Ligand-Independent Canonical Wnt Activity in Canine Mammary Tumor Cell Lines Associated with Aberrant LEF1 Expression

Pet dogs very frequently develop spontaneous mammary tumors and have been suggested as a good model organism for breast cancer research. In order to obtain an insight into underlying signaling mechanisms during canine mammary tumorigenesis, in this study we assessed the incidence and the mechanism of canonical Wnt activation in a panel of 12 canine mammary tumor cell lines. We show that a subset of canine mammary cell lines exhibit a moderate canonical Wnt activity that is dependent on Wnt ligands, similar to what has been described in human breast cancer cell lines. In addition, three of the tested canine mammary cell lines have a high canonical Wnt activity that is not responsive to inhibitors of Wnt ligand secretion. Tumor cell lines with highly active canonical Wnt signaling often carry mutations in key members of the Wnt signaling cascade. These cell lines, however, carry no mutations in the coding regions of intracellular Wnt pathway components (APC, β-catenin, GSK3β, CK1α and Axin1) and have a functional β-catenin destruction complex. Interestingly, however, the cell lines with high canonical Wnt activity specifically overexpress LEF1 mRNA and the knock-down of LEF1 significantly inhibits TCF-reporter activity. In addition, LEF1 is overexpressed in a subset of canine mammary carcinomas, implicating LEF1 in ligand-independent activation of canonical Wnt signaling in canine mammary tumors. We conclude that canonical Wnt activation may be a frequent event in canine mammary tumors both through Wnt ligand-dependent and novel ligand–independent mechanisms.


Introduction
The Wnt signaling is one of the key players during normal mammary gland development as well as during mammary tumorigenesis [1]. The canonical, b-catenin-mediated Wnt signaling is activated by secreted Wnt ligands through activation of transmembrane frizzled (Fzd) receptors and LDL-receptor related protein 5 or 6 (Lpr5/6) co-receptors. This triggers Dishevelled-dependent disruption of the b-catenin-destruction complex in the cytoplasm that is composed of multiple proteins including, glycogen synthase kinase 3b (GSK3b), adenomatous polyposis coli (APC), Axin1 (or Axin2) and casein kinase 1 (CK1). Consequently, stabilized b-catenin translocates to the nucleus where it can associate with the T-cell factor (TCF)/Lymphoid enhancer-binding factor (LEF)-family of transcription factors to regulate the expression of specific target genes. The Wnt signal strength can be negatively regulated by secreted factors that competitively bind Wnt ligands (e.g. secreted Fzd related protein (sFRP)) or by factors that disrupt the assembly of Fzd/Lpr5/6 coreceptor complex (e.g. dikkopf-1) [2].
Many studies have documented active Wnt signaling in mammary tissue based on presence of stabilized b-catenin protein and often in combination with aberrant expression of target genes (e.g. Axin2). Based on these criteria, over 50% of human breast tumor tissue samples assessed, showed signs of aberrant canonical Wnt activity (reviewed in [3]). In contrast, reports on the signaling activity in cultured human breast cancer cells are somewhat contradicting. A number of studies have used the presence of active (phosphorylated) or uncomplexed b-catenin in cell lysates as an indicator of active canonical Wnt signaling (Table 1 and references therein). Other studies have applied a more quantitative manner to assess Wnt activity in cultured cells using a TCFreporter assay. In this assay, the ratio of luciferase signal from reporters containing a promoter with either functional of mutated TCF response elements can be taken as a measure of canonical Wnt activity, with ratios higher than 1.0 indicating an active signaling. Although not always consistent between different studies, a subset of human breast cancer cell lines was found to have moderate TCF-reporter activity (Table 1 and references therein). In addition, a number of studies showed that the effect of canonical Wnt signaling could be attenuated by Wnt inhibitors sFRP1 or Dkk1 [4,5] or by blocking the receptor Fzd7 [6]. This indicates a ligand-dependent mechanism of canonical Wnt pathway activation in human breast cancer. In concordance, mutations in downstream signaling components (e.g. APC and bcatenin) are rarely found in human primary mammary tumors. Similarly, in cultured cells only one out of 24 screened human breast cancer cell lines (DU4475) had a truncating mutation in APC and none in b-catenin [7]. In contrast, epigenetic silencing of APC and Wnt ligand inhibitors (sFRP1 and