Hepatocyte nuclear factor 1b suppresses canonical Wnt signaling through transcriptional repression of lymphoid enhancer–binding factor 1

Hepatocyte nuclear factor-1b (HNF-1b) is a tissue-specific transcription factor that is required for normal kidney development and renal epithelial differentiation. Mutations of HNF-1b produce congenital kidney abnormalities and inherited renal tubulopathies. Here, we show that ablation of HNF-1b in mIMCD3 renal epithelial cells results in activation of b-catenin and increased expression of lymphoid enhancer–binding factor 1 (LEF1), a downstream effector in the canonical Wnt signaling pathway. Increased expression and nuclear localization of LEF1 are also observed in cystic kidneys from Hnf1b mutant mice. Expression of dominant-negative mutant HNF-1b in mIMCD3 cells produces hyperresponsiveness to exogenous Wnt ligands, which is inhibited by siRNA-mediated knockdown of Lef1. WT HNF-1b binds to two evolutionarily conserved sites located 94 and 30 kb from the mouse Lef1 promoter. Ablation of HNF-1b decreases H3K27 trimethylation repressive marks and increases b-catenin occupancy at a site 4 kb upstream to Lef1. Mechanistically, WT HNF-1b recruits the polycomb-repressive complex 2 that catalyzes H3K27 trimethylation. Deletion of the b-catenin–binding domain of LEF1 in HNF-1b–deficient cells abolishes the increase in Lef1 transcription and decreases the expression of downstream Wnt target genes. The canonical Wnt target gene, Axin2, is also a direct transcriptional target of HNF-1b through binding to negative regulatory elements in the gene promoter. These findings demonstrate that HNF-1b regulates canonical Wnt target genes through long-range effects on histone methylation at Wnt enhancers and reveal a new mode of active transcriptional repression byHNF-1b.

Hepatocyte nuclear factor-1b (HNF-1b) is a tissue-specific transcription factor that is required for normal kidney development and renal epithelial differentiation. Mutations of HNF-1b produce congenital kidney abnormalities and inherited renal tubulopathies. Here, we show that ablation of HNF-1b in mIMCD3 renal epithelial cells results in activation of b-catenin and increased expression of lymphoid enhancer-binding factor 1 (LEF1), a downstream effector in the canonical Wnt signaling pathway. Increased expression and nuclear localization of LEF1 are also observed in cystic kidneys from Hnf1b mutant mice. Expression of dominant-negative mutant HNF-1b in mIMCD3 cells produces hyperresponsiveness to exogenous Wnt ligands, which is inhibited by siRNA-mediated knockdown of Lef1. WT HNF-1b binds to two evolutionarily conserved sites located 94 and 30 kb from the mouse Lef1 promoter. Ablation of HNF-1b decreases H3K27 trimethylation repressive marks and increases b-catenin occupancy at a site 4 kb upstream to Lef1. Mechanistically, WT HNF-1b recruits the polycomb-repressive complex 2 that catalyzes H3K27 trimethylation. Deletion of the b-catenin-binding domain of LEF1 in HNF-1b-deficient cells abolishes the increase in Lef1 transcription and decreases the expression of downstream Wnt target genes. The canonical Wnt target gene, Axin2, is also a direct transcriptional target of HNF-1b through binding to negative regulatory elements in the gene promoter. These findings demonstrate that HNF-1b regulates canonical Wnt target genes through long-range effects on histone methylation at Wnt enhancers and reveal a new mode of active transcriptional repression by HNF-1b.
Hepatocyte nuclear factor-1b (HNF-1b) is a tissue-specific, homeodomain-containing transcription factor that is expressed in epithelial organs, such as the kidney, liver, and pancreas (1). Human full-length HNF-1b consists of 557 amino acids comprising an N-terminal dimerization domain, DNA-binding POU/homeodomain, and C-terminal transactivation domain (2). HNF-1b binds to genomic DNA containing the consensus sequence (59-GTTAATNATT-AAC-39) as a homodimer or heterodimer with the paralogous protein HNF-1a and regulates gene transcription (3). HNF-1b is required for the embryonic development of the kidney where it plays important roles in branching morphogenesis, nephron patterning, and tubular epithelial differentiation (4). In the adult kidney, HNF-1b is highly expressed in all renal tubules, including proximal tubules, loops of Henle, distal tubules, and collecting ducts (5). HNF-1b activates kidney-specific gene transcription by binding to its cognate consensus sequence in the proximal promoters of genes such as Cdh16 and Pkhd1 and recruiting coactivators including Zyxin, CBP, and P/CAF (6)(7)(8)(9). Although HNF-1b has also been shown to inhibit transcription of genes such as SOCS3, the mechanism of transcriptional repression is poorly understood (10).
Mutations of HNF1B encoding human HNF-1b were first identified as a cause of maturity-onset diabetes of the young type 5 (MODY5), an autosomal dominant disorder characterized by early-onset diabetes associated with congenital kidney abnormalities (11). Subsequently, germline mutations of HNF1B have been shown to cause congenital anomalies of the kidney and urinary tract, autosomal dominant tubulointerstitial kidney disease, and hypoplastic glomerulocystic disease (4). A common phenotype of HNF1B mutations is the formation of cystic kidneys and progressive impairment in kidney function. Disturbances in renal epithelial transport produce hypokalemia, hyperuricemia, and hypomagnesemia. Extrarenal manifestations of HNF1B mutations include pancreatic abnormalities, abnormal liver function tests, cognitive impairment, and hyperparathyroidism (12). Genomewide association studies have linked HNF1B to prostate cancer, chromophobe renal cell carcinoma, and clear cell ovarian cancer (13).
To understand the pathophysiology of HNF-1b-associated kidney diseases, we have generated Hnf1b mutant mice by kidney-specific deletion of Hnf1b or transgenic expression of dominant-negative HNF-1b mutants (7,14). Similar to humans with HNF1B mutations, mutant mice develop cystic kidneys and kidney failure. Molecular analysis of mutant kidneys has revealed that HNF-1b regulates a network of cystic disease genes, including PKD2, PKHD1, KIF12, and UMOD (7,(14)(15)(16). Collecting duct-specific deletion of Hnf1b produces a more slowly progressive disease and has revealed roles of HNF-1b in urinary concentration and renal fibrosis (17). Transcriptomic analysis of Hnf1b mutant cells suggests that HNF-1b regulates a broad range of genetic networks involving cellular metabolism, cancer, and fibrosis (8,18). Recently, new functions of HNF-1b in the regulation of noncoding RNAs, cholesterol metabolism, and epithelial-mesenchymal transition have been identified (17)(18)(19).
Wnt signaling is an essential signal transduction pathway that is required for normal embryonic development and tissue homeostasis (20). Wnts are secreted glycolipoproteins that bind to cell surface receptors and signal through at least three different pathways (21). The canonical (b-catenin-dependent) pathway consists of three main steps: 1) Wnt ligands activate the Frizzled (Fzd) and low-density lipoprotein-related protein complex, 2) sequestration of the Axin-APC-GSK3b destruction complex leads to accumulation of cytosolic b-catenin, and 3) b-catenin translocates to the nucleus where it interacts with TCF/LEF1 transcription factors bound to Wnt-responsive elements (WREs) and promotes Wnt enhanceosome formation to activate or repress Wnt target genes (22). Canonical Wnt signaling plays important roles in normal kidney development and is dysregulated in kidney diseases (23). We have recently shown that ablation of HNF-1b in renal epithelial cells produces hyperresponsiveness to Wnt ligands and activation of canonical Wnt/b-catenin signaling (24). We identified one mechanism for Wnt pathway activation involving competition between HNF-1b and b-catenin/TCF/LEF1 for binding to a novel composite DNA element that is present in a subset of Wnt target genes. Loss of HNF-1b promotes b-catenin/TCF/LEF1 binding and leads to overexpression of Wnt target genes. Here, we identify a second mechanism for activation of the Wnt pathway. We show that HNF-1b functions as a transcriptional repressor of the Wnt effector LEF1 through long-range chromosomal effects on histone methylation and b-catenin binding. Ablation of HNF-1b in kidney epithelial cells leads to increased accumulation of nuclear LEF1, which contributes to Wnt hyperresponsiveness through a feed-forward loop mechanism.

