Renal Carcinoma-associated Transcription Factors TFE3 and TFEB Are Leukemia Inhibitory Factor-responsive Transcription Activators of E-cadherin*

Translocations of the genes encoding the related transcription factors TFE3 and TFEB are almost exclusively associated with a rare juvenile subset of renal cell carcinoma and lead to overexpression of TFE3 or TFEB protein sequences. A better understanding of how deregulated TFE3 and TFEB contribute to the transformation process requires elucidating more of the normal cellular processes in which they participate. Here we identify TFE3 and TFEB as cell type-specific leukemia inhibitory factor-responsive activators of E-cadherin. Overexpression of TFE3 or TFEB in 3T3 cells activated endogenous and reporter E-cadherin expression. Conversely, endogenous TFE3 and/or TFEB was required for endogenous E-cadherin expression in primary mouse embryonic fibroblasts and human embryonic kidney cells. Chromatin precipitation analyses and E-cadherin promoter reporter gene assays revealed that E-cadherin induction by TFE3 or TFEB was primarily or exclusively direct and mitogen-activated protein kinase-dependent in those cell types. In mouse embryonic fibroblasts, TFE3 and TFEB activation of E-cadherin was responsive to leukemia inhibitory factor. In 3T3 cells, TFE3 and TFEB expression also induced expression of Wilms' tumor-1, another E-cadherin activator. In contrast, E-cadherin expression in model mouse and canine renal epithelial cell lines was indifferent to inhibition of endogenous TFE3 and/or TFEB and was reduced by TFE3 or TFEB overexpression. These results reveal new cell type-specific activities of TFE3 and TFEB which may be affected by their mutation.

Deregulated expression of the related transcription factors TFE3 and TFEB is associated with rare, juvenile forms of the malignancy renal cell carcinoma (RCC), and TFE3 mutation with alveolar soft part sarcoma (for review, see Ref. 1). The genetic lesions are translocations that lead to dramatic over-expression of TFE3 or TFEB protein sequences. Five different genetic loci have been identified as translocation partners for TFE3, which lead to the creation of a chimeric protein containing the translocation partner at the N terminus fused to the C-terminal portion of TFE3 which includes its DNA binding and multimerization domains (1). TFEB translocations result in promoter substitution and do not change the coding sequence (2,3). Although there is evidence that TFE3 fusion partners contribute oncogenic properties to the fusion protein (4,5), TFEB overexpression in RCC 1 and the ability of normal TFE3 to promote clonotypic growth of melanoma cells suggest that overexpression of the TFE3 protein sequence is a critical oncogenic force (2). Moreover, TFE3, and to a lesser extent, TFEB, have been implicated in several cytokine signaling pathways that control cell growth and differentiation, but the precise mechanisms by which their dysregulation contributes to renal oncogenesis are not clear.
TFE3 and TFEB are closely related members of the Mi/TFE3 (MiT) transcription factor family that includes TFEC and the microphthalmia (mi) transcription factor Mitf (6). TFE3 or TFEB overexpression is predicted to sabotage proper regulation of MiT family target genes that control normal growth and differentiation. MiT proteins bind the same cognate E3 DNA sequence (CANNTG) via nearly identical basic regions that requires homo-or heterodimer formation between MiT family members mediated by conserved helix-loop-helix and leucine zipper domains (7). As an essential transcriptional effector of the c-Kit pathway, Mitf is critical for mast cell and melanocyte development (8 -10). Mitf expression can be deregulated in melanoma (11), and its ectopic overexpression in fibroblasts activates melanocyte-specific genes (12). Consistent with the prediction, tumors overexpressing TFEB or a subset of TFE3 fusion proteins express some melanocyte markers (1,2), which may be a reflection of known functional similarities. TFE3, like Mitf, can synergize with the Wnt transcriptional effector LEF-1 to activate melanocyte-specific promoters (13), and overexpression of Mitf or TFE3 rescues growth of melanoma cells in which the Wnt pathway is blocked (2). TFE3 or Mitf is also required for osteoclast development because they are redundant transcriptional mediators of the m-Csf pathway (14,15). c-Kit-and m-Csf-dependent MAPK phosphorylation transcriptionally activates both TFE3 and Mitf, which is necessary for their developmental functions. In addition, TFE3 has distinct activities not known to be shared by other MiT members. TFE3 synergizes with Smad3 to induce expression of the TGF␤ signaling inhibitor Smad 7 (16) and other TGF␤-responsive genes encoding extracellular matrix proteins (17,18). TFE3 complexes with E2F3 to induce p68, a subunit of DNA polymerase ␣, in response to serum (19). Thus, several potentially important target genes and growth regulatory pathways could be affected by deregulated TFE3 activity.
Nevertheless, it remains unclear why TFE3 and TFEB mutations are found almost exclusively in RCC. Although the TGF␤ and Wnt pathways are important for normal kidney development, it is not straightforward to extrapolate paradigms for the relationship of TFE3 to those pathways established in other cell types to renal cells or their transformation given the known gene-and cell type-specific activities of the different MiT proteins (e.g. 20). For example, whereas Mitf overexpression induced melanocyte-specific genes in fibroblasts, TFE3 did not (12), and RCCs containing TFE3 fusion proteins do not always express melanocyte markers (1). Moreover, much of the essential biology of TFE3 and TFEB remains to be elucidated. TFEB-deficient embryos die at embryonic day E10 because of a failure of placental vascularization (21), but its role in the adult is not known, and in general the biological relationship of TFEB to pathways in which TFE3 and/or Mitf participates is not clear. In contrast, germ line TFE3-deficient mice are normal and fertile without any reported defects (14). However, given their extensive sequence similarities, there is likely functional redundancy between TFE3 and TFEB in some tissues where they are both expressed; if so, their contribution could only be revealed by inactivation of both genes.
