Transcription factor regulatory networks in mammary epithelial development and tumorigenesis

Abstract

The mouse mammary gland is composed of three epithelial cell types, which include ductal, alveolar and myoepithelial cells. A hierarchy in which the mammary stem cell compartment gives rise to progressively restricted progenitors that ultimately form the luminal (ductal and alveolar) and myoepithelial lineages is now emerging. Although very little is known about the mechanisms controlling the differentiation of the myoepithelial cell lineage, a growing body of work reveals that the luminal cell fate is specified by a network of transcription factors. The precise roles of specific transcription factors in promoting differentiation of luminal progenitors into ductal or alveolar cells are now being elucidated. This review will discuss the importance of these recent observations and place them within the context of other transcription factor networks involved in mammary gland development and tumorigenesis.

Introduction

The isolation of epithelial stem cells from both mouse (Shackleton et al., 2006; Stingl et al., 2006a) and human (Ginestier et al., 2007; Eirew et al., 2008; Lim et al., 2009) mammary glands, and the assessment of their ability to regenerate this tissue in vivo, has begun to build the framework for a differentiation hierarchy in the adult mammary gland (Stingl et al., 2006b; Molyneux et al., 2007; Asselin-Labat et al., 2008; Visvader, 2009). In the mouse mammary gland, a single stem cell that is characterized as CD29^hi/CD49f^hi/CD24^+/mod/Sca-1⁻ can give rise to all epithelial cell types required to produce a functional mammary gland and possesses self-renewal properties (Shackleton et al., 2006; Stingl et al., 2006a). It is believed that this mammary stem cell gives rise to a common progenitor that then splits into two lineages, which includes the myoepithelial progenitors that produce myoepithelial cells and the luminal progenitors that generate both ductal luminal epithelial cells and alveolar luminal epithelial cells (Asselin-Labat et al., 2008).

As in other cell compartments, such as the hematopoietic system, the differentiation potential of mammary epithelial cells within this hierarchical structure appears to be controlled by the action of specific transcription factors. Previous work has demonstrated an important role for steroid hormone transcription factors in regulating distinct phases of mammary gland development, with the estrogen receptor having critical roles in ductal elongation and the progesterone receptor for directing proper lobuloalveolar development during pregnancy (Brisken, 2002; Sternlicht et al., 2006). In addition to this class of transcription factors, the CCAAT/enhancer-binding protein, C/EBPbeta, also has a critical function in promoting proper lobuloalveolar development in the mammary gland (Grimm and Rosen, 2003; Lamarca et al., 2010). Although these factors are clearly necessary for the proper development of the virgin mammary gland ductal tree and subsequent development during pregnancy, it is not yet clear how these proteins control the specification of distinct cell types within the emerging differentiation hierarchy of the mammary gland. In this respect, recent work has begun to focus on elucidating the roles and interactions of several key transcription factors during normal mammary epithelial differentiation of specific mammary epithelial populations (Kouros-Mehr et al., 2006; Asselin-Labat et al., 2007; Oakes et al., 2008; Choi et al., 2009; Yamaji et al., 2009). In addition to terminal differentiation, another set of transcription factors are involved in the involution phase of mammary gland development, notably STAT3. This review will focus on the roles played by several of these; including STAT5, STAT3, Elf-5 and GATA-3 in both mammary gland development and control of a differentiation hierarchy as well as their potential roles in breast cancer initiation and progression.

STAT5A has a critical role in the establishment of luminal alveolar progenitor cells

