Contributed equally.

Signal transducer and activator of transcription (STAT) 5A and STAT5B are members of a well-known group of transcription factors [1-3]. STATs were fi rst identifi ed as members of a non-tyrosine kinase-containing cytokine receptor-activated signaling pathway [4]. Th ere are seven identifi ed components of the STAT transcription factor family: STAT 1 to 4, 5A, 5B, and 6. STAT5A and STAT5B are closely related family members thought to be a result of gene duplication [1,5]. During normal mammary gland development, STAT5A plays the more prominent role [6-10], whereas both STAT5A and STAT5B have been des cribed as contributing to breast cancer pathophysiology [11-17]. After interactions of cytokines, hormones, and growth factors with their respective cell surface receptors , STATs are activated, primarily by phosphorylation at tyrosine residues, and then dimerize and translocate to the nucleus. Once in the nucleus, STATs can initiate gene transcription [2]. Th e activity of STAT proteins can be infl uenced by serine phosphorylation [18], regulated dephosphorylation [19], and interactions with cellular proteins, including adhesion and basement membrane molecules. In mammary epithelial cells, the STAT5A/B (STAT5) pathway modulates three diff erent cellular outcomes: diff erentiation, survival, and proliferation (Figure 1). Th e two STAT5 proteins, STAT5A and STAT5B, can homodimerize and heterodimerize. Th e relative impact of STAT5 on the three diff erent cellular outcomes varies in normal as compared with malignant mammary epithelial cells as well as between diff erent types of normal and malignant mammary epithelial cells. In mammary epithelial cells, STAT5 can be activated through a few diff erent and sometimes interacting signal-ing pathways (Figure 2). Many of the pathways that mediate normal development and lactational diff erentia-tion (Figure 3) also regulate STAT5 activation in breast cancer cells (Figure 4) but with a few distinctions. Janus kinase (JAK) 2, a tyrosine kinase, is a key signaling node for STAT5 activation in both normal and malignant mammary epithelial cells [17,20,21]. In normal mammary epithelial cells, this is predominantly mediated through prolactin (PRL) acting through the PRL receptor (PRLR) upstream of JAK2/STAT5 [22-25]. Interestingly, in breast cancer cells, JAK1 activation through a PRLR-JAK2 interaction can enhance STAT5 signaling [26]. Human cellular-Src (c-Src) is a second key signaling node for STAT5 activa tion in normal and malignant mammary epithelial cells. Loss of c-Src interrupts STAT5 activation during preg nancy, and in malignant cells c-Src has been shown to mediate STAT5 activation downstream of estrogen/estrogen receptor-alpha (ERα) …

Signal transducer and activator of transcription (STAT) 5A and STAT5B are members of a well-known group of transcription factors [1][2][3]. STATs were fi rst identifi ed as members of a non-tyrosine kinase-containing cytokine receptor-activated signaling pathway [4]. Th ere are seven identifi ed components of the STAT transcription factor family: STAT 1 to 4, 5A, 5B, and 6. STAT5A and STAT5B are closely related family members thought to be a result of gene duplication [1,5]. During normal mammary gland development, STAT5A plays the more prominent role [6][7][8][9][10], whereas both STAT5A and STAT5B have been des cribed as contributing to breast cancer pathophysiology [11][12][13][14][15][16][17]. After interactions of cytokines, hormones, and growth factors with their respective cell surface receptors, STATs are activated, primarily by phosphorylation at tyrosine residues, and then dimerize and translocate to the nucleus. Once in the nucleus, STATs can initiate gene transcription [2]. Th e activity of STAT proteins can be infl uenced by serine phosphorylation [18], regulated dephosphorylation [19], and interactions with cellular proteins, including adhesion and basement membrane molecules. In mammary epithelial cells, the STAT5A/B (STAT5) pathway modulates three diff erent cellular outcomes: diff erentiation, survival, and proliferation ( Figure 1). Th e two STAT5 proteins, STAT5A and STAT5B, can homodimerize and heterodimerize. Th e relative impact of STAT5 on the three diff erent cellular outcomes varies in normal as compared with malignant mammary epithelial cells as well as between diff erent types of normal and malignant mammary epithelial cells.