Expression of Lef1 is increased in HNF-1b-deficient renal epithelial cells and HNF-1b mutant kidneys
Previous studies from our laboratory have shown that ablation of Hnf1b in kidney epithelial cells and kidney-specific knockout of Hnf1b in transgenic mice lead to activation of the canonical Wnt pathway (24). To identify Wnt pathway genes that are directly regulated by HNF-1b, we compared our previous RNA-Seq and ChIP-Seq data sets (18) and identified Lef1, a downstream effector in the canonical Wnt signaling pathway, as a potential HNF-1b target gene. To determine whether expression of Lef1 depends on HNF-1b, we performed quantitative RT-PCR analysis on mIMCD3 cells in which Hnf1b was deleted by CRISPR/Cas9 gene editing (18). Under basal conditions, the expression of Lef1 was 4.7-fold higher in HNF-1bdeficient cells compared with WT mIMCD3 cells. Treatment with exogenous Wnt3a ligand for 240 min to stimulate canonical Wnt signaling produced a 3.9-fold increase in Lef1 transcripts in WT cells and a significantly greater 37-fold increase in HNF-1b-deficient cells (Fig. 1A). Analysis of our previous RNA-Seq data (GSE130164) showed that Lef1 was the most upregulated member of the TCF/LEF1 family following treatment with Wnt3a (Fig. S1A).
Stimulation of the canonical Wnt pathway leads to cytosolic accumulation of N-terminal dephosphorylated b-catenin, which translocates to the nucleus where it binds to TCF/ LEF1 transcription factors and activates Wnt target genes (25). We performed subcellular fractionation and immunoblot analysis to measure the abundance of LEF1 protein and activated (dephosphorylated) b-catenin in HNF-1b-deficient cells. Treatment with Wnt3a produced a greater accumulation of LEF1 protein in the nuclear fraction in HNF-1b-deficient cells compared with WT cells (Fig. 1B). Treatment with Wnt3a also resulted in greater accumulation of activated b-catenin in the cytosol and nucleus of HNF-1b-deficient cells.
To validate these findings in vivo, we analyzed kidneys from Hnf1b mutant mice in which the expression of HNF-1b in renal tubular epithelial cells was ablated by Cre/loxP recombination. Previous studies have shown that kidney tubule-specific deletion of Hnf1b results in the postnatal formation of kidney cysts and leads to kidney failure (14). To measure Lef1 expression, we performed qRT-PCR analysis on kidneys at postnatal day 28 (P28), an age at which cysts are present. The levels of Lef1 transcripts were increased 1.6-fold in Hnf1b mutant kidneys compared with WT kidneys (Fig. 1C). Immunohistochemical staining showed increased LEF1 protein in the nuclei of cyst-lining epithelial cells ( Fig. 1D and Fig. S1B). Collectively, these results showed that ablation of HNF-1b in renal epithelial cells led to activation of canonical Wnt signaling associated with increased nuclear expression of LEF1.

