TCF21 hypermethylation regulates renal tumor cell clonogenic proliferation and migration

We recently identified hypermethylation at the gene promoter of transcription factor 21 (TCF21) in clear cell sarcoma of the kidney (CCSK), a rare pediatric renal tumor. TCF21 is a transcription factor involved in tubular epithelial development of the kidney and is a candidate tumor suppressor. As there are no in vitro models of CCSK, we employed a well‐established clear cell renal cell carcinoma (ccRCC) cell line, 786‐O, which also manifests high methylation at the TCF21 promoter, with consequent low TCF21 expression. The tumor suppressor function of TCF21 has not been functionally addressed in ccRCC cells; we aimed to explore the functional potential of TCF21 expression in ccRCC cells in vitro. 786‐O clones stably transfected with either pBABE‐TCF21‐HA construct or pBABE vector alone were functionally analyzed. We found that ectopic expression of TCF21 in 786‐O cells results in a trend toward decreased cell proliferation (not significant) and significantly decreased migration compared with mock‐transfected 786‐O cells. Although the number of colonies established in colony formation assays was not different between 786‐O clones, colony size was significantly reduced in 786‐O cells expressing TCF21. To investigate whether the changes in migration were due to epithelial‐to‐mesenchymal transition changes, we interrogated the expression of selected epithelial and mesenchymal markers. Although we observed upregulation of mRNA and protein levels of epithelial marker E‐cadherin in clones overexpressing TCF21, this did not result in surface expression of E‐cadherin as measured by fluorescence‐activated cell sorting and immunofluorescence. Furthermore, mRNA expression of the mesenchymal markers vimentin (VIM) and SNAI1 was not significantly decreased in TCF21‐expressing 786‐O cells, while protein levels of VIM were markedly decreased. We conclude that re‐expression of TCF21 in renal cancer cells that have silenced their endogenous TCF21 locus through hypermethylation results in reduced clonogenic proliferation, reduced migration, and reduced mesenchymal‐like characteristics, suggesting a tumor suppressor function for transcription factor 21.

We recently identified hypermethylation at the gene promoter of transcription factor 21 (TCF21) in clear cell sarcoma of the kidney (CCSK), a rare pediatric renal tumor. TCF21 is a transcription factor involved in tubular epithelial development of the kidney and is a candidate tumor suppressor. As there are no in vitro models of CCSK, we employed a well-established clear cell renal cell carcinoma (ccRCC) cell line, 786-O, which also manifests high methylation at the TCF21 promoter, with consequent low TCF21 expression. The tumor suppressor function of TCF21 has not been functionally addressed in ccRCC cells; we aimed to explore the functional potential of TCF21 expression in ccRCC cells in vitro. 786-O clones stably transfected with either pBABE-TCF21-HA construct or pBABE vector alone were functionally analyzed. We found that ectopic expression of TCF21 in 786-O cells results in a trend toward decreased cell proliferation (not significant) and significantly decreased migration compared with mock-transfected 786-O cells. Although the number of colonies established in colony formation assays was not different between 786-O clones, colony size was significantly reduced in 786-O cells expressing TCF21. To investigate whether the changes in migration were due to epithelial-to-mesenchymal transition changes, we interrogated the expression of selected epithelial and mesenchymal markers. Although we observed upregulation of mRNA and protein levels of epithelial marker E-cadherin in clones overexpressing TCF21, this did not result in surface expression of E-cadherin as measured by fluorescence-activated cell sorting and immunofluorescence. Furthermore, mRNA expression of the mesenchymal markers vimentin (VIM) and SNAI1 was not significantly decreased in TCF21-expressing 786-O cells, while protein levels of VIM were markedly decreased. We conclude that reexpression of TCF21 in renal cancer cells that have silenced their endogenous TCF21 locus through hypermethylation results in reduced clonogenic proliferation, reduced migration, and reduced mesenchymal-like characteristics, suggesting a tumor suppressor function for transcription factor 21.

