Wilms’ tumor (WT) morphologically resembles the embryonic kidney, consisting of blastema, epithelial and stromal components, suggesting tumors arise from the dysregulation of normal development. β-Catenin activation is observed in a significant proportion of WTs; however, much remains to be understood about how it contributes to tumorigenesis. Although activating β-catenin mutations are observed in both blastema and stromal components of WT, current models assume that activation in the blastemal lineage is causal. Paradoxically, studies performed in mice suggest that activation of β-catenin in the nephrogenic lineage results in loss of nephron progenitor cell (NPC) renewal, a phenotype opposite to WT. Here, we show that activation of β-catenin in the stromal lineage non-autonomously prevents the differentiation of NPCs. Comparisons of the transcriptomes of kidneys expressing an activated allele of β-catenin in the stromal or nephron progenitor cells reveals that human WT more closely resembles the stromal-lineage mutants. These findings suggest that stromal β-catenin activation results in histological and molecular features of human WT, providing insights into how alterations in the stromal microenvironment may play an active role in tumorigenesis.
Wilms' tumor, or nephroblastoma, is an embryonal tumor classically consisting of triphasic histology, with blastemal/nephron progenitor, epithelial and stromal components thought to arise from disruptions in normal fetal nephrogenesis (Treger et al., 2019; Rivera and Haber; 2005; Hohenstein et al., 2015). During normal kidney development, nephron progenitor cells (NPCs) are maintained through renewing cell divisions. However, a subset of these cells simultaneously lose their progenitor cell identity and undergo mesenchymal-to-epithelial transition (MET) to form an immature tubule that will become a nephron, the functional unit of the kidney. A balance between self-renewal and differentiation is essential for generating kidneys with a sufficient number of nephrons necessary for function. This process is highly regulated and is known to rely on signals emanating from the epithelial cells of the ureteric bud as well as the surrounding stromal/interstitial signaling to the NPCs (Das et al., 2013; Fetting et al., 2014; Hum et al., 2014).
In a normal human kidney, the NPCs are exhausted prior to birth. However, in WT, blastemal cells/NPCs persist and continue proliferating well into the postnatal period. Still, much remains to be understood regarding the mechanism of this malignant transformation and resultant triphasic histology. Of particular relevance, the contribution of the stroma to Wilms’ tumorigenesis remains largely unknown. While the stroma/interstitium plays multiple roles in supporting normal tissue development and homeostasis, it has also been shown to contribute to tumor formation, progression and metastasis in many cancers (Clark and Vignjevic; 2015; Li et al., 2007; Valkenburg et al., 2018). Tumor ‘stroma’ refers to all components of the interstitium, including fibroblasts, immune cells and vasculature (i.e. endothelium and endothelial-associated mural cells), as well as the basement membrane and extracellular matrix (Bremnes et al., 2011). In this study, we will focus on the non-immune, non-vascular, cellular components of the stroma, given that immature stroma/interstitial fibroblast cells abnormally persist and proliferate in WT.
Activating mutations in the gene encoding β-catenin, CTNNB1, occur in about 15% of WTs (Li et al., 2004; Maiti et al., 2000). However, nuclear accumulation of β-catenin is observed in up to 50% of tumors (Koesters et al., 2003), suggesting that aberrant activation of this pathway is crucial in a significant proportion of WTs. β-Catenin is a component of the canonical Wnt signal transduction pathway. In the absence of a WNT ligand, β-catenin is phosphorylated, sequestered in the cytoplasm and tagged for degradation. However, in the presence of a WNT ligand, the cytoplasmic complex that phosphorylates β-catenin dissociates, freeing β-catenin to translocate into the nucleus and promote transcription. The mutations in the β-catenin gene observed in WT render the protein insensitive to degradation, thus resulting in a constitutively stabilized form lacking regulation of its transcriptional activity.
In normal kidney development, Wnt/β-catenin signaling regulates multiple aspects of nephrogenesis, including nephron progenitor renewal, mesenchymal-to-epithelial transition, ureteric bud progenitor renewal and differentiation of the interstitium (Boivin and Bridgewater; 2018; Boivin et al., 2016; Park et al., 2007; Sarin et al., 2014; Marose et al., 2008; Karner et al., 2011; Ramalingam et al., 2018). Which of these processes is perturbed in WT is still unclear.
