Introduction

The mammary gland comprises a heterogeneous collection of epithelial cells that, in response to hormones and through interactions with stromal elements, undergo dynamic changes during the lactational cycle (reviewed in LaMarca and Rosen, 2008). Defining the molecular and cellular basis of this process is essential for understanding mammary development and associated defects, including breast cancer. The fact that breast cancer stem cells share some characteristics with mammary stem cells (Al-Hajj et al., 2003; Visvader and Lindeman, 2008) and that mutations in a number of the genes associated with mammary development have also been implicated in breast cancer (Usary et al., 2004; Bogorad et al., 2008) suggests that studies on mammary development may help define the early events that precede breast tumourigenesis.

Experimental animal models have been invaluable in mammary gland research (reviewed in Hennighausen and Robinson, 2001), enabling the identification of mammary stem (Smalley et al., 2005; Shackleton et al., 2006; Stingl et al., 2006); and luminal progenitor cells (Sleeman et al., 2006, 2007; Asselin-Labat et al., 2007) and the molecular changes that are associated with the different phases of post-pubertal mammary development (Master et al., 2002; Clarkson et al., 2004; Stein et al., 2004). Mouse models displaying defects in these processes have also been very useful, for example, Gata-3, Elf-5 and Stat5a knockout mice have helped define the cellular events involved in the development of luminal and secretory cells, respectively (Asselin-Labat et al., 2007; Oakes et al., 2008; Yamaji et al., 2009), whereas PrlR and Gal knockout mice have shed light on the molecular events associated with the secretory activation stage of lactation (Naylor et al., 2005).

Disruption of the breast cancer susceptibility gene BRCA1 results in defects in mammary epithelial differentiation in vitro (Kubista et al., 2002; Furuta et al., 2005) and defective lobular-alveolar development in the mammary gland in vivo, based on analysis of breast tissue from BRCA1 mutation carriers or mice in which the Brca1 gene has been deleted from the mammary gland or in which the mammary gland expresses a dominant-negative Brca1 mutant (Jernstrom et al., 1998; Xu et al., 1999; Russo et al., 2001; Brown et al., 2002; Triplett et al., 2008). Tumours arising in BRCA1 mutation carriers and Brca1 conditional knockout mice have a distinct histopathological and molecular phenotype, including high grade, high mitotic index and expression of basal and stem cell-associated genes (Lakhani et al., 1998; Sorlie et al., 2003; Liu et al., 2008; Shakya et al., 2008; Wright et al., 2008a, 2008b). It has been suggested that BRCA1-associated tumours arise in the basal or stem cell compartments of the mammary gland (Foulkes et al., 2003), although there is also evidence to suggest that that these characteristics may arise as a consequence of BRCA1 loss in other mammary cell types (Liu et al., 2008; Smart et al., 2008). Although the detailed pathway between BRCA1 disruption and breast tumour development has yet to be fully elucidated, recent findings suggest that tumours arising in human BRCA1 mutation carriers may arise from mammary luminal progenitor cells of the mammary gland (Lim et al., 2009).

To further investigate the role of BRCA1 in mammary development and cancer, we examined the cellular and molecular consequences of mouse Brca1 loss in vitro and in vivo. Here, we show that loss of Brca1 results in a defect in mammary epithelial morphogenesis and lobular-alveolar development that is associated with an accumulation of luminal progenitor cells and differential expression of multiple genes, including c-kit. Functional studies on c-kit suggest that, although it can be regulated by Brca1, it is unlikely to be responsible for the Brca1-associated defect in lobular-alveolar development, and thus is more likely to be a marker than a mediator of the phenotype of these mice.

Results

Accumulation of luminal progenitor cells in the mammary glands of MMTV-Cre Brca1Co/Co mice during pregnancy

Brca1 has been shown to play an important role in mouse mammary epithelial differentiation in vitro (Kubista et al., 2002). To confirm and extend previous studies, we examined the consequences of Brca1 suppression in two mouse mammary epithelial cell lines HC11 and SCp2, and examined the morphological and cellular changes in the mammary glands of MMTV-Cre Brca1Co/Co mice. As shown in Figures 1a and b, repression of Brca1 in both cell lines caused a defect in mammosphere formation. Histological analysis of SCp2 mammospheres further revealed disorganised acini-like structures, relative to controls (Figure 1b). Consistent with this and with previous reports, mammary-specific depletion of Brca1 in vivo resulted in defective lobular-alveolar development, characterised by distended lumen and disorganised alveoli (Xu et al., 1999; Brown et al., 2002; Figure 1c).

