|Home | About | Journals | Submit | Contact Us | Français|
The mammary gland is an epidermal appendage that begins to form during embryogenesis, but whose development is only completed during pregnancy. Each mammary gland begins as a budlike invagination of the surface ectoderm, which then gives rise to a simple duct system by birth. Subsequent development occurs during sexual maturation and during pregnancy and lactation. In this review, we outline the distinct stages of embryonic mammary development and discuss the molecular pathways involved in the regulation of morphogenesis at each stage. We also discuss the potential relevance of embryonic breast development to the pathophysiology of breast cancer and highlight questions for future research.
The mammary gland is a specialized epidermal appendage, whose development begins during embryogenesis but is not fully completed until the beginning of lactation. Creating a mammary gland entails a series of developmental tasks that require the coordination of many cell biological and cell signaling processes, all of which have parallels during the development of breast cancer. In this review, we will address the beginnings of this process, the formation of the rudimentary gland in embryos. We will first outline the morphological events leading to the formation of the embryonic gland. We will then discuss what is known of the molecular regulation of the different stages of this process. Given that most recent research has been done in mice, we will limit our discussions to this species. Finally, we will highlight some emerging issues and questions for future research.
The development of the mammary gland begins with the formation of the milk or mammary lines at the ventral aspect of the Wolfian ridge on day 10 of gestation (Sakakura 1987). These are bilateral ectodermal ridges that stretch in a rostral–caudal orientation between the fore and hind limb buds. In rabbits, this line has been noted to protrude above the surface of the embryo on scanning electron micrographs (Propper 1978; Sakakura 1987). However, in mice this is not the case and the line has, instead, been defined by the expression of several molecular markers of the Wnt signaling pathway (Chu et al. 2004; Veltmaat et al. 2004; Veltmaat et al. 2006). Ectodermal cells within the mammary lines become columnar and multilayered as compared to the otherwise flat, single-layered epithelial sheet that comprises the surrounding epidermis (Sakakura 1987; Veltmaat et al. 2004).
Within 24–36 h of its formation, the mammary line resolves into five pairs of placodes (three thoracic and two inguinal) in characteristic locations along the ventral–lateral border of the embryo (Fig. 1). Each pair of placodes develops symmetrically but the individual pairs develop asynchronously and in a distinct order (Veltmaat et al. 2003; Veltmaat et al. 2004). The first to become visible is placode 3, followed by placode 4, and then by placodes 1 and 5, which develop together. The last placode to form is number 2. Histologically the placodes are lens-shaped thickenings of the surface ectoderm consisting of several layers of cells that are larger and more columnar in appearance than are those in the surrounding epidermis. The orientation of the placodal cells is not uniform, which may represent the fact that they are formed by migration of cells from the mammary line (Propper 1978; Hens and Wysolmerski 2005; Robinson 2007).
Each mammary placode expands and invaginates into the underlying mesenchyme to form a mammary bud (Fig. 2), a process that is essentially completed by embryonic day 14 (Sakakura 1987; Watson and Khaled 2008). As the cells from the placode begin to invaginate they pile up at the surface and begin to array themselves in a concentric orientation so that the early bud is elevated above the plane of the surrounding epidermis. As a result, at E12, the developing buds are visible as prominent knobs on the ventral surface of the embryo, which then submerge beneath the plane of the skin (Chu et al. 2004; Veltmaat et al. 2004). The formation of the epithelial bud is also accompanied by dramatic changes in the underlying mesenchyme (Sakakura 1987; Dunbar et al. 1999; Foley et al. 2001). Cells surrounding the epithelium elongate, condense, and orient their long axis in a radial fashion around the developing epithelial cells. At the end of this process, the mature bud consists of a sphere of concentrically arrayed epithelial cells connected to the skin surface by a stalk of epidermal-like cells. Both ball and stalk are, in turn, surrounded by three to five layers of condensed mesenchyme, referred to as the primary mammary mesenchyme. In female mice the mammary buds remain relatively quiescent from E14 to E16, when ductal morphogenesis commences. However, in male mice, the mesenchyme around the stalk enlarges and condenses further, eventually severing the connection to the skin (Heuberger et al. 1982; Dunbar et al. 1999; Veltmaat et al. 2003) (Fig. 2). This is followed by apoptosis of the mammary mesenchyme and many cells within the epithelial bud as well (Dunbar et al. 1999). Consequently, male mice have either no mammary epithelial ducts or only very rudimentary ones.
