SNF5/INI1 Deficiency Redefines Chromatin Remodeling Complex Composition during Tumor Development

This article is featured in Highlights of This Issue, p. 1533

Malignant rhabdoid tumors (MRT) are a highly aggressive pediatric cancer commonly found in the kidneys (RTK) and brain (AT/RT). These tumors have a median onset of 11 months and 80% to 90% of children die from the disease within a year of diagnosis (1). MRTs are characterized by the loss of SNF5/BAF47/INI1/SMARCB1, a core member of the SWI/SNF complex, hereafter referred to as SNF5 (2). Interestingly, SNF5 loss does not affect genetic stability in MRT, but causes epigenetic instability (3). Thus, this consistent and specific molecular change, SNF5 inactivation, provides a unique model for understanding the role of epigenetics in human tumor development. Numerous studies have shown that SNF5 loss affects multiple signaling pathways (4). However, little else is known about the etiology of this cancer.

The SWI/SNF complex has multiple variants and is classified by their mutually exclusive ATPase subunits, BRM and BRG1. BRG1-containing SWI/SNF complexes can be subdivided into two mutually exclusive groups—BAF180/PBRM1- and BAF250A/ARID1A-containing complexes. Although SNF5 appears in all the SWI/SNF complexes, little is known about the effects of SNF5 loss upon SWI/SNF activity. One previous study by Doan and colleagues (5) showed that the SWI/SNF complex appeared intact in MRT cell lines despite the loss of SNF5. They also demonstrated that BRG1-dependent genes did not show altered expression in MRT cell lines (5). However, they did not address the stoichiometry of the complex components in MRT cell lines or the interaction of the complex with chromatin. Other studies have indicated that loss of a SWI/SNF complex member could affect the stability of other complex members in mammalian cells and in flies (6–12). Furthermore, a growing number of studies have identified multiple SWI/SNF complexes defined by the presence or absence of different components (4, 13–15). Importantly, these different complexes can promote either growth or differentiation depending upon their composition (4, 13–15).

To address whether SNF5 loss affects SWI/SNF complex composition during MRT development, we examined the effects of SNF5 expression in MRT cell lines. Our studies demonstrate that reexpression of SNF5 in MRT cell lines altered total protein levels of many components, whereas knockdown of SNF5 in normal human fibroblasts showed the opposite effect without altering mRNA expression. After SNF5 reexpression, many components were recruited into SWI/SNF complexes that are generally associated with induction of differentiation. Therefore, SNF5 loss may promote MRT development by preventing a switch from SWI/SNF complexes that promote proliferation to those associated with differentiation. These studies emphasize the need to resolve the scope and composition of SWI/SNF complexes among different tissues and may account for the low incidence of these deadly childhood cancers.

Our laboratory previously described the A204.1, D98OR, and G401.6 cell lines (16–18). The MCF-7, A673, Jurkat, DAOY, and RD cell lines were obtained from the American Type Tissue Collection (ATCC). The TTC642 and TTC549 cell lines were kindly provided from Dr. Timothy Triche, Children's Hospital of Los Angeles. Dr. Peter Houghton, Nationwide Childrens Hospital, kindly provided the BT-12 AT/RT cell line. 293FT cells were kindly provided by Dr. Inder Verma, Salk Institute. All cell lines were cultured in RPMI-1640 plus 10% fetal bovine serum (Gibco; or Benchmark FBS, Gemini Biosciences) and were used within 30 passages of their initial arrival to minimize the chances of cross-contamination.

The Ad/pAdEasyGFPINI-SV+ adenoviral vectors expressing hSNF5 and coexpressing green fluorescent protein (GFP; designated Ad-hSNF5) and the Ad/pAdEasyGFP expressing GFP (designated Ad-GFP) were previously published (19). To generate adenovirus expressing HA-tagged SNF5, we used the pAdEasy system (20). We first subcloned the insert from the pMDK225 lentivirus (21) expressing SNF5 with a c-terminal triple-HA tag into the pShuttle vector. The University of North Carolina (UNC) Vector Core generated Ad-SNF5-HA following the pAdEasy system protocol. The SNF5-HA was sequenced to verify the wild-type gene and the triple HA tag at the c-terminus. To achieve infection of more than 90% of cells, we infected at a multiplicity of infection (MOI) of 20 for the A204.1 cell line, 10 for the G401 cell lines, 10 for the TTC549 cell line, and 200 for the TTC642 cell line. Infection of cell lines were carried out as previously described (22).