In mammary epithelial cells, STAT5 can be activated through a few diff erent and sometimes interacting signaling pathways ( Figure 2). Many of the pathways that mediate normal development and lactational diff erentiation ( Figure 3) also regulate STAT5 activation in breast cancer cells ( Figure 4) but with a few distinctions. Janus kinase (JAK) 2, a tyrosine kinase, is a key signaling node for STAT5 activation in both normal and malignant mammary epithelial cells [17,20,21]. In normal mammary epithelial cells, this is predominantly mediated through prolactin (PRL) acting through the PRL receptor (PRLR) upstream of JAK2/STAT5 [22][23][24][25]. Interestingly, in breast cancer cells, JAK1 activation through a PRLR-JAK2 interaction can enhance STAT5 signaling [26]. Human cellular-Src (c-Src) is a second key signaling node for STAT5 activa tion in normal and malignant mammary epithelial cells. Loss of c-Src interrupts STAT5 activation during preg nancy, and in malignant cells c-Src has been shown to mediate STAT5 activation downstream of estrogen/estrogen receptor-alpha (ERα) and epidermal growth factor (EGF) signaling and to contribute to activation through the PRL/PRLR/JAK2 pathway [4,16,[27][28][29]. One example of the impact of these interacting signaling nodes is shown by the dual activation of STAT5 by estrogen and EGF [28]. When activated by estrogen in breast cancer cells, STAT5 requires c-Src and EGF, but if the EGF receptor (EGFR)/ c-Src pathway is hyperactivated, the responsive ness to changes in the estrogen pathway activation is altered and this can contribute to the development of endo crine therapy resistance. A second example lies behind the deleterious eff ect of recombinant human erythro poietin (rHuEPO) on the sensitivity to trastuza mab in breast cancer cells [30,31]. Trastuzamab is a monoclonal antibody that targets the human EGFR 2/Neu receptor, an important growth stimulus for a subset of breast cancers. Resistance to trastuzumab occurs when rHuEPO stimulates JAK2, leading to the activation of both c-Src and STAT5. In normal mammary epithelial cells, growth hormone (GH) acting through the GH receptor and JAK2 can activate STAT5 [32]. Th e EGFR and ErbB4 -or verb-b2 erythroblastic leukemia viral oncogene homolog 4, neuro/glioblastoma-derived onco gene homolog (avian) -are reported to directly associate with and activate STAT5 in mammary epithelial cells [33][34][35]. Th e IGF pathway plays an important role in STAT5 activation in mammary epithelial cells, and signaling through the IGF receptor can be processed through the JAK/STAT pathway [36].
Activity of the STAT5 pathway in mammary epithelial cells is also regulated at the level of STAT5A and STAT5B gene expression. Expression levels can be modifi ed by changes in the activity of estrogen and progesterone signal ing pathways [37][38][39] and by either EGF or IGF stimulation [40,41]. E74-like factor 5 (ets domain transcription factor) (ELF5) appears to modulate both expression levels and activation of STAT5, perhaps through changes in expression levels of suppressor of cytokine signaling (SOCS) family members [42].
Th e impact of STAT5, a known transcription factor, on the cell lies, at least in part, through regulation of gene expression. During lactation, STAT5 activation contributes to the high expression levels of milk protein genes. STAT5 activation also has been linked to regulat ing expression of the cell cycle control protein cyclin D1 both directly and indirectly [12,15,16,51,[60][61][62] and expres sion of receptor activator of nuclear factor-kappa-B ligand (RANKL) [62,63]. Signifi cantly, changes in STAT5 expres sion and activity can also modify the expression and activity of other STAT family members, most prominently STAT3 [60,64,65]. Th ese changes in STAT3 expression or activity then can modify gene expression patterns as a secondary eff ect.