Previous work from Hennighausen group had demonstrated that targeted disruption of STAT5A resulted in impaired alveolar differentiation during pregnancy and a failure to lactate (Liu et al., 1997; Miyoshi et al., 2001). To better understand the nature of this developmental block, a conditional STAT5A/5B allele was generated to permit mammary gland-specific ablation of both these key transcription factors following mammary epithelial expression of Cre recombinase (Yamaji et al., 2009). Mammary epithelial disruption of both STAT5A and STAT5B did not negatively impact the mammary stem cell population or their ability to contribute to normal ductal development. Loss of STAT5A/5B did result in a reduced number of CD61⁺ luminal progenitor cells, but this population did not increase during pregnancy. This led the authors to conclude that two types of CD61⁺ luminal progenitors exist, those destined to form ductal epithelium and those that undergo alveolar differentiation. The loss of STAT5A/5B did not affect the population of CD61⁺ luminal ductal progenitors, but caused a reduction in CD61⁺ luminal alveolar progenitor cells. Thus, transplants of STAT5A/5B-deficient stem cell-enriched populations could reconstitute a normal ductal tree but failed to generate functional alveoli during pregnancy. Strikingly, normal alveolar development could be restored upon inducible expression of STAT5A. Collectively these data argue that STAT5A is both necessary and sufficient for specification of a luminal alveolar progenitor population. These results now add additional information regarding how specific transcription factors control the differentiation potential of mammary epithelial cells during mammary gland development and builds on previous work in this field (Figure 1).

Figure 1

Transcription regulatory network in mammary gland development. (a) Schematic illustrating the roles of specific transcription factors in mammary gland development. GATA-3 is involved in generation of both ductal mammary epithelial cells and fully differentiated alveolar epithelial cells. STAT5A is involved in the specification of CD61⁺ alveolar progenitors but is dispensable for the genesis of CD61⁺ luminal ductal progenitor cells. Like GATA-3, Elf-5 is required for the induction of the terminally differentiated alveolar epithelial cell population. The question marks denote the lack of information regarding the transcription factor(s) that promote both the specification and differentiation of myoepithelial cells. (b) Schematic illustrating the mammary gland phenotypes of virgin and lactating glands derived from ablating the indicated transcription factors. Deletion of GATA-3 early in mammary gland development (MMTV–Cre) results in a severe defect in mammary gland development in virgin animals, whereas deletion of STAT5A/5B and Elf-5 has little impact on ductal outgrowth and branching morphogenesis. In contrast, deletion of GATA-3 (WAP–Cre), STAT5A/5B and Elf-5 have a major impact on the development of terminally differentiated alveolar cells during pregnancy.

GATA-3 specifies luminal cell fate in the mammary gland

Another class of transcription factors that has recently been implicated in luminal development is GATA-3. GATA-3 belongs to a family of zinc-finger transcription factors that have important roles in cell-fate specification and has recently emerged as a critical determinant of luminal epithelial cell differentiation within the mammary gland (Naylor and Ormandy, 2007; Kouros-Mehr et al., 2008b). GATA-3 expression is localized to the initial sites wherein mammary buds form, implying a role for this factor in the mammary primordium (Asselin-Labat et al., 2007). Deletion of GATA-3 at this early stage, through the expression of K14-Cre, resulted in the failure to develop mammary placodes and a subsequent lack of a mammary tree in the majority of cases (Asselin-Labat et al., 2007).

At later stages of mammary development, GATA-3 expression is restricted to luminal epithelial cells and is absent from the myoepithelial compartment of the developing and adult mammary gland (Kouros-Mehr et al., 2006; Asselin-Labat et al., 2007). Conditional loss of GATA-3 in the luminal mammary epithelial compartment, by MMTV–Cre-mediated excision of a conditional GATA-3 allele, resulted in the impairment of ductal elongation (Kouros-Mehr et al., 2006; Asselin-Labat et al., 2007). This result demonstrates a role for GATA-3 in establishing progenitor cells that form the luminal ductal epithelium. If GATA-3 was deleted at a later developmental stage, through the use of WAP–Cre mice, lobuloalveolar development was impaired during pregnancy, resulting in a lactation-deficient phenotype. Analysis of the mammary epithelial compartment in these animals revealed an increased number of luminal cells that lacked differentiation markers, such as caseins (Kouros-Mehr et al., 2006; Asselin-Labat et al., 2007). Excision of GATA-3, either by MMTV–Cre or WAP–Cre, led to an expansion of the CD61⁺ progenitor cell population, indicating a role of GATA-3 in the specification of luminal epithelial cells. Conversely, GATA-3 overexpression in mammary stem cell-enriched populations was sufficient to induce the expression of luminal differentiation markers, including WAP and β-casein (Asselin-Labat et al., 2007). These data position GATA-3 as a key regulatory factor in the differentiation of mammary progenitor cells; both in the formation of mature ductal epithelial cells during arborization of the mammary fat pad and in subsequent alveolar differentiation during pregnancy.

