WNT3a Signaling Inhibits Aromatase Expression in Breast Adipose Fibroblasts—A Possible Mechanism Supporting the Loss of Estrogen Responsiveness of Triple-Negative Breast Cancers

Estrogen-dependent breast cancers rely on a constant supply of estrogens and expression of estrogen receptors. Local biosynthesis, by aromatase in breast adipose fibroblasts (BAFs), is their most important source for estrogens. Triple-negative breast cancers (TNBC) rely on other growth-promoting signals, including those from the Wnt pathway. In this study, we explored the hypothesis that Wnt signaling alters the proliferation of BAFs, and is involved in regulation of aromatase expression in BAFs. Conditioned medium (CM) from TNBC cells and WNT3a consistently increased BAF growth, and reduced aromatase activity up to 90%, by suppression of the aromatase promoter I.3/II region. Database searches identified three putative Wnt-responsive elements (WREs) in the aromatase promoter I.3/II. In luciferase reporter gene assays, promoter I.3/II activity was inhibited by overexpression of full-length T-cell factor (TCF)-4 in 3T3-L1 preadipocytes, which served as a model for BAFs. Full-length lymphoid enhancer-binding factor (LEF)-1 increased the transcriptional activity. However, TCF-4 binding to WRE1 in the aromatase promoter, was lost after WNT3a stimulation in immunoprecipitation-based in vitro DNA-binding assays, and in chromatin immunoprecipitation (ChIP). In vitro DNA-binding assays, ChIP, and Western blotting revealed a WNT3a-dependent switch of nuclear LEF-1 isoforms towards a truncated variant, whereas β-catenin levels remained unchanged. This LEF-1 variant revealed dominant negative properties, and most likely recruited enzymes involved in heterochromatin formation. In addition, WNT3a induced the replacement of TCF-4 by the truncated LEF-1 variant, on WRE1 of the aromatase promoter I.3/II. The mechanism described here may be responsible for the loss of aromatase expression predominantly associated with TNBC. Tumors with (strong) expression of Wnt ligands actively suppress aromatase expression in BAFs. Consequently a reduced estrogen supply could favor the growth of estrogen-independent tumor cells, which consequently would make estrogen receptors dispensable. In summary, canonical Wnt signaling within (cancerous) breast tissue may be a major factor controlling local estrogen synthesis and action.


Introduction
In postmenopausal women, the production of estrogens is located mainly in extragonadal tissue, preferentially in breast adipose fibroblasts (BAFs) [1]. Estrogen synthesis from androgens depends on three consecutive oxidation steps, which are catalyzed by the cytochrome P-450 enzyme aromatase, encoded by the CYP19A1 gene [2,3]. Estrogens are the most important female sex hormones, but they can also act as important growth factors To clarify whether secretion of factors promoting the proliferation of BAFs is mainly a property of TNBC cells, the growth promoting activities of four TNBC cell lines, and four lines expressing various combinations of the respective receptors, were compared (Table 1). Whereas all TNBC cell line CM significantly stimulated BAF growth, the effects of CM from receptor-positive cell lines varied considerably ( Figure 1B Subgroup mean values were calculated from the mean values obtained for the triple-negative and receptor-positive cell lines, respectively. (C) WNT3a CM induced a concentration-dependent increase in vital cells, whereas 30% serum-free CM from L-M(TK-)-null cells, not expressing WNT3a, had no such effect. All values represent means ± SEM, and were normalized to 100% M199/fetal bovine serum (FBS) (*, p < 0.05; **, p < 0.005; versus M199/FBS or 0% CM, respectively).
To clarify whether secretion of factors promoting the proliferation of BAFs is mainly a property of TNBC cells, the growth promoting activities of four TNBC cell lines, and four lines expressing various combinations of the respective receptors, were compared (Table 1). Whereas all TNBC cell line CM significantly stimulated BAF growth, the effects of CM from receptor-positive cell lines varied considerably ( Figure 1B). These results are in accordance with the microscopic assessment of the cultures immediately before the FDA assay was started (Appendix A, Figure A1). When the mean effects of the subgroups were compared, TNBC cell lines solidly stimulated BAF growth over a wide range of CM concentrations, whereas the receptor-positive cell lines only marginally stimulated BAF growth ( Figure 1B).
We tested additional features of the cell lines for their influence on BAF proliferation: source of cell line (primary breast cancer, pleural effusion), tumor type (adenocarcinoma, invasive ductal carcinoma, ductal carcinoma), and (over)expression or mutation of cellular tumor antigen p53 (TP53). None of these classifications is associated with stronger or weaker promotion of BAF growth.  To test whether Wnt signaling may contribute to the growth promoting effect of tumor cell CM, WNT3a CM, obtained from L-M(TK-)WNT3a cells, was used [24]. WNT3a was chosen because it is a prototypical ligand stimulating the canonical Wnt signaling pathway [6], and because this producer cell line provides high titers of bioactive ligand. Indeed, WNT3a CM also dose-dependently induced significant BAF proliferation, resulting in a more than 2.5-fold stimulation with 50% WNT3a CM ( Figure 1C). Microscopy confirmed these findings (Appendix A, Figure A1).