Figure 1
figure 1

Suppression of Brca1 results in defective mammary epithelial morphogenesis in vitro and accumulation of luminal progenitor cells in vivo. (a) Brca1 knockdown by siRNA is confirmed in HC11 cells by real-time PCR (top graph). Brca1 mRNA levels were compared in HC11 cells transfected with scrambled (control) and Brca1 siRNA constructs. Bars represent the mean of three independent experiments and error bars represent s.e. Knockdown of Brca1 reduces dome formation in the HC11 in vitro differentiation assay (bottom panel). Dome formation was calculated at day 8 after initiation of the assay and shown here relative to the scrambled control. Error bars represent s.e. (b) Functional disruption of Brca1 impairs prolactin-induced mammary differentiation of SCp2 cells. Differentiation of SCp2 cells transfected with pZeoSV vector (left column) and trBrca1 (right column). Phase contrast images ( × 20 magnification) are shown of SCp2 cells that were grown on matrigel, four days (top row) and six days (middle row) after the addition of lactogenic hormones. H&E-stained sections (bottom row) of cell aggregates at day 6 were photographed using a light microscope at × 200 magnification. Arrows indicate defective morphogenesis. (c) H&E-stained sections of day 1 lactation mammary glands taken from Brca1Co/Co (left column) and MMTV-Cre Brca1Co/Co (right column) mice taken at × 40 (top) and × 100 (bottom) magnifications. Arrows indicate defective morphogenesis. (d) Sorted mammary epithelial cells from MMTV-Cre Brca1Co/Co mice exhibit increased luminal progenitor cells at day 12 of pregnancy. Histograms showing percentage of CD61+ cells in the CD29loCD24+ population isolated from virgin or 12.5-day pregnant Brca1Co/+and MMTV-Cre, Brca1Co/Co mammary glands (mean±s.e.m. of 3–6 animals per group). *P<0.05 (Student's t-test).

Full size image

Although the cellular basis of the in vivo phenotype is unknown, the observed expansion of the luminal progenitor cell population in the breast tissue of human BRCA1 mutation carriers (Lim et al., 2009) led us to hypothesise that there may be an accumulation of luminal progenitor cells in the mammary glands of MMTV-Cre Brca1Co/Co mice. We therefore examined CD61+CD29loCD24+ luminal progenitor cells isolated from virgin and pregnant mammary glands of MMTV-Cre Brca1Co/Co and control mice. These time points were selected, as the luminal progenitor (CD61+) population declines dramatically during pregnancy and is virtually undetectable in late pregnancy and lactation. No significant difference in the number of CD61+CD29loCD24+ progenitor cells was observed in the resting virgin mammary glands of MMTV-Cre Brca1Co/Co mice compared with controls (Figure 1d, left bars). During pregnancy, however, when luminal progenitors are induced to differentiate and transgene activity is augmented, we found that the proportion of CD61+CD29loCD24+ cells was significantly increased in MMTV-Cre Brca1Co/Co mice, relative to littermate controls (Figure 1d, right bars).

Significant changes in the expression profile of mammary glands from MMTV-Cre Brca1Co/Co mice at day 1 of lactation associated with Brca1 deficiency

To begin to investigate the molecular basis of this observation, gene expression profiling was performed on three pools of four mammary glands from MMTV-Cre Brca1Co/Co and control Brca1Co/Co mice at 10 weeks of age (virgin) and day 1 of lactation (Figure 2a), for which Brca1 recombination and depleted expression was verified.

Figure 2
figure 2

Differential expression of genes in the mammary glands of MMTV-Cre Brca1Co/Co mice at day 1 of lactation. (a) Experimental design. Gene expression profile analysis was performed on RNA extracted from the mammary glands of Brca1Co/Co and MMTV-Cre Brca1Co/Co mice at 10 weeks of age (virgin) and day 1 of lactation. Three pools of four individual mice were used to represent each stage and genotype. Data analysis revealed the indicated number of differentially expressed genes. (b) Results of ingenuity analysis showing the most significantly altered cellular and disease processes in MMTV-Cre Brca1Co/Co mice. (c) Comparison of gene expression data in day 1 lactation MMTV-Cre Brca1Co/Co mice compared with controls by microarray (fold change (white bars)) and validating quantitative real-time PCR (relative expression (grey bars)) of eight independently analysed individual mice. Error bars on microarray data represent s.e. of the three pools and on real-time PCR data represent the s.e.m. of eight biological replicates. Statistical analysis was performed using a Student's t-test with ***P<0.001 and **0.001<P<0.01. NS, not significant; qRT–PCR, quantitative real-time PCR.