Introduction
Clear cell sarcoma of the kidney (CCSK) is an uncommon pediatric renal tumor that comprises between 2% and 5% of all primary renal tumors in children (Furtwangler et al., 2013;Gooskens et al., 2012;Mullen et al., 2014). This tumor is observed most often in children between 2 and 4 years of age and is characterized by a high malignant potential (Argani et al., 2000;Gooskens et al., 2012). Apart from internal tandem duplications within exon 16 of BCOR (reported in the majority of CCSKs) and translocation t(10;17)(q22;p13) leading to a fusion of YWHAE and NUTM2B/E (reported in about 10% of CCSKs), the genome of CCSK seems to be rather stable (Astolfi et al., 2015;Gooskens et al., 2015;Karlsson et al., 2015;Roy et al., 2015;Ueno-Yokohata et al., 2015). This paucity of large genomic imbalances and the few detected somatic mutations in CCSK prompted the investigation of the epigenome of CCSKs. Our study as part of the NCI-initiated therapeutically applicable research to generate effective treatment project (TARGET) identified hypermethylation of the 5 0 region of transcription factor 21 (TCF21) in all studied CCSK samples, except for samples harboring the YWHAE-NUTM2 fusion transcript (Gooskens et al., 2015). This hypermethylation was negatively correlated with TCF21 expression. Other tested pediatric renal tumor samples and normal kidney samples showed significantly lower TCF21 methylation levels.
TCF21 (also referred to as Pod-1, capsulin, and epicardin) encodes a basic helix-loop-helix transcription factor that binds DNA and regulates cell differentiation and cell-fate specification during development (Hidai et al., 1998). It is expressed in embryonic mesenchymal cells surrounding areas of epithelial development in the kidney, heart, lung, and gastrointestinal tract (Hidai et al., 1998;Lu et al., 1998;Quaggin et al., 1998). TCF21 expression rapidly decreases in postnatal tissues with the exception of a subset of interstitial cells in organs including the kidney, heart, lung, and spleen (Plotkin and Mudunuri, 2008). Antisense inhibition of TCF21 has been reported to disrupt epithelial differentiation and branching morphogenesis of the epithelium in murine embryonic kidney, suggesting a role for TCF21 in epithelial-mesenchymal interactions (Quaggin et al., 1999). Gene deletion studies in chimeric mice have shown that loss of TCF21 in the kidney results in decreased glomerulogenesis and tubulogenesis (Cui et al., 2003). Of note, suppression of TCF21 expression by siRNA within a mouse kidney progenitor cell line that endogenously expresses TCF21 resulted in increased cell proliferation and migration, as well as reduced expression of smooth muscle genes and myofibroblast secreted proteins (Plotkin and Mudunuri, 2008).
Currently, no CCSK cell lines or models are available to functionally verify the role of TCF21 hypermethylation in this renal tumor type. Therefore, we searched for an alternative model. A literature search revealed that TCF21 hypermethylation is also present in clear cell renal cell carcinomas (ccRCC): renal tumors with another biology and phenotype, which most often occur in adults (Costa et al., 2011;Ye et al., 2012). Previously performed clinical studies on the prognostic impact of TCF21 expression in ccRCC tissue revealed that TCF21 expression levels significantly correlated with Fuhrman nuclear grade and cancer-specific survival of ccRCC patients (Ye et al., 2012). TCF21 methylation levels in urine samples were significantly correlated with tumor size, Fuhrman grade, and clinical stage (Xin et al., 2016). However, no studies have functionally addressed the tumor suppressor activity of TCF21 in renal cancer cells. Therefore, the aim of this study was to explore the functional potential of TCF21 expression in the tumorigenesis of ccRCC in vitro.

Vector construction and transfection
The coding sequence of TCF21 (including HA-tag) was cloned out of a pCS2+-TCF21 construct [kindly provided by Prof. Christopher Plass and Khalifa Arab, Division of Epigenomics and Cancer Risk Factors, German Cancer Research Center (DKFZ), Heidelberg, Germany], amplified, and recloned into a pBABE-puro vector. Plasmid DNAs were sequence-confirmed. Twenty-five micrograms of pBABE-TCF21-HA or pBABE-puro vector alone was transfected into cells of the 786-O cell line using electroporation. Electroporation was performed in a 4-mm gap cuvette (#165-2088; Bio-Rad Laboratories, Hercules, CA, USA) using a Gene Pulser (Bio-Rad, M€ unchen, Germany) with electric parameters 24 kV with 1000 uF capacitance; a single exponential decay pulse was applied. Selection medium containing puromycin was added to the cells after 48 h of recovery, and colonies grew after 2 weeks of culture. Eight colonies were selected for functional assays: four from pBABE-puro mock-transfected 786-O cells (N1F4, B5F1, N1G4, and B6D10) and four expressing HA-tagged exogenous TCF21 (2B12, 5D2, 9D12, and 9H9). Unless the clones are specifically named, data from 'pBABE-mock' or 'pBABE-TCF21' contain pooled data from all four clones.

