Parity induces differentiation and reduces Wnt/Notch signaling ratio and proliferation potential of basal stem/progenitor cells isolated from mouse mammary epithelium

Introduction Early pregnancy has a strong protective effect against breast cancer in humans and rodents, but the underlying mechanism is unknown. Because breast cancers are thought to arise from specific cell subpopulations of mammary epithelia, we studied the effect of parity on the transcriptome and the differentiation/proliferation potential of specific luminal and basal mammary cells in mice. Methods Mammary epithelial cell subpopulations (luminal Sca1-, luminal Sca1+, basal stem/progenitor, and basal myoepithelial cells) were isolated by flow cytometry from parous and age-matched virgin mice and examined by using a combination of unbiased genomics, bioinformatics, in vitro colony formation, and in vivo limiting dilution transplantation assays. Specific findings were further investigated with immunohistochemistry in entire glands of parous and age-matched virgin mice. Results Transcriptome analysis revealed an upregulation of differentiation genes and a marked decrease in the Wnt/Notch signaling ratio in basal stem/progenitor cells of parous mice. Separate bioinformatics analyses showed reduced activity for the canonical Wnt transcription factor LEF1/TCF7 and increased activity for the Wnt repressor TCF3. This finding was specific for basal stem/progenitor cells and was associated with downregulation of potentially carcinogenic pathways and a reduction in the proliferation potential of this cell subpopulation in vitro and in vivo. As a possible mechanism for decreased Wnt signaling in basal stem/progenitor cells, we found a more than threefold reduction in the expression of the secreted Wnt ligand Wnt4 in total mammary cells from parous mice, which corresponded to a similar decrease in the proportion of Wnt4-secreting and estrogen/progesterone receptor-positive cells. Because recombinant Wnt4 rescued the proliferation defect of basal stem/progenitor cells in vitro, reduced Wnt4 secretion appears to be causally related to parity-induced alterations of basal stem/progenitor cell properties in mice. Conclusions By revealing that parity induces differentiation and downregulates the Wnt/Notch signaling ratio and the in vitro and in vivo proliferation potential of basal stem/progenitor cells in mice, our study sheds light on the long-term consequences of an early pregnancy. Furthermore, it opens the door to future studies assessing whether inhibitors of the Wnt pathway may be used to mimic the parity-induced protective effect against breast cancer.


Introduction
Pregnancy is the most significant modifiable factor known for breast cancer risk in women. Although an initial increase in risk occurs immediately after parturition in women older than 25 years, the overall lifetime risk of breast cancer decreases after pregnancy [1,2]. This protective effect is > 50% if a full-term pregnancy has occurred before the age of 20 years [1]. Similarly, pregnancy and pregnancy-mimicking hormones have a strong protective effect against mammary tumors in rodents. This is true both for carcinogen-induced mammary tumors [3] and for genetically engineered mouse models of breast cancer [4].
The cellular and molecular mechanisms underlying the breast cancer-protective effect of early pregnancy remain unclear. Frequently raised hypotheses involve cell nonautonomous mechanisms such as systemic changes in circulating hormones and/or changes in the stromal composition of the mammary gland [5,6], and cell autonomous processes such as changes in the differentiation state of mammary epithelial cells [7]. Furthermore, numerous parity-induced changes in gene expression have been identified in genome-wide expression profiles of entire lobular breast tissues of women or entire mammary glands of rats and mice [8][9][10]. However, it is not known to what extent these tissue studies reflect alterations in gene-expression profiles of distinct mammary epithelial cell subpopulations. Hence, given that breast cancers arise from specific subpopulations of mammary epithelial cells [11], investigations of early parity-induced gene-expression changes in distinct mammary epithelial cell subpopulations are warranted.
The mammary epithelium is hierarchically organized into differentiated luminal and basal (myoepithelial) cells, luminal and basal progenitor cells, and mammary stem cells [12,13]. Whereas the latter were originally thought to lie exclusively in the basal compartment and to be multipotent (able to form both luminal and basal epithelial cells), recent lineage-tracing experiments indicated the existence of unipotent basal and luminal mammary stem cells and identified multipotent mammary stem cells solely in the embryonic and possibly in the pregnant gland [14,15]. Distinct mammary epithelial cell subpopulations, including luminal progenitor and basal stem/progenitor cells can be isolated with fluorescence-activated cell sorting (FACS) by using specific cell-surface markers from both parous and virgin mice [16][17][18][19][20][21]. Whereas progenitor cells in general can be characterized in vitro by their colony-forming potential [16,22,23], the basal stem/progenitor cell subpopulation has the additional capacity to repopulate deepithelialized mouse mammary fat pads in vivo [16][17][18]. Although previous studies in total mammary epithelial cells indicated either no change or a decrease in the mammary repopulating capacity after parity [20,21], the consequences of parity on the transcriptome and functionality of specific mammary epithelial cell subpopulations have not been investigated.
Therefore, we examined in this study whether pregnancy alters the gene-expression profiles ("gene signature") and the differentiation/proliferation potentials of the various mammary epithelial cell subpopulations. The results indicate that early parity decreases Wnt4 expression in luminal epithelial cells, leading to a reduction in the Wnt/Notch signaling ratio specifically in basal stem/ progenitor cells. As expected, the decrease in the Wnt/ Notch signaling ratio is associated with a concomitant strong prodifferentiation and antiproliferation phenotype in basal stem/progenitor cells. Because a decrease in Wnt signaling is known to have an anticarcinogenic effect [24,25], the findings support the hypothesis that a reduction in the Wnt/Notch signaling ratio in basal mammary stem/progenitor cells plays a role in the mitigating effect of early pregnancy on breast tumorigenesis.

Animals and animal experimentation
All experiments were conducted in genetically homogenous FVB/NHanHsd mice purchased from Harlan Laboratories. The mice were bred and maintained in the animal facility of the Friedrich Miescher Institute, according to the Swiss guidelines on animal experimentation. All experiments were performed under permit 2159-2, in accordance with the animal-welfare ordinance and approved by the cantonal veterinary office of Basel Stadt, Switzerland. For the early-pregnancy protocol, mice were time-mated when 42 days old and allowed to lactate for 21 days. The postweaning period until cell harvest was 40 days, unless stated otherwise. To control for the estrus cycle in the transcriptome analyses, at least five mice were grouped for gland harvesting. For immunohistochemical analyses and determination of blood progesterone levels, mice in estrus were used, as assessed by the presence of a vaginal plug after an overnight mating. Age-matched virgin control mice were maintained under the same conditions as parous mice.