Knockdown of Lef1 attenuates canonical Wnt signaling in renal epithelial cells expressing dominant-negative mutant HNF-1b
To confirm the above findings using a different experimental model, we measured Wnt pathway activity in cells expressing a dominant-negative mutant form of HNF-1b (DN-HNF-1b) that lacks the C-terminal activation domain but retains the dimerization and DNA-binding domain (10). This mutant is similar to a disease-causing HNF-1b mutant in humans (26). We used a stable cell line derived from mIMCD3 cells (53A) in which expression of the DN-HNF-1b mutant can be induced by treatment with mifepristone. RNA profiling showed that the expression of Lef1 was 2.1-fold higher in 53A cells treated with mifepristone compared with uninduced cells (data not shown). To measure Wnt pathway activity, 53A cells were transfected with an 83 TOPFlash reporter plasmid containing eight copies of a TCF/LEF1-binding site linked to a luciferase reporter gene. Induction of the DN-HNF-1b mutant increased both basal and Wnt3a-stimulated luciferase activity compared with uninduced cells ( Fig. 2A). Treatment with LiCl, which inhibits the phosphorylation of b-catenin by GSK3b, further increased luciferase activity. Mutation of the TCF/LEF1-binding sites in the 83 FOPFlash reporter plasmid largely abolished the stimulation of luciferase activity (Fig. 2B).

HNF-1b represses Lef1 transcription
The significantly higher luciferase activity in DN-HNF-1bexpressing cells treated with LiCl compared with uninduced cells suggested that the Wnt pathway was activated downstream to GSK3b, possibly at the level of LEF1. To test this possibility, we inhibited LEF1 expression in 53A cells using siRNA. Knockdown of LEF1 in cells expressing DN-HNF-1b reduced the magnitude of the increased luciferase activity following treatment with Wnt3a or LiCl (Fig. 2C). Collectively, these results demonstrated that the activation of the canonical Wnt pathway observed in HNF-1b mutant cells was mediated in part by increased expression of Lef1.

Identification of two HNF-1b regulatory elements at the Lef1 locus
We previously performed genomewide ChIP-Seq analysis to identify HNF-1b-binding sites in native chromatin from mIMCD3 renal epithelial cells (8). Inspection of the HNF-1b ChIP-Seq data using the IGV program revealed two novel HNF-1b-binding sites at the Lef1 locus (see Fig. 4A). One site was located 94 kb upstream to the transcription start site (TSS), and a second site was located 30 kb downstream within the third intron. DNA sequence alignment showed that the HNF-1b-binding sites were evolutionarily conserved in vertebrates (Fig. S2, A and B). Binding peaks at these locations were not detected when ChIP-Seq was performed using control IgG. To validate these findings, we performed quantitative ChIP using an antibody against HNF-1b (Fig. 3, B-D). These studies confirmed that HNF-1b binds to the 94-kb upstream site and the 30-kb downstream site in chromatin from mIMCD3 cells. In contrast, no significant binding was detected to an irrelevant Lef1 site or when ChIP was performed using isotype control IgG. As an additional test of specificity, no significant binding was detected when ChIP was performed using an HNF-1b antibody and chromatin from HNF-1b-deficient mIMCD3 cells.
To test whether the HNF-1b genomic binding sites were functional, we cloned the corresponding DNA sequences into a plasmid containing a luciferase reporter gene driven by a minimal promoter. The reporter plasmids were cotransfected with expression plasmids encoding WT or mutated HNF-1b into HeLa cells, which do not express endogenous HNF-1b. Luciferase measurements showed that reporter plasmids containing the upstream 94-kb enhancer or downstream 30-kb enhancer were transactivated by full-length HNF-1b (Fig. 3, E-G). In contrast, no transactivation was observed using a reporter plasmid containing an irrelevant Lef1 sequence. Cotransfection with an expression plasmid encoding the DN-HNF-1b mutant also failed to stimulate luciferase activity. These results indicated that both distal HNF-1b-occupied sites were transcriptionally active.

