Background
Triple negative breast cancer (TNBC) is defined by the lack of both estrogen receptor (ER) and progesterone receptor (PR) expression as well as overexpression or amplification of the
human epidermal growth factor receptor HER2 [
1‐
3]. Patients suffering from TNBC are not eligible for endocrine or HER2 targeted therapies, thus rendering chemotherapy the only therapeutic option, which may be accompanied by antiangiogenic approaches such as bevacizumab in the palliative setting [
2,
4,
5]. Up to 15% of all breast cancer patients are diagnosed with TNBC [
3]. Due to high recurrence rates and an increased risk of visceral and cerebral metastases these patients have a poorer prognosis in comparison to other breast cancer subtypes [
6‐
8]. However, patients suffering from TNBC do have an increased probability of positive response to anthracycline/taxane- containing neoadjuvant chemotherapy. Thus, by achieving a pathologic complete response after neodajuvant chemotherapy the prognosis is as good as in other breast cancer subtypes [
9]. Consequently, as chemotherapy sensitivity is one of the most important prognostic factors, it is inevitable to identify biomarkers and potential mediators of chemotherapy sensitivity in patients with TNBC.
The scientific goal of this study was to identify biomarkers, which may serve as mediators of chemotherapy sensitivity in TNBC. By using global gene expression profiles of patients receiving neoadjuvant chemotherapy we could identify secreted frizzled receptor protein 1 (SFRP1) as being correlated with the triple negative breast cancer subtype. Furthermore, we found a positive correlation of SFRP1 expression and response to neoadjuvant chemotherapy.
SFRP1 has been described to antagonize canonical Wnt signaling by binding to Wnt proteins or Wnt receptors, thereby inhibiting their downstream signaling activity [
10]. In a plethora of solid tumors, including colorectal cancer, ovarian cancer, prostate cancer and lung cancer, it has been shown that SFRP1 is inactivated by promoter hypermethylation [
11‐
15]. In breast cancer, hypermethylation of the SFRP1 promoter has been correlated to poor prognosis, presumably due to elevated levels of Wnt signaling [
16,
17].
By analyzing the molecular role of SFRP1 in triple negative breast cancer cells via siRNA mediated knockdown we found changes in carcinogenic properties of breast cancer cells, e.g. increased migration and invasion potential as well as reduced apoptotic events. Furthermore, we observed an increased resistance to standard cytostatic agents. Surprisingly, although SFRP1 is known to act via canonical Wnt signaling, our data suggests that its influence on triple negative breast cancer cells is apparently not mediated via this pathway.
In summary, we could show that tumor suppressor and Wnt signaling antagonist SFRP1 is correlated with the most aggressive subtype of breast cancer, i.e. triple negative breast cancer; but also with positive response to neoadjuvant chemotherapy. This makes SFRP1 a potential biomarker for future stratification of triple negative breast cancer patients. Additionally, SFRP1 seems to be involved in regulatory processes necessary for tumorigenic cancer cells, e.g. regulation of apoptosis as well as migration and adhesion processes. Surprisingly, however, these mechanisms are not mediated by canonical Wnt signaling.
Discussion
Triple negative breast cancer is the most aggressive breast cancer subtype associated with poor prognosis as well as high recurrence rates. Since patients suffering from TNBC do have an unfavorable prognosis mostly due to limited therapeutic options, this cancer subtype has gained much attention regarding the development of novel targeted therapies. However, a high number of TNBC patients do positively respond to neoadjuvant chemotherapy. As a result, achieving a pathologic complete response (pCR) is correlated with good prognosis similar to other breast cancer subtypes [
9]. Thus, the identification of patients responding to neoadjuvant chemotherapy would greatly improve the therapeutic options for TNBC. Other TNBC patients, however, would need to be treated differently, e.g. by anti-angiogenic treatment.
By using a published dataset analyzing differential gene expression profiles as well as response to neoadjuvant chemotherapeutic treatment of breast cancer patients we could show a correlation of SFRP1 expression and the triple negative breast cancer subtype.
We could demonstrate that breast cancer cell lines showing increased SFRP1 expression are associated with the triple negative phenotype, similar to published results showing higher expression in basal like cancer cell lines compared to luminal cell lines [
14]. Interestingly, we mainly found increased expression of SFRP1 in triple negative breast cancer cell lines, which molecularly belong to basal A subtype described by Neve et al. (HCC1937, MDA-MB-468 and BT20) [
27]. Cell lines of basal A subtype display more epithelial characteristics, whereas basal B cell lines are shown to be more invasive presumably due to spindle-like morphology displaying mesenchymal as well as stem/progenitor-like characteristics.