Wif1) have often been reported in primary human mammary tumors and in human breast cancer cell lines [8][9][10]. The ligand-dependent nature of canonical Wnt activation in human breast cancer and benign BC3 yes [4] BT20 no [40] BT474 yes [41] BT483 no [42] BT549 no [40,42] DU4475 yes [40] EVSA-T no [40] HCC1187 yes [25] HCC1395 yes [25] HCC1937 no [42] Hs578T no [40] MCF7 no/yes [4]/ [41,43] no/yes [40]/ [24] MDA-MB-134 no [4] MDA-MB-134VI no [40] MDA-MB-157 yes [4] MDA-MB-175 no [4] MDA-MB-231 no/yes [41]/ [4,43] no/yes [25,40,44]/ [5] MDA-MB-361 no [4] no [42] MDA-MB-415 no [4] no [42] MDA-MB-435 no [4] MDA-MB-453 no [40] no [4,43] MDA-MB-468 no [4] no [42] JIMT-1 yes [41] OCUB-F no [40] SkBr3 no [41] SK-BR-3 no [40] SK-BR-5 no [40] SK-BR-7 no [42] SUM159 no [42] SUM185 no [42] T47D yes [41,43] no [40,44] UACC893 no [42] ZR-75-1 yes [41] no [42] doi:10.1371/journal.pone.0098698.t001 breast lesions is further supported by frequent overexpression of Wnt ligands [3]. Pet dogs have been suggested as a valuable breast cancer model for preclinical research due to the high incidence and spontaneous nature of the tumor development, shared environmental risk factors, strong genetic similarity with humans and shared aspects of mammary tumor biology [11,12]. Moreover, development of canine mammary tumors is highly dependent on steroid hormone exposure, with progesterone being the main risk factor [13,14].
The activity of canonical Wnt signaling in canine mammary tumorigenesis has not been quantitatively assessed so far. Previous studies have only addressed the expression of b-catenin protein in spontaneous canine mammary tumors in relation to E-cadherin and/or APC [15][16][17]. However, comparative gene expression profiling of human and canine mammary tumors has implicated a significant similarity in deregulation of multiple cancer-related pathways, including Wnt signaling [18]. In this study we aimed to assess the activation of canonical Wnt signaling in canine mammary tumors using a panel of canine mammary cell lines. We report that subsets of canine mammary tumor cell lines exhibit moderate, ligand-dependent-, and high, ligand-independentmechanisms of canonical Wnt activation. Moreover, we show that the ligand-independent activation of canonical Wnt signaling is coupled to the overexpression of LEF1.
Canine mammary surgery was performed as a part of a necessary medical treatment due to the presence of mammary tumor. This was done under the common rules for veterinary surgery for which owners asked for medical treatment of their pets. In contrast to medical intervention in laboratory animals no external permission was necessary other than that the surgery is done by qualified veterinary surgeons. The dog's owners were informed and gave their consent that the collected tissues can be used for research purposes. Histopathology of all tumor and paired normal tissue was evaluated by Prof. E. Hellmen ( Table 2). Pictures of cell morphology were captured using an Olympus microscope (Zoeterwoude, The Netherlands) with 10610 magnification.

TCF-reporter assay
Transfection was performed in FBS-free medium using 3 ml Lipofectamine 2000 (Invitrogen), 800 ng pTOPFLASH (TOP) or pFOPFLASH (FOP) (gift from Dr. Marc van de Wetering, Hubrecht Institute, The Netherlands) and 0.5 ng human b-actinpromoter renilla construct [22] as an internal control. Cells were seeded 48 h before transfection at a density optimal for transfection, according to the manufacturer's protocol in a 24 siRNA Canine sequence-specific LEF1 (synonym: TCF1-alpha, Genbank: XM_003434032) and b-catenin (CTNNB1, Genbank: NM_001137652) siRNA was designed on the website http:// www.dharmacon.com/designcenter/designcenterpage.aspx (Dhar-maconRNAi technologies, ThermoScientific, USA). Universal MOCKsiRNA (ON-TARGET plus non-targeting pool species H, M, R) was used as the negative control for siRNA experiments. There was no cross-silencing of non-target genes checked by blasting the siRNA designed sequences against the canine genome database. The sequence of the LEF1 siRNA duplex is as follows: sense GAAGAGGAAGAGAGAGAAAUU and antisense UUU-CUCUCUCUUCCUCUUCUU, and for b-catenin sense GAAC-GAAGGUGUAGCAACAUU and antisense UGUUGCUACAC-CUUCGUUCUU. Cell transfections were first optimized with siGLO (Dharmacon, Colorado) (data not shown). 80.000 CMT-U27 cells were transfected with 1 ml DharmaFECT Duo as transfection reagent (Dharmacon, Colorado), 50 nM siRNA and 0.5 mg DNA (TOP or FOP) in 24 wells plates (Primaria, BD, The Netherlands). After 24 h and 48 h incubation in DMEM:F12 and 10% FCS cells were harvested for RNA isolation or TCF-reporter assay.
