Introduction
Growth factors of the wingless and integration site growth factor (WNT) family are secreted, glycosylated, and palmitoylated peptides that interact with seven-transmembrane receptors of the Frizzled (FZD) family. Diverse signaling pathways are activated upon WNT/FZD binding. The ligand/receptor interaction has been shown to induce the phosphorylation of scaffolding proteins of the Dishevelled (DVL) family by casein kinase Iε and -2 and PKCα [
1‐
3]. This event was reported to be a component of all WNT-induced signaling pathways [
4,
5]. The so-called canonical WNT signaling pathway leads to stabilization of β-catenin through inactivation of a protein complex consisting of, amongst others, the tumor suppressors APC and Axin. This destruction complex normally triggers rapid β-catenin phosphorylation, inducing its ubiquitination and degradation. In the presence of canonical WNT ligands, β-catenin is stabilized, binds transcription factors of the LEF-1/T-cell factor (TCF) family, and stimulates target gene transcription [
6].
Aberrant activation of the WNT signaling pathway plays an important role in the development of many human cancer types. In colorectal cancer (CRC), mutations in APC, axin, or β-catenin itself promote β-catenin stabilization and transcription of target genes encoding cancer-associated proteins [
7]. In contrast to CRC, WNT pathway mutations rarely, if ever, are detected in breast tumors [
8]. However, various lines of evidence suggest that, in breast cancer, the WNT pathway may be de-regulated by loss of expression of negative pathway regulators. For example, expression of the extracellular inhibitor of WNT signaling, secreted Frizzled-related protein 1 (sFRP1), which competes with FZD receptors for ligand binding, is downregulated in many breast tumors and is associated with poor prognosis [
9‐
11]. Furthermore, many studies have reported that WNT ligands and FZD receptors are expressed in human breast cancer cell lines and primary tumors [
7,
12‐
14]. Finally, β-catenin is frequently found stabilized and nuclear in human breast tumors and this finding has been associated with poor prognosis [
15]. Taken together, these observations suggest that WNT signaling may frequently be de-regulated in breast cancer.
We have previously described a novel crosstalk between WNT signaling and epidermal growth factor receptor (EGFR) [
16]. The mechanism, which we have shown to involve activation of zinc-dependent membrane-associated metalloproteases [
16] that control the cleavage and availability of ERBB ligands [
17], appears to be analogous to that described for transactivation of EGFR triggered by stimulation of G protein-coupled receptors (GPCRs) [
18]. GPCR-mediated EGFR transactivation involves various heterotrimeric G protein α subunits, activation of PKC and/or Src kinase, as well as ADAMs (A Disintegrin And Metalloprotease) (reviewed recently in)[
19]) or matrix metalloprotases (MMPs) [
20].
In this study, we provide evidence for constitutive autocrine WNT signaling in human breast cancer cells. We show that sFRP1 blocks proliferation of many breast tumor cell lines through interference with pathway activation that is presumably driven by endogenous WNT ligands. Thus, our study clearly demonstrates that sFRP1 fulfills its proposed tumor suppressor function [
21]. Downstream of the WNT ligand/FZD receptor interaction, knockdown of DVL expression using short interfering RNA (siRNA) also results in a proliferative reduction and the induction of apoptosis in many human breast cancer cell lines. Our results, showing that Wnt1 transactivates EGFR in tumor cell lines, imply that, in breast cancer, constitutive WNT signaling might impact not only on the canonical pathway, but also on EGFR activity by stimulating ligand availability. Considering that constitutive ERBB receptor activation is an important mechanism promoting cancer cell proliferation, migration [
22,
23], and sensitivity to anti-cancer therapies [
24], approaches to target WNT pathway activity might be appropriate as an anti-cancer strategy.
