An Integrated Genome Screen Identifies the Wnt Signaling Pathway As a Major Target of WT1

An integrated genome screen identifies the Wnt signaling pathway as a major target of WT1

Marianne K.-H. Kima,b, Thomas J. McGarryc, Pilib O´ Broind, Jared M. Flatowb, Aaron A.-J. Goldend, and Jonathan D. Lichta,b,1

aDivision of Hematology/Oncology and cFeinberg Cardiovascular Research Institute, Division of Cardiology, Northwestern University Feinberg School of Medicine, Chicago, IL 60611; bRobert H. Lurie Cancer Center, Northwestern University, Chicago, IL 60611; and dDepartment of Information Technology, National University of Ireland, Galway, Republic of Ireland

Edited by Peter K. Vogt, The Scripps Research Institute, La Jolla, CA, and approved May 18, 2009 (received for review February 12, 2009) WT1, a critical regulator of kidney development, is a tumor suppressor control. A first generation of in vitro cotransfection and DNA for nephroblastoma but in some contexts functions as an oncogene. binding assays have given way to more biologically relevant exper- A limited number of direct transcriptional targets of WT1 have been iments where manipulation of WT1 levels has been accompanied identiﬁed to explain its complex roles in tumorigenesis and organo- by examination of candidate or global gene expression. Classes of genesis. In this study we performed genome-wide screening for direct WT1 targets thus identified consist of inducers of differentiation, WT1 targets, using a combination of ChIP–ChIP and expression arrays. regulators of cell growth and modulators of cell death. WT1 target Promoter regions bound by WT1 were highly G-rich and resembled genes include CDKN1A (22), a negative regulator of the cell cycle; the sites for a number of other widely expressed transcription factors AREG (23), a facilitator of kidney differentiation; WNT4 (24), a such as SP1, MAZ, and ZNF219. Genes directly regulated by WT1 were stimulator of renal development; and SPRY1 (25), a critical intra- implicated in MAPK signaling, axon guidance, and Wnt pathways. cellular regulator of receptor tyrosine kinase signaling and kidney Among directly bound and regulated genes by WT1, nine were development. Examination of gene expression in WT1 null and identiﬁed in the Wnt signaling pathway, suggesting that WT1 mod- WT1 replete Wilms tumor specimens identified a gene (IFI16) not ulates a subset of Wnt components and responsive genes by direct normally regulated or even expressed in the same subcellular binding. To prove the biological importance of the interplay between compartment as WT1 but regulated in a pathological manner in WT1 and Wnt signaling, we showed that WT1 blocked the ability of tumors (21). Wnt8 to induce a secondary body axis during Xenopus embryonic To comprehensively identify WT1 target genes we manipulated development. WT1 inhibited TCF-mediated transcription activated by WT1 levels in a Wilms tumor cell line by conditional overexpression Wnt ligand, wild type and mutant, stabilized ␤-catenin by preventing and shRNA-mediated knockdown. Using gene expression profiling TCF4 loading onto a promoter. This was neither due to direct binding and chromatin immunoprecipitation followed by microarray anal- of WT1 to the TCF binding site nor to interaction between WT1 and TCF4, ysis (ChIP–ChIP) we identified genes bound and regulated by WT1. but by competition of WT1 and TCF4 for CBP. WT1 interference with Wnt WT1 targets included genes of the Wnt pathway, which is required signaling represents an important mode of its action relevant to the for normal renal development (26) and is deregulated in Wilms suppression of tumor growth and guidance of development. tumors (5). WT1 inhibited Wnt function during Xenopus development and interfered with Wnt-mediated transcription through the ChIP–ChIP ͉ microarray ͉ tumor suppressor CREB binding protein (CBP) cofactor. Collectively these data suggest that one critical role of WT1 in development and tumori- he genetic etiology of Wilms tumorigenesis is heterogeneous genesis is to modulate the Wnt signaling pathway. Tincluding loss-of-imprinting of IGF2 (1), deletions and muta- Results tions of WT1 (reviewed in ref. 2) and the recently identified WTX (3) genes. In addition, somatic mutations in ␤-catenin leading to a Identification of Direct Targets of WT1 by ChIP in Combination with stabilized protein are found in 15% of cases and curiously almost Promoter Microarray Analysis. ChIP–ChIP was initially performed all of these mutation cases are found in patients lacking functional in Wilms tumor-derived CCG99–11 cells, which express low levels WT1 alleles (4–7). of wild-type WT1 protein. The resulting hybridization signals were The WT1 tumor suppressor gene encodes a zinc finger transcrip- low and only 8 promoter regions were identified on all 3 NimbleGen tion factor, possibly yielding up to 32 different isoforms (8). The HG18 RefSeq arrays and 57 promoters by 2/3 arrays (Fig. S1A). major isoforms differ in the presence or absence of amino acids Previously identified direct targets of WT1 such as IFI16 (21) and KTS in the zinc finger region and the presence or absence of a 17-aa MKP3 (13) were not scored by this experiment. To increase the stretch in the middle of the protein. The ϪKTS isoforms have been sensitivity of the ChIP–ChIP assay we used CCG-5.1 cells engi- linked to DNA binding-mediated transcriptional control whereas neered to stably express additional WT1-A in an inducible manner the ϩKTS isoforms have been implicated in RNA processing as well (Fig. S1B). In contrast to the apoptotic phenotype we (27) and (9). Renal agenesis in the Wt1 knockout mouse and the presence of others (16) observed upon WT1 induction in Saos2 osteosarcoma cells, further induction of WT1 in CCG-5.1 did not yield apoptosis constitutional mutations in WT1 in a number of renal developmen- or cell cycle arrest. ChIP–ChIP analysis of CCG-5.1 cells identified tal syndromes indicate a critical role for WT1 in kidney develop- 643 promoters found by 3/3 arrays and 2415 promoters found in 2/3 ment. WT1 is also involved in the development of other organs arrays (Fig. 1A). Hence the higher signal generated by WT1 including spleen, heart, and liver (10–12). In cell culture systems WT1 generally acts as a growth suppressor (13–18) but its overex-

