Cellular pathways that contribute to adult human mammary gland architecture and lineages have not been previously described. In this study, we identify a candidate stem cell niche in ducts and zones containing progenitor cells in lobules. Putative stem cells residing in ducts were essentially quiescent, whereas the progenitor cells in the lobules were more likely to be actively dividing. Cells from ducts and lobules collected under the microscope were functionally characterized by colony formation on tissue culture plastic, mammosphere formation in suspension culture, and morphogenesis in laminin-rich extracellular matrix gels. Staining for the lineage markers keratins K14 and K19 further revealed multipotent cells in the stem cell zone and three lineage-restricted cell types outside this zone. Multiparameter cell sorting and functional characterization with reference to anatomical sites in situ confirmed this pattern. The proposal that the four cell types are indeed constituents of an as of yet undescribed stem cell hierarchy was assessed in long-term cultures in which senescence was bypassed. These findings identify an adult human breast ductal stem cell activity and its earliest descendants.
The mammary gland is a highly dynamic organ that undergoes a series of changes from intrauterine life to senescence. In humans, growth in adulthood commences at puberty where the parenchymal cells branch from a few blunt ending primary and secondary ducts into an elaborate tree with multiple terminal ducts and lobules. With each menstrual cycle, breast proliferation fluctuates, and, in the luteal phase, the growth fraction can become as high as 35% (Potten et al., 1988; Shetty et al., 2005). Accordingly, during pregnancy, there is both a 10-fold increase in the number of alveoli per lobule as well as de novo formation of lobules by lateral budding from existing terminal ductules, leaving behind a small amount of connective tissue space (for review see Russo and Russo, 2004). These cellular dynamics led Taylor-Papadimitriou et al. (1983) to postulate the existence of a population of precursor cells in the adult human breast that are capable of giving rise to new lobules.
From studies mainly in other species, it is known that adult stem cells are generally focal in distribution and not necessarily colocalized with the bulk of transiently amplifying cells (for review see Fuchs et al., 2004). In mice, the location of immature mammary gland stem cells was narrowed down to the peripheral cap cells of the terminal end buds (for review see Woodward et al., 2005). In humans, however, in which end buds are not prominent structures (Howard and Gusterson, 2000), the identification of a candidate stem cell zone has had to rely on a detailed and microscopically directed sampling of well-defined segments of the organ followed by functional assays. By combining microdissection with colony-forming ability, candidate stem cells in the human hair follicle were prospectively identified in the bulge region more than a decade before the bulge was unequivocally proven to be the epithelial stem cell niche in the skin (Yang et al., 1993; Moll, 1995; Tumbar et al., 2004). The proximal ducts of the prostate were also shown to harbor stem cells by this method (Tsujimura et al., 2002). Recently, a side population exhibiting Hoechst dye efflux properties was isolated from the human mammary gland (for review see Smalley and Clarke, 2005), and self-renewing cells were enriched for by the use of nonadherent mammosphere cultures (Dontu et al., 2003). Whereas in mice, the ultimate evidence for the existence of mammary stem cells is the clonal repopulating ability and greater morphogenic capacity within the cleared fat pad (Stingl et al., 2006a; Shackleton et al., 2006), such experiments cannot be performed in humans. Fortunately, surrogate assays conducted with mice and human cells demonstrate that putative mammary stem cells in 3D laminin-rich ECM (lrECM) gels function as they do in vivo in terms of several morphogenetic criteria across the species (Gudjonsson et al., 2002b; Dontu et al., 2003; Clarke et al., 2005; Stingl et al., 2006a). However, the existence of a hierarchy as well as a correspondence between stem cell markers and activity with specific regions of the gland has not been described.
In this study, we examine in situ whether the resting human mammary gland exhibits stem cell markers, which could identify a stem cell zone, as has been done for other tissues (Blanpain et al., 2004; Trosko et al., 2004; Burger et al., 2005; Kim et al., 2005; Moore and Lemischka, 2006). We took advantage of Ki-67 and laminin-2/4 as biomarkers for cellular turnover and differentiation (Simon-Assmann et al., 1994; Belair et al., 1997; Feuerhake et al., 2000; Fleischmajer et al., 2000; Maslov et al., 2004; Kingsbury et al., 2005; Shetty et al., 2005). Further dissection of regions of interest and microcollection allowed us to interrogate functional stem cell properties in culture. These included high clonal proliferative capacity on tissue culture plastic, self-renewal in suspension, and mammary morphogenesis in 3D lrECM (Petersen et al., 1992; Hudson et al., 2000; Gudjonsson et al., 2002b; Tsujimura et al., 2002; Dontu et al., 2003; Benitah et al., 2005; Kim et al., 2005; Ohyama et al., 2006; Stingl et al., 2006a).
Multipotency was determined based on fluorescence imaging in situ and in clonal primary cultures using two lineage markers, keratins K14 and K19, in which K14 marks the myoepithelium and K19 is hypothesized to be a neutral switch keratin that permits the changeover of one type of cytoskeleton to the other (Stasiak et al., 1989). Importantly, double-positive K14 and K19 transitional cells are known also to codistribute with the stem cell zone in the prostate (Hudson et al., 2001).
To compare mouse and human markers, we isolated mammary stem cell activity on the basis of surface markers CD49f and EpCAM (Stingl et al., 2006b). The cells thus isolated were further characterized using additional putative stem and progenitor markers. Finally, we took advantage of human papilloma virus (HPV) 16 E6/E7 to bypass cellular senescence (Band et al., 1990; Gudjonsson et al., 2002b) and to interrupt the normal differentiation of stem and progenitor cells to generate cell lines that maintain their phenotypes in culture. The absence of senescence allowed us to follow the developmental hierarchy of the progenitor cells in the breast and made it possible to isolate stable cell lines representing the different stem and progenitor stages, as done previously for other cell types (Reznikoff et al., 1994; Roecklein and Torok-Storb, 1995; Okamoto et al., 2002; Osyczka et al., 2002; Akimov et al., 2005; Mori et al., 2005). Our results have defined a human breast epithelial stem cell zone in vivo and a progenitor hierarchy both inside and outside this zone. Additionally, we have developed several stem and progenitor cell lines that will aid our understanding of the possible role of these cells in breast cancer.