Expression of WNT1 and WNT3A in Breast Cancer Cell Lines
To verify the expression of Wnt isoforms, RNA was isolated from all cell lines after three days of media conditioning-both in the presence and absence of serum. WNT1 mRNA expression was detectable in MDA-MB468, HCC-1143, MDA-MB231, MCF-7, and T-47D cells kept in fetal bovine serum (FBS)-containing media (Appendix A, Figure A2A). WNT3A mRNA expression was detectable in all investigated cell lines in the presence of FBS (Appendix A, Figure A2B). The expression patterns of WNT1 and WNT3A were only marginally altered under serum-free conditions (Appendix A, Figure A2C,D), indicating constitutive expression of these genes in the cell lines. Correlation analysis of the BAF proliferation stimulating activity of CM with WNT gene expression yielded no positive association of WNT1 or WNT3A expression with BAF growth (Appendix A, Figure A2E,F). This indicates that both, WNT1 and WNT3a, may contribute to the growth promotion, but are not the sole factors in the CM doing this.

Aromatase Activity and Expression in BAFs Treated with MDA-MB231 CM or WNT3a CM
Aromatase activity in BAFs was stimulated with forskolin, in order to mimic the tumor-stroma situation in the vicinity of breast tumors [1]. Under these conditions, the aromatase activity of BAFs was inhibited dose-dependently by MDA-MB231 CM, to less than 40% of controls ( Figure 2A). WNT3a CM (50%) had no major effect on basal aromatase activity in BAFs. By contrast, it revealed a strong inhibition (90%) of forskolin stimulated aromatase activity ( Figure 2B). This effect was strongly dose-dependent ( Figure 2C). transcriptional level. Furthermore, the expression levels of aromatase mRNA transcripts with 5'-ends, typical for transcription controlled by promoters I.3 or II, decreased similarly to that of the full-length aromatase gene expression level ( Figure 2E,F). This means that WNT3a massively antagonizes the breast cancer relevant mechanism of aromatase induction in BAFs, here experimentally mimicked by forskolin stimulation.  WNT3a CM inhibited full-length aromatase mRNA expression by up to 90% in BAFs ( Figure 2D), indicating that WNT3a exerts its effect on aromatase gene expression at the transcriptional level. Furthermore, the expression levels of aromatase mRNA transcripts with 5'-ends, typical for transcription controlled by promoters I.3 or II, decreased similarly to that of the full-length aromatase gene expression level ( Figure 2E,F). This means that WNT3a massively antagonizes the breast cancer relevant mechanism of aromatase induction in BAFs, here experimentally mimicked by forskolin stimulation.

Aromatase Activity Is Inhibited by Canonical Wnt Signaling and Histone Deacetylases
Canonical Wnt signaling is activated by inhibition of glycogen synthase kinase-3β (GSK-3β). Indeed, inhibition of GSK-3β by BIO ( Figure 2G) or lithium chloride ( Figure 2H) dose-dependently reduced aromatase activity to less than 50% and 40%, respectively. Toxic effects of the inhibitors were excluded by FDA tests. In summary, this indicates an involvement of the canonical Wnt signaling pathway, and the β-catenin/TCF transcription complex, in inhibition of aromatase expression in BAFs.
The activity of the β-catenin/TCF transcription complex is modulated by multiple interaction partners associated with epigenetic regulation, including histone acetyltransferases (HATs) and histone deacetylases (HDACs) [25,26]. In this context we observed that aromatase activity in BAFs increased significantly after nonselective HDAC inhibition by panobinostat under breast cancer mimicking conditions (forskolin stimulation, Figure 2I). A less pronounced effect was seen under basal conditions (without forskolin). Therefore, HDACs must be involved in promoter I.3-and II-dependent aromatase expression. Importantly, WNT3a stimulation led to an inhibition of aromatase activity in BAFs-even in the presence of an HDAC inhibitor, i.e., in a state of de-repression of transcription ( Figure 2I). Thus, both WNT3a-treatment and active HDACs, resulted in an inhibition of aromatase activity.

