Author contributions: P.E.: conception and design, collection of data, data analysis and interpretation, and manuscript writing; N.K. and F.V.: collection of data and data analysis and interpretation; D.J.H.F.K.: data analysis and interpretation; G.J.L. and J.E.V.: conception and design and manuscript writing; J.T.E.: provision of study material; C.J.E.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS November 30, 2011.
Elevated aldehyde dehydrogenase (ALDH) expression/activity has been identified as an important biomarker of primitive cells in various normal and malignant human tissues. Here we examined the level and type of ALDH expression and activity in different subsets of phenotypically and functionally defined normal human mammary cells. We find that the most primitive human mammary stem and progenitor cell types with bilineage differentiation potential show low ALDH activity but undergo a marked, selective, and transient upregulation of ALDH activity at the point of commitment to the luminal lineage. This mirrors a corresponding change in transcripts and protein levels of ALDH1A3, an enzyme involved in retinoic acid synthesis and the most highly expressed ALDH gene in normal human mammary tissue. In contrast, ALDH1A1 is expressed at low levels in all mammary epithelial cells. These findings raise interesting questions about the reported association of ALDH activity with breast cancer stem cells and breast cancer prognosis. STEM CELLS 2012; 30:344–348.
Much attention is now being focused on elucidating the molecular mechanisms that regulate the different early stages of normal mammary cell differentiation, with the goal of identifying those relevant to the cells that maintain human breast cancers or that may serve as useful markers of these cells. The aldehyde dehydrogenases (ALDHs), which are encoded by a large gene family (http://www.aldh.org), are of interest in this regard , although the results of analyses of their expression in normal and neoplastic human breast cells remain controversial (reviewed in ). Here we analyzed the ALDH activity and isoform expression in different phenotypically and functionally defined subsets of normal human mammary cells. We show that ALDH activity in the normal human mammary gland is largely due to ALDH1A3 and is low or absent in normal human mammary stem cells but is then elevated when these transition into progenitor cells committed to the luminal lineage.
MATERIALS AND METHODS
These are detailed in Supporting Information.
We first analyzed the ALDH activity present in each of the five biologically and phenotypically distinct populations that are found in dissociated normal human mammary tissue and can be separately isolated based on their differential expression of CD49f, epithelial cell adhesion molecule (EpCAM, also known as CD326), CD31, and CD45 (Fig. 1A) [3–5]. Three of these populations are subsets of mammary epithelial cells: (a) a basal (CD49f+EpCAM−/lowCD31−CD45−) fraction that contains all of the mammary stem cells, all of the bilineage- and myoepithelial-restricted progenitors, and all of the mature myoepithelial cells; (b) a primitive luminal (or “luminal precursor-enriched”) (CD49f+EpCAM+CD31−CD45−) fraction that contains all of the luminal-restricted progenitors and some of the more differentiated luminal cells, and (c) a mature luminal (CD49f−EpCAM+CD31−CD45−) fraction that consists of fully differentiated luminal cells. Because mammary stem cells cannot be identified directly, they are referred to as “mammary repopulating units” (MRUs) based on their ability to individually regenerate a full mammary gland structure within 4–6 weeks when transplanted into an appropriate environment in immunodeficient mice [3, 4]. Mammary progenitors are likewise referred to operationally as mammary “colony-forming cells” (CFCs), based on their ability to produce colonies within 8–10 days in vitro that contain differentiated adherent cells of one or both mammary lineages. CD49f/EpCAM/CD31/CD45 staining also yields another two populations of cells that are present in high numbers in normal breast tissue but are not epithelial, that is, a stromal (CD49f−EpCAM−CD31−CD45−) fraction and a fraction that contains a mixture of hematopoietic (CD45+) and endothelial (CD31+) cells.