Isolation of nuclear protein and Western blotting were carried out as described previously (16, 23). Western blot analyses for protein expression were carried out using the antibodies listed in Supplementary Table S1. Densitometry was carried out using the Bio-Rad Image Lab 4.1 software.

RNA was extracted using the RNeasy Mini Kit (Qiagen), and 1 μg was used for cDNA synthesis primed with random primers (Invitrogen). cDNA was analyzed using TaqMan (Applied Biosystems) quantitative real-time reverse transcription PCR (qPCR) analysis, with β-actin as the reference gene in each reaction. Reactions were performed on an ABI 7900 HT sequence detection system (Applied Biosystems), and relative quantification was determined using the 2^−ΔΔCt method (Chai and colleagues, ref. 16). The TaqMan gene expression assay primer/probe sets used in these studies are listed in Supplementary Table S2.

Cells were infected with adenovirus expressing either HA-tagged SNF5 (Ad-SNF5-HA) or empty vector (Ad-CMV). Proteins were isolated from the infected cells using IP buffer [50 mmol/L Tris, 400 mmol/L NaCl, 2 mmol/L EDTA, 10% Glycerol, 1% NP-40, 0.5% sodium deoxycholate, 0.1 mmol/L PMSF, 4 mmol/L sodium fluoride, 40 nmol/L sodium orthovanadate, and 1× complete Mini protease inhibitor cocktail (Roche Diagnostics)]. The isolated protein was quantified and adjusted to 1 mg/mL. One milligram of protein from each sample was incubated with BRG-1 (A303-877A; Bethyl), or normal rabbit IgG (sc-2027; Santa Cruz Biotechnology) rotating overnight at 4°C with 30 μL of a 50% slurry of protein A/G Sepharose beads. The beads were then washed three times with IP wash buffer (1× PBS, 10% glycerol, and 1% Triton) and then suspended in 1× NuPAGE LDS Loading Buffer supplemented with 125 mmol/L DTT and boiled for 5 minutes. The supernatants were then run on 4% to 12% Bis-Tris polyacrylamide gel, transferred to polyvinylidene difluoride (PVDF) membranes and probed with anti-SNF5, anti-BRG1, anti-BAF155, anti-BAF180, anti-BAF200, anti-BAF250A, anti-BAF170, anti-BAF60A, or anti-BAF57 as described above (Supplementary Table S1).

The determination and quantification of SWI/SNF complex components by mass spectrometry was carried out on immunoprecipitated samples prepared as in the immunoprecipitation protocol above following our previously published protocols (24). The resulting mass spectra were searched against the SwissProt sequence database (downloaded 07/23/13) using MaxQuant (Version 1.4.1.2; ref. 25) with a static carbamidomethyl modification on cysteines, variable oxidation of methionine, variable acetylation of N-termini, and a fully tryptic digestion. MaxQuant's LFQ algorithm was used for label-free quantification. Experiments with and without doxycycline treatment were each performed in biologic duplicate, with each biologic sample analyzed in technical duplicate. Technical replicates' LFQ intensities were averaged. Ratios for each batch of biologic duplicates were determined by the LFQ intensity of the doxycycline-treated sample divided by the LFQ intensity of the untreated control sample.

Lentivirus was generated using 293FT cells following the protocol of Xu and colleagues (26). Either pLKO.1, a nontargeting small hairpin RNA (shRNA) control vector (SHC002; Sigma), or SNF5 shRNA lentiviral transduction particles (TRCN 39585 and 39587) were cotransfected with the packing construct ΔNRF (from Dr. Tal Kafri, University of North Carolina, Chapel Hill, NC; ref. 26) and the VSV-G envelope expression plasmid (pMDK64; from Dr. Matthias Kaeser, Salk Institute) into 293FT cells using calcium phosphate transfection. pLKO.1 is a negative control containing an inert sequence that does not target any human or mouse gene but will activate the RNAi pathway. For infection, cells were incubated with lentiviral particles and polybrene and then selected with puromycin.