Identification of Putative Wnt Response Elements in the Aromatase Promoter I.3/II Sequence
Putative target DNA elements of (canonical) Wnt signaling in aromatase promoter I.3/II were identified in silico, by MatInspector (Genomatix, Munich, Germany) database searches, revealing three Wnt response elements (WRE1, WRE2, WRE3) up to 495 bp upstream of the promoter II transcriptional start site ( Figure 3). The sequence matching best was WRE1 (position −495/−480; MatInspector score: 0.981), and is located directly downstream of an AP-1 element. WRE3 (position −346/−330) overlapped with a C/EBP1 element. Both the AP-1 and C/EBP1 elements are known to be involved in activation of the aromatase promoter I.3/II region [1]. WRE2 presents as a combination of two binding sites (position −408/−387), and is located between WREs 1 and 3. . Transcription factor binding sites and Wnt response elements (WRE) in the aromatase promoter I.3/II. MatInspector (Genomatix) database searches revealed three putative WREs in the aromatase promoter I.3/II region (GenBank Nc_000015.10 chromosome 15 reference GRCh38 primary assembly). The core binding element positions are indicated with respect to the promoter II transcriptional start site (TSS). In addition, their genomic localization is given. WRE core binding elements are highlighted in black; full-length WRE sequences are in bold type and underlined. WRE2 combined two core binding sequences in forward and reverse direction, respectively. WRE search results (lower table) were evaluated by MatInspector´s specific matrix similarity score. The higher the score, the higher the sequence identity with the ideal WRE sequence (maximum score 1). The promoter elements in boxes are derived from a review of Chen et al. [1].  ) were evaluated by MatInspector's specific matrix similarity score. The higher the score, the higher the sequence identity with the ideal WRE sequence (maximum score 1). The promoter elements in boxes are derived from a review of Chen et al. [1].

Evaluation of Putative WREs in the Aromatase Promoter I.3/II Region In Vitro
As an established model for studies on the regulation of aromatase [27], and due to their unlimited availability, murine 3T3-L1 preadipocytes were used for detailed evaluation of the putative WREs. In nuclear extracts from 3T3-L1 cells, mediators of Wnt signaling, TCF-4, LEF-1 and β-catenin were detectable by Western blotting ( Figure 4A). For TCF-4, the smaller isoform (apparent MW 60 kDa) increased after WNT3a stimulation. An even more pronounced change of isoform expression during WNT3a stimulation was found for LEF-1. The larger isoform markedly decreased in intensity, whereas expression of a short variant of LEF-1 increased strikingly. tion of the putative WREs. In nuclear extracts from 3T3-L1 cells, mediators of Wnt signaling, TCF-4, LEF-1 and β-catenin were detectable by Western blotting ( Figure 4A). For TCF-4, the smaller isoform (apparent MW 60 kDa) increased after WNT3a stimulation. An even more pronounced change of isoform expression during WNT3a stimulation was found for LEF-1. The larger isoform markedly decreased in intensity, whereas expression of a short variant of LEF-1 increased strikingly.
To elucidate whether native LEF-1 and TCF-4 were able to bind to the putative WREs identified in the aromatase promoter I.3/II region in vitro, an immunoprecipitation-based oligonucleotide binding assay was established. In contrast to the Western blot experiments, epitopes of antibodies used for immunoprecipitation were located within the Nterminal regions, to avoid interference of antibody binding with DNA-binding. TCF-4 and LEF-1 immunoprecipitates bound all three WREs ( Figure 4B-D). The specific DNA-binding was inhibited by nonfluorescent WRE competitor oligonucleotides, with the same sequences. Remarkably, WNT3a treatment inhibited specific DNA-binding of LEF-1 and TCF-4 immunoprecipitates to WRE1 and WRE2, whereas this effect was not detectable with WRE3. This indicates that, at least WREs 1 and 2 are responsive to Wnt signaling. To elucidate whether native LEF-1 and TCF-4 were able to bind to the putative WREs identified in the aromatase promoter I.3/II region in vitro, an immunoprecipitation-based oligonucleotide binding assay was established. In contrast to the Western blot experiments, epitopes of antibodies used for immunoprecipitation were located within the N-terminal regions, to avoid interference of antibody binding with DNA-binding. TCF-4 and LEF-1 immunoprecipitates bound all three WREs ( Figure 4B-D). The specific DNA-binding was inhibited by nonfluorescent WRE competitor oligonucleotides, with the same sequences. Remarkably, WNT3a treatment inhibited specific DNA-binding of LEF-1 and TCF-4 immunoprecipitates to WRE1 and WRE2, whereas this effect was not detectable with WRE3. This indicates that, at least WREs 1 and 2 are responsive to Wnt signaling.

WNT3a Treatment Triggers TCF-4 Replacement by LEF-1 on WRE1 of the Aromatase Promoter I.3/II Region in BAFs
The evidence obtained so far indicated a possible role of WRE-bound transcription factors of the TCF-4/LEF-1 family in the WNT3a-induced inhibition of transcription from the aromatase promoter I.3/II region. To analyze their role in vivo, chromatin immunoprecipitation (ChIP) experiments were performed with forskolin stimulated BAFs, in the presence or absence of WNT3a CM. The antibodies used for the immunoprecipitations were those used for Western blotting, thus allowing differentiation between the large and small variants of TCF-4 and LEF-1, respectively. Because WRE1 is almost identical to the WRE-consensus sequence (see Figure 3), primer sets for polymerase chain-reaction (PCR) were constructed, to differentiate WRE1-mediated binding of proteins from binding to the other WREs ( Figure 5A).  With primer set 1, TCF-4 binding to the WRE region of aromatase promoter I.3/II significantly decreased upon WNT3a stimulation ( Figure 5B,C). In contrast, LEF-1 binding tended to increase under WNT3a stimulation. When the ratios of band intensities obtained for WNT3a-treated and -untreated BAFs were calculated for each antibody target examined, both the reduction in TCF-4 binding, and the increase in LEF-1 binding, triggered by WNT3a were statistically significant ( Figure 5D). When transcription factor binding was analyzed analogously with primer set 2 lacking WRE1, no effect of WNT3a treatment was observed ( Figure 5E,F). For β-catenin binding, no effect of WNT3a treatment was detectable, using either primer set. In summary, TCF-4 and LEF-1 bind to WREs in aromatase promoter I.3/II region in vivo. On WRE1 TCF-4 binding dominates under nonstimulated conditions, whereas LEF-1 binding dominates after WNT3a stimulation.

