Genomic characterization of Wilms' tumor suppressor 1 targets in nephron progenitor cells during kidney development

The pediatric kidney malignancy Wilms' tumor has an incidence of 1 in 10 000 in North America (Matsunaga, 1981), making it the most common solid tumor in childhood (Bennington and Beckwith, 1975). Wilms' tumor is thought to arise from a single transformed pluripotent nephron progenitor cell whose progeny fail to undergo normal differentiation. WT1 was the first gene identified as mutated in Wilms' tumors (Call et al., 1990; Gessler et al., 1990; Haber et al., 1990), and inactivating mutations in Wt1 are responsible for ~10% of sporadic Wilms' tumor cases. WT1 also plays a crucial role during embryogenesis, including development of the kidneys and gonads (Kreidberg et al., 1993; Pelletier et al., 1991). At the onset of kidney development, Wt1 is weakly expressed in the uninduced metanephric mesenchyme and increases in the cap of condensed nephrogenic progenitors surrounding the tips of the branching ureteric bud as development proceeds (Pritchard-Jones et al., 1990). Targeted mutation of Wt1 in mice results in bilateral renal agenesis, characterized by apoptosis of the metanephric mesenchyme and failure of ureteric bud invasion into the metanephric mesenchyme (Kreidberg et al., 1993). Importantly, nephron progenitor incompetence is the primary defect in Wt1^−/− embryos, as evidenced by the failure of isolated Wt1^−/− mesenchymal rudiments to differentiate when co-cultured with wild-type ureteric bud cells (Donovan et al., 1999). In the complementary gain-of-function experiment, microinjection of Wt1-expressing plasmids into isolated embryonic kidneys stimulates nephron development (Gao et al., 2005). Collectively, these results strongly suggest that Wt1 promotes nephron progenitor differentiation during kidney development in vivo.

Wt1 encodes a transcription factor with four Krüppel-type (Cys₂-His₂) zinc finger domains. In mammals, an alternative splice donor site at exon 9 inserts the amino acids lysine, threonine and serine (KTS) between the third and fourth zinc fingers, significantly diminishing the DNA-binding affinity of WT1 (Gessler et al., 1992; Haber et al., 1991). Thus, (+KTS) WT1 isoforms have a high affinity for RNA (Bor et al., 2006), whereas (−KTS) isoforms bind DNA with high affinity and function in transcriptional regulation. Several EGR1-like GC-rich DNA sequences and (TCC)n consensus sequences have been identified as cognate WT1 (−KTS) binding sites in vitro (Drummond et al., 1994; Rauscher et al., 1990), including a high-affinity 10-bp EGR1-like (WT1) motif (GCGTGGGCGG) associated with WT1-dependent gene transcriptional activation in vitro (Hamilton et al., 1995; Nakagama et al., 1995).

The identification of direct WT1 target genes will be essential to our understanding of WT1 function in the developing kidney and other developing organs. Therefore, to gain a deeper insight into the WT1-mediated regulatory networks that control kidney development in vivo, we initiated a systematic effort to define the genes directly regulated by WT1 during renal development. WT1-directed ChIP products isolated from mouse embryonic kidneys were used to interrogate mouse promoter arrays without PCR amplification prior to hybridization, to avoid the introduction of PCR amplification bias. A large number of in vivo target sites were identified and were validated by bioinformatics analysis, ChIP-PCR, and a novel, modified WT1 morpholino loss-of-function model in embryonic kidney explants. Essential kidney development genes Bmp7, Pax2 and Sall1 were identified as WT1 transcriptional target genes and might explain the renal agenic phenotype observed in Wt1^−/− embryos. Further, our data identified numerous WT1 target genes not previously studied in the developing kidney co-expressed with Wt1 in nephron progenitors and potentially mediating its function in nephron progenitor differentiation in vivo. Collectively, these data provide novel insights into the signaling networks and biological processes that might be regulated by WT1 in nephron progenitors during kidney development in vivo.

Location analysis was performed as previously described (Lee et al., 2006), using pooled polyclonal anti-WT1 antibodies directed against the C-19 and N-180 terminal amino acids of WT1 (Santa Cruz), and optimized for chromatin extraction from embryonic mouse kidney tissue. Hybridization to Agilent mouse 244K promoter tiling arrays was performed using Agilent SureHyb hybridization chambers according to the manufacturer's Mammalian ChIP-on-chip v.9.0 Protocol (http://www.chem.agilent.com/Library/usermanuals/Public/G4481-90010_ManmalianProtocol_10.11.pdf) (Whitehead Institute Genome Technology Core, Cambridge, MA, USA). Raw ChIP-chip fluorescence intensity data were analyzed using ChIP Analytics 1.3.1 software (Boyer et al., 2005). Settings included spatial detrending of extracted array data, and dye-bias intra-array Lowess normalization. Normalized log-intensity and log ratio histograms followed a normal and symmetric distribution (see Fig. S1 in the supplementary material), indicating that the ChIP data are of high quality. The complete data set is available in the Array Express database (http://www.ebi.ac.uk/microarray-as/ae/, Accession No. E-TABM-872). All coordinates in this manuscript are reported in mm9.