RNA isolation, cDNA synthesis, sequencing and quantitative RT-PCR From all cell lines, total RNA was isolated from two different passages. From canine mammary tissue and from the mammary cell lines total RNA was isolated and treated with deoxyribonuclease using the RNeasy mini kit (Qiagen, Venlo, The Netherlands) according to the manufacturer's protocol. Quality of mRNA from tissue samples were assessed using a 2100 Bioanalyzer (Agilent Technologies, Amstelveen, The Netherlands) and RNA integrity number (RIN) of each sample is presented in Table 2. cDNA synthesis was performed using iScript kit (Bio-Rad Laboratories) according to manufacturer's protocol. Specific primer sets were used to amplify gene products for quantitative RT-PCR (Table 3) and sequencing (Table 4). Quantitative RT-PCR was performed using Bio Rad MyIQ detection system (Bio-RAD Laboratories) with SYBR Green Fluorophore. Relative target gene expression was normalized to that of the reference gene RPS19 using a delta Ct method [23], and relative induction of gene expression was statistically assessed using paired, two tailed Student's t test in Microsoft Office Excel. For comparison of relative gene expression between three sets of cell line, REST-MCS beta software was used (http://www.gene-quantification.de/ rest-mcs.html). For the sequence reactions we used a standard amplification with Phusion-Hot Start Taq (Finnzymes, Espoo, Finland) according to manufacturer's protocol. DNA sequence reactions were performed using BigDye v3.1 according to the manufacturer's protocol (Applied Biosystems, Foster City, CA). All amplifications were performed on an ABI 3130XL (Applied Biosystems, Foster City, CA) and analyzed in Lasergene (version 9.1 DNASTAR). The obtained sequences were compared with DNA sequences in databases using BLASTn (http://blast.ncbi. nlm.nih.gov/Blast.cgi).

Protein extraction and Western blot
For whole cell lysis cells were washed with cold HANK's balanced salt solution and scraped in cold RIPA buffer (6.5 mM Na2HPO4, 1.5 mM KH2PO4, 137 mM NaCl, and 2.7 mM KCl (pH 7.4); 1% sodium dodecyl sulfate (SDS) (vol/vol), 1% Igepal (vol/vol), 0.5% Na-deoxycholate (wt/vol), 1 mM phenylmethylsulfonylfluoride,1 mM Na-orthovanadate and 1 mg/ml aprotinin). After 20 min incubation on ice, samples were centrifuged for 15 min at 16,000 g at 4uC. Protein concentration was determined using Bio-Rad Dc Protein Assay (Bio-Rad Laboratories). Fifty microgram protein of total cell lysates was subjected to SDS-PAGE and analyzed by Western blot. For extraction of cytoplasmic and nuclear protein fractions NE-PER Reagent kit (Thermo Scientific, Breda, The Netherlands) was used according to the manufacturer's protocol. Ten mg protein was subsequently subjected to western blot analysis. Primary antibodies used in this study were directed against APC (AB-1) Mouse (FE9) ( And as secondary antibody goat anti-mouse HRP-conjugated (HAF007, R&D Systems, Abingdon, UK) was used. HRP was visualized using Advance TM_Enhancedchemiluminescence (ECL, Amersham, GE Healthcare, Eindhoven, The Netherlands) and analyzed using GelDoc2000 (BioRad).

Results and Discussion
Canonical Wnt signaling is aberrantly active in a subset of canine mammary tumor cell lines To quantitatively assess canonical Wnt activity we tested a panel of 12 canine mammary cell lines using a TCF-reporter assay (Fig. 1). Three cell lines (CMT1, CMT-U27 and CMT9) showed high TCF-reporter activity. Four cell lines (P114, CHMp, CNMp and CNMm) showed moderate reporter activity comparable to previously reported activity in human mammary cell lines [24,25]. The remaining five cell lines (CMT-U229, CMT-U335, CHMm, CIPp and CIPm) with the TOP/FOP ratio around 1, lacked canonical Wnt activity.
