Introduction
The breast undergoes dynamic changes over the lifetime of a woman, from initial development at puberty, to proliferation and apoptosis during the menstrual cycle in response to hormonal fluctuations, and culminating with full lobuloalveolar development for lactation. Studies from murine mammary glands have demonstrated the presence of an epithelial hierarchy that serves to prevent exhaustion of progenitors necessary for tissue homeostasis and glandular regeneration. These extensive studies have identified and localized lineage-restricted progenitor cells that can be assayed using both
in vitro and
in vivo techniques (for review [
1]-[
4]).
There are numerous differences between the human breast and the mouse mammary gland that preclude the direct translation of rodent studies to human breast development. The human breast is composed of 11 to 48 central ducts that radiate outward from the nipple [
5], where circulating hormones and localized growth factors likely coordinate the growth of the terminal ductal lobular units (TDLU) that emanate from primary ducts. Each lobe is organized as heterogeneous lobular structures, each one representing a sequential developmental stage [
6]-[
8]. In contrast, mouse mammary glands are composed of a simple ductal tree that lack TDLU, and only exhibit strain-specific rudimentary alveolar budding in the absence of pregnancy. Thus, the anatomical and physiological equivalency of lobules and TDLU is unclear in the mouse.
The mammary gland is composed of a bi-layered epithelium; basal/myoepithelial (ME) cells express cytokeratin (CK) 14 surrounding a luminal layer that stains positively for CK8/18. In the mouse, CK expression is specific for each epithelial layer, which has enabled the use of these markers for lineage tracing studies to establish the presence of lineage-restricted progenitor cells within each layer of the mammary epithelium [
9]. Unlike the mouse, little is known about the identity and dynamics of progenitor cells in the human breast, and details about their activity and the mechanisms that regulate their numbers and differentiation remain poorly understood. Interestingly, contiguous regions of human breast lobules contain cells showing identical X-chromosome inactivation patterns suggesting that they were likely derived from a common uncommitted stem cell [
10].
The limited understanding of human breast development and stem cell biology has largely been due to the lack of appropriate model systems and assays to detect, analyze, and characterize stem cell properties. In recent years, we and others have developed and optimized various
in vivo and
in vitro tools to study the biology and mechanisms governing human breast development [
1],[
11]-[
16]. Using these approaches we sought to dissect the epithelial hierarchy of the human breast and identify the anatomic locations of progenitor cells within the breast. In doing so, we reveal that human breast tissues contain two types of phenotypically distinguishable progenitor cells localized to the luminal and basal lineages, respectively, which contribute to different anatomical structures. Further, we show that immature lobules within the breast harbor distinct types of progenitor cells.
Materials and methods
Animal studies
All animal procedures in this study were approved by the Tufts University Institutional Animal Care and Use Committee (IACUC), and all animal procedures were conducted in accordance with this approved protocol. Colonies of NOD/SCID mice were maintained in house. Mice were given food and water ad libitum.
Primary tissue isolation and culture
All human breast tissues were obtained in compliance with the laws and institutional guidelines, as approved by the Institutional Review Boards (IRB) from Tufts Medical Center and Beth Israel Deaconess Medical Center. For these studies, we utilized de-identified, non-cancerous breast tissues from patients undergoing elective reduction mammoplasty surgery under the Category 4 exemption of the IRB Review. As de-identified patient samples were utilized for this study, patient consent was not required. The tissue was sampled to make whole mounts and paraffin blocks, and snap frozen for molecular analyses, and the remainder was enzymatically digested to epithelial organoids as previously described [
12],[
17]. The epithelial organoids were aliquoted in 1:1 DMEM/Hams-F12 media (Invitrogen, Grand Island, NY USA) supplemented with 5% calf serum, 10 ng/mL insulin, 10 μg/mL epidermal growth factor (EGF), 10 μg/mL hydrocortisone, and 10% dimethyl sulfoxide (DMSO) and stored in liquid nitrogen for later use Insulin, hydrocortisone, EGF, and DMSO were all obtained from Sigma-Aldrich Corporation (St. Louis, MO, USA). To humanize mice, mammary epithelium was removed from the fourth mammary glands of 3-week-old NOD/SCID females, and RMF-EG cells were injected into the fat pad as described [
12]. Two weeks post-humanization, epithelial organoids were dissociated to single cells (100,000 cells), co-mixed with primary breast stromal cells (2.5 × 10
5 cells per gland) in a 1:1 mixture of collagen and Matrigel (BD Biosciences, San Jose, CA, USA) and injected into humanized fat pads. RMF-EG cells and primary breast stromal cells were grown in high glucose DMEM (Invitrogen) supplemented with 10% calf serum and 1% penicillin/streptomycin at 37°C and 5% carbon dioxide.