A candidate stem cell zone resides in ducts
To test whether a stem cell zone could be defined in the human breast, we stained for several postulated surrogate stem cell markers, including stage-specific embryonal antigen-4 (SSEA-4), keratin K6a, keratin K15, keratin K5, Bcl-2, and chondroitin sulfate (Bocker et al., 2002; Legg et al., 2003; Luna-More et al., 2004; Dravida et al., 2005; Schmelz et al., 2005; Ohyama et al., 2006). Three of the markers examined, SSEA-4 (Fig. 1), K15 (Fig. 1) and K6a (see Fig. 5 a), localized focally to discrete clusters of cells in ducts, including terminal ducts, and were essentially absent from lobules. Specifically, although 73–78% of ducts stained with K6a, SSEA-4, and K15, the frequency of lobules with stained cells ranged from 6 to 13% (n = 19). Keratin K5, Bcl-2, and chondroitin sulfate were more widely distributed but still were more prominent in ducts than lobules (Fig. S1 A). The frequency of SSEA-4–positive cells was further determined by FACS analysis of trypsinized uncultured organoids. Single cells identified as epithelial and myoepithelial by dual staining with epithelial antigen EpCAM and β4-integrin in the SSEA-4hi gate comprised 0.5–0.7% of the total population, which also included the stromal cells (Fig. S1 B). Importantly, SSEA-4hi cells were highly enriched for keratin K6a- and K15-positive cells (Fig. S1 B).
Because quiescence is a general property of stem cells in their niche, we stained ducts and lobules with Ki-67. The level of staining in lobules varied markedly between biopsies, as 6/12 had moderate to strong staining (up to 50% of cells per lobule), and the other six were negative. However, ducts showed a low but constant level of staining: a mean of 2.8% (range of 1.3–4.7%) of cells stained in 10 biopsies (Fig. 2, a and b). The Ki-67 staining pattern was confirmed by staining for minichromosome maintenance protein 2 (Gonzalez et al., 2003; unpublished data). Furthermore, lobules rather than ducts stained specifically for the α-2 chain of laminin-2/4, a basement membrane component surrounding proliferating epithelial cells (Fig. 2, c and d; Simon-Assmann et al., 1994; Chenard et al., 2000; Fleischmajer et al., 2000; Laine et al., 2004). RT-PCR for the different chains of laminin using purified populations of myoepithelial cells and fibroblasts revealed that whereas myoepithelial cells contributed α-1, α-3, and α-5 chains of laminin in addition to α-2 chains of laminin-2/4, fibroblasts expressed α-2 chains of laminin-2/4 only (Fig. S1 C). These data show that a candidate stem cell zone resides in ducts that are enriched in cells identified as being SSEA-4hi/K5+/ K6a+/K15+/Bcl-2+ cells, which are generally quiescent and are surrounded by chondroitin sulfate. The more frequently proliferating progenitors are found outside this region and are often surrounded by laminin-2/4.
Cells with a capacity for clonal growth, self-renewal, and bipotency are derived only from ducts
We assessed the proliferative and morphogenic capacity of primary cells on tissue culture plastic, in mammosphere suspension cultures, and in a 3D lrECM assay. The growth of the cells derived from ducts or lobules were compared on tissue culture plastic. Only ductal-derived cells formed colonies that were considerably larger than 100 cells/colony (60 ± 2% of colonies had >100 cells; n = 3 × 50 colonies) compared with lobular-derived cells (0% of colonies had >100 cells), indicating that as expected for stem cell activity in culture, the duct-derived cells have a higher proliferative potential than those derived from lobules (Fig. 3 A, a′–b′′). Similar patterns were obtained from two additional biopsies when placed in culture. To further test for self-renewal, ducts were compared with lobules using the mammosphere assay (Fig. 3 B). Whereas ducts gave rise to relatively large mammospheres (>70 μm), those derived from lobules were small and irregular (<70 μm; Fig. 3 B, a–b′). Primary mammospheres were then trypsinized and replated to derive secondary mammospheres. In our hands, the frequency of primary and secondary mammosphere formation was ∼3/1,000 cells, a figure comparable with that described originally by Dontu et al. (2003). We assessed the presence of prospective multipotent cells in mammospheres by multicolor imaging based on the combined staining for myoepithelial keratin K14 and the luminal or switch keratin K19 (Stasiak et al., 1989). The K14+/19+ phenotype was most often observed in duct-derived mammospheres (26/48 colonies; 54%) as opposed to lineage-restricted lobule-derived mammospheres (5/48 colonies; 10%; Fig. 3 B, a′′ and b′′). Similarly, mammospheres trypsinized and plated at clonal density from ducts and lobules gave rise to K19/K14 double-stained cells in 56% of those derived from ducts (n = 92 colonies) and only 6% of those derived from lobules (n = 92).
The morphogenic potential of mammospheres was assessed in a 3D lrECM assay in which 2/49 duct-derived second passage mammospheres developed into terminal duct lobular unit (TDLU)–like structures. The formation of TDLUs was not observed from lobule-derived mammospheres (0/45); these only gave rise to spherical structures (Fig. 3 B, a′′′ and b′′′). A similar pattern was observed with or without a feeder layer of primary epithelial cells. Essentially, this behavior was recorded in mammospheres derived from three randomly collected biopsies from reduction mammaplasty. Finally, immunostaining of plated mammospheres for SSEA-4 from a representative biopsy further revealed that 24% of duct-derived colonies contained positive cells compared with 10% of those derived from lobules. Thus, ducts are enriched for a subpopulation of epithelial cells that, in culture, exhibit a highly proliferative, self-renewing, and morphogenic capacity indicative of stem cell activity.
