|Home | About | Journals | Submit | Contact Us | Français|
The GATA family of transcription factors plays fundamental roles in cell-fate specification. However, it is unclear if these genes are necessary for the maintenance of cellular differentiation after development. We identified GATA-3 as the most highly enriched transcription factor in the mammary epithelium of pubertal mice. GATA-3 was found in the luminal cells of mammary ducts and the body cells of terminal end buds (TEBs). Upon conditional deletion of GATA-3, mice exhibited severe defects in mammary development due to failure in TEB formation during puberty. After acute GATA-3 loss, adult mice exhibited undifferentiated luminal cell expansion with basement-membrane detachment, which led to caspase-mediated cell death in the long term. Further, FOXA1 was identified as a downstream target of GATA-3 in the mammary gland. This suggests that GATA-3 actively maintains luminal epithelial differentiation in the adult mammary gland, which raises important implications for the pathogenesis of breast cancer.
A fundamental aspect of development is the specification and maintenance of the differentiated cell state. The specification of cell fate is mediated in part by the establishment of hierarchical networks of transcription factors and the cis-regulatory elements that control their expression (Davidson et al., 2002; Swiers et al., 2006). Transcription factors are organized into structurally similar multigene families, such as the GATA and FOX families. These genes play essential roles in activating target genes of specific cell fates and also in repressing target genes of alternate cell fates (Singh et al., 2005). Key mechanisms that underlie target gene activation and repression are chromatin remodeling and DNA methylation, which modulate the accessibility of transcriptional activation complexes to target gene loci. Though the roles of transcription factors in specifying cell fate have been established, it is unclear how cells maintain the differentiated state in the adult organism. That is, it is unclear whether the genetic regulatory networks that establish cell specification are necessary to maintain cell fate after development has completed.
The GATA family of transcription factors is a group of six highly conserved transcription factors that play fundamental roles in cell-fate specification (Patient and McGhee, 2002). GATA factors bind the DNA sequence (A/T)GATA (A/G) via a DNA binding domain containing one or two zinc-finger domains. One of the best-characterized functions of GATA factors is the essential role of GATA-1 in erythropoeisis. GATA-1 is essential for the specification of erythroid cells from myelo-erythroid progenitors, whereas the Ets family transcription factor PU.1 is essential for the specification of myeloid cells (Cantor and Orkin, 2001). GATA-1 and PU.1 exhibit transcriptional crossantagonism, thereby blocking the ability of a specified cell to adopt the alternate cell fate. Accordingly, the loss of GATA-1 causes a conversion of erythropoiesis to myelopoiesis, leading to a deficiency of erythroid cells and an expansion of myeloid cells (Galloway et al., 2005). A similar phenomenon has been observed for GATA-3 in T helper(Th)2 cell-fate specification. GATA-3 and the T box family transcription factor T-bet play essential roles in the specification of naive T helper cells into Th2 and Th1 effector cells, respectively (Grogan and Locksley, 2002). GATA-3 and T-bet also exhibit transcriptional crossantagonism to repress the alternate cell fate. Interestingly, it was recently suggested that GATA-3 expression must be sustained to maintain the differentiated Th2 cell type (Pai et al., 2004).
The mammary gland is a ductal epithelial organ that, like the Th1/Th2 system, contains two mature epithelial cell types: the luminal epithelial cells, which line the ductal lumen and secrete milk proteins, and the myoepithelial cells, which line the basal surface of the luminal cells. These differentiated cell types arise from a multipotent progenitor population that has been recently characterized (Shackleton et al., 2006). Prior to the onset of puberty, the mammary gland is a rudimentary organ consisting of a primitive network of ductal epithelium. Shortly after the onset of puberty, specialized structures known as TEBs develop at the invading epithelial tips of the mammary epithelium. TEBs contain an outer layer of cap cells, which are believed to be myoepithelial progenitors, and a multilayered inner core of body cells, which may be luminal cell progenitors (Smalley and Ashworth, 2003). The TEBs proliferate, bifurcate, and invade into the fatty stroma of the mammary gland in a process known as branching morphogenesis. The process continues until 10–12 weeks of age, at which point the TEBs have invaded the entire length of the fat pad and then regress, having given rise to a mature ductal tree (Sternlicht et al., 2006).