Elf-5 controls alveolar differentiation

The ETS family transcription factor, Elf-5, has also recently been shown to have a pivotal role in lobuloalveolar differentiation during pregnancy (Oakes et al., 2008). In contrast to GATA-3, loss of Elf-5 has no impact on ductal elongation and branching in mammary epithelial explantation experiments. However, lobuloalveolar development was severely impaired in Elf-5^−/− transplants. Conversely, forced expression of Elf-5 in luminal epithelial cells of virgin mice resulted in decreased proliferation, impaired ductal outgrowth and precocious differentiation/secretion within the terminal end buds. In Elf-5^−/− transplants, an increase in CD61⁺ luminal progenitor cells was observed, which positions Elf-5 as an important factor within the luminal alveolar lineage (Oakes et al., 2008). More recently, the importance of Elf-5 in lobuloalveolar differentiation was supported through the analysis of Elf-5 conditional knockout mice (Choi et al., 2009). Like STAT5A-deficient mammary glands, mammary-specific disruption of Elf-5 led to a complete block in alveolar differentiation (Choi et al., 2009). Taken together, these studies support Elf-5 as a master regulator of the alveolar switch that occurs during pregnancy.

Building a transcription factor network in mammary gland development

The analysis of mammary gland phenotypes derived from the loss of specific transcription factors has begun to shed light on how lineage commitment during mammary gland differentiation is controlled. A clear relationship exists between STAT5A and Elf-5 during the commitment and differentiation of luminal alveolar cells. Recent work demonstrates that deletion of STAT5A/5B in the mammary glands results in a severe defect in lobuloalveolar differentiation without effects on ductal outgrowth (Yamaji et al., 2009). Similarly, Elf-5^−/− glands are also capable of normal ductal elongation and branching morphogenesis but fail to undergo proper lobuloalveolar differentiation (Oakes et al., 2008; Choi et al., 2009). On its face, these results suggest that STAT5A/5B and Elf-5 control similar aspects of lobuloalveolar differentiation. Indeed, the presence of STAT transcription factor-binding sites have been identified within the Elf-5 proximal promoter and Elf5 expression was lost in STAT5A/5B-deficient glands (Yamaji et al., 2009), arguing that STAT5A lies immediately upstream of Elf-5 in a transcription factor hierarchy. This regulation is likely more complex, given that the block in alveolar differentiation in Elf-5 null mammary glands may, in part, be attributed to decreased transcriptional activation of STAT5 through Elf-5 transcription binding sites within the STAT5A proximal promoter (Choi et al., 2009). However, one important difference between the phenotypes observed in STAT5A/5B and Elf-5 deficient mammary glands is that the Elf-5 deficient glands exhibit an increase in the CD61⁺ luminal progenitor population (Oakes et al., 2008); whereas the STAT5A/5B-deficient glands exhibit a dramatic deficiency in CD61⁺ progenitor cells. One possible interpretation of this phenotype is that STAT5A is required for the induction of CD61⁺ luminal progenitors that are destined to form alveolar cells, whereas Elf-5 is dispensable for the induction of any CD61⁺ lineages. Under this scenario, the increase in CD61⁺ luminal progenitors observed in Elf-5 deficient glands is because of the requirement of Elf-5 for conversion of these cells into fully differentiated alveolar cells. Taken together, these results suggest that a transcriptional regulatory network involving STAT5A and Elf5 has a critical role in the establishment of alveolar cell identity.