Mammary cell preparation
The fourth mammary glands from virgin and parous mice were collected after lymph node removal and pooled. Mammary epithelial organoids were prepared as described [26]. Adipocytes were removed by repeated centrifugations (300 g). Red blood cells (RBCs) were eliminated by incubation with 8.3 g/L ammonium chloride (pH 7.5) for 5 minutes. The number of fibroblasts in gland extracts was reduced by their attachment to polystyrene cell-culture flasks (Corning, Buchs, Switzerland) during a 45-minute incubation step in Dulbecco modified Eagle medium (Invitrogen, Zug, Switzerland) with 10% FCS (Sigma, Buchs, Switzerland) at 37°C/5% CO 2 [26]. The epithelial organoids were directly processed to single-cell suspensions by digestion in Hyclone HyQ-Tase (ThermoScientific, Lausanne, Switzerland) with gentle pipetting for 15 minutes at 37°C. The cell suspension was filtered through a 40-μm cell strainer (BD Falcon, Basel, Switzerland), and the final cell suspension pelleted at 650 g for 4 minutes.

In vitro colony formation assay and quantification
Freshly sorted cells of each subpopulation (500 cells) were plated onto irradiated 3T3L1 feeder cells (5,000 cells) in 24well Primaria plates (Becton Dickinson, Basel, Switzerland). The cells were cultured in DMEM/Ham F12 (Invitrogen, Zug, Switzerland) with 10% FCS, 100 IU/ml penicillin, 100 μg/ml streptomycin (Invitrogen, Zug, Switzerland), 5 μg/ml bovine pancreatic insulin (Sigma, cell culturetested solution, Buchs, Switzerland), 10 ng/ml mouse EGF (BD Biosciences, Basel, Switzerland), and 10 ng/ml cholera toxin (Sigma, Buchs, Switzerland). After 24 hours, the medium was renewed, and 4 days later, the colonies were fixed with acetone/methanol (1:1), washed, and rehydrated with PBS and 0.05% NaN 3 Figure 1 The CD24/Sca1 and CD49f High /CD24 flow-cytometry profiles of parous and age-matched virgin mice are similar. (A) Schematic illustration of the cell-isolation strategy and representative flow-cytometry pseudocolor plots of mammary cells from age-matched virgin control mice. After depletion of CD45 + white blood cells, luminal and basal mammary epithelial cells were separated on the basis of CD24 and Sca1 expression. Further separation of basal cells into myoepithelial and basal stem/progenitor cell subpopulations was based on the expression of CD24 and CD49f. The isolated mammary epithelial cell subpopulations included luminal Sca1 + (CD24 +High Sca1 + ) cells, luminal Sca1 -(CD24 +High Sca1 -) cells, basal CD49f High (CD24 +Low Sca1 -CD49f High ) or basal stem/progenitor cells, and basal myoepithelial (CD24 +Low Sca1 -CD49f Low ) cells. (B) Outline of the mouse mating, parturition, weaning, and involution protocol. (C) Representative flow-cytometry pseudocolor plots of mammary cells from parous mice. The gates applied were the same as those for agematched virgin controls. (D) Bar graph showing the distribution of mammary epithelial cell subpopulations comparing cells from parous with age-matched virgin control mice. Data represent the mean ± SEM of seven cell-isolation experiments with a minimum of 10 mice per experiment. The proportion of luminal Sca1 + cells was reduced by approximately 50% in parous mice (P = 0.02 with a two-tailed unpaired Student t test). the incubation medium was supplemented with recombinant mouse Wnt4 (R&D Systems, Abingdon, UK) at 500 ng/ml. For quantification, colonies were immunofluorescently stained with Krt18 and Krt14 antibodies and with Hoechst 33342. The stained colonies were imaged with a MacroFluo Z16 microscope (Leica, Heerbrugg, Switzerland) at 2× magnification, and a Z1 microscope (Zeiss, Feldbach, Switzerland) at 5× magnification. The feeder cells served as negative controls for Krt18 and Krt14 staining. The number of colonies (that is, clusters of more than three cells [23]) per well was determined manually. Colonies defined as Krt18/Krt14 double-positive were double positive over a minimum of 20% of colony area.

Mammary fat pad transplantation
Freshly sorted basal stem/progenitor cells were resuspended in PBS with 50% Matrigel (BD Biosciences, Basel, Switzerland) and injected (30 μl) in limiting dilution numbers into the cleared fourth mammary glands of 3-week-old syngeneic female mice. Glands from recipient mice were harvested 8 to 9 weeks after transplantation, processed as whole mounts (see earlier), and scanned with an Epson Expression 1600 Pro scanner. The number of outgrowths was counted, with an outgrowth defined as an epithelial structure with ducts starting from a central point and with lobules and/or terminal end buds (see Additional file 2) [27].

