Key signaling nodes in mammary gland development and cancer: β-catenin

β-Catenin plays important roles in mammary development and tumorigenesis through its functions in cell adhesion, signal transduction and regulation of cell-context-specific gene expression. Studies in mice have highlighted the critical role of β-catenin signaling for stem cell biology at multiple stages of mammary development. Deregulated β-catenin signaling disturbs stem and progenitor cell dynamics and induces mammary tumors in mice. Recent data showing deregulated β-catenin signaling in metaplastic and basal-type tumors suggest a similar link to reactivated developmental pathways and human breast cancer. The present review will discuss β-catenin as a central transducer of numerous signaling pathways and its role in mammary development and breast cancer.


Introduction
β-Catenin plays a signifi cant role in the regulation of mammary development and tumorigenesis. At the plasma membrane, β-catenin aids cadherins in maintain ing mammary epithelial integrity. In the nucleus, β-catenin regulates gene expression programs that are essential for mammary stem cell biology during mammary development. Loss of β-catenin from cell-cell adhesive junctions or elevation in the cytosol and nucleus can occur independently by numerous routes. Importantly, both events predispose the breast to cancer. In the present article we review pathways relevant to breast that feed into the β-catenin signaling node, the role of β-catenin signaling in normal mammary development, and the eff ects of deregulating β-catenin on breast cancer.

Cadherin association
Breast epithelium is bilayered, comprising an inner luminal layer and a surrounding basal layer. In normal breast, the majority of β-catenin localizes to cell-cell adherens junctions through association with cadherins and only a minor portion is involved in signaling.
Th e importance of the cadherin/catenin complex for epithelial integrity and survival has been demonstrated by conditional deletion of E-cadherin. E-cadherin null cells undergo apoptosis but in the absence of p53 these cells survive and recapitulate many features of lobular breast cancer [1][2][3]. In humans, lobular breast cancer is associated with irreversible E-cadherin loss and transient changes in cadherin expression facilitate metastasis in many ductal breast cancers [4].

Canonical Wnt signaling
While cadherin/catenin complexes are critical to mammary integrity, cytosolic and nuclear β-catenin serve a distinct role, transducing signals from multiple pathways into cell-context-specifi c gene expression patterns that are essential for mammary development. Th is role has been studied most intensively within the context of canonical Wnt signaling.
Wnts are secreted morphogens that provide positional information to neighboring cells through canonical and noncanonical signaling routes (for recent reviews, see [21,22] -primary references and a detailed description of Wnt pathways can be accessed online: http://www. stanford.edu/~rnusse/wntwindow.html). In brief, the canonical Wnt pathway operates by stabilizing cytosolic βcatenin ( Figure 1). In the absence of Wnt signals, a multiprotein destruction complex -including the tumor suppressor adenomatous polyposis coli (APC) and Axinfacilitates serine/threonine phosphorylation of the Nterminal domain of β-catenin, which leads to its ubiquitination and proteosomal destruction. Wnt ligands prevent this sequence of events by binding to Frizzled receptors and signaling through the associated low-density lipoprotein-related proteins, LRP5/6. LRP signal ing via Disheveled and Axin inhibits the destruction complex and thus induces cytoplasmic accumulation of β-catenin. Th is accumulation promotes β-catenin trans loca tion to the nucleus, where it associates with T-cell factor/Lef DNA binding proteins, displaces trans crip tional repressors, and recruits B-cell lymphoma-9/legless, pygo pus homolog 2 (Pygo2) and other factors to induce cell-context-specifi c gene expression. Canonical Wnt signal ing is highly conserved and, importantly, regulates stem cell dynamics in many tissues, including the breast [23].