Keratin profiles reveal that K19+/K14+ cells reside in ducts
When tissue sections were double stained for keratins K19 and K14, four different populations could be distinguished: K19+/K14+ (+/+), K19+/K14− (+/−), K19−/K14− (−/−), and K19−/K14+ (−/+). The +/+ cells were found as scattered single cells or small groups of cells in ducts (Fig. 4, a and inset). The +/− cells were present both in ducts (not depicted) and in lobules (Fig. 4 b). The −/− cells were rare, but, when present, they appeared generally throughout an entire TDLU (Fig. 4 c), indicating their clonal origin. The −/+ cells were observed in the myoepithelial cell layer as expected (Fig. 4 d). These profiles reflect the unique subsets of breast epithelial cells that reside in the mammary gland, as was subsequently revealed in cloned primary cultures from crude collagenase digests (Fig. 4, a′–d′). Furthermore, this pattern was confirmed in microcollected ducts and lobules. Thus, whereas lobules gave rise almost exclusively to colonies of +/−, −/+, and a few −/− cells, only ducts also gave rise to colonies of +/+ cells (Fig. 4, e and f). The predominance of +/+ cells in cultures of ducts was observed in three experiments from two independent biopsies. Similar data were obtained after recloning. Thus, whereas cells from lobules either did not clone out or formed small abortive +/− clones, the ducts formed large +/+ clones (9/10 clones in an experiment from one biopsy; Fig. S2). We conclude that ducts house a subpopulation of epithelial cells that exhibit four important attributes of epithelial stem cells: they express stem cell markers in situ; they are slow cycling in vivo; they exhibit a high proliferative, self-renewal, and morphogenic capacity in culture; and they are bipotent.
Enrichment of K19+/K14+ cells in a Lin−CD49f+EpCAMhi population of ductal origin
To find the in situ equivalent of ductal/lobular cells isolated by FACS analysis, we defined a set of anatomical markers. Specifically, within the luminal lineage, we used keratin K6a for ductal cells and BCA-225 for lobular cells; within the myoepithelial lineage, we used keratin K17 for ductal cells and WT1 against lobular cells (Fig. 5). Before FACS analysis, we first removed the substantial stromal component of the human breast by depleting endothelial, fibroblastic, lymphocytic, and monocytic lineages using a CD31/1B10/CD34/CD45 immunomagnetic column, allowing a flow through of lineage-negative (Lin−) epithelial cells (Shackleton et al., 2006). Primary cells were then sorted based on their expression of CD49f (α6-integrin) and EpCAM (Stingl et al., 2006b). Four subpopulations, which were identified by gates I–IV, were sorted and then analyzed by immunofluorescence for the expression of anatomical selective markers (Fig. 6 and Fig. S3 C; see also Materials and methods). Similar FACS profiles were revealed in cells from six independent biopsies. Subpopulations isolated in gates I and II represented the luminal epithelial lineage, and gates III and IV contained the myoepithelial lineage (Fig. 6, a and b). A further subdivision into lobular cells (gates I and III) and ductal cells (gates II and IV) could be established by the anatomical markers (Fig. 6, b and c). As expected, the K19+/K14+ cells cosorted with ductal Lin−CD49f+EpCAMhi cells within the luminal epithelial lineage (gate II; Fig. 6, a and b). In addition, triple color staining showed that SSEA-4hi expression was observed in the +/+, Lin−CD49f+EpCAMhi population (Fig. S3).
Functional assays for stem cell activity demonstrated that colony-forming ability was significantly restricted to the Lin−CD49f+EpCAMhi population (gate II; Fig. 6 d), also suggesting the isolation of multipotent progenitor activity. Furthermore, single-cell sorting into 96-well dishes based on SSEA-4 staining within a CD49f/EpCAM context revealed that the SSEA-4hi subpopulation had a cloning efficiency of three colonies per 96-well dish as opposed to zero from the SSEA-4neg gate, which suggests that the SSEA-4hi subpopulation represents the majority of colony-forming activity. However, our experimental design so far does not allow us to account for the considerable variation in cloning efficiency between donor samples, so we cannot conclude definitively that the SSEA-4hi subpopulation is significantly different from the total Lin−CD49f+EpCAMhi population. Nevertheless, multicolor imaging revealed that all three clones were +/+ (Fig. S3). Most importantly, when cells from gates I–IV were cultured in 3D lrECM at clonal density, only the Lin−CD49f+EpCAMhi population (gate II) formed budding (TDLU-like) structures (gate II; Fig. 6, e and f). Thus, whereas the cells isolated from other gates morphologically consisted mainly of small spheres (acinus-like) or large, solid spheres as previously described (Fig. 6, e and f; Petersen et al., 1992) and were lineage restricted, those isolated from gate II were TDLU-like budding structures and formed +/+, +/−, and −/+ as revealed by staining the whole mounts of gels (Fig. 6 g). Collectively, analysis of primary breast tissue reveals that all detectable stem cell–like or stemlike activities are restricted to duct-derived cells.
