Background
WWOX (WW domain-containing oxidoreductase) was originally cloned by our laboratory because it was observed to reside in a chromosomal region (ch16q23) commonly affected by deletions in breast cancer [
1]. Subsequently, it was concluded that the second most common chromosomal fragile site, FRA16D, spans the same locus as
WWOX[
1,
2]. It was determined that FRA3B (
FHIT) and FRA16D (
WWOX) loci rank second and third respectively, only after the
CDKN2A (
p16) locus, as the chromosomal sites most commonly affected by hemi- and homozygous deletions in a genome wide study of over 740 cancer lines [
3]. The high frequency of deletions affecting
WWOX in multiple solid tumors is well documented [
4‐
6]; additionally, translocations affecting
WWOX are common in multiple myeloma [
7]. Loss of WWOX expression is frequent in multiple tumor types including breast cancer. Importantly, it has been determined that over 70% of estrogen receptor alpha (ER) negative breast cancers express little or no WWOX protein, suggesting an inverse association between WWOX expression and increasing breast cancer aggressiveness [
8,
9].
WWOX behaves as a suppressor of tumor growth in some cancer lines [
10‐
12]. Contradictory results were reported with
Wwox KO mice that suffer from early life lethality; Aqeilan
et al. reported osteosarcoma development in some
Wwox KO newborn mice [
13] whereas no neoplasias were detected in
Wwox KO mice generated by our laboratory [
14]. Furthermore, we recently demonstrated that no tumors develop spontaneously in mice targeted for conditional deletion of
Wwox in the mammary gland [
15]. Interestingly,
Wwox ablation led to a significant inhibition of mammary gland ductal branching and impaired alveologenesis. Based on these studies, we concluded that
WWOX does not behave as a classical tumor suppressor gene in the normal mammary gland. Therefore, in order to gain a better understanding of the role of WWOX in breast epithelium we investigated the cellular and molecular effects of modulating WWOX expression levels in normal, immortalized human breast cells.
Methods
Cell culture and reagents
All cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and validated by DNA fingerprinting. MCF10 cells (ATCC #CRL-10318) were cultured in DMEM/F12 supplemented with 5% fetal bovine serum, 100 μg/mL hydrocortisone, 10 μg/mL insulin, 20 ng/mL EGF, 1 ng/mL cholera toxin and 1% penicillin-streptomycin. MCF7 cells (ATCC #HTB-22) were cultured in modified IMEM supplemented with 10% fetal bovine serum. 184B5 cells (ATCC #CRL-8799) were cultured in MEBM. Recombinant human TGFβ1 was purchased from R&D Systems.
Cells were infected with the following shRNA-expressing GIPZ lentiviruses (Open Biosystems) at an MOI of 5: scrambled control shRNA (RHS4348), shWWOX-A (V2LHS_255213); shWWOX-B (V2LHS_255229) or shWWOX (V2LHS_255213 and V2LHS_255229). Cells were infected according to manufacturer’s instructions. Stably WWOX silenced cells and controls were selected with 2 μg/ml puromycin and WWOX protein level was assayed by western blot.
Doxycycline-inducible WWOX expression system and other transient transfections
pLVX-Tight-Puro from Clontech’s Tet-on advance system (Clontech, Mountain View, CA) was used to construct inducible WWOX expression. Full-length human WWOX cDNA was amplified and inserted using BamH1/EcoR1 restriction enzyme sites. Lentiviral stocks were made according to manufacturer’s protocol. MCF10 cells were either stably or transiently infected by the lentiviruses carrying the target cassettes and subjected to selection with 2 μg/ml puromycin. One μg/ml of doxycycline were used to induce WWOX expression.
Transient transfections were performed using FuGene 6 transfection reagent (Promega) and plasmids used were: pCMV5b-FLAG-
SMAD3 (Addgene plasmid 11742) [
16], 3TP-LUX (Addgene plasmid 11767) [
17], pRL Renilla luciferase and pcDNA-Myc-
WWOX.