Other pathways regulating β-catenin signaling
Although canonical Wnt signaling is the most wellknown path to β-catenin stabilization, numerous other signaling pathways -many of which are aberrant in breast cancer -also regulate β-catenin stability and trans criptional activity ( Figure 2). For example, the prolyl isomerase Pin1 prevents association of β-catenin with APC, providing a major Wnt-independent stabilization mechanism [24]. Akt and ILK also stabilize β-catenin by inhibiting GSK3β activity. Importantly, the tumor suppressor phos phatase and tensin homolog (PTEN) inhibits both path ways [25,26]. Additionally, members of the p53-DNA damage pathway, Siah-1 (RING domain protein; trans criptional target of p53) and Ebi (F box protein) interact with APC to form an alternative destruction complex that recognizes and destroys βcatenin regardless of its phosphorylation status [27][28][29]. Th e p53 homolog, ΔNp63, was shown to be a target of p53-dependent destruction. During cancer, loss of p53 permits the aberrant interaction of ΔNp63 with protein phosphatase (PP2A), which inhibits GSK3β leading to βcatenin stabilization [30]. Finally, critical kinases of the NF-κB signaling pathway (IκB kinase (IKK)α and IKKβ) diff erentially regulate β-catenin trans crip tion as well as activity [31]. IKKα phosphorylation increases stability and transcriptional activity of β-catenin, while phosphory lation by IKKβ decreases both [31]. In certain cell contexts, Wnts stimulate alternative eff ectors termed non canonical pathways that can anta go nize the β-catenin pathway ( Figure 1). A web of regula tory pathways thus con verges on cytosolic β-catenin and becomes trans duced into a series of thresholds that exquisitely regulate the timely development and cyclical remodeling of the mammary gland. (See Table 1 for a summary of the phenotypes of transgenic and knockout mice used to study Wnt signaling in mammary develop ment and cancer.)

β-catenin signaling and mammary stem cells
In many epithelia, including breast, Wnt pathway activity is critical for stem cell self-renewal and multipotency [23]. β-Catenin signaling exerts powerful eff ects on mammary stem cells and progenitor cells. In mice, mammary stem cells reside among diff erentiated myoepithelial cells of the basal layer and alveolar progenitors are scattered among the diff erentiated cells of the luminal layer [32,33]. Recent studies have shown that Lrp5 is expressed in the basal epithelium and cells with high expression have a 200-fold greater ability to regenerate a mammary tree when transplanted into cleared mammary fatpads. Conversely, Lrp5 -/cells lack this capability and show senescence in culture, demonstrating that canonical Wnt signaling is a critical for normal mammary stem cell maintenance [34]. Support ing this concept, exogenous Wnt 3A protects mammary stem cells from senescence and the loss of multipotency associated with long-term culture. Furthermore, MECS derived from mice where Wnt signaling is mildly derepressed outcompete wild-type MECS in mammary reconstitution assays [23]. Th e diverse roles of β-catenin signaling during mammary development, which we will now review, may thus be understood in a unifi ed way when viewed through the lens of the highly conserved role of canonical Wnt signaling in regulating stem cell dynamics.

Wnt/β-catenin signaling during embryonic mammary development
Wnt signaling is essential for specifi cation and morphogenesis of the mammary gland [35]. Canonical Wnt signaling reporters defi ne the mammary lines and later become restricted to cells forming placodes [36]. Th e functional importance of Wnt signaling is demonstrated by experiments showing that Dkk1 suppresses early canonical Wnt signaling, subsequent Wnt10b induction along the mammary line and all placodal development [36]. Targeting other positive acting elements of the Wnt pathway, such as Lrp6, Lrp5, Lef1 and Pygo2, also results in placodal impairments, ranging from loss to reduced size and degeneration, while stimulating β-catenin signaling produces the converse eff ect -acceleration, expansion and induction of placodes and placodal markers (Wnt10b and T-box transcription factor-3) [37][38][39][40]. Candidate activators of early β-catenin signaling include Wnt3, Wnt6 and Wnt10b, which are expressed diff usely throughout the ectoderm as well as a genetic hierarchy of factors, including ventral bone morphogenetic protein-4, dorsal T-box transcription factor-3, neuregulin-3 and somitic Fgf10, which function upstream to defi ne the dorsal-ventral position of Wnt10b and Lef1 induction along the mammary line [35,[41][42][43][44][45][46][47][48]. While these up stream factors are body-site specifi c, Wnt/β-catenin signaling is required universally by all mammary placodes -and indeed by all ectodermal appendages [36]. Downstream targets remain obscure but probably include ectodysplasin-A, which promotes placodal cell fate along the mammary line when overexpressed [49].