Immortalization stabilizes the differentiation repertoire in accordance with a breast stem cell hierarchy
We reasoned that if the mammary epithelial cell types identified above by K14 and K19 expression were connected in a hierarchy, only one cell type should give rise to the other lineage-restricted progenitors. To examine this we used HPV16 E6/E7 (Band et al., 1990) to bypass senescence so cells could be followed for long periods of time. After transducing cells only once with retroviruses harboring E6/E7, we successfully recovered several sublines and clones of the four cell types that could be distinguished by K19/K14 expression as described in Fig. 4. We designated these cell lines +/+E6/E7, +/−E6/E7, −/−E6/E7, and −/+E6/E7 (Fig. 7). HPV16 E6/E7 conferred extended life span or immortality to all four subtypes, although the −/+ myoepithelial cells senesced after passage 30. +/+E6/E7 cells' unique ability to give rise to the other three subtypes (Fig. 7 a) and their ability to form TDLU-like structures in 3D lrECM cultures (Fig. 7 a′) confirmed that +/+E6/E7 cells exhibited stemlike activity. Quantification of morphogenesis using the 3D lrECM assay revealed 54 ± 2.5% (n = 3) of TDLU-like structures in cultures of +/+E6/E7 cells, whereas +/−E6/E7, −/−E6/E7, and −/+E6/E7 gave rise to cells of their own subtypes only (Fig. 7, b–d) and did not form TDLU in lrECM (Fig. 7, b′–d′). Relative to the other E6/E7 cell types, the multipotent +/+E6/E7 cells were unique in their combined expression of SSEA-4, keratin K15, Oct-4 (Tai et al., 2005), and Musashi-1 (Clarke et al., 2005) as well as by their capacity to differentiate into more restricted luminal-like and myoepithelial-like cells in the presence of serum. The latter were characterized by a comprehensive panel of lineage markers (Fig. S4, A–C; and Table S1). These data provide evidence that a stem cell zone in ducts marked by K19+/K14+ can give rise to K19+/K14−, K19−/K14−, and K19−/K14+ lineage-restricted progenitors.
HPV16 E6/E7 transduction reveals that lineage-restricted cells are progenitors
We asked two questions: whether the integration site of the viral genes affected the stability or the expression of the cells and whether lineage-restricted cells represented the end of their differentiation repertoire or whether they could specialize further. We performed a mass transduction of lineage-restricted flow-sorted mammary epithelial cells based on their expression of surface markers. As in other hierarchical tissues, specialization of the lineages in the human breast is defined by the acquisition of certain differentiation markers that are characteristic of luminal epithelial and myoepithelial cells. RT-PCR of five lineage-specific markers revealed that this general pattern carried through to all of the isolated cell lines (Fig. 8 A). Thus, whereas the stemlike cells (+/+E6/E7) had a low expression of three luminal and two myoepithelial markers, lineage-restricted progenitors (+/−E6/E7, −/−E6/E7, and −/+E6/E7) displayed a strong expression of these markers in a mutually exclusive manner (Fig. 8 A). Furthermore, staining of the lineage-restricted progenitors for surface markers revealed that they stained with either the luminal marker (MUC1) or the myoepithelial marker (Thy1). As such, we could use these markers in a large-scale cell sorting to examine the robustness of lineage maintenance after E6/E7 immortalization (acquired self-renewal). Primary cultures were sorted into luminal epithelial (+/− and −/−) and myoepithelial (−/+) lineages based on staining with MUC1 and Thy1 (Fig. 8 B; before E6/E7, crude). The profile of primary cultures was identical to that of uncultured cells (unpublished data). Further plating and cultivation in a second passage of sorted cells did not shift this phenotype appreciably (Fig. 8 B; before E6/E7 transduction, sorted). Sorted cells were transduced with HPV16 E6/E7 and selected in the presence of G418. More than 70 different clones emerged during selection, and these were pooled and reanalyzed by the same criteria. The general patterns of MUC1+/Thy1− and MUC1−/Thy1+ were sustained, albeit with a slight drifting in the population as a whole in spite of the different proviral insertion sites and the large number of pooled clones (Fig. 8 B; after E6/E7, pooled clones). This was confirmed directly by restriction site (RS) PCR and sequence analysis of chromosomal DNA in the unpooled +/−E6/E7 (MUC1+/Thy1−) and −/+E6/E7 (MUC1−/Thy1+) clones to identify individual integrated proviruses (Fig. 8 B, clonal analysis). These data demonstrate that lineage phenotype is essentially sustained irrespective of the retroviral integration site in these lineage-restricted pooled clones.
The fact that retroviral transduction also immortalizes the lineage-restricted cells suggested the possibility that these cells may be derived from the lineage-restricted progenitor cells. To show that this is the case, we searched for lineage-restricted markers in cells on the ductal-lobular junction. The search yielded new intralineage combinations of markers, demonstrating clearly that the three lineage-restricted cell lines were able to differentiate into additional cellular phenotypes within their respective lineages. Accordingly, +/− luminal progenitors can give rise to all combinations of the descendant cells expressing the luminal markers E29 and BCA-225 (Fig. 8 C, a). Similarly, −/− and −/+ cells give rise to descendants expressing combinations of either luminal markers K8 and CDw75 or myoepithelial markers WT1 and K17, respectively (Fig. 8 C, b and c). In all three cases, the cellular phenotypes were similar to their counterparts in situ (Fig. 8 C, a′–c′). These findings support the hypothesis that the lineage-restricted cells are progenitors rather than the ultimate end of their respective differentiation lineages. The data presented in this study provide the first evidence for the existence of a stem cell zone in mammary ducts marked by K19+/K14+, which can give rise to K19+/K14−, K19−/K14−, and K19−/K14+ lineage-restricted progenitors.
The existence of stem and progenitor cells in the human mammary gland has been widely postulated; however, until now, neither the location nor the candidate cellular entities have been definitively identified and functionally characterized. Previously, we had described two cell lines that were derived from the human breast using magnetic sorting and HPV16 E6/E7 immortalization (Gudjonsson et al., 2002b). One of the immortalized clones, MUC1−/ESA+(E6/E7), was able to generate itself as well as luminal epithelial and myoepithelial cells; it further expressed keratin K19 and formed TDLU-like structures inside a 3D lrECM. The other, MUC1+/ESA+(E6/E7), was lineage restricted, keratin K19 negative, and formed acinus-like spheres within the 3D lrECM. Although this was an important advance, several essential questions remained: did these virally transduced cells have identical counterparts in vivo, or were these a result of immortalization? If counterparts existed, where were they located in situ? Were there additional progenitor cells we had not been able to immortalize? Could we develop a procedure whereby one could isolate the untransduced counterparts reproducibly? Were there differences between ducts and lobules, and, intriguingly, was there a hierarchy?