pression in a number of tumor types such as colon and thyroid (19, Author contributions: M.K.-H.K. and J.D.L. designed research; M.K.-H.K. and T.J.M. per- 20) suggests that it may act as an oncogene as well. Accordingly formed research; M.K.-H.K. contributed new reagents/analytic tools; M.K.-H.K., P.O., J.M.F., knockdown of WT1 led to decreased growth of a Wilms tumor cell and A.A.-J.G. analyzed data; and M.K.-H.K. and J.D.L. wrote the paper. line suggesting a permissive role in growth in some cases of Wilms The authors declare no conﬂict of interest. tumor (21). This article is a PNAS Direct Submission. WT1 activates or represses gene transcription depending on the 1To whom correspondence should be addressed. E-mail: [email protected]. cellular, developmental and promoter context. WT1 targets have This article contains supporting information online at www.pnas.org/cgi/content/full/ been sought to explain its function in development and growth 0901591106/DCSupplemental.

11154–11159 ͉ PNAS ͉ July 7, 2009 ͉ vol. 106 ͉ no. 27 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901591106 Downloaded by guest on September 28, 2021 A Table 2. Gene ontology of the genes bound and differentially B regulated by WT1 KEGG pathway term Genes

MAPK signaling pathway DUSP16, DDIT3, MAPKAPK2, PPP3CB, NLK, RRAS2, JUN, DUSP6, DUSP5 Axon guidance NRP1, PPP3CB, ROBO2, CXCR4, ITGB1, GNAI3, EFNA1 C Wnt signaling CCND2, LEF1, BTRC, PPP3CB, NLK, JUN, DKK2, TBL1X, DACT1*

*Not recognized in DAVID functional annotation system and manually added.

the SP1, MAZ, and ZNF219 transcription factors. The MatInspec- tor tool (33) revealed 340 overlapping WT1/SP1 sites, 186 over- Fig. 1. Identification of WT1-bound promoter regions and motif analysis. (A) lapping WT1/EGR1 sites and 149 sites where all three factors may Genes identified by ChIP–ChIP at WT1-induced level in CCG-5.1 are shown in Venn diagrams. (B) A total of 199 bound and differentially regulated genes by WT1 compete for binding (examples shown in Fig. S2). There were a total (shown in bold) were identified in the comparison from the LOF (by shRNA) and of 675 overlapping sites spread over 437 of the 643 promoters. This GOF (CCG-5.1) expression array sets and ChIP–ChIP. (C) The representative motif information suggests that WT1 target genes may have a complex for each cluster was generated using the STAMP platform. The SOMBRERO motifs mode of regulation that depends on the presence and/or activity of were created by clustering the original redundant set of 68 motifs produced by WT1, and other transcription factors. Furthermore, among previ- the algorithm. ously identified in vitro WT1 motifs, EGR1 site was more common than others such as WRE and WTE sites (Table S1). overexpression identified many more putative WT1 targets but Gene Expression Profiles in Gain-of-Function and Loss-of-Function could obscure occasionally, because the enhanced noise caused by Systems. To identify genes differentially regulated by WT1, we WT1 induction led the peak detection program to miss several performed microarray analysis in both gain-of-function (GOF) peaks in this analysis (Fig. S1C). However, some peaks identified in (Fig. S1B) and loss-of-function (LOF) (21) systems. In the initial CCG99–11 were absent without any increase in baseline hybrid- screen for bound and regulated genes, we applied a low stringency ization to other probe sets of the promoter. This could represent cut-off (P Ͻ 0.005, Ն1.5-fold change). Microarray profiling showed redistribution of WT1 among genomic sites upon high-level ex- that 455 genes were regulated in the biological triplicates. Genes pression. All 16 tested genomic regions identified by ChIP–ChIP activated in this system were enriched in axon guidance, whereas were validated by ChIP-PCR (Fig. S1D). Consistent with the down-regulated genes were enriched in aminoacyl-tRNA syntheta- ChIP–ChIP data, peaks in the RHOH and SH3BP5L promoters ses and amino acid metabolism (Table S2). Knockdown of WT1 in identified at endogenous levels of WT1 showed significant enrich- the CCG99–11 cells results in apoptosis and cell cycle arrest. The ment without WT1 induction by quantitative PCR. Nevertheless, WT1 shRNA targets all 4 major isoforms including WT1-A–D (21). many genes were identified at both high and low levels of WT1 Gene expression profiling was performed when the maximum expression, which suggested that induction of WT1 could be used knock down of WT1 protein was achieved. Many genes (1,129) were to identify target genes relevant to Wilms tumors. This was sup- differentially regulated after WT1 depletion (P Ͻ 0.005, Ն1.5-fold ported by gene ontological analysis of the 2415 genes bound by WT1 change) in the biological triplicates. Functional classification re- from CCG5.1-ChIP–ChIP set. WT1-bound targets were enriched vealed that up-regulated genes after WT1 knock-down were en- in MAPK signaling, focal adhesion, regulation of actin cytoskeleton riched in axon guidance, ATP synthesis, focal adhesion, and Wnt and Wnt signaling pathways (Table 1). pathways, whereas down-regulated genes were enriched in cell CELL BIOLOGY cycle, pyrimidine metabolism, one carbon pool and amino acid Motif Analysis on WT1-Bound Regions. To determine the common metabolism pathways (Table S3). Furthermore, a total of 81 genes motifs among DNA sequences bound by WT1 in vivo, we analyzed were regulated in both the GOF and LOF systems and of the 81 the 643 promoters identified after WT1 induction in all 3 ChIP– genes only 9 were identified by ChIP–ChIP, suggesting that the remain- ChIP arrays (Fig. 1A), using 3 de novo motif prediction tools: ing 72 genes might be indirectly regulated by WT1 (Fig. 1B, Table S4). AlignACE (28), MEME (29) and SOMBRERO (30). The 627,000 Collectively, these data show that WT1 manipulation affects only a bp from the WT1-bound regions were first filtered using SOM- small subset of targets directly at a given time and condition. BRERO to increase the signal to noise ratio. All three tools To prioritize WT1 targets, we compared genes differentially identified highly G-rich sequences in 628 of 643 promoters. When regulated by WT1 in the either the GOF or LOF systems with genes compared against known binding sites in the Transfac (31) and identified in at least 2/3 ChIP–ChIP arrays, identifying a total of 199 Jaspar (32) databases, the consensus motifs also resembled sites for genes both bound and regulated upon manipulation of WT1 levels (Fig. 1B and Table S5). Gene ontological analysis showed that pathways for axon guidance, Wnt signaling, and MAPK signaling Table 1. Functional classification of WT1 targets identified from were significantly enriched in this gene set (Table 2). For example, ChIP-ChiP (2,415 gene symbols) ROBO2 (34), a regulator in branching morphogenesis of the kidney, KEGG pathway term Count P plays a role in axon guidance. We previously showed that WT1 negatively regulates MAPK signaling through activation of SPRY1 MAPK signaling 62 0.000048 (25) and DUSP6 (MKP3) (13), supporting the validity of our system Regulation of actin cytoskeleton 45 0.002035 and experimental approach. In the current experiments WT1 was Focal adhesion 45 0.003266 recruited to the DUSP6 promoter but curiously was up-regulated in Long-term potentiation 17 0.009817 both the LOF and GOF systems. Thus, our comprehensive ap- Axon guidance 30 0.012828 proach seemed successful in identifying genes that could be regu- ECM-receptor interaction 21 0.020187 lated by WT1 but did not necessarily indicate whether they were Wnt signaling 30 0.050041 activated or repressed by WT1. We believe this is due to the G-rich