Progenitor cell assays
For colony formation, 40,000 human mammary epithelial cells (MEC) were plated on non-adherent plates or adherent plates in 2 mL of epithelial basal media supplemented with bovine pituitary extract (52 μg/mL), hydrocortisone (0.5 μg/mL), human EGF (10 ng/mL), and insulin (5 μg/mL; Lonza, Allendale, NJ, USA) for 7 days. Mammospheres or floating colonies were plated in triplicate and quantified using a Multisizer3 Coulter Counter (Beckman Coulter, Danvers, MA, USA). Colonies growing on adherent plates were fixed with methanol and stored at -20°C until stained for cytokeratins 8 and 14. Mammospheres and floating colonies were cytospun onto glass slides, methanol fixed, and stored at -20°C until used for analyses. For growth on collagen, 1 mg/mL rat tail collagen (pH = 7; BD Biosciences) polymerized for 30 minutes at 37°C on four- or eight-well chamber slides (Corning, Corning, NY, USA). MECs (n = 5,000) were plated in 1 mL of complete epithelial basal media supplemented with 2% Matrigel (BD Biosciences) in duplicate. Colony growth was observed using Nikon Eclipse Ti and quantified based on morphology. For mammosphere and floating colony growth on collagen, MECs were grown as mammospheres and floating colonies for 7 days in triplicate. One replicate was quantified, and two replicates were plated on collagen. All conditions were plated on collagen in duplicate.
Mammary whole mounts
Tissue isolated from elective reduction mammoplasty surgery from eight patients was used to make whole mounts. Each tissue was sampled in 5 to 10 different areas to make whole mounts, which were formalin-fixed overnight and stained with carmine. The tissue was sliced into 2- to 3-mm sections, dehydrated in graded alcohols, and incubated in xylenes to remove fat. The mammary sections were imaged using a Nikon Eclipse80i microscope, and lobules were identified as described [
6]. Briefly, the number of ductules/lobule were counted and categorized as Type I (average (avg) 11 ductules/lobule), Type II (avg 47 ductules/lobule), and Type III (avg 81 ductules/lobule). Lobules of each type were quantified from multiple sections of the reduction mammoplasty tissue, bluntly dissected with a razor, paraffin-embedded, and sectioned at 5 μm for immunohistochemical and immunofluorescent staining.
Immunofluorescence and immunohistochemistry
Human-in-mouse normal outgrowth tissue sections were stained with an antibody for α-smooth muscle actin (1:1000; Vector Laboratories, Burlingame, CA, USA). Sorted breast lobules and sections from human-in-mouse normal outgrowths were stained with antibodies for cytokeratin (CK) 8/18 (1:250, Vector Laboratories) and CK 14 (1:250, Thermo Scientific, Tewksbury, MA, USA). Breast lobules were additionally stained with p63 (1:500; Santa Cruz Biotechnologies, Dallas, TX, USA), estrogen receptor alpha (1:500, F10 clone, Santa Cruz Biotechnologies), progesterone receptor (1:250; Cell Signaling Technologies, Danvers, MA, USA), and EpCAM (1:500; Stem Cell Technologies, Vancouver, BC, Canada).
For immunohistochemistry, tissue sections were incubated with 2% hydrogen peroxide to quench endogenous peroxidase, antigen retrieved in 0.1 M citrate (pH = 6.0), and blocked with 1% BSA and 1.5% goat serum in PBS for 1 hour. Tissue sections were incubated with primary antibodies in 1% BSA in PBS overnight at 4°C, followed by secondary antibodies for 30 minutes at room temperature (1:250 biotinylated anti-mouse or anti-rabbit; Vector Laboratories). Staining was detected using Vectastain ABC kit followed by ImmPact DAB kit (Vector Laboratories), and sections were counterstained using hematoxylin. For adherent colonies on plastic plates, colonies were fixed in methanol at −20°C for 10 minutes, and cells were permeabilized with 0.1% triton in PBS for 10 minutes. Cells were blocked with 1% bovine serum albumin (BSA) and 1.5% goat serum in PBS for 1 hour, incubated with anti-rabbit CK 14 at 4°C overnight, and stain was detected as described above. Following 3,3-diaminobenzidine (DAB) stain, colonies were blocked with the avidin/biotin blocking kit (Vector Laboratories) per the manufacturers’ instructions, with 1% BSA in PBS for 1 hour, and with anti-mouse CK 8/18 overnight. Specific staining was detected as described above, except the substrate used for detection was the VIP kit (Vector Laboratories). Colonies were air-dried and quantified by morphology and cytokeratin expression as described [
16].