In summary, STAT5 can be activated by diff erent and sometimes interacting signaling pathways in mammary epithelial cells. Th e prolactin, EGF, estrogen, GH, and signaling networks dominate in STAT5 activation in normal mammary gland development with contributions from growth hormone (GH), insulin growth factor (IGF), estrogen, epidermal growth factor (EGF), and ErbB4 signaling. PRL and GH work predominantly through their respective receptors prolactin receptor (PR) and growth hormone receptor (GHR) through Janus kinase 2 (JAK2) and are key mediators of pregnancy-induced mammary gland development. Estrogen and EGF acting through respective receptors estrogen receptor-alpha (ERα) and EGF receptor (EGFR) initiate pubertal mammary gland development and contribute to pregnancy-induced development. They can interact through human cellular-Src (c-Src) pathways. Transforming growth factor-alpha (TGF-α) is the second ligand from the EGF family to be shown to infl uence STAT5 activation levels in normal and cancer cells. IGF signaling through insulin growth factor-related receptors (IGFRs) may also include c-Src and, under some circumstances, JAK2 in both puberty-and pregnancy-induced development. The contribution of ErbB4 to STAT5 signaling is most prominent during lactation. In breast cancer, EGF and estrogen pathways acting through c-Src can drive proliferation and survival. JAK1 has been shown to increase PR/JAK2 activation in some settings. When the erythropoietin receptor (EPOR) is expressed in breast cancer cells and erythropoietin (EPO) is present, they can signal through JAK2 to STAT5 to promote resistance to trastuzumab therapy. ErbB4, v-erb-b2 erythroblastic leukemia viral oncogene homolog 4, neuro/glioblastoma-derived oncogene homolog (avian); STAT5, signal transducer and activator of transcription 5.  Research to date has identifi ed several factors that contribute to higher levels of STAT5 activation or, conversely, reduce activation levels in normal mammary epithelial cells. E74-like factor 5 (ets domain transcription factor) (ELF5) acts in the PRL/JAK2 signaling pathway to increase levels of STAT5 activation in normal mammary epithelial cells. Other signaling pathway components that increase STAT5 activation are EGF, ErbB2, transforming growth factor-alpha (TGF-α), estrogen acting with the progesterone pathway acting downstream, GH/JAK2, IGF, and c-Src. The TGF-β signaling pathway can decrease STAT5 activation levels. Cellular proteins that contribute to increased STAT5 activation include PI 3-kinase enhancer A (PIKE-A), serine/ threonine protein kinase Akt 1 (AKT-1), p21-activated kinase 1 (Pak1), the phosphotyrosine phosphatase Shp2, beta-integrin, and dystroglycan. Cellular proteins that have been shown to decrease STAT5 activation levels include caveolin-1 and suppressor of cytokine signaling 3 (SOCS-3). c-Src, human cellular-Src; EGF, epidermal growth factor; ErbB4, v-erb-b2 erythroblastic leukemia viral oncogene homolog 4, neuro/glioblastomaderived oncogene homolog (avian); GH, growth hormone; IGF, insulin growth factor; JAK2, Janus kinase 2; PRL, prolactin; Shp2, Src homology region 2 domain-containing phosphatase-2; STAT5, signal transducer and activator of transcription 5. execute the diff erent actions of STAT5 on cellular diff eren tiation, survival, and proliferation. Th ese gene expression changes may include both direct eff ects and secondary changes due to STAT5-mediated eff ects on other transcription factors, including other STAT family members such as STAT3.

NORMAL DEVELOPMENT AND LACTATIONAL
The STAT5 signaling node in normal mammary gland development STAT5 plays a critical role in the development and diff eren tiation of the normal mammary gland toward lactational competence [7]. It is essential for the generation of luminal progenitor cells that diff erentiate into alveolar Research to date has identifi ed several factors that can contribute to higher levels of STAT5 activation or, conversely, reduce activation levels in breast cancer cells. As in normal mammary epithelial cells, PRL/JAK2, c-Src, EGF/EGFR, ErbB4, TGF-α, estrogen and progesterone, and IGF pathways can increase STAT5 activation. In breast cancer cells, EPO/EPOR and HOXA1 also have been shown to increase STAT5 signaling. Two cellular proteins shown to increase STAT5 activation in breast cancer cells are breast tumor kinase (Brk) and transcription factor proto-oncogene v-Myb myeloblastosis viral oncogene homolog (avian) (c-Myb). As in normal mammary epithelial cells, caveolin-1 and SOCS-3 can downregulate STAT5 activation; however, in breast cancer cells, PTPN9 (protein tyrosine phosphatase, non-receptor type 9) also has been shown to downregulate STAT5 signaling. c-Src, human cellular-Src; EGF, epidermal growth factor; EGFR, epidermal growth factor receptor; EPO, erythropoietin; EPOR, erythropoietin receptor; ErbB4, v-erb-b2 erythroblastic leukemia viral oncogene homolog 4, neuro/glioblastoma-derived oncogene homolog (avian); HOXA1, homeobox A1; IGF, insulin growth factor; JAK2, Janus kinase 2; PRL, prolactin; SOCS-3, suppressor of cytokine signaling 3; STAT5, signal transducer and activator of transcription 5; TGF-α, transforming growth factor-alpha. cells, although it is not required for the appearance of mammary ductal cells or the production of mammary stem cells [21,66]. STAT5A is the more prominent STAT family member expressed in the mammary gland. During pubertal development, its absence results in defects in secondary ductal and side branching [62] and delayed diff er entiation of terminal end buds when coupled with mammary epithelial cell-targeted ERα overexpression [43]. During pregnancy, it plays an essential role in mammary gland lactational development and in diff erentiation and expression of milk protein genes [6]. However, in its absence, STAT5B can be upregulated, especially with serial pregnancies, and substitute for STAT5A [9]. STAT5 activation is downregulated at the onset of post-lactational involution [8].