The next phase of development involves the onset of ductal branching morphogenesis on E16. A solid cord of epithelial cells emerges from the mammary bud and grows down from the primary mammary mesenchyme and into a second stromal compartment, termed the mammary fat pad (Sakakura 1987; Hens and Wysolmerski 2005). The embryonic fat pad consists of a loose collection of preadipocytes that originate from a mesenchymal condensation formed on E14 (Sakakura et al. 1982). Once the primary cord or mammary sprout reaches the adipocytes of the fat pad, it begins to branch in a characteristic dichotomous fashion to form the initial ductal tree. This initial round of branching growth extends from E16 to the perinatal period and results in a primary duct and about 10–15 initial branches. The epithelial duct system then grows slowly until puberty, when a second round of rapid expansion takes place (Watson and Khaled 2008). Concurrent with this outgrowth, two other important morphological processes occur. First, a ductal lumen is formed. Second, the skin overlying the primary mammary mesenchyme is remodeled into the typical nipple structure. This involves a thickening of the epidermis, the suppression of hair follicle development and the invagination of a concentric ring of keratinocytes to form the nipple sheath surrounding the origin of the primary duct as it connects to the surface (Foley et al. 2001). The initial primary mesenchyme becomes a dense connective tissue associated with the nipple while the ducts themselves are thereafter contained within the loose fatty connective tissue of the mammary fat pad. At the conclusion of embryonic development, the mammary gland consists of a short primary duct ending in a small, branched ductal tree that is embedded in one end of larger mammary fat pad. This rudimentary duct system forms the scaffold on which further development during puberty and pregnancy will eventually produce the mature milk-producing glands found during lactation (Figs. 1 and and2)2) (Table 1).
Wnt10b is the earliest (E11.25) focal endogenous marker of the ectodermal mammary line and is also expressed in axillary and inguinal streaks around the limb buds that contribute to buds one, two, and five (Veltmaat et al. 2004). The Wnt signaling reporter, TOP-Gal, is activated in an identical pattern but, surprisingly, appears in both ectoderm and mesenchyme a day earlier than Wnt10b suggesting that some earlier expressed Wnt must regulate its expression (Chu et al. 2004). The functional importance of this early Wnt signal has been clearly showed by experiments involving ectopic ectodermal expression of a secreted Wnt inhibitor, Dkk1. Dkk1 extinguishes the expression of TOP-Gal, Wnt10b, and Tbx3, a T-box-containing transcription factor also important for mammary development (see later discussion). Dkk1 expression also abolishes all morphologic evidence of mammary development (Chu et al. 2004). These data indicate that canonical Wnt signaling is essential for mammary line specification and acts in two waves during this process.
The identity of the first Wnt signal has yet to be determined. Candidates include Wnt 3, 6, and 10b, which are expressed at earlier stages in a broad band of flank ectoderm (Chu et al. 2004; Veltmaat et al. 2004). Wnt5a and 11 are expressed broadly within the flank mesenchyme and could also serve as the first mesenchymal signal (Chu et al. 2004). Embryonic mammary development proceeds normally in Wnt5a−/− mice (Chu et al. 2004; Roarty and Serra, 2007). Therefore, Wnt 5a is not required for mammary line specification, but it might function redundantly with Wnt11. The diffuse expression pattern of these Wnts points to the existence of additional mechanisms that restrict Top-Gal response to the mammary line. Wnt5a and Wnt11 are more frequently associated with noncanonical signaling raising the intriguing possibility that noncanonical antagonism of canonical Wnt signaling may focally restrict the canonical Wnt response to the mammary line (Roarty and Serra, 2007).
Other pathways also contribute toward defining the mammary line (see Fig. 3). For example, two components of the fibroblast growth factor pathway, Fgf10 and Fgfr2b, have been suggested to function early in mammary line specification. Fgf10−/− and Fgfr2b−/− mice fail to develop all but mammary bud #4 (Mailleux et al. 2002). Fgf10 is expressed at the center and ventral tip of the thoracic somites on E10.25 when the mammary line is forming (Veltmaat et al. 2006). This region of mesenchyme, the hypaxial dermomyotome, has been shown to instruct the fate of the ectoderm and lies directly underneath the mammary line, thus being well placed to provide vertical, localizing signals (Sakakura 1987). Supporting the concept that a somitic signal determines the D/V position of the mammary line, a wave of progressive ectodermal stratification was found to coincide with ventrally extending thoracic somites (Veltmaat et al. 2006). Furthermore, Pax3 mutants, which fail to undergo ventral extension of the thoracic somites, show dorsalization of the mammary line (Veltmaat et al. 2006). Poland syndrome in humans, which is characterized by hypoplasia of breast, pectoral muscles, and skeletal structures derived from thoracic somites, further suggests a connection between somite and mammary gland development (Baban et al. 2009). Fgf10's candidacy as the critical somitic factor is supported by the fact that Fgf10 hypomorphs and Gli3xt/xt mutants, which both have normal somitic development but show reduced Fgf10 expression, fail to express TOP-Gal and Wnt10b and lack ectodermal stratification within the thoracic mammary line region leading to loss of buds #3 and #5 (Hatsell and Cowin 2006; Veltmaat et al. 2006). Gli3 is a transcription factor that acts as downstream mediator of the hedgehog signaling pathway. Somitic Fgf10 expression is unaffected by Wnt inhibitors placing Fgf10 signals upstream of or parallel to the first Wnt signal (Chu et al. 2004). Notably, the punctate pattern of TOP-Gal and Wnt10b aligns with the segmentation pattern of the thoracic somites (Veltmaat et al. 2006). Collectively these data have led to a model in which somitic Gli3 regulates expression of Fgf10, which in turn signals to ectodermal Fgfr2b and thence to Wnt10b (Mailleux et al. 2002; Veltmaat et al. 2006).