To identify sequence motifs in WT1-bound regions detected by ChIP-chip, sequences of all 4953 WT1-bound regions were extracted and extended 200 bp from both ends. The hypothesis-driven motif discovery algorithm THEME (Macisaac et al., 2006) tested a compendium of 233 unique sequence motifs derived from ~400 mammalian transcription factor binding specificities in the TRANSFAC and JASPAR databases, to identify the enrichment of any of these motifs in WT1-bound sequences, using 5490 randomly sampled, unbound sequences for background comparison. Motifs were ranked by their mean cross-validated prediction error on held-out-bound and unbound test sequences. The top-ranking motif identified by THEME (Fig. 1A), henceforth called the WT1 matrix, comprises multiple 8-12 bp permutations of a motif consistent with the previously reported WT1 consensus sequence: GCG(T/G)(G/A)GG(C/A)G(G/T) (Hamilton et al., 1995; Nakagama et al., 1995). The statistical significance of the WT1 matrix site was determined by randomly permuting bound and unbound sequence labels and rerunning the algorithm 25 times to obtain an empirical null distribution of cross-validation errors of ~ 0.38 and an associated z-score of27.5, indicating that the WT1 matrix site is highly statistically significant. To map the WT1 binding site to our bound data set, each bound region was scanned for matches to the WT1 consensus sequence, permitting the defined substitutions denoted above in the fourth, fifth, eighth and tenth positions (substitutions noted in parentheses).

ChIP followed by site-specific PCR (ChIP-PCR) was performed according to published protocols (Lee et al., 2006) to confirm binding of WT1 to a panel of ~ 40 WT1 target loci identified by ChIP-chip (see Table S1 in the supplementary material). Each PCR experiment comprised 5 PCR reactions: no DNA, rabbit IgG ChIP, RNA Polymerase II ChIP, WT1 ChIP and input DNA, using equimolar amounts (30 ng) of starting DNA per reaction. The following antibodies were used to perform ChIP: anti-rabbit IgG and anti-WT1 (Santa Cruz), and anti-RNA polymerase II (Upstate). The linear range of amplification was determined for each PCR reaction, with the optimal cycle number semi-quantitatively determined as 2 cycles prior to the plateau phase of amplification (see Table S2 in the supplementary material).

Mouse embryonic kidneys were excised from E12.5 pregnant CD1 mice, and kidneys that had undergone 2-3 branching events were selected and transferred onto a 0.4 μM polyethylene terephthalate membrane (Falcon). Explants were cultured for 24 hours as previously described (Piscione et al., 2001), in media supplemented with either 10 μM Wt1 antisense vivo-morpholino (5′-CAGGTCCCGCACGTCGGAACCCATG-3′) or 10 μM five-base-pair-mismatched vivo-morpholino control (5′-CAGcTCCgGCACcTCGcAACCgATG-3′) (Gene Tools).

Immunofluorescence was performed using anti-WT1 (Santa Cruz) and anti-cytokeratin (Sigma) primary antibodies and anti-rabbit Texas Red and donkey anti-mouse FITC secondary antibodies (Jackson ImmunoResearch). TUNEL staining was performed with the Apoptag Plus Fluorescein in situ Apoptosis Detection Kit, as per manufacturer's instructions (Millipore). In situ hybridization was performed as previously described (Mo et al., 1997). The following probes were generated by PCR amplification (see Table S3 in the supplementary material) and subsequently cloned into the pCRII-TOPO vector (Invitrogen): HeyL (exon 5), Cxxc5 (exon 3), Lsp1 (5′UTR), Pbx2 (exon 9), Plxdc2 (5′UTR), Rps6ka3 (exon 14-20), Scx (exon 1-2) and Sox11 (3′UTR). Probes encoding Wt1 (Gao et al., 2005), Bmp7 (Lyons et al., 1995), Pax2 (Dressler et al., 1990) and Sall1 (Nishinakamura et al., 2001) have been previously described.

To quantify apoptosis, the number of TUNEL-labeled cells in the general nephron progenitor cap region was first standardized to regional surface area as previously described (Hartwig et al., 2005). This was accomplished by overlaying each digital Pax2- and TUNEL-labeled confocal image with a grid and counting the number of boxes filled by the cap region. Regional apoptosis was then quantified as the ratio of TUNEL-positive cells to the total number of filled boxes in the grid, with values presented as mean fold change in WT1-morphants versus control morphant explants (n=3). Mean differences were examined using a Student's t-test (two-tailed) and significance was taken at P<0.05.

