal epithelial phenotype in this system is not clear. However, it was the central mass of cells in the mammospheres that exclusively expressed luminal cell markers. It is unlikely that the EHS matrix takes an active role (e.g. via ligand-receptor interactions) in the maintenance of luminal cell identity, as these cells were not in contact with it.

Despite not demonstrating conversion from luminal epithelial cells to fully differentiated myoepithelial cells, the ability of mouse mammary luminal epithelial cells to generate clones possessing a basal cell layer lacking luminal cytoskeletal markers provides evidence that these cells are, or contain, a pluripotent compartment. These findings were supported by studies in which a subset of mature human breast luminal epithelial cells were shown to give rise to myoepithelial cells over an extended culture period [74]. Whether or not all luminal layer-derived cells could generate multilayer clones under the appropriate ideal conditions is impossible to say. We therefore could not ascertain whether the capacity for de novo generation of a basal cell layer is a property common to all mouse luminal epithelial cells or whether it is a property specific to a subpopulation of these cells.

This model, wherein cell-cell interactions reinforce luminal cell identity and contact with an optimal amount of ECM is required for the de novo generation of basal epithelial cells from luminal epithelial cells, also had to be reconciled with our data that only the outermost cell layer of mammospheres cultured in a lactation medium (containing hydrocortisone and prolactin) produced ß-casein whereas the more fully luminally differentiated central cells did not [40]. In vivo, myoepithelial cells form a basket-like network around the alveolus (the functional unit of lactation in vivo), allowing the underlying secretory luminal cells limited contact with the matrix [75]. Similarly, mouse mammary luminal epithelial cells isolated in vitro require contact with ECM for functional differentiation [76]. It is therefore likely that ß-casein expression only occurs in the outer cell layer of mammospheres as only these cells are in contact with a (thin) layer of ECM. Clearly, therefore, the luminal epithelial phenotype as defined by expression of appropriate cytoskeletal markers can be dissociated to some degree from the functional alveolar phenotype as defined by milk protein expression, although the possibility that lactogenic hormones could not diffuse into the interior of the mammospheres could not be definitively excluded as a reason why the central cells did not produce β-casein.

Prospective Isolation of Pluripotent Mammary Cells—The SP Question

We next addressed whether it was possible to prospectively isolate the cells with pluripotent differentiative potential—candidate stem/progenitor cells—for which the in vitro culture systems had provided evidence.

The side population (SP) was first reported by Goodell et al. in 1996 [77], who used flow cytometry for the simultaneous visualization of fluorescence emissions from Hoechst 33342-stained whole bone marrow at two wavelengths to reveal a unique cell staining pattern. The subsequent isolation and characterization of the different cell subsets exposed by this staining profile revealed a group of cells that had cell surface markers characteristic (Sca-1+/linneg/low) of haematopoietic stem cells (HSCs) and that was enriched at least 1000-fold for HSC activity. Being well separated from the rest of the bone marrow, this group of cells was termed the side population. It was speculated that this SP profile was due to an elevated level of Hoechst efflux from the HSCs. The multidrug resistance protein (Mdr or p-glycoprotein) is responsible for the pumping of vital dyes out of cells. When stained in the presence Mdr inhibitory drug verapamil the SP disappeared [77]. However, a comparison of wild type mice with mice deficient for both Mdr1a and Mdr1b showed normal numbers of bone marrow SP cells to be present in the latter. Four members of the Adenosine Binding Cassette (ABC) transmembrane pump family (similar to Mdr) were later identified as being expressed in SP cells. Of these, Abcg2 or Breast Cancer Resistance Protein 1 (Bcrp1) was expressed at much higher levels than the others and further examination of cell lines expressing the four ABC family members showed that only expression of Bcrp1 conferred the SP phenotype. In addition, high levels of Bcrp1 expression were found in bone marrow stem cells isolated using the CD34/ckit+/Sca-1+/lin phenotype. Bcrp1 expression was therefore proposed to be the molecular determinant of the SP phenotype. The overexpression of Bcrp1 in normal bone marrow cells was found to increase the proportion of SP cells in a haematopoietic culture system while at the same time causing a reduction in the numbers of maturing progeny [78]. Furthermore, the bone marrow side population was found to be lost in Bcrp1 knockout mice. These animals retained normal numbers of haematopoietic stem cells, however, they were more sensitive to DNA damaging drugs such as mitoxantrone [77]. It was therefore suggested that Bcrp1 might have a role in protecting stem cells from DNA damage.