Full size image

The choice of the day 1 lactation time point was based on the time at which the mammary phenotype was most evident in MMTV-Cre Brca1Co/Co mice, and that it is technically easier to precisely identify mice at day 1 of lactation rather than a specific day in pregnancy. Such precision is essential for molecular profiling studies, given the dynamic molecular changes occurring in the mammary gland during lactational development. Brca1Co/Co mice were used as a control to ensure that our data were not confounded by effects of the Brca1Co/Co locus on the expression of Brca1 splice isoforms. The lactational phenotype is not observed in MMTV-Cre mice (Supplementary Figure 4), indicating that this phenotype is unlikely to be a result of Cre recombinase expression only in the gland.

A robust statistical analysis of microarray data (see Materials and methods) identified 1379 probes that were differentially expressed between MMTV-Cre Brca1Co/Co and Brca1Co/Co mice at day 1 of lactation (Supplementary Table 1). Of these, 786 were upregulated (44 with a greater than twofold change) and 593 were downregulated (3 with greater than twofold change) in MMTV-Cre Brca1Co/Co mice. In contrast, only 38 genes were differentially expressed in the 10-week old virgin mice (Supplementary Table 2).

Validation of multiple differentially expressed genes, including c-kit

To validate individual genes corresponding to the probes identified in the microarray data, quantitative real-time PCR analysis of selected genes was performed, using RNA isolated from 8 of the 12 individual mice that comprised the pools used in the array analysis and normalising to hypoxanthine phosphoribosyltransferase 1 (HPRT). The genes selected for validation spanned a range of fold changes and B-scores, and were components of functional categories that Ingenuity pathway analysis indicated were significantly overrepresented in our gene list. The relative gene changes seen using quantitative real-time PCR and the fold changes given by the microarray analysis were similar, demonstrating the robustness of the microarray data (Figure 2c).

The highest B-score that validated was 16.3 (p21) and the lowest B-score that validated was 1.19 (CD24a), a range of B-scores attributed to over 500 transcripts (Supplementary Table 1). Quantitative real-time PCR was also used to confirm that Brca1 transcript levels were significantly reduced in the mammary glands of MMTV-Cre Brca1Co/Co compared with control Brca1Co/Co mice (data not shown), further validating the experimental design and expression data.

Given the fold increase of c-kit observed in MMTV-Cre Brca1Co/Co glands, together with its likely importance in Brca1-associated tumourigenesis in humans (Nielsen et al., 2004; Lim et al., 2009), our initial studies focused on the expression, regulation and function of c-kit in mammary epithelial cells.

An increase in c-kit immunopositivity in lactating mammary glands of MMTV-Cre Brca1Co/Co mice at day 1 of lactation

As the expression profiling experiments were performed using RNA from whole-mammary glands, the observed changes in gene expression may reflect molecular events in any of the cellular compartments of the mammary gland. To determine where c-kit protein is normally expressed in the lactating mammary gland and the spatial consequences of Brca1 loss, immunofluorescence analysis of c-kit was performed on frozen sections of optimal cutting temperature (OCT) media-embedded glands of MMTV-Cre Brca1Co/Co (n=6) and Brca1Co/Co (n=4) mice at day 1 of lactation. Control Brca1Co/Co mice displayed a distinct lack of c-kit expression, with only sporadic staining in the luminal epithelium of the mammary ducts (Figure 3a). In contrast, the mammary glands of MMTV-Cre Brca1Co/Co mice demonstrated heterogeneous c-kit immunostaining throughout the mammary epithelium (Figure 3a). Some c-kit immunopositivity was observed in the interstitium of mice of both genotypes, perhaps reflecting the presence of mast cells.

Figure 3
figure 3

An increase in c-kit immunopositivity in the mammary glands of MMTV-Cre Brca1Co/Co mice. Immunofluorescence analysis of c-kit (green) and DAPI (blue) nuclear counterstain in frozen sections of day 1 lactating mammary glands. Images represent glands from Brca1Co/Co mice and MMTV-Cre Brca1Co/Co mice. Images were taken at × 200 magnification. Scale bars represent 100 μm.