Proliferation assay
A total of 2 9 10 4 786-O pBABE-mock cells (four different clones) and 786-O pBABE-TCF21 cells (four different clones) were plated in 6-well plates in two separate experiments in triplicate in selection medium and were counted after 2, 4, and 5 days. Cells were counted in duplicate with the Bio-Rad TC20 Automated Cell Counter (Hercules, CA, USA) and averaged.

Colony formation assay
A total of 200 786-O pBABE-mock cells (four different clones) and 786-O pBABE-TCF21 cells (four different clones) were plated in 6-well plates on five different occasions in triplicate in selection medium until 786-O pBABE-mock cells formed sufficiently large colonies (> 50 cells per colony, 10-14 days). Cells were then washed with PBS and fixed and stained using a mixture of 6.0% glutaraldehyde and 0.5% crystal violet for 30 min at room temperature, before being washed in water. Wells were individually photographed and analyzed by the Colony Counter plugin for Image J software. Numbers of colonies and colony area were calculated in Image J.

Migration assays
Boyden chambers (Transwell Permeable Supports; Cole-Parmer; BergmanLabora AB, Danderyd, Sweden) were used to evaluate the migratory capacity of 786-O pBABE-mock cells (N1G4 and B6D10 clones) and 786-O pBABE-TCF21 cells (9D12 and 5D2 clones). 3 9 10 4 cells were seeded in the upper section of a Boyden chamber containing serum-free medium. The lower section, separated from the upper one by a membrane with 8-lm pores, contained medium with 10% serum. Cells were grown at 37°C for 16 h, fixed in methanol, and stained with DAPI (1 : 2000). Unmigrated cells on top of the membrane were removed with a cotton tip. The number of migrated cells on the bottom of the membrane was evaluated by counting cells in 10 different views per membrane using microscopy at 109 magnification. The experiments were performed in duplicate and were repeated two times (first time N1G4/B6D10 versus 9D12/5D2, second time N1G4 versus 5D2).
For the real-time migration assay, in each experiment, a modified Boyden chamber setup was used, with 10% serum in the lower well as chemoattractant. Migration was recorded in real-time at 5-min intervals over 24 h using the xCELLigence Real-Time Cell Analysis system (Acea Biosciences, San Diego, CA, USA), which quantifies cell numbers in the lower compartment by electrical impedance as reported by Gremmels et al. (2014). Rate of migration was quantified by calculating a series of moving slopes over 12 interval periods and taking the maximum slope in each migration curve. The moving slope takes the slope over small sections of the S-curve, 12 intervals of 5 min. So one takes the slope over the 0to 60-min interval, 5-to 65-min, 10-to 70-min, 15-to 75-min interval, and so on until the last point. Of the values, the highest value is the maximum slope for a given curve, which we used for analysis.
Within each experiment, four technical replicates per clone were recorded and the experiment was repeated on three separate days. Data were analyzed using a hierarchical linear mixed model using experimental day as a random variable.

Zebrafish injections
An incross of vhl+/À zebrafish embryos (van Rooijen et al., 2009) carrying a transgene for GFP-labeled cadherin-17 (kidney-specific) at the 1-2 cell stage were injected with 1-2 nL of mRNA (concentration 50 ngÁlL À1 ) transcribed from the linearized pCS2+-TCF21 plasmid (SP6 mMessage mMachine kit; Ambion, Oudeschoot, the Netherlands) in pure water with 0.05% Phenol Red using a nanoject2000 microinjector (World Precision Instruments, Sarasota, FL, USA). Pronephros volume (determined by GFP fluorescence) was measured at 5.75 days postfertilization and imaged with an LSM700 confocal microscope (Zeiss, Oberkochen, Germany). All embryos were genotyped after imaging as previously described (van Rooijen et al., 2009), and only confirmed vhlÀ/À embryos were analyzed. All zebrafish experiments were approved by the Animal Care Committee of the University Medical Center Utrecht, the Netherlands.