There are several aspects of the above pathway that suggest additional complexity. For example, Hedgehog signaling is absent and Gli3 has been shown to act as a repressor, whereas Fgf10 serves as an activator, raising the possibility that an additional repressor must intervene (Hatsell and Cowin 2006). The insolubility of FGFs also makes it unlikely that they diffuse across the proposed distance between somites and ectoderm. Delamination of hypaxial cells expressing Fgf10 has been suggested as a mechanism to traverse this gap (Veltmaat et al. 2006). Alternatively, it has been suggested that intervening molecules may relay Fgf10 signals through the mesoderm. Neuregulin 3 (Nrg3), a member of the EGF family, has been proposed to serve this function (Howard et al. 2005). Hypomorphic NrgSka mutants show no disturbance in Fgf10 or mesenchymal Tbx3 expression and thus may lie downstream of these factors. Nrg3 coated beads have been shown to induce both mesenchymal and epithelial canonical Wnt signaling (indicated by BAT-Gal, another Wnt-reporter transgene) and ectopic placode development suggesting that Nrg3 augments or facilitates Wnt signaling (Howard et al. 2005). Additional Fgf/Fgfr1 signals are also required early in the initiation of mammary development (Eblaghie et al. 2004). Experiments documenting marker expression changes in response to beads soaked in Fgf8, Wnt, or inhibitors of Fgfr1 signaling have indicated that Wnt and Fgf/Fgfr1 pathways converge to induce and maintain mesenchymal Tbx3 (Eblaghie et al. 2004).
Tbx3 is associated with the ulnar-mammary syndrome in humans. In this syndrome, haploinsufficiency for Tbx3 causes severe mammary hypoplasia, and sometimes complete loss of mammary glands, coupled to upper limb deficiencies (Bamshad et al. 1997). Similarly, Tbx3+/- mice form mammary placodes but the three thoracic buds are not maintained and development of the remaining inguinal glands is hypoplastic (Davenport et al. 2003; Jerome-Majewska et al. 2005). Tbx3−/− mice have a more severe phenotype and fail to develop most mammary placodes all together. Tbx3 is expressed very early, at approximately E10.25, in a thin line in the mesenchyme under the presumptive mammary line (Jerome-Majewska et al. 2005). This expression precedes ectodermal preplacodal Tbx3 accumulation (at E10.5) and is essential for later Wnt 10b and Lef1 (at E11.5) expression (Davenport et al. 2003; Eblaghie et al. 2004; Jerome-Majewska et al. 2005). Lef1 is a downstream transcriptional mediator of Wnt/β-catenin signaling but is also itself a Wnt target gene. Reciprocal antagonism between early Tbx3 expression and ventral bone morphogenic protein 4 (Bmp4) has been proposed to establish a D/V boundary that defines the mammary line by positioning Lef1 expression (Cho et al. 2006). Experimental manipulations have shown that ectopic Tbx3 expands Lef1 and Wnt10b expression (Cho et al. 2006). Ectopic Bmp4 appears to ventralize the flank by severely constricting Tbx3 and abolishing dorsal Tbx15 (Cho et al. 2006). Thus Bmp4 appears to control the ventral extent of Tbx3 expression. A similar mechanism has been shown to determine the D/V boundary for coat color (Candille et al. 2004). Bmp4, however, is also essential to maintain Lef1 expression (Cho et al. 2006). Thus establishment of a Bmp4/Tbx3 boundary appears to determine an essential level of Tbx3. Levels of Tbx3 are critical for the induction of placodes and for the maintenance of thoracic buds as shown by the haploinsufficient phenotypes in mice and humans.