In this study, we have answered all of the aforementioned questions: we have demonstrated the existence of four distinct human mammary epithelial cell types in situ, two of which most certainly are precursors to the two immortalized cell lines we had isolated previously using E6/E7. The cells are distributed in a stem cell zone in ducts and outside this zone in lobules. The size of the stem cell zone varies somewhat with the marker used, the most restricted and presumably most specific being the presence of keratin K15 and SSEA-4. Ducts and lobules were dissected, collected under the microscope, cloned, and characterized. We believe it is critical to meticulously dissect lobules distal to the intralobular terminal duct. Failure to do so could explain previously published data on the reported absence of a difference in growth patterns between ducts and lobules (O'Hare et al., 1991). Permanent cell lines were established, cloned, and characterized further with respect to the profiles defined in situ and in primary cultures. The procedures we have developed, while painstaking and time consuming by necessity because they are for normal human tissues, are nevertheless robust. We show that these four cell types are hierarchically connected such that only one cell type can give rise to all others, which themselves are lineage-restricted progenitors.
An important implication of these findings is that for the first time, a stem cell zone containing one or more mammary stem cells was identified in the human breast. Mouse mammary stem cells have recently been isolated prospectively based on various FACS strategies (Shackleton et al., 2006; Stingl et al., 2006a). However, so far, FACS profiles have not translated into the cell of origin or location in situ. In this study, by the use of anatomical markers, we demonstrate that the previously described gates for bipotent progenitors, colony-forming cells, and mammary repopulating units in mice (Shackleton et al., 2006; Stingl et al., 2006a) enrich specifically for cells of ductal origin in the human breast. Older studies had suggested that the mammary ducts of rodents may harbor such a stem cell niche. For instance, in the mammary gland from virgin mouse, candidate stem cells (stained with antibodies JB6 and JsE3) were found in ducts rather than in alveoli, and it was postulated that these served to regenerate ductal epithelium as well as forming new alveolar buds (Sonnenberg et al., 1986). This postulate was born out recently in an experiment with rudimentary ducts from postgestational mice transplanted to cleared fat pads of TGF-β1 transgenic mice, which retained the capacity to reactivate lobular structures at late pregnancy (Boulanger et al., 2005). Apparently, the mouse mammary gland ductal niche responds specifically to the MMTV–c-myc transgene by amplification of the stem cell compartment (Chepko et al., 2005). This implied strongly that in mice, an entire TDLU at any time represents the progeny from a single early ductal progenitor. Indeed, seminal studies of X chromosome inactivation in the human breast had demonstrated the presence of contiguous patches of normal mammary epithelium suggestive of being derived from single stem cells (Tsai et al., 1996). These patches were explained to be the result of some developmentally important long-term primitive stem cells, which were presumed to be estrogen receptor negative, as opposed to shorter term estrogen receptor–positive stem cells important for adult tissue homeostasis (Clarke, 2005). The hierarchy we describe here is essentially compatible with the aforementioned findings. However, it is important to recognize that the present identification of a stem cell zone in ducts does not necessarily exclude the existence of stem cells at other locations, although despite an extensive search, we have not observed such additional candidates as of yet.
The possibility of one stemlike cell population in an organ, which precedes all of the other stemlike cells during development, appears to be a general phenomenon because such multiple cell-type niches were also previously described for the hair follicle (Blanpain et al., 2004). The division of labor into that of the maintenance of tissue homeostasis and the more elaborate development or regenerative remodelling between stem cell compartments may also be a general phenomenon because in the skin, homeostasis is maintained exclusively by the epidermal proliferative unit, but wound repair depends on bulge cells from hair follicles (Ito et al., 2005). In analogy to hair follicles, we would propose that mammary tissue homeostasis is maintained by proliferative units in the TDLUs, whereas more elaborate structures, including the formation of new TDLUs from ductal alveolar buds, depend on the recruitment of cells from ducts.
Until recently, very few markers were assigned directly to a candidate mammary stem cell pool in humans or, indeed, even in rodents, but scattered single cells within the ducts of adult mice were shown to be positive for keratin K6a (Buono et al., 2006), a marker originally found in putative stem cells in the terminal end bud (Smith et al., 1990). In the present study, we found that the ductal stem cell zone was characterized by the accumulation of K19+/K14+ cells. If these traits were even indirectly related to stem cell activity, we would expect them to fluctuate with the size of the stem cell compartment. One of the most important pathways responsible for mammary stem cell activity is canonical Wnt signaling (Lindvall et al., 2006). Interestingly, Wnt1- or β-catenin–induced mammary hyperplasia and tumorigenesis in mice correlate with the accumulation of K6-positive cells even though K6 in itself does not appear to be essential for mammary gland development, at least in embryonic knockouts (Grimm et al., 2006). It has been shown that caveolin-1 deficiency in mice, which conveys mammary hyperplasia and tumorigenesis along with an increased stem cell activity, is characterized by the accumulation of keratin K6-positive cells (Sotgia et al., 2005). However, the mammary glands of mice that are unable to signal through the Notch pathway most closely mimic the profile of the human mammary stem cell zone (Buono et al., 2006; Wang et al., 2006). Under these conditions, there is a remarkable ductal accumulation of cells expressing luminal keratins coordinately with keratins K14 and K6 (Buono et al., 2006). A similar stem cell–related profile has been recorded in the prostate (Hudson et al., 2001; Wang et al., 2006). The appearance of +/+ cells in the suprabasal position has been interpreted as an accumulation of intermediate immature luminal cells (Buono et al., 2006). Thus, it cannot be excluded that the stem cell zone described here contains bona fide ductal basal cells, which, under the current available culture conditions, are not colony forming.