Furthermore, the basal A subtype has been correlated with increased response to neoadjuvant chemotherapy when compared to the basal B subtype [
28,
29]. In line with these observations we could show that expression of SFRP1 is also correlated with the achievement of a pathologic complete response after neoadjuvant chemotherapy. Thus, SFRP1 might become a useful biomarker to stratify triple negative breast cancer patients, which might benefit from neoadjuvant treatment. However, for clinical applications using SFRP1 expression as a prognostic biomarker, a proper platform is needed e.g. immunohistochemistry (IHC) or fluorescent in situ hybridization (FISH) followed by conversion into dichotomous status [
30].
SFRP1 belongs to the family of 5 secreted frizzled receptor proteins, which show homology to the frizzled proteins, surface receptors for Wnt proteins [
31]. SFRP1 has been described to antagonize canonical Wnt signaling by binding to either Wnt ligand proteins or frizzled receptors, thereby inhibiting the downstream signaling cascade [
17]. SFRP1 has been linked to a number of solid tumors, e.g. colon cancer, ovarian cancer, prostate cancer or breast cancer [
11‐
17]. It has been shown that the SFRP1 promoter is hypermethylated in these entities, thereby inactivating SFRP1 expression and its protein translation.
When analyzing the influence of loss of SFRP1 in triple negative breast cancer cells we found an increase of tumor-associated characteristics, e.g. increase in migration and invasion capacity, reduced apoptotic events as well as resistance to cytotoxic chemotherapy. Increased Wnt signaling is known to regulate tumor progression mechanisms as well as resistance to chemotherapy or radiation [
32‐
34]. Thus, we initially hypothesized that knockdown of SFRP1 might result in increased Wnt signaling activity, thereby promoting tumor-associated properties of cells. This would be in line with previously published data showing reduced xenograft growth after SFRP1 overexpression in breast cancer cells presumably due to blockade of canonical Wnt signaling activity [
35].
Surprisingly, however, we were neither able to detect Wnt activation by TOPFlash luciferase assays, changes in cellular localization of β-Catenin, nor detect any significant upregulation of known Wnt target genes. Therefore, we propose a different mechanism at which SFRP1 influences tumorigenic properties like invasion potential or resistance to chemotherapeutic agents. However, a Wnt dependent effect may also occur in triple negative cancers as in vivo Wnt signals may be supplied from the tumor stroma. Thus, Wnt dependent effects may also be contributing to the in vivo responsiveness of triple negative breast cancer patients to chemotherapy.
In addition to its known role in Wnt signaling, recent reports also demonstrate novel roles for SFRP1 signaling. One report showed binding of SFRP1 to thrombospondin-1, thereby inhibiting cancer cell adhesion and migration. This binding was conducted via the netrin related motif of SFRP1 [
36]. These data are in accordance with our observation of increased invasiveness of SFRP1-depleted cells. Another study demonstrated an increased sensitivity of cells to TGF-β signaling upon SFRP1 reduction [
37]. The TGF-β pathway is involved in epithelial to mesenchymal transition (EMT) as well as cellular migration in later stage mammary tumors, despite its known function as a tumor suppressor in early stage malignancies [
38‐
40]. However, as expression of a majority of known EMT related genes are not substantially altered upon SFRP1 knockdown (Figure
4E), EMT may be of minor relevance in our experimental system. Recently, another study demonstrated a link between reduction of SFRP1 and reduction of apoptosis in vitro [
41]. By using global gene expression profiles after SFRP1 knockdown, gene ontology analyses revealed upregulation of genes involved in migration processes, whereas genes involved in the positive regulation of apoptosis were downregulated (Tables
4,
5). Thus, chemotherapy might be reinforced by inhibition of apoptosis after SFRP1 knockdown. This view is supported by our observation of a slight decrease of apoptotic events upon SFRP1 depletion (Figure
3C, D). Apparently, pathways different from Wnt signaling presumably regulate processes that lead to increase of tumorigenic properties of cancer cells.
Conclusions
Our study sheds light on the complex regulatory network of mammary tumorigenesis and tumor progression, proposing a model in which SFRP1 regulates either invasive processes via canonical Wnt signaling but also via different pathways, e.g. TGF-β signaling as well as apoptotic processes via so far unknown mechanisms (Figure
4F).