Full size image

Reciprocal regulation of Brca1 and c-kit in mammary epithelial cells

Although expression profiling and immunohistochemical analysis provides important information on the consequences of Brca1 loss, it does not allow conclusions to be made about the molecular relationship between Brca1 and the differentially expressed genes including c-kit within a given cell. To address this, experiments were performed using the mouse mammary epithelial cell line HC11. In the first instance, the endogenous expression of Brca1 and c-kit was determined during differentiation using real-time PCR. As shown in Figure 4a, Brca1 expression levels were reduced by ∼threefold as the cells underwent differentiation. In contrast, the levels of c-kit were induced by ∼200-fold. This result raises the possibility that Brca1 and c-kit are coordinately, albeit reciprocally regulated.

Figure 4
figure 4

Reciprocal regulation of Brca1 and c-kit in mammary epithelial cells. (a) Endogenous Brca1 and c-kit mRNA expression was measured in HC11 cells throughout 8-day in vitro differentiation with lactogenic stimulus. Bars represent the mean of three independent experiments and error bars represent s.e., and significance was shown to be P<0.05 between days 2 and 4 and <0.005 between days 2 and 8, as determined by Student's t-test. (b) Brca1 knockdown in HC11 cells was achieved with transfection of Brca1 siRNA (grey bars) and compared with scrambled controls (white bars). Knockdown of Brca1 in proliferation was confirmed by quantitative PCR of the Brca1 mRNA and the effect on c-kit mRNA was also quantified. Expression is shown relative to HPRT, and bars represent the mean of three independent experiments. Error bars represent s.e., and significance was determined by Student's t-test with ***P<0.001, **0.001<P<0.01.

Full size image

To address the hypothesis that the increase in c-kit expression is a consequence of Brca1 repression in mammary epithelial cells, Brca1 levels were repressed using small interfering RNA (siRNA). As shown in Figure 4b, repression of Brca1 was associated with an increase in the levels of c-kit mRNA. These data indicate that induction of c-kit may be a consequence of Brca1 repression in the epithelial compartment of the mammary gland.

c-kit does not regulate HC11 mammary epithelial differentiation

Induction of c-kit as a consequence of Brca1 repression in mammary epithelial cells suggests that c-kit may be a mediator of some of the effects of Brca1 loss, including the observed defects in mammary epithelial differentiation. To begin to address this hypothesis, the effect of perturbing c-kit expression on HC11 differentiation was determined. Overexpression, achieved by retroviral transduction and confirmed by real-time PCR, had no effect on the number of HC11 domes formed (Figure 5a) or the expression of β-casein, a marker of terminal differentiation (data not shown). Similarly, knockdown of expression, achieved by siRNA and confirmed by real-time PCR, similarly had no significant effect on the number of HC11 domes (Figure 5b) or the expression of β-casein (data not shown). These data indicate that c-kit does not regulate the differentiation of HC11 mammary epithelial cells. However, this does not preclude a physiological role for c-kit in the mouse mammary gland.

Figure 5
figure 5

c-kit does not regulate HC11 differentiation. Ectopic c-kit expression (a) and knockdown (b) was confirmed in HC11 cells by real-time PCR (left-hand graphs). c-kit mRNA levels were compared in HC11 cells transfected with (a) pBABE (control) and pBABE-c-kit or with (b) scrambled (control) and c-kit siRNA constructs. Bars represent the mean of three independent experiments (relative HPRT) and error bars represent s.e. Dome formation was unaffected by ectopic c-kit expression but was reduced with c-kit knockdown (right-hand graphs). Dome formation was calculated at 8 days after initiation of the assay and shown here relative to the pBABE (a) or scrambled (b) controls. Error bars represent s.e.

Full size image

Discussion

Disruption of the breast cancer susceptibility gene BRCA1 results in defective lobular-alveolar development of the mammary gland and predisposes to breast tumourigenesis, but the molecular and cellular basis of this phenotype is not well understood and has been investigated in this manuscript. Here, we show that loss of Brca1 results in a defect in mammary epithelial morphogenesis that is associated with an accumulation of luminal progenitor cells and with a significant number of gene expression changes.