Microarray analyses
Microarray analyses were performed on unsorted total mammary cell suspensions (see Figure 2) and on FACSsorted mammary cell subpopulations (see Figures 3 and 4, as well as Table 1). For unsorted mammary cells, RNA was extracted by using the RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). Genomic DNA was removed by using gDNA Eliminator Mini Spin Columns (Qiagen, Hilden, Germany). The RNA concentration was measured with a Nanodrop 1000 machine, and RNA quality determined with an Agilent 2100 bioanalyzer and RNA Nano Chips. Aliquots of 100 ng of isolated total RNA were amplified by using the Ambion WT Expression kit (Ambion, Zug, Switzerland). For FACS-sorted mammary epithelial cell subpopulations, total RNA was isolated by using the Arcturus PicoPure Frozen RNA Isolation Kit (Life Technologies, Zug, Switzerland). Genomic DNA was removed by using RNase-Free DNase (Qiagen, Hilden, Germany), the RNA concentration determined by using the RiboGreen Assay, and RNA quality assessed by using an Agilent 2100 bioanalyzer and RNA Pico Chips. Total RNA was used as the input for synthesis of amplified cDNA with the NuGen Ovation Pico WTA System (NuGen Inc., Leek, The Netherlands).
Resulting double-stranded cDNA was fragmented and labeled by using the Affymetrix GeneChip WT Terminal V ir g in s 3 V ir g in s 2 V ir g in s 4 P a ro u s 1 P a ro u s 4 P a ro u s 2 P a ro u s 3  calculated with R/Bioconductor's (v2.14) "affy" package by using the rma() function. Differential gene expression between experimental and control samples was determined by using linear modeling as implemented in the R/Bioconductor package "limma." For general analysis of gene expression in total mammary cell suspensions and in FACS-sorted mammary cell subpopulations, we used the cut-off linear fold change > 1.5, adjusted P value < 0.05, and average linear expression between conditions greater than 4. To determine the 10 most up-or downregulated genes in FACS-sorted mammary epithelial cell subpopulations, a cut-off of linear fold change > 2.0, adjusted P value < 0.001, and average linear expression between conditions ave > 4 was used. For FACS-sorted mammary epithelial cell subpopulations both resulting , and downregulated genes, as negative (-). Differentiation genes were upregulated (blue), Wnt target genes were downregulated (green), Wnt inhibitor Sfrp1 was upregulated (green) and overall Notch signaling (orange) was increased in basal stem/progenitor cells from parous mice. (C) qPCR validation of the changes in gene expression in basal stem/progenitor cells of parous mice. All classic Wnt target genes were downregulated, including Lgr5, Axin2, and versican (Vcan), whereas the more ubiquitously regulated target Myc was unchanged. In all cases, fold changes are shown relative to cells from age-matched virgin control mice. Ct values were normalized to the reference genes Hprt and Ubc [63]. Data represent the mean ± SEM of three independent experiments with 10 mice (five virgins; five parous) per experiment.
(D) Representative images and quantification of immunostaining for the Wnt target gene versican in mammary gland sections from agematched virgin and parous mice in estrus. Scale bar, 50 μm. Quantitative data represent the mean ± SEM from 60 randomly selected images from three virgin and three parous mice.
Percentage of Epithelial Cells  . Activity is reported as Z scores (positive Z score, upregulation; negative Z score, downregulation) by using a cut-off of linear fold change > 1.5 and P value < 0.05. (E) BioFunctions most strongly downregulated (blue) or upregulated (orange) in basal stem/ progenitor cells, as calculated by Ingenuity IPA [40]. The color of the fields represents the Z score; the size of the fields represents the number of genes. A stringent cut-off of linear fold change > 2 and P value < 0.01 was used to minimize the number of false positives.
lists of differential genes were imported into Ingenuity IPA (Ingenuity, content version 12710793) for pathway analysis. Gene-set enrichments in FACS-sorted mammary epithelial cell subpopulations were determined by using the JAVA application from the Broad Institute v2.0 [28,29] and gene sets v2.5 and v3.0, as well as custom gene sets (see Additional file 3).
The microarray data from this publication have been submitted to the NCBI Gene Expression Omnibus [30] and are deposited as GSE40875 (mouse mammary cell subtypes), GSE40876 (total mammary epithelial cells in mice), and GSE40877 (both total mammary epithelial cells and mammary cell subtypes in mice).

Quantitative PCR
RNA was isolated as described earlier and converted to cDNA by using the WT-Ovation Exon Module Version 1.0 (NuGen Inc., Leek, The Netherlands). Real-time PCR was performed on the unamplified cDNA corresponding to the specified number of cells or on 25 ng of amplified cDNA (by using the NuGen Ovation Pico WTA System (NuGen Inc., Leek, The Netherlands) for amplification). The TaqMan probe-based system was applied in combination with the TaqMan Universal PCR Master Mix (Applied Biosystems, Zug, Switzerland). The probe IDs are given later. Cycling was performed with 7500 Fast and Step OnePlus Real-Time PCR Systems (both from Applied Biosystems, Zug, Switzerland).

Probe IDs for quantitative PCR
These were the probes:

Immunofluorescent staining
For single-cell staining, freshly sorted cells were allowed to air-dry on poly-L-lysine-coated slides and stored at -20°C. The dried cells were blocked with PBS, 2.5% goat serum, and 0.05% NaN 3 for 60 minutes under UV light. The UV treatment was used to attenuate residual fluorescence from bound FACS antibodies. Primary antibody staining was performed overnight at 4°C by using Krt18 antibody (1:1,000) and Krt14 antibody (1:500) as luminal and basal cell markers, respectively. Secondary antibody staining was carried out for 60 minutes at room temperature by using anti-guinea pig Ig-Alexa488 (1:500) and anti-rabbit Ig-Alexa546 (1:1,000). Hoechst 33342 (0.5 μg/ml; Invitrogen, Zug, Switzerland) staining was performed for 10 minutes at room temperature. The stained cells were mounted with ProLong Gold antifade reagent (Invitrogen, Zug, Switzerland) and imaged with a Z1 microscope (Zeiss, Feldbach, Switzerland) at 63× magnification. No primary antibody was added as a negative control.