Materials and methods
Reagents
The following antibodies were used in this study: extracellular signal-regulated kinase 1/2 (ERK1/2), p-ERK1/2, total β-catenin, poly(ADP-ribose)polymerase (PARP), EGFR, p-EGFR (tyrosine [Tyr] 845), and p-Tyr-100 (Cell Signaling Technology, Inc., Danvers, MA, USA); c-MYC (9E10), DVL2 and DVL3, EGFR 528, 1005, and R1 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA); Wnt1 and DVL1 (R&D Systems Europe, Abingdon, UK); active β-catenin (anti-ABC; Upstate, now part of Millipore Corporation, Billerica, MA, USA); and α-Tubulin (Lab Vision Corporation, Fremont, CA, USA). As secondary antibodies, α-rabbit and α-mouse (GE Healthcare, Little Chalfont, Buckinghamshire, UK, and LI-COR Biosciences, Lincoln, NE, USA) or α-goat (DAKO A/S, Glostrup, Denmark) coupled to horseradish peroxidase (HRP) were used. Mouse Wnt1 in the retroviral vector pLNCX was obtained from Andrew McMahon (Harvard University, Cambridge, MA, USA); the cDNA encoding human sFRP1 in pcDNA was provided by Jeffrey Rubin (National Cancer Institute, Bethesda, MD, USA). The retroviral vector for the expression of short hairpin RNA (shRNA) constructs pSUPERretro Neo green fluorescent protein (GFP) was provided by Francois Lehembre (DKBW, Basel, Switzerland). PKI166 and AEE788 were provided by Peter Traxler (Novartis Pharma AG, Basel, Switzerland); CGP77675 was provided by Jonathan Green and Mira Susa Spring (Novartis Pharma AG), and CGS27023A was provided by Ulf Neumann (Novartis Pharma AG). 4-Hydroxytamoxifen (4-HT) was purchased from Sigma-Aldrich (St. Louis, MO, USA).
Cell culture, transfections, and retroviral infections
The human breast cancer cell lines T47D, MCF-7, ZR-75.1, SkBr3, BT474, and MDA-MB-231 (American Type Culture Collection, Manassas, VA, USA) and JIMT-1 (DSZM, Braunschweig, Germany) were cultivated in Dulbecco's modified Eagle's medium (DMEM), 10% heat-inactivated fetal calf serum (FCS) (Amimed, Allschwil, Switzerland) supplemented with penicillin and streptomycin. HC11 and HC11/Wnt1 cells were maintained in RPMI 1640, 10% FCS, penicillin/streptomycin, epidermal growth factor (EGF) (Collaborative Research Co., Bedford, MA, USA) and insulin (Sigma-Aldrich). HC11/Wnt1 cells were kept under selection in 1 mg/mL G-418 (Life Technologies, Inc., now part of Invitrogen Corporation, Carlsbad, CA, USA). HEK 293 cells were transfected with a vector encoding myc/HIS-tagged human sFRP1 using Lipofectamine according to the manufacturer's guidelines. Cells were kept for 3 weeks in medium containing 1.5 mg/mL G-418, and clones were selected. T47D and SkBr3 cells were stably transfected with Wnt1 or empty pLNCX as control by Lipofectamine Reagent (Invitrogen Corporation) according to the manufacturer's instructions. Clones of Wnt1-expressing cells were selected with 0.5 mg/mL G-418. The expression of Wnt1 ligand was verified by Western blotting, and biological activity was assayed in a co-culture assay with HEK 293/8× SUPERTopFlash cells, using 300,000 cells each in a six-well overnight culture before the assay was performed. Knockdown of β-catenin was achieved by retroviral infection [
25] with pSUPERretro Neo GFP containing a short-hairpin targeting β-catenin [
26]. A construct targeting bacterial LacZ (sense strand gCggCTgCCggAATTTACCdTdT) was used as control. Clones and a pool of cells with low levels of β-catenin were analyzed for their response to Wnt1 condition medium (CM). Src
-/- mouse embryonic fibroblasts (MEFs), provided by Kurt Ballmer (Paul Scherrer Institut, Villigen, Switzerland), were transfected with empty vector or a c-Src-expressing vector, and clones were selected. Src re-expressing MEFs were generated by Monilola Olayioye (University of Stuttgart, Germany).
siRNA transfections
Five hundred thousand cells per well were seeded in a six-well plate the day before transfection and were transfected with either 50 nM control RNA duplex targeting bacterial LacZ (sense strand gCggCUgCCggAAUUUACCdTdT) or a mixture of two siRNA duplexes (25 nM each; Qiagen GmbH, Hilden, Germany) targeting bases 1420 to 1440 (gCUCAACAAgAUCACCUUCUdTdT) in human DVL1 (NM_004421) and bases 1754 to 1774 and 1579 to 1599 (gUCAACAAgAUCACCUUCUdTdT) in human DVL2 (NM_004422) and DVL3 (NM_004423), respectively, using HiPerfect (Qiagen GmbH) according to the manufacturer's instructions. The DVL target sequences were chosen based on the high conservation in all three human DVL homologues. The cells were cultured for 72 hours, and knockdown efficiency was monitored by Western blotting.