Little is known about the differentiation of the luminal cells, which are the principal cells implicated in breast cancer. Efforts to understand luminal cell differentiation have focused on the estrogen receptor (ERα), which is found in a subset of luminal cells that exhibit low proliferation (Cheng et al., 2004). ERα is found in only half of mature luminal cells and is found in fibroblasts and other stromal cells. Mice lacking ERα exhibit defects in mammary development, including an inability to form TEBs and defects in ductal invasion, though differentiation markers such as milk proteins are still expressed in its absence (Mallepell et al., 2006). In this report, we carry out a genome-wide screen and identify GATA-3 as the most highly enriched transcription factor in the mammary ductal epithelium of pubertal mice. We show that GATA-3 is localized to the body cells of TEBs and to the mature luminal cells of mammary ducts. Using a loss-of-function approach, we show that GATA-3 is necessary in vivo for mammary development and for the maintenance of the differentiated luminal epithelium.
We devised a microarray strategy to identify novel regulators of mammary development. β-actin-GFP reporter mice were used to visualize the mammary epithelium in vivo (Hadjantonakis et al., 1998). TEBs, mature ducts, and epithelium-free stromal microenvironments were surgically microdissected from 5-week-old β-actin-GFP mice, and RNA was immediately isolated for analysis (Figure 1A). The RNA expression profiles of the TEB and mature-duct microenvironments were compared to the epithelium-free stroma by using long-oligonucleotide spotted microarrays with 19,500 features. An analysis of the most highly enriched genes in the mammary epithelium revealed that members of the keratin family (keratins 8 and 19) were the most highly enriched genes in the TEB microenvironment, whereas members of the casein family (casein k and casein α) were the most highly enriched genes in the mature-duct microenvironment (Table 1). In this analysis, we identified GATA-3 as the most highly enriched transcription factor in both the TEB (13.9-fold change) and mature-duct (10.6-fold change) microenvironments. A full analysis of the microarray data can be found elsewhere (Kouros-Mehr and Werb, 2006).
We further investigated the localization of the GATA-3 protein by immunofluorescence. GATA-3 was found in all luminal cells of mature ducts, both within pubertal (5-week-old) and adult (12-week-old) virgin mice (Figures 1C and 1D). Costaining with α-smooth muscle actin (SMA) revealed that SMA-positive myoepithelial cells were GATA-3 negative, suggesting that GATA-3 was restricted to the luminal cells of the mammary gland (Figure 1D). GATA-3 was also expressed in the body cells of TEBs; however, the intensity of GATA-3 staining was lower in the distal end of the TEB (i.e., nearest to the invading front) compared to the proximal end (Figure 1B). GATA-3 was not found in the cap cells of the TEB (Figure 1B, arrowhead).
We used a loss-of-function strategy to study the role of GATA-3 in the mammary gland. GATA-3 null mice display embryonic lethality between days 11 and 12 postcoitum, so we were unable to use these mice to study mammary development (Pandolfi et al., 1995). Instead, we used the Cre/loxP recombination system to generate mice that were conditionally deficient in GATA-3. Mice with a floxed GATA-3 locus were obtained, as described elsewhere (Pai et al., 2003). In the floxed construct, exons 4 and 5 of the native GATA-3 locus were replaced by a cassette containing exons 4, 5, and 6, flanked by loxP sites. The loxP sites were further flanked by a phosphoglycerokinase (PGK) promoter and an enhanced green fluorescent protein (GFP) cassette, such that Cre-mediated recombination led to the placement of the GFP immediately downstream of the PGK promoter. We thus used GFP as a marker of Cre-mediated recombination to assess the efficiency of GATA-3 deletion.
We first crossed the floxed GATA-3 mouse with the constitutively active MMTV-Cre line, where Cre expression has been reported in the mammary gland, skin, salivary glands, and other sites (Wagner et al., 2001). Homozygous floxed mice carrying the MMTV-Cre transgene (MMTV-Cre; GATA-3flox/flox) were born in normal Mendelian ratios compared to their littermates (data not shown). This suggested that the conditional deletion approach rescued the embryonic lethality that had been observed in GATA-3 null mice. Whereas heterozygous floxed mice expressing MMTV-Cre (MMTV-Cre; GATA-3flox/+) appeared in all respects normal compared to non-Cre littermate controls (Non-Tg; GATA-3flox/x), MMTV-Cre; GATA-3flox/flox exhibited severe defects in their skin and mammary glands beginning at puberty. MMTV-Cre; GATA-3flox/flox mice began to lose hair shortly after puberty, and by adulthood, these mice displayed total alopecia. The mice also exhibited other major skin defects, including epidermal hyperplasia, dermal fibrosis, and hyperkeratosis (data not shown).