Ligand-dependent and -independent mechanism of canonical Wnt activation in canine mammary cell lines IWP-2 is a small molecule inhibitor that impairs Wnt ligand palmitoylation and secretion [26] and could, therefore, be used to discriminate between ligand-dependent and -independent mechanisms of canonical Wnt activation in cells. To assess the activity of IWP-2 in canine cells, P114 cell line was transiently cotransfected with canonical Wnt ligand Wnt3a ( Fig. 2A) and treated with increasing concentration of IWP-2. IWP-2 effectively inhibited Wnt-3a-dependent TOP-flash activity, but not the FOP-flash activity (Fig. 2B), confirming its specificity as canonical Wnt inhibitor in canine cells. The effect of IWP-2 treatment on the basal canonical Wnt activity was subsequently evaluated in all cell lines with active Wnt signaling. In cell lines with moderate basal canonical Wnt activity (i.e. P114, CHMp, CNMp and CNMm), IWP-2 was able to efficiently inhibit the TCF-reporter activity. Treatment with 10 mM IWP-2 resulted in TOP/FOP ratios around 1, suggesting a full ligand-dependency in the cell lines with moderately activate Wnt signaling (Fig. 2C). Moreover, we have assessed the expression of several Wnt ligands previously reported as activators of canonical signaling and/or being expressed in mammary tissue and cell lines [27,28]. Ligand-dependent activation of the pathway in these cell lines is further supported by the high expression of multiple Wnt ligands (especially Wnt5b and Wnt7a) and undetectable levels of the inhibitor sFRP1 (Fig. 3). IWP-2 treatment in CMT1, CMT-U27 and CMT9 cells, however, had no or only a minor effect on the TCF-reporter activity (Fig. 2C). These three cell lines are therefore expected to have a ligand-independent component for canonical Wnt activation. Recently, it has been reported that conditioned medium of tumor-associated macrophages or co-culture with macrophages mediate a switch from canonical to non-canonical Wnt signaling in multiple canine mammary cell lines, including P114 and CMT-U27 [29]. Inhibition of canonical Wnt signaling (demonstrated by downregulation of cytoplasmic and nuclear b-catenin protein levels) was associated with exposure of cells to increased levels of non-canonical Wnt ligands and canonical Wnt inhibitor Dkk-1. Our data supports the responsiveness of P114 cell line to Wnt ligands and inhibitors. However, the insensitivity of basal TCFreporter activity in CMT-U27 cells to treatment with IWP-2 suggests that the reported inhibition of canonical Wnt signaling [29] is not mediated by altered Wnt ligand or inhibitor expression, but most probably is caused by other mechanisms.

High canonical Wnt activity is not associated with a lack of functional E-cadherin
The majority of the canine mammary cell lines used in this study had spindle-cell-like morphology, except for CMT1, CMT-U27 and CMT9, which grow as attached cells but also as partially rounded cells (Fig. 4A). A partially rounded morphology has been associated with E-cadherin mutations in human breast cancer cell lines [30]. As the loss of E-cadherin protein can stimulate canonical Wnt activity [31], we analyzed its coding sequence and mRNA and protein levels in all 12 canine mammary cell lines. Sequencing analysis of the whole CDH1 coding region revealed no mutations in any of the cell lines (Table S1). Moreover, CMT1,  CMT-U27 and CMT9 highly expressed mRNA and mature protein of E-cadherin (Fig. 4B), suggesting a different mechanism of canonical Wnt activation in these cell lines.

High canonical Wnt activity is not associated with defects in b-catenin destruction complex
In multiple tumors, elevated canonical Wnt activity has been shown to result from mutations in components of b-catenin destruction complex [32]. Mutational analysis of coding sequences of APC, b-catenin, GSK3b, CK1a and Axin1 in canine mammary cell lines revealed, however, no mutations that were restricted to cell lines with the active canonical Wnt signaling (Table S1). As APC is also known to be epigenetically silenced or proteolytically cleaved in tumors [33,34], its mRNA and protein expression were additionally assessed. All 12 canine mammary cell lines expressed comparable levels of APC mRNA (Fig. 5A). Analysis of protein expression in CMT1, CMT-U27 and CMT9 revealed that APC was expressed as a full-length protein (Fig. 5B). We next asked whether high canonical Wnt activity in these cells is a consequence of a defect at the level of b-catenin destruction complex. Canonical Wnt signaling in cells in which b-catenin destruction complex function is fully impaired is expected to be insensitive to further stimulation of the pathway by Wnt ligands or to treatment with GSK3b inhibitors [35]. CMT1, CMT-U27 and CMT9 cells, however, responded potently to GSK3b inhibitor, LiCl (Fig. 5C) as well as to Wnt3a transfection (Fig. 5D). To determine whether the high Wnt activity is associated with increased stabilization of bcatenin protein we assessed total, cytoplasmic and nuclear levels of b-catenin by western blot. GAPDH was used as a marker of cytoplasmic proteins to assess the purity of extracts from different cell fractions. When compared to a cell line lacking canonical Wnt activity (CHMm), CMT1, CMT-U27 and CMT9 cells did not show evidence of marked cytoplasmic or nuclear b-catenin protein stabilization (Fig. 5E), implying no major defect in the b-catenin destruction complex function.