β-catenin signaling during pubertal mammary development
Hormones temporally regulate postnatal development of the mammary gland, but their actions are relayed spatially by local paracrine factors. During puberty, estrogen, growth hormone and local IGF stimulate cell proliferation within terminal end buds to produce ductal extension. No direct connection has been established as yet between estrogen and canonical Wnt signaling. Several Wnts peak in expression during puberty, however, and Wnt5a and Wnt7b mRNAs are particularly enriched in terminal end buds and Wnt2 in the surrounding stroma [58][59][60][61][62]. Most reporters of canonical Wnt Figure 2. β-Catenin is at the hub of multiple signaling pathways. Many signaling pathways regulate the stability or binding interactions of β-catenin. In the Wnt pathway, glycogen synthase kinase-3β (GSK3β) and casein kinase 1 (CK1) phosphorylate the N-terminal degron sequence of β-catenin to facilitate its destruction. The phosphatidylinositol-3-kinase (PI3K) and phosphatase and tensin homolog (PTEN) pathways also impinge upon β-catenin phosphorylation by regulating GSK3β activity. In addition, p53 induces the degradation of β-catenin through protein interactions involving Seven in absentia homolog 1 (Siah1), Siah interacting protein (SIP) and EBV-induced G-protein coupled receptor (Ebi), resulting in ubiquitination and degradation of β-catenin. Pin1 binds to β-catenin phosphorylated on S246P to prevent its association with adenomatous polyposis coli (APC). In the NF-κB pathway, IκB kinase (IKK)α and IKKβ phosphorylate β-catenin throughout the protein to activate and inhibit transcription, respectively, although the N-terminus is essential for IKKα regulation. Some proteins, such as those within the transforming growth factor beta (TGFβ) pathway and Sox17, regulate β-catenin in the nucleus by modulating its interaction with transcriptional co-activators Tcf/Lef. Other proteins, like enhancer of zeste homolog 2 (EZH2), interact with β-catenin to promote its translocation into the nucleus. A number of tyrosine kinases phosphorylate (both membrane bound and cytosolic) β-catenin to prevent its binding to the cadherin complex at the cell membrane. Src, epidermal growth factor receptor (EGFR), and erythroblastic leukemia viral oncogene-2 (ErbB2) have been shown or implicated to phosphorylate β-catenin on Y654, while Abelson tyrosine kinase (Abl) phosphorylates Y489. β-TrCP, beta-transducing repeat-containing protein; DKK, Dickkopf; Lgs, legless; LRP, low-density lipoprotein-related protein; MAPK, mitogen-activated protein kinase; NLK, Nemo-like kinase; Pygo2, pygopus homolog 2; sFRP, secreted frizzled-related protein; WIF, Wnt inhibitory factor. signaling remain silent, but weak Conductin-lacZ expression is seen in basal cells and stroma surrounding the terminal end buds [23]. Conductin/Axin2 is an endogenous target gene that is faithfully expressed in response to canonical Wnt signaling and serves as a nega tive feedback loop to downregulate the pathway [63]. Th ese observations indicate that β-catenin signaling occurs at low levels. Genetically reducing canonical Wnt receptors produces mild eff ects: Lrp5 -/mutants and Lrp6 +/mutants show fewer terminal end buds, Lrp5 -/mutants show delayed ductal extension, and Lrp6 +/mutants show reduced ductal branching. Compound Lrp6 +/-;Lrp5 -/mutants show no ductal outgrowth, however, demon strating a requirement for canonical Wnt pathway activity [34,39,40]. Current evidence suggests that noncanonical Wnt antagonism of canonical Wnt pathways is important. For example, loss of Wnt5a, which stimulates noncanonical signaling pathways during other polarized morphogenic events, increases cytoplasmic and nuclear β-catenin and accelerates ductal outgrowth [64]. Importantly, Wnt5a is an essential mediator of TGFβ, suggesting that low thresholds of β-catenin signaling are maintained during pubertal ductal morphogenesis through TGFβ and Wnt5a antagonism.