The most significant implication of identifying Bcrp1 as the determinant of the stem cell phenotype was that its expression, and therefore the SP phenomenon itself, might be a universal stem cell marker. Having previously been identified in murine skeletal muscle and neural cell preparations [79, 80], SP was now identified in murine embryonic stem (ES) cells and rhesus monkey bone marrow cells [78]. With the exception of the neural cell preparations, the SPs of each of these cells were shown to express high levels of Bcrp1. However, the expression of Bcrp1 in the non-SP population of the ES cells suggested that Bcrp1 expression is not sufficient for the SP phenotype to be observed in some cell types. Regardless of this, the fractionation of ES cells into SP and non-SP and their subsequent injection into murine blastocysts showed the SP fraction to produce a much greater level of chimerism (70–80% by coat color) than the non-SP cells (1–5%) [78]. So, not only was it apparent that Bcrp1 was the molecular determinant of the SP phenotype (in most cells at least), it also appeared that the SP may provide a means for the identification of stem cell populations from various sources, including the mammary epithelium, without the need for cell-type-specific markers.

The presence of a SP was reported in human breast epithelium preparations by us [81] and in mouse mammary epithelium by us and others [29, 81]. Our initial study showed that the SP composed 0.45 ± 11% (n = 17) of mouse mammary cells. Using RT-PCR we found mouse mammary SP cells to have high levels of Bcrp1 expression and when placed under mouse mammary epithelial clonal conditions they grew as typical cytokeratin-expressing mammary epithelial cell clones [81, 82]. Thus, the mouse mammary SP cells did contain mammary epithelial cells with in vitro progenitor capacity.

To address the in vivo developmental potential of mouse mammary SP cells and the question of whether they were genuinely enriched for mammary epithelial stem cells, as the studies of Zhou et al. would predict [78], we used limited cell dilution cleared fat pad transplantation. We did not use competitive repopulation [29], but rather opted for transplantation of pure SP cells freshly isolated from the mouse mammary gland. This approach eliminated the possibility of changes in Bcrp expression (induced by placing them in culture) but also reduced the numbers of cells available and did not allow for interactions with potential niche cells. The viability of the SP cells isolated for transplant was >70% and enough cells could be isolated without the use of an intervening culture period for transplantation of 37 cleared fat pads with 2,000–5,000 freshly isolated pure SP cells per fat pad. Similar numbers of non-SP cells were used in 25 fat pad transplantations. After 5–8 weeks, to allow for ductal outgrowth, animals were mated to stimulate lobuloalveolar development and the fat pads were then wholemounted. Outgrowths were observed in only five of the 37 SP transplants and, of these, four showed lobuloalveolar development only (although these structures did contain both myoepithelial and luminal epithelial cells). The remaining successful transplant showed both ductal and lobuloalveolar development. The non-SP transplants had outgrowths in six of the 25 transplants carried out [81]. Therefore, this study provided evidence that SP cells might be lobuloalveolar progenitors. Recent data have demonstrated that Bcrp1 is expressed and functional in the mature secretory alveolar luminal epithelium of the lactating gland [83] and a novel estrogen response element has been identified in the human BCRP promoter [84]. Together, these data raise the possibility that epithelial SP cells of the virgin gland are either (a) derived from alveolar epithelial cells transiently forming during the estrous cycle or (b) alveolar stem/progenitor cells. Either way, we suggest that Bcrp1 may be a lineage marker of cells destined to form the secretory alveolar epithelium.

Using time-course studies as part of our investigations of the SP phenotype, we proposed a new criterion for defining the SP, which would enable data from different laboratories to be more comparable [82]. We observed that the SP region is simply the region in which the red and blue fluorescent emissions of the DNA-bound Hoechst 33342 dye are in proportion to each other. Increasing time and dye loading leads to red fluorescence emissions exceeding the blue and the formation of the non-SP region. This fluorescence shift is the result of an increasing concentration of DNA-bound Hoechst 33342 molecules [85]. Thus, the Hoechst 33342-staining profile is a function of dye loading. The rate of dye loading is slowed by the activity of the efflux pump Bcrp1. The time-course experiments clearly showed that taking a sample at a single incubation point does not indicate whether the population of cells in the SP region includes large numbers of cells that will shortly be moving into the non-SP region. Therefore, we proposed that a new objective criterion of the ‘stabilized SP percentage’ should be established and defined this as ‘the percentage of cells in the SP region at the time point at which the change in SP percentage between that time point and the time point preceding it is 0.1% or less.’