Kim et al. PNAS ͉ July 7, 2009 ͉ vol. 106 ͉ no. 27 ͉ 11155 Downloaded by guest on September 28, 2021 Fig. 3. WT1 negatively affects Wnt signaling in Xenopus embryos. Xwt1 inhibits the secondary axis formation induced by Xwnt8. Xwnt8 (10 pg) was coinjected into the ventral side of 4-cell Xenopus embryos with 500, 2000 pg of GFP or Xwt1(ϪKTS) mRNA. Xwt1 (2,000 pg) alone was injected as a control. Injected embryos were scored blindly for secondary axis formation when they reached the swimming tadpole stage (Nieuwkoop stage 40). P values from Fisher’s exact test Fig. 2. Direct targets of WT1 identiﬁed in the Wnt signaling pathway. Genes (or * from ␹2 test) are shown. bound and regulated by WT1 are indicated in red bold. PPP3CB, a candidate WT1 direct target, is involved in noncanonical Wnt/Ca2ϩ pathway (not shown here). Activation and repression by WT1 are shown with up and down arrows, Wnt signaling (35) and a repression target of WT1 in Wilms tumors respectively. (36), in HEK 293–5.1. Transcriptional activities from these promoters were induced by transient transfection of wild-type ␤-cate- nature of the WT1 binding site, which can also be recognized by nin, a hyperactive stabilized ␤-catenin S37A mutant or Wnt3a and other transcription factors such as SP1, MAZ, and EGR1. these activities were suppressed Ϸ50% when WT1 was induced by In this study, we further focused on the WT1’s effect in the Wnt doxycycline addition (Fig. 4A). Dose-dependent activation of the signaling pathway. The potential interplay between WT1 and Wnt TOPFLASH reporter by Wnt3a-conditioned media (Wnt3a-CM) signaling is significant in light of increasing evidence of activation- was consistently inhibited by WT1 induction as well (Fig. 4B). ing mutation of Wnt/␤-catenin pathways in Wilms tumor, particu- Next, we investigated how this WT1-mediated transcriptional larly in tumors without functional WT1 (4–7). This led us to inhibition occurred. Because we observed that WT1 induction hypothesize that WT1 might inhibit Wnt signaling. ChIP–ChIP and affected neither total nor active ␤-catenin levels in both 293–5.1 and expression analysis identified 8 direct targets in the canonical Wnt CCG-5.1 cells, this interference was not due to the decrease in pathway (Fig. 2) and PPP3CB in the noncanonical Wnt/Ca2ϩ ␤-catenin level (Fig. 4B and Fig. S4). Electrophoretic mobility shift pathway. Some but not all Wnt-related genes were regulated by assays (EMSA) of nuclear extracts from Wilms tumor or 293 cells WT1 in our systems in a manner indicating that WT1 opposed Wnt readily detected a TCF-DNA complex that was completely unaf- signaling. This may be because of cell context and because other fected by induction of WT1 (Fig. S4). transcription factors bind to the GC-rich WT1 motif. Because the in vitro EMSA assay may not reflect complex assembly in vivo, we performed ChIP to measure dynamic inter- WT1 Inhibits the Wnt Signaling in Xenopus Embryonic Development. actions of WT1, TCF4 (TCF7L2), and the CBP coactivator. ␤-cate- We sought to determine how WT1 might affect Wnt signaling in nin interacts with CBP to activate transcription and is required for vivo. However, CCG-5.1 and HEK 293–5.1 showed no biological the activation of Wnt-responsive genes in Xenopus (37). CBP also changes in properties such as proliferation or morphology upon complexes with WT1 to enhance its transcriptional activity (38). Wnt3A-CM treatment, even in the absence of WT1. Therefore, we Consistent with transient reporter assays, the recruitment of TCF4 used the Xenopus system in which exogenous Wnt signals cause axis to the TOPFLASH promoter was consistently inhibited upon WT1 duplication and the ability of the WTX protein to antagonize the induction (Fig. 5A). However, WT1 was not recruited to the Wnt signaling was shown (3). Four-cell embryos were injected on TOPFLASH promoter, suggesting an indirect mechanism of inhi- the ventral side with a mixture of Xwnt8 with either GFP or bition, potentially through competition between WT1 and the Xwt1(ϪKTS) RNAs and WT1 expression was checked (Fig. S3B). TCF/␤-catenin complex for CBP. Accordingly, the decrease in Coinjection of Xwt1(ϪKTS) mRNA with Xwnt8 mRNA dramati- TCF4 recruitment onto the TOPFLASH promoter upon WT1 cally reduced the number of embryos that form a complete sec- induction was abolished when CBP was overexpressed (Fig. 5B). ondary axis and most embryos formed only a partial secondary axis, CBP recruitment to the TOPFLASH promoter slightly increased when compared with coinjection of GFP RNA with Xwnt8 as a upon CBP cotransfection, although measuring the indirect binding control (Fig. 3, Fig. S3A). The ability of WT1 to inhibit Xwnt of cofactor on the transiently transfected artificial promoter in a activity was similar in magnitude to that reported for WTX (See SI consistent manner may be a technical challenge. Using the same Text). Injection Xwt1 mRNA by itself had no affect on secondary ChIP samples in immunoblotting, we observed that the endogenous axis formation. Xwt1(ϩKTS) also significantly inhibited axis induc- CBP interacted with both WT1 and TCF4 (Fig. 5C). Upon WT1 tion by Xwnt8, but the effect was not as pronounced as with the induction, WT1 showed a very weak interaction with TCF4 and ϪKTS isoform. Xwt1(ϪKTS) also inhibited secondary axis forma- conventional immunoprecipitation could not demonstrate a WT1/ tion mediated by Xwnt3a. TCF4 interaction. In addition CBP overexpression, in a dose- dependent manner, reversed the repression activity of WT1 on the WT1 Negatively Affect TCF-Mediated Transcriptional Activity. WT1 TOPFLASH promoter (Fig. 5D). Collectively, these data suggest may regulate Wnt signaling in part by binding and regulating genes that the ability of WT1 to interfere with TCF-mediated transcrip- encoding components and targets of the Wnt pathway. To deter- tional activity is due to its ability to bind to CBP rather than direct mine whether other mechanism may be at work as well, we tested interaction with TCF binding sites or TCF4 itself. the effect of WT1 on TCF-mediated transcription, using an artificial TCF/LEF binding site containing reporter (TOPFLASH) and CBP Is Recruited to the WT1 Targets. Although CBP was shown to on the reporter of c-myc promoter, a natural activation target of physically interact with WT1 in vitro and contribute to the syner-