The goal of the present study was to determine whether additional and common activities of TFE3 and TFEB could be identified which might elucidate new aspects of their essential biology and consequently help an understanding of the basis for their common involvement in renal malignancy. Here, we show that TFE3 and TFEB are cell type-specific leukemia inhibitory factor (LIF)-responsive transcription factors that directly activate E-cadherin expression in fibroblasts. In contrast, E-cadherin expression in epithelial cell lines was not dependent on endogenous TFE3 or TFEB and was reduced by TFE3 or TFEB overexpression. When overexpressed in transformed 3T3 fibroblasts, TFE3 and TFEB also activated WT1 expression and decreased expression of the E-cadherin repressor Snail. These studies reveal the existence of a new, cell type-specific receptortarget gene pathway in which TFE3 and TFEB are involved and which may become perturbed by TFE3 and TFEB overexpression in malignancy.

MATERIALS AND METHODS
cDNA Construction and Expression Vectors-Cloning of full-length mouse TFEB, TFE3, and transdominant negative (TDN) cDNAs are described in Supplemental Data S1. A schematic illustration of these proteins is shown in Supplementary Data S2. cDNAs were subcloned into the eukaryotic plasmid expression vector pEBB (22) and retroviral vector MIG (23).
Transient Transfection and Reporter Gene Assays-HEK293 cells were transiently transfected by calcium phosphate precipitation. 3T3, MEFs, mIMCD-3, and MDCK cells were transiently transfected with SuperFect transfection reagent (Qiagen). The E-pal wt and E-pal mut luciferase reporter plasmids were derived from pGL2 (24); the E3 wt and E3 mut plasmids were derived from pGL3 (25). Both plasmid sets contain the same E-cadherin promoter fragment (mouse Ϫ178 to ϩ92) cloned upstream from the luciferase gene. Firefly luciferase (Luc) and Renilla reniformis luciferase (RLluc) activities were measured from cell extracts with the Dual Luciferase Reporter Assay System (Promega) and TD-20/20 Luminometer (Turner Designs) 48 h after transfection. Luciferase activity was always normalized to RLluc activity. In all experiments, the total amount of pEBB expression vector DNA was the same by balancing cDNA-containing pEBB with empty pEBB. For all luciferase data, n Ն 3, and standard errors are shown.
Retroviral Infection-Retroviral supernatants were prepared by recovering the medium from HEK293 cells 48 -72 h after transfection with MIG viral constructs and the ecotropic helper plasmid pECO; cells were infected following Pui et al. (23) and kept for 24 -48 h before assay.
Induction of E-cadherin after Replating Epithelial Cells-Subconfluent mIMCD-3 cells were infected with retroviruses as described above. Just prior to reaching confluence (at 48 -72 h), cells were treated with trypsin-EDTA and harvested for replating. A small aliquot of cells was taken for the 0 time point, and 2 ϫ 10 6 , 1 ϫ 10 6 , and 0.5 ϫ 10 6 collected cells were seeded onto 60-mm dishes for collection 4, 6, and 24 h later, respectively. Lysates were prepared and equal amounts of protein loaded onto each lane of a 10% SDS-polyacrylamide gel.
Cell Staining and Confocal Microscopy-The protocol of cell staining for immunofluorescence microscopy (BD Biosciences) was followed. Cells were incubated with 15 g/ml biotinylated anti-mouse E-cadherin antibody (R&D Systems, BAF748) for 1 h at room temperature, then incubated with streptavidin-PE (BD Biosciences) or streptavidin-rhodamine (gift of Dr. S. Mirra, SUNY-Downstate Medical Center), added with the Slow-Fade Antifade Kit (Molecular Probes) before processing with the MRC-1024 krypton/argon confocal imaging system (SUNY-Downstate Medical Center Confocal Facility).
Cell Cycle Analysis-2 ϫ 10 6 cells were harvested and washed twice in cold phosphate-buffered saline, and cell pellets were vortexed while ice-cold 70% ethanol was added and were left at 4°C overnight to fix. Cells were pelleted, resuspended in propidium iodide staining solution (50 g/ml propidium iodide and 100 units/ml RNase in phosphatebuffered saline), and rocked for Ն30 min at room temperature before flow cytometry analysis (FACScan, BD Biosciences).
Chromatin Immunoprecipitation (ChIP) Analyses-ChIP analyses were performed with modifications of previously published methods (26). Nuclear lysate was prepared from 1 ϫ 10 8 cells and then sonicated on ice to shear chromatin into an average length of 500 -700 bp (Branson 450 digital sonifier). Chromatin was diluted with immunoprecipitation buffer (2% Triton X-100, 300 mM NaCl, 20 mM Tris, pH 7.4, 2 mM EDTA, 2 mM EGTA, pH 8.0, 0.4 mM sodium orthovanadate, and protease inhibitor mixture), precleared with protein A/G-agarose (Roche Applied Science), incubated with 2 g of each antibody: anti-TFE3 (BD Biosciences), anti-TFEB (abcam), and control anti-IL7 receptor (BD Biosciences) with protease inhibitors at 4°C overnight, and then incubated with protein A/G-agarose for 1 h. DNA recovered from immunoprecipitates was purified with a QIAquick PCR purification kit (Qiagen) and eluted in 30 l of Tris-Cl, pH 8.0, for semiquantitative PCR. PCR conditions were within the linear range of product amplification (not shown). PCR products were electrophoresed on 5% polyacrylamide gels. Each experiment was performed at least three times, and representative data are shown. Oligonucleotides for ChIP assays are descried in Supplemental Data S1.
Stem-Loop RNA Interference (slRNAi)-TFE3 and TFEB mRNA target sequences for RNAi were based on published TFE3 and TFEB sequences and were designed by Qiagen (Supplemental Data S1). Double-stranded oligonucleotides encoding the slRNAi were inserted into pLentiLox 3.7 plasmid (27).
Reverse Transcription (RT)-PCR-Cells were grown on plates until confluent. Total RNA was extracted from 2-5 ϫ 10 7 cells with TRI reagent (Molecular Research Center, Inc.). RT was performed with the StrataScript First Strand Synthesis System (Stratagene) according to the manufacturer's specifications. Primers are listed in Supplemental Data S1. RT-PCR analyses were performed for at least three separate experiments, and representative data are shown in the figures.