Immunohistochemistry
The fourth mammary glands of parous or age-matched virgin control mice were collected 40 days after weaning from mice in estrus. The glands were fixed in 4% PFA and embedded in paraffin. For immunostaining with ERα and PR, the sections were dewaxed and subjected to antigen retrieval by boiling in 10 mM citrate buffer for 10 minutes. Subsequently, the sections were cooled to room temperature, quenched for 10 minutes with PBS and 3% H 2 O 2 , and blocked for 30 to 60 minutes with PBS and 2.5% NGS. ERα and PR primary antibody staining was performed overnight at 4°C at a 1:1,000 (ERα) and a 1:200 (PR) dilution. Secondary antibody staining was carried out for 30 to 60 minutes at room temperature with biotinylated anti-rabbit IgG. Gene-set enrichment analysis (GSEA) for Wnt and Notch signaling, as defined by the expression of Wnt or Notch target genes (see Methods) showed a decrease in the Wnt/Notch signaling ratio (strongly negative/positive normalized enrichment scores (NES)), which was specific for the basal stem/progenitor cells. Significance was determined by a cut-off of a nominal P value < 0.01, a false-discovery rate (FDR) < 0.05, and a family-wise error rate (FWER) < 0.05, as suggested by the GSEA homepage [28]. The gene set for Wnt signaling consisted of the reported Wnt targets in mammalian systems (derived from the Stanford Wnt homepage) and is depicted in Additional file 3. The gene set for Notch signaling was derived from the Broad Institute v3.0 and is termed Nguyen Notch1 targets. The number of permutations was set to 1,000, and the permutation type was set to "gene set," because fewer than seven samples existed per phenotype [28]. Otherwise, the default settings were used.
Immunohistochemistry for versican, β-catenin, and p21 was performed on the Ventana DiscoveryXT instrument (Roche Diagnostics, Rotkreuz, Switzerland) by using the Research IHC Dap Map XT procedure. In brief, dewaxing was performed in the machine, and slides were pretreated with mildCC1 (versican and p21) or standardCC1 (β-catenin) (Roche Diagnostics, Rotkreuz, Switzerland). Primary antibodies were incubated at 37°C for 1 hour at the following dilutions: versican, 1:50; p21, 1:50; and β-catenin, 1:500. As secondary antibody, Immpress, an anti-rabbit HRP conjugated polymer, was used. All slides were counterstained with Hematoxylin II (Roche Diagnostics, Rotkreuz, Switzerland) and Bluing Reagent (Roche Diagnostics, Rotkreuz, Switzerland). Images were captured at 20-fold (ERα or PR) or 40-fold (versican, β-catenin, p21) magnification by using an Eclipse E600 microscope (Nikon, Egg, Switzerland). For quantification of ERα and PR positivity, at least 2,000 epithelial cells per mouse were counted. For quantification of nuclear β-catenin and p21, at least 500 epithelial cells per mouse were counted. The quantification of versican was performed with the MATLAB software by using color segmentation based on Mahalanobis distance to determine the pixels with a particular RGB-color distribution [31].

Determination of blood progesterone concentrations
Blood was collected from the right atrium of mice in estrus in EDTA-covered tubes (Sarstedt, Numbrecht-Rommelsdorf, Germany), and plasma was extracted by centrifugation at 1,500 g for 15 minutes at 4°C. Progesterone concentrations were assessed with ELISA, as specified by the manufacturer's guidelines (DRG, catalog no. EIA-1561, Marburg, Germany).

Statistics
The limited dilution transplantation data were analyzed statistically as published previously [18]. The two-tailed unpaired Student t test was used to determine statistical significance of comparisons.

Motif Activity Response Analysis (MARA)
The MARA model [32] combines knowledge of geneexpression levels (measured by microarray) with transcription-factor binding sites to answer the question of which transcription factors are driving expression changes in mammary stem/progenitor cells in parous as compared with age-matched virgin control mice. Specifically, logexpression levels of all genes present on the microarray were modeled as linear combinations of transcription factor activities. The coefficients of these combinations were determined by the number of transcription-factor binding sites in the proximal promoter regions. For each transcription-factor binding motif m and each sample (microarray) s, we estimated the activity A ms with the corresponding error. Furthermore, we quantified the significance of activity change of each binding motif in parous as compared with virgin control mice.