Functional Consequences of TCF-4 or LEF-1 Binding to WREs in the Aromatase Promoter I.3/II Region
The evidence obtained so far indicated a major role of WRE1-bound transcription factors of the TCF-4/LEF-1 family in the WNT3a-induced inhibition of transcription at the aromatase promoter I.3/II region. The functional relevance of putative WREs was analyzed further in luciferase reporter gene assays, in 3T3-L1 cells transfected with reporter constructs containing wildtype or WRE-mutated promoter sequences. Starting from the plasmid pGL3-PII-522, where luciferase expression is under the control of the aromatase promoter regions I.3 and II, constructs with individually mutated WREs were generated. These mutations were designed so as to preclude TCF/LEF-binding. Mutation in WRE1 or WRE2 increased promoter activity in WNT3a-stimulated cells, which is in agreement with a role of these WREs in transduction of the inhibitory effect of WNT3a on aromatase promoter I.3/II activity ( Figure 6A). Interestingly, in the absence of WNT3a, TCF/LEF binding to WRE2 seems to significantly contribute to full forskolin-dependent activation ( Figure 6A). This suggests that WNT3a stimulation might switch WRE1 and WRE2 from an activating to an inhibitory mode. Mutation of WRE3 had no effect on firefly luciferase activity. For an in-depth analysis of their roles, expression plasmids for full-length or N-terminally truncated variants of TCF-4 or LEF-1 were co-transfected with the aromatase promoter I.3/II reporter plasmids. TCF-4, or ∆N-TCF-4, overexpression resulted in significantly decreased firefly luciferase activities in forskolin stimulated cells, both without and with WNT3a treatment ( Figure 6B). This inhibition was also observed, when WREs in the aromatase promoter were individually mutated ( Figure 6C). In summary, the inhibitory function of TCF-4 is independent from its N-terminal β-catenin binding region, and is mediated by more than a single WRE (i.e., at least two WREs mediate inhibition by TCF-4).
The ChIP experiments suggested that aromatase promoter I.3/II inhibition might be triggered by increased LEF-1 binding to WRE1. We used LEF-1 constructs, fused to the VP16 transactivation domain from Herpes simplex virus. Previous studies have shown that these constructs activate Wnt target gene transcription, independent of β-catenin [28]. In contrast to TCF-4, full-length LEF-1-VP16 overexpression induced a significant increase in aromatase promoter activity in WNT3a-treated, but not in untreated cells ( Figure 6D). Deletion of the N-terminal β-catenin binding region (in the construct ∆N-LEF-1-VP16) eliminated this activating effect of LEF-1-VP16, and transformed it into an inhibitory factor, which acts independently from WNT3a-treatment. In contrast, overexpression of the ∆∆N-LEF-1-VP16 construct, with an in addition deleted context-dependent regulatory domain, increased luciferase activity (up to 400%) ( Figure 6D). This construct contains the DNA-binding domain of LEF-1 fused to the transactivation domain of VP16. Taken together, this indicates that the context-dependent regulatory domain (which is present in ∆N-LEF-1 but absent in ∆∆N-LEF-1) is responsible for inhibition of reporter gene activity. The LEF-1 part of the ∆N-LEF-1 construct, therefore, should functionally resemble the lower molecular weight variant upregulated in response to Wnt3a treatment (see Figure 4A).
Unlike in cases of TCF-4 overexpression, the effects of LEF-1-VP16 and ∆N-LEF-1-VP16 overexpression under WNT3a treatment depended on a single WRE. Mutation of WRE1 (almost) eliminated the stimulatory action of full-length LEF-1-VP16 on the aromatase reporter gene (Figure 6E), and it (more than) abolished the inhibitory action of ∆N-LEF-1-VP16 ( Figure 6F). Taken together, WRE1 is responsible for the antagonistic actions of LEF-1 isoforms.
To evaluate the effects of TCF-4 and LEF-1-VP16 overexpression on an independent reporter system, the function of both proteins was analyzed by co-transfection of 3T3-L1 preadipocytes with the TOPflash reporter vector, where multiple optimized WREs control luciferase expression. As expected, WNT3a stimulation significantly increased luciferase activity ( Figure 6G,H). Co-transfection of LEF-1-VP16 massively increased WNT3a-dependent and -independent luciferase activity in transfected 3T3-L1 cells ( Figure 6G), whereas TCF-4 overexpression did not further increase the luciferase activity ( Figure 6H). These results indicate that, at least in the 3T3-L1 cell model, a truncated isoform of LEF-1 is the critical factor for Wnt signaling.