TFE3 and TFEB Activate E-cadherin Expression in 3T3
Cells-To understand better normal TFE3 and TFEB function and how their deregulated expression could contribute to malignancy, we sought to identify new cellular processes and target genes regulated by both TFE3 and TFEB in different cell types. The strategy was to focus on the short term consequences of overexpression of exogenous and inhibition of endogenous TFE3 and TFEB, rather than create stable cell lines, so as to avoid a prolonged selection period in which the more immediate effects of TFE3 and/or TFEB overexpression or inactivation could be lost. To accomplish this, TFE3 and TFEB were expressed by retroviruses that also expressed the marker green fluorescence protein (GFP) so that infected cells could be analyzed by microscopy and flow cytometry. For comparison, a truncated TFE3 allele ⌬NTFE3 (28) was analyzed in parallel; ⌬NTFE3 lacks a MAPK phosphorylation site necessary for transcription activation in certain contexts (29 -32) and resembles the portion of TFE3 which remains in many of the TFE3 fusion proteins.
Strikingly, 3T3 fibroblasts transduced with GFP-retroviruses that expressed TFE3, TFEB, or ⌬NTFE3 underwent dramatic morphological changes within 24 -48 h of infection (Fig.  1A). Cells became larger and multinucleated and had a Ն4N DNA content (Fig. 1A, DNA Content). At that time, retroviral expression increased TFE3 and TFEB levels between 20-and 40-fold above endogenous at 1-2 days postinfection (Fig. 1B), levels comparable with those reported in RCCs with TFE3 or TFEB translocations (1). In contrast, 3T3 cells infected with the control virus MIG expressing only GFP retained fibroblast morphology indistinguishable from uninfected cells.
The change in morphology was reminiscent of terminal osteoclast differentiation, a Mitf-and TFE3-dependent process in which multinucleation via cell fusion depends on E-cadherin expression (25,33). In cell culture, exogenous expression of wt Mitf was shown to induce E-cadherin in Mitf-defective mi/mi osteoclast precursors (25). Reporter gene studies showed that Mitf directly activated E-cadherin promoter via a cognate E3 site (25). Although TFE3 was not tested previously for this property, we reasoned that TFE3 and TFEB exhibited this property in 3T3 cells. Indeed, immunofluorescence microscopy revealed that 3T3 cells infected with the TFE3, TFEB, or ⌬NTFE3 virus expressed E-cadherin, whereas control infected cells did not (Fig. 1A). Western blot analysis confirmed Ecadherin expression in these cells (Fig. 1C).
We then used the E-cadherin reporter gene assay to determine whether exogenous TFE3 and TFEB were directly activating the E-cadherin promoter (24,25). The reporter plasmid carried a wt mouse E-cadherin promoter fragment (Ϫ178 to ϩ92) containing critical positive (including the E3 and WT1 sites) and negative (E-pal site) regulatory elements linked to luciferase and was cotransfected with TFE3, TFEB, or control expression plasmids. Using this strategy, exogenous TFE3, TFEB, and ⌬NTFE3 were all shown to activate the E-cadherin promoter in a dose-dependent manner in 3T3 cells (Fig. 1D, panel 1). In the reporter assay, TFE3 and TFEB exhibited similar levels of activity, whereas the induction of endogenous E-cadherin was more responsive to exogenous TFE3 than TFEB (Fig. 1C); the basis for this difference remains to be determined. Nevertheless, the ability of both proteins to activate the promoter was attenuated by E3 site point mutations known to abrogate Mitf binding and activation (25; Fig. 1D, panel 2). In contrast, point mutation of the E-pal site that blocked repressor binding had no effect on the ability of TFE3 and TFEB proteins to activate the E-cadherin promoter (24; Fig. 1D, panel 3). These results demonstrate the independent ability of TFE3 to activate directly E-cadherin inferred from genetic studies of Mitf and reveal this as a new transcriptional activity of TFEB.
In 3T3 Cells, TFE3 and TFEB Overexpression Also Induced WT1-Although the E3 site mutation significantly reduced the degree of activation of the E-cadherin promoter by TFE3 and TFEB, some activity remained, albeit at lower levels. One possibility to account for this remaining activity was that, in addition to acting directly via the E3 site, overexpressed TFE3 and/or TFEB also affected the expression of other Ecadherin regulators. These include WT1, an E-cadherin activator not expressed in 3T3 cells (34), and Snail, a repressor that is expressed in 3T3 cells (24,35). Snail and other negative regulators bind the E-pal site (24,(35)(36)(37).
Interestingly, RT-PCR analysis of mRNA prepared from infected cells showed that retrovirally expressed TFE3, TFEB, and ⌬NTFE3 induced WT1 mRNA in 3T3 cells up from undetectable levels ( Fig. 2A). Induction of WT1 protein was confirmed by Western blot analysis (Fig. 2B). In contrast, steadystate Snail expression decreased ( Fig. 2A). No effect was observed on GAPDH, which served as a control ( Fig. 2A). In all samples, 85-100% of cells were infected with each construct as determined by GFP expression (not shown). Thus, TFE3 and TFEB overexpression each affected the relative expression levels of these other regulators of E-cadherin in 3T3 cells in a manner consistent with E-cadherin activation.

Inhibition of Endogenous TFE3 and TFEB Decreases Endogenous E-cadherin Expression in a Subset of E-cadherin-express-
ing Cells-Although TFE3 and TFEB directly activated the E-cadherin promoter when overexpressed in 3T3 cells, the next important goal was to determine whether endogenous TFE3 and/or TFEB contributed to the expression of endogenous Ecadherin in other cells that normally express E-cadherin under physiological conditions. For this analysis, we compared three E-cadherin-expressing cells: multipotent MEFs from wt and TFE3Ϫ/Ϫ mice (14); HEK293 cells, which were chosen because they exhibit fibroblast and some epithelial characteristics (38); and mIMCD-3 cells, a murine renal epithelial cell line established from the inner medullary collecting duct (39), which express the highest amount of E-cadherin (Fig. 3A). Endogenous TFE3 protein levels were highest in MEFs and 3T3 cells, low in mIMDC-3 cells, and undetectable in HEK293 cells and TFE3Ϫ/Ϫ MEFs (Fig. 3A). In contrast, all cell types expressed TFEB, with highest levels in HEK293 cells. MDCK cells, an E-cadherin-expressing canine kidney epithelial cell line used in some experiments, also expressed polypeptides that corresponded to TFE3 and TFEB proteins (Supplemental Data S3).