Results
Early pregnancy decreases luminal Sca1 + cells, but does not change the proportions of the other mammary epithelial cell subpopulations To investigate the influence of early parity on the proportions of mammary epithelial cell subtypes, we first established FACS profiles of epithelial cell subpopulations in virgin FVB control mice [21,26]. Luminal CD24 +High Sca1 + cells, luminal CD24 +High Sca1cells, basal CD24 +Low Sca1 -CD49f High cells, and basal CD24 +Low Sca1 -CD49f Low myoepithelial cells were isolated from these mice ( Figure 1A). Use of the established cell markers keratin 18 (Krt18) and keratin 14 (Krt14) confirmed the luminal and basal origin of the isolated cell subpopulations, and qPCR for CD49f and Sca1 affirmed the purity of the isolated cell subpopulations (see Additional file 4). Because basal CD49f High cells are considered to be enriched for basal mammary stem/ progenitor cells in virgin mice [19,21], the term "basal CD49f High cells" is used synonymously with "basal stem/ progenitor cells" throughout this article. To extend these analyses to parous mice, we confirmed that involution was complete 28 and 40 days after weaning (see Additional file 5). To allow a margin of safety, 40 days after weaning was used for all subsequent cell-isolation experiments in a standardized parturition protocol with mating at 42 days ( Figure 1B). The FACS profiles of epithelial cell subpopulations from parous mice and age-matched virgin control mice were similar ( Figure 1C, D; Additional file 1) with the exception of luminal Sca1 + cells, which decreased by about 50% in parous mice (P = 0.02). Of note, luminal Sca1 + cells have been shown to be enriched for hormone receptorpositive cells [17]. This was verified by qPCR (see Additional file 6). These data demonstrate that the adopted experimental procedure permits the isolation of all epithelial cell subpopulations from parous mice, including basal stem/progenitor cells, at levels adequate for transcriptomic and functional analyses.
Parity upregulates differentiation genes in all cell subpopulations and decreases the Wnt/Notch signaling ratio in the basal stem/progenitor cell subpopulation Next we investigated the effects of early pregnancy on the gene-expression profiles of the isolated mammary epithelial cell subpopulations. To control for the effect of the cell-isolation procedure on gene expression, we performed first a transcriptome and cluster analysis in non-FACS-sorted total mammary cell suspensions from age-matched virgin and parous mice (Figure 2A). The analysis showed that pregnancy induces an upregulation of many immunoglobulin and differentiation genes (for example, whey acidic protein Wap, and carboxyl ester lipase Cel) and a downregulation of growth factors (for example, amphiregulin Areg, betacellulin Btc, tumorassociated calcium signal transducer 2 Tacstd2), and extracellular matrix (ECM) elements (for example, laminin, gamma 2 Lamc2, desmocollin 2 Dsc2). These data are consistent with the published pregnancy-induced gene signature determined in snap-frozen rodent mammary glands [9,10]. Thus, the gene signature was not lost during the isolation procedure, confirming the validity of our experimental system. Furthermore and most interestingly, a novel 3.4-fold downregulation of the Wnt signaling protein Wnt4 was observed in cells from parous mice (Figure 2). These microarray data were validated by qPCR for four genes, including Wnt4. Importantly, the expression of the luminal marker Krt8 was not altered on parity, demonstrating that the observed changes in gene expression were independent of unspecific alterations in total luminal cell numbers ( Figure 2B).
Because the number of isolated basal stem/progenitor cells for gene-profiling analysis was limited compared with the other cell subpopulations (that is, myoepithelial, luminal Sca1 -, and luminal Sca1 + cells), it was important to evaluate the influence of different cell numbers on transcriptome analysis. As shown in Additional Material, we found that it is valid to use cell numbers in the range of 2,000 to 50,000 cells for comparison of transcriptomes from different mammary cell subpopulations (Additional file 6). Furthermore, qPCR analysis of nonamplified and amplified cDNA for known basal (Krt14) and luminal (Krt8; Krt19) cell marker genes confirmed that the amplification process of the microarray analysis was unbiased (Additional file 6).
In the subsequent transcriptome analysis of FACSsorted mammary epithelial cell subpopulations from agematched virgin control and parous mice, all mammary epithelial cell subpopulations, except immune celldepleted stromal cells, showed parity-induced changes in gene expression with by far the most prominent effects being scored for basal stem/progenitor cells ( Figure 3A). Furthermore, although differentiation genes were upregulated in all epithelial cell subpopulations, the strongest prodifferentiation effects were seen in basal stem/progenitor cells from parous mice. For example, casein alpha 1 (Csn1s1), casein beta (Csn2), and lactotransferrin (Ltf) were upregulated 41-, 19-, and 14-fold, respectively ( Figure 3B). Apart from these differentiation genes, the Notch co-activator Maml2 was found among the 10 most-upregulated genes (by fold change and/or by P value) in basal stem/progenitor cells from parous mice. In contrast, the 10 most-downregulated genes included the Wnt target and co-receptor Lgr5, the Wnt target and epithelial stem cell marker keratin 15 (Krt15) [33], and the Wnt targets versican (Vcan) and Igfbp3 [34] ( Figure 3B). Extension of the analysis to all data for signaling-pathway genes revealed further Wnt target genes, which were downregulated (for example, CD44 and follistatin (Fst) [35,36]), and the Wnt inhibitor Sfrp1, which was upregulated in basal stem/progenitor cells from parous mice. Moreover, in the same epithelial cell subpopulation of parous mice, the Notch ligand Jag2 and the Notch target genes Gata3, Id3, and Dusp1 were upregulated, whereas the Notch inhibitor Itch and the Maml2 RBP-J complex inhibitors Cdk8 and Pak1 [37] were downregulated ( Figure 3B). These gene-expression profiling data were validated with qPCR for several selected genes ( Figure 3C). Furthermore, one of the classic Wnt target genes (versican) was examined on the protein level and found to be strongly downregulated in the basal compartment of mammary glands from parous mice ( Figure 3D). Thus, the data demonstrate strong upregulation of differentiation genes, a downregulation of Wnt target genes, and an increase in Notch signaling in the basal stem/progenitor-cell subpopulation of parous mice.
Further verification of these conclusions came from Motif Activity Response Analysis (MARA) of transcription factor activities (see Methods) [38]. The canonical Wnt transcription factor LEF1/TCF7 was shown to have significantly decreased activity in basal stem/progenitor cells from parous mice ( Figure 4A), which is consistent with the observed decrease in canonical Wnt signaling in the previous gene-expression profile analysis.
Final confirmation of parity-induced downregulation of Wnt signaling in mammary glands was provided by immunohistochemical staining of mammary gland sections from age-matched virgin control and parous mice for β-catenin. As illustrated in Figure 4B, nuclear β-catenin was observed in basal but not in luminal mammary epithelial cells of virgin mice. In parous mice, the proportion of basal mammary epithelial cells positive for nuclear β-catenin was significantly decreased ( Figure 4C). Because nuclear β-catenin is necessary for Wnt target gene expression, this finding represents an additional verification of parityinduced downregulation of Wnt signaling (Figure 3).