11156 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901591106 Kim et al. Downloaded by guest on September 28, 2021 A screens (39, 40). Given our finding that many genes directly regulated by WT1 were also Wnt target genes, we compared the 412 high-confidence TCF4 peaks (40) with WT1 peaks. Of the 412 clusters, as reported previously, nine were found in gene promoters, whereas 213 were found within transcribed regions. Of those 9 genes, 5 genes contain the predicted WT1 binding sites with no overlaps between predicted TCF/LEF and WT1 sites. However, when we compared TCF peaks (converted to 166 gene symbols) with WT1 peaks (2415 gene symbols) there were 29 genes common even though the TCF4 and WT1 ChIP assays were performed in different cell types and using different techniques, suggesting that there are indeed some WT1 and Wnt genes regulated in common. Both TCF4 studies reported that only Ϸ15% of TCF binding sites were found in the 5Ј flanking sequences of genes (2.5 kb upstream and downstream of the transcriptional start site) and tended to be located at large distances from the body of a gene. Future studies assaying the entire genome for WT1 binding sites could find more overlap with TCF binding regions. Collec- tively, we conclude that WT1 and TCF4 can share common target genes through distinct DNA binding elements given that there was no overlap between WT1 binding sites and TCF4-GST clusters.

B Discussion Canonical Wnt/␤-catenin signaling and its precise regulation in a temporal and spatial manner are essential for mesenchymal- to-epithelial transition (MET) during kidney development and especially for molecular dynamics of tubule formation (reviewed in refs. 41 and 42). ␤-catenin has a dual role involved in cell–cell adherent junctions and in gene transcription as a Wnt signaling effector. Identification of Wnt4 as a WT1 target demonstrated a positive role of WT1 on the Wnt signaling during kidney development (24). De-regulation of Wnt/catenin signaling is implicated in the pathogenesis of Wilms tumor (reviewed in ref. 43) and overexpression of activated ␤-catenin in the developing kidney precludes normal nephron development (44). Interest- ingly, WT1 null Wilms tumors, frequently containing GOF mutations in ␤-catenin, have a stromal-predominant histology, whereas WT1 replete tumors have triphasic and blastemal- and epithelial predominant histology (45). One way to explain the Fig. 4. WT1 negatively regulates TCF-mediated transcription. (A) TCF-mediated ␤ transcriptional activity was determined using TOPFLASH construct containing 4 occurrence of -catenin mutation only in WT1 null tumors TCF binding sites and the specificity was confirmed using the negative control would be that wild-type WT1 inhibits Wnt signaling and our data counterpart FOPFLASH. A c-myc promoter activity was measured as a natural support this idea. counterpart. Transcriptional activity in 3 replicates was measured after cotrans- First, WT1 directly regulated many Wnt signaling components CELL BIOLOGY fecting a reporter plasmid (200 ng) with 800 ng of pEGFP-C1 (empty vector), and Wnt target genes in a Wilms tumor cell line (Fig. 2). We used wild-type ␤-catenin, ␤-catenin S37A, or WNT3a expressing vector in the absence an integrated genome-wide approach of ChIP–ChIP and microar- or presence of WT1 induction in 293–5.1 cells. The luciferase activity was mea- ray analysis to identify WT1 target genes. Because many previously sured at 48 h after transfection. The graphs shown represent 3 independent identified WT1 targets contained WT1 responsive elements within experiments. (B) Either 200 ng of TOPFLASH or FOPFLASH was cotransfected with 2 kb from transcriptional start sites, we used the promoter arrays. 800 ng of pEGFP-C1 in 6-well plates. Transfection mix was removed after4hof incubation and fresh medium was added without and with 2 ␮g/mL doxycycline. It is possible that more direct targets of WT1 will be identified using After 40 h after transfection, Wnt3a-CM was added to induce the Wnt signaling genome-wide tiling arrays or the ChIP-seq technique. Nevertheless, for 4 h before assays. Twenty microliters of each lysate was used for luciferase WT1 target genes are enriched for those involved in Wnt signaling. assay, and 30 ␮L was loaded on 10% SDS/PAGE to check expression levels of This may be one way by which WT1 opposes Wnt signaling. Second, ␤-catenin (detected by ␤-catenin E-5, Santa Cruz) and WT1. GFP was cotrans- Xwt1 inhibited the Wnt action in the Xenopus system. A recent fected to monitor the transfection efficiency and used as a loading control on the study showed that WT1 negatively regulates Wnt signaling in other Western blot. The graphs shown represent 3 independent experiments. contexts. In Sertoli cells, deletion of WT1 leads to up-regulation of ␤-catenin itself, and Sertoli-specific Wt1 deletion phenocopied the effect of stabilized ␤-catenin in the testis (46). Third, WT1 induc- gistic activation in reporter assays (38), the importance of CBP for tion inhibited Wnt-mediated transcription by preventing the re- WT1 function in vivo is less certain. ChIP assay showed that WT1 and CBP were present together on genes both up-regulated (ETV5 cruitment of TCF onto a promoter, an effect overcome by over- and NRP1) and down-regulated (ATF3 and PAG1) by WT1 (Fig. expression of CBP. Cofactor competition was reported in other 6A). Intriguingly we noted that CBP recruitment to these WT1 systems. For example, nuclear receptors inhibit AP1-dependent target promoters was more evident when cells were grown in the transcription by competing for limited amount of CBP (47), presence of Wnt3a-CM (Fig. 6B). This is more evidence of interplay whereas the IFN-␥ and TGF-␤ signaling pathways compete for between WT1 and Wnt signaling. The highly related p300 protein was limiting amounts of the CBP-related factor p300 (48). Unlike not recruited to any of the WT1 target promoters. previous studies showing that WT1 triggered ␤-catenin degradation in breast cancer cells (15) and Sertoli cells (46), WT1 did not affect TCF4 and WT1 Share Common Targets but Not Binding Sites. Recently, ␤-catenin levels in our system. Last, we discovered that WT1 and in vivo binding sites for TCF4 were discovered from 2 genome-wide TCF4 share common target genes although different cis-acting

Kim et al. PNAS ͉ July 7, 2009 ͉ vol. 106 ͉ no. 27 ͉ 11157 Downloaded by guest on September 28, 2021 Fig. 5. WT1 inhibited the DNA binding of the TCF complex through cofactor CBP in vivo. (A and B) The 293–5.1 AB cells were transfected with TOPFLASH (1 ␮g) and Wnt3a (4 ␮g) expression vectors (A), and TOPFLASH (1 ␮g), Wnt3a (4 ␮g) and mCBP (4 ␮g) expression vectors (B) in 150-mm plates. Transfection mix was removed after4hofincuba- tion and fresh medium was added without and with 2 ␮g/mL doxycycline. At 48 h after transfection, cells were processed for ChIP. Enrichment was calculated from Ct numbers and shown as relative to % input signal in 2 independent experiments. (C) Western blot analysis of chromatin immunoprecipitated samples. The same amount (5–8 ϫ 106 cell equivalents) of ChIP samples from CD the same transfection used in Fig. 5A was boiled for 20 min after mixing with SDS-sample loading dye and loaded on the 10% SDS gel. To avoid the signal from IgG heavy chain, True Blot ULTRA HRP-conjugated anti-mouse or anti-rabbit IgG (eBioscience) was used as secondary antibody. (D) TOP- FLASH reporter construct was cotransfected with either empty or pSG5-mCBP expression vector into 293–5.1 cells in triplicates. Transfection mix was removed after4hofin- cubation and fresh medium was added without and with 2 ␮g/mL doxycycline. Wnt3a-CM was added for 4–5 h before harvest. The luciferase assay was done at 48 h after transfection and standard deviation was calculated from 3 independent transfections. The graph represents 5 independent similar experiments.