The development of the mammary gland was severely disrupted in MMTV-Cre; GATA-3flox/flox mice. Prior to puberty at day 19 postpartum, MMTV-Cre; GATA-3flox/flox mice contained a rudimentary mammary gland that appeared to be similar to littermate controls by whole-mount inspection (Figures 2A and 2B). However, with the onset of puberty, the mammary glands of MMTV-Cre; GATA-3flox/flox mice failed to develop TEBs (Figures 2C and 2D). As a result, by 5 weeks postpartum, the ductal epithelium failed to invade into the fatty stroma and remained in a compact and highly defective state (Figures 2E, 2F, and 2I). At 8 weeks postpartum, a few sporadic TEBs emerged in MMTV-Cre; GATA-3flox/flox mammary glands, but these TEBs were incapable of filling the mammary stroma by 30 weeks postpartum, suggesting that GATA-3 played an essential role for normal TEB function (Figures 2G–2I). Moreover, the 8-week-old null outgrowths displayed gross structural defects, including irregular luminal diameters and deficiencies in side branching (Figure 2H).
Histologic analysis of the MMTV-Cre; GATA-3flox/flox outgrowths further demonstrated an essential role for GATA-3 in mammary development. Hematoxylin and eosin (H&E) staining showed profound defects in the luminal epithelium of 8-week-old MMTV-Cre; GATA-3flox/flox outgrowths (Figure 3A). Whereas MMTV-Cre; GATA-3flox/+ ductal epithelium contained a single layer of luminal cells, the MMTV-Cre; GATA-3flox/flox ductal epithelium contained regions that lacked luminal cells and regions that contained a multilayered luminal epithelium (Figure 3B, arrowheads). Interestingly, immunostaining revealed that these defective outgrowths remained GATA-3 positive despite having undergone recombination (GFP+) (Figure 3B). The multilayered epithelium in the null outgrowths contained a basally located population of GATA-3/GFP double-positive luminal cells and a distinct population of GFP-positive, GATA-3-negative cells in the ductal lumen (Figure 3B, yellow arrowhead). A similar histologic profile was observed in 19-day-old MMTV-Cre; GATA-3flox/flox outgrowths (data not shown).
To determine the mechanism for the expression of GATA-3 in the 8-week-old MMTV-Cre; GATA-3flox/flox outgrowths, we purified genomic DNA from the recombined (GFP+) cells of MMTV-Cre; GATA-3flox/+ and MMTV-Cre; GATA-3flox/flox mammary glands. PCR analysis using primers specific to wild-type, floxed, and deleted GATA-3 loci revealed that the null outgrowths contained a nondeleted GATA-3 allele despite being GFP+ (Figure 3C). Thus, there was a selective pressure to retain a functional GATA-3 allele in the surviving outgrowths, which further suggested that GATA-3 is essential for mammary development.
To further analyze the molecular function of GATA-3 in the mammary gland, we crossed the floxed GATA-3 mouse with the mammary-specific, doxycycline-inducible Cre line WAP-rtTA-Cre. Cre expression in the latter is both highly specific to the mammary gland and tightly regulated by doxycycline administration (Utomo et al., 1999). This enabled us to study the role of GATA-3 in the mammary gland while minimizing possible secondary effects from GATA-3 deletion in other organs. Furthermore, this approach enabled us to delete GATA-3 in the adult mammary gland after development had taken place.