Two methods were used to inactivate endogenous TFE3 and TFEB in cells: expression of a TDN protein and slRNAi. The TDN is derived from a portion of TFE3 which contains the dimerization motifs but lacks the basic region and thus blocks MiT family activity by forming inactive heterodimers (Supplemental Data S2). Although the TDN blocks both TFE3 and TFEB, the slRNAi is engineered specifically to knock down expression of each individual protein. The efficacy of the TDN was established in HEK293 cells via biochemical and reporter gene assays (Supplemental Data S2) and in 3T3 cells by assaying its effect on endogenous Smad7, a known target for TFE3 in other cell types (16). Steady-state Smad7 RNA levels dropped dramatically in 3T3 cells expressing the TDN ( Fig. 2A), whereas it had no effect on the expression of WT1 or Snail genes (Fig. 2, A and B). These results are consistent with the Smad7 gene being a bona fide TFE3 target in 3T3 cells and show the efficacy of the TDN approach.
In MEFs, TDN-mediated inactivation of TFE3 and TFEB and slRNAi-mediated knock-down of TFE3 resulted in a decrease of endogenous E-cadherin (TDN, Fig. 3B, lanes 1 and 2; slRNAi, Fig. 3C, lanes 1-3). Importantly, a decrease in endogenous E-cadherin was observed in TFE3Ϫ/Ϫ MEFs only after introduction of TFEB slRNAi but not in the TFE3 slRNAitransfected control (Fig. 3C, lanes 4 -6), demonstrating the ability of TFEB to contribute independently to E-cadherin ex-  2N and 4N indicate peaks corresponding to cells in G 1 /G 0 and G 2 /M, respectively, with N being haploid DNA content. Numbers shown are the percent of total GFPϩ cells with Ն4N DNA content. B and C, Western blot analysis of extracts prepared from cells infected with the indicated retroviruses listed in A, for TFE3 and TFEB (B) and E-cadherin expression (C). In B, extracts from 3T3 cells infected with TFE3 or TFEB MIG retrovirus were serially diluted into extraction buffer (dilution indicated above each lane) to approximate the degree of TFE3 and TFEB overexpression compared with the endogenous proteins (MIG). In all cases Ն85% cell in the culture were GFPϩ (not shown), and GAPDH expression was used as a loading control. Data shown are representative of at least three independent infections. D, luciferase reporter gene assays that show TFE3 and TFEB activate the E-cadherin promoter via the E3 site. Panels 1-3, bar graphs plotting luciferase values from 3T3 cells transfected with an E-cadherin promoter reporter plus the indicated pEBB expression vector. Panel 1, 3T3 cells transfected with 2.5 g of wt E-cadherin promoter reporter plus the indicated amount of pEBB-TFE3 or TFEB expression vector. The total amount of pEBB expression vector in each sample was 5 g, balanced with empty pEBB. Panels 2 and 3, 3T3 cells transfected with 2.5 g of wt (black bars) or the indicated mutant E-cadherin reporters (white bars) plus 5 g of the indicated pEBB-TFE3 or TFEB vector. Panel 2, the E3 mut reporter contains point mutations of the E3 site in the E-cadherin promoter that abrogates Mitf binding (25). Panel 3, the E-pal mut contains a mutation of the E-Pal site that abrogates Snail and Slug binding (24,35). In all cases a control vector expressing Renilla was cotransfected, and luciferase values were normalized and expressed relative to Renilla values. Standard error bars are shown in all figures.
pression and the specificity of the slRNAi approach in reducing expression of the target proteins. Similarly, a reduction in endogenous E-cadherin expression was observed in HEK293 cells transiently transfected with a TDN-expressing plasmid vector compared with control plasmid (Fig. 3B, lanes 3 and 4). Mitf and TFE3 are not expressed in HEK293 cells (Ref. 10; this study and data not shown), again suggesting that TFEB was contributing independently to basal E-cadherin expression in those cells. In contrast, neither the TDN nor slRNAi knockdown in mIMCD-3 cells resulted in any consistent change in endogenous E-cadherin protein or steady-state mRNA levels (Fig. 3B, lanes 5 and 6; 3C, lanes 7-9; and data not shown).
Similarly, E-cadherin protein levels in MDCK cells were unaffected by stable expression of the TDN (data not shown). These differences were not the result of differences in cell density or infection/transfection efficiency: 85-100% of cells analyzed contained the TDN-retroviral or slRNAi plasmid vectors in all cases (data not shown) as determined by analysis of GFP expression from the TDN and slRNA vectors, which contained an internal GFP expression cassette as a vector marker. These results indicate that endogenous TFE3 and TFEB proteins have a direct and limiting role in activating endogenous Ecadherin expression in MEFs and that TFEB independently contributes to E-cadherin expression in HEK293 cells. In contrast, steady-state E-cadherin expression in mIMCD-3 and MDCK cells did not depend on endogenous TFE3 and TFEB.
Parallel results were obtained with the E-cadherin luciferase reporter assay. Basal E-cadherin promoter activity in HEK293 cells was reduced dramatically by a mutation of the E3 site, but a mutation of the E-pal site had no effect (Fig. 3D, panel 1). Moreover, transcription activity of the E-pal mutated promoter was as sensitive to repression by the TDN as the wt promoter, whereas the residual activity of the E3 site mutated promoter was insensitive to the TDN. In contrast, TDN expression had no effect on the activity of the wt or E3 site mutated Ecadherin promoters in mIMCD-3 cells or MDCK cells (Fig. 3D,  panels 2 and 3). Although wt and E3 site mutated promoters had comparable activity in mIMCD-3 cells (Fig. 3D, panel 2), the E3 site mutated promoter was 2-fold less active than wt in MDCK cells (Fig. 3D, panel 3); however, this activity was insensitive to TDN expression, suggesting that TFE3 and TFEB were not involved. Thus, in both contexts the promoter reporter system mirrored the TFE3-and TFEB-responsive behavior of endogenous E-cadherin in the corresponding cell types.