Collectively, these experiments lead to a working model (see Fig. 3) in which ventral Bmp4 restricts dorsal Tbx3 to establish a D/V boundary that focally confines competence to respond to the first broadly expressed Wnt ligand (Chu et al. 2004; Cho et al. 2006). Vertical somitic Fgf10/Fgfr2B and dermomyotomal Fgf/Fgfr1 signals, possibly conveyed through mesenchymal Nrg3 converge with the initial Wnt signals to cause Tbx3 accumulation (Eblaghie et al. 2004; Howard et al. 2005; Jerome-Majewska et al. 2005; Veltmaat et al. 2006). A focal line of Wnt10b expression is the end downstream result of this specification process (Veltmaat et al. 2004). Wnt10b in turn goes on to further amplify canonical Wnt signaling, up-regulating Lef1 and maintaining Tbx3 expression leading to ectodermal cell-fate change and pseudostratification of the mammary line.
Wnt10b and TOP-Gal expression become restricted around E11.5 to cells that are actively rearranging to form placodes (Chu et al. 2004). TOP-Gal is expressed predominantly within epithelial cells at this stage. Both loss- and gain-of-function experiments show that β-catenin signaling is critical for placodal formation and maintenance (Chu et al. 2004; Boras-Granic et al. 2006). Canonical Wnt signaling may be important to direct cell movements into the placode at this stage. Moreover, as loss of β-catenin signaling affects all ectodermal appendage development it is likely to causes multipotent ectodermal cells to adopt a generic placodal cell fate (Zhang et al. 2008). Blocking Wnt reception with Dkk completely suppresses all placode development. Conversely, stimulating canonical signaling accelerates, expands, and induces placodes and placodal markers (Wnt10b and Tbx3) within the ventral–lateral ectoderm (Chu et al. 2004). Further supporting the requirement for canonical Wnt signaling in placodal development, Lef1−/− mice lack two pairs of placodes and the three pairs that do form are small and degenerate (van Genderen et al. 1994; Boras-Granic et al. 2006). The fact that loss of Lef1 results in a milder phenotype than overexpression of Dkk may be caused by redundancy among the Tcf/Lef transcriptional partners. Placodes are also absent or smaller in mice mutant for other Wnt pathway genes such as Pygopus 2 (Pygo2, a Wnt modifier) and Lrp5 and Lrp6 (two Wnt coreceptors) (Lindvall et al. 2006; Gu et al. 2009; Lindvall et al. 2009).
Factors that may lie upstream of Wnt signaling at this stage appear to be body-site specific (Mailleux et al. 2002; Davenport et al. 2003; Howard et al. 2005; Jerome-Majewska et al. 2005) (Table 1). Individual placodes are differentially regulated by upstream somitic and mesenchymal signals: FGF10/FGFR2b critically regulates placode 1, 2, 3, and 5 formation and Lef1 and TOP-Gal-F expression; Tbx3 is essential for placodes 1, 3, 4, and 5 formation and Lef1 expression. Nrg3 critically regulates the size and position of buds 3 and 4, Wnt 10b and Lef1 expression, and appears to be involved in placode specification. NrgSka mutants show both supernumerary and missing placodes particularly around placodes #3 and #4, and ectodermal Nrg3 overexpression can induce the formation of extra placodes along the mammary line. Downstream targets of β-catenin signaling at this stage are poorly defined but likely include positive and negative regulators Eda and Dkk. Eda is a member of the tumor necrosis factor (TNF) superfamily and signals via its receptor, Edar. Like Nrg3, overexpression of Eda, a Wnt target, promotes placodal cell fate along the entire mammary line (Panchal et al. 2007; Pummila et al. 2007). In other appendages, the Wnt pathway converges with BMP and Edar pathways to delimit placodal size by augmenting circumferential expression of Wnt inhibitors such as DKK (Fliniaux et al. 2008).
As noted previously, the fully formed mammary bud is essentially a sphere of concentrically arrayed mammary epithelial cells hanging from the skin by a stalk of epidermal-like cells and surrounded by condensed mammary mesenchyme (see Fig 2B) (Hens and Wysolmerski 2005). The formation of this structure is a continuum with the formation of the placode, and it involves a series of sequential and reciprocal interactions between the epithelial cells and the surrounding mesenchyme. There are likely also specific morphogenetic movements required by the epithelial cells for invagination of the bud. The mitotic index of the cells within the bud has been shown to be lower than that of the surrounding epidermal cells and, thus, the bud does not form as a result of cell division (Balinsky 1950; Propper 1978; Chu et al. 2004; Heckman et al. 2007). As with placode specification, the formation of the individual bud pairs display differential requirements for specific signaling pathways (Veltmaat et al. 2006; Robinson 2007; Watson and Khaled 2008). There are now several molecular pathways that have been found to be important for this process, including members of the PTHrP, Wnt, and IGF1 signaling pathways as well as the combination of two homeodomain transcription factors, Msx1 and Msx2.