elements, further supporting the interplay between WT1 and WNT factors with larger and less degenerate binding sites such as p53 for pathways. example might more cleanly bind and activate a set of genes, with The absence of ␤-catenin mutations in WT1 replete tumors, the less overlap with widely expressed transcription factors. The re- ability of WT1 to bind and regulate Wnt pathway genes, to suppress dundant G/GC rich nature of the WT1 recognition sequence may duplex axis formation in Xenopus embryos and to inhibit TOP- also explain why microarray studies have identified divergent sets of FLASH activity by wild-type or activated mutant ␤-catenin, all WT1-regulated genes in different cell types. The nature of the GC suggest that inhibition of Wnt pathways is a major tumor suppressor sequence binding protein present in the cell likely affects the basal function of WT1. How Wilms tumors develop in the presence of state of gene expression. How readily GC binding factors are WT1 is still uncertain but may be related to the recently discovered released and replaced by WT1 on a promoter could determine the WTX gene that can be mutated in WT1 replete tumors (3). ability of WT1 to regulate the putative target gene. Of note, because Although GC-rich and TCC repeat sequences were identified as the knockdown of WT1 would lead to loss of all WT1 isoforms, in vitro binding sites, the nature of the WT1 binding site in vivo has whereas only WT1(ϪKTS) isoform was induced in the GOF been less clear. Our unbiased searches identified consensus G-rich system, our study would not score genes whose activation or motifs among WT1-bound regions in vivo. We also observed that repression depended on the presence of other alternatively spliced WT1 bound G-rich sequences frequently overlapped or coincided WT1 isoforms and might miss genes that are exclusively bound by with EGR1 and SP1 binding sites. Similarly, WT1 binding sites in those isoforms. In addition, some of the genes maximally activated the promoters of 11 genes coordinately expressed in prostate cancer or repressed in the presence of endogenous levels of WT1 would be cells overlapped with SP1 and EGR1 sites (49). Collectively, these missed. Nevertheless, through this genome-wide screen, we found data indicate that G/GC-rich sequences are common WT1 binding that the action of WT1 in development and growth suppression may sites in vivo. This may explain the finding that at times both be in part explained through its ability to suppress Wnt signaling. induction and knockdown of WT1 activates a WT1 target. In the former case WT1 might occupy the promoter and engage the Materials and Methods transcriptional machinery. In the knockdown case, WT1 might Plasmids and Cell Cultures. See SI Text. represent a repressor or weak activator of the gene and WT1-loss might be accompanied by the binding of EGR1, SP1, or other ChIP and ChIP–ChIP. ChIP was performed as in ref. 21, using WT1 antibody (C19) GC-binding factors of the KLF family. By contrast transcription (Santa Cruz Biotechnology). Input and WT1-ChIP samples were amplified by ligation-mediated PCR, using JW102 5Ј-GCGGTGACCCGGGAGATCTGAATTC and JW103 5Ј-GAATTCAGATC oligos described at http://genomics.ucdavis.edu/ farnham/protocol.html. Three biological input replicates were labeled with Cy3 A B and the corresponding WT1-ChIP samples were labeled with Cy5. Hybridization onto HG18࿝RefSeq Promoter array was performed by NimbleGen. The peaks were identified with a cut-off of false discovery rate of Ͻ0.2 using Nimblegen SignalMap software. For motif analysis genes had to be identified by all 3 ChIP–ChIP experiments. For gene ontology we used a less stringent cut-off and included genes if 2 of 3 replicates detected a significant peak. An independent ChIP was performed for peak validation. Input and ChIP-samples were purified using the Qiagen PCR purification kit. Quantitative PCR was performed using Stratagene Mx3000 machine in a 25-␮L reaction volume (21). Antibodies for ChIP include normal rabbit IgG (Santa Cruz, or Zymed in Figs. 5B and 6), WT1(C-19) (Santa Cruz), TCF4 (US Biological) and CBP(A-22) (Santa Cruz). PCR primer se- Fig. 6. Cofactor CBP is recruited to the WT1-bound regions in Wilms tumor cells. quences are shown in Table S1. WT1 in CCG5.1 cells was induced for 48 h and Wnt3a-CM treatment was done for 4 h before cross-linking. Polyclonal antibodies of rabbit IgG (Zymed), WT1 and Immunoblotting and EMSA. See SI Text. CBP were used for ChIP. Four promoter loci of WT1 candidate genes identified from ChIP–ChIP and expression arrays in this study were tested in the absence (A) Microarray and Bioinformatic Analysis. For gain-of-function microarray analysis, and presence (B) of Wnt3a-CM with and without WT1 induction. RNA was isolated from CCG5.1 cells before and 48 h after WT1-induction, using