Flag-tagged SNF5 was amplified from an existing plasmid, pcDNA3-fSNF5, by PCR (T7 promoter F primer, hSNF5 ORF R primer—TTA CCA GGC CGG CGT GTT; ref. 16). The resulting PCR product was TOPO-cloned into a Gateway vector using the pCR8/GW/TOPO TA Cloning Kit (45-0642; Invitrogen). The fSNF5 was transferred into the pINDUCER20 vector that provided a c-terminal HA tag using Gateway LR Clonase II Plus Enzyme Mix (12538-120; Invitrogen). The identity of the fSNF5HA insert was confirmed by DNA sequencing.

To identify β-galactosidase–positive cells, we used the Senescence β-Galactosidase Staining kit (Cell Signaling Technology; #9860). Cells were prepared and stained following the kit protocol as previously described (9).

To inactivate the proteasome, cells were treated with 10 μmol/L MG132 (1748; Boston Biochem) or DMSO vehicle control for 6 hours. Cells were harvested and prepared lysates were analyzed by immunoblotting, as described above.

We first examined whether component protein levels differed between SNF5-deficient MRT and AT/RT cell lines and SNF5-positive non-MRT cell lines, including pediatric (A673 Ewing sarcoma, DAOY medulloblastoma, and RD rhabdomyosarcoma) and adult (D98OR, a HeLa cell derivative, Jurkat acute T-cell lymphoma, and MCF-7 breast cancer) tumor cell lines. As expected, we found a clear loss of SNF5 protein in the four MRT cell lines and one AT/RT cell line in comparison with non-MRT cancers (Fig. 1A; ref. 23). We also found no expression of BRM/SMARCA2 protein or mRNA, consistent with previous reports (Fig. 1A and B; ref. 27, 28), whereas levels of the other ATPase, BRG1/SMARCA4, appeared similar among all cell lines. Surprisingly, the other component proteins showed generally lower levels in MRTs compared with other pediatric and adult tumor cell lines ranging from nearly complete absence (ARID1A, SMARCC2, and SMARCD2) to moderate reduction (ARID2; Fig. 1A).

We next asked whether the reduced protein levels reflected changes in mRNA expression. Thus, we assessed component mRNA levels using qPCR normalized to the HeLa-derived D98OR cell line, a cell line often used for the purification of SWI/SNF complex members (29, 30). Consistent with the Western blot analysis data, SNF5 mRNA expression was undetectable in MRT cell lines, with the exception of TTC642, a cell line that contains a nonsense mutation in the SNF5 gene (23). In contrast, other than SMARCA2/BRM, we could not observe consistent changes in the mRNAs of other complex components in MRT cell lines that could account for their lower protein levels (Fig. 1B)

To extend these findings to primary MRTs, we examined protein expression in nuclear extracts from a series of primary tumors derived from MRTs as well as other prototypical childhood cancers, including Wilm tumors (WT), and rhabdomyosarcomas (RMS; ref. 23). Because of the limited amount of starting material, we could only assess a subset of SWI/SNF components. Protein levels in these nuclear extracts were normalized to histone H3 protein levels and aggregated by MRT versus non-MRT tumor type. Loss of SNF5 protein again was apparent in the MRTs compared with the other tumor types (Fig. 2A). The average ratio of BAF180 and BAF57 to H3 in MRTs was significantly lower than in non-MRT samples, whereas the BRG1 levels remained similar (Fig. 2B). Thus, the primary tumor data for MRTs recapitulate the findings derived from cell lines.

We next examined the effects of SNF5 reexpression on SWI/SNF complex members in MRT cell lines. To carry out these experiments, we used adenovirus to reexpress SNF5 in three representative MRT cell lines. We focused on the MRT cell lines, G401, TTC549, and TTC642 and omitted the soft tissue–derived A204 to maintain our focus on renal-derived MRTs. We examined SWI/SNF complex component protein expression by Western blotting and mRNA levels by qPCR before and at 24 hours after infection with adenovirus. As shown in Fig. 3A, we observed a consistent increase in the protein level BAF180 as a result of SNF5 reexpression in all MRT cell lines. In contrast, BRG1 and BAF200 protein levels did not change significantly after SNF5 reexpression. Other complex members showed differential changes, with BAF155 increasing in the G401 and TTC549 cell lines, BAF250A showing an increase in the TTC549 cell line, and BAF60A and BAF57 showing increases in the TTC642 cell line. These findings seem consistent with the observations in primary tumors (Fig. 2A) in which SNF5-negative MRTs demonstrated decreased BAF180 and BAF57 levels but not BRG1 in comparison with WT and RMS. Similar to the results in Fig. 1A and B, changes in protein levels did not result from altered mRNA expression (Fig. 3B), consistent with posttranscriptional regulation.