Confluent HC11/Wnt1 or parental HC11 cells were cultured for 3 days in RPMI 1640, 10% FCS, penicillin/streptomycin without EGF, insulin, and G-418. The supernatant was filtered through a 0.25-μm syringe filter (Sarstedt AG & Co., Nümbrecht, Germany). Biological activity of Wnt1 CM and control CM was assayed by their ability to induce β-catenin/TCF-dependent luciferase reporter activity in HEK 293/8× SUPERTopFlash cells (provided by Feng Cong, Novartis Institutes for BioMedical Research, Cambridge, MA, USA).
sFRP1 CM was obtained from HEK 293 cells transfected with myc/HIS-tagged human sFRP1 cDNA. CM was collected and sFRP1 activity was assayed by testing its ability to block the activation of β-catenin/TCF-driven transcription in a co-culture of T47D/Wnt1 cells and HEK 293/8× SUPERTopFlash cells and the reduction of DVL3 phosphorylation in T47D/Wnt1 cells. For treatment of breast cancer cell lines, confluent sFRP1-expressing HEK 293 cells were treated overnight with 10 mM sodium butyrate in 0.1% FCS to increase sFRP1 expression. The CM was concentrated, and sodium butyrate was removed by filtration with a Centricon Plus-70 filtration unit (Millipore Corporation). The resulting concentrate was diluted to the starting volume or used as a 2× concentrate and adjusted to 10% FCS accordingly. Cell proliferation was measured either by counting cell numbers manually or with a Vi-Cell XR cell viability analyzer (Beckman Coulter, Nyon, Switzerland), Cell Proliferation Kit I (MTT; Roche Diagnostics GmbH, Mannheim, Germany), or YOPRO cell viability assay (Invitrogen Corporation) according to manufacturer instructions.
Hybridoma cells secreting the EGFR monoclonal antibody C225 were cultured in DMEM, 10% FCS. Collected medium was cleared by centrifugation, filtered, and used undiluted on target cells for 2 hours prior to collection of cell lysates.
Purification of sFRP1
sFRP1 was purified by fast performance liquid chromatography from sFRP1 CM. After 1:10 dilution in 50 mM sodium phosphate loading buffer pH 7.0, the solution was loaded on a 1 mL HiTrap-HIS column (GE Healthcare) that was previously loaded with 1 mL 0.5 M NiSO4 and washed with 10 column volumes of loading buffer. Elution was performed using 50 mM sodium phosphate, 100 mM NaCl pH 7.0 elution buffer with a 3-minute step-gradient of 10 to 500 mM imidazole. Fractions were collected, and 1-μL aliquots were analyzed by Western blotting using a c-MYC antibody for detection of the MYC-tag. Biological activity was assayed as previously described for sFRP1 CM, and the identity of the purified protein was determined by mass spectrometry.
Cells were lysed in lysis buffer (1% Nonident P-40, 50 mM Tris pH 7.5, 120 mM NaCl, 5 mM EDTA [ethylenediaminetetraacetic acid], 1 mM EGTA [ethylene glycol tetraacetic acid], 2 mM sodium vanadate, 20 mM β-glycerophosphate, 10 μg/mL aprotinin, 10 μg/mL leupeptin, 0.5 mM PMSF [phenylmethylsulphonyl fluoride], 50 mM NaF, and 1 mM dithiothreitol) for 5 minutes on ice, and lysates were collected. For a Western analysis, loading buffer was added to 30 to 50 μg of protein and the samples were denatured for 10 minutes at 95°C prior to separation on 10% polyacrylamide gels and blotting by semi-dry transfer for 90 minutes on polyvinylidene fluoride membrane (Millipore Corporation). Blots were pre-blocked using 10% horse serum in TBS-T buffer for 1 hour (0.2 M NaCl, 25 mM Tris pH 7.5, 0.5 mL/L Tween-20) and incubated with primary antibodies at room temperature for 1 hour or at 4°C overnight, followed by 30 minutes of incubation with secondary antibodies: α-rabbit-HRP, α-mouse-HRP, or α-goat-HRP (1:5,000). Detection of luminescence was carried out using ECL (enhanced chemiluminescence) (GE Healthcare) or SuperSignal West Dura (Pierce, Rockford, IL, USA) according to manufacturer instructions. Immunoprecipitations (IPs) and Western analyses were performed using standard procedures [
24,
27]. EGFR IP was performed with α-EGFR 528 and R1. Quantifications of Western blots were carried out using the ImageQuant TL version 2005 software package from Amersham Biosciences (now part of GE Healthcare).