Parathyroid hormone-related protein (PTHrP) is a peptide growth factor that is ancestrally related to parathyroid hormone (PTH) and that shares with PTH the use of a common G protein-coupled receptor known as the Type 1 PTH/PTHrP receptor (PTHR1) (Strewler 2000). The PTHrP gene is prominently expressed by mammary epithelial cells from the placodal stage through birth and the PTHR1 gene is expressed broadly in the subdermal mesenchyme, including the dense mammary mesenchyme (Wysolmerski et al. 1998). Disruption of either gene results in a similar phenotype, the formation of relatively normal appearing mammary buds, but a failure of further development (Wysolmerski et al. 1998). A series of studies has shown that PTHrP is necessary for the full differentiation of the mammary mesenchyme; in its absence, mesenchymal cells condense around the epithelial bud, but do not express markers typical of the dense mammary mesenchyme (Wysolmerski et al. 1998; Dunbar et al. 1999; Foley et al. 2001). As a result, the epithelial cells loose their mammary identity and, instead, differentiate down an epidermal pathway. There is also a loss of sexual dimorphism because PTHrP from the epithelial cells is necessary for the expression of androgen receptor (AR) in the mesenchyme. In addition to the AR, PTHrP signaling has now been shown to regulate the expression of at least 40 different molecules in the mammary mesenchyme, although many of these may likely require coexpression of ventrally restricted BMP4 to be induced by PTHrP (Dunbar et al. 1999; Foley et al. 2001; Hens et al. 2007; Hens et al. 2009). These studies underscore the significance of epithelial/mesenchymal crosstalk in the formation of the mammary bud and show the importance of PTHrP as an epithelial signal that shapes the nature of the primary mesenchyme.
P190-B is a RhoGTPase activating protein (RhoGAP) family member, which serves to inhibit Rho activity (Burbelo et al. 1995). It is expressed primarily within the epithelium of the mammary bud at E12.5, but also at lower levels in the surrounding mesenchyme (Heckman et al. 2007). P190-B−/− embryos have small mammary buds with a disorganized mammary mesenchyme that fails to express AR (Heckman et al. 2007). Given that the p190-B RhoGAP pathway has been shown to intersect with IGF-1 signaling, Heckman and colleagues also examined the function of insulin receptor substrates 1 and 2 (IRS1and IRS2) and the IGF-1 receptor (IGF-1R) in embryonic mammary development (Heckman et al. 2007). IRS 1 and 2 are expressed in both the epithelial cells and mesenchyme of the mammary bud and the combined loss of both IRS 1 and 2 phenocopied the loss of p190-B. Interestingly, disruption of the IGF-1R gene resulted in small epithelial buds but a normal mammary mesenchyme. The defect in the epithelial component of the bud appeared to be related to a failure to recruit p63-positive bud progenitor cells, whereas the defects in mammary mesenchyme were associated with reduced proliferation of mesenchymal cells around the bud. These data show that the IGF-1/p190-B pathway is vital for formation of the mature buds.
In addition to its role in mammary line and placode formation, Wnt signaling also contributes to the transition of the placodes to buds. Different Wnt reporter transgenes have shown that the canonical pathway is active in both epithelial and mesenchymal cells (Chu et al. 2004; Boras-Granic et al. 2006; Lindvall et al. 2006; Lindvall et al. 2009). Canonical Wnt signaling is transduced by a coreceptor complex consisting of a member of the frizzled family of seven-pass, transmembrane receptors combined with either Lrp5 or Lrp6 (Kikuchi et al. 2007). Disruption of the Lrp5 or Lrp6 genes has been reported to result in abnormally small mammary buds (Lindvall et al. 2006; Badders et al. 2009; Lindvall et al. 2009). Similarly, deletion of the pygo2 gene, also leads to small buds that have a decreased rate of epithelial proliferation (Gu et al. 2009). Lef1 is expressed first in epithelial cells in mammary placodes and then in both epithelial and mesenchymal cells in the buds. Lef1−/− embryos form placodes 1, 4, and 5 but they never transition to buds (Kratochwil et al. 1996; Boras-Granic et al. 2006). Instead, the mesenchymal cells undergo apoptosis and the epithelial cells fail to invaginate (Boras-Granic et al. 2006). A similar phenotype is also observed in Gli3 mutant embryos (extratoes, Gli3xt/xt); bud 2 forms but fails to invaginate, remains protruding from the epidermis and loses mammary identity (Hatsell and Cowin, unpublished observations). These data suggest that both epithelial and mesenchymal Wnt pathways contribute to the formation of the mature mammary buds.