In contrast to a specific role for STAT5A and Elf-5 in alveolar differentiation, GATA-3 function appears to be required at distinct stages of mammary gland development, first in the specification of the luminal cell fate and second during alveolar differentiation. It does not appear that GATA-3 lies downstream of STA5A/5B during alveolar differentiation given that GATA-3 levels are unchanged in STAT5A/5B-deficient glands (Yamaji et al., 2009). It remains possible that GATA-3 may lie upstream of STAT5A/5B or Elf-5 during alveolar differentiation. In this regard, the analysis of transcription factors expressed in terminal end-buds and mature ductal epithelium, a pattern that mimicked GATA-3 expression, identified Elf-5 (Kouros-Mehr et al., 2006). Whether GATA-3 functions within the same cells and can directly regulate the expression of STAT5A or Elf-5 requires further experimentation. It has also been shown that the expression patterns of GATA-3 and Elf-5 within the luminal epithelial compartment are largely distinct. In the case of CD61+ cells, GATA-3 and Elf-5 have both been shown to be expressed in this population; however, it has been suggested that these transcription factors may not be coexpressed within the same CD61⁺ epithelial cell (Oakes et al., 2008). Further experimental work will be required to dissect the functional relationship between GATA-3, STAT5A and Elf-5 in alveolar cell differentiation.

Bridging mammary gland development and breast cancer

Gene expression profiling has categorized breast cancers into several subtypes including luminal A, luminal B, HER2, basal and normal-like (Perou et al., 2000; Sorlie et al., 2003), which are correlated with distinct clinical outcomes (Sorlie et al., 2001; Sotiriou et al., 2003). These original subtypes will undoubtedly be revised as additional subtypes emerge, such as the putative ‘claudin-low’ subtype (Herschkowitz et al., 2007; Hennessy et al., 2009). An open question in the field of breast cancer is whether the different sub-types of breast cancer, which are defined by gene-expression profiling, reflect the transformation of specific stem cell or progenitor cell compartments responsible for development of the normal mammary gland (Prat and Perou, 2009). Using tissue from human breast reduction mammoplasties, cells were enriched for specific markers and were divided into one of four groups, including mammary stem-cell-enriched (MaSC), luminal progenitor, mature luminal and stromal populations (Lim et al., 2009). When the profiles of these four populations was compared with the previously defined breast cancer subtypes, it was found that the mammary stem-cell-enriched signature was most prominent in the ‘claudin-low’ and normal-like subtypes, the mature luminal signature was found highly represented in the luminal A, luminal B and HER2 subtypes, whereas the stromal signature was associated with the ‘claudin-low’ subtype. Perhaps the most surprising finding was the correlation between the luminal progenitor signature and the basal subtype (Lim et al., 2009). This study suggests that the differentiation hierarchy in the developing mammary gland might indeed have connections with the different subtypes of cancer that arise from this tissue. Moreover, emerging data suggests that those transcription factors that have critical roles in the specification of different mammary epithelial progenitors might also have important affects on cancer progression.

GATA-3 is a marker of the luminal subtype in breast cancer

As described above, GATA-3 specifies commitment of mammary stem cell/luminal progenitor cells to a mature luminal cell fate (Kouros-Mehr et al., 2006; Asselin-Labat et al., 2007). Thus, GATA-3 represents a strong candidate marker for the luminal breast cancer subtypes. Indeed, GATA-3 expression is strongly linked to ER-positive status, the luminal A subtype and good patient prognosis in numerous gene expression studies of primary human breast cancer (Mehra et al., 2005; Fang et al., 2009). Recent studies with the PyV mT transgenic mouse model (Guy et al., 1992; Lin et al., 2003) have examined potential roles played by GATA-3 in breast cancer development and progression. Comparison of gene expression signatures from early adenomas to those obtained from late carcinomas in the PyV mT model revealed that the expression of genes associated with luminal cell differentiation, including GATA-3, was dramatically impaired during this transition (Kouros-Mehr et al., 2008a). Conversely, when GATA-3 was ectopically expressed in PyV mT-expressing tumors, the GATA-3 expressing tumors possessed a well-differentiated phenotype that correlated with a dramatic reduction in their ability to metastasize when compared with parental PyV mT tumors (Kouros-Mehr et al., 2008a). However, when GATA-3 was deleted in well-differentiated tumors, it did not result in the progression to a more metastatic phenotype. Rather, breast cancer cells that lost GATA-3 underwent caspase-mediated cell death (Kouros-Mehr et al., 2008a). These observations have led to the hypothesis that, in the PyV mT transgenic model, GATA-3 promotes the formation of well-differentiated breast cancers that have a low metastatic potential. The eventual emergence of highly metastatic tumor cells in late carcinomas may reflect the expansion of a GATA-3 negative stem cell like population. The notion that GATA-3 expression is associated with good prognosis through its ability to promote differentiation in breast cancers may not fully explain how GATA-3 functions to suppress the metastatic properties of breast cancers. It has recently been demonstrated that GATA-3 overexpression in MDA-MB-231 breast cancer cells, which are a basal human breast cancer cell population, could impair mammary tumor growth and lung metastasis without promoting the differentiation of these cells (Dydensborg et al., 2009). Whether GATA-3 directly modulate the expression of genes involved in breast cancer metastasis, independently of its effects on breast cancer differentiation, will require further investigation.