HNF-1b inhibits b-catenin occupancy and represses Lef1 transcription
In addition to its function as a downstream Wnt effector, Lef1 itself is also a well-characterized canonical Wnt target gene (27). Previous studies have suggested that the 59 UTR of the human LEF1 gene contains several WRE that are occupied by b-catenin/TCF/LEF1 and mediate transcriptional activation in response to Wnt signaling (28). We have recently performed ChIP-Seq to identify b-catenin-occupied sites in WT and HNF-1b-deficient renal epithelial cells treated with Wnt3a (24). Surprisingly, inspection of the ChIP-Seq data showed only a minor increase in b-catenin occupancy in the 59 UTR of mouse Lef1 (Fig. 4A). Instead, we identified several novel b-catenin-binding peaks located ;4-kb upstream to the Lef1 TSS.
Three evolutionarily conserved consensus TCF/LEF1-binding motifs were identified within this region (Fig. S3). Occupancy of the 4-kb upstream sites by b-catenin was strongly induced by Wnt3a in HNF-1b-deficient cells. In contrast, minimal b-catenin occupancy was observed in WT cells. Together with the results shown in Fig. 3, these findings indicate that HNF-1b normally binds to the 94-kb upstream site and the 30-kb downstream site, where it inhibits b-catenin occupancy and represses Lef1 transcription. Ablation of HNF-1b induces b-catenin occupancy at the novel 4-kb upstream site and derepresses Lef1 transcription.
To further elucidate the mechanism of transcriptional repression by HNF-1b, we compared histone methylation in WT and HNF-1b-deficient cells. Trimethylation of lysine 27 of histone H3 (H3K27me3) is a characteristic epigenetic mark of gene repression (29). HNF-1b-deficient cells and WT cells were treated with Wnt3a or vehicle, and quantitative ChIP was performed using an antibody that recognizes H3K27me3. Treatment of WT cells with Wnt3a produced a modest 22.3% reduction in H3K27me3 at the 4-kb upstream b-catenin-occupied site (Fig. 4B). In contrast, treatment of HNF-1b-deficient cells with Wnt3a produced a substantially larger 81.3% reduction in H3K27me3, consistent with derepression of Lef1 transcription in the absence of HNF-1b. Importantly, H3K27me3

HNF-1b represses Lef1 transcription
at the HNF-1b-binding site located 30 kb downstream in intron 3 was also decreased in HNF-1b-deficient cells following treatment with Wnt3a (Fig. 4C). Loss of HNF-1b did not affect global histone H3K27 trimethylation (Fig. S4A) or methylation at an irrelevant region of Lef1 (Fig. S4B), indicating that the effects on methylation at the b-cateninand HNF-1boccupied sites were specific.
Monomethylation, dimethylation, and trimethylation of lysine 27 on histone H3 (H3K27me1, H3K27me2, and H3K27me3) are mediated by the polycomb-repressive complex 2 (PRC2) (30). PRC2 comprises the core subunits SUZ12 and EED and the methyltransferase EZH2 or EZH1. To determine whether WT HNF-1b interacts with PRC2, we immunoprecipitated endogenous PRC2 subunits under chromatin cross-linked conditions and performed immunoblot analysis using an antibody against HNF-1b. Immunoprecipitation of either EZH2 or SUZ12 resulted in coprecipitation of HNF-1b (Fig.  4D). As a positive control, immunoprecipitation of PRC2 using an antibody against EZH2 resulted in coprecipitation of the SUZ12 subunit. The interaction between HNF-1b and EZH2 was specific because no coprecipitation was seen in HNF-1b-deficient cells or using control IgG. Taken together, these results suggest that HNF-1b recruits PRC2, which increases H3K27 trimethylation via its catalytic subunit EZH2, thereby generating a more repressive chromatin environment. Mutation of HNF-1b results in the loss of repressive histone marks both locally and at distal enhancers, which leads to transcriptional derepression.

Deletion of the b-catenin-binding domain of LEF1 partially rescues the hyperresponsiveness to Wnt3a
To confirm that up-regulation of Lef1 underlies the activation of the canonical Wnt pathway in HNF-1b-deficient cells, we used CRISPR/Cas9 gene editing to delete the N-terminal domain of LEF1 that is required for its interaction with b-catenin (31). Two guide RNAs (sgRNAs) complementary to sequences in exon 1 and intron 1 were cotransfected with Cas9 into HNF-1b-deficient cells to delete the b-cateninbinding domain by nonhomologous end joining (Fig. 5A). Deletion-spanning PCR confirmed the partial removal of target sequences from the Lef1 locus (Fig. 5B). We were unable to isolate clones with complete deletion of Lef1, presumably because interference with b-catenin binding confers a growth disadvantage (32). However, two independent cell lines with partial deletion of LEF1 were established (named CRISPR KD#1 and CRISPR KD#2). Importantly, the genomic deletion within the Lef1 gene did not affect the Wnt-responsive  HNF-1b represses Lef1 transcription elements located upstream to the transcription start site; therefore, expression of the deletion mutants can be used as a readout of endogenous Lef1 transcription. As shown above, the expression of Lef1 mRNA transcripts was increased in HNF-1b-deficient cells compared with WT mIMD3 cells. This increase was abolished by deletion of the b-catenininteracting domain of LEF1 (Fig. 5C). Immunoblot analysis confirmed that LEF1 protein levels were increased in Wnt3atreated HNF-1b-deficient cells compared with WT mIMCD3 cells. Deletion of the b-catenin-interacting domain reduced LEF1 protein to WT levels and abolished the increase in response to Wnt3a (Fig. 5D). These findings indicated that the increased expression of LEF1 in HNF-1b-deficient cells resulted from a positive feedback loop that was blocked by deletion of the b-catenin-binding domain. We have previously shown that ablation of HNF-1b increases the expression of canonical Wnt target genes, including Axin2, Rnf43, and Ccdc80 (24). Fig. 5 (E-G) shows that concomitant deletion of the b-catenin-binding domain of LEF1 reduced the Wnt3ainduced overexpression of Axin2, Ccdc80, and Sp5 in HNF-1b-deficient cells. Taken together, these findings demonstrated that Wnt hyperresponsiveness in HNF-1b-deficient cells was partially dependent on LEF1 and its interaction with b-catenin.