Discussion
In triple-negative breast cancers (TNBC), active Wnt signaling [15][16][17] is associated with poor prognosis [29,30]. Furthermore, Wnt signaling in neighboring adipose tissue may lead to cellular de-differentiation and stabilization of a developmental state of breast adipose fibroblasts (BAFs) [31,32]. Therefore, it is assumed that Wnt signals contribute to the desmoplastic reaction in breast cancers. In this respect, we observed that WNT3aconditioned media induced an increased growth rate of human BAFs. Similar effects were obtained with conditioned media from all TNBC cell lines. In contrast, conditioned media from receptor-positive cell lines induced heterogenous behavior. Whereas the ERpositive MCF-7 cell line and the HER2-positive SK-BR-3 cell line had no significant effects on cell growth, the ER-positive T-47D and BT-474 cell lines stimulated the growth of BAFs. Therefore, we conclude that canonical Wnt signaling induces BAF accumulation, not solely by forced de-differentiation of adipocytes [31,32], but apparently, in addition, directly promotes BAF proliferation, which would intensify the desmoplastic reaction in the microenvironment of TNBC.
Furthermore, clinical trials revealed a reduced relapse-free period in cases of stromal cell accumulation in TNBC [33], whereas stromal accumulation in ER-positive breast cancers predicted better survival [34]. Hence, the size of the stromal compartment has predictive value regarding the long-term outcome in both of these breast cancer types.
The developmental mechanisms underlying the etiology of diverse breast cancer entities have been increasingly elucidated in recent years, and it has become clear that Wnt signaling is massively involved in normal mammary gland development, as well as in oncogenic dysregulation, as reviewed in, e.g., [35]. For ER-positive tumors, effective treatments are well established. On the other hand, their recognized limitations (e.g., development of endocrine resistance) lead to further optimization of therapies [36]. However, the mechanism(s) responsible for the loss of ERα (and estrogen-dependent growth) in TNBC is (are) currently not clear. These tumors rely on other signaling pathways for growth stimulation, e.g., combined Wnt and Met signaling [37]. In addition, there is a massive discrepancy between ER-positive tumors and triple-negative tumors concerning local estrogen metabolism. In ER-positive tumors the intra-tumoral estrogen concentration can be 10-fold higher than the blood concentration of estrogens [38], and in most cases there is a gradient of aromatase expression towards the tumor in the affected breast (reviewed in [1]). In contrast, in triple-negative cancers, aromatase expression is found only in a minority of samples (and surprisingly is associated with strong androgen receptor expression) [19].
With this background we reasoned that factors driving the growth of TNBC might also be involved in the suppression of aromatase expression in these tumors. Therefore, based on its growth-promoting activities discussed above, and its well-established role in breast (tumor) development [15][16][17], we tested WNT3a for its effect on aromatase induction. Indeed, WNT3a-conditioned medium led to a strong inhibition of aromatase activity in human BAFs. This inhibitory effect correlates with a reduction in aromatase mRNA levels of a similar magnitude, which results from a proportionate decrease in the transcription regulated by the cAMP-dependent aromatase promoters I.3 and II. cAMP-mediated transcriptional activation of aromatase is typical in the vicinity of ER-positive breast cancers [39,40]. Moreover, conditioned medium from triple-negative MDA-MB231 (WNT3a-secreting) cells also inhibited forskolin-induced aromatase activity, and promoter I.3/II mediated aromatase gene expression, however in a less potent manner. This could be due to lower WNT3a levels compared to the conditioned medium from overexpressing L-M(TK-) cells, which were selected for their high WNT3a secretion. Besides that, MDA-MB231 cells may express other canonical Wnt ligands, which are more or less strongly expressed in other breast cancer cell lines [21]. In addition, the known secretion of glucocorticoid-dependent aromatase stimulating factors by MDA-MB231 cells [41], acting via promoter I.4 in target cells (BAFs), may partially antagonize WNT3a-mediated inhibition.
Our results suggest that breast cancer-associated aromatase activity, and estrogen production, not only depend on activating factors from different sources [1], but also on the absence of inhibitory signaling molecules, such as WNT3a. Such a bifunctional model of regulation of breast cancer-associated aromatase expression has not yet been clearly described in the literature. However, it should be noted that a limited number of factors inhibiting aromatase induction in BAFs under certain conditions have been reported. Progesterone can act as a physiological antagonist for glucocorticoid-mediated aromatase induction, via promoter I.4 [42]. In addition, pharmacological doses of RU486 [43] or thiazolidinedione drugs [44] have been shown to repress promoter I.4-and I.3/II-mediated aromatase transcription. Furthermore, some cytokines partially (at best 50%) inhibit induction at of these promoters [1]. However, up to now, no physiological factor has been reported that equals the potency of WNT3a in aromatase inhibition observed in this study.
As a consequence of the results discussed above, it can be concluded that the absence of WNT3a-induced signaling (or effective antagonism, for example by non-canonical Wnts [21]) towards BAFs appears to be of crucial importance for aromatase expression in ER-positive breast cancers. By analogy, this should also hold true for other activators of the canonical Wnt pathway. For an estrogen-dependent tumor, in consequence this implies that the secretion of factors leading to activation of promoter I.3/II-mediated aromatase expression in BAFs is not sufficient to secure a constant supply of estrogens for the tumor cells. Thus, this suggests that ER-positive tumors promote a desmoplastic reaction via factors that concomitantly induce aromatase [1,40], whereas triple-negative tumors drive the desmoplastic reaction predominantly via factors that inhibit aromatase induction. Such a mechanism of differential growth factor secretion also may support a facilitated loss of ERα in initially estrogen-dependent (ER-positive) breast cancers, thereby promoting them to develop a typical TNBC signature. So, if Wnt signaling is activated in BAFs, in a tumor micro-environment rich in non-estrogenic growth factors, the resulting estrogen starvation would favor the growth of cells relying on other growth factors, which would reciprocally make the ER dispensable.