Our data thus far demonstrated that global SWI/SNF complex component levels increase following SNF5 overexpression. Next, we determined if this global change is mirrored within the SWI/SNF complex. To carry out these experiments, we developed a tet-inducible SNF5 vector (pIND20-fSNF5-HA) using the pINDUCER system (31) to develop MRT cell lines with inducible SNF5 expression (Fig. 4). As shown in Fig. 4, the tet-inducible cell lines validated the results from the adenovirus infections studies in the previous section (Fig. 3A).

To determine the composition of the SWI/SNF complex in the presence and absence of SNF5, we initially carried out label-free quantitative mass spectrometry on the TTC642 cell line. We isolated SWI/SNF complexes from this cell line ± SNF5 from whole-cell lysates by immunoprecipitation with an antibody against BRG1. Because the MRT cell lines used in this study do not express the mutually exclusive BRM ATPase (Fig. 1A), functional complexes must possess the BRG1 ATPase. As shown in Fig. 5A, reexpression of SNF5 resulted in significant changes in the components associated with BRG1.

To confirm the semi-quantitative mass spectrometry results, we assessed individual component protein levels by Western blotting after immunoprecipitation with our BRG1 antibody. Because BRG1/SMARCA4 levels did not change or are decreased after SNF5 reexpression (Figs. 3 and 4), we used the ratio of complex subunits to BRG1 to measure changes in component levels within the complex. Whole-cell protein extracts from MRT cell lines, G401, TTC549, and TTC642 cells, either untreated or induced with doxycycline for 24 hours, were harvested, immunoprecipitated with α-IgG or α-BRG1 antibodies, and Western blotted to determine the protein levels of BRG1 and other complex members.

Whereas the overall levels of most component proteins increased, BRG1 levels either remained the same or decreased after SNF5 reexpression (compare input lanes in each cell line in Fig. 5B). Furthermore, without the confounding issue of viral infection, we found increased expression of BAF57 and BAF170 in all cell lines. As shown in Fig. 5B, the ratio of most component proteins to BRG1 increased in all MRT cell lines after SNF5 reexpression. Surprisingly, we could not detect several components, such as BAF250A/ARID1A and BAF170/SMARCC2, in the complex in the absence of SNF5. Other components showed a differential pattern among the cell lines. For example, the complexes in the TTC549 and TTC642 cell lines contain BAF200/ARID2 after SNF5 reexpression, whereas it is absent in complexes from the G401 cell line. On the basis of these data, the increased component protein levels observed in MRT cell lines after SNF5 reexpression coincides with increased association with SWI/SNF complex members.

Although our studies indicated that restoration of SNF5 in MRT cell lines increased expression of SWI/SNF complex components, they did not address whether SNF5 loss in normal cells would lead to decreased protein levels. To address this issue, we infected normal human fibroblasts (NHF-1) with two different shRNAs against SNF5 and assessed component mRNA and protein levels. Because the cell of origin for MRTs remains unknown, we used NHFs as a well-characterized normal human cell line model. As shown in Fig. 6A, we achieved an approximately 80% decrease in SNF5 expression 48 hours after lentiviral infection with two different shRNAs. The SNF5 reduction corresponded to decreases in all SWI/SNF complex components with the exception of BRG1, ARID200, and BAF155. The nontargeting shRNA control, pLKO.1, did not show any effect with the exception of a moderate decrease in ARID1A levels. Thus, the decreases in component protein levels after SNF5 loss correlated well the opposite changes observed in the MRT cell lines after SNF5 reexpression. However, similar to our results with the MRT cell lines, the change in component protein levels did not mirror its mRNA expression (Supplementary Fig. S1). However, using β-galactosidase staining, we observed replicative senescence only in the cells with SNF5 knockdown (Supplementary Fig. S2).

The levels of many cellular proteins are regulated through ubiquitination and subsequent degradation via the proteasome machinery (32). To assess whether a posttranscriptional mechanism might contribute to changes in component levels in the MRT cell lines, we examined the effects of MG132, a potent proteasome inhibitor, on protein expression in the TTC642, G401, and TTC549 cell lines (33). We treated the cell lines with DMSO or 10 μmol/L MG132. After 6 hours, we assessed representative SWI/SNF complex members as well as c-FOS (positive control) and RAN (negative control/loading control) protein levels by Western blotting. As shown in Fig. 7, MG132 treatment increased protein levels of many components as well as or better than SNF5 reexpression, suggesting posttranscriptional regulation.

