FOXD1 promotes nephron progenitor differentiation by repressing decorin in the embryonic kidney

Forkhead transcription factors regulate fundamental biological functions, including fertility, metabolism and immunity (Behr et al., 2007; Fontenot et al., 2003; Nakae et al., 2002; Shih et al., 1999). During embryogenesis, they are required for node and notochord formation and organogenesis of multiple systems such as liver, lung, kidney and central nervous system (Ang and Rossant, 1994; Hatini et al., 1996; Lai et al., 2001; Lee et al., 2005; Shu et al., 2005; Wan et al., 2005). This profoundly important and evolutionarily conserved family of transcription factors is defined by similarity of the DNA-binding forkhead domain (Hannenhalli and Kaestner, 2009). DNA-binding motifs are similar for many forkhead transcription factors and functional diversity is determined by flanking domains that interact with distinct co-factors (Gajiwala and Burley, 2000; Overdier et al., 1994; Pierrou et al., 1994). In vivo transcriptional targets and mechanisms of regulation have not yet been determined for the majority of forkhead transcription factors, although many family members are known to associate with the Groucho co-repressor (Yaklichkin et al., 2007a; Yaklichkin et al., 2007b). In this article, we focus on understanding the mechanism of action of FOXD1, which is required for kidney development (Hatini et al., 1996; Levinson et al., 2005).

In the developing mouse kidney, Foxd1 expression is restricted to cortical interstitial cells, which give rise to glomerular mesangial cells and the interstitium of the mature kidney (Hatini et al., 1996; Humphreys et al., 2010). Importantly, this lineage contributes to scarring in experimental chronic kidney injury, and gene expression analysis indicates that Foxd1 may be dynamically regulated in kidney injury and repair (Humphreys et al., 2010). In the developing mouse metanephros, ablation of cortical interstitium and inactivation of Foxd1 both result in accumulation of undifferentiated nephron progenitor cells (cap mesenchyme), demonstrating an essential role of the interstitial cell niche in regulating progenitor cell differentiation (Das et al., 2013; Hatini et al., 1996; Levinson et al., 2005).

Cap mesenchyme cells are arranged in a series of compartments (Mugford et al., 2009). Transition from the CITED1+ SIX2+ compartment to the CITED1-SIX2+ compartment sensitizes them to the inductive effects of canonical WNT signaling (Brown et al., 2013). This transition between progenitor compartments depends on SMAD-mediated bone morphogenetic protein (BMP) signaling. We find that the majority of cap mesenchyme cells in Foxd1^-/- kidneys remain in the CITED1+ compartment, indicating that the initiating step in nephron progenitor differentiation is inhibited. We performed comparative transcriptome analyses of Foxd1^-/- versus wild-type kidney tissue to identify direct FOXD1 transcriptional targets that may underlie this interesting compartmental skewing, identifying five candidates for which regulation by FOXD1 could be validated in cell culture. The small leucine-rich proteoglycan decorin (DCN) was of particular interest because of its role in modulating growth factor signaling (Iozzo and Schaefer, 2010). To test its role in the Foxd1^-/- phenotype, we generated Foxd1;Dcn compound mutant mice. Dcn inactivation partially reversed the blockage in differentiation of CITED1+ progenitors. SMAD-mediated BMP signaling increases in Foxd1^-/-;Dcn^-/- kidneys, suggesting that elevated DCN reduces SMAD-dependent transition of cap mesenchyme cells out of the CITED1+ state in the Foxd1^-/-. Indeed, treatment of primary renal progenitor cells with recombinant DCN blocks BMP-mediated transition of cap mesenchyme cells out of the CITED1+ state and inhibits epithelial induction. These findings provide a novel mechanism for the skewing of progenitor compartments in the Foxd1^-/- whereby loss of FOXD1 leads to de-repression of Dcn in cortical interstitial cells, resulting in DCN accumulation in the nephrogenic zone, which blocks the differentiation of CITED1+ cap mesenchyme cells.

The Foxd1^-/- kidney lacks epithelial differentiation at embryonic day (E) 12.5-15.5, with dramatic expansion of the PAX2+ cap mesenchyme surrounding mislocalized collecting duct (CD) tips (Hatini et al., 1996; Levinson et al., 2005). Cells within the cap mesenchyme are subdivided into distinct compartments (Brown et al., 2013; Mugford et al., 2009). Functional analyses indicate that the compartment expressing CITED1 and SIX2 is refractory to WNT-mediated epithelial induction by the CD, whereas the more distal compartment that loses CITED1 while maintaining SIX2 is sensitized to WNT-mediated induction (Brown et al., 2013). It is not known in which of these compartments progenitor cells are retained in the Foxd1^-/-, or if they maintain a normal distribution across compartments. Therefore, we compared expression of CITED1 and SIX2 between wild-type and Foxd1^-/- kidneys at E15.5. CITED1 localizes to cap mesenchyme adjacent to the CD tips at the cortex of wild-type kidneys (Fig. 1A). By contrast, CITED1 is expressed in large mesenchymal cell aggregates surrounding CDs throughout the kidney in mutants (Fig. 1B). SIX2 is expressed throughout the cap mesenchyme and pretubular aggregates of wild-type kidneys. In the mutant, mesenchymal clusters surrounding CD tips are SIX2+ (Fig. 1D). These data show that nephron progenitor cells in the Foxd1^-/- are maintained in the CITED1+ state (Fig. 1C).