Finally, the homeodomain-containing transcription factors, Msx1 and Msx2 are both expressed in the epithelial cells in mammary buds and Msx2 is also expressed in the mammary mesenchyme (Phippard et al. 1996; Satokata et al. 2000; Hens et al. 2007). Loss of either gene alone has no effect on the formation of the mammary buds, although loss of Msx2 does affect nipple formation and bud outgrowth (see discussion later) (Satokata et al. 2000; Hens et al. 2007). However Msx1/Msx2 double knockout mice form small placodes that never develop into full buds (Satokata et al. 2000). Therefore, these two transcription factors have necessary but redundant effects on mammary bud formation.
Sprouting of the mammary buds initiates the formation of the rudimentary ductal tree. Although many of the cellular tasks involved in the extension of the ducts and the formation of the initial branch-points are likely to be shared with the later phase of ductal morphogenesis during puberty, there are important differences and, overall, much less is known about this phase of development than about pubertal development. The most obvious difference between the two phases of ductal development is that morphogenesis during puberty is regulated by systemic hormones but, during embryogenesis, it proceeds independent of hormonal input. Formation of the initial ductal tree occurs normally in mice deficient in estrogen receptors (α and β), progesterone receptors, prolactin receptors, and growth hormone receptors, all of which contribute to later stages of ductal development (Hennighausen and Robinson 2001; Hovey et al. 2002). In addition, mice deficient in the epidermal growth factor receptor, a key mediator of estrogen action during puberty, also appear to initiate mammary sprouting and ductal development, although the neonatal duct system may be somewhat smaller overall (Sternlicht et al. 2005). Until recently, very little was known of what actually did contribute to the regulation of this phase of mammary development. However, in the past several years, three signaling pathways, PTHrP, BMP, and Wnt, have been implicated in triggering and/or sustaining ductal outgrowth from the mammary buds. In addition, the homeobox genes Msx2 and Hoxc6 have both been shown to contribute to bud outgrowth. Each of these factors will be discussed later.
Disruption of either the PTHrP or PTHR1 genes in mice results in the failure of epithelial ductal morphogenesis (Wysolmerski et al. 1998). Null mutations of the PTHR1 gene in humans results in a condition known as Bloomstrands chondrodysplasia and affected fetuses also lack any evidence of mammary epithelial ducts (Wysolmerski et al. 2001). As noted previously, mammary buds form in PTHrP−/− and PTHR1−/− mice, but there is a failure of ductal morphogenesis and little or no growth occurs from the mammary buds. There is also a failure of nipple skin development (Foley et al. 2001). These findings show that PTHrP signaling from epithelium to mesenchyme represents an important event in the transition from mammary bud to duct system. They also underscore the importance of proper epithelial–mesenchymal crosstalk for the initiation of ductal and nipple morphogenesis and suggest that the primary mammary mesenchyme may initiate ductal outgrowth.
An obvious question is how PTHrP signaling enables the mammary mesenchyme to support ductal outgrowth. Recent data suggest that BMP4 is an important downstream mediator of PTHrP's actions (Hens et al. 2007; Hens et al. 2009). Both BMP4 and BMP2 are expressed in the developing bud epithelial cells, and from E11.5 to E14.5, BMP 4 is also expressed in the underlying ventral mesenchyme (Phippard et al. 1996; Hens et al. 2007). The BMP1A receptor is expressed in the mesenchyme and its expression is up-regulated by PTHrP signaling. Hens et al showed that both PTHrP and BMP4 can rescue the outgrowth of PTHrP−/− mammary buds cultured ex vivo, and that the BMP antagonist Noggin could inhibit the outgrowth of normal buds in culture and block the ability of PTHrP to rescue the outgrowth of PTHrP−/− buds (Hens et al. 2007). These data suggest that autocrine BMP4 acts downstream of PTHrP in the mammary mesenchyme to trigger the initiation of ductal morphogenesis. More recently, analysis of gene expression changes in PTHrP−/− versus WT mammary buds at E15 identified 35 genes that were regulated by PTHrP signaling in the primary mammary mesenchyme (Hens et al. 2009). Of these 35 genes, at least six are known targets of BMP signaling, reinforcing the notion that PTHrP action in these cells leads to activation of the BMP pathway. Thus far, two genes activated by the combination of PTHrP and BMP4 have been shown to mediate morphogenetic events in the mammary bud. Msx2 is a known target of BMP signaling and its expression in the mammary mesenchyme has been shown to require PTHrP signaling (Hens et al. 2007). Mesenchymal expression of Msx2 contributes to the lateral inhibition of hair follicle formation around the primary duct and nipple and it may contribute to ductal outgrowth as well. The combination of BMP4 and PTHrP also activates matrix metalloproteinase 2 (MMP2) activity in mesenchymal cells and MMP inhibitors have been shown to block PTHrP-dependent mammary bud outgrowth in culture (Hens et al. 2009). However, it should be noted that ductal outgrowth occurs in MMP2−/− mice, so that although MMP2 activity likely participates in ductal outgrowth in vivo, it is not required (Wiseman et al. 2003). In summary, the emerging model is that PTHrP signaling acts to sensitize the primary mesenchyme to respond to pre-existing BMP4 expressed in the ventral dermis. This induces the expression of a subset of specific genes in the mesenchyme (Msx2, MMP2, and others), which, in turn, mediate the various morphogenetic tasks involved in initiating ductal morphogenesis from the bud. Defining this entire program should teach us much more about ductal outgrowth.