HNF-1b binds to Wnt negative regulatory elements in the Axin2 gene
Axin2 is the most highly up-regulated gene transcript in Wnt3a-treated HNF-1b-deficient cells (24). Expression of Axin2 is regulated by multiple WRE located upstream to the translation start site (33). In addition, two 11-bp Wnt negative regulatory elements (NREs) have recently been identified in the Axin2 distal promoter (34). Using ChIP-Seq, we identified HNF-1b-binding peaks in the distal promoter and first intron of Axin2 (Fig. S5A). ChIP-Seq using antibodies against b-catenin showed occupancy of multiple sites upstream to the translation start site, consistent with previous work (33). Occupancy by b-catenin was induced by Wnt3a and was only modestly higher in HNF-1b-deficient cells compared with WT cells (Fig.  S5A). These results suggested that HNF-1b may bind to two regulatory regions to repress Axin2 transcription largely independent of b-catenin/TCF/LEF1 occupancy (Fig. 6A). Consistent with this notion, treatment of WT cells with Wnt3a produced a slight 9.2% reduction in histone H3K27me3 at the Axin2 promoter, whereas similar treatment of HNF-1b-deficient cells produced a greater 56.7% reduction in H3K27me3 (Fig. S5B). The less repressive chromatin at the Axin2 promoter enables a higher level of transcription in the absence of HNF-1b.
Detailed inspection of DNA sequences underlying the HNF-1b-binding peak in the distal promoter of Axin2 revealed two canonical HNF-1b consensus motifs (named BSI and BSII). Both binding sites were evolutionarily conserved in higher vertebrates (Fig. S6, A and B). Surprisingly, the consensus HNF-1b sequences contained the NREs (underlined in red in Fig. 6A) that had previously been identified in the Axin2 distal promoter (42). These results suggested that HNF-1b might bind to the NREs in the Axin2 distal promoter. To test this possibility, we performed electrophoretic mobility shift assays (EMSAs) (Fig.  6, B-F). Incubation of IRD700-labeled DNA probes containing BSI or BSII with reticulocyte lysates programmed with HNF-1b produced retarded bands (Fig. 6, B and C). No retarded bands were seen using lysates programmed with LEF1, TCF-1, or TCF-4. In contrast, EMSA using a probe corresponding to the b-catenin/LEF peak in the first intron showed binding to LEF1, TCF-1, and TCF-4 but no binding to HNF-1b (Fig. 6D). Binding to a known HNF-1b site in the Pkhd1 promoter was used as a positive control (Fig. 6E). Taken together, these results identify HNF-1b as a transcriptional repressor that binds to the NREs in the Axin2 locus.