It is interesting that different lines of MMTV-Cre transgenic mice sometimes display different morphological phenotypes, including defective lobular-alveolar development (Cui et al., 2002; Pylayeva et al., 2009); however, this is usually attributable to the well-documented differences in spatial and temporal expression of the MMTV-Cre transgene and thus recombination-mediated deletion of the target gene (Wagner et al., 2001), rather than to effects of the Cre recombinase. In the case of Brca1, defective lobular-alveolar development is consistently observed using two independent MMTV-Cre lines: line F (Figure 1a) and line A (Triplett et al., 2008), and WAP-Cre Brca1Ko/Co mice (Xu et al., 1999; Triplett et al., 2008), arguing strongly that the observed mammary phenotypes are a consequence of Brca1 loss. This is further supported by our supplementary data showing that the mammary glands of MMTV-Cre mice at day 1 of lactation are indistinguishable from wild-type control mice (Supplementary Figure 4). In the case of more subtle molecular changes, however, we cannot completely exclude a potential contribution of MMTV-Cre on the expression of some genes. This is one of the many limitations of Cre-loxP technology (reviewed in Maddison and Clarke, 2005).

Our data suggest that a differentiation block exists in luminal progenitor cells of the mammary glands of MMTV-Cre Brca1Co/Co mice. These observations support the finding that Brca1-deficient ER-negative primary mammary progenitor cells are defective in their ability to differentiate into ER-positive luminal cells in vitro (Liu et al., 2008) and that aberrant luminal progenitor cells are expanded in breast tissue from human BRCA1 mutation carriers (Lim et al., 2009). Defects in luminal progenitor cells have been observed in other mouse models, including MMTV-Cre Gata-3fl/+ mice, which exhibit an expansion of the CD61+ progenitor population during puberty and pregnancy (Asselin-Labat et al., 2007), and Elf5−/− and Stat5a−/− mice, which show an increase in these progenitors at mid-pregnancy (Oakes et al., 2008; Yamaji et al., 2009). Interestingly, non-parous precancerous tissues from BRCA1 mutation carriers harbour an aberrant luminal progenitor population (Lim et al., 2009), whereas young virgin mice do not (Figure 1d). This is most likely to be due to differences in age and parity, although a more thorough analysis of this, including analysis of mice at different ages and after pregnancy cycles, will be necessary to firmly establish this.

Many of the observed molecular changes, including repression of milk protein genes, are consistent with previously reported expression changes in mammary glands and epithelial cell lines in which Brca1 function has been diminished (Brown et al., 2002; Furuta et al., 2005) and in mammary glands lacking Brca1. However, a comparison of our gene list with the expression changes that occur during normal lactational development of the mouse mammary gland, and with other mouse models of lactational delay, indicate, however, that the molecular changes induced by loss of Brca1 are not solely a reflection of the observed defect in lobular-alveolar development, and that many of the observed changes are likely to reflect other consequences of Brca1 loss. These data, therefore, provide a resource for the identification and analysis of novel functions of Brca1, and for better understanding how Brca1 contributes to normal development and disease.

Genes upregulated more than fivefold in the mammary glands of MMTV-Cre Brca1Co/Co mice at day 1 of lactation include Cdkn1A (p21, up 16.3-fold), a cell cycle regulated that has previously been shown to be upregulated upon Brca1 repression (Hakem et al., 1996) and that has been implicated in breast cancer (Weiss, 2003); interferon-stimulated gene 15 (Isg15, up 10.9-fold) and the gene encoding the Isg15-interacting protein Ubp43 (up eightfold), the former being a ubiquitin-like gene implicated in immune response to viruses and in breast cancer (Bektas et al., 2008); oligoadenylate synthetase 1G (up 9.2-fold) and 1A (up 5.7-fold), an interferon-induced enzyme implicated in nucleotide metabolism and expressed in the mammary gland (Maia et al., 2008) and induced upon irradiation of human breast cancer cell lines (Tsai et al., 2007); and c-kit (up 5.58-fold), a hematopoietic stem cell marker that is overexpressed in undifferentiated mammary tumours (Tsuda et al., 2005) and functions as an oncogene in mammary epithelial cells (Tsuda et al., 2005; Raafat et al., 2007). Taken together, these genes suggest a role for Brca1 in the immune response and cancer.

The finding that multiple cancer-associated genes are overrepresented in the expression profile of MMTV-Cre Brca1Co/Co mice is also interesting. This is in contrast to other mouse models with a mammary phenotype (Naylor et al., 2005), Ingenuity pathway analysis has identified multiple functional themes with statistical significance greater than those associated with cancer (data not shown). This provides additional evidence that the observed changes in gene expression in MMTV-Cre Brca1Co/Co mice are consequences of Brca1 loss rather than those associated with the mammary phenotype. It is also raises the possibility that some of the observed changes also represent early changes in Brca1-associated tumourigenesis.