Signals that induce BAFs to secrete WNT proteins, in addition to tumor cells, could lead to some basal autocrine Wnt signaling [45]. Nuclear accumulation of β-catenin is dependent on an active canonical Wnt signaling pathway [6], and was observed in cells irrespective of treatment with WNT3a-conditioned media. This suggests that aromatase expression in the vicinity of breast tumors is controlled by a rather labile signaling environment, where Wnt signaling above a critical threshold will result in a switch-off of estrogen responsiveness/aromatase expression.
Experiments with GSK-3β inhibitors indicated that activation of canonical Wnt signaling is involved in the suppression of aromatase induction in BAFs. It could lead to silencing of aromatase transcription mediated by the promoter region I.3/II via any (combination) of three in silico identified putative Wnt responsive elements (WREs), in this promoter region. Surprisingly, WNT3a treatment, and associated signaling, did not result, as expected, in enhanced association of β-catenin to the WREs, when analyzed by ChIP. In contrast, both in reporter gene assays, and in ChIP experiments, the decisive step for the WNT3a effect on aromatase induction was a switch in WRE1 occupancy from TCF-4 to LEF-1.
Western blots with nuclear extracts from controls and WNT3a-treated cells revealed a change of the expression patterns of TCF-4 and LEF-1, specifically of a WNT3a-induced increase in the levels of an alternative, lower molecular weight LEF-1 variant. The increased expression of the small LEF-1 variant was accompanied by a similar reduction in the amount of the larger variant.
DNA binding assays, with immunoprecipitated nuclear transcription factors, proved that each of the putative WREs can be bound by TCF-4 or LEF-1 in vitro. Unexpectedly, in immunoprecipitates from WNT3a-treated cells, binding to WRE1 and WRE2 was apparently lost. This effect could be traced back to the antibodies used for these immunoprecipitations, which were directed against the N-termini of the proteins, and therefore are not able to bind N-terminally truncated variants. The endogenous full-length TCF-4 or LEF-1 proteins from WNT3a-treated cells seem to lack sufficient ability to bind to WRE1 and WRE2, and preferentially bind to WRE3. In light of the ChIP results, this strongly suggests that Wnt signaling induces preferential binding of an N-terminally truncated LEF-1 variant to WRE1.
Taken together, the findings discussed so far do not perfectly fit to a direct role of canonical Wnt components in the suppression of aromatase induction [6][7][8][9]. Therefore, we systematically analyzed the role of WREs, and various variants of TCF-4 and LEF-1, in reporter gene assays. The emphasis was on N-terminally truncated variants, because these are known for potential antagonistic activities, in comparison to the full-length proteins [46]. Western blot results indicated a WNT3a-induced switch from the full-length LEF-1 isoform to a shorter isoform, which must be truncated N-terminally (because the antibodies used for Western blotting bind more C-terminal regions of their targets than the antibodies discussed in the preceding paragraph). Therefore, N-terminally deleted variants of TCF-4 and LEF-1 were tested for their effects on aromatase promoter I.3/II. Overexpression of full-length LEF-1 resulted in promoter activation, whereas overexpression of ∆N-LEF-1 suppressed luciferase reporter gene activity, both via WRE1. This was the only combination of full-length/truncated factors with a WRE that revealed a switch of the mode of action.
How does alteration of the LEF-1 isoforms produce that switch? Here, comparison of ∆N-LEF-1 with a further truncated variant, ∆∆N-LEF-1, which had lost the β-catenin binding domain, together with the context-dependent regulatory domain, is instructive. The latter domain is crucial for transducin-like enhancer of split (TLE) repressor binding [25,[47][48][49]. The inhibitory effects on gene expression of LEF-1 are lost if the association of TLE together with histone deacetylases (HDACs) [50] is lost. The VP16-fusion proteins were used in order to make this effect visible. Consistent with this, the WNT3a-induced suppression of aromatase activity was partially abolished by HDAC inhibition. Therefore, inhibitory HDAC activity, which is very often associated with TLE, cooperates with inhibitory Wnt signaling on the aromatase promoter I.3/II region in BAFs.
In contrast to LEF-1, both the full-length and an N-terminally truncated variant of TCF-4 suppress aromatase promoter I.3/II-dependent reporter gene activity. TCF-4 lost WRE binding ability upon WNT3a treatment of 3T3-L1 cells, or BAFs, in the immunoprecipitation in vitro binding assay (WRE1 and WRE2), and in the ChIP experiments (at least WRE1). Furthermore, using the TOPflash Wnt reporter system, overexpression of TCF-4 does not increase luciferase activity, whereas LEF-1-VP16 does. Taken together, TCF-4 function must be modified in a WNT3a-dependent way, in both 3T3-L1 cells and BAFs.
Finally, the still open question is, "How is this switch from TCF-4 to LEF-1 mediated?" We assume that WNT3a-induced signaling will affect not only aromatase expression in BAFs, but will also induce further changes. In this respect, Wnt signaling was shown to regulate differential expression of LEF-1 and a dominant-negative N-terminally shortened (dnLEF-1) variant. Activation of the Wnt pathway was shown to trigger the switching from promoter 1 utilization (full-length) in the LEF1 gene, to promoter 2 activation (dnLEF-1) [46]. Although we could not directly verify the identity of the N-terminally shortened LEF-1, and the ∆N-LEF-1 or dnLEF-1 (both lacking the β-catenin binding domain), or the way in which the shortened variant is generated. Our data fit into a unifying model for the mechanism underlying WNT3a-triggered suppression of aromatase expression in BAFs (Figure 7).