We then asked whether MG132 treatment would restore expression of SWI/SNF components after SNF5 knockdown in NHF-1 hTERT cells. The pLKO.1 nontargeting lentivirus served as a negative control. We used the hTERT-expressing cells because they abrogate the problem of the limited lifespan of normal human fibroblasts in culture. In general, we saw good agreement between the effects of SNF5 knockdown (SNF5-KD) on NHF-1 and NHF-1 hTERT (Fig. 6). However, some differences appeared, including the levels of ARID2, BAF155, and BAF60A. We again saw differential effects on protein stability after MG132 treatment with some components increasing (BAF180, BAF60B, and BAF57), while BAF60A demonstrated decreased expression (Fig. 6B). BRG1 levels were unaffected, consistent with the previous results, while BAF250A was not detected. These results implicate a proteasome-dependent mechanism in lowering some component protein levels after SNF5 knockdown.

Previous reports have shown the effects of loss of individual SWI/SNF complex members upon the organization and stability of the remaining complex components. For example, studies by Chen and Archer (6) posited that BAF155 serves as a scaffolding element required for stability of other SWI/SNF complex, including BRG1 (34) and BAF57. Another report demonstrated that ARID2/BAF200 regulated BAF180 expression at the level of mRNA expression (35). Other reports have shown increased expression of BRM upon loss of BRG1 expression (36–38). Recent evidence has also shown decreased BAF57 and SNF5 levels in BAF155-deficient cell lines (9). Only a few studies have addressed the effects on complex formation in the absence of different complex members (5). Doan and colleagues previously showed assembly of the complex in the absence of SNF5 with no effect on expression of BRG1-dependent genes (5). Our data demonstrate increases in SWI/SNF complex members binding to the complex after SNF5 ectopic expression, implying that SNF5 plays a role in the stability of the complex members but not the corresponding mRNA. This role does not contradict prior research, but further expands the functions of SNF5 within the SWI/SNF complex.

We have previously shown that SNF5 reexpression in MRT cell lines causes increased p21^WAF1/CIP1 expression associated with preferential recruitment of SNF5 to the p21 transcription start site (TSS; ref. 39). This information, in the context of our data, suggests that the constant presence of BRG1 containing complexes lacking SNF5 along the p21 promoter maintains transcription but cannot remodel nucleosomes at the TSS. This obstacle could prevent the elongation of RNA polymerase II, resulting in promoter pausing and minimal p21 expression (40, 41). Similar effects have been observed for the regulation of developmental genes in Drosophila (12, 42). SNF5 reexpression could stabilize the SWI/SNF complex, resulting in targeting to TSSs and increased gene transcription. Alternatively, our results could support a model in which SNF5-deficient complexes are already recruited to TSSs and maintain a basal level of gene expression. However, gene expression levels remain low due to the short half-life of most complex members in the absence of SNF5 or the type of SWI/SNF complex present at the promoter. SNF5 expression would stabilize the complex or recruit a different type of SWI/SNF complex causing an increased gene expression.

Our finding that a subset of SWI/SNF complex components appears degraded in a proteasome-dependent manner seems consistent with previous reports indicating their instability in the absence of other complex members (6, 34). However, SNF5 loss correlates with degradation of multiple complex members, in contrast with the limited numbers affected by loss of other complex components (6, 34, 35). The identity of the proteasome pathway responsible for the degradation of complex components and whether it acts directly or indirectly remains unknown. The observation that BAF60A levels decrease after MG132 treatment in the NHFs supports an indirect mechanism (Fig. 6B). However, our observations seem consistent with previous results that cells maintain tight control over the protein levels of SWI/SNF complex members (36, 43, 44).