We incubated normal human mammary cells with Aldefluor (a fluorescent reagent that is activated by ALDH to produce an intracellularly retained fluorescent derivative, bodipy aminoacetate [BAA]) and then with fluorescently conjugated antibodies. Comparison of the resulting fluorescence profiles obtained in the presence and absence of diethylaminobenzaldehyde (DEAB), a specific inhibitor of ALDH enzyme activity, showed that the cells in all fractions had DEAB-sensitive ALDH activity, as indicated by their decreased fluorescence when DEAB was present (Fig. 1B, Supporting Information Fig. 1). However, in all tissue samples examined (four fresh and seven previously cryopreserved), we found the median fluorescence intensity (MFI) to be much higher (on average, 18-fold, p < .001) in the primitive luminal fraction when compared with the basal fraction (Fig. 1C) and this difference was sustained when CD10 was used as an additional marker to discriminate the luminal progenitor-enriched and basal fractions [6, 7] (with a 17-fold higher MFI in CD49f+EpCAM+CD10− population on average when compared with the CD49f+EpCAM−/lowCD10+ cells; Supporting Information Fig. 2). The mature luminal cells exhibited intermediate activity (on average, fourfold higher MFI than the basal cells, p < .001, Fig. 1C). In the same cell suspensions, the MFI of the stromal cell fraction was also much higher (on average, 13-fold, p < .001) than the MFI of the basal fraction. Thus, the elevated ALDH activity within bulk populations of human mammary cells is determined predominantly by the prevalence of primitive luminal cells and stromal cells, both typically present in high numbers. As a result, even after fractionation, minor contamination of the more primitive basal fraction with either of these subpopulations would give false-positive readings.
Examination of two published global gene expression datasets [5, 8] for these subpopulations of cells that used microarray or LongSAGE technologies suggested that ALDH1A3 is the most prominently expressed ALDH of all the members of this gene family in the normal human breast and, in both datasets, ALDH1A3 expression is highest in the CD49f+EpCAM+ (or CD24+) primitive luminal populations (Supporting Information Fig. 3A). We also compared these datasets with another gene expression microarray dataset generated on similarly fractionated cells obtained from 3-day cultures of dissociated primary mammoplasty tissue cells . The results were largely confirmatory, ALDH1A3 remaining as one of the most highly expressed ALDH family members in these cultured cells, with highest ALDH1A3 expression again in the CD49f+EpCAM+(CD133/Muc1)+ luminal progenitor-enriched fraction (Supporting Information Fig. 3B). However, some notable differences were seen (e.g., marked upregulation of ALDH1B1, 3B1, 3B2, 5A1, and 16A1 expression in the cultured vs. freshly isolated mammary cells), highlighting the likelihood that mammary cells maintained in culture may not retain all of the phenotypic properties they possess in vivo.
To obtain a more accurate assessment of the changes in expression of ALDH1A3 during the differentiation of human mammary stem cells in vivo, we performed quantitative reverse trancriptase polymerase chain reaction and Western blotting analyses for ALDH1A3 transcripts and protein on cells isolated from three normal breast tissue samples. The results (Fig. 2A, 2B) consistently paralleled the ALDH activity profiles determined by Aldefluor staining (Fig. 1B, 1C). In contrast, the ALDH1A1 transcript and protein levels were consistently low in all epithelial subsets (Fig. 2A, 2B). Interestingly, we found that the stromal cells present in mammary tissue express high levels of ALDH1A1 transcripts and protein compared with any of the subsets of mammary epithelial cells and low levels of ALDH1A3 compared with the luminal subsets (Fig. 2A, 2B).
To determine directly the ALDH activity of CFCs and MRUs, we subjected cells dissociated from another eight normal breast tissue samples to the same surface marker- and Aldefluor-based fractionation strategy (Fig. 3A, 3B, left and middle panels) and then further subdivided the cells as follows. In the first set of experiments, all of the three mammary epithelial subsets plus the stromal cells were separated as a single group into an ALDH+ and ALDH− fraction using a single stringent gate to discriminate between DEAB-sensitive and DEAB-insensitive BAA fluorescence. In the second set of experiments, we subdivided the individually defined primitive luminal and basal subsets (only) into low, medium, and highly BAA-fluorescent fractions. The cells from each of the eight fractions from both designs (2 + 6) were then isolated by fluorescence-activated cell sorting (FACS) and assayed for CFC activity in vitro and MRU activity in vivo (the latter using a subrenal capsule xenotransplantation assay that facilitates sensitive quantitative measurements of human MRU frequencies [4, 9]). The results we obtained from the first experiments showed that more than 90% of the MRUs as well as the bilineage and myoepithelial CFCs are in the ALDH− (BAAlow-med) fraction, whereas most (76 ± 10%) of the luminal CFCs are in the ALDH+ (BAAhigh) fraction (right panel of Fig. 3A; Supporting Information Table 1). The results of the second experiments confirmed the segregation of luminal versus basal CFCs and MRUs in the EpCAM+ and EpCAM− subsets of CD49f+ cells, respectively (right panel of Fig. 3B; Supporting Information Table 2). In addition, they showed the luminal CFCs to be split between the three ALDH activity fractions in a ratio of 5:62:33 (BAAlow:BAAmed:BAAhigh), with a reverse distribution (75:24:1 and 81:18:1) for the myoepithelial plus bilineage CFCs and the MRUs, respectively (right panel of Fig. 3B; Supporting Information Table 2).