During normal development, cortical interstitial cells of the Foxd1 lineage lose expression of Foxd1 and differentiate to medullary interstitium and pericytes, including mesangial cells (Humphreys et al., 2010). However, the Foxd1 locus remains active in interstitial cells of the Foxd1^-/- (Levinson et al., 2005). To determine whether cortical interstitial cells are blocked in their differentiation similar to cap mesenchyme, we assessed localization of PDGFRβ, which is expressed in medullary interstitium and pericytes (Lindahl et al., 1998) (Fig. 1E). Intriguingly, interstitial cells in the Foxd1^-/- are PDGFRβ+ irrespective of their cortical or medullary location (Fig. 1F), indicating that these cells are not blocked in their differentiation, but adopt a medullary interstitial cell/pericyte fate.

To explore which genes FOXD1 regulates in the developing kidney, we conducted a transcriptome analysis comparing Foxd1^-/- and wild-type kidneys using the Affymetrix Mouse Gene 1.0 ST Array. We analyzed RNA from E14.5 Foxd1^-/- and wild-type littermate kidneys from three biological replicates each containing pooled RNA from two pairs of whole embryonic kidneys from different litters. Data were normalized using XRAY (Biotique Systems). As an initial quality control, we generated a list of differentially expressed candidates (supplementary material Table S2). Genes exceeding 1.5-fold or -1.5-fold difference with a P value <0.05 were considered for further analysis. qRT-PCR was performed on seven candidates using RNA from independent kidney samples, and fold changes were compared with those detected in the microarray analysis (Fig. 2A). All seven transcripts showed an identical directionality of change and a similar fold change by microarray and qRT-PCR, demonstrating the reproducibility of the microarray data in an independent assay.

Foxd1 is expressed in cortical interstitial cells of wild-type kidneys and is downregulated as they differentiate to medullary interstitial cells (Hatini et al., 1996). FOXD1-regulated genes will therefore be differentially expressed in cortical versus medullary interstitial cells, and comparing these two cell types should yield a gene list enriched for FOXD1 targets. We generated a list of 2708 genes by comparing cortical and medullary interstitial cell datasets from GUDMAP (Harding et al., 2011; McMahon et al., 2008) (Fig. 2B). This was subsequently compared with the list of genes differentially regulated between Foxd1^-/- and wild-type kidneys. Forty-seven overlapping genes were identified and, to further enrich for direct FOXD1 targets, were screened for FOXD1-binding sites within 5 kb upstream of the promoter using Genomatix MatInspector. Five genes containing predicted FOXD1-binding sites were identified: decorin (Dcn), zinc finger E-box binding homeobox 2 (Zeb2), latent transforming growth factor beta binding protein 1 (Ltbp1), discoidin domain receptor family member 2 (Ddr2) and collagen type XIV alpha 1 (Col14a1). To test the ability of FOXD1 to modulate target gene transcription, we compared gene expression in cells transfected with either the Foxd1 expression construct CMV-BF2 or with pCX-eGFP. We chose to perform this analysis in MES13 cells because as mesangial cells, they derive from the Foxd1-expressing cortical interstitium. Although mesangial cells do not express Foxd1 in vivo, qRT-PCR analysis reveals that MES13 cells do express Foxd1 at a low level, suggesting that they may represent a partially de-differentiated state (supplementary material Fig. S1). We therefore expect genes characteristic of cortical interstitium to be expressed at some level in these cells, and FOXD1 to be able to regulate its targets. We used qRT-PCR to assess the gene expression changes resulting from increased Foxd1 expression in these cells. Expression of Dcn, Zeb2 and Ddr2 is significantly reduced upon transfection with Foxd1 (P<0.05) (Fig. 2C), whereas Col14a1 and Ltbp1 levels are essentially unchanged. FOXD1 transfection does not affect expression of the mesangial marker Pdgfrb, indicating that gene changes are not due to further de-differentiation.

We focused on Dcn because it is both the most highly upregulated of our candidates with a 2.49-fold increase in Foxd1^-/- kidneys and because its transcription is the most strongly inhibited by FOXD1 in MES13 cells (Fig. 2B,C). Furthermore, comparison of GUDMAP transcriptional profiles of cortical versus medullary interstitial cells shows that Dcn expression is elevated 4.9-fold in medullary interstitium, concomitant with loss of Foxd1 expression (Fig. 2B). To more stringently define Dcn regulatory elements, we compared mouse-human orthology of predicted FOXD1-binding sites using the rVista ECR browser. Dcn contains two conserved FOXD1-binding sites, and is therefore a strong candidate for direct FOXD1 regulation.