The parity-induced decrease in Wnt/Notch signaling ratio is specific for the basal stem/progenitor cell subpopulation
To assess whether the observed decrease in canonical Wnt and increase in Notch signaling were specific for basal stem/progenitor cells, we next investigated the enrichment of Wnt/Notch signaling genes over all genes altered in the various FACS-sorted mammary epithelial cell subpopulations from parous as compared with virgin control mice. Such analysis showed that canonical Wnt signaling was significantly downregulated in basal stem/progenitor but not in luminal Sca1or luminal Sca1 + cells from parous mice (Table 1; see Additional file 7). In myoepithelial cells, a trend toward a decrease in Wnt signaling was not significant when applying a very stringent cut-off (see Table 1 legend) and was probably due to contamination of the myoepithelial cell subpopulation with basal stem and/or progenitor cells. Similarly, Notch signaling was found to be significantly upregulated in basal stem/progenitor cells but in no other mammary epithelial cell subpopulation (Table 1). Whereas downregulation of Wnt and upregulation of Notch signaling were specific for basal stem/progenitor cells, similar enrichment analyses for genes involved in other signaling pathways revealed that the previously reported upregulation of the p53-p21 pathway [39] occurred in all epithelial cell subpopulations tested (Additional file 7).
Further bioinformatics analysis of the data with different software [40] provided a second line of verification of the specific decrease in Wnt signaling in basal stem/ progenitor cells after parity. Performing transcriptionfactor activity analyses based on target gene expression, we found TCF3, an inhibitor of the canonical Wnt signaling pathway [41], to be the transcription factor with the highest z-score (Z = 3.521) and the protooncogene MYC to have a very low z-score (Z = -4.108) in basal stem/progenitor cells but not in other mammary epithelial cell subpopulations of parous mice ( Figure 4D, z-score defined as (x-μ)/sd and used as a measure for transcription factor activity based on the expression levels of target genes). These findings confirm the results obtained in the MARA analysis. Furthermore, inhibition of MYC leads to upregulation of Sfrp1, which in turn inhibits canonical Wnt signaling [42]. Upregulation of Sfrp1 was observed in the microarray analysis and validated with qPCR ( Figure 3B, C), thus directly reflecting the expected effects of the bioinformatic predictions.
Downregulation of canonical Wnt signaling and MYC activity would be expected to decrease the propensity for cancer, and indeed, in an analysis of biofunctions, a marked and consistent decrease in cancer-associated functions, was observed for basal stem/progenitor cells ( Figure 4E), but no other mammary epithelial cell subpopulation. This potential anticancer phenotype of basal stem/progenitor cells was underscored by gene-enrichment analyses on all available pathway gene sets (see Additional file 8), which showed a strong downregulation of proliferation-and tumorigenesis-associated gene sets.
Parity decreases the in vitro clonogenic potential to the greatest extent in the basal stem/progenitor cell subpopulation Because decreased Wnt signaling and increased Notch signaling have been shown to decrease in vitro and in vivo proliferation of basal stem/progenitor cells [43,44], we next assessed the in vitro colony-formation capacities of mammary epithelial cell subpopulations from parous and agematched virgin mice ( Figure 5A). In virgin control mice, luminal Sca1cells had the highest colony-formation capacity, with an average of 107 colonies per well ( Figure 5B). This strong clonogenic potential suggests a pronounced progenitor identity of luminal Sca1cells and is consistent with previous observations in younger virgin mice [16,17]. A high colony-formation capacity was also observed for the basal stem/progenitor cells of virgin mice (63 colonies per well) ( Figure 5B), which is consistent with the notion that CD49f High cells contain a high proportion of basal progenitor cells as well as putative mammary stem cells [45]. With the exception of luminal Sca1cells, the colonyformation capacities of all epithelial cell subpopulations were lower in parous mice than in age-matched virgins ( Figure 5B). Thereby, by far the most pronounced difference was observed in basal stem/progenitor cells ( Figure 5B). A substantial decrease in the colony-formation capacity was also seen for the myoepithelial cell subpopulation, which also contains basal progenitor cells. Of note, basal stem/progenitor cells from parous mice did not die but remained as quiescent single cells or divided only once during 5 days of culture ( Figure 5C).
The reduced progenitor potential of basal mammary epithelial cell subtypes and especially of basal stem/ progenitor cells from parous mice was further confirmed by phenotypic analysis of colonies by colony size (cell number) and immunophenotyping of luminal and basal markers Krt18 and Krt14, respectively. The largest difference in colony size was observed for the basal stem/ progenitor cell subpopulation, where parity induced a decrease in colonies of ≥ 20 cells from 66% to 9% (Table 2A; Figure 5C). Considerably smaller parityinduced reductions in colony size were observed for the other epithelial cell subpopulations, although a substantial decrease was also seen for basal myoepithelial cells (Table 2A). With regard to Krt18/Krt14 double positivity, parity induced a reduction in Krt18/Krt14 doublepositive colonies derived from basal stem/progenitor cells from 26% to 0 (Table 2A). In contrast, although Krt18/Krt14 double-positivity decreased by 27%, 57% of colonies derived from luminal Sca1cells of parous mice maintained double positivity (Table 2A). No effects of parity on Krt18/Krt14 double positivity were observed for colonies derived from luminal Sca1 + or basal myoepithelial cells (Table 2A), whereby myoepithelial cells did not give rise to double-positive colonies even when originating from virgin control mice. Thus, consistent with the basal mammary stem/progenitor cell subpopulation-specific reduction in the Wnt/Notch signaling ratio, parity decreased colony-formation capacity, in vitro proliferation potential (colony number and size), and Krt18/Krt14 double positivity most prominently in the basal stem/progenitor cell subpopulation.

Parity decreases the in vivo reconstitution efficiency of the basal stem/progenitor cell subpopulation
To test their proliferation potential in vivo, we transplanted basal stem/progenitor cells into deepithelialized mammary glands ("cleared fat pads") [12]. It already was demonstrated that basal CD49f High cells have the highest mammary gland reconstitution ability of all mammary cell subpopulations in virgin mice [16,19]. Transplantation of FACS-isolated basal stem/progenitor cells into cleared fat pads demonstrated a significant decrease in the number of large outgrowths (≥ 10% and ≥ 25% of fat pad filled), indicating a decrease in the in vivo proliferation potential. Interestingly, when assessing also for rudimentary outgrowths (≥ 3% of fat pad filled), no significant difference in the number of outgrowths was observed between parous and virgin donors (Table 2B). Apart from the change in size, no qualitative differences were apparent between outgrowths from virgin and parous donors. In both cases, ductal as well as lobular structures were formed. Hence, early parity led to a reduction in in vivo mammary repopulating efficiency of FACS-isolated basal stem/progenitor cells, whereas their ability to differentiate into different mammary epithelial structures was maintained.
Decrease in the proportion of Wnt4-secreting cells after early parity can explain decreased Wnt signaling and reduced proliferation capacity in basal mammary stem/progenitor cells Having observed most of the expected functional consequences of a decrease in the Wnt/Notch signaling ratio in basal stem/progenitor cells from parous mice, we finally examined the possible cause of parity-induced reduction in canonical Wnt signaling and proliferation capacity in basal stem/progenitor cells. Because parity induced a greater than threefold decrease in Wnt ligand Wnt4 gene expression (Figure 2A, B), and Wnt4 is known to be secreted in response to progesterone by hormone-sensing luminal cells [46], thus inducing canonical Wnt signaling in mammary stem/progenitor cells [47], a parity-induced decrease in estrogen/progesterone-sensitive luminal cells could explain the overall decrease in Wnt signaling in mammary stem cells. This hypothesis is supported by the reduction in the proportion of luminal Sca1 + cells isolated from parous mice ( Figure 1D) and by the demonstration that luminal Sca1 + cells are hormone receptor positive (Additional file 6) [17]. Furthermore, immunohistochemical analysis of mammary gland sections for estrogen receptor alpha (ERα) and its target progesterone receptor (PR) showed a twofold decrease in ERα-and a threefold decrease in PR-positive cells in parous compared with agematched virgin control mice ( Figure 6A, B). These results were additionally verified by qPCR in total mammary cell suspensions ( Figure 6C). Notably, expression of the luminal marker Krt8 was similar in cell suspensions from parous and age-matched virgin control mice, supporting the conclusion of a specific decrease in hormone receptorpositive cells rather than a general cell loss after pregnancy. Furthermore, parity-induced reduction in progesterone-stimulated Wnt4 expression was independent of blood progesterone, because average blood progesterone concentrations were similar in parous mice and agematched virgin control mice in estrus (see Additional file 9). Finally, supplementation of the culture medium with recombinant Wnt4 stimulated in vitro proliferation capacity of basal myoepithelial and basal stem/progenitor cells from parous mice by +138% ± 22% and +140% ± 17%, respectively ( Figure 7A). In contrast, no significant effects of recombinant Wnt4 on the colony-formation capacity of luminal Sca1 -(+3.7% ± 3.0%) and luminal Sca1 + cells from parous mice (-5.1% ± 2.9%) were observed. These data strongly suggest a causal relation between reduced number of luminal progesterone receptor-positive/Wnt4-secreting cells and decreased Wnt/ Notch signaling and proliferation potential of basal stem/ progenitor cells after early pregnancy ( Figure 7B).
In conclusion, all collected data indicate that early parity induces a decrease in luminal hormone-sensing ERαand PR-positive cells, which leads to decreased Wnt4 expression levels and to reduced Wnt signaling in basal stem/progenitor cells. Consistent with the decrease in the Wnt/Notch signaling ratio, differentiation is promoted and proliferation inhibited in basal stem/progenitor cells of parous mice.

