Kaiso depletion attenuates transforming growth factor-β signaling and metastatic activity of triple-negative breast cancer cells

Triple-negative breast cancers (TNBCs) represent a subset of breast tumors that are highly aggressive and metastatic, and are responsible for a disproportionate number of breast cancer-related deaths. Several studies have postulated a role for the epithelial-to-mesenchymal transition (EMT) program in the increased aggressiveness and metastatic propensity of TNBCs. Although EMT is essential for early vertebrate development and wound healing, it is frequently co-opted by cancer cells during tumorigenesis. One prominent signaling pathway involved in EMT is the transforming growth factor-β (TGFβ) pathway. In this study, we report that the novel POZ-ZF transcription factor Kaiso is highly expressed in TNBCs and correlates with a shorter metastasis-free survival. Notably, Kaiso expression is induced by the TGFβ pathway and silencing Kaiso expression in the highly invasive breast cancer cell lines, MDA-MB-231 (hereafter MDA-231) and Hs578T, attenuated the expression of several EMT-associated proteins (Vimentin, Slug and ZEB1), abrogated TGFβ signaling and TGFβ-dependent EMT. Moreover, Kaiso depletion attenuated the metastasis of TNBC cells (MDA-231 and Hs578T) in a mouse model. Although high Kaiso and high TGFβR1 expression is associated with poor overall survival in breast cancer patients, overexpression of a kinase-active TGFβR1 in the Kaiso-depleted cells was insufficient to restore the metastatic potential of these cells, suggesting that Kaiso is a key downstream component of TGFβ-mediated pro-metastatic responses. Collectively, these findings suggest a critical role for Kaiso in TGFβ signaling and the metastasis of TNBCs.


INTRODUCTION
Breast cancer is the most common female cancer and a leading cause of female deaths worldwide. 1 Of the five major breast cancer subtypes, 2,3 the triple-negative breast cancers (TNBCs) have the worst prognosis because of their limited treatment options and highly metastatic nature. 4,5 Several studies suggest a role for the epithelial-to-mesenchymal transition (EMT) program in the metastatic propensity of TNBCs. 6,7 Indeed, increased expression of various EMT proteins (for example, Vimentin, ZEB1) has been reported in many TNBC cases, where they appear to correlate with increased invasiveness and poor disease-free survival. 7,8 EMT is a complex and tightly regulated process that confers mesenchymal properties (for example, increased motility and invasiveness) to epithelial cells (reviewed in Kalluri and Weinberg 9 ). The switch in cellular behavior and characteristics during EMT is accomplished mostly by EMT-associated transcription factors (for example, Snail/Slug, ZEB1/2) that function to promote the loss of epithelial components (for example, E-cadherin) and gain of mesenchymal proteins (for example, Vimentin). These EMT transcription factors are activated by many cytokines or growth factors including the transforming growth factor-β (TGFβ) (reviewed in Puisieux et al. 10 ).
The TGFβ pathway controls many normal and pathological processes in addition to EMT. 11,12 TGFβ signals are transduced either via the canonical cascade involving Smad proteins (for example, Smad2/3) or the noncanonical cascade involving non-Smad proteins (for example, phosphatidylinositol 3 kinase/ AKT, extracellular signal-regulated protein kinase-1/2; reviewed in Zhang 12 and Heldin et al. 13 ). Depending on the cellular context, TGFβ suppresses or promotes tumor progression in breast cancers (BCa). In early-stage BCa, TGFβ is a potent inhibitor of uncontrolled cell proliferation; however, in advanced BCa, TGFβ promotes metastasis as the cells become refractory to TGFβ growth inhibition. 14 The mechanism underlying the switch in TGFβ function from a tumor suppressor to tumor promoter is not well understood but studies implicate the TGFβ receptors (TGFβR1 and 2) as critical determinants of the functional specificity of the TGFβ signaling cascade. [15][16][17] A metastasis-associated TGFβ response signature that includes expression of several EMT-associated genes was recently identified in breast tumors, 18 further highlighting the importance of TGFβ signaling in EMT induction and malignant progression of BCa.
Recently, the transcription factor Kaiso was identified as a regulator of E-cadherin expression and EMT in prostate and breast tumors. 19,20 Kaiso is a unique dual-specificity transcription factor that recognizes and binds a consensus Kaiso-binding sequence (KBS), TCCTGCNA, or methylated CpG-dinucleotides. 21 Most Kaiso target genes (for example, CCND1, S100A4, MMP7, CDH1) identified to date are linked to tumor onset, progression and metastasis. [22][23][24][25] Thus, not surprisingly, Kaiso is implicated in various human cancers (breast, colon, lung, prostate), and appears to have both tumor suppressive and promoting roles. 19,20,[26][27][28][29][30] Indeed, high Kaiso expression correlates significantly with estrogen receptor-α negativity, basal/TNBCs and poor prognosis in patients with infiltrating BCa. 20,29 More recently, Kaiso was implicated as a potential drug target in glucocorticoid-combined chemotherapy in breast cancer. 30 However, the precise roles and mechanism of action of Kaiso in tumorigenesis remain poorly understood. Here, we report that high Kaiso expression in BCa patients correlates with high expression of the TGFβ signalsome and shorter metastasis-free survival. Silencing Kaiso expression in TNBC cells attenuates TGFβ signaling and TGFβR1 expression, and induces an EMT reversal concomitant with decreased EMT protein expression. More importantly, silencing Kaiso strongly inhibited TNBC cell metastasis in two mouse metastasis models. However, although expression of a constitutively active TGFβR1 in Kaiso-depleted TNBC cells rescued TGFβ signaling, this was insufficient to restore the metastatic abilities of these cells. Our results present the first evidence linking Kaiso to TGFβ signaling and BCa metastasis in vivo, and highlight a clinically relevant role for Kaiso in the metastasis of aggressive breast tumors.