Quantitative RT-PCR
Total RNA was extracted and purified using the RNeasy Mini Kit (74106; QIAGEN, Hilden, Germany). cDNA was synthesized from 1 lg of RNA by the iScript cDNA Synthesis Kit (170-8891; Bio-Rad) according to the supplier's protocol. A cDNA concentration of 5 ngÁlL À1 was used for the quantitative RT-PCRs; samples were run with iQ SYBR Green Supermix (170-8880; Bio-Rad) and CFX96 Touch Real-Time PCR Detection System (Bio-Rad); 95°C for 3 min, followed by 40 cycles of 10 s of 95°C, 10 s of 50-65°C and 20 s of 72°C, then 10 s at 95°C followed by a melt of the product from 65 to 95°C. Values were normalized to the housekeeping gene GAPDH. All experiments were performed in triplicate on two different dates, from independent mRNA harvests. Relative expression ratios of target genes were calculated using the comparative cycle threshold method (Vandesompele et al., 2002). Primer sequences are provided in Table S1.

Fluorescence-activated cell sorting (FACS)
We trypsinized cells from each clone and stained unfixed live cells with anti-E-cadherin (Cat # 563571; BD Biosciences, Vianen, the Netherlands) and anti-CD24 (Cat # 562789; BD Biosciences). Ten thousand events were acquired in a FACSCanto II flow cytometer (BD Biosciences). Representative samples were plotted as histograms; median fluorescence intensity (MFI) of each channel was used for quantification.

Immunofluorescence
Coverslips were seeded with cells at low (50% confluent) or high density (confluent) and cultured for 24 h with medium containing serum, followed by 40 h with serum-free medium. Cells were rinsed with PBS and fixed with 4% PFA for 15 min at room temperature, followed by permeabilization with 1% bovine serum albumin/0.1% Triton X-100 in PBS. Cells were stained with rabbit anti-E-cadherin (15148, 1 : 500; Abcam) for 4 h at room temperature. Secondary antibody Alexa Fluor-488 anti-rabbit (1 : 500; Invitrogen) was incubated at room temperature for 1 h. Z-stacks for 10 cells per condition were generated using a Zeiss LSM700 confocal microscope and photographed using Zeiss Zen black edition software (Jena, Germany). Maximum projected images of representative cells are shown.

Statistical analysis
Statistical analyses were performed by the two-tailed Student's t-test using GraphPad, except for statistical analysis of the real-time migration assay (see paragraph 2.6.) and analysis of qPCR TCF21 values as shown in Fig. 1D, which were analyzed using one-way ANOVA. Differences were considered statistically significant at P < 0.05 (*) and P < 0.01 (**).

786-O ccRCC cells have low endogenous levels of TCF21
We first asked whether TCF21 methylation (Fig. 1A) could impact ccRCC cell proliferation. To this end, we generated isogenic stable cell lines for in vitro comparison of cell proliferation in 786-O cells transfected with an empty pBABE vector ('pBABE-mock' clones N1F4, B5F1, B6D10, N1G4) and 786-O cells stably expressing pBABE-TCF21 (clones 2B12, 5D2, 9D12, 9H9). TCF21 mRNA and protein expression is very low to undetectable in control 786-O cells by qPCR or by western blot (Fig. S1). Overexpression of TCF21 was examined by qPCR (Fig. 1B) and by western blot (Fig. 1C, Fig. S1). Treatment of 786-O cells with 3 lM of the demethylating agent decitabine (5-aza-2 0 -deoxycytidine) for 96 h increased TCF21 mRNA levels ( Fig. 1D), supporting data from Ye et al. (2012) that the endogenous low level of TCF21 in 786-0 cells can be at least partly explained by methylation of the TCF21 promoter.

TCF21 ectopic expression suppresses clonogenic cell proliferation
To investigate whether TCF21 expression results in a reduction in cell proliferation in the context of the ccRCC cell line 786-O, we performed standard growth curves on the four clones that were pBABEmock-transfected and compared them to standard growth curves from the four clones expressing TCF21 after 2, 4, and 5 days. Although the difference in proliferation was not significant at any of the chosen time points, we observed a mild trend toward decreased proliferation in the pBABE-TCF21 clones ( Fig. 2A). We therefore suggest that TCF21 expression moderately suppressed proliferation of this 786-O cancer cell line and that the hypothesis that proliferation is suppressed by TCF21 warranted further investigation.
We next assessed the ability of our clones to generate colonies from single cells in colony formation assays. No difference in the overall number of established colonies was observed between 786-O pBABEmock and 786-O pBABE-TCF21 cells (P > 0.5). However, individual colonies of 786-O pBABE-mock cells were larger compared with TCF21-expressing cells (Fig. 2B,C, P < 0.01). These data are partly consistent with the proliferation data and support the idea that while TCF21 overexpression does not seemingly affect cell survival (colony number), proliferation (colony size) is suppressed.