CD24 and Basal Stem Cells

As has been discussed, the direct identification of mammary gland stem cells requires appropriate markers for the separation of the luminal epithelial and myoepithelial/basal cell populations into their component subpopulations. Such separation is necessary in order to accurately assay populations of putative mammary stem cells by cleared fat pad transplantation. We have recently refined our earlier cell separation methods using the marker CD24 and, along with two other groups, have now demonstrated that it is the basal epithelial cell layer that is most highly enriched for transplantable mammary stem cells.

CD24 generated interest as one of the markers used to identify human breast cancer stem cells [5]. It consists of a small, heavily glycosylated protein core of 27 amino acids attached to the cell membrane by a phosphatidylinositol anchor [86]. First identified as the marker of B cells, it was later found to be strongly expressed in neutrophils but not in normal T cells or monocytes. In addition, CD24 expression has been observed in various haematologic malignancies and solid tumors including lung cancer, neuroblastoma, rhabdomyosarcoma, and renal cell carcinoma [87, 88]. Recently it has been reported to be expressed in tumors of the ovary, prostate and breast [8991]. CD24 serves as a specific ligand for P-selectin (CD62P), with an important role in the initial rolling of leukocytes on activated endothelial cells or when leukocytes are attached to activated platelets [92, 93]. It may also modulate integrin function [94]. CD24 expression in non small cell lung cancer [95], breast cancer [91] and prostate cancer [90] is closely related to tumor metastasis, the survival rate of patients, and the rate of tumor recurrence.

To explore the potential of the CD24 marker for isolating subpopulations of normal epithelial cells from the adult virgin mouse mammary gland, we used flow cytometry to investigate CD24 expression in mouse mammary cell preparations [44]. Staining with CD24 revealed three distinct cell populations: CD24High, CD24Low and CD24Negative (Fig. 2). Analysis of cytoskeletal antigen staining and of gene expression patterns demonstrated that these populations represented luminal epithelial, basal/myoepithelial and non-epithelial cells, respectively. Remarkably, mammary fat pad repopulation assays using freshly isolated cells revealed that the CD24Low, basal population was most strongly enriched for stem/progenitor activity, both in terms of the numbers of outgrowths generated and the extent to which they filled the fat pad. However, a limited transplantation capacity was observed in the CD24High population, with a few outgrowths of limited extent being generated by this population. Interestingly, outgrowths from both the CD24Low and CD24High, no matter how extensive, contained luminal epithelial and basal epithelial layers [44].
Fig. 2

CD24 staining pattern of mouse mammary cells. Freshly isolated primary mouse mammary cell suspensions are stained and sorted on the basis of light scatter, using TO-PRO-3 dye exclusion to remove dead cells. Live cells are next gated to remove contaminating CD45+ lymphocytes and the anti-CD24 staining pattern of the remaining cells is compared to a control sample stained with a non-specific isotype control antibody. From this comparison, CD24Negative (non-epithelial), CD24Low (basal/myoepithelial) and CD24High (luminal epithelial) populations are gated

At the same time as this work was published, the transplantation of freshly isolated single mouse mammary epithelial cells was reported [27, 28]. Stingl et al. [28] reported the phenotype of these ‘mammary repopulating units’ (MRUs) as being CD24Medium CD49fHigh [28]. The CD24Medium population of Stingl et al. is equivalent of our CD24Low. CD49f (also called α6 integrin) is a basal epithelial cell marker. Shackleton et al. reported that mammary stem cells were CD24+ CD29High [27], and did not differentiate between degrees of CD24 expression. Again, CD29 (β1 integrin) is a basal epithelial cell marker. Thus, all three reports have demonstrated that it is the basal cell compartment that is enriched for mammary epithelial cells with the ability to repopulate a cleared fat pad. How then, do we reconcile the direct evidence that mammary stem cells are basally located with our previous data pointing at pluripotent capacity for some luminal epithelial cells? The most likely explanation is that basal stem cells generate a transit amplifying or progenitor compartment which is located in the luminal compartment and which is characterized by high in vitro cloning efficiency and dual lineage differentiation capacity.