RESULTS
High Kaiso expression correlates with poor prognosis in breast cancer patients Kaiso is highly expressed in several TNBC cell lines (our unpublished data) and nuclear Kaiso expression has been linked with EMT and TNBC aggressiveness. 20,29 To determine the clinical relevance of Kaiso (ZBTB33) expression in aggressive BCa, we analyzed The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO) (GSE20685) breast cancer data sets. Consistent with an earlier study, 29 most high Kaiso-expressing tumors lacked the estrogen receptor. However, the highest and most statistically significant Kaiso expression correlated with TNBC cases (Figure 1a). Importantly, Kaplan-Meier survival curves revealed that patients with high Kaiso-expressing tumors (ZBTB33 high) had a poorer overall survival (log-rank test, P = 0.0052) and Kaiso-depleted TNBC cell lines undergo mesenchymal-to-epithelial transition As a first step to unraveling the function of Kaiso in TNBC, we generated stable Kaiso depletion in two highly invasive TNBC cell lines (MDA-231 and Hs578T) using two independent Kaiso-specific short hairpin (sh)-RNAs. As Kaiso was linked to EMT, 20  Kaiso expression positively correlates with TGFβ signaling protein expression To successfully undergo metastasis, tumor cells must activate various cellular processes in addition to EMT, to enable their extravasation, survival in the circulatory system and establishment at secondary sites. 34 To elucidate how Kaiso might potentiate the complete metastatic cascade, we analyzed the TCGA BCa dataset to correlate Kaiso expression with other genes implicated in tumor progression and metastasis. We found that high Kaiso expression positively correlates with several TGFβ signaling genes including Smad2, Smad4 and TGFβR1 (Figure 3a). Examination of the expression levels of various TGFβ signaling components in Kaiso-depleted TNBC cells revealed that silencing Kaiso attenuated the expression of TGFβR1 and TGFβR2 at both the transcript and protein levels in both cell lines (Figures 3b and c). However, there were no significant changes in Smad2 or Smad4 expression in either cell line (data not shown). Notably, TGFβR1 and TGFβR2 expression was upregulated following expression of a sh-resistant Kaiso form in Kaiso-depleted MDA-231 cells (Figure 3d). iv. ii.