Links between progesterone and β-catenin signaling during ductal maturation
A signifi cant body of data shows that progesterone not only promotes Wnt expression but also induces competence to respond to β-catenin signaling within specifi c mammary progenitors. Progesterone receptor (PR) is expressed in most luminal cells in juveniles but becomes downregulated and interspersed with PR-negative cells in adults. Th is patterning event requires PR and generates PR-negative progenitors that produce side branches and alveoli during pregnancy [65]. Expression of a stabilized form of β-catenin lacking N-terminal phosphorylation degron sequences (MMTV-ΔN89β-catenin) uniformly within the luminal epithelium induces signaling within this progenitor population, resulting in precocious develop ment at regular intervals along ducts [66,67]. In the absence of PR, however, ΔN89β is only able to produce this eff ect at ductal tips [66,67]. Th is observation suggests that PR specifi es and spatially patterns alveolar progenitors by inducing a factor that is essential for βcatenin's nuclear entry or transcriptional activity and thus renders cells poised to receive a β-catenin signal [66,67]. Proges terone also induces several paracrine factors -for example, Wnt4 and receptor activator of NF-κB ligand (RANKL) -in PR-positive luminal cells that signal to their neighbors [68][69][70]. A recent study has highlighted the importance of these paracrine eff ectors during estrus, showing that their expression correlated with β-catenin target gene expression in both luminal and basal cells and with increased stem cell activity [71].
In sharp contrast, a series of studies have shown that activation of β-signaling within specifi c hormone receptor-negative luminal cell populations promotes and is required for alveologenesis. Expression of MMTV-ΔN89β-catenin within luminal cells induces precocious alveologenesis in virgin mice [67,75]. β-Catenin signaling reporters are transiently expressed during normal alveologenesis, and inhibiting endogenous β-catenin signaling within luminal cells by expressing a negative regulator (Axin) or a dominant negative form of βcatenin (β-engrailed) prevents alveologenesis and caused alveolar degeneration [36,50,76,77]. Together, these studies establish that luminal β-catenin signaling is essential for alveologenesis and survival.
Th e upstream activator remains to be defi ned but is unlikely to involve Wnts since luminal cells express no obvious Wnt receptor and alveologenesis proceeds in the absence of Lrp5 [34]. Intriguingly, RANKL stabilizes βcatenin in other contexts and, like β-catenin, specifi cally promotes the proliferation of PR-negative cells [78,79]. RANKL -/mice and mice expressing Axin show similar impairment in alveologenesis [76,79]. Impor tantly, Pin1, which pro vides a Wnt-independent route to β-catenin stabilization, is critical for alveolar develop ment during pregnancy [24,80]. Downstream targets of β-catenin signaling remain obscure. C-myc and cyclin D 1 are reduced and increased in mice where the pathway is suppressed and is increased, respectively [67,[75][76][77].