Endogenous TFE3 and TFEB Bind to the E-cadherin Promoter in a Cell Type-specific Manner-Consistent with the above findings, ChIP analyses demonstrated that endogenous TFE3 could specifically and preferentially bind to an endogenous E-cadherin promoter fragment that contained the E3 site in MEFs, but not in 3T3 cells, which do not normally express E-cadherin, or in mIMDC-3 cells, in which inhibition of endogenous TFE3 and TFEB had no effect on endogenous Ecadherin expression (Fig. 3E, top panel, anti-TFE3 ChIP, lanes 2-5, compared with control Ab ChIP, lane 6). Similarly, a TFEB antibody could immunoprecipitate the E-cadherin promoter fragment from wt and TFE3Ϫ/Ϫ MEFs above background levels but not from mIMDC-3 cells or 3T3 cells (Fig. 3E, bottom  panel). As an additional control for specificity, the TFE3 antibody failed to immunoprecipitate the E-cadherin promoter fragment above background levels from TFE3Ϫ/Ϫ MEFs or wt MEFs that expressed the TDN protein (Fig. 3E, lanes 1 and 5). In contrast, the p68 promoter (a subunit of DNA polymerase ␣), which is known to be bound by TFE3 (19), could be preferentially immunoprecipitated at equivalent levels from 3T3, wt MEF, and mIMCD-3 extracts (lanes 2-4) compared with TFE3Ϫ/Ϫ MEFs. Therefore, endogenous TFE3 and TFEB each directly binds to the E-cadherin promoter in a cell type-specific manner consistent with their ability to maintain endogenous E-cadherin expression in those cell types. This also illustrates that TFE3 occupancy and activation from cognate sites are promoter context-and cell type-dependent.
Overexpression of TFE3 and TFEB Reduces E-cadherin Promoter Activity in HEK293 Cells and Epithelial Cell Lines-A surprising finding was that exogenous overexpression of TFE3 and TFEB could not activate the E-cadherin promoter in HEK293 cells as overexpression did in 3T3 fibroblasts and MEFs; rather, TFE3 and TFEB overexpression reduced basal promoter activity by ϳ 2-fold (compare Fig. 4A, panel 1, and  Fig. 1C, panels 2 and 3; MEF data are described in a later section; see Fig. 6A). Similarly, exogenous TFE3 or TFEB reduced the activity of the E-cadherin promoter reporter in mIMCD-3 and MCDK cells and in a manner dependent on the E3 site (Fig. 4A, panels 2 and 3). This response was mimicked by endogenous E-cadherin. As shown in Fig. 4B, the induction of endogenous E-cadherin protein after replating trypsinized mIMCD-3 cells was delayed in TFE3-and TFEB-infected cells compared with MIG controls (4 -8 h). Differences in endogenous E-cadherin levels were not the result of differences in cell density, and this effect was not observed with TDN-expressing cells (data not shown).
In light of these data, it was also unexpected that in contrast to the full-length proteins, the artificially truncated TFE3 allele ⌬NTFE3 (28, 40) could activate the E-cadherin promoter in HEK293 cells in a manner still dependent on the E3 site and could be blocked by the TDN (Fig. 4A, panel 1, and Supplemental Data S2). This property also distinguished HEK293 cells from the epithelial cell lines, in which there was no measurable difference in the activity of ⌬NTFE3 and the full-length proteins (Fig. 4A, panels 2 and 3). To determine whether this was ϪRT is the amplification of an aliquot of the MIG RNA sample in which no reverse transcriptase was added. Equivalent controls for all samples were similarly negative for a PCR product. B, Western blot analysis of extracts from 3T3 cells infected with the indicated retroviral expression construct for WT1 (top panel; the specific WT1 band is indicated with an arrow) and GAPDH (lower panel) expression. GAPDH protein was used as a loading control. The asterisk in the WT1 blot indicates a nonspecific band revealed by the anti-WT1 antibody that is present in all cell extracts analyzed and serves as an additional loading control. The Western blot is representative of two independent experiments.

FIG. 3. TFE3 and TFEB directly activate E-cadherin expression under physiological conditions in certain cell types.