Furthermore, we conclude that SFRP1 might be clinically used to stratify patients, which suffer from triple negative breast cancer for responding to neoadjuvant chemotherapy. As the reduction of SFRP1 is in line with increased aggressiveness of cancer cells, its overexpression might be an approach to explore novel therapeutic projections [
35]. Thus, an increased level of SFRP1 might sensitize triple negative breast cancer patients towards chemotherapy, thereby improving prognosis of this aggressive breast cancer subtype. Nevertheless, future analyses have to be undertaken to explore the role of SFRP1 in regulating mammary tumor progression, particularly progression of triple negative breast cancer.
Methods
Cell culture, chemicals
The human mammary epithelial cell line MCF10a and the cancerous cell lines HCC 1937, MDA-MB 468, BT-20, MDA-MB 453, HCC 1806, MDA-MB 231, SKBR-3 and MCF-7 were supplied from ATCC and cultured under recommended conditions. Medium, trypsin-EDTA, PBS, fetal calf serum and horse serum were received from PAA Laboratories.
In order to chemically stimulate Wnt signaling, cells were starved in medium without serum for 24 h followed by incubation with 10 mM LiCl (SIGMA Aldrich), which inhibits GSK3β, thereby activating Wnt signaling [
19,
21].
Microarray gene expression analyses
A published microarray dataset was used for differential gene expression analysis [
18]. For gene expression analysis in patients, triple negative breast cancer was defined using clinical measurements for ER, PR and HER2 as described previously and compared to the remaining cases merged as non-TNBC. Response to neoadjuvant chemotherapy was defined as absence of invasive breast cancer cells at the time of definitive surgery [
42] and dichotomized as either pCR (n = 34) or residual invasive disease (RD; n = 99). Among cases with TNBC, 13 cases had pCR and 14 cases had RD. Gene expression data was processed and normalized using the robust multiarray average normalization algorithm as implemented in R-package affy version 1.32.0 [
43].
Differential gene expression between patient subgroups was assessed using Welch's t statistic. The resulting p values were adjusted for control of the false discovery rate (FDR) according to Benjamini and Hochberg's method [
44]. Analyses were performed in R using the Bioconductor multitest package version 2.10.0 [
45].
For microarray experiments after SFRP1 knockdown in MDA-MB-468 cells, mRNA was converted into cDNA by using Superscript Double-Stranded cDNA Synthesis Kit (Invitrogen) according to manufactures protocol. Fluorescence labeling was performed using NimbleGen One-Color DNA Labeling Kit followed by hybridization onto arrays (NimbleGen human gene expression 12 × 135k arrays) according to protocol. By using DEVA software raw data was extracted. Further normalization was performed using GeneSpring Software. Normalized values were imported into gene ontology database DAVID (
http://david.abcc.ncifcrf.gov; [
26].
Western Blot analysis
Cells were incubated with RIPA buffer (10 mM NaF, 1 mM Na3VO4, 10 mM β-Glycerophosphate, 7.6 mM Tris pH 7.4, 52 mM NaCl, 0.4% Triton X-100, 0.8 mM EDTA, proteinase inhibitor (SIGMA Aldrich). Protein quantification was performed via BCA assay (Pierce) according to the manufacturer’s protocol. SDS page electrophoresis and blotting were performed using standard protocols. Detection was performed using SFRP1 antibody (SIGMA Aldrich, SAB2900383) and β-Actin antibody (BioLegend, clone # 2 F1-1) and SuperSignal West Pico Chemiluminescent Substrate (Pierce). Bands were visualized with AGFA developer and fixer (AGFA).
Quantitative real-time PCR
RNA isolation was performed using NucleoSpin RNA Kits (Macherey-Nagel) with on-column DNase digestion. Reverse transcription for real-time quantitative polymerase chain reaction (RT-qPCR) was performed using MMLV reverse transcriptase (USB (Affymetrix)) and Oligo-dT
15 priming at 42°C for 1 hour and at 60°C for 10 minutes. A cDNA equivalent of 50 ng total RNA was used as template in a total reaction volume of 20 μl with Power SYBR Green PCR mix (Invitrogen) on an Step One Plus cycler (ABI). Primers were added at 0.375 μM each. Calculations were based on the ΔΔCt method using two housekeeping genes for normalization. Real time primer sequences can be found in supplemental Table
1 (Additional file
1: Table S1).
siRNA mediated knockdown assays were implemented using SFRP1 siRNA (part no 4392422) and negative control siRNA (part no 4390844) (Applied Biosystems) in combination with DharmaFECT (ThermoScientific) transfection reagent according to the manufacturer’s protocol. Efficacy of knockdown was analyzed by qPCR and Western blotting 48 h – 72 h after transfection.