Nuclear expression of β-catenin, as well as cells carrying CTNNB1-activating mutations, are observed among the different cell lineages of WT including blastemal, stromal, epithelial and even heterologous components such as skeletal muscle (Corbin et al., 2009; Duhme et al., 2019; Uschkereit et al., 2007). It has long been assumed that the causal mutation in WTs occurs in the blastemal component (Charles et al., 1998); however, recent studies performed in mice demonstrate that mutations in Wilms' candidate genes (including Lin28 and Wt1/Igf2) in the blastemal/nephron progenitor component alone are not sufficient to cause WT (Urbach et al., 2014; Huang et al., 2016). It is becoming increasingly apparent that the balance between nephron progenitor maintenance and differentiation is regulated by signals from the renal stroma, and we and others have shown that perturbations in stromal differentiation result in abnormally maintained NPCs reminiscent of nephrogenic rests in WTs (Das et al., 2013; Hum et al., 2014). Given the established roles of the stroma in tumor progression and its developmental role in regulating nephron differentiation, and that WT stroma harbors β-catenin-activating mutations, we decided to interrogate the role of the stroma in WT.
Here, we characterize mice carrying an activating mutation of β-catenin specifically in the stromal progenitor population. We show that mutant kidneys form remnant epithelial structures surrounded by undifferentiated mesenchyme and spindle-shaped fibroblasts, similar to the histology of WT. Transcriptomes of mutant mouse kidneys with activating mutations in either the nephron progenitor lineage or the stromal lineage were compared with human WTs revealing that WT shares characteristics of the stromal-lineage mutants, more so than wild-type developing kidney or NPC lineage mutants. Indeed, expression of Six2, Cited1 and Ncam, diagnostic markers of WT, is observed in the stromal lineage mutants and not the NPC lineage mutants. Finally, we show that activation of β-catenin simultaneously in the nephron progenitor and stromal lineages (using dual Six2cre and Foxd1cre expression) results in severely disrupted kidney development with the formation of bone-like tissue, a phenotype that has been reported in human WT (Pritchard-Jones, 1997). These findings suggest that activation of β-catenin in the stroma contributes to WT pathogenesis. Further understanding of the lineage-specific effects of β-catenin-activating mutations, as well as a more-detailed analysis of the tumor stromal microenvironment, may aid in unraveling the molecular and cellular mechanisms underlying WT.
Activating mutations of β-catenin are found in both the stroma and NPCs of WT
Although it has long been known that WTs contain activating mutations in β-catenin, the role of these mutations (drivers versus passenger) and the cell type of origin has been somewhat controversial. Several groups have shown that activating mutations in β-catenin can be found in all cell types of WTs (Corbin et al., 2009; Duhme et al., 2019; Uschkereit et al., 2007). To confirm these observations, we isolated stromal and blastemal cells using laser capture microdissection from three different WT samples carrying activating mutations of β-catenin. In all three samples, the mutant allele was identified in both blastemal and stromal cellular components (representative data from one sample shown in Fig. 1). As the blastema and stroma arise from different cellular lineages in a normal kidney, and the two lineages appear to sort independently in early kidney development, the fact that the mutations are found in both cell populations within a tumor supports the claim that the mutagenic event occurred in a common precursor cell for both lineages (the intermediate mesoderm) and raises the possibility that activation in the stromal progenitor lineage contributes to the pathology of WT.
Activation of β-catenin within different lineages of the metanephros severely perturbs renal development, with stromal activation demonstrating histological characteristics of Wilms' tumor
β-Catenin is ubiquitously expressed in the developing kidney. Through the use of various lineage-restricted Cre lines and alleles of β-catenin that can be activated or inactivated by Cre, this factor has been shown to play a role in the balance of NPC maintenance and differentiation/mesenchymal-epithelial transition, as well as in the development of interstitial cells and the ureteric bud progenitors (Boivin and Bridgewater; 2018; Boivin et al., 2016; Park et al., 2007; Sarin et al., 2014; Marose et al., 2008; Karner et al., 2011; Ramalingam et al., 2018; Yu et al., 2009). Although previous studies have characterized kidneys carrying activated alleles of β-catenin (Catnbex3/+) within both the NPC and stromal lineages, we reanalyzed both mutants with a focus on their relationship to WT pathology. Six2Cre is active in the nephron progenitor cells, and Six2cre;Catnbex3/+ mutants lack developing nephrons and show premature loss of NPCs (Fig. 2C). Lineage-positive cells instead form aggregates of cells that persist throughout development and do not appear either mesenchymal or epithelial (Fig. 2D, Fig. S1).