To confirm that Dcn is misexpressed in interstitial cells in vivo, we performed in situ hybridization analysis on Foxd1^-/- and wild-type tissues using a probe against Dcn. In the wild-type kidney at E15.5, Dcn is expressed at low levels in the cortical interstitium/capsule and at higher levels in the medullary interstitium (Fig. 3A,B). In the E15.5 Foxd1^-/- kidney, the Dcn expression domain is dramatically increased. The entire expanded interstitium surrounding the kidneys expresses high levels of Dcn, as does the interstitium surrounding the mislocalized regions of cap mesenchyme (Fig. 3C,D). Neither cap mesenchyme nor the CDs express Dcn in either wild-type or Foxd1^-/- kidneys. Importantly, in the Foxd1^-/- kidney, Dcn-expressing cells surround all of the CITED1+ progenitor cell clusters, whereas in the wild-type kidney these cells are never adjacent to Dcn expression.

We then performed immunostaining on whole E14.5 Foxd1^-/- and wild-type kidneys with an anti-DCN antibody to confirm that the transcriptional misregulation translates into aberrant DCN protein expression. In the wild-type kidney (Fig. 3E), DCN localizes primarily to the medullary interstitium. Little to no DCN can be detected in the cortex, or within glomeruli. By contrast, DCN localizes throughout the Foxd1^-/- kidney (Fig. 3F). Thus, in the E15.5 Foxd1^-/- kidney, Dcn is misregulated in cortical interstitial cells lacking Foxd1. Progenitor differentiation blockage is partially relieved by E17.5 in Foxd1^-/- kidneys (Levinson et al., 2005). This correlates with a decrease in aberrant Dcn expression. Misexpression of Dcn is largely confined to the cortex at E17.5, and only a modest number of CITED1+ progenitors abut Dcn-expressing cells (supplementary material Fig. S2). DCN associates with collagen I and is predicted to bind collagen IV, modulating their conformation (Fleischmajer et al., 1991; Parkin et al., 2011; Schönherr et al., 1995). We next investigated whether collagens also mislocalize in Foxd1^-/- kidneys. In wild-type kidneys, both collagens I and IV localize to epithelial basement membranes. Cortical interstitium contains collagen I but not IV (Fig. 3G,H). In Foxd1^-/- kidneys, collagen I and IV are localized similarly to wild type in epithelial basement membrane, but the interstitium contains both collagens I and IV; thus, unlike wild type, nephron progenitor cells in the mutant reside in a collagen IV-rich environment (Fig. 3I,J).

The data described above led us to investigate if FOXD1 directly represses Dcn. Our genome scans identified two species-conserved FOXD1-binding sites, one 4.9 kb upstream of the UCSC genome browser-predicted Dcn transcription start site and one 17.5 kb downstream of the 3′UTR. The 5′ FOXD1-binding site is CCAAAGTCAACAGG and differs from the TRANSFAC consensus FOXD1-binding site (CTWAWGTAAACANWG) at only three residues, and at none of the five 100% conserved bases (bold) (Pierrou et al. 1994). The 3′ binding site sequence is GATAAATAAACAGT, and it differs from the consensus site at four bases and none of the 100% conserved bases (Fig. 4A,B). To confirm that FOXD1 binds the endogenous sites, we performed a chromatin immunoprecipitation (ChIP) assay using MES13 cells that had been transfected with a triple-FLAG-tagged Foxd1 expression construct (3xFLAG-FOXD1). 3xFLAG-FOXD1-associated chromatin complexes were compared with complexes from pCX-eGFP transfected cells, and assayed for enrichment of the FOXD1-binding sites in the Dcn locus using PCR. There was a sevenfold enrichment of the 5′ binding site in 3xFLAG-FOXD1-transfected cells relative to GFP-transfected control cells and a 2.5-fold enrichment of the 3′ binding site (Fig. 4C). From these data, we conclude that FOXD1 physically associates at the Dcn locus. Our rVista analysis of the 5′ binding site predicted three orthologously conserved elements, each containing multiple predicted transcription factor-binding sites (supplementary material Table S3). We generated a series of luciferase constructs with each element cloned into a pGL3-promoter vector (Fig. 4D). pDcn-luc2 contains the 5′ FOXD1-binding site. MES13 cells were transfected with pGL3-promoter or one of the three luciferase constructs and co-transfected with either pCDNA3.1 or CMV-BF2. The resulting luciferase activity was measured and compared with pGL3-promoter alone (Fig. 4D). Both pDcn-luc1 and pDcn-luc2 co-transfected with empty CMV vector activate luciferase transcription compared with pGL3 alone, whereas pDcn-luc3 has no transcriptional activity. Dcn-luc1 and Dcn-luc2 regions thus appear to contain binding sites for transcriptional activators. Co-transfection of CMV-BF2 did not modulate pDcn-luc1 or pDcn-luc3 expression, but significantly inhibited pDcn-luc2 (P<2×10^-4). To confirm that this inhibition requires the FOXD1-binding site, we generated a pDcn-luc2-mut construct in which the forkhead core sequence TCAACA was replaced with CCGGTA (Fig. 4A). Even in the absence of CMV-BF2, mutating the FOXD1-binding site increased transcriptional activity 68-fold compared with pGL3-promoter alone, suggesting that MES13 expression of FOXD1 or another factor represses pDcn-luc2 transcription through the FOXD1-binding site in MES13 cells. Reporter activity was not decreased by transfection with CMV-BF2, demonstrating that mutagenesis of the FOXD1 site abrogated the repressive activity of FOXD1 (Fig. 4D). Taken together, these data indicate that we have identified a novel enhancer element upstream of Dcn where FOXD1 binds and represses transcription.