For immunofluorescence, tissue sections were antigen-retrieved in 0.1 M citrate (pH = 6.0) and blocked with 1% BSA and 1.5% goat serum in PBS for 1 hour. Tissue sections were incubated with primary antibodies in 1% BSA in PBS overnight at 4°C, followed by secondary antibodies, for 30 minutes at room temperature (1:250; Alexa Fluor 488 or 546, Invitrogen). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) in the mounting media using Vectashield (Vector Laboratories). For double-labeling, tissue sections or mammospheres and floating colonies were incubated with rabbit primary antibodies and secondary antibodies followed by mouse primary and secondary antibodies. Images were captured with Nikon Eclipse 80i and merged using ImageJ software.
Quantification of mammary whole-mount staining
Five lobules of each type were evaluated from each patient sample. The percentage of positive cells for each marker was quantified in each high-power image and categorized into one of three percentage categories. EpCAM staining was evaluated using Allred scoring based on staining intensity and number of cells staining positively. Staining differences among lobule types was determined by chi-squared analyses.
Flow cytometry
Unsorted cells from organoid preparations were dissociated to a single-cell suspension as described above and filtered through a 20 μm nylon mesh (EMD, Millipore, Billerica, MA, USA). Endothelial, lymphocytic, monocytic, and fibroblastic lineages were depleted with antibodies to CD31, CD34, and CD45 (all from Thermo Scientific), fibroblast-specific protein IB10 (Sigma), and a mixture of pan-mouse IgG and IgM Dynabeads (Dynal; Invitrogen) according to the manufacturers’ instructions and as described previously [
11],[
18]. Lineage-depleted single-cell suspensions or cells from immunomagnetic bead-sorted populations were resuspended at 1 × 10
6 cells per mL in PBS containing 1% calf serum (fluorescence-activated cell sorting (FACS) buffer) and bound with fluorescently conjugated antibodies to human EpCAM (allophycocyanin), CD24 (FITC; BD Biosciences), CD49f (PerCP-Cy5.5; Biolegend, San Diego, CA, USA), and CD10 (phycoerythrin; Dako, Carpenteria, CA, USA) for 20 minutes at 4°C. Antibody-bound cells were washed and resuspended at 1 × 10
6 cells per mL in FACS buffer and run on a FACSCalibur flow cytometer (BD Biosciences). Flow cytometry data were analyzed with the FlowJo software package (TreeStar, Ashland, OR, USA). Cellular debris and dead cells removed from analysis through manual gating and use of propidium iodide. Gates were set on the remaining cells based on isotype controls for each antibody.
Spanning-tree progression analysis of density-normalized events (SPADE)
SPADE was used to generate tree diagrams from flow cytometry data as previously described [
19] using the SPADE 2.0 MATLAB implementation. Enzymatically dissociated, lineage-depleted breast-reduction tissue was prepared and analyzed by flow cytometry. Ungated FACS data generated from eight patient samples was used to construct the SPADE tree using three parameters (CD49f, CD24, and EpCAM). Downsampling was manually set at 10,000 cells per sample and the number of desired clusters set to 50. The SPADE tree generated was evaluated and sectioned based on characterized relationships between markers: mature luminal (EpCAM
hiCD49f
-), luminal progenitor cells (LPC) (EpCAM
hiCD49f
+), mature basal (MB) (EpCAM
loCD49f
+), basal progenitor cells (BPC) (EpCAM
-CD49f
+), and mammary lineage negative (MLN) (EpCAM
-CD49f
+). To visualize differences between tissue samples predominantly consisting of immature (Type I/II) and mature (Type III) lobules, data from four patients enriched for Type I/II lobules and 4 patients enriched for Type III lobules were concatenated in FlowJo, upsampled to the previously generated SPADE tree, and pseudocolored to indicate differences in the frequency of cells falling into each cluster.