These results establish that the most primitive normal human mammary epithelial cell types defined by their bilineage developmental potential are characterized by low ALDH activity, and that this activity is first and selectively increased by more than an order of magnitude at the time of commitment to the luminal lineage before further proliferative ability is lost. A similar pattern occurs in bovine mammary tissue where the stem cells are detected using the same in vivo transplant assay as used here for human MRUs . The present findings are also in agreement with those found in the mouse mammary gland where ALDH1A3 expression has been found to be upregulated at the point of luminal progenitor generation . ALDH1A3 transcripts are also detected at high levels in luminal progenitor-enriched fractions obtained from short-term adherent cultures of freshly isolated normal human mammary cells ; however, a more detailed study of different in vitro systems is required to investigate the stability of this phenotype.
ALDH1A3 plays a catalytic role in the biosynthesis of retinoic acid in mammary cells  and approximately two-thirds of human genes with binding sites for Gata-3 (a transcription factor that regulates the luminal differentiation program ) contain retinoic acid regulatory elements . Activation of retinoic acid signaling (using all-trans retinoic acid) in breast cancer cell lines was reported to result in a reduced frequency of cells with mammosphere-forming activity, and blocking of ALDH activity (using DEAB) had the reverse effect . These findings encourage future investigation of the possibility that ALDH activity plays a role in activating the luminal cell differentiation program in the mammary gland through retinoid signaling pathways.
The high ALDH activity reported for cells with breast cancer-initiating properties in xenograft assays [16–18] stands in direct contrast with the results for normal mammary stem cells and correlates better with normal luminal progenitors. This adds further weight to the possibility that cells with breast cancer stem cell activity in clinical samples may be more closely related to luminal progenitor cells. A case for this model has recently been made in BRCA1-associated breast cancer, in which aberrant luminal progenitor activity has been described in human breast tissue that is still phenotypically normal . It is noteworthy that more ALDH+ cells have been reported in breast tissue from BRCA1 mutation carriers , where they appear to have a luminal rather than a basal location. Alternatively, it may be that in certain breast cancers, the oncogenic process itself perturbs the control of ALDH expression and/or activity. Examples of the latter have been demonstrated in leukemia [20–23], suggesting that the picture may be equally heterogeneous in breast cancer.
Our study demonstrates that elevated ALDH activity characterizes luminal progenitor cells in the normal human mammary epithelium.
We acknowledge the excellent technical contributions of D. Wilkinson, G. Edin, and the staff of the Flow Cytometry Facilities of the Terry Fox Laboratory and the Walter and Eliza Hall Institute. Mammoplasty tissue, generously donated by patients, was obtained with the assistance of Drs. J. Sproul, P. Lennox, N. Van Laeken, and R. Warren (Canada) and the Victorian Cancer Biobank (Australia). This work was supported by the Canadian Breast Cancer Research Alliance (Grant CBCRA 019343), the U.S. Department of Defense Breast Cancer Research Program (Predoctoral Fellowship number W81XWH-06-1-0702), the Canadian Breast Cancer Foundation BC/Yukon (Fellowship to N. Kannan), the National Health and Medical Research Council (Australia, Grants #461224 and #461221), the Victorian Government through the Victorian Cancer Agency/Victorian Breast Cancer Research Consortium and an Operational Infrastructure Support grant, and the Australian Cancer Research Foundation.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.