Estrogen Receptor Positive and Negative Cells of the Mammary Epithelium

Our current work is focused on understanding the relationship between the basal stem cells and their differentiated daughters. In particular, we have addressed the identity of estrogen receptor expressing (ER+) and estrogen receptor negative (ER) luminal cells in the mammary epithelium and their relationship with the stem cell compartment.

The importance of estrogen in normal breast development as well as in mammary carcinogenesis is well documented [9698] and there is strong evidence that the cumulative exposure to this mitogen significantly influences the lifetime risk of developing breast cancer. The relationship between mammary stem cells and hormone receptor expressing cells is, therefore, a fundamental issue in breast biology. Only a subset of cells in the developing and adult gland express the estrogen receptor (ER) but there are conflicting views as to the role of these cells. They are seen by some as a stem/progenitor compartment [99102] while others suggest that ER+ cells stimulate proliferation of the stem cell compartment in a paracrine manner [103].

To resolve this issue we have now prospectively isolated and functionally analyzed ER+ and ER luminal epithelial cells from freshly prepared mouse mammary tissue and shown that they form distinct cellular populations. We have shown that ER+ cells of the mouse mammary gland are defined by the phenotype CD24High/Sca-1+/Prominin-1+ (Prominin-1 is the mouse homologue of CD133). Besides the estrogen receptor, these cells express the progesterone and prolactin receptors. They also express luminal cytokeratins at higher levels than the ER luminal cells. The latter population, which are defined as CD24High/Sca-1/Prominin-1, were enriched for expression of milk proteins (lactotransferrin and β-casein), even in the virgin animal, and contain the highest number of in vitro colony forming cells. Cleared fat pad transplantation of these luminal populations and the basal/myoepithelial cells shows that the CD24Low/Sca-1/Prominin-1 basal/myoepithelial cells are most highly enriched for cells with fat pad transplantation capacity, consistent with previous data [27, 28, 44]. Both the ER luminal compartment and the ER+ luminal compartment contain very few transplantable cells, directly demonstrating that ER+ cells are not stem cells in the mouse mammary gland [104]. Rather, the data are consistent with the CD24High/Sca-1+/Prominin-1+ population forming a differentiated hormone sensing compartment which stimulates proliferation of basal stem cells and luminal transit amplifying or progenitor cells in a paracrine manner.

We have also been able to confirm that the transplantable cells can be further purified from the basal/myoepithelial compartment by their expression of CD49f [104] as described [28]. We can now simultaneously prospectively isolate ER+ and ER luminal epithelial cells, myoepithelial cells and epithelial stem cells (while excluding non-epithelial cells) from the mouse mammary gland for analysis.

Future Questions/directions

The aim of this work is to identify ways of killing breast cancer stem cells in a clinical setting. To do this, an understanding of gene expression patterns within normal mammary stem, progenitor and differentiated cells is required so that potential drug targets can be identified. It is also important to be able to identify tumour stem cells so that they can be compared with normal stem cells and allow the presence of our potential drug targets within the tumour stem cells to be confirmed. We are actively pursuing these areas.

From a purely developmental biology point of view, however, understanding the relationship between the stem cells and their differentiated daughters is also of interest. For that reason, we are undertaking lineage analysis of the mammary epithelium to address this fundamental biological question.

Over the last 10 years, there has been a huge increase in both the interest in this field and in the reagents and technologies available for its study. We can now look forward to making the sorts of advances in our understanding of mammary stem cell biology, which have until now, been principally available only to the haematopoietic stem cell field.


The authors would like to thank all those who contributed to or advised on the studies described above, in particular Katherine Sleeman, Azra Alvi, Mike O’Hare, Trevor Dale, Tariq Enver, Clare Isacke and Alan Ashworth.

Work in the laboratory is supported by Breakthrough Breast Cancer.

Copyright information

© Humana Press Inc. 2007