To reveal if the Foxd1^-/- phenotype is due to excess DCN, we inactivated Dcn on the Foxd1^-/- background. Litters from Foxd1^+/-;Dcn^+/- intercrosses were harvested at E15.5 and at postnatal day (P) 0. Cohorts of four E15.5 and five P0 mice for each genotype were compared. Externally, Foxd1^-/-;Dcn^-/- kidneys resemble the Foxd1^-/-: they are fused and remain attached to the body wall caudal to the location of wild-type kidneys (supplementary material Fig. S3). E15.5 wild-type kidneys demonstrate the expected cortico-medullary organization, with cortically located cap mesenchymes (Fig. 5A). Foxd1^-/- kidneys are disorganized and have large clusters of cap mesenchymes in the center of the kidney, as previously described (Fig. 5B). However, E15.5 Foxd1^-/-;Dcn^-/- sections display fewer regions of mislocalized progenitor clusters, and these clusters are smaller (Fig. 5C). This difference is enhanced at P0. Medial transverse sections through pairs of Foxd1^-/-;Dcn^-/- kidneys indicate that there is partial rescue of Foxd1^-/- kidney phenotypes. The wild-type P0 kidney is subdivided into functional regions along the kidney’s cortico-medullary axis (Fig. 5D): the cortex, the outer medulla, and the inner medulla and papilla. Foxd1^-/- kidneys lack this arrangement and papillary morphology is rudimentary (Fig. 5E). Characteristic of the Foxd1^-/- kidney is the inappropriate localization of progenitor cell clusters in all kidney regions. In Foxd1^-/-;Dcn^-/- kidneys, cortical, outer medullary and inner medullary domains are visible and clear papillary structure and ducts of Bellini can be identified leading to the renal calyx (Fig. 5F). In contrast to the Foxd1^-/- kidney, progenitor cell clusters are never found in the medulla of the Foxd1^-/-;Dcn^-/- kidney.

In addition to the recovery of gross patterning, we also saw a change in the size of nephrogenic units in Foxd1^-/-;Dcn^-/- kidneys compared with Foxd1^-/-. At E15.5, we compared expression of CITED1, which labels nephron progenitor cells, and LEF1, which labels clusters of differentiating pretubular aggregates and renal vesicles, in wild-type, Foxd1^-/- and Foxd1^-/-;Dcn^-/- littermate kidneys. In all analyses, CITED1 and LEF1 cell counts were restricted to cells associated with a cross-section through a CD tip with a visible lumen. Wild-type kidneys contain ∼16 CITED1+ cells per CD tip in 5-μm sections (Fig. 6A). Foxd1^-/- kidneys contain ∼42 CITED1+ cells per CD tip (Fig. 6B). Foxd1^-/-;Dcn^-/- CDs associate with significantly fewer, an average of 26 CITED1+ cells (P<3.3×10^-32) (Fig. 6C,D). We also quantified the number of LEF1+ cells per cluster in 5-μm sections. In the wild type, LEF1+ clusters contain an average of eight cells (Fig. 6E). Foxd1^-/- kidneys have clusters with an average of five LEF1+ cells (Fig. 6F). This is rescued in Foxd1^-/-;Dcn^-/- kidneys, which have clusters with an average of eight LEF1+ cells (P<5.8×10^-17) (Fig. 6G,H). Importantly, the ratio of CITED1+ cells:LEF1+ cells, which is 2:1 in wild-type kidneys, increases to 7:1 in Foxd1^-/- kidneys. These data reflect a decrease in the rate of differentiation from the CITED1+ state to the LEF1+ state in Foxd1^-/- kidneys compared with wild type. In Foxd1^-/-;Dcn^-/- kidneys, this ratio is 3:1, indicating that cells progress from the CITED1+ state to the LEF1+ state at nearly the wild-type rate (Fig. 6I).

The same reduction in CITED1+ progenitor cells between littermates is seen at P0. In wild-type kidneys, ∼24 CITED1+ cells associate with each CD tip in a 5-μm section (Fig. 6J). However, Foxd1^-/- CD tips associate with an average of 50 CITED1+ cells (Fig. 6K). The number of CITED1+ cells in Foxd1^-/-;Dcn^-/- is significantly reduced compared with Foxd1^-/- (Fig. 6L), with ∼30 CITED1+ cells/CD tip (P<9.8×10^-25) (Fig. 6M). For each genotype, 50 CD tips from three non-adjacent sections from each of five kidneys were counted.