In mass cultures derived from reduction mammoplasty, we and others had shown previously that a subpopulation of the primary breast luminal epithelial compartment was capable of giving rise to both luminal epithelial and myoepithelial cells (Kao et al., 1995; Kang et al., 1997; Péchoux et al., 1999; Clarke et al., 2005). Furthermore, in culture and occasionally in suprabasal cells in the epithelial layer in situ, single cells could exhibit dual staining for these two cell types (Péchoux et al., 1999). Cells in this position were shown later to be K19 positive and could be immortalized occasionally in mass culture with HPV E6/E7 (Gudjonsson et al., 2002b). The discovery in the present study of a ductal stem cell zone in the human breast opens the possibility of the existence of a hierarchy among differentiated progeny at other locations. In the mammary gland, the ultimate level of differentiation is reached during lactation. Although we have not addressed the consequence of lactation on ductal differentiation, a previous study strongly suggested that ducts are indeed protected from hormone-induced differentiation (Bocker et al., 2002). This is in agreement with the presumed function of candidate stem cells in mature ducts of humans and rodents as a source of lateral branching during the formation of milk-producing lobules (Taylor-Papadimitriou et al., 1983; Cardiff and Wellings, 1999). Functionally, one of the hallmarks of stem cells is the ability to self-renew (for reviews see Watt, 1998; Mackenzie, 2006). A culture assay for self-renewal in human breast epithelial cells was successfully adopted from the field of neurobiology combined with the use of 3D laminin-rich gels (Petersen et al., 1992). Thus, cells prevented from attachment grew in suspension to form the so-called mammospheres (Dontu et al., 2003) reminiscent of the originally described neurospheres. Self-renewing stem cells were characterized by the ability to form new multipotent mammospheres even after passaging (Dontu et al., 2003). The other widely used culture assay of self-renewal in epithelial tissues is growth after plating at clonal density (for reviews see Watt, 1998; Mackenzie, 2006). We used both assays in the present study and showed that both mammospheres and multipotent clones could self-renew, although we do not yet know how the fraction of the self-renewing clones compare between the assays.
Our findings are pertinent to two poorly understood aspects of breast cancer evolution. For example, we know that breast cancers comprise at least two well-defined subtypes with distinct molecular profiles reminiscent of the luminal and basal lineages (Perou et al., 2000). This has led to a resurgence of speculation that breast cancers may arise in the different compartments within a stem cell hierarchy (Taylor-Papadimitriou et al., 1983; Rudland, 1987; Al-Hajj et al., 2003; Behbod and Rosen, 2005; Clarke, 2005). The hierarchy described here may guide the design of experiments to test the possible role of these progenitors in the origin of breast cancer subtypes, a possibility that is under investigation.
Materials And Methods
Breast tissue and cell cultures
Normal breast biopsies (n = 54) were obtained from patients undergoing reduction mammoplasty for cosmetic reasons. The use of human material has been reviewed by the Regional Scientific Ethical Committees for Copenhagen and Frederiksberg and approved with reference to (KF) (11) 263995.
Normal breast tissue was prepared as previously described (Rønnov-Jessen and Petersen, 1993). Upon collagenase treatment, epithelial organoids were either cultured as crude preparations or first manually separated under an inverted phase-contrast microscope (TMS-F; Nikon) into ducts and lobules before explantation (microcollected). Microcollected ducts and lobules were plated in collagen-coated (8 μg/cm2; Vitrogen100; Cohesion) T-25 flasks (Nunc) in the presence of chemically defined medium (CDM3; Petersen and van Deurs, 1988) and were allowed to spread for 8 d. Cells were trypsinized and cloned by limited dilution in the presence of serum-supplemented growth medium consisting of Ham's F12 medium (Invitrogen) supplemented with 2 mM glutamine, 50 mg/ml gentamycin (Biological Industries), 5% FCS (PAA Laboratories), 5 μg/ml insulin (Roche), 1 μg/ml hydrocortisone (Sigma-Aldrich), 0.1 μg/ml cholera toxin (Sigma-Aldrich), and 10 ng/ml EGF (PeproTech) at a density of 60 cells/cm2 (referred to below as F12 medium). After 7 d, some of the cultures were fixed in methanol and stained with hematoxylin. For secondary cloning experiments, cells were first plated at a density of 400 cells/cm2 and cultured for 12 d and then were subcloned at a density of 50 cells/cm2. Cultures were fixed and stained after 10 d, and the number of colonies was quantified. K19/K14 profiles were assessed by fluorescence in parallel cultures and in cloned primary cultures seeded at a density of 350 cells/cm2. The K19/K14 profiles of primary cultures derived from microcollected ducts or lobules were assessed at day 7 at densities of 800 cells/cm2 and in cultures of secondary clones at a density of 60 cells/cm2 at day 10.
For the nonadherent mammosphere assay, large ducts, terminal ducts (identified by connecting alveoli), and lobules were isolated and trypsinized for 10–15 min at 37°C on an orbital shaker to obtain a single cell suspension. Nonadherent mammosphere cultures were prepared as previously described (Dontu et al., 2003). In brief, cells were plated at a concentration of 5,000–20,000 cells/ml. The cultures were monitored for up to 12 d for the appearance of mammospheres. After 8 d, cultures were photographed, and structures derived from ducts (large and terminal) and lobules, respectively, were quantified and separated into two categories: >70 μm and <70 μm (n = 3 × 200 structures). For analysis of keratin expression, duct- and lobule-derived mammospheres were either smeared onto a glass slide and stained or trypsinized at day 9, plated at clonal density (200 cells/cm2), and propagated for 5 d in F12 medium before immunocytochemical staining. A total of 92 colonies from each segment were quantified using a fluorescence microscope (Dialux 20; Leitz) equipped with a 10× objective. Mammosphere populations derived from ducts and lobules were assessed for morphogenic potential by inoculation for 3 wk of each population in 300 μl lrECM (Matrigel; Becton Dickinson). Some cultures were conditioned by a feeder layer of primary human breast epithelial cells separated from the top gel by 200 μl of cell-free gel. The number of mammosphere-derived budding structures was assessed by phase-contrast microscopy.