MDA-231
Ctrl sh-Kaiso Lung Hs578T Figure 2. Kaiso depletion inhibits breast tumor cell metastasis to the lungs and liver. (a i-vi) Hematoxylin and eosin (H&E) staining of murine lungs and liver revealed that control MDA-231 xenografts formed extensive metastases in lungs (i) and liver (iii), whereas control Hs578T xenografts formed moderate metastases that were limited to the lungs (v) of immune-deficient mice. In contrast, Kaiso-depleted MDA-231 xenografts formed very few metastases in the lungs (ii), and no metastases in the liver (iv) of immune-deficient mice. Kaiso-depleted Hs578T xenografts also formed negligible metastases in the lung (v). (b) H&E images showing extensive metastases of control MDA-231 (i) and control Hs578T (iii) in the lungs of NSG mice after tail vein injections compared with few metastases formed by Kaiso-depleted MDA-231 (ii) and Hs578T cells (iv). Scale bar, 1000 μm. Representative images are shown.

Kaiso depletion attenuates TGFβ signaling and transcriptional responses
The TGFβR1 and TGFβR2 serine/threonine kinases are essential for activation of the TGFβ signaling cascade. 14,35,36 Hence, loss of either the expression or function of TGFβR1 or TGFβR2 perturbs TGFβ signaling. [37][38][39][40] As our Kaiso-depleted cells displayed decreased TGFβR1 and TGFβR2 expression, we hypothesized that suppressing Kaiso would attenuate TGFβ signaling. Indeed, Kaiso-depleted MDA-231 and Hs578T cells treated with recombinant human TGFβ1 had negligible levels of phosphorylated Smad2 (p-Smad2) that is indicative of active TGFβ signaling. This was in striking contrast to TGFβ1-treated MDA-231 and Hs578T control cells that exhibited increased p-Smad2 ( Figure 4a). Consistent with our in vitro results, Kaiso-depleted MDA-231 and Hs578T mouse xenografts displayed reduced p-Smad2 expression in vivo compared with control MDA-231 and Hs578T xenografts ( Figure 4b). To further validate the role of Kaiso in TGFβ-mediated signaling, we examined Kaiso-depletion effects on TGFβ-target gene expression. We chose ANGPTL4 that is involved in TGFβ-mediated breast tumor cell homing to lungs 18 as both control MDA-231 and Hs578T cells displayed a proclivity for lung metastasis. Silencing Kaiso significantly reduced TGFβ-induced expression of ANGPTL4 (Figure 4c). Similarly, Kaiso depletion also attenuated TGFβ induction of ZEB1 (Supplementary Figure 2) that participates in TGFβ-mediated EMT. 41 Unexpectedly, we observed increased Kaiso (ZBTB33) transcript levels in response to TGFβ treatment in both cell lines (Supplementary Figure 3). This increase in Kaiso transcripts was abrogated by Kaiso-specific shRNA in Kaiso-depleted cells (Figure 4c). Persistent TGFβ treatment (1-24 h) also resulted in increased Kaiso protein levels that peaked at~12 h in both cell lines ( Figure 4d). Together, these results hint at a positive feedback loop between Kaiso expression and TGFβ signaling.