This decrease in CITED1+ cells in Foxd1^-/-;Dcn^-/- kidneys correlates with increased numbers of LEF1+ pretubular aggregates and renal vesicles at P0. In wild-type kidneys, many CD tips in a 5-μm section associate with LEF1+ structures (Fig. 6N). Additional interstitial LEF1 staining surrounds CD trunks in all three genotypes. Foxd1^-/- kidneys have few LEF1+ pretubular aggregates and renal vesicles per section (Fig. 6O), whereas Foxd1^-/-;Dcn^-/- kidneys show a sizable increase in the number of LEF1+ structures per kidney section (Fig. 6P). Three sections from each of five kidneys per genotype were quantified, revealing a twofold increase in LEF1+ structures in Foxd1^-/-;Dcn^-/- compared with Foxd1^-/- (P<0.001) (Fig. 6Q).

An increase in the number of epithelialized nephron structures at P0 would also indicate improved nephron differentiation in the double null. To test this, we assessed localization of the proximal tubule marker lotus lectin. In wild-type kidneys, lotus lectin staining is strongest in proximal tubules in the outer and inner medulla with some weak staining in the CDs in the papilla (Fig. 6R). Lotus lectin staining in Foxd1^-/- kidneys is strong in scattered regions, but relatively few lotus lectin-positive proximal tubules can be found (Fig. 6S). Increased numbers of lotus lectin-positive structures are seen in Foxd1^-/-;Dcn^-/- kidneys relative to Foxd1^-/- (Fig. 6T), with a 25% increase in lotus lectin-positive area (P<0.03) (Fig. 6U). Taken together, this marker analysis shows that inactivating Dcn in a Foxd1^-/- background reduces the accumulation of CITED1+ progenitor cells in nephrogenic structures, increases the abundance of differentiated nephron structures, and improves cortico-medullary patterning of the kidney.

DCN interacts with TGFβ and receptor tyrosine kinase (RTK) growth factors through diverse mechanisms, including ligand and receptor binding (Neill et al., 2012). Our work indicates that maintaining the CITED1+ progenitor cell state requires RTK signaling and that transition out of the CITED1+ compartment into the SIX2-only compartment requires SMAD-mediated BMP signaling (Brown et al., 2011; Brown et al., 2013). To understand whether DCN retains progenitor cells in the CITED1+ state by modulating either RTK or SMAD signaling pathways, we evaluated the effect of DCN on cultured nephrogenic zone cell (NZC) differentiation (Blank et al., 2009). NZCs were treated with FGF2 to maintain Cited1 expression, followed by treatment with BMP7 with and without DCN. Changes in Cited1 expression were measured at both transcript and protein levels. Addition of DCN alone has little effect on the number of CITED1+ cells in FGF2-treated cultures (Fig. 7A,B). However, addition of DCN abrogates the decrease in CITED1+ cells seen upon BMP7 treatment (Fig. 7C,D). Quantification of CITED1+ cells in each condition supports the conclusion that DCN reduces the effect of BMP7 (Fig. 7E), and this is corroborated by gene expression analysis (Fig. 7F). Importantly, although Cited1 expression is reduced in BMP7-treated cells, the level of Six2 expression remains the same with or without addition of DCN, indicating that DCN antagonizes BMP7- and SMAD-mediated differentiation of progenitors from the CITED1+ to the SIX2-only compartment.

The pro-differentiative effect of BMP7 relies on Smad1/5/8 transcription factor signaling (Brown et al., 2013). To verify that DCN modulates this pathway, we performed a transcriptional reporter assay in NIH3T3 cells using pBRE-luc, in which luciferase expression is regulated by a Smad1/5/8-binding enhancer element (Korchynskyi and ten Dijke, 2002). DCN antagonizes Smad1/5/8-mediated signaling by BMP7, significantly reducing transcriptional reporter activation (Fig. 7G). These findings suggest that DCN promotes retention of progenitor cells in the undifferentiated CITED1+ compartment by directly antagonizing SMAD-dependent BMP7 signaling. To test the contribution of this interaction to the Foxd1^-/- phenotype, we compared nuclear phospho-SMAD1/5 expression in wild-type, Foxd1^-/- and Foxd1^-/-;Dcn^-/- kidneys at E15.5. In wild-type nephrogenic units, phospho-SMAD1/5 is detected in clusters of cap mesenchyme cells located adjacent to or just under the CD tip (Fig. 7H,I) (Brown et al., 2013). We find that only a subset of CD tips associate with cap mesenchymes containing phospho-SMAD1/5+ cells in the Foxd1^-/- kidney (Fig. 7J,K), but that this proportion increases in the Foxd1^-/-;Dcn^-/- kidney (Fig. 7L,M). Quantification of the number of CD tips associated with phospho-SMAD1/5+ cap mesenchymes in five non-adjacent 5-μm sections in each of five pairs of kidneys per genotype revealed that 93% of CD tips associate with phospho-SMAD1/5+ progenitor cells in the wild type compared with 66% in Foxd1^-/- (P<0.0008) and 83% in Foxd1^-/-;Dcn^-/- (P<0.0009 versus wild type, P<0.0007 versus Foxd1^-/-; n=400 tips per genotype) (Fig. 7N). To determine if the size of phospho-Smad1/5+ clusters also differs between genotypes, we counted the number of phospho-SMAD1/5+ cap mesenchyme cells per CD tip cross-section (determined by presence of lumens) in these same 5-μm sections. Wild-type kidneys have an average of 9.5 phospho-SMAD1/5+ cap mesenchyme cells per tip, whereas comparable Foxd1^-/- tips have 6.5 cells (Fig. 7L,M; P<1.6×10^-9). By contrast, Foxd1^-/-;Dcn^-/- kidneys contain 10.7 phospho-SMAD1/5+ cells per CD tip, a significant increase compared with both Foxd1^-/- and wild-type kidneys (Fig. 7O; P<9.6×10^-17 and P<0.003, respectively; n=150 CD tips per genotype). These data indicate that DCN reduces SMAD-dependent BMP7 signaling in E15.5 Foxd1^-/- kidneys.