A, Western blot analysis of cell extracts for the presence of endogenous TFE3, TFEB, and E-cadherin proteins. Cell types are indicated above each lane and are described in the text. The bands specific for each protein are shown. The two bands in the E-cadherin blot correspond to different to processed forms of the E-cadherin protein; their ratio varies among cell types. B, TDN-mediated inhibition of TFE3 and TFEB decreases E-cadherin expression. The TDN protein, which blocks both TFE3 and TFEB DNA binding activity, is described in detail in Supplemental Data S2 and in the text. Cells (identified above each panel) were either infected (MEFs and mIMCD-3 cells) or transfected (HEK293 cells) with control or TDN-expressing vectors. 2 days later extracts were prepared and analyzed by Western blot for E-cadherin and GAPDH expression. Transfection and infection efficiencies were nearly 100% as determined by flow cytometry of targeted cells for GFP expression (not shown), and data shown are representative of at least three separate experiments. C, slRNAi-mediated repression of endogenous TFE3 or TFEB decreases E-cadherin expression in specific cell types. Wild-type and TFE3Ϫ/Ϫ MEFs and mIMCD-3 cells were transfected with the indicated slRNAi vector. 2-3 days later cell extracts were prepared and analyzed for the indicated protein by Western blotting. In all cases, the transfection efficiency, as measured by flow cytometry for GFP on an aliquot of transfected cells, was nearly 100%. A shorter exposure of the E-cadherin blot of mIMCD-3 cells compared with other samples is shown for comparison. Data shown are representative of at least three separate experiments. D, endogenous TFEB in HEK293 cells activates the E-cadherin promoter via the E3 site. Panels 1-3 are luciferase reporter assays. In each bar graph, the cell type transfected, the E-cadherin exclusively a promoter context-dependent phenomenon or an intrinsic difference in TFE3 and TFEB activity between the cell types, the activity of the proteins was compared with [E3] 4luciferase, an artificial promoter reporter comprising four tandem E3 sites linked to the luciferase gene (41). In 3T3 cells, all proteins had comparable levels of activity and could activate the [E3] 4 reporter above basal by more than 50-fold (Fig. 4C,  panel 1). Similarly, in HEK293 cells all proteins could activate the [E3] 4 reporter, unlike the E-cadherin promoter. However, ⌬NTFE3 was at least 8-fold more active than the full-length proteins (Fig. 4A, panel 1, and Fig. 4C, panel 2) in HEK293 cells compared with 3T3 cells. Interestingly, the ⌬NTFE3 allele was ϳ2-4-fold less active than TFE3 and TFEB in activating the E-cadherin reporter in 3T3 cells than in HEK293 cells (Fig.  1D, panels 1, 2, and 3) and was more dependent on the E3 site (panel 3). Differences in activity were not the result of differences in protein levels because both exogenous TFE3 and ⌬NTFE3 proteins were expressed at equivalent amounts in these systems (data not shown). In contrast, TFE3 and TFEB activity in the epithelial cells was different from in 3T3 cells and HEK293 cells in several respects. In those cells, the overexpressed full-length and truncated proteins only modestly increased [E3] 4 reporter activity (1.5-3-fold), and there was no measurable difference in activity between them (Fig. 4C,  panels 3 and 4). These results again indicate that the activity of exogenous TFE3 and TFEB is cell type-specific and that the manifestation of the TFE3 truncation is promoter context-and cell type-dependent.
Exogenous TFE3, TFEB, and ⌬NTFE3 Differentially Activate the E-cadherin Promoter in 3T3 Cells versus HEK293 Cells promoter linked to luciferase, and the identity of the pEBB expression vector are indicated. In all cases a control vector expressing Renilla was cotransfected, and luciferase values were normalized and expressed relative to Renilla values. Panel 1, HEK293 cells were transfected with 5 g of wt or mut E-cadherin promoters (indicated at the bottom of the graph) plus 5 g of control (empty pEBB vector; black bars) or the pEBB-TDN expression vector (TDN; white bars). The E-pal wt and E3 wt are both wild-type promoters with the identical E-cadherin promoter fragment (mouse Ϫ178 to ϩ92) linked to luciferase but differ in vector backbone (pGL2 versus pGL3; see "Materials and Methods"). The corresponding mut promoters are the same as described in Fig. 1. Panels 2 and 3, mIMCD-3 cells (panel 2) or MDCK cells (panel 3) transfected with 2.5 g of wt (black bars) or E3 site mutated (E3 mut; white bars) E-cadherin reporter plus pEBB control or TDN vectors. E, ChIP assays show that TFE3 and TFEB bind directly to the E-cadherin promoter in MEFs. Chromatin immunoprecipitates were prepared from different cell types with TFE3 or TFEB antibodies and subjected to semiquantitative PCR to determine the relative amounts of DNA fragments derived from the E-cadherin or p68 promoter in each precipitate. Cell types are indicated above each lane. MEF ϩ TDN are MEF cells infected with the TDN retrovirus (nearly 100%); control antibody was anti-IL7R. CHIP, PCR products from immunoprecipitates; Input are amplified from starting material, diluted 10-fold compared with immunoprecipitates. Data shown are representative of at least three separate experiments.

FIG. 4. Overexpression of TFE3 and TFEB represses E-cadherin expression in epithelial cells. A, E-cadherin promoter reporter assays.
Wild-type (black bars) and E3 site mutated (E3 mut; white bars) E-cadherin promoter reporters were cotransfected into the indicated cell line with pEBB expression vectors. Transfections and luciferase assays were performed as in Figs. 1 and 3 Because of Differences in MAPK Pathway Activation-Given that TFE3 and by inference TFEB are MAPK-responsive transcription factors (29 -32), to explain the above findings, we tested the possibility that the MAPK pathway was able to activate overexpressed TFE3 and TFEB in 3T3 cells but was insufficiently active and thereby limiting the activity of TFE3 and TFEB in HEK293 cells. The ⌬NTFE3 protein lacks the MAPK site, and its activity should not be MAPK-dependent.
Consistent with this possibility, addition of the MEK inhibitor PD98059 to 3T3 cells reduced the ability of TFE3 and TFEB to activate the E-cadherin reporter (Fig. 5A, panel 1). This inhibitory effect was not observed with ⌬NTFE3, whose activity was insensitive to the drug, thus suggesting the importance of the MAPK site in mediating these effects. Furthermore, the drug had no effect on blocking the activation of the E3-mutated E-cadherin promoter by exogenous TFE3 and TFEB (Fig. 5A, panel 2). Thus, the remaining TFE3-and TFEBdependent transcriptional activity of the E3 site mutant promoter is insensitive to the drug, suggesting that the E3 site is MAPK-sensitive via TFE3 and TFEB.