BREAST CANCER CELLS
To date, a larger number of signaling pathways and proteins have been identifi ed as contributing to the activation of STAT5 during normal development and lacta tional diff erentiation than have been shown to downregulate this pathway ( Figure 3). Th is may be because inadequate levels of STAT5 activation result in the readily apparent defect of insuffi cient milk production whereas increased STAT5 activation produces the less obvious developmental abnormality of precocious alveolar diff erentiation during pregnancy and, only later, hyperplasia and cancer [55,60,66,67]. Estrogen signaling and progesterone signaling contribute to regulating normal levels of STAT5 expression in the non-pregnant gland [39,68]. Th e transcription factor ELF5 is reported to lie functionally upstream of STAT5 and downstream of prolactin [69]. Its absence compromises STAT5 expression levels and activity in mammary epithelial cells and results in loss of normal lobuloalveolar development [42]. While during normal development STAT5 is activated primarily by PRL signaling pathways, EGF and GH signal ing pathways can also contribute [32,70]. ErbB4 has a defi ned role in mediating STAT5 activation during lactation [35]. Activation of EGF signaling pathways by the introduction of expression of transforming growth factor-alpha (TGFα) through a mammary epithelial celltargeted transgene can interrupt the downregulation of STAT5 activation and promote mammary epithelial cell survival during involution [70]. PIKE-A, active in AKT signaling, associates with STAT5 and PRLR and its absence results in impaired mammary epithelial cell proliferation and lactation [51]. Pak1 interacts with and serine phosphorylates STAT5 and interruption of the normal function of Pak1 reduces lobuloalveolar growth and milk produc tion [18]. Ablation of Akt1, but not Akt2 or Akt3, interferes with STAT5 activation in late pregnancy and lactation [52,71]. Loss of one copy of the Akt2 gene coupled with ablation of Akt1 results in the loss of STAT5 activation associated with increased expression of caveolin-1 and SOCS-2, negative regulators of STAT5 [52]. Th e phosphotyrosine phosphatase Shp2 reciprocally modulates STAT5 and STAT3 activation in the mammary epithelium. Deletion of this molecule results in impaired STAT5 activation but slightly increased STAT3 activity [72]. Ablation of Src in normal mammary epithelial cells impairs STAT5 activation through the downregulation of PRLR [27]. Finally, structural molecules outside the mammary epithelial cell, including beta-integrin [73,74] and the basement membrane receptor dystroglycan [75], also play a part in establishing the normal levels of STAT5 activation required for lactational diff erentiation. Loss of dystroglycan interrupts mammary gland outgrowth and lactation competency that is correlated with decreased STAT5 activity. Negative regulators of STAT5 include TGF-β, which has been reported to downregulate prolactininduced JAK/STAT5 activation [76,77]. If caveolin-1 is absent, STAT5 activation is increased with subsequent development of mammary hyper plasia and cancer [55].
Candidate downstream mediators of STAT5 activity during normal development include RANKL [62,63] and cyclin D1 [51,62]. STAT5 also has the ability to bind to consensus sequences within the Akt1 locus which defi ne a unique promoter active only in mammary epithelial cells [61].

The STAT5 signaling node in normal mammary stem cells and cancer progenitor cells
STAT5 has a well-defi ned role in the regulation of stem and progenitor cells in hematopoietic systems [78][79][80]. Studies in the mammary gland reveal a role for STAT5 in the development of the mammary epithelial alveolar cell lineage [21,59,66] (Figure 5). Th e cellular repertoire of the mammary gland is generated by growth and diff erentiation from a stem cell component coordinated by hormones and cytokines. Mammary stem cells have a distinctive capacity for self-renewal and give rise to the three principal lineages that encompass the normal lobuloalveolar structure of the adult mammary gland: myoepithelial cells that form the basal layer of ducts and alveoli, ductal epithelial cells lining the lumen of ducts, and alveolar epithelial cells that have the capacity to synthesize milk proteins in large quantities [81]. In the absence of STAT5, mammary alveologenesis is abrogated through a reduction in the mammary luminal progenitor cell population [21,66]. STAT5 is required not only for proliferation and survival of alveolar cells but also for the generation of alveolar progenitor cells from stem cells. Importantly, the defect is limited to this lineage and loss of STAT5 does not aff ect the underlying mammary stem cell population. Ductal cells do develop in the absence of STAT5 [21], but while primary ductal extension is normal, impaired side branching is reported [62]. Consistent with the impact of STAT5 loss, gain of a constitu tively active STAT5A causes otherwise relatively quiescent ductal epithelial cells in virgin mice to undergo rapid expansion and develop into alveolar-like structures [59]. Consistent with the position of ELF5 as a STAT5 regulator, gain of ELF5 induces diff erentiation of luminal progenitor cells into alveolar cells [69]. In a reciprocal fashion, gain of STAT5 is associated with increased ELF5 expression levels [59], indicating that the two factors have a bi directional interaction.