Microarray data processing, bioinformatics and statistical analyses
Total RNA was extracted from 3 biological replicates each of MCF10 scrambled (Scr), MCF10 shWWOX-A and MCF10 shWWOX-B using the RNeasy Mini kit (Qiagen). Briefly, 2 μg of RNA from each of WWOX–silenced sublines labeled with Cy5 were individually hybridized on Agilent Whole Human Genome 4X44K microarrays to analyze ~40000 transcripts (Agilent Technologies, Palo Alto, CA, USA) using the RNA derived from the corresponding MCF10 Scr sample (labeled with Cy3) as reference. For RNA labeling, we used the Quick Amp Kit (Agilent Technologies, Palo Alto, CA) by following the manufacturer’s protocol. The hybridization steps were carried out according to the Agilent protocol and images were scanned using a Genepix 4000B microarray scanner (Axon Instruments, Foster City, CA, USA). Image analysis and initial quality control were performed using Agilent Feature Extraction Software v10.2. Raw datasets have been submitted to NCBI GEO database with accession number GSE47371. We used the limma Bioconductor package for background adjustment (normexp method), within (Loess algorithm) and between (quantiles method) arrays normalization [
18]. To identify significantly up- or down-modulated genes within the hybridized samples (MCF10 shWWOX-A vs. Scr and MCF10 shWWOX-B vs. Scr) we employed the one-class Rank Products' test (q-value < 0.05; Fold change > 2) [
19]. Statistical analyses were done with the MultiExperiment Viewer software (MeV 4.8) [
20]. Differentially expressed genes derived from both analyses were compiled into one Excel spreadsheet pivot Table for comparison of overlapping data between MCF10 shWWOX-A and MCF10 shWWOX-B WWOX sub-lines. The number and identity of genes commonly affected in both models was determined. We used the normal approximation to the binomial distribution as previously described [
21] to calculate whether the number of matching genes derived from each pairwise comparison was of statistical significance (q < 0.05). Datasets were then uploaded to IPA software for automated functional annotation and gene enrichment analysis [
22]. In addition, we employed Enrichr online resource [
23] for ChIP enrichment analysis [
24].
Clonal growth, attachment and cell motility assays
For clonal growth assays, 500 cells were plated into individual wells of a 6-well plate. After 9 days of culture, colonies were fixed and stained with crystal violet. Digital images were used to determine the number and area of growing colonies using ImageJ software 1.46 (NIH).
For attachment assays, cells (4×104 per well) were seeded in serum-free medium on fibronectin, collagen IV or laminin-coated 96-well plates (BD-BioCoat; BD) and incubated for 120 min at 37°C/5% CO2. Adherent cells were fixed at different time-points by adding a cold 10% TCA solution and then processed according to the sulforhodamine B (SRB) assay (Sigma-Aldrich).
To assess cell motility we conducted a standard wound-healing assay. Briefly, 1×106 cells were seeded in each well. After cells adhered the FBS concentration in the medium was reduced to 2% to decrease cell proliferation. Two scratch wounds were made in each well. Images of the same fields were collected at 0 and 24 hrs. Wound area expressed as percent of field of view was quantified using the ImageJ software.
Real-time Q-PCR, ELISA, Western blotting and antibodies
RNA isolation and Real-time PCR was performed as previously described [
15]. Real-time assays were performed using Sybr Green and the following primer sets: FST F 5′-GCCACCTGAGAAAGGCTACC-3′, FST R 5′-TTACTGTCAGGGCACAGCTC-3′, ANGPTL4 F 5′- CACAGCCTGCAGACACAACT -3′, ANGPTL4 R 5′- AAACTGGCTTTGCAGATGCT -3′, PTHLH F 5′-CGCTCTGCCTGGTTAGACTC-3′, PTHLH R 5′-AGAATCCTGCAATATGTCCTTGG-3′, SERPINE1 F 5′-GACCGCAACGTGGTTTTCTC-3′, SERPINE1 R 5′-CATCCTTGTTCCATGGCCCC-3′, 18S rRNA F 5′-ACGGAAGGGCACCACCAGG-3′ and 18S rRNA R 5′-CACCAACTAAGAACGGCCATGC-3′. Experiments were done in triplicate and normalized to 18S rRNA expression.
Levels of FST and ANGPTL4 proteins in conditioned medium were determined using the FST Quantikine ELISA kit and the ANGPTL4 DuoSet ELISA kit (R&D Systems) according to manufacturer’s protocols. Briefly, 4×105 cells were seeded in phenol red-free DMEM/F12 medium supplemented with charcoal-stripped serum (5%) and adequate growth factors under normal conditions for 72 hrs before collection of conditioned medium.
Western blotting was performed under standard conditions by loading 20 μg of total protein per lane and transferring to PVDF membranes. Primary antibodies used were: rabbit anti-WWOX (Aldaz lab), rabbit anti-SMAD3 (Cell Signaling), mouse anti-actin (Sigma-Aldrich) and mouse anti-Myc (Origene). Secondary antibodies used were: anti-rabbit HRP (Jackson Labs) anti-mouse HRP (K&P Labs), anti-rabbit Alexa 594 and anti-mouse Alexa 488 (Invitrogen).