Materials and Methods
All chemicals used were of analytical or cell culture grades. All oligonucleotides were from Metabion (Planegg/Steinkirchen, Germany).

Cells and Cell Culture
The 3T3-L1 cells and breast cancer cell lines were obtained from the ATCC (Manassas, VA, USA). The molecular classifications and the culture media for the breast cancer cell lines are summarized in Table 1. Several of these cell lines are known to express various Wnt ligands [21], MDA-MB231 cells are known to secrete WNT3a [18]. The 3T3-L1 cells were cultured in Dulbecco's Modified Eagle's Medium (DMEM) (Sigma, Taufkirchen, Germany) containing 10% (v/v) fetal bovine serum (FBS) (Sera Plus, PAN-Biotech, Aidenbach, Germany) and 40 µg/mL gentamicin. Furthermore, L-M(TK-) cells (parental and WNT3a expressing) were cultured in DMEM containing 10% (v/v) FBS, 40 µg/mL gentamicin, and 100 µg/mL G418, before being used for production of conditioned media. The production of high titers of WNT3a protein by these cells was verified previously [55].
Human BAFs were isolated from adipose tissue of healthy patients undergoing cosmetic breast reduction surgery. The study was conducted in accordance with the Declaration of Helsinki, and patients gave informed consent according to a protocol approved by the ethics committee of the Jena University Hospital (Ref.-Nr. 4285-12/14). BAFs were isolated and cultured in medium 199, containing 10% (v/v) FBS and 40 µg/mL gentamicin, as described previously [56]. Confluent primary human BAFs, resembling almost exclusively preadipocytes, were subcultured only once.
All cultured cells were maintained at 37 • C in a humidified atmosphere; 5% CO 2 and 95% air content were used for all cells and media, except for 3T3-L1 preadipocytes, where the atmosphere contained 7.5% CO 2 and 92.5% air, during culture in serum-containing growth medium.

Viability Assay with Fluorescein Diacetate
The fluorescein diacetate (FDA) test was carried out as described previously [57]. Two days after passaging of BAFs into 24-well plates (average 5000 cells/cm 2 ), the cells were stimulated by up to 50% conditioned media. The conditioned media were mixed with medium 199, containing 10% (v/v) FBS and 40 µg/mL gentamicin, to ensure a basal growth rate. Stimulations were repeated after 2, 4, and 6 days. On day 8, the FDA test was carried out. After 90 min FDA incubation under cell culture conditions, fluorescence was measured (excitation 480 nm; emission 525 nm, cut off 495 nm) on a SpectraMax M2 plate reader (Molecular Devices, Sunnyvale, CA, USA). All conditions were tested in triplicate or quadruplicate per experiment.