11158 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901591106 Kim et al. Downloaded by guest on September 28, 2021 the RNeasy Mini Kit (Qiagen). For loss-of-function experiments, CCG99–11 cells Department of Medicine, Oregon Health and Science University, Portland, OR) were transfected with either control shRNA or a WT1 shRNA (21), selected for 4 (40). The clusters were mapped as either being within a particular gene (between days with 0.4 ␮g/mL of puromycin and harvested at day 5 for RNA isolation. After the transcriptional start and end) or in the gene promoter (2,000 bp upstream of confirmation of RNA quality, using an Agilent 2100 bioanalyzer, RNA was con- TSS) on build 17 of the human genome. The promoters were defined as a region verted to biotin-labeled cDNA, using the GeneChIP expression 3Ј-amplification 2,000 bp upstream to 300 bp downstream of TSS as in the NimbleGen promoter reagents one-cycle cDNA synthesis kit (Affymetrix) and hybridized to Affymetrix array design. U133Plus2 chips. Cel files were imported into ArrayAssist software 5.2.2 (Strat- agene), and probe levels were normalized by the GC-RMA algorithm. Lists of Xenopus Embryo Injection and Statistical Analysis. Plasmid DNA was linearized differentially expressed genes were created using a 1.5-fold cut-off with PϽ0.005 and transcribed in vitro in a 50-␮L reaction volume containing 2.5 ␮g of DNA, 0.5 by unpaired T test in the biological triplicates. Gene ontology of all differentially mM each rNTP, 0.5 mM GpppG cap (Amersham/GE Healthcare), 10 mM DTT, 50 expressed genes, and WT1-activated and repressed genes were separately deter- units RNAsin and 40 units SP6 polymerase (Promega). Four-cell Xenopus embryos mined using DAVID (http://david.abcc.ncifcrf.gov). showing a clear pigmentation difference between the dorsal and ventral side were injected with RNA (10 nL) into the ventral (darker) vegetal blastomeres. Motif Analysis. MEME, AlignACE and SOMBRERO algorithms were used to Embryos were allowed to develop at room temperature until they reached the discover motifs of length 8–22 bp in the 643 promoters. Parameters for all swimming tadpole stage (Nieuwkoop stage 40–45) and then were scored blindly algorithms were kept as similar as possible with no predefined expectations on for secondary axis formation. If the induced axis contained either cement gland either the number of sites in the dataset or their distribution across the promot- or eye tissue it was scored as complete, otherwise it was scored as partial. ϫ ers. SOMBRERO motifs were further clustered using the STAMP (50) platform to Statistical significance was calculated by Fisher’s exact test (3 2 exact contin- ␹2 reduce the redundancy in its predicted motif set. Hierarchical agglomerative gency table) except when test was more appropriate. clustering with ungapped local alignment was used to produce a reduced set of 5 motifs, each being equivalent to the familial binding profile or average binding Transient Transfections. See SI Text. specificity for its particular cluster. These 5 representative motifs were compared against the top 5 motifs produce by MEME and AlignACE, and all were queried ACKNOWLEDGMENTS. We thank Dr. Greg Khitrov and Jin Chen (Mount Sinai against the Transfac database to ascertain their similarity to known motifs. Of the School of Medicine) for microarray hybridization and to Katie Shinnick (North- resulting matches, a further analysis was carried out to examine the distribution western University) for RNA synthesis for Xenopus embryo injection. This work was supported by National Institutes of Health Grant CA102270 (to J.D.L.), the of SP1, EGR1, and WT1 binding sites, using MatInspector. This analysis manually Northwestern Memorial Foundation (J.D.L.), American Cancer Society Postdoc- determined loci in which predicted binding sites for 2 or more these factors toral Fellowship PF-05-252-01-MGO (to M.K.-H.K.), Science Foundation Ireland overlapped. For WT1 binding site analysis in TCF4 peaks, the 412 clusters con- Grant RFP/05/CMS0001 (to P.O. and A.G.) and American Heart Association Grant taining 3 or more GSTs were kindly provided by R. Goodman (Vollum Institute and 0630290N (to T.M.).