Is STAT5 involved in mammary tumor progression?

Although a clear role for STAT5A in mammary development has been established, STAT5A function in breast cancer progression remains unclear. High levels of tyrosine-phosphorylated and nuclear-localized STAT5A can be detected in human breast cancer samples (Cotarla et al., 2004). Furthermore, reduction of STAT5A dosage impairs mammary tumor induction in mice-expressing SV40T antigen (Ren et al., 2002). Other studies have suggested that signaling elements that act upstream of STAT5A may have an important role in mammary cancer development. For example, the loss of either Prolactin or ErbB4, which are known to be potent activators of STAT5 signaling, also have major effects on alveolar differentiation (Jones et al., 1999; Clevenger, 2003). Loss of Prolactin signaling impairs mammary tumorigenesis in mice expressing the PyV mT oncogene (Vomachka et al., 2000). However, because mammary gland development is generally impaired in Prolactin-deficient mice, it is unclear whether the reduction in mammary tumor formation reflects an indirect impact on the development of the mammary epithelial cells that are targeted by PyV mT oncogene.

Although ErbB4 is a potent activator of Stat5A function, mammary epithelial disruption of ErbB4 had little impact on the ability of ErbB2 to induce mammary tumors (Jackson-Fisher et al., 2006). In fact, a recent study has suggested that catalytically inactive somatic mutations in ErbB4 can be detected in sporadic human breast cancer (Tvorogov et al., 2009). Significantly these inactivating mutations in ErbB4 result in dramatic impairment in the capacity to activate Stat5A and induce differentiation of MDA-MB-468 cells (Tvorogov et al., 2009).These data would argue that tyrosine phosphorylation of STAT5A may actually impair breast cancer development by inducing differentiation.

One of the future challenges in elucidating the role of STAT5A in breast cancer will be discriminating between the direct effects of STAT5 family signaling on tumor development and progression and the effects of STAT5A's indirect role in the development of mammary epithelial cells that are then the target for transformation by activated oncogenes. Mouse models that permit the conditional and inducible loss of STAT5A/5B in the mammary gland, like those described by the Hennighausen group (Yamaji et al., 2009), will have an instrumental role in assessing the precise role of STAT5 in both initiation and maintenance of oncogene-induced mammary tumor progression.

A STAT3 dependent transcription regulatory network is involved in mammary gland involution and metastasis

Like the STAT5A and GATA-3 transcription factors, the STAT3 transcription factor has been implicated as a key regulator of both normal mammary gland development and mammary tumor progression (Pensa et al., 2009). The development of transgenic mice that carry conditional alleles of STAT3 has been critically important for understanding STAT3's role in mammary gland biology. Using different mammary epithelial Cre-expressing transgenic reporter strains, it has been demonstrated that mammary-epithelial disruption of STAT3 results in a dramatic impairment of mammary gland involution (Chapman et al., 1999; Humphreys et al., 2002). A survey of STAT3 dependent transcriptional targets has revealed that many are involved in acute inflammation and acute-phase regulation often associated with mammary gland involution. One such gene encodes the Oncostatin M receptor (OSMR) that, when activated, results in enhanced STAT3 tyrosine phosphorylation. STAT3 activation further enhances OSMR expression, thereby establishing a positive-feedback loop. Conversely, OSMR activation results in the dephosphorylation of STAT5 (Tiffen et al., 2008). Consistent with its importance in promoting mammary gland involution, OSMR-deficient mice still retained mammary epithelium even after 10 days of involution (Tiffen et al., 2008). The reciprocal regulation of STAT5A and STAT3 by OSMR suggests that these closely related STAT transcription factors have opposing roles in mammary gland involution.