If MAPK-dependent activation of TFE3 and TFEB were limiting in HEK293 cells compared with 3T3 cells, one way this could be evident would be in differences in the saturation profiles of exogenous TFE3-and TFEB-dependent activation of the E-cadherin promoter between the two cell types. Indeed, titration of TFE3 or TFEB plasmids indicated that an ϳ20-fold lower ratio of expression plasmid to reporter vector than what was used in previous experiments could reveal a modest (1.5-2-fold) activation of the E-cadherin promoter in HEK293 cells. However, this activation capability plateaued and returned to basal or below basal levels as the ratio increased (Fig. 5B,  panels 1 and 2). In contrast, ⌬NTFE3-dependent activation of the E-cadherin promoter remained dose-dependent over the same range (Fig. 5B, panel 3), suggesting that the MAPK site was responsible. In support of this idea, the MAPK activator PMA stimulated TFE3-and TFEB-dependent activation of the E-cadherin promoter in HEK293 cells (Fig. 5B, panels 1 and 2) compared with control cells treated with PMA, and their ability to activate became dose-dependent like ⌬NTFE3. This effect was blocked by the simultaneous addition of PD98059. In this set of experiments, both PMA and PD98059 treatments reduced basal activity ϳ2-fold compared with untreated, possibly because of secondary cytotoxic effects; results shown are normalized to the 0 g of TFE3 or TFEB (basal) value for each set of four under each treatment. As an additional strategy to activate endogenous MAPK pathways, we cotransfected the v-raf gene (42). Indeed, coexpression of v-raf induced TFE3 and TFEB activation of the E-cadherin promoter in HEK293 cells compared with cells transfected with the control, which was a truncated, deactivated raf allele pFS-raf (42; Fig. 5C, panels 1 and 2). In contrast, ⌬NTFE3 activity was insensitive to v-raf expression (Fig. 5C, panel 3). Western blot analysis of the proteins from transfected HEK293 cells indicated that PMA and v-raf coexpression both increased the levels of a slower migrating form of TFE3 and TFEB, consistent with known serine phosphorylation of the MAPK site ( Fig. 5D and data not shown), whereas the ⌬NTFE3 protein did not change. These results demonstrate that TFE3 and TFEB are MAPK-dependent activators of the E-cadherin promoter and that the Nterminal domain of TFE3 containing a known MAPK phosphorylation site is critical for this regulation. They also show that excess TFE3 and TFEB proteins without sufficient MAPK-dependent activation were inhibitory rather than activating in HEK293 cells.
TFE3 and TFEB Are LIF-responsive Transcription Activators of E-cadherin via the MAPK Pathway in MEFs-Given the relationship among TFE3, TFEB, E-cadherin, and MAPK activation, we tested whether TFE3 and/or TFEB activation of the E-cadherin promoter was cytokine-responsive. LIF was chosen because it is a cytokine important for renal development which promotes the differentiation of metanephric mesenchymal cells into nephric epithelia that express E-cadherin (43). The LIF receptor complex activates STAT3 and MAPK pathways (44). To perform these experiments, LIF-responsive MEF cells were used (45). Cells were deprived of serum for 12 h prior to LIF and/or drug treatment and then harvested for analysis of Ecadherin expression 24 -48 h later. Under conditions of serum starvation, the basal activity of the E-cadherin promoter was insensitive to the TDN protein as measured by the luciferase reporter assay (Fig. 6A). In contrast, treatment of cells with exogenous LIF activated the E-cadherin promoter 2-3-fold. This effect was blocked by the expression of the TDN protein, suggesting that endogenous TFE3 and TFEB were important in this response. Both exogenous TFE3 and TFEB could independently activate the E-cadherin promoter reporter (Fig. 6A,  panel 1). Importantly, their activity increased when cells were treated simultaneously with LIF. The LIF-dependent activation of TFE3 and TFEB was attenuated by the addition of PD98059, and under these conditions TFE3 and TFEB had the same activity as ⌬NTFE3. Consistent with this, ⌬NTFE3-dependent activation of the E-cadherin reporter was insensitive to LIF. In this assay, the activity of the truncated protein was less than the full-length on activating the E-cadherin promoter (Fig. 6A, panel 1); however, the two had indistinguishable activities on the [E3 4 ] promoter in the presence of serum, similar to what was observed in 3T3 cells (Fig. 6A, panel 2), further suggesting that the impact of TFE3 truncation is promoter context-dependent.
Some of these effects were mirrored by endogenous E-cadherin expression. Under serum starvation, no change in Ecadherin protein levels was detected in control, TDN, or ⌬NTFE3-infected MEF cells with or without endogenous LIF over the duration of cell culture (Fig. 6B, lanes 1-6 and 13-15). In contrast to 3T3 cells and unlike the luciferase reporter system, exogenous TFE3 and TFEB did not induce endogenous E-cadherin under serum starvation. However, endogenous Ecadherin increased ϳ3-fold by LIF only in cells expressing retroviral TFE3 or TFEB (Fig. 6B, lanes 7, 8, 10, and 11). This effect was sensitive to MAPK inhibition because it was modestly but reproducibly reduced by PD98059 (Fig. 6B, lanes 9  and 12). Consistent with this, the E-cadherin response to ⌬NTFE3 was insensitive to the drug and LIF. Taken together, these results indicate TFE3 and TFEB are LIF-responsive activators of endogenous E-cadherin gene expression and again suggest that overexpression alone is not sufficient to activate endogenous target promoters in all cell types.

DISCUSSION
Our data support the existence of a new, cell type-specific receptor-target gene signaling pathway in which TFE3 and TFEB are direct, LIF-responsive, and MAPK-dependent activators of E-cadherin. Aspects of this pathway identified in MEFs were mimicked in transformed 3T3 fibroblasts, in which overexpression of TFE3 or TFEB activated endogenous and reporter E-cadherin expression, and in HEK293 cells, considered intermediate between fibroblasts and epithelial cells, in which inhibition of endogenous TFEB depressed endogenous E-cadherin expression. In these contexts, TFE3and TFEB-dependent activation of E-cadherin was responsive to MAPK pathway activity and was mediated primarily or exclusively through a cognate E3 binding site in the E-cadherin promoter. This pathway was evidently not operational in the epithelial cells, in which TFE3 and TFEB overexpression reduced E-cadherin promoter activity and in which the endogenous proteins were not required for endogenous E-cadherin expression. Consistent with this latter observation, the E3 site did not contribute significantly to E-cadherin promoter activity in a panel of transformed epithelial cells from various tissues (46), suggesting that this may be a general property of epithelial cells.