Discussion
This study demonstrates that a history of early pregnancy changes the gene-expression profiles and functional properties of mammary epithelial cell subpopulations in  Notch signaling ratio provides direct experimental evidence for the hypothesis that early pregnancy changes the "genomic signature" of mammary stem/progenitor cells [7], causing their differentiation and reducing their proliferation potential. Furthermore, the data indicate a novel causal relation between parity-induced reduction in hormone-sensing and Wnt4-secreting luminal cells and altered biofunctions in basal stem/progenitor cells. The basal CD49f High cells, as isolated in this study, are a subfraction of basal epithelial cells [19,21]. Basal CD49f High cells have been demonstrated previously to be enriched in mammary repopulating units (MRUs) (known as mammary stem cells (MaSCs)) [16,19] and to correspond to the Lin -CD24 + CD29 High epithelial cell subpopulation isolated by an alternative method [18,19,45]. However, the isolated basal CD49f High epithelial cell subpopulation represents a heterogeneous cell fraction containing, in addition to MaSCs, basal progenitor cells and possibly mature myoepithelial cells [45]. Progenitor cells can be characterized in vitro by their colony-formation capacity [22,23], whereas MaSCs have traditionally been defined by their in vivo regenerative capacity [16][17][18]. The observed effects of parity on the in vitro and in vivo proliferation capacity of the CD49f High cell subpopulation (Table 2) suggests that both basal progenitor cells and basal MaSCs are the target of an early pregnancy within the mammary epithelium. Although the dramatic decrease in the in vitro proliferation capacity of the CD49f High cell subpopulation ( Figure 5) indicates a predominant effect of an early pregnancy on basal progenitor cells, the additional reduction in large in vivo outgrowths (≥ 10% of fat pad filled) and the prevalence of rudimentary outgrowths (≥ 3% of fat pad filled) after parity (Table 2) suggest that isolated basal MaSCs are also affected by an early pregnancy. Because basal mammary stem and progenitor cells are closely related and likely to be interdependent in their proliferation potentials, our data do not permit a definite discrimination between basal stem and progenitor cells as primary targets of pregnancy. Therefore, we adhered to the combined term basal stem/progenitor cells throughout this study.
We found the p53-p21 pathway to be enriched to a similar extent in all mammary epithelial cell subpopulations (Additional file 7) [39], and hence, although parity-caused induction of the p53-p21 pathway may explain the relatively modest decrease in in vitro colony-formation potential of the luminal Sca1 + cell subpopulation ( Figure 5), it may contribute but cannot account for the almost complete proliferation block in basal stem/progenitor cells. The most prominent parity-induced alterations in gene expression in basal stem/progenitor cells were downregulation of the Wnt-signaling pathway, upregulation of the Notchsignaling pathway, and upregulation of differentiation genes. Decreased Wnt signaling in basal stem/progenitor cells from parous mice was verified on the protein level by measuring versican and nuclear β-catenin expression ( Figures 3D and 4B, C). Wnt signaling has been shown to promote long-term expansion of cultured Lin -CD24 + CD29 High cells and to provide a competitive advantage in mammary gland reconstitution assays [43]. The latter is especially true for the expression of the classic Wnt target Lgr5 [48], which was found in our study to be downregulated in the basal stem/progenitor cell subpopulation after parity. Notably, Wnt signaling inhibition was demonstrated to have an antiproliferative effect in CD29 High cells [43]. Furthermore, Notch signaling was observed to reduce in vitro and in vivo proliferation of CD29 High cells while promoting their differentiation [44,49]. Because the differentiation processes of stem and progenitor cells in many organs and in several model systems are dependent on the Wnt/ Notch signaling ratio [50], an overall reduction in Wnt/ Notch signaling ratio would be expected to have a dramatic antiproliferation and prodifferentiation effect in mammary basal stem/progenitor cells. This is exactly what we observed in the CD49f High cell subpopulation of parous mice. Thereby, the overall conclusion of our study is strengthened by the fact that all assays used (that is, transcriptome analysis, bioinformatics transcription factor, and gene-enrichment analyses, in vitro colony-forming assay, in vivo transplantation assay, and immunohistochemistry) pointed into the same direction. Thus, analysis of specific mammary epithelial cell subpopulations allowed the discovery of a decrease in the Wnt/Notch signaling ratio, which so far is the only plausible explanation for the observed differentiation burst and dramatic proliferation block experienced by basal stem/progenitor cells of parous mice.
As possible explanations for the parity-induced decrease in Wnt signaling, we found a marked increase in the activity of the Wnt repressor TCF3 in basal stem/progenitor cells and a more than threefold reduction in expression of the secreted Wnt ligand Wnt4 in total mammary cells from parous mice. The latter corresponded to a similar decrease in Wnt4-secreting [46] and estrogen and progesterone receptor-positive luminal cells. Notably, a similar decrease in progesterone receptor α-positive cells after parity has also been observed in human breast epithelium [51]. Hence, mechanistically, early parity decreases the hormone responsiveness of the mammary gland in mice by decreasing the number of estrogen/progesterone receptor-positive luminal cells. This reduces the paracrine signaling cascade mediated by Wnt4, inducing TCF3dependent repression [52] and/or primary downregulation [47] of canonical Wnt signaling and secondary (reactive) upregulation of Notch signaling in basal stem/progenitor cells.