Discussion
The transcription factor HNF-1b is required for the normal embryonic development and differentiation of kidney tubular epithelial cells. Mutations of human HNF1B produce a spectrum of kidney abnormalities including cystic kidney disease, renal fibrosis, renal agenesis/hypoplasia/dysplasia, and inherited renal tubulopathies (4,5). We recently identified a new HNF-1b represses Lef1 transcription role of HNF-1b in the suppression of canonical Wnt signaling (24). Ablation of HNF-1b in renal epithelial cells leads to hyperresponsiveness to Wnt ligands and increased expression of Wnt target genes. One mechanism of activation involves reciprocal binding of HNF-1b and b-catenin/TCF/LEF1 to genomic DNA. We identified a novel composite DNA element, present in a subset of Wnt target genes, to which HNF-1b and b-catenin/TCF/LEF1 bind competitively. Loss of HNF-1b leads to disinhibition of b-catenin/TCF/LEF1 binding and enables activation of Wnt target genes. Because the composite DNA element was present in only a subset of Wnt target genes, additional activation mechanisms must exist.
To further understand how HNF-1b suppresses the canonical Wnt pathway, we examined our previous genome-wide RNA-Seq and ChIP-Seq data sets (8,18,19,24) and identified Lef1 as a novel transcriptional target of HNF-1b. LEF1 belongs to the TCF/LEF1 family of high-mobility-group domain-containing transcription factors that regulate the expression of canonical Wnt target genes (35). LEF1 contains an N-terminal domain that binds b-catenin and a C-terminal transactivation domain. Expression of LEF1 is normally repressed in epithelial cells and up-regulated in mesenchymal cells (36). We found that the expression of Lef1 was abnormally up-regulated in HNF-1b mutant renal epithelial cells. Analysis of HNF-1b mutant mice confirmed that Lef1 was also up-regulated in vivo. Importantly, knockdown of Lef1 with siRNA or genetic deletion of Lef1 in HNF-1b mutant cells reduced Wnt-dependent promoter activity and reduced the overexpression of Wnt target genes. Collectively, these findings demonstrated that HNF-1b functions as a transcriptional repressor of Lef1 and that derepression Lef1 plays a key role in the activation of canonical Wnt signaling in HNF-1b mutant cells.
Overexpression of LEF1 has been shown to stimulate cell proliferation, dedifferentiation, epithelial-mesenchymal transition, and tissue fibrosis (37,38). Because these phenotypes are also seen in Hnf1b mutant kidneys, activation of LEF1 may contribute importantly to the pathogenesis of HNF-1b-related kidney diseases. Expression of LEF1 is also up-regulated in kidney cysts from humans with polycystic kidney disease caused by mutations of PKD1, suggesting that up-regulation of Lef1 observed in Hnf1b mutant mice may be common to other cystic kidney diseases (39). Deletion of the b-cateninbinding domain of LEF1 abolished a positive feedback loop that regulates Lef1 expression and leads to Wnt hyperresponsiveness in HNF-1b-deficient cells. Although Wnt/b-catenin signaling is needed for normal kidney development and promotes cell survival and tubular regeneration after acute kidney injury, sustained activation of canonical Wnt signaling is pathogenic and leads to renal fibrosis and chronic kidney disease (40,41). The derepression of the positive feedback loop described here may represent one mechanism for sustained activation of the pathway.
Mechanistically, we found that HNF-1b binds to two sites located 94 kb upstream and 30 kb downstream of the TSS of Lef1. The binding of HNF-1b at these two genomic loci may induce a more repressive chromatin conformation that inhibits Wnt-mediated b-catenin/TCF/LEF1 binding at a site located 4 kb upstream of the Lef1 TSS. Consistent with this model, we detected lower histone H3K27 trimethylation and increased b-catenin/TCF/LEF1 binding at the 4-kb upstream site in Wnt-treated HNF-1b-deficient cells compared with WT cells. Because HNF-1b binding was not detected at the 4-kb upstream site, the effect of loss of HNF-1b on b-catenin/TCF/ LEF1 binding and H3K27 trimethylation at the site must be mediated through long-range chromosomal effects. Histone H3K27 trimethylation is a repressive epigenetic mark that is generated exclusively by EZH2, a component of the polycombrepressive complex PRC2 (29). Immunoprecipitation studies showed that HNF-1b interacts with the core PRC2 subunits EZH2 and SUZ12. These results suggest that HNF-1b represses Lef1 transcription by inhibiting proximal promoter b-cateninmediated WRE enhanceosome formation through recruitment of PRC2 and induction of H3K27me3 marks (Fig. 4E). These studies highlight a new mode of transcriptional repression by HNF-1b.
The novel mechanism for HNF-1b-dependent transcriptional repression of Lef1 was also identified at the Axin2 locus. Axin 2 is a well-established canonical Wnt target gene and the most highly up-regulated gene in Wnt3a-treated HNF-1b mutant cells (24). The 5´end of Axin2 contains multiple wellcharacterized WREs that mediate transcriptional activation and are bound by b-catenin/TCF/LEF1 (33). In addition, a recent study using b-catenin ChIP-Seq in colon cancer cells identified an 11-bp NRE that binds to b-catenin/TCF/LEF1 and mediates transcriptional repression (34). Surprisingly, the two 11-bp NRE sequences identified in the Axin2 distal promoter are located at the same positions as two HNF-1b-binding sites identified by ChIP-Seq in mIMCD3 cells ( Fig. 6A and Fig. S5A). EMSA confirmed that the 11-bp NRE sequences in the Axin2 distal promoter can form DNA-protein complexes with HNF-1b but not with LEF-1, TCF-1, or TCF-4. Our results suggest that the NREs at the Axin2 locus actually represent binding sites for HNF-1b rather than b-catenin/TCF/ LEF1. Loss of HNF-1b results in only a modest increase in occupancy of the Axin2 promoter by b-catenin, which by itself would not explain the substantial increase in Axin2 expression. Instead, our studies suggest that binding of HNF-1b to NREs mediates long-range transcriptional repression, lending support to the mechanism we identified in Lef1. Moreover, we detected a larger reduction in histone H3K27 trimethylation at the Axin2 promoter in HNF-1b-deficient cells compared with WT cells, indicating that the loss of HNF-1b binding to NREs generated a less repressive chromatin state. Because similar NREs have been identified in other Wnt target genes (34), this mechanism may regulate expression of these genes as well. The regulation of Axin2 by both direct repression by HNF-1b and activation by LEF1, which is itself repressed by HNF-1b, represents an example of a coherent feed-forward loop (FFL), a frequent motif in transcription networks (42). A coherent FFL can accelerate changes in expression of target genes in response to inputs from the transcription factors, HNF-1b and LEF1 in this case. The frequent colocalization of HNF-1b and b-catenin/TCF/LEF1-binding sites that we have previously found in the mouse genome (24) suggests that the FFL described here may represent a common mechanism for regulation of Wnt target genes.