WT1 is an essential kidney development gene expressed in nephron progenitor cells that plays a crucial role in nephron progenitor differentiation in vivo. However, a comprehensive list of the transcriptional targets of WT1 that mediate its function in vivo has yet to be described. As a first step towards identifying the transcriptional targets of WT1 in nephron progenitor cells in vivo, we performed ChIP-chip on chromatin DNA isolated from embryonic mouse kidney tissue to identify genes physically bound by WT1. Chromatin immunoprecipitated with anti-WT1 antibodies (ChIP DNA) or without ChIP (input DNA) are typically amplified by PCR in order to generate sufficient DNA for hybridization to array (~10 μg DNA per channel). However, to avoid introducing bias associated with this PCR amplification step in our experiments, we processed ~1200 E18.5 kidneys, generating sufficient ChIP DNA for array hybridization without prior PCR amplification. Applying a previously published P-value cut-off of 0.001 for significant WT1 binding (Boyer et al., 2005; Lee et al., 2006), 4953 bound sequences were identified, corresponding to 1663 bound genes (see Table S4 in the supplementary material). The defined peak regions of WT1-bound promoters were largely localized (>90%) within 2 kb of the transcriptional start site (TSS; Fig. 1B). The close proximity of these cis-elements to the TSS indicate that WT1, possibly in conjunction with other co-factors, might act to stabilize general transcription factor machinery at the core promoter elements to regulate target gene transcription, as has been shown for other transcription factors that bind proximal promoter regions (Farnham, 2009).

To interpret the biological significance of WT1 binding events in the developing kidney, we used the DAVID integrative knowledgebase (Huang da et al., 2007) to identify biological processes and molecular functions enriched in our data set. DAVID identified 64 partially overlapping functional clusters (see Table S5 in the supplementary material), which were manually organized into eight unique meta-clusters (Table 1). The most highly enriched function relates to regulation of transcription, and includes transcription factors and genes involved in chromatin establishment and modification. The second-ranking meta-cluster relates to development and differentiation in multiple tissues. In fact, Wt1 is widely expressed in many developing organs, and Wt1 loss-of-function mouse models have established a requirement for Wt1 in the development of multiple organ systems, including the kidney, gonads, heart, lungs (Kreidberg et al., 1993), spleen (Herzer et al., 1999), liver (Ijpenberg et al., 2007), diaphragm (Moore et al., 1998), nervous system, vasculature (Scholz et al., 2009), brain, eye (Wagner et al., 2002), olfactory system (Wagner et al., 2005), adrenal gland and mesothelial tissues (Moore et al., 1999). The large breadth of developmental processes enriched in WT1 target genes suggests that there might indeed be a common mode of action by which WT1 regulates differentiation and development of multiple organ systems during embryogenesis. The third-ranking meta-cluster relates to cell cycle, including regulation of cell proliferation and apoptosis. The top three functional meta-clusters enriched in WT1 targets are consistent with established functions of WT1 and thus also serve as an indicator of high data set quality.

ChIP-chip experiments identify transcription factor binding events at low resolution. To improve the resolution of our data set, we used the THEME algorithm (Macisaac et al., 2006) to identify a sequence motif distinguishing WT1-bound target sequences from unbound sequences. The top-ranked motif identified by THEME (visualized by WebLogo in Fig. 1A and referred to as the WT1 matrix site) was detected in 86.1% of WT1-bound sites versus 34.8% of unbound sites and was consistent with the previously published EGR1-like WT1 consensus sequence (Hamilton et al., 1995; Nakagama et al., 1995). To map the WT1 binding site to our bound data set, the core WT1 consensus sequence G¹C²G³(T/G)⁴(G/A)⁵G⁶G⁷(C/A)⁸G⁹(G/T)¹⁰ was used to scan genomic sequences, with defined substitutions permitted at the fourth, fifth, eighth or tenth positions (substitutions noted in parentheses). Notably, the WT1 consensus sequence occurs with higher frequency in higher-ranking WT1 target genes (presence of WT1 consensus site in WT1-bound promoters: 74% in genes enriched greater than 8-fold versus 25% in genes enriched less than 8-fold; P<0.001, Fisher's exact test). These observations indicate that high-affinity WT1 sites, as predicted by the consensus sequence, are more likely to be highly occupied in vivo. Collectively, these results indicate that the WT1 matrix site is a strong predictor of WT1 binding events in our data set.

To validate our ChIP-chip results by ChIP-PCR, we selected ~40 target genes from our array list that could potentially function downstream of WT1 in nephron progenitor cells (Table 2). Using both database-assisted (DAVID) annotations, as well as manual annotations (PubMed searches), targets were selected based on their established or potential roles (i.e. established role of other highly homologous family members) in kidney development, progenitor cell fate, organogenesis, cell signaling or transcriptional regulation. Targets were chosen from across a range of array enrichment scores (3- to 25-fold), to assist in determining an enrichment cut-off value for subsequent biological validation of our array data.

To verify WT1-specific enrichment of cis-regulatory regions identified by ChIP-chip, we adapted a standard ChIP protocol to perform ChIP using anti-WT1 antibodies in embryonic mouse kidney tissue (Lee et al., 2006). Following WT1 ChIP, enrichment of defined WT1-bound peak regions in ChIP fractions was compared with input fractions by PCR (see Fig. S2 in the supplementary material). A positive validation by ChIP-PCR was semi-quantitatively defined as a greater than 2-fold enrichment of WT1 ChIP DNA versus input DNA in each PCR reaction within the linear range of amplification, as measured by densitometry (NIS Elements statistical software; see Table S6 in the supplementary material). Numerous kidney development genes, including members of the Bmp (Bmp4, Bmp7, Smad4) and Hedgehog pathways (Smo, Hhat), as well as Vegfa, Pax2 and Sall1, were validated as WT1 targets by ChIP-PCR (Fig. 2), suggesting that they might be bona fide transcriptional targets of WT1 in nephron progenitor cells in vivo.