The hallmark feature of the Foxd1 null is accumulation of cap mesenchyme cells in the CITED1+ compartment (Fig. 1A,B). To directly test the effect of DCN on differentiation of these cells, we used a three-dimensional culture assay in which tubulogenesis of aggregated CITED1+ cells, which produce endogenous active BMP7, can be induced by the GSK3β antagonist BIO (Brown et al., 2013). CITED1+ cells were aggregated in media with or without recombinant DCN and tubulogenesis was monitored by light microscopy (Fig. 7P-S). BIO-treated aggregates achieve a high level of complexity within 5 days, forming large epithelial structures with lumens (Fig. 7P,P′). However, BIO-treated aggregates cultured with DCN display reduced complexity with few identifiable epithelial structures (Fig. 7Q,Q′), showing that DCN does indeed attenuate differentiation of pure CITED1+ cells in a functional assay, through inhibition of endogenous BMP signaling.

Nephrogenesis is impaired in renal explants grown in dissolved collagen IV (Sebinger et al., 2013). Collagen IV is aberrantly expressed in Foxd1^-/- kidneys (Fig. 3G,H), such that the extracellular matrix (ECM) environment surrounding the cap mesenchyme cells contains two anti-nephrogenesis factors, collagen IV and DCN. To recapitulate this environment in vitro, we cultured BIO-treated aggregates with collagen IV and DCN. Collagen IV alone reduced the complexity of aggregates, and the combination of collagen IV and DCN had the most severe attenuating effect on epithelial differentiation (Fig. 7R-S′). From this, we conclude that we are able to replicate, in culture, part of the aberrant ECM environment found in the Foxd1^-/- and that this DCN-rich environment prevents the differentiation of epithelial structures.

In this study, we identify in vivo targets of the FOXD1 transcription factor, and provide a mechanistic explanation for the influence of the interstitial niche on progenitor cell differentiation in the Foxd1^-/- kidney. FOXD1 represses the gene encoding the small leucine-rich proteoglycan DCN in cortical interstitial cells. When Foxd1 is inactivated, DCN is expressed and inhibits BMP-SMAD signaling, which blocks transition of the neighboring nephron progenitor cells to a state in which they are susceptible to epithelial induction.

The transcriptome analysis for FOXD1 target genes identified surprisingly few signaling molecules. Instead, DAVID pathway analysis of putative FOXD1 target genes revealed that many genes, including Dcn, encode ECM proteins that specifically associate with collagen fibril formation, including multiple collagens, fibrillin1 and elastin (Huang et al., 2009a; Huang et al., 2009b). Furthermore, we demonstrated that collagen IV localization is aberrant in Foxd1^-/- kidneys. This is interesting because it suggests that FOXD1 either directly or indirectly regulates an entire cassette of genes, including Dcn, that sets up a specific ECM environment surrounding the cap mesenchyme cells. The importance of ECM components to cellular signaling responses is increasingly evident (Alexi et al., 2011; Schaefer and Iozzo, 2008). In this work, we demonstrate that the aberrant ECM is sufficient to reduce differentiation of progenitor cells to epithelial structures. Stem and progenitor cells, in particular, are sensitive to changes in the ECM environment (Ichii et al., 2012; Mathews et al., 2012a; Mathews et al., 2012b; Wang et al., 2010).

The finding that cap mesenchyme cells are trapped in the CITED1+ compartment in the Foxd1^-/- kidney is a key observation in this study. Each compartment of the cap mesenchyme differs in its responsiveness to essential signals such as WNTs and BMPs. Progenitor cells respond differently to WNT9b, choosing to proliferate or differentiate depending on whether they have high or low levels of SIX2 (Brown et al., 2013; Karner et al., 2011; Park et al., 2012). In a wild-type kidney, transition out of the CITED1+ state to the SIX2-only state requires BMP7/SMAD signaling (Brown et al., 2013). We propose that DCN reduces BMP/SMAD signaling in Foxd1^-/- kidneys by modulating both BMP7 and BMP4, expression of which is elevated in this mutant (Levinson et al., 2005). Progenitor cell transition into a WNT9b/β-catenin inducible state is thus prevented (Fig. 8). BMPs are TGFβ superfamily ligands, and their structural similarities suggest that DCN may directly bind BMP7 by the same mechanism as it binds TGFβ1 (Yamaguchi et al., 1990). BMP7 promotes cap mesenchyme cell proliferation through a SMAD-independent mechanism (Blank et al., 2009). Although Foxd1^-/- kidneys are much smaller than wild-type kidneys, proliferation within cap mesenchyme does not differ significantly, indicating that the effect of BMP on progenitor proliferation is unchanged and suggesting that signaling through SMAD1/5 is specifically affected by DCN (Levinson et al., 2005). Factors governing signaling outcomes of BMP receptor activation are incompletely understood, but include receptor complex conformation at the cell surface, which may be strongly influenced by the extracellular environment (Guzman et al., 2012). Future studies of the molecular mechanism by which DCN acts on BMP ligands and receptors will explore the possibility that it does not simply inhibit BMP signaling, but redirects its outcome.