Deregulating β-catenin signaling in mammary stem and progenitor cells induces cancer
Just as regulation of stem cells is key to understanding the function of β-catenin signaling during embryonic and adult mammary development, its proto-oncogenic properties may be viewed as arising from unbridled stem and early progenitor expansion. More than 25 years ago, retroviral insertion resulting in Wnt1 activation was found to induce mammary tumors -and, subsequently, transgenic expression of Wnt1, Wnt3 and Wnt10b was shown to produce similar eff ects [61,83,84]. A series of studies have revealed that exposure of glands to excessive Wnt1 or activated ΔN89β-catenin leads to progressive accumulation of cells displaying stem and progenitor marker profi les and enhanced colony-forming ability and bipotency in vitro [67,[85][86][87][88]. Reporter studies indicate that Wnt1 activates β-catenin signaling in a paracrine manner, increasing basally located multipotent stem cells at the expense of diff erentiated myoepithelial cells [67]. Th e pivotal role of a canonical Wnt pathway versus a noncanonical Wnt pathway in this process has been demonstrated by the fact that Lrp5 -/mice are highly resistant to Wnt1 tumor development and show signifi cantly delayed tumor onset [39,40]. Collectively, these studies indicate that paracrine signals from MMTVexpressing luminal cells promote β-catenin signaling within a basal stem cell population and induce mixed lineage tumors. In keeping with this scenario, continuous Wnt1 expression is required for tumor maintenance [89].
Mice expressing activated β-catenin from the MMTV promoter also develop adenocarcinoma with 100% penetrance [75,90]. In this mouse model, however, reporter studies have shown that β-catenin signaling occurs within a distinct subset of luminal cells scattered along ducts with alveolar progenitor marker profi les and functional characteristics [67]. Th ese studies show that deregulation of β-catenin signaling within luminal progenitor cells also produces tumors of mixed lineage [67,[85][86][87][88]. Signi fi cantly, in humans, basal-type hormone receptor-negative breast cancer is thought to arise from luminal progenitors. Target genes include matrix metalloprotein ases, cyclin D 1 and cmyc, which are upregulated in both MMTV-Wnt1 and MMTV-ΔN89β-catenin glands. Cyclin D 1 , however, is dispensable for tumor initiation [81,91]. In fact, MMTV-ΔN89β-catenin mice lacking cyclin D 1 present accelerated tumor onset and higher tumor burden [91]. Lastly, canonical Wnt pathway activity has been implicated in mediating radiation resistance in mammary Sca1-positive progenitors by inducing Survivin expression [92].
Th e response to β-catenin signaling is highly context dependent. K5-ΔN57β-catenin mice develop basal preneo plasia and invasive ductal carcinomas, while activation of β-catenin by MMTV-cre-targeted or whey acidic protein-cre-targeted exon 3 deletion shows squamous metaplasia [74,93]. Likewise, a mutant APC generated mammary adenocarcinomas when expressed in the basal layer but induced squamous metaplasia when expressed in diff er en tiating luminal cells [94]. APC mutants have also revealed tissue-specifi c dose responses. Th e APC Min/+ mutation, which results in high levels of β-catenin, predominantly causes intestinal polyposis, while APC mutants that generate intermediate levels of β-catenin preferentially induce mammary tumors [95]. Th e strain background infl uences the tumorigenic properties of APC Min/+ and genetic studies have begun to identify tissue-specifi c modifi er loci [96]. Tumor microenvironmental factors also exert powerful eff ects on the tumor phenotype. For example, inhibiting TGFβ signaling in stromal cells by expression of a dominant negative receptor (metallothionine inducible dominant-negative TGFβ receptor II transgene) increases β-catenin signaling within the epithelium of MMTV-Neu-induced and MMTV-PyVmT-induced tumors. Th is is accompanied by a remarkable change in tumor phenotype from one that is principally luminal to one with mixed luminal and basal cell types [97]. Similar eff ects are produced by eliminating the TGFβ target gene and the paracrine eff ector, Wnt5a [97]. Th e tumor microenvironment is in turn infl uenced by Wnt1, which impairs laminin deposition, disrupts cell polarity, promotes stromal hyperplasia and induces hedgehog signaling activity, suggesting important connections between these pathways in tumor onset [67,98].