To passage mammosphere cultures, mammospheres were collected, trypsinized for 10 min at 37°C while shaking, and single cells were plated at a concentration of 1,000 cells/ml (Dontu et al., 2003). Secondary mammospheres were sampled and plated in monolayer culture and analyzed for SSEA-4 expression by immunocytochemical staining (see supplemental Materials and methods). For RT-PCR analysis of laminins (see supplemental Materials and methods), myoepithelial cells and fibroblasts were purified and cultured as previously described before RNA extraction (Rønnov-Jessen and Petersen, 1993; Gudjonsson et al., 2002a).
Cell lines and clones
Cells representing K19+/K14+, K19+/K14−, K19−/K14−, and K19−/K14+ profiles were recovered in primary culture and were HPV16 E6/E7 transduced using a previously described protocol (Gudjonsson et al., 2002b). Fig. S4 D shows the culture history of each of the representative cell lines. For 3D culture, single cell suspensions of 105 cells were inoculated in lrECM as previously described (Petersen et al., 1992; Gudjonsson et al., 2002b) and were observed for a period of 12 d.
FACS analysis and cloning
For analysis of the robustness of transduced lineages, uncultured epithelial organoids or normal breast epithelial cells grown in primary monolayer cultures in chemically defined medium (CDM3; Petersen et al., 1990) were trypsinized and filtered through a 20-μm cell strainer (Miltenyi Biotec) and resuspended in Hepes buffer supplemented with 0.5% BSA (bovine fraction V; Sigma-Aldrich) and 2 mM EDTA (Merck), pH 7.5. The suspended cells were incubated for 30 min at 4°C in the presence of monoclonal primary antibodies recognizing MUC1 (CD227; 1:100) and Thy1 (CD90; 1:50). Control samples were incubated without primary antibody. Upon incubation, the cells were washed twice in Hepes/BSA/EDTA buffer and incubated for 15 min at 4°C with secondary isotype-specific fluorescent antibodies, AlexaFluor488 goat anti–mouse IgG2b, and AlexaFluor633 goat anti–mouse IgG1 antibodies (1:500; Invitrogen). After incubation, the cells were washed twice in Hepes/BSA/EDTA buffer and resuspended in a volume of 1 ml of buffer. Propidium iodide (Invitrogen) was added at a concentration of 1 μg/ml, and the cells were analyzed and sorted using a flow cytometer (FACSAria; BD Biosciences). The sorted populations were subsequently plated in collagen-coated T-25 culture flasks in chemically defined media (CDM6; Péchoux et al., 1999) for MUC1+/Thy1− cells and in CDM4 for MUC1−/Thy1+ cells. The sorted cell populations were transduced as previously described (Gudjonsson et al., 2002b). Untransduced sorted cell cultures were run in parallel as controls. Secondary cultures were analyzed for the expression of MUC1 and Thy1 by FACS analysis as described above.
To isolate putative stem cells within the lineage-negative epithelial cell population, breast organoids (n = 3 biopsy samples) were trypsinized for ∼10 min at 37°C under rotation. The solution was robustly agitated a few times during trypsination. Trypsination was stopped with FCS, and the cells were filtered through a 30-μm filter followed by filtration through a 10-μm filter to obtain a single-cell suspension. Cells were incubated with a cocktail of antibodies against stromal cells, which included CD31 (JC70A; 1:50), CD34 (QBEnd/10; 1:50), CD45 (Bra-55; 1:250), and fibroblast surface protein (1B10; 1:50). Cells were incubated at 4°C for 30 min. After incubation, cells were washed twice in Hepes/BSA/EDTA buffer and incubated for 15 min at 4°C with goat anti–mouse IgG and rat anti–mouse IgM microbeads (Miltenyi Biotec). Cells were then washed again twice and further applied to a column for immunomagnetic cell sorting (MACS; Miltenyi Biotec). The flowthrough from this column was collected and incubated with CD49f (GoH3; 1:500) and EpCAM (VU1D9; 1:100). A control solution was also prepared in which the primary antibodies were excluded. The cells were incubated at 4°C for 30 min followed by two washes and were further incubated with the secondary fluorescent antibodies AlexaFluor488 rabbit anti–rat IgG and AlexaFluor633 goat anti–mouse IgG1 (1:500; Invitrogen). Some degree of cross-reaction between CD49f and AlexaFluor633 was observed. For comparison with a directly conjugated EpCAM antibody, please see Fig. S3 C. The four FACS-sorted populations in gates I–IV were analyzed for luminal and myoepithelial markers as described below and were tested for self-renewal by limited dilution cloning in collagen-coated six-well plates (70 cells/cm2) in the presence of F12 medium. Morphogenic potential was analyzed in 24 wells in the presence of CDM3 by inoculation in lrECM at a density of 2 × 104 cells in 300 μl lrECM placed on top on a 200-μl cell-free gel solidified on top of a feeder layer of primary breast epithelial cells. The cultures were observed for 3 wk and assessed in triplicate for morphogenesis by phase-contrast microscopy in 50–80-μm spherical acinus-like structures, budding structures, and spherical colonies (>100 μm).
Immunohistochemistry and cytochemistry
Monolayer cultures and cells derived directly from collagenase digested tissue, cryostat sections of biopsies, mammospheres, or cell lines cultured in 3D lrECM were prepared and stained by immunoperoxidase or immunofluorescence as previously described (Petersen and van Deurs, 1988; Rønnov-Jessen et al., 1992; Gudjonsson et al., 2002b). For antibodies and further details, also see Table I.