Whether or not STAT5 plays a role in cancer progenitor cell pathophysiology is under investigation. Gain-offunction experiments using a Stat5 variant, cS5-F, in which serine 710 is mutated, demon strate that STAT5 overexpression in mouse models can lead to the development of ER + /PR + adenocarcinomas harboring a small fraction of CD44 + cells that are postulated to represent a population of cancer progenitor cells [66].

The STAT5 signaling node in mouse models of breast cancer
Th e impact of increasing or decreasing the STAT5 activity on mammary cancer initiation, promotion, and progression has been investigated from a mechanistic perspective by using genetically engineered mouse models ( Figure 6). Mammary epithelial cell-targeted over expression of genetically engineered STAT5 variants is suffi cient for initiation, promotion, and progression along a cancer pathway, resulting in mostly well-diff erentiated adenocarcinomas, sometimes exhibiting a papillary architecture, and a low percentage of undiff erentiated carcinomas [60,66,67]. Loss of caveolin-1 in mammary epithelial cells induces hyperactivation of STAT5A signal ing, leading to the development of mammary hyperplasias and well-diff erentiated cancers [55]. STAT5 also contri butes to mammary cancer progression initiated by mammary epithelial cell-targeted overexpression of PRL [23], simian virus 40 T antigen (TAg) [82], and TGF-α [70]. Signifi cantly, JAK2 is required to initiate mammary cancer development by PRL overexpression but is dispensable for cancer cell survival and proliferation [83], exerting a 'hit and run' eff ect in mammary carcinogenesis. Similarly, loss of STAT5A reduces the prevalence of ERα-initiated mammary preneoplasia but does not prevent the development of invasive cancer [43]. Upregulation of growth factors known to contribute to mammary carcino genesis, including ErbB2 [43,83] and cyclin D1 [43], is found in the cancers that develop in the absence of either JAK2 or STAT5a and that may be responsible for sustaining the cancer cell growth.

The STAT5 signaling node in benign human breast disease and cancer
Stat5 is expressed in a high proportion of human breast cancers [84,85]. Activation of STAT5 in breast cancers is statistically associated with higher levels of diff erentiation [85] and a better prognosis [84] and response to endocrine therapy when co-expressed with ERα [86]. Consistent with these results, higher expression levels of STAT5 target genes such as IGF-1 and SOCS-2 also correlate with a better prognosis [57]. Decreased expres sion levels of STAT5A are found more frequently in high-grade breast cancers as compared with normal breast tissue or ductal carcinoma in situ lesions. Th e absence of activated STAT5 correlates with a higher probability of not responding to endocrine therapy [87]. Th ese lower levels of STAT5A are associated with increased expression levels of proto-oncogene B-cell chronic lymphocytic leukemia/lymphoma 6 (BCL6), a transcriptional repressor that recognizes similar DNA target sequences [88]. Interestingly, a PRLR gain-of-function mutant resulting in increased STAT5 signaling is associated with the presence of multiple benign breast fi broadenomas [89].