Co-immunoprecipitation, GST-pulldowns and Luciferase assays
For co-immunoprecipitation, cells were lysed with a buffer containing 50 nM Tris–HCl pH 7.4, 100 mM NaF, 10 mM EDTA, 10 mM Na
3VO
4, 2 mM PMSF, 1% NP-40 and 0.5% TritonX-100. Immunoprecipitations were carried out with Protein A/G beads and washed five times in the same buffer. Construction and purification of GST fusion proteins was performed as previously described [
25]. Pull-down assays were performed using immobilized purified GST or GST fusion proteins incubated with total cell lysate from MCF10 cells transfected with 1 μg of pCMV5b-Flag-SMAD3 plasmid for 48 hours.
For luciferase assays, MCF10 cells stably infected with the described Dox-inducible WWOX expression system were exposed to 1 μg/mL doxycycline for two days (or no treatment). Cells were then co-transfected with 3TP-LUX and pRL Renilla luciferase expressing control vector. Serum-free media was applied and cells were then exposed to 10 ng/mL TGFβ1 (or vehicle treatment) for 8 hours. Luciferase assays were performed according to Dual-Luciferase Assay protocol (Promega).
Chromatin immunoprecipitation (ChIP)
MCF10 cells transiently infected with the described Dox-inducible WWOX expression system were exposed to 1 μg/mL Dox for one day (or no treatment), changed to serum-free media for 16 hours then exposed to 10 ng/mL TGFβ1 for 4 hours (or vehicle treatment). ChIP was performed as described elsewhere [
26].
Real-time PCR was performed to assay SMAD3 occupation at promoter elements via the percent input method. Primers used for ChIP qPCR for the region 2000 bases upstream from the
ANGPTL4 transcriptional start site were: F: 5′-GATTTGCTGTCCTGGCATCT-3′ and R: 5′-CTCCAAGCCAGCTCATTCTC-3′. Primers for the SMAD binding element of the
SERPINE1 promoter were: F: 5′-GGGAGTCAGCCGTGTATCAT-3′ and R: 5′-TAGGTTTTGTCTGTCTAGGACTTGG-3′ [
27].
Confocal microscopy
Cells transiently transfected with pcDNA-Myc-WWOX were seeded on round, glass coverslips in 12-well plates, serum starved for 12 hours, treated with 20 ng/μL TGFβ1 for 1 hour, fixed for 15 min in 4% PBS-buffered paraformaldehyde, permeabilized with 0.05% Triton X-100 in PBS (PBS-T) for 5 min, blocked with 1% bovine serum albumin (BSA), and incubated with rabbit anti-SMAD3 (Cell Signaling) overnight at 4°C then mouse anti-Myc (Origene) for one hour at room temperature. AlexaFluor-conjugated secondary antibodies were applied for 2 hours at room temperature. Cells were washed three times in PBS-T, DAPI solution applied, washed three more times then mounted in Prolong Gold Anti-Fade (Invitrogen) on a microscope slide. Confocal microscopy was done on a Zeiss LSM510 META confocal microscope with 100X plan-apochromatic objective and oil immersion. Images were acquired in sequential mode and single-color controls were used to verify absence of crosstalk and bleed-through.
WWOX and ANGPTL4expression meta-analysis in breast cancer datasets
To perform a comparative analysis of
WWOX and
ANGPTL4 expression in breast cancer, we analyzed 819 primary carcinomas obtained from three independent studies available in public databases. The fRMA preprocessed expression matrixes of the studies GSE26639 (n = 226), GSE21653 (n = 266), and GSE20685 (n = 327) were downloaded from the InSilico database [
28]. These gene expression profiles were obtained using the Affymetrix HG U133 Plus2 platform (GPL570).
WWOX and
ANGPTL4 mRNA expression levels were estimated by using the mean expression values of the Affymetrix probes for each gene. We employed the Gaussian Mixture Model to identify bimodal distributions in the expression levels of both genes [
29]. Heatmap visualization of
WWOX and
ANGPTL4 expression profiles was done with the MultiExperiment Viewer software (MeV 4.8).
Discussion
It is clear that expression of WWOX is lost in breast cancer and that this loss becomes more frequent as the disease progresses [
8,
9,
35,
36]. Thus, we feel it is important to understand the functions of WWOX in normal breast cells and the effects of loss of expression of this protein in breast cancer progression. In this study, we have described the multiple consequences of
WWOX silencing in normal human breast cells.
WWOX knockdown leads to a pro-transformation phenotype with increased proliferation, decreased attachment to ECM substrates and increased cell motility. These phenotypes were supported by corresponding changes in gene expression as genes involved in cell cycle, DNA damage response and cell motility were found deregulated in WWOX silenced cells.