The four CD49f/EpCAM cell populations were smeared onto glass slides and stained for BCA-255, K6a, K17, and WT1 by immunoperoxidase and for K19/K14 by fluorescence staining with isotype-specific AlexaFlour488 goat anti–mouse IgG3 and AlexaFlour568 goat anti–mouse IgG2a and were quantified (n = 3 × 100 cells; in triplicate from a representative biopsy). As controls, sorted cells were smeared without further staining to check for residual background fluorescence, or cells were stained after having switched the secondary antibodies to use different colors than were used previously. For staining of whole mounts of 3D cultures of cloned FACS-sorted cells, gels were fixed in methanol/acetone (1:1) for 30 min at −20°C followed by a 2-h incubation with blocking buffer and were incubated with 400 μl of antibody solution (K14 [1:25] and K19 [1:50]) overnight at 4°C, washed for 3–4 h, and incubated with secondary antibodies at 4°C overnight. Immunofluorescence stainings were evaluated using a laser-scanning microscope (LSM 510; Carl Zeiss MicroImaging, Inc.).
RNA isolation and RT-PCR
RNA isolation and PCR reactions were performed as previously described (Gudjonsson et al., 2002a,b). Specifically, primers included EpCAM forward (AGTGTACTTCAGTTGGTGCACAAA) and EpCAM reverse (AGTGTCCTTGTCTGTTCTTCTGAC; TA 56°C for 29 cycles); estrogen receptor forward (CCCTACTGCATCAGATCCAAGG) and estrogen receptor reverse (CTGCAGGAAAGGCGACAGC; TA 60°C for 40 cycles); and tenascin forward (TCCTGCTGACTGTCACAATC) and tenascin reverse (TGCTCACATACACATTTGCC; TA 60°C for 30 cycles).
To determine the integration sites of the HPV16 E6/E7–expressing vector, an RS-PCR assay was used (Sarkar et al., 1993; Ragin et al., 2004) The HPV-specific primers were modified to locate to the E6/E7 ORF, and additional RS oligonucleotides were added to the assay. In brief, genomic DNA was extracted from each cell line using the DNeasy tissue kit (QIAGEN). Previously used HPV-specific primers (Gudjonsson et al., 2002b) were supplemented with the HPV765-24D/HPV790-25D primer sets (Thorland et al., 2000). The RS oligonucleotides contain a T7 phage promoter, 10 random nucleotides, and a specific RS recognition sequence.
Amplification of vector-genome hybrid sequence was used by a seminested PCR reaction. The PCR reactions were performed using the Expand High Fidelity PCR System (Roche) in a 20-μl volume using 1 U polymerase enzyme, 1× PCR buffer, 200 μM deoxynucleotide triphosphate, 2 pmol HPV primer, and 20 pmol RS oligonucleotide primer. 1 μl from the first round of PCR was used as a template for the second round of amplification. Cycling conditions for the first round of PCR were 94°C for 2 min followed by 30 cycles of 94°C for 30 s, annealing for 30 s at 65°C descending 0.5°C each cycle, and extension at 72°C for 2 min. This was followed by 15 more cycles at a fixed TA of 55°C. The conditions for the second round of PCR were similar, and products were run on agarose gels, bands of interest were cut out, and specific products were extracted using a QIAquick Gel Extraction kit (QIAGEN). The presence of an E6- or E7-specific sequence in the selected products was verified by PCR reamplification using internal E6/E7 primers. Sequencing using a genetic analyzer (ABI PRISM 310; Applied Biosystems) was performed as previously described (Rønnov-Jessen et al., 2002).
Online supplemental material
Fig. S1 shows a more detailed characterization of the ductal stem cell zone. Fig. S2 shows that clonal colonies from ducts, in contrast to lobule-derived colonies, are multipotent. Fig. S3 shows that SSEA-4hi cells cosort with the Lin−CD49f+EpCAMhi population and that these cells give rise to multipotent clonal colonies. Fig. S4 shows that the stemlike cells (+/+E6/E7) express surrogate stem cell markers and are bipotent. A flow chart describing the isolation of one stemlike cell and three lineage-restricted progenitor cells is also shown. Table S1 shows that bypass of senescence maintains the differentiation repertoire of target cells with respect to a panel of markers for differentiated luminal and myoepithelial cells. Supplemental materials and methods provides further details about cell lines and clones, FACS analysis and cloning, immunohistochemistry and cytochemistry, and RNA isolation and RT-PCR.
We thank Tove Marianne Lund for expert technical assistance, Naja Slot Bundgaard and Mimi Birkelund for help with immunostainings and RT-PCR, and Keld Ottosen for photographic assistance. We also thank Søllerød Privathospital, the Private Clinic, and Dr. A. Aasted for providing the normal breast biopsy material.
This work was supported by the Danish Cancer Society, Dansk Kræftforskningsfond, Danish Research Council, Fru Astrid Thaysens Legat for Lægevidenskabelig Grundforskning, The Meyer Foundation, The Toyota Foundation, the Novo Nordic Foundation, the European Commission Research Directorates (contract HPRN-CT-2002-00246), Bernhard Rasmussen og hustru Meta Rasmussens fond (to L. Rønnov-Jessen), the US National Cancer Institute (grant CA-64786 to M.J. Bissell and O.W. Petersen), and the Office of Biological and Environmental Research (OBER) of the US Department of Energy (grant to M.J. Bissell). M.J. Bissell is the recipient of an Innovator award from the US Department of Defense and a Distinguished Fellow award from the OBER office of the US Department of Energy. M.A. LaBarge was supported by a postdoctoral fellowship from the American Cancer Society. T. Gudjonsson is supported by the Research Grant of Excellence from the Icelandic Research Council.
R. Villadsen and A.J. Fridriksdottir contributed equally to this paper.
Abbreviations used in this paper: HPV, human papilloma virus; lrECM, laminin-rich ECM; RS, restriction site; SSEA-4, stage-specific embryonal antigen-4; TDLU, terminal duct lobular unit.