A correlation between increased STAT5 expression and cellular transformation has also been shown in mammary epithelial cells in which activation levels of STAT5 are increased. Overexpression of the homeobox A1 (HOXA1) gene in these cells induces increased STAT5B expression that is associated with increased cell prolifera tion, survival, and oncogenic transformation [90]. Interestingly, STAT3, but not STAT5A, is simultaneously induced by HOXA1, and downregulation of either STAT5B or STAT3 is suffi cient to abrogate the phenotype. STAT5 Figure 6. Increased STAT5 initiates cancer and decreased STAT5 diminishes cancer promotion in genetically engineered mouse models. Increased levels of STAT5 expression and/or activation either directly by transgene-mediated STAT5 overexpression or indirectly through genetically engineered loss of caveolin-1 leads to initiation, promotion, and progression to mammary cancer. Decreased levels of STAT5 expression reduce cancer promotion initiated by transgene-mediated overexpression of prolactin, transforming growth factor-alpha (TGF-α), and estrogen receptoralpha (ERα) as well as mammary-targeted simian virus 40 T antigen. STAT5, signal transducer and activator of transcription 5. activation mediated by artifi cial induction of the EPO receptor into the benign non-invasive rat mammary cell line, Rama 37, leads to increased colony formation, invasion, migration, and changes in adhesion associated with increased extracellular signal-regulated kinase (ERK) and AKT [91]. Th e observations in human breast cells coupled with the pathophysiological studies in mouse models raise interesting questions about the role of STAT5 in human breast cancer. It would appear that, when the mechanistic mouse studies are taken together with the descriptive human studies, STAT5 activation is defi nitively compatible with cancer cell growth and increased levels of STAT5 activation may contribute to cancer progression. However, the cancers exhibiting STAT5 expression are more diff erentiated and statistically more likely to respond to endocrine therapy and perhaps are therefore associated with a better prognosis. Experiments in human breast cancer cell lines have demonstrated that this may be because STAT5 is able to maintain some of its ability to promote cellular diff erentiation in cancer cells, as it does during normal development. Th ese studies have identifi ed specifi c cellular behaviors and downstream genes that can be infl uenced by changes in STAT5 expression and activation (Figure 7).

MAMMARY CANCER DEVELOPMENT IN MOUSE MODELS
In some experimental systems, STAT5 has been shown to contribute to diff erentiation of breast cancer cells and this contribution is reminiscent of its role in normal mammary gland development. In BT-20 and T47D breast cancer cells, the combination of STAT5 and Jak2 overexpression induces a mesenchymal-to-epithelial transition when the cells are grown in a three-dimensional matrigel assay. Th e cells shift to a luminal epithelial cell phenotype and become less invasive [92]. Paralleling these results, diff erentiation as measured in a threedimensional culture assay is impaired when kinase-dead ErbB4 mutants that reduce STAT5 activation levels are placed into MDA-MB-468 breast cancer cells [33]. STAT5 has been found to attenuate prolactin signaling to activating protein-1 (AP-1), perhaps through direct binding. Loss of STAT5 in T47D cells increases prolactininduced AP-1 signaling, matrix metalloproteinase-2, and invasive behavior [93], and forced increased expression levels of STAT5 can inhibit motility of MCF-7 and T47D cells [13].
In other experiments, STAT5 activation in cancer cells enhances behaviors conventionally associated with advanced malignancy. Th e same experiments that demonstrated that forced expression of STAT5 in MCF-7 and T47D cells suppresses cell motility showed that it enhanced cell survival and anchorage-independent growth [13]. In MDA-MB-231 and BT-549 cells, knockdown of STAT5B can inhibit beta-1-integrin-mediated cell migration [14].
BCL6 represents a gene whose upregulation is associated with loss of diff erentiation of breast cancer cells when STAT5 expression is reduced [88]. STAT5A, but not STAT5B, is able to repress BCL6 expression through a prolactin-induced mechanism. Whereas STAT5 represses BCL6 expression, STAT3 increases BCL6 expression. However, STAT5 is dominant over its related family member STAT3 in regulating BCL6 expression levels in T-47D and SK-BR-3 cells [65]. Moreover, in MDA-MB-468 cells, the same authors demonstrated that simul taneous STAT5 and STAT3 activation resulted in decreased proliferation and increased sensitivity to paclitaxel and vinorelbine as compared with cells with STAT3 activation alone.
STAT5A signaling in breast cancer cells can be modifi ed by interactions with c-Myb [53] and Brk [54]. c-Myb and STAT5A associate in a PRL-inducible manner in T47D and MCF7 breast cancer cells, stimulate expression of STAT5A downstream genes, and are associated with increased PRL-induced cell proliferation. Brk can phos phorylate STAT5 through a mechanism involving signal-transducing adaptor protein 2 (STAP-2). Knockdowns of STAT5B, Brk, and STAP-2 equivalently reduce prolifera tion of T47D breast cancer cells.