Aromatase Activity Testing
The in vivo evaluation of aromatase activity in BAFs was performed using the tritium water release assay, in 24-well plates. The method was carried out as described previously [56,59]. After 18 h in 500 µL serum-free stimulation medium, 1 µCi/well (80 nM) [1β-3 H]androstenedione (PerkinElmer, Rodgau, Germany) was added as a substrate for the aromatase enzyme 6 h before the incubation was terminated. Aromatase activity was given as pmol androstenedione used per 6 h and mg protein. All conditions were tested in triplicate per experiment.

Preparation of Soluble Nuclear Extracts
The preparation of soluble nuclear extracts was based on a method published by Wilde et al. [27]. The protein concentration was quantified by the Bradford method [60].

Immunoprecipitation-Based DNA-Binding Assay
Soluble nuclear extract proteins (50 µg) were pre-incubated with 20 µL pre-cleared protein G-Sepharose 4 Fast Flow (GE Healthcare, Freiburg, Germany), in soluble nuclear extract buffer, at 4 • C, in a rotator, to eliminate proteins binding nonspecifically to protein G. After centrifugation of the pre-incubated samples (20 s, 12,000× g, 4 • C), the supernatants were transferred into new tubes and incubated with antibodies for 24 h, at 4 • C, under constant rotation. The antibodies were, mouse anti-TCF-4 (L40C3) (directed against a peptide around Glu81 of human TCF-4; Cell Signaling Technology, Frankfurt, Germany) or mouse anti-LEF-1 (2D12) (directed against amino acids 1-85 of human LEF-1; nanoTools, Teningen, Germany). After antibody incubation, 20 µL of pre-cleared protein G-Sepharose 4 Fast Flow was added and incubated for 4 h, at 4 • C, under constant rotation. For final iso-lation of TCF-4 or LEF-1 immunoprecipitates, respectively, the samples were washed three times in a three-fold volume of DNA-binding buffer C (20 mM HEPES pH 7.9, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM PMSF), with centrifugation after each step (20 s, 12,000× g, 4 • C). The final immunoprecipitates were resuspended in 8 µL buffer C. The DNA-binding reaction was a modification of the sample preparation protocol for electrophoretic mobility shift assays described by Taylor et al. [61]. Immunoprecipitates in buffer C (8 µL) were mixed with 1.7 µL 10-fold binding buffer (500 mM Tris/HCl pH 7.5, 1 M NaCl, 1 mM EDTA, 50 mM β-mercaptoethanol). For normal binding reactions, the premix was added to 2 µL Cy5-labeled double-stranded oligonucleotides (25 pmol/µL; WRE1: 5 -GAGTCACTTTGAATTCAAT-3 , WRE2: 5 -ACTTACTATTTTGATCAAAAAAGTCATT-3 , WRE3: 5 -CTTTTTGTTTTGAAATTGATTTGGCTTCA-3 , only sense sequences given) and 5.3 µL water. For binding reactions in the presence of competitor, 8 µL immunoprecipitate, 2 µL fluorescence-labeled double-stranded oligonucleotides, and 5.3 µL unlabeled competitor double-stranded oligonucleotides (250 pmol/µL; same sequences like fluorescencelabeled oligonucleotides) were mixed. After incubation for 30 min with rotation, at room temperature in the dark, the samples were washed three times in a three-fold sample volume of wash buffer (50% buffer C, 10% 10-fold binding buffer, 40% water), followed by a 20 s centrifugation at 12,000× g at room temperature. Finally, the oligonucleotide bound immunoprecipitates were resuspended in 17 µL wash buffer, and transferred to a well of a 96-well plate for fluorescence measurement (excitation 600 nm; emission 670 nm, cut-off 630 nm). As a control, unspecific binding of fluorescent oligonucleotides to protein G-Sepharose 4 Fast Flow beads treated as described above, in the absence of antibodies, was analyzed, resulting in negligible fluorescence signals. All conditions were tested in triplicate per experiment.

Statistical Analyses
Statistical analyses of all experiments, and creation of diagrams, were performed with the SigmaPlot 13 or 14 software (Systat, Erkrath, Germany). Data are presented as means ± SEM or using box plots, where appropriate. Initially, normal distribution of values was tested by the Shapiro-Wilk method. Normally distributed values were compared to another group by the two-tailed Student's t-test (if not explicitly indicated otherwise in text/legends). In the case of non-normally distributed values, two groups were compared by the Mann-Whitney rank sum test, if indicated. For all tests, the significance criterion p < 0.05 was used. All numbers of replications (n) in figure legends refer to biological replicates.

Conclusions
Canonical Wnt signaling toward BAFs can induce change in a breast tumor environment in two ways: it can initiate/enhance the desmoplastic reaction, and thus increase the amount of altered stroma; and it can suppress local estrogen production in the BAFs. Therefore, breast tumors, which secrete Wnt ligands may cut themselves off from a sufficient estrogen supply for growth promotion. Lacking estrogen signaling consequently will make ERα dispensable, and thus supports development into a hormone receptor-negative tumor.    Figure 1B).