Bead sorting
MEC were plated briefly in serum (1 to 2 h) to deplete mammary fibroblasts from the organoid fraction. The organoids remaining in suspension were dissociated by trypsinization and filtered with a 40 μm filter (BD Biosciences) to remove residual clustered cells. Single-cell suspensions of breast epithelial cells were sorted with CELLection pan-mouse IgG magnetic beads (Dynal; Invitrogen) conjugated to an anti-CD10 antibody (clone SS2/36; Santa Cruz Biotechnology) according to the manufacturer’s instructions. CD10+ cells were released from the beads by DNase treatment with occasional agitation at 37°C for 10 minutes. Cells that did not bind to the CD10 beads were further sorted with magnetic beads conjugated to an anti-EpCAM antibody (clone VU-ID9; Abcam, Cambridge, MA, USA and AbD Serotec, Raleigh, NC, USA). Positive cells were again released by DNase treatment. EpCAM+ bead-sorted cells were further sorted by binding of CD49f antibody (clone 450-30A; AbD Serotec) followed by binding of pan-mouse IgG CELLection beads. Beads were released from positively sorted cells as described above. Viable cells (verified by trypan blue exclusion) from unsorted, basal progenitor cells (BPC, EpCAM-CD10+), mature basal (MB; EpCAM+/loCD10+), luminal progenitor cells (LPC, EpCAM+CD49f+) and mature luminal (ML) (EpCAM+CD49f-) cells were used for collagen assays.
Statistics
Differences between two groups were detected with Student’s t-test, and differences among multiple groups were detected using repeated measures analysis of variance (ANOVA) with multiple comparisons post hoc test. Statistical analyses were performed using Graph Pad Prism software (La Jolla, CA, USA).
Discussion
The human breast undergoes extensive remodeling following birth through pregnancy and lactation, suggesting the presence of a cellular hierarchy to functionally expand the epithelium. Here, we demonstrate that alveolar and ductal structural progenitors exist and reside within luminal and basal lineages of the breast, respectively. Our findings strongly support a model of the breast epithelial hierarchy in which two types of phenotypically distinguishable progenitors contribute to specific structural elements of the mammary tree (ducts or alveoli) and that the combination of these progenitor cell populations is necessary for development and maintenance of mammary tissues. Luminal alveolar progenitors are more abundant in transplant assays and in vitro collagen assays than basal ductal progenitors, and SPADE demonstrated that the luminal lineages show increasing complexity with lobule maturity. These results suggest that different types of luminal alveolar progenitors may exist within the breast to functionally expand the lobules.
Stem/progenitor cells have been an important area of investigation for understanding the cellular origin of breast cancer (for review, [
2]). Localization of stem/progenitor cells within human tissue has been challenging due to limitations on tissue availability, patient heterogeneity, and differences among techniques to measure stem/progenitor activity and cell surface markers used in studies. Multiple studies have examined stem/progenitor activity in populations defined by cell surface markers [
11],[
16],[
20],[
22],[
30],[
31], but have reported conflicting results about whether luminal or basal lineages were enriched for progenitor activity. Here, we show the relationships among the assays used to detect stem/progenitor activity, as well as their relationships with the cell surface markers most commonly used. Although there are important functional differences among the lineages, both the luminal and basal lineages retain distinct progenitor competency. Further, we show that Type I lobules, which are the least developed of the breast lobules and most closely associated with terminal ducts, are enriched in progenitor activity. Although we did not directly assess collecting ducts, our results are consistent with those found by Villadsen and colleagues who previously mapped progenitor activity to cells within main and terminal ducts, through use of collagenase digestion of breast epithelial organoids and microdissection [
22].
Recent studies of the mouse epithelial hierarchy have incorporated numerous cell surface markers to characterize specific epithelial populations. However, determining the relationships among the cells identified by these markers has proven challenging. For the first time, using SPADE we identified specific differences between breasts enriched for Type I/Type II and Type III lobules that were obscured by traditional flow cytometry gating. Epithelial cells isolated from tissues enriched for Type I/Type II lobules demonstrated increases in basal lineages and mammary lineage negative cells, as well as different clusters of LPCs, compared with those from Type III lobules. Additionally, epithelial cells enriched from Type III lobules demonstrated increased clusters of mature luminal cells, which clustered into a more complex hierarchy when examined individually. Although recent studies have begun the integration of more cell surface markers to delineate the specific functions of different luminal epithelial cell types [
30], the relationship between different luminal progenitor cells and lobule types were not examined. Identification and localization of these cell types in the context of lobule maturity may clarify functional studies of breast epithelial cells, where inherent breast heterogeneity often obscures consistent results.