Downstream STAT5 genes, including SOCS-3, can feedback to regulate STAT5 activity in breast cancer cell lines. For example, overexpression of SOCS-3 in T47D cells reduces PRL-induced STAT5 phosphorylation and this is correlated with decreased cell proliferation [94]. Expression of PTPN9 (protein tyrosine phosphatase, non-receptor type 9) is reported to reduce STAT5 activation coincident with growth inhibition as measured in soft agar assays by using SKBR3 and MDA-MB-231 breast cancer cell lines, perhaps (in whole or in part) through regulation of ErbB2 and EGFR phosphorylation [95].
STAT5 regulates expression of genes that promote cell survival and proliferation in breast cancer cells. For example, STAT5 can induce expression of heat shock protein 90-A (HSP90A), a protein that can promote cancer cell survival. In SKBR3 breast cancer cells, PRL increases HSP90A, and STAT5B activates the HSP90A promoter [96]. IGF was found to lie downstream of STAT5 in breast cancers, including ERα-negative breast cancer cells [12,57]. Cyclin D1 also lies downstream of STAT5 in breast cancers [12,15,16,38]. It should be understood that, while HSP90A, IGF, and cyclin D1 lie downstream of STAT5 activation, their expression patterns, activity, and impact on cellular growth can be modifi ed by other cellular factors and signaling molecules expressed in breast cancer cells [36,[97][98][99].
STAT5 activation has been correlated with response to endocrine therapy, although currently there is a disconnection between available in vivo and in vitro data.
Descriptive studies using human breast cancer tissue demonstrate a positive correlation between STAT5 activa tion and response to endocrine therapy [86,87], whereas constitutive activation of STAT5B in T47D Figure 7. STAT5 expression in human breast cancer is generally associated with increased diff erentiation. In human breast cancer tissue, STAT5 expression has been reproducibly associated with increased diff erentiation and a better prognosis and response to endocrine therapy. Experiments that alter STAT5 expression levels in breast cancer cell lines revealed a link between the presence of STAT5 and increased diff erentiation in BT-20, T47D, and MDA-MB-468 breast cancer cell lines. Increased cell motility, invasiveness, and migration are behaviors that can be found in association with decreased diff erentiation. In MCF-7 andT47D cells lines, STAT5 inhibits motility, and in the T47D cell line loss of STAT5 increases invasiveness and this is consistent with the correlation between STAT5 and increased diff erentiation. However, in MDA-MB-231 and BT-549, loss of STAT5 actually inhibits migration. Increased sensitivity to therapy is considered a good prognostic sign, and in the MDA-MB-468 cell line, STAT5 is correlated with increased sensitivity to paclitaxel and vinorelbine. However, in SKBR3, MCF-7, and T47D cell lines, experiments have correlated the presence of STAT5 with increased cell survival and, in T47D cells, resistance to tamoxifen and, in SKBR3 cells, a decreased response to trastuzumab. In SKBR3 and MD-MBA-231 cells, loss of STAT5 is actually correlated with growth inhibition. These sometimes consistent and sometimes confl icting results in diff erent breast cancer cell lines indicate that the relative impact of STAT5 on cell diff erentiation, survival, and proliferation can be cell linespecifi c. BCL6 gene expression and activating protein-1 (AP-1) signaling are reduced by increased STAT5 signaling and increased by reduced STAT5 signaling. Identifi ed downstream STAT5 genes in breast cancer cells include heat shock protein 90-A (HSP90A), insulin growth factor (IGF), and cyclin D1. BCL6, proto-oncogene B-cell chronic lymphocytic leukemia/lymphoma 6; STAT5, signal transducer and activator of transcription 5. breast cancer cells is reported to induce tamoxifen resistance [16].

Development of a unifying hypothesis of STAT5 action and future directions
An examination of the diff erent results from both normal and malignant mammary epithelial cells, in mice and in women, reveals some commonalities in the diff erent experimental systems employed. One is the mechanistic impact of STAT5 activation on diff erentiation in the mouse and cell line studies coupled with its correlation with diff erentiation in the human breast cancer tissue studies. In both normal and malignant mammary epithelial cells, STAT5 is positively associated with diff erentiation. Similarly, STAT5 contributes to improving cell survival and increasing cell proliferation in both normal and malignant cells. However, cellular diff erentiation appears to be the most frequently altered feature found when STAT5 activation or expression levels are high and therefore one signifi cant factor in a unifying hypothesis of STAT5 action.