We administered doxycycline to 12-week-old WAP-rtTA-Cre; GATA-3flox/flox and WAP-rtTA-Cre; GATA-3flox/+ mice and analyzed GFP expression to determine the timing and efficiency of recombination. GFP was not detected in the absence of doxycycline administration, indicating that there was no leakiness of Cre expression. Although GFP was not detectable after a 3 day course of doxycycline, by 5 days of doxycycline, high levels of GFP were observed throughout the ductal epithelium of both WAP-rtTA-Cre; GATA-3flox/flox and WAP-rtTA-Cre; GATA-3flox/+ mammary glands (Figure 4A). GFP was induced in a large fraction of luminal cells but was absent in myoepithelial and stromal cells, suggesting that WAP-rtTA-Cre was highly specific to the luminal epithelium (Figure 4C). Interestingly, the WAP-rtTA-Cre; GATA-3flox/flox mammary glands showed greatly reduced GFP expression after a 14 day course of doxycycline, whereas the WAP-rtTA-Cre; GATA-3flox/+ glands maintained a high level of GFP expression (Figure 4A). Both the relative intensity and the distribution of GFP along the ductal epithelium were reduced (Figure 4A), suggesting that these cells had been lost.
On histologic examination, the 5 and 14 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands showed severe cellular defects in the luminal epithelium. After 5 days of doxycycline treatment, WAP-rtTA-Cre; GATA-3flox/flox mammary glands developed a disorganized multilayered epithelium (Figure 4B). There appeared to be a substantial increase in cell number and heterogeneity of nuclear size and orientation. Detachment of single cells into the ductal lumen was observed, but only sporadically (see below). After 14 days of doxycycline treatment, WAP-rtTA-Cre; GATA-3flox/flox mammary glands showed additional defects in the luminal epithelium, including cell detachment into the ductal lumen, disruption of the ductal architecture, and widespread cell death (Figure 4B). In some areas, individual luminal cells detached from the basement membrane, whereas in other areas, large portions of ductal epithelium detached (Figure 4B). No phenotype was observed in 5 or 14 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/+ mammary glands (Figure 4B).
Immunostaining of WAP-rtTA-Cre; GATA-3flox/flox mammary glands verified that GATA-3 was lost after doxycycline administration. In 5 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands, we identified two populations of luminal cells: a GATA-3-positive population lining the basement membrane and a GATA-3-negative population that had detached from the basement membrane (Figure 4C, white arrow). Both populations expressed similar levels of GFP, suggesting that only the subset of recombined luminal cells that lost the GATA-3 protein had detached from the basement membrane at this time point. In 14 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands, we detected a single population of GATA-3-positive, GFP-negative cells. We did not detect GFP-positive cells in histologic analysis, suggesting that recombined cells had undergone negative selection by this time point (Figure 4C). Taken together, these data suggest that GATA-3 is necessary in adult mammary glands to maintain the integrity of the luminal epithelium.
To characterize the immediate consequences of GATA-3 deletion in the adult mammary gland, we further immuno-stained the 5 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands with structural and differentiation markers. Coimmunostaining with the luminal marker keratin 18 and the myoepithelial marker keratin 14 revealed that the multilayered epithelium in 5 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands was exclusively keratin 18 positive, suggesting that they retained luminal character and had not trans-differentiated into myoepithelial cells (Figure 5A).
Despite being a luminal cell population, the 5 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox epithelium expressed significantly reduced levels of luminal differentiation markers (Figures 5B–5D). There was a substantial reduction of β-casein immunostaining in the null mammary glands compared to littermate controls (Figure 5B). There was also a reduction of E-cadherin and an absence of ERα in the null cells that exhibited detachment from the basement membrane (Figures 5C and 5D). To verify these findings, we compared the RNA expression profiles of 5 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox and WAP-rtTA-Cre; GATA-3flox/+ whole mammary glands by microarray (Figure 5E). The data confirmed that there was a relative increase in the expression levels of the luminal keratins 18, 19, and 8 and a decrease in various luminal differentiation markers (including members of the casein, cadherin, and estrogen receptor families) in the null mammary glands compared to heterozygous controls.
These data suggested that the acute loss of GATA-3 led to an expansion of a luminal cell population that lacked markers of differentiation. To confirm that this phenomenon indeed represented cellular proliferation, we analyzed BrdU and PCNA staining in the 5 day doxycycline-treated mammary glands. WAP-rtTA-Cre; GATA-3flox/flox mammary glands showed a significant increase of PCNA and BrdU-positive luminal cells compared to WAP-rtTA-Cre; GATA-3flox/+ controls, suggesting that the acute loss of GATA-3 led to cell-cycle progression (Figures 5F and 5G).