Kaiso binds the TGFβR1 and TGFβR2 promoter endogenously As Kaiso depletion attenuated TGFβR1 and TGFβR2 expression, we next assessed whether Kaiso promotes TGFβ signaling through regulation of TGFβR1 and TGFβR2. We performed electrophoretic mobility shift assay analyses using purified GST-Kaiso-ΔPOZ fusion proteins as previously described, 42,43 and oligonucleotides derived from the TGFβR1 (KBS 1-4) and TGFβR2 (KBS1-4) promoters that each contains several KBS and/or CpGs (Tables 1 and 2). Kaiso bound the core KBS in proximal TGFβR2 oligonucleotides (TβR2-KBS-2, 3, 4) but not the distal TβR2-KBS1 probe (Figures 5a  and b). However, Kaiso binding was abolished upon introduction of a point mutation in the core KBS in these probes (Figure 5b; Supplementary Figure 4). Surprisingly, despite the strong correlation between Kaiso and TGFβR1 expression in the TCGA BCa dataset, no binding was observed between Kaiso and any of the TβR1-KBS-1-4 probes even after methylation of the CpG sequences found in the TGFβR1 (KBS2-4) probes (data not shown).
Chromatin immunoprecipitation (ChIP) experiments subsequently revealed that Kaiso bound the endogenous TGFβR2 promoter containing core KBS in MDA-231 and Hs578T cells ( Figure 5c). Intriguingly, despite no direct interaction between Kaiso and the minimal TGFβR1 promoter region in vitro, we found that Kaiso associated with the TGFβR1 promoter endogenously ( Figure 5d). As the amplified TGFβR1 promoter region contained a CpG dinucleotide in addition to several core KBS (Table 1), we repeated the ChIP-PCR experiments using chromatin from MDA-231 and Hs578T cells treated with the demethylating agent 5′-aza-cytidine. Demethylation slightly abolished binding of Kaiso to the TGFβR1 promoter in MDA-231 cells but had no effect on Kaiso binding in Hs578T cells (Figure 5d). The specificity of Kaiso binding to the TGFβR1 and TGFβR2 promoters was confirmed using primers designed against a distal region of both promoters lacking KBS or CpG sites (Supplementary Figure 5). Collectively, these results implicate both TGFβR1 and TGFβR2 as Kaiso target genes, and suggest that Kaiso may regulate TGFβR1 expression indirectly, whereas it may directly regulate TGFβR2 expression.
TGFβR2 expression (data not shown) correlated with poor prognosis in BCa patients, although not significantly. Remarkably, increased Kaiso and TGFβR1 expression, but not increased Kaiso and TGFβR2 expression, correlated significantly with poor overall survival in BCa patients (Figures 6a and b). Kaiso may thus drive metastasis through TGFβR1 but not TGFβR2.
Kinase-active TGFβR1 rescues TGFβ signaling but not the metastatic abilities of Kaiso-depleted MDA-231 cells Based on the above findings, we questioned whether restoration of TGFβ signaling in Kaiso-depleted cells would restore their metastatic abilities. To address this, we overexpressed a constitutively kinase-active TGFβR1 (TRI 204D ) in Kaiso-depleted MDA-231 and Hs578T cells. TβRI 204D overexpression in Kaiso-depleted cells restored TGFβ signaling as evidenced by increased p-Smad2 and other non-Smad proteins (pAkt) compared with MDA-231-sh-K cells (Figure 7a). Remarkably, although TRI 204D overexpression restored TGFβ signaling, it was insufficient to restore the metastatic potential of the Kaiso-depleted cells (compare with metastatic foci generated by MDA-231-Ctrl cells in the lungs of injected mice) (Figure 7b). This suggested that Kaiso expression is important for TGFβ-mediated breast tumor metastasis.