VhlÀ/À zebrafish pronephros phenotype is not rescued by TCF21
786-O cells are derived from a ccRCC which, like most ccRCCs, harbors biallelic somatic mutations of the von Hippel-Lindau (VHL) tumor suppressor gene (Gnarra et al., 1994). We have previously generated a zebrafish model with biallelic constitutive vhl mutations and have reported the embryonic renal phenotype as exhibiting an alveolar hyperplastic morphology (van Rooijen et al., 2011). Not much is known about the endogenous TCF21 expression in zebrafish although lineage tracing experiments suggest a role for TCF21 in epicardial cell fate (Kikuchi et al., 2011). Given the apparent effect of TCF21 on 786-O vhlÀ/À cells, we questioned whether injection of TCF21 RNA into our vhlÀ/À zebrafish embryos would partially or completely rescue the pronephric phenotype. To that end, 17 vhlÀ/À zebrafish embryos were injected with human TCF21 RNA (n = 8), or mock-injected with dye only as a control (n = 9) within an hour after fertilization, and imaged 4 days later, when the pronephros phenotype in vhlÀ/À is very clear, with z-stacks on a confocal microscope. All embryos were genotyped for confirmation after the experiment and only images of confirmed vhlÀ/À were analyzed. No differences were observed in pronephros size between mock-injected or TCF21 mRNA-injected zebrafish embryos (Fig. S2).

TCF21 expression suppresses migration
Because the effect of the colony formation assays could only be partially explained by the modest cell proliferation differences observed in Fig. 2A, we hypothesized that the colony area differences (Fig. 2C) might as well be attributed to altered cell migration. Initial analysis of migration using the chemotaxisbased Boyden chamber assay on two pBABE-mock clones (N1G4 and B6D10) and two pBABE-TCF21 clones (5D2 and 9D12) showed a trend toward inhibition of migration in TCF21-expressing cells (analysis repeated for N1G4 and 5D2 clones). Fifty-nine percent fewer of the TCF21-expressing cells migrated on average (Fig. 3A). To verify these findings, we subsequently serum-starved all four pBABE-mock and four pBABE-TCF21 clones for 24 h and then exposed them to either empty medium or 10% FBS for each clone in triplicate and measured impedance changes 12 times per hour for 24 h using an xCELLigence system. We observed a decrease in cell migration as represented by impedance measurements in the isogenic clones expressing TCF21 (P = 0.0495, Fig. 3B,C). Taken together, these data indicate that ccRCC cells become less likely to migrate upon reconstitution of TCF21 expression.