We determined the long-term effects of GATA-3 loss by further characterizing the 14 day doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands. In contrast to the 5 day doxycycline treatment, the 14 day doxycycline treatment led to a relative reduction in the number of keratin 18-positive cells in WAP-rtTA-Cre; GATA-3flox/flox mammary glands compared to controls (Figures 6A–6C). There was also a reduction of E-cadherin and ERα-positive cells after 14 days of doxycycline (Figures 6B and 6C). To determine if the reduction in cell number was due to cell death, we immunostained with the apoptosis marker M30, which recognizes a caspase-cleaved epitope of keratin 18. We observed widespread M30-positive luminal cells in 14 day, but not 5 day, doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mammary glands (Figures 6D and 6E). This suggested that cell death was not a primary event after GATA-3 loss but was rather a long-term consequence of GATA-3 loss.
The lactational competence of doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mice was assessed to determine the functional consequence of GATA-3 loss. WAP-rtTA-Cre; GATA-3flox/flox mice that had received a 14 day course of doxycycline were bred and sacrificed at day 10 postpartum. The mammary glands of WAP-rtTA-Cre; GATA-3flox/flox mothers exhibited severe defects in lactation as a functional consequence of GATA-3 loss, including a reduction in the number and size of milk-producing alveolar units (Figures 6F and 6G). As a consequence of the lactational defect, the pups of null mothers displayed a significant reduction in weight at day 10 post-partum (Figure 6H). Doxycycline-treated WAP-rtTA-Cre; GATA-3flox/+ mothers showed normal lactation and exhibited a heterogeneous pattern of GFP expression, indicating mosaicism of recombination in the WAP-rtTA-Cre line (Figure 6F). In contrast, the ductal structures in doxycycline-treated WAP-rtTA-Cre; GATA-3flox/flox mothers were GFP negative, which suggested there was a negative selection for recombined cells in the WAP-rtTA-Cre; GATA-3flox/flox mammary glands (Figure 6F).
To identify potential downstream effectors of GATA-3 signaling, we performed bioinformatic analysis of our mammary microarray data. We identified 22 transcription factors that were highly enriched in TEBs and mature ducts of the mammary gland, a pattern that mirrors GATA-3 expression (Figure S1A in the Supplemental Data available with this article online). These genes included FOXA1, FOXP4, MSX2, TRPS1, ELF5, EHF, RUNX1, as well as several members of the Id, Irx, Sox, and TCFAP-2 transcription factor families. We then analyzed four independent microarray datasets to identify whether any of these genes were consistently correlated with GATA-3 expression in breast cancer samples. The transcription factor FOXA1 emerged as one of the best predictors of GATA-3 expression in these datasets. Of ~24,000 genes, FOXA1 emerged as the most highly correlated gene with GATA-3 in two offour datasets, and it was the second and sixth most highly correlated gene in the other two datasets (Bonferroni-adjusted p < 0.0001 in all datasets, (Figure S1B). The strong positive correlation between GATA-3 and FOXA1 suggested a potential interaction between these transcription factors.
To explore the possible relationship between GATA-3 and FOXA1, we analyzed the 50 kb promoter sequences of mouse and human FOXA1 and identified three conserved GATA binding sites lying 0.8, 1.7, and 3.6 kb upstream of the first exon (rVista 2.0). We performed chromatin immunoprecipitation (ChIP) of primary mammary epithelial cultures with a GATA-3 antibody and then PCR amplified the regions containing these GATA binding sites. The GATA binding site lying 0.8 kb upstream of the FOXA1 first exon coimmunoprecipitated with the GATA-3 antibody, suggesting that GATA-3 binds to the FOXA1 promoter in this region (Figure 6I).