Cell migration / invasion assay
For migration assays, cell culture inserts equipped with 8 μm membranes were used (Falcon). For invasion assays, BioCoat Matrigel invasion chambers (BD Biosciences) were used according to the manufacturer’s protocols. Briefly, 24 – 48 h after transfection, 2 – 5 ×104 cells were seeded into cell culture inserts in medium without serum. The lower chamber was filled with medium containing serum as chemoattractant. 48 – 96 h after seeding cells, which passed the membranes, were fixed and stained using Diff-Quik staining set according to manufacturer’s protocol (Medion Diagnostics). Stained filters were mounted on microscope slides with VitroClud (Langenbrinck). Quantitative analysis was done by cell counting using Image J software.
Luciferase assay
Luciferase assays were performed using TOPFlash or FOPFlash plasmids (addgene plasmid numbers 1256 and 12457, respectively) along with renilla normalization construct (pRL-TK, Promega) using Lipofectamine 2000 (Invitrogen). Luciferase constructs were transfected 48 h after siRNA transfection. Starvation medium as well as normal medium was changed the next day. After additional 24 h, cells were harvested and processed according to the DualGlo Luciferase protocol (Promega). Relative Luciferase activity was normalized to the activity of the FOPFlash mutant vector control.
Immunohistochemistry/Immunocytochemistry
The study was approved by the local ethical review committee (Research ethics committee of the Medical Association Westfalen-Lippe and Westphalian Wilhelms University; ethical vote: 2013-156-f-S). We used tissue microarrays of 362 patients. Of these, 37 were diagnosed as being TNBC by missing expression of ER, PR and HER2. Immunohistochemistry of formalin-fixed, paraffin-embedded tissue microarrays was performed using primary antibody (SFRP1, Epitomics, clone# EPR7003) and biotinylated secondary antibodies (DAKO). Detection was performed using Chromogen Red (DAKO) and H&E (Merck). Slides were embedded with Scientific Cytoseal (Thermo Scientific Fisher).
For immunocytochemistry, cells were fixed with phosphate buffered formalin. Cells were blocked with 10% Aurion (DAKO) in PBS for 1 h. Cells were washed and incubated with primary antibody (β-Catenin, Cell Signaling, # 9587) diluted with Dako REALTM Antibody Diluent (overnight at 4°C). Fluorescent visualization was carried out using suitable Alexa Fluor-conjugated secondary antibody (1:600) together with 4′,6-diamidino-2-phenylindole (1:400) in in Dako REALTM Antibody Diluent) for 1 h at RT.
Chemotherapy sensitivity assay
For analysis of chemotherapy sensitivity, cells were incubated with cytotoxic agents using decreasing concentrations: paclitaxel (10 pM - 1 μM), doxorubicin hydrochloride (500 pM - 50 μM), cis-diamineplatinum II dichloride (50 nM - 5 mM). After 96 hours, cell viability was determined via MTT (Thiazolyl Blue Tetrazolium Bromide) (all substances were received from SIGMA Aldrich) according to the manufacturer’s protocol. Measurements were performed at least in triplicates. Significance was calculated via one-side Welch’s t-test.
Flow cytometry
Following transfection cells were stained for apoptosis as well as apoptosis/necrosis using the annexin V test kit from Becton Dickinson (San José, USA). Flow cytometric cell analysis and quantification of cell death took place on a flow cytometer (CyFlow Space, Partec, Germany) as described previously [
46,
47].
Consent
Written informed consent was obtained from the patients for the publication of this report and any accompanying images.
Competing interests
The authors declare no conflict of interests.
Authors’ contribution
C.B. conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript, C.H. conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing; C.R. data analysis and interpretation; S.S. and L.B. provision of study materials; G.H. and M.G. data analysis and interpretation; B.G. and P.J.B. collection and/or assembly of data, data analysis and interpretation; L.K. conception and design, financial support; C.L.: conception and design, data analysis and interpretation. All authors read and approved the final manuscript.