One of the most significantly overrepresented genes (up 5.58-fold), participating in the cancer network, is c-kit (Figure 2c). c-kit has recently been identified as a marker of luminal progenitor cells in the human breast (Lim et al., 2009). Although making comparisons across species and at different stages of lactational development must be done with caution, it is interesting to note the overrepresentation of c-kit in the mammary glands of MMTV-Cre Brca1Co/Co mice at day 1 lactation is consistent with our fluorescence activated cell sorting (FACS) analyses showing an accumulation of luminal progenitor cells during pregnancy (Figure 1b). Although analysing c-kit expression through other developmental time points would facilitate comparisons between the molecular and cellular profiling experiments, these data do support the notion that luminal progenitor cell population is expanded in Brca1-deficient mammary glands. It would also be interesting to examine the luminal progenitor population in mammary glands after involution to determine how stable the observed cellular changes are.

To begin to investigate whether there is a molecular link between Brca1 and c-kit, we analysed the expression of both genes during mammary epithelial differentiation, and the consequences of Brca1 repression on c-kit expression, in vitro. Our data suggest that there is a regulatory link between Brca1 and c-kit, albeit probably indirect. The c-kit promoter and 3′ UTR regions have been well characterised by others and many regulators, including the transcription factor Myb (Vandenbark et al., 1996) and the microRNA 221 (Felli et al., 2005) have been described. Future studies aimed at elucidating the molecular basis of Brca1-mediated regulation of c-kit are currently underway.

Although Brca1 repression was clearly associated with the accumulation of luminal progenitor cells and increased c-kit levels, and Brca1 in turn regulated the expression of c-kit in mammary epithelial cells, it was unclear whether the observed overexpression of c-kit was a bystander or mediator of the observed defect in lobular-alveolar development in Brca1-deficient mice. Experiments in HC11 cells in which c-kit was overexpressed or knocked down indicated that c-kit may not have a direct role in regulating epithelial differentiation, at least in these cells. There are at least two possible explanations for this result, first, it may be that a role for c-kit in mammary differentiation is p53 dependant and thus a phenotype would not be observed in p53 mutant cell lines such as HC11. Alternatively, the observed defect in lobular-alveolar development may be mediated by molecules other than c-kit, potentially including other differentially expressed genes identified in this study (Supplementary Table 1). This data further suggest that Brca1-mediated induction of c-kit is more likely to contribute to other changes in the phenotype of luminal progenitor cells. c-kit mediated hyperproliferation of luminal progenitor cells is a reasonable hypothesis that is supported by the literature (Kijima et al., 2002) and the observed imbalance of these cells in the Brca1 deficient mammary gland (Figures 1 and 3). c-kit induction may also induce other cellular changes. Studies by other investigators have shown that repression of c-kit results in a decrease in anchorage-independent growth (Raafat et al., 2007), raising the possibility that Brca1 induction of c-kit might contribute to the oncogenic transformation of luminal progenitor cells rather than a block in their differentiation. Consistent with this notion is the finding that luminal progenitor cells from human BRCA1 mutation carriers display factor-independent growth (Lim et al., 2009).

Overrepresentation of c-kit in Brca1-deficient mammary glands is also consistent with the finding that c-kit is overexpressed in a proportion of human and mouse Brca1-associated mammary tumours (Herschkowitz et al., 2007; Lim et al., 2009). This supports our previous hypothesis that luminal progenitor cells, rather than basal or stem cells, are the target of BRCA1-associated tumourigenesis (Lim et al., 2009), which in turn is consistent with our previous studies showing no mammary phenotype when Brca1 is specifically deleted in non-luminal compartments of the mammary gland (Smart et al., 2008).

In summary, we show that loss of Brca1 in mouse mammary epithelial cells results in a block in differentiation and an accumulation of luminal progenitor cells. This is associated with overexpression of c-kit, a marker of luminal progenitor cells. Based on in vitro experiments, c-kit may simply serve as a cell marker rather than a mediator of the observed defect in lobular-alveolar development. However, analysis of c-kit mutant mice will be required to further address the role of c-kit, and may reveal a role in regulating commitment to the luminal cell lineage. Future priorities will also include defining the role of c-kit in Brca1-associated tumourigenesis and establishing whether targeting c-kit before tumour formation can interfere with this process.