As a final consequence, proliferation is repressed and basal stem/progenitor cells differentiate. This mechanistic model is supported by the ability of recombinant Wnt4 to rescue the proliferation defect of basal stem/ progenitor cells from parous mice in vitro (Figure 7).
Our findings in specific mammary epithelial cell subpopulations are in part consistent with and in part contradictory to studies in entire breasts/mammary glands, total mammary cells, or total mammary epithelial cells. With respect to the transcriptome analysis, our studies in total mammary cells agree with previous reports in the entire mammary glands [9,10]. However, in intact mammary glands or total mammary cells, the additional presence of stromal and dominant epithelial cell subtypes might mask the detection of key signaling-pathway changes. Indeed, our study demonstrates that isolation of specific mammary epithelial cell subpopulations is a prerequisite for the detection of a decrease in Wnt/ Notch signaling ratio in basal stem/progenitor cells. A similar masking effect by stromal and dominant epithelial cell subtypes (for example, strong clonogenic luminal Sca1cells) might also explain why a previous study with a similar early-pregnancy protocol did not observe a parity-induced reduction in in vitro proliferation of total mixed mammary cells [20]. Controversial results have also been reported with respect to the effect of parity on the in vivo mammary-repopulating capacity. Hence, although Britt et al. [21] found no effects of late pregnancy (9 weeks) on mammary-repopulating units (MRUs) in total mammary epithelial cells, Siwko et al. [20] (early-pregnancy protocol) observed a parityinduced reduction in the mammary-repopulating capacity of total mixed mammary cells. With a similar earlypregnancy protocol and the same cut-off for mammary gland outgrowth (≥ 10% of fat pad filled), our findings in isolated mammary basal stem/progenitor cells appear consistent with the observations of Siwko et al. However, given that our studies were performed with isolated mammary basal stem/progenitor cell subpopulations, our findings are not directly comparable with and neither confirm nor contradict previous studies using total mammary (epithelial) cells [20,21]. Furthermore, the fact that the number of smaller outgrowths (≥ 3% of fat pad filled) was unchanged after parity suggests that MRUs survive after pregnancy despite their reduced reconstitution efficiency. This conclusion is in line with the recent demonstration that Wnt-responsive mammary epithelial stem cells persist after parity in Axin2 reporter mice [15].
It is intriguing to speculate that marked growth inhibitory effects and the downregulation of canonical Wnt signaling in basal stem/progenitor cells account, at least in part, for the cancer-protective effect of early pregnancy. Increases in canonical Wnt signaling have been linked repeatedly to oncogenesis [53,54]. Moreover, downregulation of the Wnt inhibitory protein Sfrp1 and overexpression of the Wnt target versican have been associated with carcinogenesis [55,56]. In the transcriptome analysis reported here, Sfrp1 was upregulated and versican downregulated, thus supporting a parity-induced anticarcinogenic effect. Moreover, Li et al. [57] showed that transgenes encoding components of the Wnt signaling pathway preferentially induce mammary cancers from progenitor cells. Hence, the contribution of decreased canonical Wnt signaling in basal stem/progenitor cells to the cancer-protective effect of early pregnancy may be in conjunction with other tumor-suppressing mechanisms, such as parity-induced induction of p53 [58,59]. Thereby, the increase in the TCF3 repressor activity in basal stem/ progenitor cells is expected to elevate the threshold further for the activation of tumorigenic Wnt signaling [52]. In addition to decreased Wnt signaling in basal stem/progenitor cells, decreased ERα-and PR-positive cells could also be a mechanism for the breast cancer-protective effect of an early pregnancy. This is especially relevant, given the specific protective effect of pregnancy against ER/PR-positive tumors [60].
Furthermore, mammary epithelial cell differentiation per se has been suggested to exert a breast cancer-protective effect. This has been challenged, however, by the observation that differentiation-causing agents such as placental lactogen and perphenazine failed to protect against carcinogenesis in rodents [3,61]. Also, the hypothesis of a potential breast cancer-protective effect of mammary stromal cells [62] is not supported by our study, because stromal cells exhibited by far the fewest parity-induced gene-expression changes. However, our stromal cell subpopulation was not homogenous and devoid of immune cells, which may have masked some parity-induced alterations. In any case, the parity-induced downregulation of the Wnt/Notch signaling ratio in basal stem/progenitor cells represents a possible important mechanism for the breast cancer-protective effect of early pregnancy.

Conclusions
This study identified downregulation of the Wnt/Notch signaling ratio in basal stem/progenitor cells as the dominant early parity-induced alteration of gene expression in mice. This change in gene expression is specific for basal mammary stem/progenitor cells, is associated with proliferation defects in vitro and in vivo, and is probably caused by an early parity-induced decrease in hormone-sensitive and Wnt4-secreting luminal cells. Importantly, because a similar reduction in progesterone receptor α-positive luminal cells has been reported in women [51], parityinduced alterations in Wnt/Notch signaling pathways may also occur in human basal stem/progenitor cells. Testing whether Wnt inhibitors mimic early parity-induced breast cancer protection warrants further investigation.

Additional material
Additional file 1: Flow-cytometric separation of mammary epithelial cells from virgin control and parous mice. Representative flowcytometry pseudocolor plots depicting the first steps in the gating strategy used to eliminate doublets, cell clumps, dead cells (DAPI bright) and white blood cells (CD45 + ) during the procedure for isolating mammary epithelial cell subpopulations from virgin control (A) and