Discussion
De-regulation of WNT signaling is a well-established hallmark of certain types of human cancer, such as CRC and melanoma, in which a high percentage of mutations in the β-catenin destruction complex components
APC and
AXIN or in
β-catenin itself have been described [
42]. Although mutations of this type are rarely observed in breast cancer, we show here that many breast cancer cell lines have autocrine activity of WNT signaling and that blocking this pathway has multiple biological effects. In breast cancer, activation of the Wnt pathway is likely due to co-expression of WNT ligands and FZD receptors (T Schlange, unpublished observations) [
7,
12‐
14]. WNT ligands play different roles in cancer biology depending on the downstream pathways activated. Whereas canonical Wnt signaling is required for G
1 cell cycle progression in CRC [
43], the non-canonical ligand WNT5A negatively regulates proliferation [
44,
45] but promotes migration in various cancer types [
46,
47]. One potential mechanism contributing to pathway activity might be loss of negative modulators of WNT signaling [
48], as decreased expression of sFRP1 is well documented in human breast cancer [
10,
11,
29]. Furthermore, the loss of sFRP1 expression was recently shown to synergize with c-MYC-induced tumorigenesis [
49]. Extending the analysis of Bafico and colleagues [
21], we assayed the activation of WNT signaling by DVL phosphorylation, the most proximal read-out of FZD receptor activation, and found autocrine WNT activity in a panel of human breast cancer cells with diverse genetic alterations.
We show here that treatment of many breast tumor cell lines with sFRP1 has a consistently negative effect on their proliferation by affecting the canonical WNT pathway. In addition, ERK1/2 pathway activity is also decreased by sFRP1 treatment in the majority of the cancer cells, with SkBr3 cells being particularly sensitive. SkBr3 cells have high levels of ERBB activity. The fact that sFRP1 decreases p-ERK1/2 levels suggests that WNT-mediated ERBB transactivation has an important role in maintaining ERK1/2 signaling in these tumor cells. As SkBr3 cells have essentially no active β-catenin, sFRP1 effects on ERK1/2 activity might be the main cause for their decreased proliferation in response to sFRP1 treatment. A similar dependence on a non-canonical WNT signal was observed in β-catenin-deficient mesothelioma cells [
50], in which siRNAs against WNT1 and DVL induced apoptosis in a JNK (c-jun N-terminal kinase)-dependent manner. This finding is particularly interesting given the inhibition of proliferation and induction of apoptosis we observe in response to the knockdown of all three DVL homologues in different breast cancer cell lines. Interfering with WNT signaling at the DVL level should block all autocrine activation [
5]. sFRP1, on the other hand, most likely binds only a subset of WNT ligands [
30,
51], which might explain why sFRP1 treatment could not completely block β-catenin stabilization or WNT-induced ERK1/2 activity. In fact, compared with sFRP1 treatment, DVL knockdown elicited a stronger negative effect on ERK1/2 activity in the breast cancer cell lines. BT474 and MCF-7 cells are most resistant to both sFRP1 treatment and DVL knockdown when compared with the other cell lines analyzed. In the case of BT474, this is in line with relatively low levels of DVL phosphorylation, indicating that this cell line is mostly independent of autocrine WNT signaling. This shows that there is differential sensitivity of human breast cancer cells with different oncogenic pathways activated (for example, ERBB2 overexpression, estrogen dependence) to inhibition of autocrine WNT signaling.
Recently, blocking the FZD/DVL interaction using a small molecule targeting the PDZ domain of DVL was explored and shown to inhibit the proliferation of cancer cell lines derived from different types of cancer [
52]. Our observations imply that targeting this interaction or the use of a 'ligand trap' like sFRP1 might be a valid approach to treat breast cancer by interfering with the canonical WNT pathway as well as the EGFR/ERK1/2 pathway. Inhibition of more than just one WNT ligand or FZD receptor may overcome the problem of functionally redundant expression of several family members when specific antibodies are used [
53‐
57]. In summary, our observations on blocking autocrine WNT activity in human breast cancer cells suggest an important role for WNT-induced EGFR transactivation in the control of ERK1/2 signaling and of proliferation.