Materials and methods

HC11 and SCp2 mammary epithelial morphogenesis assays

HC11 cells were supplied by Chris Ormandy (Garvan Institute, Sydney, NSW, Australia) and cultured as previously described (Naylor et al., 2005). Brca1 repression was achieved by Brca1 siRNA (L-040545-00) and non-targeting (D-001810-10-05) purchased from Dharmacon RNA Technologies (Lafayette, CO, USA), and used at 100 nM. HC11 cells were transfected with lipofectamine 2000 as described in Naylor et al. (2005) and incubated for 24 h. Knockdown was confirmed by real-time PCR, as described below. HC11 differentiation assays were performed as previously described (Naylor et al., 2005). SCp2 cells were supplied by Mina Bissell (Lawrence Berkeley National Laboratory, Berkeley, CA, USA). Repression of Brca1 function was achieved by ectopic expression of a truncated (dominant-negative) mouse Brca1 as described previously (Brown et al., 2002). Six independent pools of 7 × 105 trBrca1-expressing or control SCp2 cells were plated on Engelbreth-Holm-Swarm-derived extracellular matrix (Matrigel, Becton Dickinson, Franklin Lakes, NJ, USA) in growth media containing reduced fetal calf serum (FCS) (2%). Cells were allowed to attach for 24 h and then stimulated to differentiate in serum-free media containing 3–5 μg/ml ovine prolactin (AF Parlow, Los Angeles, UCLA, CA, USA), 1 μg/ml hydrocortisone and 5–10 μg/ml insulin. SCp2 cells were fixed in neutral-buffered formalin (Sigma-Aldrich, St Louis, MO, USA) and processed to paraffin wax. Sections were analysed by staining with haematoxylin and eosin.

Mice

MMTV-Cre (Wagner et al., 1997) and Brca1Co/Co mice (Xu et al., 1999) were obtained from the NCI, Frederick, MD, USA. F1 and F2 matings, and genotyping were performed as described (Xu et al., 1999). The MMTV-Cre mice used were transgenic line F (NCI catalogue #01XA9). To minimise the effects of expression changes associated with stage of the oestrus cycle, multiple independent groups of six mice were used from the same female-only cage, and the vulva of each mouse was checked to ensure comparable vascularisation. All procedures were approved by The University of Queensland Animal Ethics Committee. Mammary gland dissections for RNA and whole-mount analysis were performed as previously described (Smart et al., 2008).

Mammary luminal progenitor cell analysis

The thoracic and inguinal mammary glands were dissected from virgin or 12.5-day pregnant Brca1Co/+ and MMTV-Cre Brca1Co/Co female mice. Mammary epithelial cell suspensions were prepared as previously described (Shackleton et al., 2006; Asselin-Labat et al., 2007). Cells were stained with anti-CD31 and anti-CD45 antibodies conjugated to biotin (BD Pharmingen, Franklin Lakes, NJ, USA), followed by streptavidin-allophycocyanin (APC)-Cy7 (BD Pharmingen) or streptavidin-Alexa-594 (Molecular Probes, Invitrogen, Carlsbad, CA, USA). Anti-CD24 conjugated to phycoerythrin (PE) (BD Pharmingen), anti-CD29 conjugated to fluorescein isothiocyanate (clone HMβ1.1 kindly provided by Dr H Yagita, Juntendo University, Tokyo, Japan) and anti-CD61 conjugated to APC (Caltag, Invitrogen) were used. Cell sorting and analysis was carried out on a FACS ARIA or FACSDiVa (Becton Dickinson).