Chu et al reported that the TOP-Gal Wnt reporter transgene was expressed in a subset of mammary epithelial cells during the outgrowth of the primary duct system at E18.5 (Chu et al. 2004). More recently, Lindvall et al have also documented expression of another Wnt reporter transgene, BAT-Gal, in mammary epithelial ducts during the initiation of ductal development (Lindvall et al. 2006). Several lines of evidence support the functional importance of canonical Wnt signaling during ductal outgrowth. While loss of Lrp5 or Lrp6 results in smaller mammary buds, only loss of Lrp6 impairs the formation of the primary duct system (Lindvall et al. 2006; Badders et al. 2009; Lindvall et al. 2009). At E18.5, Lrp6−/− embryos have only a rudimentary primary duct, a small fat pad and no ductal branches. Loss of Lrp6 also leads to a loss of BAT-Gal activity showing that the ductal defects are associated with disruption of canonical signaling. A very similar phenotype was also reported by Gu and colleagues, who showed that pygo2 is necessary for full ductal outgrowth during embryogenesis (Gu et al. 2009). Pygo2 is primarily expressed in mammary epithelial cells and in some mammary mesenchyme cells. As with Lrp6−/− embryos, outgrowth of the duct system is markedly stunted in Pygo2−/− embryos, consisting of a small primary duct and few if any branches. Finally, initiation of overexpression of Dkk1 after the formation of the mammary buds impairs outgrowth of the initial mammary ducts as well (Sarah Millar, unpublished data). Therefore, it is now clear that canonical Wnt signaling is important to the outgrowth of initial ducts from the bud. Given that there is a primary duct that develops in each of these genetic models, it may be that Wnt signaling does not initiate mammary sprouting, but rather is necessary for the initiation of ductal branching and the formation of the mammary fat pad. At least some of these defects may be caused by a failure to sustain and/or expand progenitor cells, as loss of either Lrp6 or pygo2 is associated with a decrease in mammary progenitor numbers (Gu et al. 2009; Lindvall et al. 2009).
Msx2 and Hoxc6 (another homeodomain containing transcription factor) have both been reported to be necessary for mammary ductal outgrowth. As noted previously, just before ductal sprouting, Msx2 is expressed in the mammary mesenchyme as a result of PTHrP/BMP4 signaling (Hens et al. 2007). Sakota and colleagues reported that mammary buds formed but no ductal outgrowth occurred in Msx2−/− embryos (Satokata et al. 2000). However, Hens and colleagues noted that, in their hands, although ductal outgrowth was delayed in these embryos, it did occur, and post natal Msx2−/− mice had relatively normal appearing epithelial ducts (Hens et al. 2007). Hoxc6−/− embryos develop mammary buds, which begin to grow out and form nipple sheaths and primary ducts (Garcia-Gasca and Spyropoulos 2000). However, similar to the Lrp6−/− and pygo2−/− embryos, they do not form a branched ductal tree and have smaller mammary fat pads. Interestingly, this phenotype is only present in the thoracic mammary buds; embryonic mammary development is normal in the inguinal glands. It is not clear where Hoxc6 is expressed in the embryonic mammary buds or how it might be involved in ductal outgrowth. However, given the similarity in the respective phenotypes, it is tempting to speculate that it may intersect with Wnt signaling, at least in the thoracic buds. There is evidence to support such an interaction during prostate cancer development and during xenopus gastrulation (McCabe et al. 2008; In der Rieden et al. 2010).