β-catenin in human breast cancer
Studies in mice strongly suggest that deregulated βcatenin signaling increases the breast cancer risk by inducing stem and early progenitor cell accumulation [67,85]. Consistent with this hypothesis, a gene expression signature derived from MMTV-Wnt1 tumorinitiating cells was found to have prognostic value in determining poor outcome associated with specifi c breast cancer types, including basal-like and hormone receptor-negative cancers [99]. Genes of the canonical Wnt pathway featured prominently in a breast-to-lung metastasis signature that was signifi cantly correlated with poor prognosis and reduced overall survival. Finally, Wnt1 tumors share transcriptional patterns with BRCA1 +/tumors and human basaloid tumors, and several studies have linked Lrp6 and nuclear β-catenin with triple-negative breast cancer [40,[100][101][102][103].
Stabilizing mutations in the β-catenin N-terminal sequence found in other human cancers are uncommon in breast cancer and, until recently, were only reported in benign fi bromas [104,105]. Such mutations have now been found in 25% of metaplastic breast cancers [106]. In contrast to the rarity of mutation, β-catenin expression and/or localization are often abnormal in human breast cancer [5,24,[106][107][108][109]. Increased cytoplasmic and nuclear β-catenin levels have been documented in 40% of primary breast cancers and correlated with poor prognosis and worse patient survival [24,107,[109][110][111][112][113][114]. Th ese observations suggest that upstream elements of Wnt and/ or other pathways that stabilize β-catenin catenin are deregulated in breast cancer. Indeed, a signifi cant number of breast tumors develop in an environment of unopposed Wnt ligand activity [112,[115][116][117][118][119][120]. Inactivation of genes encoding secreted Wnt pathway inhibitors, such as secreted frizzled-related protein (sFRP), Wnt inhibitory factor and DKK, occurs frequently in breast cancer. sFRP1 levels are diminished in infi ltrating (or invasive) ductal carcinoma, and gain and loss of sFRP1 suppresses and promotes invasive ness of breast cancer, respectively [112,116,117,119,120]. sFRP1, sFRP5, Wnt inhibitory factor-1 and DKK3 are transcrip tionally silenced by promoter methy lation in 27 to 100% of breast cancers [118,[121][122][123][124][125]. Silencing of sFRP1, in particular, is a prognostic indicator of poor overall survival. One study found DKK1 to be preferentially expressed in hormone receptor-negative cancers; as DKK is a Wnt target gene, however, this elevation may actually indicate augmented pathway activity [126].
Th ere is also mounting evidence for excess Wnt ligand production, aberrant receptor activation and loss of noncanonical Wnt antagonism in human breast cancer. Wnt2, Wnt4 and Wnt10b mRNA upregulation has been reported [112,[127][128][129][130][131][132][133][134]. Wnt5a, which negatively regulates β-catenin activity during mammary development, is lost in a proportion of human infi ltrating (or invasive) ductal carcinoma and is a powerful predictor of recurrence [135]. FRZ1 and FRZ2 upregulation occurs and a splicing mutation in LRP5 that induced β-catenin activity was found in 85% of tumors [136,137]. Two studies have reported increased LRP6 expression in human breast cancer, particularly in triple-negative basal-like subtypes [40,102]. One study showed that Mesd, an inhibitor of Lrp5/6 folding, suppressed growth of MMTV-Wnt1 tumors despite the presence of excess ligand and without adverse eff ects on normal Wnt-driven processes, suggesting that Lrp6 may have potential as a therapeutic target [102].
An increasing body of evidence suggests that stabilization of β-catenin by Wnt-independent routes, such as Pin1, p53, PTEN/Akt and NF-κB pathways, plays a signifi cant role in breast cancer. For example, Pin1 upregulation correlates with increasing cancer grade (25% in ductal carcinoma in situ, 71.4% in grade II tumors, and 89.5% in grade III tumors) [164], and is associated with increased expression of β-catenin in breast tumors [24]. Components of the NF-κB pathway, which can diff erentially regulate β-catenin, are also activated in estrogen receptor-negative/ErbB2-positive breast cancers (86%) [165].
Loss of PTEN and p53 tumor suppressors has also been linked to induction of β-catenin in breast cancer. Germline PTEN mutations are common in patients with Cowden's syndrome, showing a lifetime risk of 25 to 50% for developing breast cancer, and sporadic PTEN loss correlates with invasive breast carcinomas (48%, n = 151) [166,167]. Signifi cantly, overexpression of PTEN partially inhibits Wnt-induced tumorigenesis; conversely, loss of PTEN and activation of Akt caused an expansion of normal and malignant breast stem cells via activation of β-catenin [26,167,168]. Th ese data suggest that the ability of PTEN to act as a tumor suppressor is mediated partly through the Wnt pathway. In addition to PTEN, inhibitor of diff erentiation binding-1, Slit-2 and melanoma diff erentiation association gene-7 proteins all regulate βcatenin in breast cancer cells through the phosphatidylinositol-3-kinase pathway [169][170][171].
Mutations in the p53 gene cause the heritable disease Li-Fraumeni syndrome, which is characterized by an 18-fold enhancement in susceptibility to breast cancer. p53 mutation occurs in 50% of all human cancers and in 20% of sporadic breast cancer (for a review see [172]). β-Catenin is downregulated by p53 and is also degraded through a p53-dependent mechanism, which included Siah1, SIP and Ebi [27][28][29]. Adding to the complexity, Siah1 has splice variants that diff erentially aff ect βcatenin in breast cancer cells [173,174]. MMTV-ΔNβcatenin tumors induced in p53 +/mice are more malignant and metastatic than those in p53 +/+ mice [175].

Conclusions
Evidence from murine models show that deregulated βcatenin signaling in basal stem cells or luminal progenitors induces breast tumors. Th is event correlates with disturbances in the homeostatic balance between stem cell and early progenitor cell self-renewal and the production of diff erentiated progeny. Th ere are numerous pathways, in addition to the canonical Wnt pathway, which regulate β-catenin. Mutations in β-catenin are rare but deregulated β-catenin signaling occurs frequently in human breast cancer and may be particularly relevant for basal-type and triple-negative subtypes, suggesting important links between reactivation of developmental pathways and this breast cancer subtype.

Competing interests
The authors declare that they have no competing interests.