MAPK-dependent activation of TFE3 and TFEB was an important and cell type-specific determinant of TFE3 and TFEB activity in our system. Phosphorylation of the N-terminal MAPK target site in TFE3, which is conserved in TFEB, allows binding and recruitment of p300/HAT chromatin remodeling complexes to target promoters (30,31,47). Consistent with this, the MAPK pathway stimulated TFE3 and TFEB activity in fibroblasts and HEK293 cells, and deletion of the N-terminal domain that contained the MAPK phosphorylation site rendered TFE3 unresponsive to MAPK activity (Figs. 5 and 6). Interestingly, the reporter gene assays showed that the truncation had different effects in each cell type: it was an activating mutation in HEK293 cells (Fig. 4, A and C), whereas the truncated protein was less active than correspondingly overexpressed full-length TFE3 in fibroblasts (Figs. 1D and 6A). The effect of truncation in fibroblasts was promoter-dependent, evident with the E-cadherin promoter but not with the artificial E3 site promoter (Figs. 1D and 4C). In contrast, overexpressed TFE3 activity in the epithelial cells was indifferent to this truncation and in a manner irrespective of promoter context (Fig. 4, A and C). These observations suggest that additional factors other than MAPK activation also influence the ability of TFE3 (and likely TFEB) to transactivate depending on the cell type, although the responsible factors are not known.
Cell type-specific differences in activity are properties of other known activators of E-cadherin. TFE3 and TFEB share the ability to activate directly E-cadherin expression with WT1 and the Rb-c-myc-AP-2 complex (34,48), each of which binds to the E-cadherin promoter via a discrete cognate site. However, whereas TFE3, TFEB, and WT1 can active E-cadherin expression in 3T3 cells (34), the Rb-c-myc-AP2 complex could only coordinately activate the E-cadherin promoter in epithelial cells but not in 3T3 cells, as revealed in cell culture systems (48). Interestingly, in 3T3 cells, exogenous TFE3 and TFEB were also shown to induce the expression of WT1 (Fig. 2). This scenario is reminiscent of c-myc, which can induce expression of AP-2 (48). We speculate that endogenous WT1 induced by TFE3 and TFEB could account for the remaining ability of overexpressed TFE3 and TFEB to activate the mutated Ecadherin promoter lacking the E3 site in 3T3 cells (Fig. 1), although this remains to be determined experimentally.
In addition to induction of E-cadherin and WT1, 3T3 cells overexpressing TFE3 or TFEB expressed less vimentin and Snail and increased ␤-catenin ( Fig. 2 and data not shown). These types of changes are consistent with cells undergoing a mesenchymal-to-epithelial transition, as was reported for 3T3 cells that ectopically expressed WT1 (49). Therefore, WT1 expression induced by ectopic TFE3 and TFEB is consistent and likely responsible for these other effects, although it is not yet clear to what extent TFE3 and TFEB may directly and independently participate and how their overexpression may affect WT1-dependent changes. Cells overexpressing TFE3 and TFEB did not otherwise resemble epithelial cells in morphology, localization of E-cadherin, and in being multinucleated (Fig. 1). Another possibility is that aspects of an osteoclast terminal differentiation program were induced in 3T3 cells overexpressing TFE3 or TFEB, given that this process requires Mitf or TFE3 and involves direct activation of E-cadherin by those proteins. However, a broader survey of TFE3 and TFEB target genes and the range of cell type-specificity of these activities is necessary to determine the generality of the phenomena reported here and the extent to which TFE3 and TFEB may regulate other genes involved in different cell lineage determination programs.
It is noteworthy that LIF and WT1 have critical roles in the differentiation of mesenchymal cells into tubular epithelial cells of the adult nephron (50). E-cadherin is induced during this differentiation process, and epigenetic loss of E-cadherin expression is a hallmark of tumor progression and metastasis. Dysregulation of TGF␤ responsiveness is associated with renal fibrosis (51). However, it is not yet known whether TFE3 and TFEB are important and/or the pathways defined in this study are normally operational in the developing kidney or other organs: no renal or other intrinsic developmental defects were reported in TFE3Ϫ/Ϫ mice or E10 TFEBϪ/Ϫ embryos (14). However, it is likely TFE3 and TFEB are functionally redundant to each other, and to reveal a contribution would require organ or cell type-specific inactivation of both genes. Moreover, Smad7, like E-cadherin, can be activated by other transcription factors in addition to TFE3, depending on the cell type (e.g. 52,53). Nevertheless, our observations may help explain why superphysiological levels of TFE3 and TFEB caused by translocation could be particularly disruptive to renal tubular epithelial cells and so far found almost exclusively in a subset of RCCs of tubular epithelial cell origin (54). The data suggest that overexpression of TFE3 or TFEB without the concomitant and sufficient activation of signaling pathways such as the MAPK pathway may render the overexpressed molecules in-hibitory in some contexts because the overexpressed proteins lack modifications necessary for the activation of particular target genes. Such an inhibitory activity may adversely affect the differentiation of mesenchymal cells into nephric epithelium or destabilize the epithelial state. Moreover, changes in sensitivity to TGF␤ caused by TFE3-and TFEB-dependent alterations of Smad7 regulation could compound these effects.
Based on the existing and current data, we favor a pathogenetic model for TFE3 and TFEB mutation in which the combined effects of overexpression on the activation of genes controlling growth and survival, and the inhibition or destabilization of expression of other genes such as WT1 and E-cadherin, would be responsible for the full oncogenic consequences of TFE3 or TFEB deregulation. Otherwise, as activators of WT1 and E-cadherin, TFE3 and TFEB would be predicted to have a tumor suppressor-like activity, but no phenotype corresponding to that property was reported in TFE3Ϫ/Ϫ or TFEBϩ/Ϫ mice (14). It is still possible that any putative tumor suppressor functions are redundant and would require early inactivation of both genes to reveal them. However, given that ectopic TFE3 expression promoted growth and survival of melanoma cells (2), inappropriate activation of genes promoting cell growth by TFE3 or TFEB overexpression must be a critical part of the transformation process. These effects would not be recouped in strict loss-of-function systems. Taken together, these results now reveal new molecular pathways and target genes responsive to TFE3 and TFEB which may be deregulated by TFE3 and TFEB mutation in renal cells and perhaps common to other RCCs.