Experimental procedures
Transgenic mice Hnf1b f/f mice containing loxP sites flanking exon 1 of Hnf1b and Ksp/Cre mice that express Cre recombinase under the control of the Ksp-cadherin (Cdh16) promoter have been described previously (14). Hnf1b f/1 mice were crossed with Ksp/Cre mice, and the bitransgenic progeny were bred to generate Ksp/Cre;Hnf1b f/f mice (HNF-1b mutant mice). Cre-negative or Ksp/Cre;Hnf1b f/1 littermates were used as negative controls. Mice of both sexes were used for the experiments.

Cell culture
HeLa cells and mIMCD3 cells were obtained from ATCC and maintained in Dulbecco's modified Eagle's medium supplemented with 10% FBS and antibiotics (100 units/ml penicillin and 100 mg/ml streptomycin). The cells were cultured at 37°C under 5% CO 2 .

CRISPR/Cas9 gene editing
HNF-1b mutant cells were seeded on 6-well plates (Corning) at a density of 200,000 cells/well, expanded overnight, and then transfected when 60-70% confluency with 2 mg of vector encoding Cas9 and U6-sgRNA (PX459, Addgene#48139) using Lipofectamine 3000 (Life Technologies) according to the manufacturer's directions. The Lef1 sgRNA sequences are shown in Table S1. After incubation for 72 h, transfected cells were lysed using QuickExtract DNA extraction kits (Epicentre) according to the manufacturer's directions. Genomic DNA was amplified using PCR kits (Quanta Biosciences). The sequences of PCR primers are shown in Table S2.

Quantitative real-time PCR
Quantitative real-time PCR was performed as described previously (18). Briefly, total RNA was extracted from kidney epithelial cells or kidneys using TRIzol (Invitrogen) or RNeasy mini kits (Qiagen), respectively, according to the manufacturers' directions. cDNA was synthesized using a cDNA synthesis kit (Roche), and quantitative real-time PCR was performed with PerfeCTa SYBR Green FastMix (Quanta Biosciences) using a CFX Connect real-time system (Bio-Rad). Gene expression levels in cell lines or whole kidneys were normalized to 18S rRNA and b2 microglobulin, respectively. Fold changes in mRNA expression levels were calculated using the comparative Ct method as described (43). Primers used for qRT-PCR are shown in Table S3.

ChIP
Quantitative ChIP was performed as described previously (43). The following antibodies were used: anti-HNF-1b (Santa Cruz, catalog no. sc-22840) and anti-tri-methyl-histone H3K27 (Cell Signaling, catalog no. 9377S). Isotype-specific IgG (Santa Cruz, catalog no. sc-2027) was used as a negative control. ChIP-enriched DNA was quantified by real-time PCR using PerfeCTa SYBR Green FastMix (Quanta Bioscien-ces). The sequences of the gene-specific primers are shown in Table S4.

Immunoprecipitation
The cells were seeded in 10-cm dishes at a density of 5 3 10 6 /dish and grown to 80-90% confluency. The cells were cross-linked with 1% paraformaldehyde for 10 min at room temperature, neutralized with 0.125 M glycine for 5 min, and washed in PBS. Nuclear protein lysates were collected and immunoprecipitated using the Simple ChIP enzymatic chromatin IP kit (Cell Signaling, catalog no. 9003S). The following antibodies were used for immunoprecipitation: anti-EZH2 (Cell Signaling, catalog no. 5246) and anti-SUZ12 (Cell Signaling, catalog no. 3737). Isotype-specific IgG (Cell Signaling, catalog no. 2729) was used as a negative control. Instead of eluting DNA, immunoprecipitated protein complexes were directly eluted with 20 ml of Laemmli sample buffer and heated at 95°C for 10 min. All eluted proteins were subjected to immunoblot analysis using anti-HNF-1b antibody (Abcam, catalog no. ab128912) as primary antibody and horseradish peroxidase (HRP)-conjugated mouse anti-rabbit IgG as secondary antibody (Cell Signaling, catalog no. 93702). The blot was then stripped and reprobed with anti-SUZ12 primary antibody (Cell Signaling, catalog no. 3737).

Immunoblot
The cells were lysed in Laemmli buffer and subjected to immunoblot analysis as described previously (44). Immunoblots were incubated with primary antibodies overnight at 4°C and then incubated with the appropriate HRP-conjugated secondary antibodies at room temperature for 1 h. Immunoblots were developed by incubation with chemiluminescence reagent (Amersham Biosciences) and exposed to X-ray film. The antibodies are listed in Table S5.