We show here for the first time that GATA-3 is a critical regulator of luminal cell differentiation in the mammary gland. This finding adds to the growing body of evidence that GATA factors are important mediators of cell-fate decisions in developmental systems. Examples include the essential roles of GATA-1 in erythrocyte differentiation, GATA-3 in T cell and Th2 cell differentiation, GATA-3 in sympathetic neuron differentiation, GATA-4 in gastric epithelial differentiation, and GATA-6 in distal lung epithelial differentiation (Jacobsen et al., 2002; Tsarovina et al., 2004; Yang et al., 2002). Though the role of GATA factors in specifying cell fate has been established, our data suggest that GATA factors maintain cellular differentiation after development has taken place. We have found that GATA-3 actively maintains luminal epithelial differentiation in the adult mammary gland.
The maintenance of differentiation is likely governed by GATA-3 operating in a hierarchical network of transcriptional activators and repressors, as has been observed in other developmental systems (Davidson et al., 2002; Swiers et al., 2006). We have identified FOXA1 as a putative member of the GATA-3 genetic regulatory network. Expression of GATA-3 and FOXA1 is strongly correlated in mammary and breast cancer microarray datasets, and FOXA1 contains a functional GATA-3 binding site. Moreover, the overexpression of GATA-3 was shown to upregulate FOXA1 expression, suggesting that GATA-3 activates gene expression (Usary et al., 2004). Interestingly, FOXA1 is essential for estrogen signaling in mammary cells and is required for the direct interaction of ER to chromatin sites (Carroll et al., 2005; Laganiere et al., 2005). It is possible that FOXA1 mediates crosstalk between GATA-3 and ER signaling in the mammary gland. Because GATA-3 is not an estrogen-responsive gene (Hoch et al., 1999), we hypothesize that GATA-3 acts upstream of ER to regulate luminal gene expression. This hypothesis explains why acute GATA-3 loss leads to ER loss and also explains the similarity in phenotypes between GATA-3 and ERα null mammary glands (Mallepell et al., 2006). Further work will be necessary to define the interactions between GATA-3, FOXA1, and ER and to further define the GATA-3 genetic regulatory network in the mammary gland.
Our data suggest that, in addition to maintaining the luminal cell fate in adulthood, GATA-3 plays a role in the specification of the luminal cell fate during development. We showed that GATA-3 is necessary for the formation of TEBs, which perform the essential developmental functions in this organ. Within the TEBs, GATA-3 is restricted to the body cells, which include putative luminal progenitor cells (Smalley and Ashworth, 2003). The inability of TEBs to develop in the absence of GATA-3 may be due to a failure in luminal progenitor specification. However, as direct evidence establishing the body cell as the bona fide luminal progenitor cell is lacking, a characterization of the body and cap cell populations of TEBs will be necessary for future studies on GATA-3-mediated luminal cell specification.
We observed that the acute loss of GATA-3 in adult mammary glands led to widespread cellular proliferation. This suggests that GATA-3 may play a direct role in maintaining the quiescent state of differentiated luminal cells. Other studies have reported the connection between GATA factors and cell-cycle control. In one study, the inactivation of GATA-3 caused an uncontrolled cellular proliferation in the nephric duct of the embryonic kidney (Grote et al., 2006). In another study, GATA-2 inactivation enhanced the proliferative capacity of embryonic neuroepithelial cells, whereas the overexpression of GATA-2 was sufficient to inhibit the proliferation of neural progenitors (El Wakil et al., 2006). Interestingly, in the mammary gland, we observed a gradient of GATA-3 expression such that highly proliferative cells (i.e., the body cells of TEBs) expressed low levels of GATA-3, whereas the nonproliferating (differentiated) population expressed high levels of GATA-3. Further work will be necessary to determine whether the establishment of GATA-3 expression plays a causal role in cell-cycle control.
Though acute loss of GATA-3 led to cellular proliferation, the long-term consequence of GATA-3 loss was caspase-mediated cell death. This finding contrasts with the results of other loss-of-function studies involving GATA factors. The loss of GATA-1 in mature erythroid cells and GATA-3 in committed Th2 cells led to transdifferentiation into alternate cell types (Galloway et al., 2005; Pai et al., 2004). In our study, the loss of GATA-3 in committed luminal cells did not lead to a transdifferentiation into myoepithelial cells; instead, these cells retained partial luminal character and exhibited cell death. In the mammary gland, differentiated luminal cells rely on survival signals from the basement membrane (Wiseman and Werb, 2002). Detachment of luminal cells causes a loss of these survival signals, resulting in caspase-mediated cell death (Frisch and Screaton, 2001). Thus, the cell death that we observed after long-term loss of GATA-3 is likely to be a secondary event following basement-membrane detachment and is not a primary consequence of GATA-3 loss.