1. Ogawa O, et al. (1993) Relaxation of insulin-like growth factor II gene imprinting impli- 27. Morrison DJ, English MA, Licht JD (2005) WT1 induces apoptosis through transcriptional cated in Wilms’ tumour. Nature 362:749–751. regulation of the proapoptotic Bcl-2 family member Bak. Cancer Res 65:8174–8182. 2. Lee SB, Haber DA (2001) Wilms tumor and the WT1 gene. Exp Cell Res 264:74–99. 28. Roth FP, Hughes JD, Estep PW, Church GM (1998) Finding DNA regulatory motifs within 3. Rivera MN, et al. (2007) An X chromosome gene, WTX, is commonly inactivated in Wilms unaligned noncoding sequences clustered by whole-genome mRNA quantitation. Nat tumor. Science 315:642–645. Biotechnol 16:939–945. 4. Maiti S, Alam R, Amos CI, Huff V (2000) Frequent association of beta-catenin and WT1 29. Bailey TL, Elkan C (1994) Fitting a mixture model by expectation maximization to discover mutations in Wilms tumors. Cancer Res 60:6288–6292. motifs in biopolymers. Proc Int Conf Intell Syst Mol Biol 2:28–36. 5. Koesters R, et al. (1999) Mutational activation of the beta-catenin proto-oncogene is a 30. Mahony S, Hendrix D, Golden A, Smith TJ, Rokhsar DS (2005) Transcription factor binding common event in the development of Wilms’ tumors. Cancer Res 59:3880–3882. site identification using the self-organizing map. Bioinformatics 21:1807–1814. 6. Uschkereit C, et al. (2007) Different CTNNB1 mutations as molecular genetic proof for the 31. Wingender E, et al. (2000) TRANSFAC: An integrated system for gene expression regula- independent origin of four Wilms tumours in a patient with a novel germ line WT1 tion. Nucleic Acids Res 28:316–319. mutation. J Med Genet 44:393–396. 32. Sandelin A, Alkema W, Engstrom P, Wasserman WW, Lenhard B (2004) JASPAR: An 7. Li CM, et al. (2004) CTNNB1 mutations and overexpression of Wnt/beta-catenin target open-access database for eukaryotic transcription factor binding profiles. Nucleic Acids genes in WT1-mutant Wilms’ tumors. Am J Pathol 165:1943–1953. Res 32:D91–D94. 8. Scharnhorst V, Dekker P, van der Eb AJ, Jochemsen AG (1999) Internal translation initiation 33. Quandt K, Frech K, Karas H, Wingender E, Werner T (1995) MatInd and MatInspector: New generates novel WT1 protein isoforms with distinct biological properties. J Biol Chem fast and versatile tools for detection of consensus matches in nucleotide sequence data. 274:23456–23462. Nucleic Acids Res 23:4878–4884. 9. Bickmore WA, et al. (1992) Modulation of DNA binding specificity by alternative splicing 34. Piper M, Georgas K, Yamada T, Little M (2000) Expression of the vertebrate Slit gene family of the Wilms tumor wt1 gene transcript. Science 257:235–237. and their putative receptors, the Robo genes, in the developing murine kidney. Mech Dev 10. Herzer U, Crocoll A, Barton D, Howells N, Englert C (1999) The Wilms tumor suppressor 94(1–2):213–217. gene wt1 is required for development of the spleen. Curr Biol 9:837–840. 35. He TC, et al. (1998) Identification of c-MYC as a target of the APC pathway. Science 11. Wagner N, Wagner KD, Scholz H, Kirschner KM, Schedl A (2006) Intermediate filament 281:1509–1512. CELL BIOLOGY protein nestin is expressed in developing kidney and heart and might be regulated by the 36. Udtha M, Lee SJ, Alam R, Coombes K, Huff V (2003) Up-regulation of c-MYC in WT1-mutant Wilms’ tumor suppressor Wt1. Am J Physiol Regul Integr Comp Physiol 291:R779–R787. tumors: Assessment of WT1 putative transcriptional targets using cDNA microarray ex- 12. Ijpenberg A, et al. (2007) Wt1 and retinoic acid signaling are essential for stellate cell pression profiling of genetically defined Wilms’ tumors. Oncogene 22:3821–3826. development and liver morphogenesis. Dev Biol 312:157–170. 