Notably, lower-ranking WT1 binding sites enriched less than 8-fold on array were generally not validated by ChIP-PCR (4 were validated out of 15 ChIP-PCR reactions; see Fig. S2A-O in the supplementary material; Table 2). However, ChIP-PCR confirmed WT1-specific enrichment in more than 90% of higher-ranking target genes enriched greater than 8-fold (24 validated out of 26 ChIP-PCR reactions; see Fig. S2P-O′ in the supplementary material). This ChIP-PCR validation threshold of 8-fold enrichment correlates with the inflexion point of the curve identified at ~7.75-fold enrichment (R Statistical Software), when all 1663 genes were plotted against their fold-enrichment scores (see Fig. S2P′ in the supplementary material). Interestingly, among the target genes enriched greater than 8-fold, the only 2 target genes not validated by ChIP-PCR, Pbx1 and Pbx2 (see Fig. S2W,D′ in the supplementary material), do not contain the WT1 consensus site (Table 2). In all subsequent experiments, we therefore focused on the cohort of WT1 targets exhibiting enrichment scores greater than 8-fold by array (n=202).

The absence of kidneys in Wt1^−/− mice has precluded conventional approaches of investigating WT1 function during early renal development in vivo. Therefore, to gain insight into the role of WT1 during early kidney development, we used a novel, modified antisense, ‘vivo-morpholino’ delivery system (Gene Tools) to examine the consequences of WT1 knock-down in cultured E12.5 kidney explants. Each vivo-morpholino is a fusion moiety comprising a standard morpholino covalently fused to an octaguanidium dendramer transporter at its 5′ end, permitting morpholino penetration into mouse tissue. Embryonic E12.5 mouse kidneys were cultured for 24 hours in media supplemented with WT1 antisense or 5-mismatch control vivo-morpholinos at varying dosages (10-20 μM). Anti-WT1 immunofluorescent staining was used to determine the efficacy of WT1 antisense vivo-morpholinos in blocking WT1 protein translation, and ureteric bud branching was visualized using anti-cytokeratin antibodies (Fig. 3). After 24 hours of culture, explants cultured with 10 μM control vivo-morpholino (henceforth called control morphants) had undergone extensive ureteric bud branching, similar to explants cultured in media alone (Fig. 3C,C′ versus A,A′). Control morphants exhibited characteristic WT1 expression in the cap of nephron progenitors surrounding the ureteric bud tips, as well as in pretubular aggregates, also similar to explants cultured in media alone. By contrast, WT1 expression in nephron progenitors was markedly reduced in explants treated with 10 μM WT1 vivo-morpholino (hitherto referred to as WT1 morphants; n=6; Fig. 3D,D′ versus C,C′). Decreased WT1 expression in WT1 morphants was associated with moderately reduced ureteric bud branching and overall explant size.

To examine the effects of increased WT1 morpholino concentration, we next treated explants with a high dose (20 μM) of vivo-morpholino. Control morphants treated with 20 μM control vivo-morpholino exhibited a moderate reduction in ureteric bud branch number associated with altered terminal ureteric bud tip morphology compared with explants cultured in media alone, indicative of a mild, general cytotoxic effect of vivo-morpholino treatment at this dosage, independent of effects on WT1 expression. Nevertheless, control morphants retained strong WT1 expression in nephron progenitors and epithelial tubules and underwent branching (Fig. 3E,E′). By contrast, WT1 expression was not detected in explants treated with 20 μM antisense WT1 vivo-morpholino, and loss of WT1 was associated with arrested ureteric bud branching and reduced overall explant size. In all subsequent experiments, 10 μM of WT1 vivo-morpholino was used, a dosage at which WT1 expression is reduced in nephron progenitors while maintaining the presence of progenitor cells.

Having demonstrated that WT1 expression is reduced in embryonic kidney explants treated with WT1 vivo-morpholinos, we proposed to use this vivo-morpholino system to examine whether WT1 target genes identified by array are indeed regulated by WT1. Loss of target gene expression in nephron progenitors of WT1 morphants could be attributable to a loss of specific regulation by WT1 or due to a general loss of progenitor cells. Therefore, we first determined whether nephron progenitor cells are still present in WT1 morphant explants. Explants were cultured for 24 hours with either 10 μM WT1 antisense or control vivo-morpholino (n=10) and Six2 mRNA expression was assessed by RNA in situ hybridization (Fig. 4A,A′). Six2 is a cap-specific marker of nephron progenitors surrounding ureteric bud tips (Kobayashi et al., 2008) and is downregulated in pretubular aggregates (Self et al., 2006). Six2 was identified as a WT1 target gene (enriched 3.3-fold) by ChIP-chip. However, Six2 was not validated by ChIP-PCR (Fig. 2A), and we therefore examined Six2 expression in morpholino-treated explants as a negative control. Characteristically discrete Six2 mRNA expression was detected in the cap of nephron progenitors of control morphants, tightly condensed around ureteric bud tips (Fig. 4A,A′). Six2 expression was also detected in WT1 morphant explants (Fig. 4B,B′ versus A,A′). This was confirmed by qRT-PCR for Six2, which revealed no significant difference in Six2 mRNA levels (1.1-fold higher in morphants; s.d. for WT1 morphants=0.158; s.d. for control morphants=0.136 normalized to GAPDH; n=3). However, the Six2 expression domain appeared less compact adjacent to ureteric bud tips in these explants, possibly reflecting a failure of nephron progenitors to undergo condensation in the absence of WT1. Although we are not able to rescue WT1 expression in morpholino-treated explants, to completely eliminate the possibility of a toxic effect specific to the WT1 morpholino, the continued expression of Six2 makes it unlikely that the WT1 morpholino is considerably more toxic than the control morpholino.