The mechanisms by which SNF5 loss initiates MRT development remain unresolved. Recent reports have identified at least nine different forms of the SWI/SNF complex, based upon protein composition, that promote diverse biologic functions, including growth and differentiation (4, 15). Our current study implicates changes in SWI/SNF complex composition after SNF5 inactivation as a mechanism for MRT development. This makes an attractive model because it accounts for several hallmarks of this cancer. Presumably, the transition from a growth promoter configuration of the SWI/SNF complex to a differentiation-inducing one occurs within a narrow window of development. Therefore, SNF5 loss would only exert an effect if it happened within this time frame. This strict requirement for timing could account for the relative paucity of these tumors. Second, if MRTs arise from retention of growth-promoting complexes, affecting gene expression, one would expect little genomic instability in these tumors. In agreement with this notion, a recent report from Lee and colleagues demonstrates a lack of significant changes in MRTs (45). The dramatic effects on ARID1A protein expression in the presence or absence of SNF5 compared with changes in BAF180 suggest that PBAF complexes remain more stable during MRT development than BAF complexes. Finally, while the loss of SNF5 from a stem-like population may prevent differentiation, KD on SNF5 in differentiated cells may result in SWI/SNF complexes with growth-inhibitory properties. This would account for the paradoxical result that SNF5 reexpression in MRT cell lines and SNF5 KD in primary NHFs both cause growth arrest (16, 46, 47).

The changes in gene expression observed after SNF5 reexpression in MRT cell lines or its inactivation in normal cells may arise from differential binding of the SWI/SNF complexes present under each condition (47, 48). Therefore, future ChIP-Seq experiments should identify additional unique and mutual binding sites for SWI/SNF complex members with and without SNF5. These data, in conjunction with gene expression analyses, will allow for further insights into the role of SWI/SNF complex activities after SNF5 inactivation in MRT development. It will also be important to investigate the activities of SWI/SNF complexes that do form in the absence of SNF5. These complexes may account for the observation of Wang and colleagues that lymphoma development after SNF5 inactivation requires BRG1 expression (38).

The existing protocols for treatment of MRT include tumor resection, followed by adjuvant chemotherapy and/or radiation (2). These current protocols suffer from several inadequacies, including the difficulty of resection due to the tumor size and the contraindication of radiation in young patients (2). Understanding the molecular mechanisms behind MRT development and growth will allow for improved patient treatment and survival rate. For example, the posttranscriptional regulation of SWI/SNF complex components suggests that they can be potential targets for therapeutic intervention. Further screening of inhibitors of the 20S proteasome could prove fruitful in finding a drug candidate to stabilize the complex in the absence of SNF5. In addition, identifying the pathways regulated by the SWI/SNF complexes present in the absence of SNF5 should reveal additional therapeutic targets. Furthermore, our results, showing that BAF180 and BAF250 stability inextricably depends upon SNF5's presence, suggest a potential mechanistic link between SNF5 loss in MRT and BAF180 and BAF250A losses in renal cell carcinoma and ovarian carcinomas, respectively. Future studies will further our understanding of chromatin-remodeling functions and provide knowledge for translation into the clinic for improving the outcomes of patients with MRTs and other cancers with SWI/SNF complex mutations.

No potential conflicts of interest were disclosed.

Conception and design: D. Wei, M.B. Major, B.E. Weissman, Y. Kuwahara

Development of methodology: D. Wei, M.B. Major, B.E. Weissman, Y. Kuwahara

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Wei, S. Song, C. Cannon, F. Yan, D. Sakellariou-Thompson, M.B. Major

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Wei, D. Goldfarb, S. Song, D. Sakellariou-Thompson, M. Emanuele, M.B. Major, B.E. Weissman

Writing, review, and/or revision of the manuscript: D. Wei, D. Goldfarb, S. Song, M.B. Major, B.E. Weissman, Y. Kuwahara

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Wei, D. Sakellariou-Thompson, M.B. Major, B.E. Weissman

Study supervision: M.B. Major, B.E. Weissman, Y. Kuwahara

The authors thank the members of the Weissman laboratory for their help, advice, and support, Drs. Ramon Parsons, Dale Ramsden, Jiandie Lin, and Zhuoxian Meng for antibodies, Drs. Dennis Simpson, William Kaufmann, Peter Houghton, Inder Verma, and Tim Triche for cell lines, and Dr. Guang Hu, NIEHS, for the kind gift of the pINDUCER20 vector. The authors also thank the University of North Carolina Vector Core for generation of the adenovirus expressing SNF5-HA.

The studies received financial support, in part, from Lineberger Comprehensive Cancer Center development funds, the Sidney Kimmel Foundation for Cancer Research (to M.B. Major), and the NIH (1-DP2-OD007149-01, to M.B. Major; R01CA91048, to B.E. Weissman; P30 CA016086 and T32ES007126, to D. Wei).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.