RNA extraction and gene expression analysis

RNA was extracted from the abdominal mammary glands (after removal of lymph nodes) of 12 mice per genotype and stage (virgin and lactation). One entire abdominal mammary gland for each mouse was ground on dry ice for RNA extraction using an RNAeasy Extraction kit (Qiagen, GmbH, Hilden, Germany) according to the manufacturers’ instructions. The RNA quantity and integrity was assessed with a Bioanalyser (Agilent Technologies, Forest Hill, VIC, USA). Three pools of four individual mice per genotype and stage were then generated, creating a total of 12 pools. The Illumina Totalprep RNA amplification kit (Applied Biosystems, Foster City, CA, USA) was used to amplify and biotinylate the 12 pooled RNA samples. Biotinylated RNA was hybridised overnight at 55 °C to Illumina Mouse Sentrix 6 v1.1 BeadChips containing >46 000 probes (Illumina Inc., San Diego, CA, USA). The microarrays were washed, stained with streptavidin-Cy3, and then scanned with an Illumina BeadArray Scanner as per manufacturer's guidelines. All microarray data is available for download from the NCBI at GEO (Accession number: pending). Raw data was imported into Illumina Beadstudio v2.6, then further analysed with Genespring v7.3 (Agilent Technologies) and R Bioconductor packages BeadArray (Dunning et al., 2007) and Limma (Smyth et al., 2005). Data were normalised with quantile normalisation and filtered using an Illumina detection score of 0.95, yielding 23 891 probes. Unsupervised cluster analysis was performed within the GeneSpring package (Agilent, Santa Clara, CA, USA) using the Pearson's correlation for the distance metric on genes that differed two-fold from the mean in at least one sample. Genes differentially expressed between genotype, or stage, were determined using the Limma package (WEHI Bioinformatics, Melbourne, VIC, Australia) in ‘R’. A total of 1379 differentially expressed genes (B-score>0) at the lactational stage were examined by Ingenuity Pathway Analysis Systems (www.ingenuity.com). Overrepresented biological themes were scored by significance. A threshold was determined by inputting the same number of random Illumina Mouse Sentrix 6 v1.1 probe codes as in the microarray results in triplicate and using the highest −log(P-value) as the background level of significance.

Real-time PCR

Primers were designed using Primer Express (Applied Biosystems) and supplied by Geneworks (Australia, Hindmarsh, SA, Australia) (Supplementary Table 4). Analysis of complementary DNA (cDNA) from virgin and lactating glands was performed using SYBR Green Master Mix (Applied Biosystems) and parous glands using the FastStart Universal SYBR Green Master Mix (Roche Applied Sciences, Castle Hill, NSW, Australia). Gene expression was quantified on an ABI 7900HT Fast Real-time PCR system with ABI 7900HT SDS software (v3.3.3, Applied Biosystems), normalising to HPRT. Cycling conditions were 2 min (50 °C), followed by 10 min (95 °C) and then onto 45 cycles of 15 s (95 °C) and 1 min (58 °C). The comparative ΔCt method (Applied Biosystems) was used to determine relative mRNA expression (Livak and Schmittgen, 2001). The values were then further normalised to the wild-type average to ascertain the expression difference between genotypes. Statistical analysis used a two-tailed Student's t-test. A P-value of 0.05 or less was considered statistically significant.

Immunofluorescence analysis

In all, 7 μm sections of OCT-embedded mammary glands were fixed with 100% cold acetone. Sections were blocked in 5% fetal calf serum, 1% bovine serum albumin and 0.1% Tween-20 in phosphate buffered saline for 30 min and then incubated with 1:100 dilution of rat anti-murine c-kit antibody (Cedarlane, Burlington, Ontario, Canada) in the same blocking buffer for 1 h at room temperature. Goat anti-rat Alexa-Fluor 488 (Molecular Probes) secondary antibody at 1:400 dilution was used to visualise protein localisation together with a DAPI (Sigma-Aldrich) nuclear counterstain used at 0.1 μg/ml, incubated on sections for 30 min at room temperature. Sections were rinsed and mounted in vectashield-mounting media before visualisation and image capture using a Zeiss Confocal LSM 710 (Zeiss, Oberkochen, Germany).

Knockdown and overexpression of c-kit

Kit siRNA (L-042174-00) and non-targeting (D-001810-10-05) were purchased from Dharmacon RNA Technologies and used at 100 nM to transfect HC11 cells as described above. Full-length mouse c-kit complementary DNA was PCR amplified from mice complementary DNA using the following primers c-kit FOR: 5′-AGCCACCGCGATGAGAGGC-3′, REV: TGCTCAGGCATCTTCGTGCAC using Phusion polymerase (Finnzymes Oy, Espoo, Finland) and cloned into pGEM-T easy (Promega, Madison, WI, USA), and then subcloned into pBABE puromycin (Morgenstern and Land, 1990), which was used to transfect BOSC-23 packaging cells to produce ecotropic retroviruses. Filtered viral supernatant plus 4 μg/ml polybrene was used to infect recipient, the murine HC11 cell line. Stably expressing cells were selected for using 1.5 μg/ml of puromycin (Sigma-Aldrich) for 10 days, following which mRNA overexpression was verified using quantitative real-time PCR. HC11 dome assays were performed as described above.