Breast cancers are luminal epithelial cell neoplasias that can be subdivided into ER+ and ER− tumors. ER+ breast tumors tend to be morphologically well differentiated and exhibit a relatively good prognosis, whereas the ER− tumors are poorly differentiated and exhibit a relatively poor prognosis. To date, a series of eleven independent microarray profiling studies (1009 patients in total) has analyzed the gene expression signatures of ER+ and ER− human breast tumors (Bertucci et al., 2000; Farmer et al., 2005; Gruvberger et al., 2001; Hoch et al., 1999; Mehra et al., 2005; Perou et al., 2000; Sorlie et al., 2003; Sotiriou et al., 2003; van ‘t Veer et al., 2002; Wang et al., 2005; West et al., 2001). In all studies, GATA-3 emerged as a strong and independent prognostic indicator of breast cancer. Low GATA-3 expression was strongly associated with higher histologic grade, positive lymph nodes, larger tumor size, ER and PR-negative status, and HER2 overexpression. Meta-analysis of the microarray data revealed that the prognostic utility of GATA-3 exceeded that of conventional variables such as ER status (Mehra et al., 2005). In one study, GATA-3 was identified as the second-best predictor of breast cancer survival among 8024 genes (Jenssen et al., 2002). These studies have established GATA-3 status as being an important prognostic factor in breast cancer, one that clearly warrants further investigation.
Given the fundamental role of GATA-3 in maintaining the differentiation of the luminal epithelial cell, we hypothesize that GATA-3 is causally involved in the pathogenesis of breast cancer. Well-differentiated tumors with high GATA-3 expression have a low propensity for metastasis, whereas poorly differentiated tumors with low GATA-3 expression have a greater propensity for metastasis. We postulate that loss of GATA-3 may play a causal role in loss of tumor differentiation and malignant conversion in breast cancer. The loss of GATA-3 in the mammary gland causes luminal cell proliferation and basement-membrane detachment. The loss of GATA-3 in breast cancer may similarly lead to cellular proliferation and basement-membrane detachment, which may result in the acquisition of metastatic capability. Our future work is directed at determining what role GATA-3 plays in breast cancer progression and whether it is a suitable target for screening and drug therapy.
The floxed GATA-3 mouse in the C57Bl/6 strain was kindly provided by I-Cheng Ho and Sung-Yun Pai (Harvard Medical School). MMTV-Cre and WAP-rtTA-Cre mice in the C57Bl/6 strain were obtained from the Mouse Models of Human Cancer Consortium (MMHCC), National Cancer Institute. β-actin-GFP reporter mice in the FVB/n strain were from Jackson Laboratory.
TEB microenvironments, mature-duct microenvironments, and distal stroma regions of mammary glands 2–5 were microdissected from 5-week-old β-actin-GFP reporter mice. RNA was extracted with Trizol Reagent (Tel-Test), reverse transcribed in the presence of amino-allyl-dUTP, and coupled to CyScribe dyes (Amersham). Unamplified Cy5-labeled TEB or mature duct cDNAs and Cy3-labeled stromal cDNAs were hybridized onto 70-mer oligonucleotide microarrays with 19,500 features (Operon, mouse version 2.0), as described elsewhere (Barczak et al., 2003). The most highly enriched genes were determined by calculating M = log2(Cy5/Cy3) for the TEB versus distal stroma (n = 6) and mature duct versus distal stroma (n = 6) microarrays. A full analysis of the microarray data can be found elsewhere (Kouros-Mehr and Werb, 2006).