Ectopic LEF1 expression contributes to the high canonical Wnt activity in CMT1, CMT-U27 and CMT9 Upon stabilization, b-catenin translocates to the nucleus and interacts with members of TCF/LEF family of transcription factors (TCF1, LEF1, TCF2 and TCF4). Interestingly, although all TCF/LEF family members were expressed in canine mammary cell lines, LEF1 showed high mRNA expression specifically in CMT1, CMT-U27 and CMT9 (Fig. 6A). Different LEF1 isoforms have been described, resulting from alternative splicing of LEF1 transcript [36]. Sequencing of LEF1 coding region in CMT1, CMT-U27, CMT9 and CIPp showed that the first three cell lines express LEF1 transcript lacking exon 6. Lack of exon 6 in Xenopus LEF1 has been shown to lower its transcriptional potential on TOPFlash reporter in HEK293 cells [37]. Lack of exon 6 in LEF1 in CMT1, CMT-U27 and CMT9 can therefore not be attributing to the high canonical Wnt activity in these cell lines. As LEF1 is known to be a direct target gene of canonical Wnt signaling [38] we asked whether high expression of LEF1 in these three cell lines could be a cause or a consequence of high canonical Wnt activity. For this purpose b-catenin and LEF1 knock down was performed in CMT-U27 cells. Knock-down of bcatenin resulted in a potent inhibition of TCF-reporter activity (Fig. 6C) and down regulation of Axin2 target gene (Fig. 6B), but it had no effect on the expression of LEF1 (Fig. 6B). Knock-down of LEF1 did not affect b-catenin expression (Fig. 6B) but was able to significantly inhibit TCF-reporter activity (Fig. 6C). Altogether, this suggests that the ectopic expression of LEF1 in CMT1, CMT-U27 and CMT9 is not a consequence but rather a contributing factor to the high canonical Wnt activity in these cell lines. LEF1 was recently shown to affect the viability, invasion and migration of breast cancer cells [39]. The correlation between LEF1 overexpression and high canonical Wnt activity in canine mammary tumor cell lines prompted us, therefore, to assess LEF1 expression in spontaneous canine mammary tumors. To assess the tumor-specific overexpression of LEF1, each canine mammary tumor was compared to the corresponding normal tissue from the same patient. Interestingly, five out of seven malignant tumor samples showed overexpression of LEF1 (Fig. 6D). However, the sample size will need to be increased in order to test the statistical significance of these results. A challenging task remains to assess whether LEF1 overexpression in canine mammary tumors is associated with high canonical Wnt activity. Considering that canine mammary tumor cell lines with high LEF1 expression do not seem to show marked overstabilization of b-catenin protein (Fig. 5E), an alternative marker for canonical Wnt activity in tissue samples is needed. In this regard, Axin2 mRNA levels were suggested to correlate with mutations in the Wnt signaling pathway in a panel of human cancer cell lines [32]. However, in canine mammary cell lines, basal Axin2 mRNA levels do not correlate with the canonical Wnt activity (Fig. S1), implicating that Axin2 expression is also not a reliable canonical Wnt activity marker in canine mammary tumors. On a further note, LEF1 knock-down was not able to fully inhibit TCF-reporter activity in canine mammary cell lines. This may be a consequence of insufficient knock-down of LEF1 mRNA but it may also argue for involvement of additional canonical Wnt activating factors. To test whether other TCFs may be compensating for the knock-down of LEF1, the levels of all TCF-family members were assessed 24 h after knock-down of LEF1 (Fig. S2). Expression of neither of TCFs showed, however, signs of compensation. Lastly, the use of transient transfection system prevented us from investigating a relationship between LEF1 overexpression and the cellular morphology. For this purpose stable transfection of inducible LEF1 knock-down system should be employed.

Conclusions
Altogether, this study provides evidence for moderate, liganddependent canonical Wnt activation in canine mammary tumors that is comparable to human breast cancer. In addition, we report a novel ligand-independent mechanism involving LEF1 overexpression, which results in high canonical Wnt activity. Our further studies aim to explore this ligand-independent mechanism extensively and to identify the underlying gene mutations.