Our finding that DCN-mediated antagonism of BMP signaling skews progenitor cell compartments in Foxd1^-/- kidneys suggests that there should be similarities between kidney phenotypes seen upon inactivation of Bmp7 and Foxd1. Although the kidneys of these mutants are morphologically distinct at E14.5, an important similarity is that both mutants retain CITED1-expressing cells and that differentiation to the SIX2-only compartment is blocked (Brown et al., 2013). The Foxd1 mutant phenotype is highly complex, and our work implicates separate origins of the horseshoe and the progenitor cell retention phenotypes. We speculate that the lack of separation of the Foxd1^-/- kidney from the body wall imposes severe constraints on the capacity for centrifugal growth of the kidney, compressing CD tips and associated cap mesenchyme cells into a small volume resulting in the characteristic Foxd1^-/- morphology.

Genetic inactivation of Dcn in the Foxd1^-/- background partially repairs the cortico-medullary patterning defects seen in Foxd1^-/- kidneys. Intriguingly, BMP signaling is essential for appropriate cortico-medullary patterning, and medullary hypoplasia is seen upon inactivation of the BMP receptor ALK3 (BMPR1A - Mouse Genome Informatics) in the CD (Hartwig et al., 2008; Yu et al., 2009). Thus, in Foxd1^-/- kidneys, DCN modulation of BMP signaling may inhibit appropriate patterning. An alternative explanation for the Foxd1^-/- cortico-medullary patterning deficiency is that excess DCN, combined with high levels of the other misexpressed ECM molecules, might form a rigid matrix resulting in a physical impediment to both the outgrowth of the CD tips and the elongation of the CD stalks. Removing DCN from the matrix would reduce its inflexibility, much as loss of Dcn reduces the tensile strength of skin, permitting the directional growth of CDs and nephron tubules (Danielson et al., 1997).

The nephrogenic zone has been viewed as a stem cell-stroma niche, in which cortical interstitial support cells promote the self-renewal and/or differentiation of cap mesenchyme cells. However, lineage tracing reveals that cortical interstitial cells are progenitor cells in their own right, which self-renew and differentiate into essential cell types, including mesangial cells and peritubular pericytes (Humphreys et al., 2010). Our gene expression data and PDGFRβ marker analyses reveal that Foxd1 regulates the progenitor cell status of cortical interstitial cells. Loss of Foxd1 leads to an increase in transcription of genes normally expressed in cells derived from cortical interstitium, including Rgs5 and Mgp (mesangium) and Dcn (medullary interstitium) (supplementary material Table S2). We propose that Foxd1 acts as a ‘stemness’ gene for the cortical interstitium, and that in its absence the cortical interstitium prematurely adopts its differentiated fates. The demonstration that interstitium throughout the Foxd1^-/- kidney expresses PDGFRβ supports this notion. Premature differentiation of cortical interstitial cells correlates with a lack of cap mesenchyme cell differentiation in the Foxd1^-/- kidney (Fig. 1E,F). We propose that the renal nephrogenic zone comprises a niche composed of interdependent progenitor cell populations, rather than progenitors and stromal support cells. Interestingly, the nephrogenic zone is home to two additional populations of progenitor cells: the Flk1+ (Kdr - Mouse Genome Informatics) endothelial progenitor cells that give rise to the glomerular capillary tufts, and the Ret+ CD tip cells (Chi et al., 2009). Spatial integration of these four stem cell compartments enables the synchronized formation of nephrons containing epithelial cells, endothelial cells and mesangial cells. Defining the axes of communication between these distinct components of the niche is essential for our understanding of nephrogenesis, and for ongoing attempts at de novo generation of kidney tissue.

Foxd1^lacZ/+ mice (Hatini et al., 1996) were on an ICR background and Dcn^TM1Ioz mice (Danielson et al., 1997) on BL/6. Noon of the day of the vaginal plug was designated E0.5. Animal care was in accordance with the National Research Council Guide for the Care and Use of Laboratory Animals and all experiments were approved by the Maine Medical Center IACUC.