Subcellular fractionation
Subcellular fractionation was performed as described previously (44). Briefly, the cells were seeded in 6-cm dishes at a cell density of 5 3 10 5 /dish. The cells were grown to 80-90% confluency and treated with vehicle or 100 ng/ml Wnt3a (R&D, catalog no. 1324-WN) in Dulbecco's modified Eagle's medium containing 10% FBS for 16 h. The cells were washed in PBS, harvested in hypotonic buffer (250 mM sucrose, 20 mM HEPES, pH 7.4, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA), and incubated for 20 min on ice. The cells were lysed by 20 passages through 25-gauge needles. The cytosolic fraction (supernatant) was collected by centrifugation at 720 3 g for 5 min at 4°C. The nuclear pellet was washed twice by resuspending in 500 ml of hypotonic buffer followed by 20 passages through 25-gauge needles. The nuclei were pelleted by centrifugation at 3000 3 g for 10 min at 4°C. Isolated cytosolic and nuclear fractions were resuspended in Laemmli buffer, boiled for 5 min, and then subjected to immunoblot analysis.

HNF-1b represses Lef1 transcription Immunohistochemical staining
Immunohistochemical staining was performed as described previously (18). The mice were euthanized, and kidneys were fixed by perfusion and incubation in 4% paraformaldehyde for 48 h. Tissues were embedded in paraffin and sectioned at 4 mm. The slides were incubated in heat-induced antigen retrieval 13 Reveal Decloaker (Biocare Medica, RV100M) for 30 min, treated with 3% hydrogen peroxide for 10 min, and blocked in 100% rodent block M (Biocare Medical, catalog no. RBM961) for 1 h. LEF1 primary antibody (Sigma, catalog no. HPA002087 1: 200) was added in 10% rodent block and incubated overnight at 4°C. The slides were incubated in Rabbit-on-Rodent HRPpolymer (Biocare Medical, catalog no. RMR622) for 30 min and then treated with DAB (Biolegend, catalog nos. 926603 for buffer A and 926503 for buffer B) for 5 min. The sections were counterstained with hematoxylin. The images were captured using a Leica DM5500 B upright microscope with DFC7000 T camera.

Luciferase reporter gene assays
Luciferase reporter assays were performed as described previously (24). 83 TOPFlash and 83 FOPFlash luciferase reporters were obtained from Dr. Randall Moon's laboratory. 53A cells were transfected with 20 nM siRNA against mouse Lef1 (Dharmacon On-Targetplus siRNA) for 36 h and stimulated with 1 mM mifepristone or ethanol (vehicle) as a control. The cells were incubated for an additional 16 h before harvesting. Genomic DNA regions flanking the Lef1 upstream 94-kb enhancer and downstream 30-kb enhancer were amplified from mIMCD3 genomic DNA by PCR and cloned into pGL4.23 vector (Promega). PCR primer sequences are shown in Table S6. Expression vectors encoding full-length HNF-1b and DN-HNF-1b have been described previously (19). Luciferase reporters containing Lef1 enhancer DNA sequences, SV40-Renilla (Promega), and cytomegalovirus-driven full-length HNF-1b and DN-HNF-1b were cotransfected into Hela cells for 36 h before harvesting. The cells were lysed, and Dual Luciferase assays were performed according to the manufacturer's recommendations (Promega). Photinus (firefly) luciferase activity was normalized to Renilla luciferase activity.
EMSA EMSA was performed as described previously (43). Briefly, IRD700-labled duplex DNA probes were prepared by annealing complementary synthetic DNA oligonucleotides. T7-driven expression vectors encoding full-length HNF-1b and Lef1 have been described previously (24). T7 promoter-driven expression constructs pcDNA-HA-TCF1 (TCF7, catalog no. 40620) and pcDNA-myc-TCF4 (TCF7L2, catalog no. 16512) were obtained from Addgene. Binding reactions were carried out with 2 ml of TNT lysate and 2.5 nM IRD700-labeled EMSA probe using an Odyssey IR EMSA kit (LI-COR catalog no. 829-07910) according to the manufacturer's directions. Binding reactions were resolved in 4% polyacrylamide gels with Trisglycine electrophoresis buffer (50 mM Tris, 380 mM glycine, 2 mM EDTA). The gels were scanned in an Odyssey scanner (LI-COR) at 700 nm. The images were recorded and quantified using Image Studio software (LI-COR). The sequences of IR700-labelled EMSA oligonucleotides are shown in Table S7.

Statistical analysis
The data shown are the means 6 S.E. or means 6 S.D. as indicated. Student's two-tailed unpaired t test was used for pairwise comparisons. One-way analysis of variance followed by Tukey's test was used for multiple comparisons. p , 0.05 was considered significant.

Study approval
All experiments involving animals were approved by the Institutional Animal Care and Use Committee at the University of Minnesota.

Data availability
All analyzed data are included in this article. Raw data, extra information, reagents, and plasmids are available upon request from the first author, Dr. Siu Chiu Chan (University of Minnesota Medical School, Minneapolis, Minnesota, USA). The ChIP-Seq and RNA-Seq data sets were previously deposited in the Gene Expression Omnibus database under accession numbers GSE71250 (HNF-1b ChIP-Seq) and GSE130164 (b-catenin ChIP-Seq and RNA-Seq).