Having established an ex vivo model in which to modulate WT1 expression in embryonic kidney explants while maintaining the presence of nephron progenitor cells, we next determined whether expression of essential early kidney development genes identified as WT1 targets by array are indeed regulated by WT1. In contrast to Six2, WT1 targets identified by array, including Pax2 (enriched 9.3-fold), Sall1 (enriched 9.3-fold), Bmp7 (enriched 12.9-fold) and HeyL (enriched 25.1-fold), were all validated by ChIP-PCR (Fig. 2L,M,P,R) and are regulated by WT1 (Fig. 4C-J′). In control morphants, Bmp7 (Lyons et al., 1995) and Pax2 (Dressler et al., 1990) mRNA expression was strongly detected in the ureteric bud and nephron progenitors and weakly expressed in pretubular aggregates (Fig. 4C,C′,E,E′). Sall1 expression was detected predominantly in nephron progenitors and weakly in nephrogenic tubules of control morphants (Fig. 4G,G′) (Nishinakamura et al., 2001), whereas HeyL was expressed in nephron progenitors and ureteric bud cells (Leimeister et al., 2003), with strongest expression in pretubular aggregates (Fig. 4I,I′). In all cases, WT1 morphants exhibited a marked and specific reduction in target gene expression in nephron progenitors (Fig. 4D,D′,F,F′,H,H′,J,J′ versus C,C′,E,E′,G,G′,I,I′), whereas residual Bmp7 and Pax2 expression was retained in the ureteric bud.

Importantly, the demonstration that Six2-expressing nephron progenitor cells are indeed present in WT1 morphants (Fig. 4B,B′) indicates that loss of Bmp7, Pax2, Sall1 and HeyL expression in nephron progenitors is not attributable to the absence of nephron progenitor cells. Rather, WT1-deficient nephron progenitors are present, but do not express these WT1 target genes. Collectively, these findings, which recapitulate both array and ChIP-PCR results, strongly suggest that Bmp7, Pax2, Sall1 and HeyL are bona fide transcriptional targets of WT1 in embryonic kidneys. Notably, both Pax2^−/− and Sall1^−/− mice exhibit renal agenesis phenotypes. Our observations that these essential kidney development genes are regulated by WT1 suggest that loss of Pax2- and Sall1-dependent signaling in Wt1^−/− embryos might account for the renal agenic phenotype in these embryos.

WT1 vivo-morpholino treatment of embryonic kidney explants reduces expression of essential kidney development genes Bmp7, Pax2 and Sall1. Increased mesenchymal apoptosis is observed in Bmp7^−/− (Luo et al., 1995), Sall1^−/−(Nishinakamura et al., 2001) and Wt1^−/− embryos (Kreidberg et al., 1993) and it has been proposed that these genes might control renal development in part via anti-apoptotic effects. To determine whether apoptosis is altered in WT1 morphants, we performed TUNEL staining in WT1 morphant explants, together with Pax2 immunostaining to identify ureteric bud and nephron progenitor cells and visualized explants by confocal microscopy (Fig. 5). In explants treated with 10 or 20 μM control vivo-morpholino (Fig. 5A,B), the highest concentration of TUNEL-positive cells (green) was observed in the peripheral region of metanephric mesenchyme, with low numbers of TUNEL-positive cells also observed in the ureteric bud, nephron progenitors and stromal compartment. The number of apoptotic cells in the cap region of explants treated with 10 μM control vivo-morpholino was not significantly different from explants treated with 20 μM control vivo-morpholino (P=0.62, n=3). In WT1 morphant explants, Pax2 expression was reduced in the cap region (Fig. 5C,D), consistent with our previous results demonstrating reduced Pax2 mRNA expression in WT1 morphant explants (Fig. 4F,F′ versus E,E′). In WT1 morphants, the peripheral apoptotic zone was expanded towards the center of the explant and a dose-dependent increase in the number of TUNEL-positive cells is observed in the cap region of nephron progenitors at ureteric bud tips. Explants treated with 10 μM WT1 vivo-morpholino exhibited a 2.5-fold increase in TUNEL-positive cells in the cap region (P=0.03) compared with explants treated with 10 μM control morpholino (Fig. 5C versus A; n=3). Explants treated with 20 μM WT1 vivo-morpholino exhibited a 4.22-fold increase in TUNEL-positive cells in the cap region (P<0.0001) compared with explants treated with 20 μM control vivo-morpholino (Fig. 5D versus B; n=3). Thus, the loss of WT1 in WT1 morphant explants is associated with reduced expression of Bmp7, Pax2 and Sall1, together with increased apoptosis in the cap region and ureteric bud.