All animal experiments were performed with protocols approved by the UCSF IACUC. Floxed mice were bred with MMTV-Cre or WAP-rtTA-Cre mice to obtain Cre; GATA-3flox/+ mice. Cre; GATA-3flox/+ mice were then bred with GATA-3flox/flox mice to obtain Cre; GATA-3flox/flox, Cre; GATA-3flox/+, and nontransgenic; GATA-3flox/x mice. Mice were genotyped as described (Pai et al., 2003). Number 4 (inguinal) mammary glands of MMTV-Cre; GATA-3flox/flox, MMTV-Cre; GATA-3flox/+, and Non-Tg; GATA-3flox/+ mice were analyzed by carmine whole mount, as described (Jones et al., 1996). Ductal invasion was measured from the proximal end of the primary duct to the most distal TEB. Genomic DNA from the recombined cells of 30-week-old MMTV-Cre; GATA-3flox/flox, and MMTV-Cre; GATA-3flox/+ mice was purified by collagenase digestion of mammary tissue, purification of single cells, and FACS sorting of GFP+ cells (Welm et al., 2002). Genomic DNA was analyzed with PCR primers specific to the wild-type GATA-3, floxed GATA-3, and deleted GATA-3 loci, as described (Pai et al., 2003). For the WAP-rtTA-Cre experiments, 12-week-old mice were placed on a 3, 5, or 14 day course of doxycycline feed (Bio-Serv). A subset of mice that were on a 14 day course of doxycycline was mated with wild-type males and maintained on doxycycline until day 10 of lactation, at which time the mice were sacrificed.
Mammary glands were fixed overnight in 4% paraformaldehye and then paraffin embedded. Sections (5 μm) were rehydrated, processed with microwave antigen retrieval, and blocked with the M.O.M kit (Vector Biolabs). Primary antibodies were diluted in M.O.M. diluent and incubated on sections overnight at 4°C. Primary antibodies used were anti-GATA-3 (1:25, HG-31, Santa Cruz Biotechnologies), anti-Estrogen Receptor (1:30, clone 1D5, DAKO), anti-E-cadherin (1:500, 610181, Becton-Dickinson), Cy3-anti-α-smooth muscle actin (1:200, 1A4, Sigma), anti-β-casein (1:100, gift of M.J. Bissell laboratory), anti-Cyto-keratin 14 (1:500, PRB-155P, Covance), anti-cytokeratin 18 (1:50, Ks 18.04, Progen), anti-PCNA (1:500, PC10, DAKO), anti-caspase-cleaved cytokeratin 18 (1:100, M30, Roche) and anti-GFP (1:1000, ab290, Abcam). Sections were then incubated with Alexa 488 and Alexa 564 secondary antibodies (1:300, Molecular Bioprobes) for 1 hr. For BrdU analysis, mice received a 300 μl intraperitoneal injection of 5 mg/ml BrdU and were sacrificed 3 hr later. Sections were then immunostained with FITC-anti-BrdU (1:10, 447583, Becton-Dickinson). Analysis and three-dimensional rendering were performed by using the Volocity software package.
Primary cultures of adult mammary gland were used in the analysis, as described (Welm et al., 2005). Adult mammary gland DNA was immunoprecipitated with the GATA-3 antibody, control mouse immunoglobulin (Jackson ImmunoResearch), or no primary antibody control, as described (ChIP assay kit, Upstate Biotechnology). PCR amplification of a 200 bp fragment 0.8 kb upstream of the FOXA1 first exon (primers 5′-TACGAGGGCAAGCCACTAAC-3′ and 5′-ATCTGCCACGCTAAAT GAGG-3′) was performed with the following cycle parameters: 95°C for 5 min, 35 cycles of 94°C for 30 s, 55°C for 30 s and 72°C for 30 s, and a 10 min 72°C final extension (HotStarTaq, Qiagen).
Supplemental Data include Supplemental References and one figure and can be found with this article online at http://www.cell.com/cgi/content/full/127/5/1041/DC1/.
We thank I-Cheng Ho and Sung-Yun Pai for kindly providing the floxed GATA-3 mouse. We thank Andrea Barczak, Agnes Paquet, and David Erle of the Sandler/UCSF Genomics Core Facility for their assistance with microarray profiling and data analysis. We also thank Andrew Ewald for critically reading the manuscript. This work was supported by grants (CA057621 and ES012801) from the National Cancer Institute and National Institute of Environmental Health Sciences. H.K.-M. was supported by the UCSF Medical Scientist Training Program (MSTP), a California Breast Cancer Research Program (CBCRP) dissertation award, and the Paul and Daisy Soros Fellowship for New Americans.
Accession Numbers Microarray data were submitted to the NCBI Gene Expression Omnibus under the accession numbers GSE2988 and GSE5602.