Lineage tracing to physiologically identify and characterize progenitor cells in the mouse mammary gland has demonstrated that following birth, the expansion and maintenance of the luminal and basal lineages is ensured by the presence of lineage-restricted progenitor cells [
9],[
32]. Elegant studies have demonstrated that expression of both CK14 and CK8 were restricted to basal and luminal cells, respectively [
9]. However, within human breast tissue, the presence of CK14
+ epithelial cells within the luminal layer have been previously described [
33]-[
36], although the localization of CK14
+ luminal cells was not evaluated with respect to lobule types. Here, we show that expression of CK8 and CK14 was variable in Type I lobules, with luminal expression of CK14 and basal expression of CK8 as well as co-expression of both cytokeratins in the luminal and basal layers. These results suggest that either human epithelial cell lineages are not as restricted as in the murine gland, or that CK8 and 14 expression is less specific to the luminal and basal lineages in the human as it is in the mouse. Although CK14 was present in luminal cells in Type I lobules, we did not identify any p63
+ epithelial cells in this layer, suggesting that CK14 may not always be a precise marker for basal epithelial cells.
Russo and Russo have classified the degree of breast complexity into Type I, Type II, and Type III lobules [
6]-[
8]. The increasing lobular complexity during pregnancy in preparation for lactation suggests that the lobules become successively more mature in the progression from Type I to Type III. This implies that breast tissue from multiparous women contain an increased number of Type III lobules compared to the breasts of nulliparous women. Although mice undergo complete involution following pregnancy, resulting in parous mammary glands that are structurally very similar to nulliparous glands, the degree of involution in humans following pregnancy has not been well-characterized. Additionally, Type III lobules have been identified in very young nulliparous breast tissue [
37],[
38], which suggests that considerable heterogeneity exists among all patient samples, regardless of parity. Interestingly, in breast tissue from women undergoing elective reduction mammoplasty surgery, we observed surprising variability among lobules of the same patient for steroid receptor expression and proliferation, suggesting an underlying mechanism of regulation for lobule expansion. Even under the differentiating conditions of pregnancy, studies have noted that it is not unusual to see acini that are devoid of proliferative markers, while other acini demonstrate a dramatic proliferative response [
37],[
38]. Clearly, the mechanisms regulating the response of individual lobule types to hormonal stimuli require further investigation to determine why one lobule may remain quiescent, while others expand.
In this study, we did not observe any clear differences between Type I, II, and III lobules and age or parity, however, the number of patients examined in this study was small. Although we tried to obtain detailed reproductive information about the patients used in this study, we were not able to obtain information regarding the phase of the menstrual cycle or phase of oral contraceptive use at the time of surgery. Given the critical role that steroid hormones play in breast development [
3],[
6], changes circulating levels of progesterone and estrogen during the course of the menstrual cycle may influence the complexity of the epithelial populations present within the mammary gland as well as the structures of the lobules. A large study conducted with detailed questionnaires in addition to quantification of breast lobule types and epithelial cell surface markers would need to be conducted in order to fully address this issue.
Type I lobules have been characterized as being the least mature of the lobules types of the breast [
6],[
8]. Our results support this hypothesis, given the heterogeneity of expression of multiple markers within this lobule type. Interestingly, a recent study has demonstrated that breast tissue of aging women shows an increase in luminal CK14 expression [
39], however, the localization of these cells was not directly examined within the lobules. Following menopause, the breast undergoes involution [
40], which may increase the number of Type I lobules present within the tissue. Given that women older than 50 years account for the majority of new breast cancer diagnoses [
40], examining the effect of aging on lobule composition may provide insight into the origin of breast cancer in this population. Studies examining the effects of carcinogens on tumorigenesis using rodent models have supported the idea that transformation of cells in Type I lobules give rise to the most common breast malignancies, whereas dysregulated growth in more mature lobules become benign breast lesions [
41],[
42]. However, a nested case-control study in the Nurses’ Health Studies demonstrated that women with predominantly Type I lobules within their breast had a decreased risk of breast cancer compared with those who had no Type I lobules or mixed lobule types [
43]. Unraveling the normal growth regulation of specific lobules within the breast may provide insight into the dysregulated growth during cancer. As our data suggest underlying differences in the epithelial hierarchy between breasts with immature and mature lobules, studies examining the correlation between the lobule types present in the normal tissue matched with the specific breast tumor subtype could potentially clarify the differences between rodent and human models of breast cancer.
Competing interests
The authors declare that they have no competing interests.