In a final series of proof-of-principle experiments, we attempted to identify novel kidney development WT1 target genes from our array target list that could mediate WT1 function in nephron progenitor differentiation. Of the 202 WT1 target genes enriched greater than 8-fold, a small cohort were selected based on established or potential roles in progenitor cell fate, development, or cell-cell signaling/cell migration. In E18.5 mouse kidneys, Wt1 expression is not restricted to nephron progenitors, but is also strongly detected in presumptive podocytes occupying the posterior portion of developing nephrons (Pritchard-Jones et al., 1990) and in mature podocytes (Mundlos et al., 1993). As we had performed ChIP-chip in E18.5 mouse kidneys, our array target list included genes expressed in nephron progenitors, induced nephrogenic structures and mature podocytes. In order to select genes specifically expressed in early renal development, we screened the mRNA expression pattern of selected genes using the Genepaint Mouse Expression Database (http://www.genepaint.org/index.html) (Alvarez-Bolado and Eichele, 2006) to further select for genes expressed in E14.5 kidneys. We thus identified a panel of WT1 target genes enriched greater than 8-fold and co-expressed with WT1 in nephron progenitors in embryonic E14.5 kidney tissues, including Cxxc5, Lsp1, Pbx2, Plexdc2, Rps6ka3 (Rsk2), Scx and Sox11 (Table 3).

The mRNA expression pattern of Cxxc5 (enriched 20.9-fold), Lsp1 (enriched 24.4-fold), Pbx2 (enriched 12.8-fold), Plxdc2 (enriched 10.0-fold), Rps6ka3 (enriched 11.9-fold), Scx (enriched 14.7-fold) and Sox11 (enriched 15.5-fold) in control and WT1 morphant kidney explants is shown in Fig. 6. With the exception of Plxdc2, strong WT1 target gene expression was detected in nephron progenitors of control morphant explants, as well as in other cell lineages (Fig. 6). The expression of Plxd2 appeared restricted to the outer population of metanephric mesenchyme in control morphants, a more weakly Wt1-expressing domain, and did not appear to be expressed in cap or ureteric bud cells (Fig. 6G,G). Similar to our previous observations (Fig. 5), WT1 morphants, in all cases, exhibited marked and specific reductions in target gene expression in nephron progenitors (Fig. 6) or metanephric mesenchyme (Fig. 6H,H′ versus G,G′). Collectively, these findings illustrate the predictive quality of our array output in identifying novel kidney genes expressed in nephron progenitors of the developing kidney that might act as mediators of WT1 function in nephron progenitors in vivo.

In the present study, we performed ChIP-chip in embryonic mouse kidney tissue without PCR amplification to identify transcriptional targets of WT1 in nephron progenitor cells during renal development in vivo. Using the THEME motif discovery algorithm, a single high-affinity WT1 binding motif was identified in 86% of the WT1-bound sequences (versus 34.8% of unbound sequences) on the array (Fig. 1). Consistent with a role for WT1 in transcriptional regulation during development, biological processes most highly enriched in WT1 target genes relate to transcription as well as development and differentiation of multiple tissues (Table 1). ChIP-PCR and biological methods validated WT1 targets enriched greater than 8-fold by array, a threshold value that closely corresponds to the inflexion point (~7.75) of the fold-enrichment curve for our data set (see Fig. S2P′ in the supplementary material).

Biological validation was performed using a novel vivo-morpholino WT1 loss-of-function ex vivo model (Fig. 3, Fig. 4, Fig. 5 and Fig.6). Embryonic kidney organ culture is a well-established model of both organogenesis in general and early kidney development in particular. WT1 vivo-morpholino treatment of embryonic kidney explants provides a powerful new biological method for validating WT1 candidate genes and characterizing both WT1 and WT1 target-gene function during renal development ex vivo. In a broader context, the core vivo-morpholino strategy should be applicable to the study of other genes that, similar to Wt1, exhibit early embryonic lethal organ agenesis or complex phenotypes.

Following WT1 vivo-morpholino treatment of embryonic kidney explants, RNA in situ hybridization was performed to characterize changes in renal developmental expression patterns of WT1 target genes enriched greater than 8-fold by array, including both established kidney development genes (Fig. 4), as well as genes not previously characterized in the kidney (Fig. 6). In all cases, target gene mRNA expression overlapped with the expression domain of WT1 in nephron progenitor cells and WT1 morphants uniformly exhibited decreased target gene expression in nephron progenitor cells. Collectively, these analyses suggest with considerable confidence, that WT1 target genes enriched greater than 8-fold by ChIP-chip represent bona fide WT1 transcriptional targets during renal development. Further, these results indicate a high predictive quality of our data set to identify novel WT1 target genes co-expressed with WT1 in the developing kidney that might mediate WT1 function during nephron progenitor differentiation in vivo.