A second feature of a unifying hypothesis of STAT5 action is that, although many of the signaling pathways that activate STAT5 and cellular proteins that interact with it are similar in normal and malignant mammary epithelial cells, diff erences in the degree to which the diff erent components contribute exist. In normal mammary epithelial cells, the PRL/PRLR/JAK2 pathway dominates, whereas in cancer cells signifi cant STAT5 activation can occur through EGF and c-Src with contributions from estrogen and progesterone signaling pathways and even the EPO pathway. Another diff erence between normal and malignant mammary epithelial cells lies in the relative contributions of STAT5A and STAT5B. In normal mammary epithelial cells, STAT5A is the dominant actor, whereas in malignant mammary epithelial cells, STAT5B can be the more signifi cant contributor.
A third feature for a unifying hypothesis for STAT5 action is its ability to contribute to cancer initiation, promotion, and progression and at the same time be a critical component of normal mammary gland development. In this respect, it shares the challenges of other growth factor pathways involved in normal mammary gland development, including estrogen, progesterone, EGF, and IGF pathways. When properly regulated, these pathways mediate normal growth, cellular proliferation, and diff erentiation but can contribute to neoplastic transition and unrestrained cellular proliferation when their activation becomes unbalanced or associated with oncogenic changes in the cell or both.
Finally, the contribution of STAT5 to the specifi cation of progenitor cells has to be considered in any unifying hypothesis of STAT5 action. Here, there is clear defi nition of its role in specifying alveolar cell lineage diff erentiation during normal development, but its contribution to cancer progenitor cell biology is less well defi ned. One can speculate that it may 'specify' a type of breast cancer progenitor cell, perhaps a more diff erentiated one that shares morphological or biological features with alveolar cells. In support of this conjecture, STAT5 overexpression in mice is associated with the develop ment of papillary adenocarcinomas, a more diff er entiated histological type that is not commonly found in genetically engineered mouse models of cancer.
At present, a unifying hypothesis of STAT5 action in mammary epithelial cells would include the following statements. STAT5 is a protein with dominant eff ects on cellular diff erentiation and lineage specifi cation that lead to proliferation and survival of specifi c mammary cell types. It has the ability to contribute to growth and survival in both normal and cancer cells. Well-established mammary cellular growth factors and hormones from outside the cell acting through their receptors are the most frequent activators of STAT5. Th e degree to which the diff erent pathways contribute to STAT5 activation varies between normal and malignant mammary cells and between diff erent types of breast cancer cells.
Future directions for study build upon this broad hypothesis. While the role of STAT5 in lineage specification of normal mammary epithelial cells appears clear, it is important to know whether STAT5 also can specify specifi c lineages of breast cancers. Human breast cancer is not a unitary disease but rather is composed of diff erent types with some unique pathological and molecular features. One hope is that we can build upon our previous success with targeted endocrine and ErbB2targeted therapy to develop a more eff ective treatment that is more specifi cally designed for specifi c breast cancer types. If STAT5 were to specify a particular type of breast cancer, this knowledge could be used for better treatment and prevention. Related to this is the possibility that the activity level of STAT5 could be used as a marker to help select individuals who might respond better to particular therapies or indicate women who might require closer follow-up or a diff erent approach because their risk of non-response would be increased. Th e most immediate need for additional research in this area might be in regard to endocrine therapy and to the combination of trastuzumab with EPO-type drugs. In regard to endocrine therapy, we need to know why activation of STAT5 is correlated with a higher response rate. Is this a simple correlation with ERα status and higher diff erentiation levels, or does STAT5 activation play a mechanistic role in improving the response? Th e fact that an interaction between EPO and trastuzumab has been identifi ed raises the possibility that other competing interactions between signaling pathways may occur and should be sought. STAT5 is actually two proteins, STAT5A and STAT5B, embedded within a network of cellular signaling pathways. Future studies will have to try a more global analysis of this network as opposed to focusing on a limited number of readouts. Th is is true for all three major types of experiments that examine STAT5 action: mouse model studies, tissue culture cell studies, and human tissue studies. In each of these approaches, investi gators will need to examine not only STAT5 activation but also associated family members and the constella tion of interacting proteins and signaling path ways that impact STAT5. Finally, STAT5 is described as a transcription factor. Whereas some of its downstream genes are characterized, the transcriptional networks that lead to the diff erent cellular behaviors associated with STAT5 activation are not fully defi ned. Further defi nition of the genetic networks lying downstream of STAT5 remains a highly relevant research goal.

Compet ing interests
The authors declare that they have no competing interests.
Authors' information REN, SM, ESD-C, and MCC contributed equally and are listed in reverse alphabetical order.