37. Takemaru KI, Moon RT (2000) The transcriptional coactivator CBP interacts with beta- 13. Morrison DJ, Kim MK, Berkofsky-Fessler W, Licht JD (2008) WT1 induction of mitogen- catenin to activate gene expression. J Cell Biol 149:249–254. activated protein kinase phosphatase 3 represents a novel mechanism of growth suppres- 38. Wang W, Lee SB, Palmer R, Ellisen LW, Haber DA (2001) A functional interaction with CBP sion. Mol Cancer Res 6:1225–1231. contributes to transcriptional activation by the Wilms tumor suppressor WT1. J Biol Chem 14. Fraizer G, et al. (2004) Suppression of prostate tumor cell growth in vivo by WT1, the Wilms’ 276:16810–16816. tumor suppressor gene. Int J Oncol 24:461–471. 39. Hatzis P, et al. (2008) Genome-wide pattern of TCF7L2/TCF4 chromatin occupancy in 15. Zhang TF, Yu SQ, Guan LS, Wang ZY (2003) Inhibition of breast cancer cell growth by the colorectal cancer cells. Mol Cell Biol 28:2732–2744. Wilms’ tumor suppressor WT1 is associated with a destabilization of beta-catenin. Anti- 40. Yochum GS, et al. (2007) Serial analysis of chromatin occupancy identifies beta-catenin cancer Res 23(5A):3575–3584. target genes in colorectal carcinoma cells. Proc Natl Acad Sci USA 104:3324–3329. 16. Englert C, et al. (1995) WT1 suppresses synthesis of the epidermal growth factor receptor 41. Vainio SJ, Uusitalo MS (2000) A road to kidney tubules via the Wnt pathway. Pediatr and induces apoptosis. EMBO J 14:4662–4675. Nephrol 15(1–2):151–156. 17. Haber DA, et al. (1993) WT1-mediated growth suppression of Wilms tumor cells expressing 42. Schmidt-Ott KM, Lan D, Hirsh BJ, Barasch J (2006) Dissecting stages of mesenchymal-to- a WT1 splicing variant. Science 262:2057–2059. epithelial conversion during kidney development. Nephron Physiol 104:56–60. 18. Luo XN, et al. (1995) The tumor suppressor gene WT1 inhibits ras-mediated transforma- 43. Nusse R (2007) Cancer. Converging on beta-catenin in Wilms tumor. Science 316:988–989. tion. Oncogene 11:743–750. 44. Park JS, Valerius MT, McMahon AP (2007) Wnt/beta-catenin signaling regulates nephron 19. Oji Y, et al. (2003) Overexpression of the Wilms’ tumor gene WT1 in primary thyroid cancer. induction during mouse kidney development. Development 134:2533–2539. Cancer Sci 94:606–611. 45. Fukuzawa R, et al. (2004) Myogenesis in Wilms’ tumors is associated with mutations of the 20. Oji Y, et al. (2003) Overexpression of the Wilms’ tumor gene WT1 in colorectal adenocar- WT1 gene and activation of Bcl-2 and the Wnt signaling pathway. Pediatr Dev Pathol cinoma. Cancer Sci 94:712–717. 7:125–137. 21. Kim MK, et al. (2008) A pathologic link between Wilms tumor suppressor gene, WT1, and 46. Chang H, et al. (2008) Wt1 negatively regulates beta-catenin signaling during testis IFI16. Neoplasia 10:69–78. development. Development 135:1875–1885. 22. Englert C, Maheswaran S, Garvin AJ, Kreidberg J, Haber DA (1997) Induction of p21 by the 47. Kamei Y, et al. (1996) A CBP integrator complex mediates transcriptional activation and Wilms’ tumor suppressor gene WT1. Cancer Res 57:1429–1434. AP-1 inhibition by nuclear receptors. Cell 85:403–414. 23. Lee SB, et al. (1999) The Wilms tumor suppressor WT1 encodes a transcriptional activator 48. Ghosh AK, Yuan W, Mori Y, Chen S, Varga J (2001) Antagonistic regulation of type I collagen of amphiregulin. Cell 98:663–673. gene expression by interferon-gamma and transforming growth factor-beta. Integration at 24. Sim EU, et al. (2002) Wnt-4 regulation by the Wilms’ tumour suppressor gene, WT1. the level of p300/CBP transcriptional coactivators. J Biol Chem 276:11041–11048. Oncogene 21:2948–2960. 49. Eisermann K, et al. (2008) Evolutionary conservation of zinc finger transcription factor 25. Gross I, et al. (2003) The receptor tyrosine kinase regulator Sprouty1 is a target of the tumor binding sites in promoters of genes co-expressed with WT1 in prostate cancer. BMC suppressor WT1 and important for kidney development. J Biol Chem 278:41420–41430. Genomics 9:337. 26. Bridgewater D, et al. (2008) Canonical WNT/beta-catenin signaling is required for ureteric 50. Mahony S, Benos PV (2007) STAMP: A web tool for exploring DNA-binding motif similar- branching. Dev Biol 317:83–94. ities. Nucleic Acids Res 35:W253–W258.

Kim et al. PNAS ͉ July 7, 2009 ͉ vol. 106 ͉ no. 27 ͉ 11159 Downloaded by guest on September 28, 2021