DISCUSSION
Most cancer-related deaths are because of tumor metastasis to vital organs. 45 The recent association of Kaiso with EMT 19,20 coupled with its misexpression in several aggressive cancers (prostate, breast) implicates Kaiso in metastasis. In this study we report for the first time that Kaiso depletion attenuated the metastatic ability of highly invasive TNBC cells (MDA-231 and Hs578T) in mouse models of metastasis. As our in vitro studies showed that Kaiso-depleted cells underwent mesenchymal-toepithelial transition and exhibited a more epithelial phenotype (that is, increased E-cadherin and ZO-1 but decreased Slug, ZEB1 and Vimentin expression), the effect of Kaiso depletion on the metastatic potential of breast tumor cells may be partially attributed to the attenuated EMT phenotype observed in these cells.
EMT is itself regulated by several distinct signaling pathways. 35 Thus, it was intriguing to find that Kaiso expression positively correlates with the expression of several members of the TGFβ signalsome. Importantly, Kaiso associates with proximal TGFβR1 and TGFβR2 promoter regions, and Kaiso depletion results in reduced TGFβR1 and TGFβR2 expression, and attenuated TGFβ signaling. Consequently, TGFβ-dependent activation of target genes like ANGPTL4 and ZEB1 that are known to promote tumor dissemination and invasiveness 18,46 was impaired by Kaiso silencing. As the TGFβ pathway is highly implicated in BCa metastasis, the effect of Kaiso depletion on the metastasis of MDA-231 and Hs578T cells may be due to attenuation of TGFβ signaling in these cells, that is, loss of Kaiso-dependent regulation of TGFβR1/2 expression.
Several studies suggest that expression levels of the TGFβ receptors (high vs low) may determine the biological specificity of the TGFβ signaling cascade and the differential activation of Smad vs non-Smad signaling pathways. [15][16][17] Our finding that Kaiso regulates expression of both TGFβR1 and TGFβR2 raises the possibility that Kaiso plays a central role in TGFβ-mediated tumorigenic effects. Consistent with this theory, our studies revealed that high Kaiso and TGFβR1 but not TGFβR2 expression is associated with poor overall survival in BCa patients. As metastasis accounts for poor overall survival in cancer patients, we surmise that Kaiso-dependent regulation of TGFβR1 but not TGFβR2 promotes TNBC metastasis.
Our unexpected finding that TGFβ treatment increased Kaiso expression in breast tumor cells suggests that TGFβ signaling may positively regulate Kaiso expression, and thus form a positive feedback loop that enhances TGFβ-mediated signaling and metastasis (Figure 8a). Intriguingly, Kaiso may itself be required for TGFβ signaling or participate in other pathways implicated in BCa metastasis as overexpression of a kinase-active TGFβR1 in Kaiso-depleted MDA-231 cells was insufficient to rescue their metastatic abilities. Such findings are consistent with our model (Figure 8b), and other studies that have implicated increased Kaiso  expression in the aggressiveness and overall survival of BCa patients. 20,29 However, it remains to be determined whether increased TGFβ signaling first induces high Kaiso expression or vice versa.
Collectively, these data implicate Kaiso as an important factor in TNBC aggressiveness and metastasis and suggest that it may be a relevant target for the development of therapies that will restrain the metastasis of aggressive breast cancers such as those of the TNBC subtype. Our finding that Kaiso can modulate TGFβ signaling further suggests that targeting Kaiso will alter the prometastatic phenotype associated with TGFβ signaling in advanced breast cancers.