E-cadherin is upregulated but mislocalized upon TCF21 expression
Epithelial-to-mesenchymal transition (EMT) is a reversible process by which fully differentiated cells lose their epithelial features and acquire a migratory mesenchymal phenotype, with a concomitant increased expression of mesenchymal-associated proteins and decreased expression of epithelial markers. EMT is known to contribute to the metastasis of RCC (Mikami et al., 2014), although the underlying cellular and molecular mechanisms have not been clarified yet. The inhibition of migration in ccRCC cells by reexpression of TCF21 (Fig. 3) suggests that TCF21 might function in the renal epithelium by consolidating epithelial characteristics, or alternatively, in repressing mesenchymal characteristics. We investigated mesenchymal and epithelial marker expression in the eight 786-O clones expressing pBABE-mock versus pBABE-TCF21 by qPCR. No changes were observed in mRNA levels of the mesenchymal markers VIM and snail (SNAI1), but we did observe significant upregulation of epithelial marker E-cadherin (CDH1) both at the mRNA and at the protein level in the clones expressing ectopic TCF21 (Fig. 4A,B). These data could suggest that cells with hypermethylated TCF21 promoters are more predisposed to EMT, thereby potentially contributing to the tumor suppressor function of TCF21 in renal tumors. In a melanoma cell line (C8161), re-expression of TCF21 was described to activate expression of the metastatic suppressor KISS1 and consequently inhibit motility of the cells (Arab et al., 2011;Zhang et al., 2012). We investigated KISS1 expression levels in 786-O clones expressing pBABE-mock and clones expressing pBABE-TCF21; expression levels were undetectably low in both untransfected and transfected 786-O cells, while being robustly expressed in control human placenta (average C q value 21.4).
As consistent upregulation of genes involved in the Sonic hedgehog signaling pathway and Akt cell proliferation pathway has been reported in both CCSKs and ccRCCs (Cutcliffe et al., 2005;Dormoy et al., 2009;Guo et al., 2015), we investigated expression of a few genes involved in these pathways [epidermal growth factor receptor (EGFR), smoothened (SMO), CCND1] in pBABE-mock and pBABE-TCF21 cells; no significant difference in expression was observed (Fig. 4A).
To test the functional relevance of increased CDH1 expression, we performed fluorescence-activated cell sorting (FACS) of extracellular epithelial differentiation markers CDH1 (E-cadherin) and CD24. Each 786-O clone expressing either pBABE-mock or pBABE-TCF21 was stained by directly conjugated fluorescence antibodies for the endogenous extracellular epitope of CDH1 or CD24 and sorted on the strength of the fluorescent signal and side scatter. In contrast to the qPCR data, we observed no differential CDH1 or CD24 protein expression in the pBABE-TCF21 clones compared to the pBABE-mock clones (Fig. 4C).
Because the upregulation of CDH1 did not correlate with increased CDH1 extracellular epitope, we investigated the subcellular localization of E-cadherin using immunofluorescence. We observed that although upregulation of E-cadherin is also observed in 786-O clones expressing ectopic TCF21, it is not localized at the plasma membrane in correspondingly high levels in both low-density and confluent cultures (Fig. 4D). We conclude that TCF21 directly affects expression of CDH1, although possibly the cancer cell line we used does not have the intrinsic ability to incorporate significantly more E-cadherin into the plasma membrane. Lastly, we examined levels of mesenchymal marker VIM by western blot and observed that 786-O clones expressing TCF21 (Fig. S1B) display lower levels of VIM (Fig. 5).