There are many parallels between embryonic mammary development and breast cancer. During embryonic development, breast cells proliferate, migrate, and invade from one stromal compartment into another. Many of these cellular tasks are also required for breast cancer to metastasize from a primary tumor to distant sites. Signaling pathways such as the Wnt, Hedgehog, FGF, EGF, IGF-1, and PTHrP systems discussed in this review have all been implicated in the pathogenesis of metastatic breast cancer. Therefore, the mechanisms by which these signaling pathways regulate embryonic morphogenesis may inform how they might contribute to tumor growth and metastases. During normal development, these pathways must be constrained so that morphogenesis is coordinated properly. Understanding the normal feedback constraints on cell division and mammary epithelial cell invasion during normal development may then suggest strategies to restrain the proliferative and metastatic behavior of breast cancer cells. Tumor cell biology may also inform our understanding of normal development. Recently, it has been suggested that tumor cells and especially tumor “stem” cells may manifest features of epithelial-mesenchymal transition (EMT). EMT can be triggered by the expression of several transcription factors such as slug or twist, and is characterized by the loss of homotypic cell adhesion and a more mobile phenotype. Interestingly, both snail and slug have been shown to be expressed in the epithelial cells of the embryonic mammary bud. Although one could envision that embryonic mammary epithelial cells might undergo an EMT during the formation of the mammary placode or during the outgrowth of the mammary sprout, this issue has not been examined in detail in the developing embryonic gland.
There are also intriguing similarities between the dense mammary mesenchyme and the tumor stroma surrounding breast cancers. In recent years, it has become clear that the stroma around tumors is fundamentally different than that surrounding the normal breast epithelium (Finak et al. 2008; Liao et al. 2009; Teissedre et al. 2009). Furthermore, it appears that alterations in the stromal microenvironment may contribute to the progression and metastasis of breast cancer (Finak et al. 2008; Liao et al. 2009). In mice, the mature mammary ducts grow out from the embryonic mammary mesenchyme to reside in the fatty stroma of the mammary fat pad. Transplantation studies from Sakakura and colleagues (Sakakura et al. 1982) showed that recombination of normal epithelium with embryonic mammary mesenchyme produced hyperplasia, whereas recombination with the mammary fat pad allowed for the development of normal ducts . These data suggest that an embryonic-type stroma may support tumor progression and raise the question as to the similarity between embryonic mammary mesenchyme and breast cancer associated stroma. The Wysolmerski laboratory has begun to address this question by comparing the gene expression profile of the mammary mesenchyme with that of tumor stroma as defined in the study by Finak et al (Finak et al. 2008), using the Oncomine search engine (Rhodes et al. 2004). Of the 35 genes defined as being regulated in the mammary mesenchyme by PTHrP, 22 were also differentially expressed in breast tumor stroma as compared to normal breast stroma (Hens et al. 2009). This suggests that the capacity of the mammary mesenchyme to support ductal outgrowth in embryos may be mirrored by the ability of the peri-tumor stroma to support malignant progression. It remains to be seen if breast cancer cells remodel their surrounding stroma through pathways akin to those used by embryonic mammary epithelial cells to induce the condensation of the mammary mesenchyme.
The last decade has witnessed an explosion of knowledge regarding the molecular signals that regulate embryonic mammary development. In the past several years in particular, we have gained great insight into the specification of the mammary line and the formation of the initial placodes. We are also just beginning to understand which signaling pathways contribute to the initiation of ductal development. Many more details require further elucidation, but there also remain many relatively untouched questions. We offer only a few for consideration: (1) What is the nature of the cell movements involved in the formation of the placodes and buds? (2) Are there similarities between these cell movements and the processes involved in metastasis? (3) What distinguishes mammary buds from other embryonic skin appendages? (4) What controls the development of the mammary fat pad and how do adipocytes change the nature of ductal morphogenesis? (5) Are the embryonic mammary epithelial cells similar to the recently defined mammary stem cell? If so, how are they established during embryonic development and how are they distributed during ductal outgrowth? (6) Are embryonic signaling pathways reactivated in breast cancer and do they contribute to pathogenesis? These are just a few of the many interesting problems that remain to be tackled to better understand the initial development of the mammary gland. We look forward to future surprises.
The authors would like to thank Drs. Kata Boras-Granic and Minoti Hiremath for valuable discussions. We appreciate Dr. Sarah Millar's willingness to share unpublished results. Also thanks to Drs. Minoti Hiremath and Sarah Hatsell for skillful artwork. This work was supported by National Institutes of Health grants DK055501, DK069542 and DK073941 to JW and CA129905 to PC, and DOD-BRCP-BC093088 to PC.
Editors: Mina Bissell, Kornelia Polyak, and Jeffrey Rosen
Additional Perspectives on Mammary Gland Biology available at www.cshperspectives.org