It is also noteworthy that there is differential phosphorylation of DVL isoforms in the panel of breast cancer cell lines. Perhaps DVL family members are not redundant and might be activated by different WNT/FZD complexes. Furthermore, DVL isoform levels vary substantially in different breast cancer cell lines. Therefore, it might be worth analyzing whether aspects of tumor biology like proliferation and migration are differentially regulated by these scaffolding proteins, potentially providing a paradigm for the differentiation of non-canonical versus canonical WNT signaling.
We show here that, in addition to activating the canonical Wnt/β-catenin pathway, Wnt1 transactivates EGFR and stimulates ERK1/2 activity in many human breast cancer cells. This Wnt1-mediated response is similar to EGFR transactivation induced by many GPCRs)[
19]. In fact, various lines of evidence, including the GPCR-like heptahelical structure of the FZD receptor family and genetic data from
Drosophila, suggest that these receptors have biological similarities [
58‐
60]. Although we could not block Wnt1-induced ERK1/2 activation using pertussis toxin (PTX) to block Gα
i/o proteins, this still leaves the possibility that PTX-insensitive Gα proteins mediate the effects of WNT/FZD signaling. Indeed, it was shown that Gα
q/11 group proteins contribute to the activation of canonical WNT signaling [
60,
61].
Our results also show that c-Src has an important role in Wnt1-driven EGFR transactivation. Wnt1 was able to transactive EGFR in Src-expressing MEFs, but not in Src knockout MEFs. Furthermore, an Src kinase inhibitor abolished the effects of Wnt1 on ERK1/2 activation in human breast cancer cell lines and Src kinase activation was increased in SkBr3/Wnt1 cells. Src kinase has also been implicated in GPCR-mediated EGFR transactivation)[
19]. Src kinase might act directly downstream of GPCRs and FZD receptors via its interaction with ADAMs and MMPs [
62‐
65]. Association of Src kinases with these enzymes might regulate their proteolytic activity and subcellular localization [
66], leading to an increase in ERBB ligand shedding and autocrine receptor activation. Since we observed that neither metalloprotease inhibitors nor an EGFR-blocking antibody completely blocked Wnt1-induced ERK1/2 activation, this might reflect a direct effect of Src kinase on EGFR activity due to its ability to phosphorylate the receptor at Tyr 845 [
67]. The involvement of WNT-induced Src activity on EGFR activation is corroborated by our observation that the knockdown of DVL decreased the level of Tyr 845 phosphorylation in several breast cancer cell lines.
WNT signaling has previously been linked to the activation of Src and ERK1/2 in NIH3T3 cells and in osteoblast progenitors [
68‐
70], and recently EGFR was shown to be involved in ERK1/2 activation downstream of purified Wnt3a [
71]. However, these studies rely on overexpression or treatment with recombinant proteins and did not link the transactivation to autocrine signaling processes. It was recently shown that Wnt1 is induced by progesterone receptor (PgR) signaling in T47D breast cancer cells and that it is required for EGFR transactivation by a PgR agonist in an Src- and metalloprotease-dependent manner [
72]. These results are interesting to consider in light of the data presented in this paper. It is possible that the rapid effects of steroid hormones leading to sustained proliferation or survival of breast tumor cells proceed by establishing an autocrine loop of EGFR activity that is linked, in part, to Wnt1 production. It will be important to see whether results from the T47D breast cancer model are clinically relevant in primary breast tumors, many of which overexpress Wnt1 [
13].
EGFR activity is known to play a role in endocrine therapy resistance (for example, in MCF-7 cells) [
73]. In fact, there are increased β-catenin levels and increased expression of WNT pathway target genes in these resistant cells [
74], further implicating WNT pathway activity in endocrine resistance. Our data also show the potential importance of autocrine WNT signaling in response to anti-hormonal therapies. Wnt1 treatment of the ER
+ MCF-7 and T47D cells rescued them from the anti-proliferative action of 4-HT, and this was blocked by treatment with an EGFR TKI, showing the importance of autocrine EGFR signaling in the Wnt1 rescue.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
TS designed and carried out the experiments, unless otherwise specified, and wrote the manuscript. YM analyzed the specificity of the act. β-catenin antibody. SL assisted TS in carrying out the experiments. AH purified sFRP1 protein. NEH participated in designing the experiments and in writing the manuscript. All authors read and approved the final manuscript.