ChIP enrichment analysis identified SMAD3 as one of the most over-represented transcription factors responsible for many of the observed gene expression changes. Well known SMAD3 target genes such as
FST, ANGPTL4, PTHLH and
SERPINE1 were found significantly upregulated upon WWOX silencing. Interestingly,
ANGPTL4,
PTHLH and
SERPINE1 have all been shown to be involved in breast cancer progression and metastasis [
33,
37,
38]. We observed that these specific gene expression changes detected in WWOX knockdown cells can be reverted upon WWOX re-expression. Furthermore, we showed that WWOX protein expression significantly decreases SMAD3 promoter occupancy at target DNA elements and significantly decreases the response of a TGFβ luciferase reporter.
These observations lead us to investigate whether WWOX and SMAD3 physically interact with each other. Indeed, we demonstrate for the first time that WWOX is able to bind SMAD3 via the first WW domain and likely modulates SMAD3 transcriptional activity by cytoplasmic sequestration.
The effect of TGFβ signaling in breast cells has been described as paradoxical since it acts as an inhibitor of growth in normal mammary epithelium [
39] but transitions to being an enhancer of tumor progression in advanced breast cancer stages [
40‐
42]. The mechanisms behind this dichotomous behavior are poorly understood [
43]. In normal mammary epithelial cells TGFβ inhibits cell growth by inducing the expression of cell cycle inhibitors such as
CDKN2B (
p15) and
CDKN1A (
p21) and repressing the expression of cell cycle activators such as
MYC[
44‐
46]. On the other hand, in advanced-stage breast cancer the growth inhibitory effects of genes such a p15 and p21 are no longer effective and different subsets of pro-oncogenic and pro-metastatic genes are activated by TGFβ [
40‐
42]. In fact the majority of breast cancers demonstrate active signaling through the TGFβ pathway and some tumors secret high levels of TGFβ [
40].
SMAD protein family members are known to be regulated by a number of WW-domain containing proteins such as YAP, PIN1, NEDD4L and SMURF1/2 [
47,
48]. YAP and PIN1 interact with SMADs in a phosphorylation-dependent manner and stabilize SMAD-cofactor binding at promoter elements to enhance transcriptional effects [
47]. NEDD4L and SMURF1/2 are E3 ubiquitin ligase proteins responsible for SMAD protein turnover [
43,
47]. WWOX, also a WW domain containing cytoplasmic protein, is known to physically interact with the PPXY motif of various transcription factors via such domains and it has been postulated that one of its mechanisms of action is to impede nuclear translocation, thus regulating their transcriptional activity [
5,
49]. In this study, we propose that
via the same mechanism WWOX acts as an inhibitor of TGFβ signaling by binding to SMAD3 and modulating nuclear translocation of this transcription factor, thus reducing promoter occupation and transcriptional activation. In the absence of WWOX, a condition that emulates advanced breast cancer, SMAD3 can enter the nucleus uninhibited. Promoter specificity and activation of pro-metastatic genes such as
ANGPTL4,
PTHLH and
SERPINE1, depends on SMAD3 interaction with specific transcriptional co-activators such as RUNX2. RUNX2 is a SMAD3 coactivator that has been shown to induce EMT [
50] and pro-metastatic genes such as
ANGPTL4[
33] in a TGFβ-dependent manner. Interestingly, it has been previously demonstrated that WWOX also binds to RUNX2 (Figure
5A) and modulates its transcriptional activity [
32]. The ability of WWOX to affect the transcriptional activity of not only SMAD3 but also of a key transcriptional cofactor such as RUNX2 suggests that the presence or absence of WWOX could be critical for modulating TGFβ signaling and, more importantly, for the activation or repression of specific transcriptional targets known to be associated with tumor progression. Interestingly, our breast cancer gene expression meta-analysis indicates an inverse correlation between
WWOX and
ANGPTL4. Furthermore, tumors with the
WWOX
lo
/ANGPTL4
hi
signature correlate with breast cancer subtypes characterized by poor prognosis. Thus, the
WWOX
lo
/ANGPTL4
hi
breast cancer subset could represent good candidates for exploring anti-TGFβ therapeutic approaches.
Conclusions
Loss of WWOX expression leads to significant upmodulation of SMAD3 transcriptional activity leading to overexpression of multiple gene targets associated with breast cancer progression. WWOX directly binds SMAD3 via WW domain 1 and inhibits its transcriptional activity by sequestering this transcription factor in the cytoplasmic compartment.
In summary, we hypothesize that the progressive loss of WWOX expression in advanced breast cancer contributes to deregulating the TGFβ pathway and, more importantly, may explain some of the pro-metastatic effects resulting from TGFβ/SMAD3 hyperactive signaling in advanced breast cancer.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CMA and BWF contributed the conception of the project and the design of all experiments. Experiments were conducted by BWF, XG, MJZ and JL. CRJ contributed confocal microscopy expertise. MCA carried out all bioinformatic analyses. BWF, MCA and CMA wrote the main body of the manuscript. All authors read and gave their final approval for the manuscript.