Immunoblot analysis
Immunoblot analysis was performed as previously described. 48

Rescue experiments
A pCDNA3 vector expressing the sequence encoding the murine Kaiso cDNA (mKaiso) that is not targeted by the Kaiso-specific shRNA was utilized for Kaiso rescue experiments. Transient transfection of the pCDNA3 mKaiso vector into MDA-231-sh-K1 was achieved using Turbofect. At 3 weeks post transfection, whole-cell lysates obtained from the pCDNA3-mKaiso and pCDNA3-empty (control) transfected cells were subjected to immunoblot analysis of specified proteins after transient selection in media containing 0.8 μg Puromycin and 1000 μg Geneticin. For rescue of TGFβ signaling, constitutively active TGFβR1 (TGFβR1 T204D ), hereafter TβRI 204D , was stably transfected into Kaiso-depleted MDA-231 cells using Turbofect. At 48 h post transfection, cells were treated with selection media containing 1000 μg Geneticin and Puromycin (Invitrogen) at 0.8 μg/ml to select for stable TRI 204Doverexpressing Kaiso-depleted clones. Total protein isolated from control and experimental (TβRI 204D ) Kaiso-depleted cells was used for immunoblot analysis. Where applicable, all experiments were performed in triplicate.

Reverse transcription-PCR (RT-PCR)
RNA was isolated from control and Kaiso-depleted breast cancer cells using the RNeasy mini kit (Qiagen, Hilden, Germany). cDNA synthesis and RT-PCR analysis were performed using the Superscript One-Step RT-PCR with Platinum Taq kit (Invitrogen) and the primers are indicated in Table 3. RT-PCR reactions were performed using the Eppendorf-Thermal cycler (Eppendorf, Hauppauge, NY, USA) under the following conditions: reverse transcription at 50°C for 30 min, followed by initial denaturation at 95°C for 5 min, 30 cycles of denaturation at 95°C for 30 s, annealing at the specified temperature as indicated in Table 3 for 30 s, extension at 72°C for 30 s, followed by a final extension at 72°C for 10 min. Then, 10 μl of each RT-PCR reaction was electrophoresed on 1% agarose/ethidium bromide gels and images captured using the Bio-Rad ChemiDoc MP imaging system. All experiments were performed in triplicate.  Table 4. The expression of each target was determined using a standard curve and normalized to the expression levels of β-actin. Statistical significance (using t-test and oneway analysis of variance with Tukey's test where appropriate) was determined using data obtained from at least three trials.
Electrophoretic mobility shift assay Double-stranded oligonucleotides corresponding to the specified KBS in the TGFβR1 and TGFβR2 promoters (see Table 1    ChIP and ChIP-PCR MDA-231 and Hs578T cells were cultured to achieve~80% confluency before chromatin isolation. Treatment with the demethylating agent, 5-azacytidine, ChIP and ChIP-PCR experiments were performed as previously described. 42,49 The To study Kaiso depletion effects on in vivo breast tumor metastasis, we injected 4.5 × 10 6 Kaiso-shRNA or control-shRNA MDA-231 or Hs578T breast tumor cells in a Dulbecco's modified Eagle's medium/serum-free media/Matrigel mixture under the fourth mammary fat pad of the right abdominal mammary gland of~5-8-week-old female NSG mice. No randomization was used in our studies as we used similar-aged pups obtained from the same breeding pair for each experiment. Most experiments were performed using at least five mice/treatment condition. Non-invasive monitoring of mice was performed weekly, and increased to 2-3 times weekly upon tumor appearance. Tumor growth was monitored externally with vernier calipers and tumor volume (in mm 3 ) measured using the following formula (length/2 × width 2 ) 2-3 times weekly. Mice were killed when tumor volume reached end point (~3300 mm 3 ), and necropsies performed blindly by a veterinary pathologist to detect macrometastases. Tissues were perfused and fixed in 10% formalin before harvest and histological examination.

Immunohistochemical staining of xenograft tissues
Harvested xenografts were embedded in paraffin before the preparation of 5 μM thick tissue sections on slides that were either stained with H&E, mouse anti-Kaiso 12H monoclonal (1:800) 50 or p-Smad2 (CST-138D4; 1:200 for MDA-231 xenografts and 1:50 for Hs578T xenografts) primary antibodies overnight at 4°C. Briefly, xenograft tissues were dewaxed by warming on a slide warmer at 60°C for 20 min followed by immersion in xylenes 3 × 5 min. All other steps were performed as previously described, 31 but we utilized PBS in place of TBS. Images were obtained using the Aperio Slide scanner (Leica Biosystems, Concord, ON, Canada).
Gene expression analysis of TCGA and GEO data sets Level 3 IlluminaHiSeq_RNASeqV2 expression (Illumina, iNC., San Diego, CA, USA) and associated clinical data were downloaded for all available patients from the TCGA data portal 51 (19 March 2014; n = 977). We used RSEM-quantified gene expression values to represent gene expression. 52 For consistency, we used transcript levels of the genes ESR1 and ERBB2 to assign estrogen receptor and HER2 status to each patient. Transcript profiling data from the GEO dataset, GSE20685 (n = 327), was performed on Affymetrix U133 Plus 2.0 gene chips (Affymetrix, Santa Clara, CA, USA) and downloaded from the GEO website. 35 Robust Multi-Array was used to preprocess the dataset and gene expression values were calculated based on median expression of all probe sets mapping to a given gene based on Unigene ID. All genomic data processing was completed using R software.

Statistical analysis
All statistical tests were completed using GraphPad Prism statistical software (GraphPad Software, Inc., La Jolla, CA, USA), and P o0.05 indicated significance. Data are presented as means ± s.e.m. Unpaired Student's t-test was used for statistical analysis of two data sets, whereas one-way analysis of variance with Tukey/Newman-Keuls test was used for analysis of more than two data sets.