Another important transcriptional target of STAT3 was C/EBPδ (Thangaraju et al., 2005; Zhang et al., 2007). Like STAT3, germ-line disruption of C/EBPδ resulted in a profound defect in mammary gland involution (Thangaraju et al., 2005). Thus, genetic and biochemical analyses suggest that a STAT3-dependent inflammatory cascade is essential for the normal involution of the mammary gland (Pensa et al., 2009). STAT3 acts to invoke a massive inflammatory response that is required for efficient clearance of apoptotic mammary epithelium during involution. In contrast to STAT5, which acts in a cell-autonomous manner to determine the luminal alveolar cell fate, STAT3 functions in non-cell autonomous manner to modulate the mammary microenvironment during mammary gland involution.

STAT3's role as the regulator of a pro-apoptotic program in normal mammary gland development is in apparent contradiction to its documented pro-survival functions in mammary tumor progression (Guo et al., 2006; Zhou et al., 2007). A potential explanation may lie in recent work demonstrating that the involuting mammary gland is a particularly good microenvironment for promoting mammary tumor metastasis (McDaniel et al., 2006; Schedin et al., 2007). Recently published data suggests that the pro-inflammatory functions of STAT3 during mammary gland involution may also have a critical role in mammary tumor progression. In this work, mammary epithelial disruption of STAT3 in transgenic mice expressing the ErbB2 oncogene provide compelling evidence that a STAT3-dependent pro-inflammatory program is required for the metastatic phase of tumor progression (Ranger et al., 2009). Although mammary tumor onset was not affected by ablation of STAT3, the resulting tumors rarely metastasized to the lung (Ranger et al., 2009). Remarkably, analyses of gene expression profile of these STAT3-deficient ErbB2 tumors revealed that the acute inflammatory response was dramatically impaired because of the reduced expression of C/EBPδ, OSMR, SAA1 and SAA2. The fact that this same STAT3-dependent transcriptional regulatory network was involved in regulating mammary gland involution (Pensa et al., 2009), suggests that STAT3 may use the same network in mammary tumor progression. Consistent with this concept, mammary specific disruption of the STAT3 co-activator SRC-1 suppresses tumor metastasis without affecting the primary growth of PyV mT mammary tumors (Wang et al., 2009). Taken together, these observations argue that, as in normal mammary gland involution, STAT3's primary function during mammary tumor progression is to alter the immediate tumor microenvironment.

Future perspectives

One of the key issues stemming from a comparison of these transcription factor networks in normal mammary gland development is to what extent they are utilized during tumor progression. Emerging data suggests that distinct progenitor cells in the developing mammary gland, which are specified through the action of these transcriptional regulatory networks, are targeted by oncogenic initiating events which ultimately give rise to the different breast cancer subtypes (Lim et al., 2009; Prat and Perou, 2009). However, the tremendous heterogeneity that occurs within a given breast cancer suggests that these tumor cells are under considerable evolutionary pressures (Campbell and Polyak, 2007; Polyak, 2007a, 2007b). Accumulated genetic lesions within a breast cancer cell will have an impact on how these networks interact and will thus ultimately shape the tumor phenotype. In addition to epithelial cell-autonomous consequences, activation of these transcription networks can also have an impact on other cell types within the tumor microenvironment and thereby influence their capacity to metastasize (Ranger et al., 2009). The future elucidation of the complex relationship between the normal and pathological outputs of these networks may provide important therapeutic insights into the treatment of breast cancer.