The transcriptional function of WT1 has been controversial. Primarily using promoter-reporter constructs in cell culture models of WT1 gain-of-function, WT1 has been shown to bind and repress GC-rich promoters of several growth-promoting genes expressed in the kidney including Pax2 (Ryan et al., 1995), N-myc (Zhang et al., 1999) and Egfr (Englert et al., 1995). However, the biological relevance of many of these regulatory events is unknown, as GC-rich sequences are found in the promoter regions of thousands of genes, including those that have CpG islands. Moreover, major discrepancies were found when evaluating WT1 function using promoter-reporter constructs versus studies on the regulation of endogenous genes: in several cases, WT1 transfection failed to repress expression of the native gene, despite potently repressing the activity of its corresponding promoter-reporter constructs in transfection assays (Scharnhorst et al., 2001). For example, in the case of Egfr, where both endogenous gene expression and promoter activity were downregulated following WT1 transfection in Saos2 osteosarcoma cells (Englert et al., 1995), results could not be confirmed in HEK293 cells (Thate et al., 1998). Thus, WT1 transcriptional repressor activity in vitro appears to depend on promoter architecture, experimental conditions, cell line and type of expression vector (Reddy et al., 1995).

In the present study, WT1 vivo-morpholino treatment in embryonic kidney explants resulted in reduced target gene expression. These findings are consistent with the in vitro transcriptional activation function associated with the WT1 consensus site (Hamilton et al., 1995; Nakagama et al., 1995) enriched in our data set. Together with our results demonstrating physical binding of WT1 (or a WT1-containing complex) to the proximal promoter regions of these target genes, these observations indicate that WT1 can act as a transcriptional activator during embryonic renal development in vivo.

WT1 has been shown to bind several GC-rich and (TCC)n consensus sequences in vitro (Drummond et al., 1994; Rauscher et al., 1990). In fact, the last three zinc fingers of WT1 are highly homologous with the three zinc fingers of EGR1, which each recognize a 3 bp consensus sequence. WT1 binding specificity is influenced by specific nucleotide substitutions within the core 9-bp WT1 site GCGTGGGCG in vitro. For example, adenosine and thymidine substitutions in the eighth and tenth positions (GCGTGGGAGT), respectively, have been shown to confer 20- to 30-fold higher affinity for WT1 binding compared with EGR1 in vitro (Nakagama et al., 1995). The motif discovery algorithm THEME identified a WT1 matrix site that is highly enriched in WT1-bound sequences (present in 86.1% of WT1-bound sequences versus 34.8% of unbound sequences), consistent with the previously reported, EGR1-like, WT1 consensus site (Hamilton et al., 1995; Nakagama et al., 1995). This WT1 consensus site occurs with higher frequency in higher-ranking WT1 target genes on the array (present in 74% of genes enriched greater than 8-fold versus 25% in genes enriched less than 8-fold), indicating that highly occupied in vivo WT1 targets identified by ChIP are more likely to correspond to high affinity binding sites. However, we did not observe enrichment of WT1 consensus site variants containing adenosine at the eighth position and/or thymidine at the tenth position among higher-ranking targets on the array. This might, in part, be explained by additional factors affecting recruitment of WT1 to its targets, which could include structural characteristics of the binding site that are not well-modeled by simple matrix motif models.

In addition, the observation that a significant portion (14%) of WT1-bound genes lacked the WT1 matrix site altogether, indicates that additional factors independent of cis-sequence recognition probably mediates WT1 binding to target genes in vivo. It is possible that, similar to SWI/SWF-DNA- (Fairall et al., 1993; Schwabe et al., 1993) and GLI-DNA-binding events (Pavletich and Pabo, 1993; Vokes et al., 2007), WT1 zinc finger-DNA interactions might involve a combination of cis and trans DNA-binding sequences. In addition, cooperative or competitive interactions with other transcription factors, histone modifications and the chromatin structure in the vicinity of the binding site probably play an important role in bringing WT1 to its target site and increasing WT1 binding affinity in vivo.

DAVID identified eight major functional groups enriched in WT1 targets (Table 1). The three most enriched functions relate to transcription, multi-organ development and cell cycle, terms consistent with the established role of WT1 during development and in tumor biology. WT1 targets are also enriched for specific families of proteins containing conserved zinc-finger DNA-binding motifs including BZIP, BHLH, Dwarfin/CTF/NF-1, Chromo, PhD, C₂H₂ and BTP/POZ domains, as well as zinc-finger families of nuclear receptors. Non-DNA-binding protein domains including WH1/EVH1, GPCR, Frz/PDZ, WW, SH3 and pleckstrin homology domains were also enriched in WT1 target genes. That WT1 binds conserved families of proteins suggests a model wherein WT1-dependent transcriptional regulation of a small number of ancestral genes might have been retained following gene duplication events, chromosomal segregation, and other genome reorganization events that both increased gene number and diversified gene function during evolution (Pal and Hurst, 2000).