Discussion
In the present study, functional experiments demonstrate that reconstitution of TCF21 expression in 786-O cells results in decreased clonogenic proliferation and suppressed migration in vitro. These data support a tumor suppressor function for TCF21 in the context of ccRCC, in accordance with several studies that have demonstrated that TCF21 acts as a tumor suppressor gene in different tumor types. For example, re-expression of TCF21 reduced cell growth and colony formation in lung cancer cells in vitro and in vivo (Smith et al., 2006), reduced cell proliferation, promoted apoptosis and suppressed cell invasion and migration in colorectal cancer in vitro (Dai et al., 2016), and reduced cell proliferation and epithelial-mesenchymal transition in breast cancer cells in vitro (Wang et al., 2015). In addition, downregulation of TCF21 through hypermethylation has been reported to be associated with poor outcome in patients with ccRCC and metastatic melanoma (Ye et al., 2012).
As reported before, and confirmed in the current study, expression of TCF21 is (at least partly) repressed by methylation in 786-O cells (Ye et al., 2012). TCF21 hypermethylation has also been demonstrated in other ccRCC cell lines and human ccRCC tissue samples (Costa et al., 2011;Xin et al., 2016;Ye et al., 2012). Likewise, in other human cancer types, such as gastric cancer, colorectal cancer, melanoma, head and neck cancer, and non-small-cell lung carcinoma, hypermethylation is described to be the predominant mechanism for TCF21 downregulation (Arab et al., 2011;Dai et al., 2016;Smith et al., 2006;Yang et al., 2015).
To increase metastatic and invasive capacities, cancer cells often show loss of epithelial and gain of mesenchymal characteristics, which permit invasion along the basement membrane, establishing an opportunity for metastasis (Birchmeier et al., 1996;Thiery, 2002). TCF21 is known to be involved in mesenchymal-toepithelial transition (MET) (Acharya et al., 2012;Quaggin et al., 1999;Smith et al., 2006;Wang et al., 2015). Although we did not detect decreased mRNA expression of mesenchymal markers in our 786-O cells transfected with TCF21, our data support a positive association of TCF21 expression and increased mRNA and protein expression of the epithelial marker E-cadherin (CDH1), which was markedly upregulated in TCF21-expressing 786-O cells. However, immunofluorescence and FACS experiments demonstrate that despite global upregulation, expression E-cadherin is not increased at the plasma membrane, which may be cell specific for 786-O cells. Inactivation of CDH1 is described to be prominently associated with tumor invasiveness, metastatic dissemination, and poor patient prognosis; the significance of CDH1 expression for metastatic potential has been shown in a variety of in vitro and in vivo models (Frixen et al., 1991;Onder et al., 2008;von Burstin et al., 2009). Therefore, increase in CDH1 expression might be a possible reason for decreased migration of TCF21-expressing 786-O cells in the current study, through alternative downstream transcriptional pathways, because we did not detect changes in the extracellular epitope for CDH1 (Canel et al., 2013;Onder et al., 2008). Furthermore, our results that overexpression of TCF21 inhibits migration are in accordance with in vitro studies in colorectal cancer cell lines (Dai et al., 2016(Dai et al., , 2017 as well as siRNA suppression of endogenous TCF21 in renal progenitor cells which increased migration (Plotkin and Mudunuri, 2008). Although re-expression of TCF21 in melanoma cells and Caki-1 ccRCC cells has been described to activate expression of the metastatic suppressor KISS1 (Arab et al., 2011;Zhang et al., 2012), in the current study KISS1 expression does not seem to be involved in the suppressed migration of 786-O cells transfected with TCF21. It would also be of value to explore several recently identified TCF21 targets expressed in renal tubules such as IL6R, SH2B3, and SMG6 and interrogate their contribution to phenotypic differences (Sazonova et al., 2015). Further studies are warranted to consolidate these data.
Recently, we reported that CCSKs bearing the BCOR internal tandem duplication showed hypermethylation of TCF21, while CCSKs bearing the translocation t(10;17)(q22;p13) showed significantly lower methylation levels of the TCF21, suggesting that internal tandem duplication of BCOR may be directly or indirectly responsible for TCF21 hypermethylation in CCSKs (Gooskens et al., 2016). To functionally validate the tumorigenic role of TCF21 hypermethylation in CCSKs, CCSK models urgently need to be developed. We explored the possibility of using a vhlÀ/À zebrafish model for reconstitution, but did not observe any phenotypic modulation. Due to the rarity of CCSK and limited availability of histologically verified fresh tissue, it is a challenge to develop such research tools as cell lines or other CCSK models. If we are able to develop efficient and physiologically relevant models through additional studies and could generate similar results in CCSK cells as in ccRCC cells, then treatment with demethylating agents might be an option for these patients. Currently, the most studied and promising demethylating agents are the DNA methyltransferase inhibitors decitabine and azacitidine (Kulis and Esteller, 2010).

Conclusions
In summary, our study revealed that restoration of TCF21 expression in 786-O ccRCC cells results in decreased clonogenic proliferation and migration. CCSK models need to be developed to fully determine the influence of reconstitution of TCF21 expression in CCSK cells. Three clones expressing pBABE-TCF21 show reduced levels of VIM on western blot as compared to four pBABE-mock clones. Clone 9D12 was not included in the analysis as it did not express TCF21 at the time lysates were prepared (Fig. S1B). N = 2 independent experiments. acquired the data. SLG, TDK, and RHG analyzed and interpreted the data. SG, TDK, RHG, and MMvdHE wrote the manuscript. Fig S1. (A) Western blot of anti-TCF21 (18 kDa) on HEK293T cell lysate (far left, positive control for endogenous protein), 786-0 parental cells (far right, negative control), and clones stably transfected with either empty pBABE-puro plasmid '(À)', or TCF21 '(+)' in pBABE-puro. Equal loading is indicated by btubulin staining (55 kDa). Experiment was performed twice independently. (B) Western blot of anti-TCF21 (18 kDa) with the same lysates used Fig. 5 shows that TCF21 was no longer stably expressed in clone 9D12, which was consequently not loaded on to the blot for VIM expression (Fig. 5). Fig. S2. Confocal images of the trunk of vhlÀ/À zebrafish embryos (5 days postfertilization, anterior is to the left in all images) in a background with green fluorescent pronephros (Tg-cdh17:GFP). Table S1. Primer sequences used for qPCR experiments.