MES13 cells (ATCC Rockville, MD, USA) were maintained per vendor instructions and transfected with CMV-BF2 (kindly provided by E. Lai, Columbia University, NY, USA) using Lipofectamine 2000 (Life Technologies) as per manufacturer instructions. After 24 hours, RNA was prepared using the Qiagen RNeasy Mini Kit. cDNA was synthesized with the qScript cDNA Kit (Quanta BioSciences), and quantitative RT-PCR reactions (three technical replicates) run using iQ Sybr Green SuperMix (Bio-Rad). ICR NZCs were isolated and cultured as described (Blank et al., 2009). Cells were treated with 50 ng/ml FGF2 (R&D Systems) for 6 hours followed by 10 μg/ml DCN (R&D Systems) and/or 50 ng/ml BMP7 (R&D Systems) for 24 hours before immunofluorescence or RNA isolation. NIH3T3 cells (ATCC) were maintained as per vendor instructions. 3D aggregate cultures were isolated and incubated as described (Brown et al., 2013) then treated with 2 μM BIO (Calbiochem), 10 μg/ml DCN and 50 μg/ml collagen IV (BD Biosciences).

Conserved elements near the predicted FOXD1-binding site were identified with rVista (rvista.dcode.org). Elements were amplified from genomic DNA using the Phusion High-Fidelity Polymerase Kit (NEB) using primers in supplementary material Table S1.

Amplicons were directionally cloned into pGL3-promoter vector (Promega) using the Fermentas Rapid DNA Ligation Kit. The FOXD1-binding site was mutated using Phusion and the amplicon directionally cloned into pDcn-luc2.

MES13 cells were transfected with either pGL3-promoter or one of the four Dcn-luc constructs along with pRL-CMV (Promega) and either pCDNA3.1 (Life Technologies) or CMV-BF2 as above, incubated for 24 hours with 5% fetal calf serum and luciferase assays were performed using the Dual-Luciferase Reporter Assay Kit (Promega). Five biological replicates were analyzed for each condition.

NIH3T3 cells were transfected with pBRE-luc (Korchynskyi and ten Dijke, 2002) and pRL-CMV (Promega) as above, serum starved for 6 hours, and treated with factors as above. After 24 hours, cells were lysed and luciferase assays were performed as above.

Embryonic kidneys were harvested from Foxd1^-/- and wild-type littermates from three E14.5 litters. Three biological replicates were generated per genotype, each containing two non-littermate kidney pairs. These were pooled and homogenized in Trizol. RNA was prepared using a Qiagen RNeasy Micro Kit. Samples were run on the Affymetrix Mouse Gene 1.0 ST Array platform by the Vermont Genetics Network Microarray Core Facility. The data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus (Edgar et al., 2002) and are accessible through GEO series accession number GSE52354 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE52354).

Microarray data were analyzed with XRAY software (Biotique Systems). Genes with at least four probes and P<0.01 were selected. FOXD1-binding site analysis was performed using Genomatix MatInspector (genomatix.de) to identify sites with 0.75 matrix similarity to the Transfac V$FREAC4.01 matrix. Binding site conservation was identified via ECR Browser (http://ecrbrowser.dcode.org).

Immunostaining was conducted on paraffin sections of paraformaldehyde-fixed tissue as previously described (Blank et al., 2009). Primary antibodies were: CITED1 (1:100; NeoMarkers Cited 1, RB9269-PO; AbCam Cited 1, AB15096); SIX2 (1:50; Santa Cruz Biotechnology SC-67834); PDGFRβ (1:250; Epitomics, 1469-1); collagen IV (1:50; Rockland Antibodies and Assays: 600-401-106-0.1); collagen I (1:50; Rockland Antibodies and Assays: 600-401-103-0.1); DBA lectin (1:250; Vector Laboratories, B-1035); Lotus lectin (1:250; Vector Laboratories, B-1325); phospho-SMAD1/5 (1:50; Cell Signaling Technology, 9516S). Secondary antibodies were: biotinylated anti-rabbit (Vector Laboratories, BA-1000); streptavidin Alexafluor 488 (Life Technologies, S11223); Alexafluor 568 goat anti-rabbit (Life Technologies, A11011); Alexafluor 568 goat anti-mouse (Life Technologies, A11004); Alexafluor 568 donkey anti-goat (Life Technologies, A11057). All secondary antibodies were diluted 1:250. DAPI counterstain was also used (1:10000; Life Technologies D1306).

Methanol-fixed whole E14.5 kidneys were processed as described (Barak and Boyle, 2011) in 1:100 anti-DCN (R&D Systems AF-1060). In situ hybridization using digoxigenin-labeled probes was carried out on paraffin sections as described (Oxburgh et al., 2004). PCR amplification from an E17.5 whole-embryo cDNA library was used to generate probe templates.

ChIP was performed with pCX-eGFP or p3xFLAG-FOXD1-transfected MES13 cells. The p3xFLAG-FOXD1 expression construct was generated by subcloning mouse Foxd1 cDNA into a mammalian expression vector along with an N-terminal triple FLAG epitope (3xFLAG). The Imprint Ultra Chromatin Immunoprecipitation Kit (Sigma-Aldrich) was used with a rabbit anti-FLAG monoclonal antibody (Sigma-Aldrich, M8823) according to the manufacturer’s instructions.

We thank Professor Liliana Schaefer for help in importing the decorin null mouse strain.