Interestingly, WT1 target genes were also enriched for actin cytoskeleton organization, actin biogenesis and binding, as well as cell adhesion, cell migration and cell-cell signaling. In fact, three of the top ten most highly enriched WT1 targets identified by array (see Table S4 in the supplementary material), namely Myo1B (enriched 27.2-fold), Lsp1 (enriched 24.5-fold) and Actn1 (enriched 23.8-fold), are expressed during early kidney development and are involved in cytoskeletal interactions. Previously, two independent transcriptome profiling reports during pre-implantation mouse development demonstrated an obligate, transient surge in expression of genes involved in actin cytoskeleton and cell-cell signaling occurring prior to the major differentiation events in early development (Hamatani et al., 2004; Mitiku and Baker, 2007). High expression of cytoskeleton-interacting genes at this juncture is thought to relate to the dramatic morphological changes and energy requirements associated with compaction, as well as the requirement of cadherin-cytoskeleton interactions for cell-to-cell contact and adhesion during compaction (Hamatani et al., 2006). Expression of cell-cell signaling genes is thought to facilitate assembly of gap and tight junctions that enable blastomeres to lose their round shape, become tightly packed together and undergo differentiation (Watson and Barcroft, 2001).

It is intriguing to hypothesize that a similar burst of expression of genes involved in cytoskeleton and cell-cell signaling might necessarily precede the major morphologic changes associated with nephron progenitor aggregation around ureteric bud tips and differentiation. In fact, isolated Wt1^−/− mesenchymal rudiments fail to aggregate in response to inductive signals from wild-type ureteric bud cells and instead undergo massive apoptosis (Donovan et al., 1999). This phenotype is distinct from Sall1^−/− rudiments, which retain their competence to respond to inducing ureteric bud signals (Nishinakamura et al., 2001), suggesting that the Wt1^−/− phenotype is not simply due to loss of Sall1 signaling. Similarly, in WT1 morphant kidney explants, WT1-deficient nephron progenitor cells fail to become tightly packed into a discrete ‘cap’ around the tips of the ureteric bud. Instead, WT1-deficient nephron progenitors are loosely arranged around the ureteric bud tips, as evidenced by a diffuse rather than cap-specific expression pattern of the nephron progenitor marker Six2 (Fig. 4B,B′ versus A,A′). It is possible that one function of WT1 during early renal development is to activate genes involved in cytoskeletal-interactions, cell-cell adhesion, migration and signaling, processes that might themselves play pivotal permissive roles in nephron progenitor differentiation. In this paradigm, loss of WT1-dependent activation of cytoskeleton-interacting genes in Wt1^−/− mice would reduce the cellular integrity and scaffolding of individual nephron progenitor cells, thus rendering them incompetent to withstand the major morphologic changes associated with aggregation and differentiation. Similarly, loss of WT1-dependent activation of cell-cell signaling genes might prevent groups of Wt1^−/− nephron progenitor cells from forming necessary cell-cell contacts. Thus, these cells would be unable to aggregate and undergo concerted group movement to form the cap of cells that surround the ureteric bud tips and undergo differentiation.

Recently, ChIP-chip location analysis was performed in immortalized CCG099-11 Wilms' tumor cells (Kim et al., 2009) and eight WT1 target genes were initially identified in three technical replicates. By overexpressing WT1 in inducible CCG-5.1 cells, the authors subsequently identified 643 promoter targets of WT1. The highest functional enrichment in WT1 target genes related to MAPK signaling, regulation of actin cytoskeleton and focal adhesion functional groups that were also observed in our data set, albeit at the lower range of enrichment. Major differences between this data set and our own might reflect differences in WT1-dependent tumor biology versus developmental function, as well as differences obtained by performing ChIP-chip in WT1-overexpressing immortalized cells versus in embryonic kidney tissues.

During renal development, WT1 is expressed in multiple nephrogenic compartments including the nephrogenic mesenchyme, nephron progenitors and developing tubules, as well as in glomerular podocytes. Whether WT1 regulates similar sets of gene targets in each of these cellular compartments, and at different stages of renal development, has not been explored. Recently, Brunskill et al. (Brunskill et al., 2008) have used laser capture microdissection and/or fluorescence-activated cell sorting followed by microarray profiling, to define a genome-wide gene expression atlas for the different cell lineages of the developing kidney at both late and early developmental time points (Brunskill et al., 2008). The integration of our WT1 ChIP-chip results with results from the kidney atlas and other gene expression databases will be a necessary and important first step towards a comprehensive description of WT1 function and gene targets at different stages of renal development in vivo.

In summary, by applying location analysis together with bioinformatics and biological approaches, we have identified WT1 target genes in embryonic mouse kidneys co-expressed with WT1 in nephron progenitor cells that could play an important role in mediating WT1 regulation of nephron progenitor differentiation in vivo. These data provide novel insights into biological processes that might be regulated by WT1, as well as the mechanisms by which WT1 binds and regulates target gene transcription in progenitor cells of the developing kidney and other developing organs in vivo.

Dr Kreidberg's laboratory is supported by the NIDDK (R01DK070901), the Joanne Webb Fund and the Harvard Stem Cell Institute. Sunny Hartwig is the recipient of a KRESCENT fellowship. The Pax2 antibody was a gift from Dr Gregory Dressler (University of Michigan). The authors acknowledge valuable discussions with members of the Kreidberg Laboratory. Deposited in PMC for release after 12 months.