Foxd1 is expressed in a population of mesenchymal cells in the cortex of the embryonic kidney, and lineage tracing studies have shown that Foxd1-derived cells give rise to non-endothelial non-immune stromal cells, including pericytes, fibroblasts, mesangium and vascular smooth muscle cells (Kobayashi et al., 2014). In comparison with activation with Six2Cre, activation of β-catenin with Foxd1Cre (Foxd1cre;Catnbex3/+, from here on referred to as stromal or interstitial activation) results in abnormal maintenance of the NPC population, with delayed nephrogenesis (Fig. 2E). Gross morphologic examination revealed that stromal mutant kidneys were fused to the body wall. E18.5 kidneys showed a complete absence of mature nephrons (Fig. 2F′) with expanded interstitial cells surrounding undifferentiated NPCs and immature epithelia resembling renal vesicles or s-shaped bodies, grossly resembling the morphology of WTs. Lineage tracing of mutant kidneys derived from the two different Cre driver strains confirmed recombination in the expected nephron progenitor and stromal lineages, with the majority of lineage-positive cells showing pathologically high levels of nuclear β-catenin (Fig. S1).
Activation of β-catenin within the stromal lineage results in expanded nephron progenitor populations with delayed mesenchymal-to-epithelial transition
WTs are characterized by the expression of a number of genes normally associated with undifferentiated NPCs, including Six2, Pax8 and Ncam. We thus sought to assess the expression of these markers in kidneys with activating mutations of β-catenin in either lineage. Previous work suggests that activation of β-catenin in the NPC lineage promotes differentiation, with decreased numbers of NPCs, an increased number of pre-tubular aggregate structures expressing Wnt4, Pax8 and Lhx1, and a blockade in MET/differentiation (Park et al., 2007). Re-analysis of these mutants shows that this ‘pre-tubular aggregate-like state’ is transient, and by E15.5, NPCs lose Six2 expression as well as expression of Lhx1, Pax8 and Ncam (Fig. 3G,H). Thus, at a molecular level, activation of β-catenin within the NPC lineage does not lead to a WT-like phenotype nor does it lead to a maintenance of pre-tubular aggregate-like structures.
Next, we further examined the abnormally maintained NPCs in stromal mutant kidneys. In contrast to activation within the NPC lineage, E15.5 Foxd1cre;Catnbex3/+ kidneys maintain expression of Six2 and Ncam (Fig. 3K,L and Fig. 4H). However, expression of pre-tubular aggregate/differentiating NPC markers is abnormal, with some markers, including C1qdc2 and Wnt4, showing expanded expression surrounding the NPCs, while others, including Pax8 and Lhx1, are not expressed (Fig. 4I-L, respectively). At E18.5, a small number of structures expressing Lhx1 are present (Fig. 4N), corresponding to histologically identifiable renal vesicles and perhaps comma and S-shape bodies (Fig. 2F). These findings suggest that stromal activation of β-catenin not only non-autonomously blocks NPC differentiation, but also fundamentally alters the molecular state of these cells, demonstrating that disruption of the normal stromal microenvironment significantly affects the differentiation of the neighboring NPC population.
Foxd1Cre-mediated activation of β-catenin disrupts stromal patterning
Previously, it has been shown that ablation of the stromal progenitor population results in abnormally maintained NPCs that are reminiscent of nephrogenic rests (Das et al., 2013). Foxd1cre;Catnbex3/+ kidneys show some hallmarks of these stroma-less kidneys, including abnormally expanded Six2-expressing NPCs surrounding the UB (Fig. 3K). Given these findings, we previously hypothesized that nephrogenic stromal cells produce a signal that promotes differentiation or blocks renewal of the NPCs. With this in mind, we examined the molecular phenotype of stromal cells upon activation of β-catenin with Foxd1Cre. Foxd1cre;Catnbex3/+ kidneys show early loss of the stromal progenitor population, as indicated by decreased expression of Foxd1, Ntn1 and Smoc2 (Fig. 5G,H,S). Instead, several genes normally expressed in the papillary stroma were precociously and ectopically expressed in the cortex, including Cpmx2, Sdc2, Dpp6 and Wnt5a, although Wnt4 papillary stroma expression was lost (Fig. 5J-L,R,T) (Shan et al., 2010). Mutant kidneys showed decreased expression of the cortico-medullary stromal markers Penk and Smoc2 (Fig. 5I,S). These findings suggest that activation of β-catenin in the stromal progenitor cells may be leading to precocious and ectopic differentiation of a more papillary stromal cell type. Given that β-catenin has previously been shown to be necessary for the development of the papillary stroma (Yu et al., 2009; Boivin and Bridgewater; 2018; Boivin et al., 2016), our findings suggest it is also sufficient.
Transcriptome profiling suggests human WT resembles mutant mouse kidneys with activation of β-catenin specific to the stromal lineage
Previous transcriptional analyses of human WT suggests the upregulation of numerous β-catenin targets genes (Gadd et al., 2017; Li et al., 2004; Zirn et al., 2006). However, as β-catenin is active in multiple lineages within the kidney and turns on unique targets in each lineage, we next sought to further characterize β-catenin target genes upregulated in human WT by investigating their expression in normal mouse kidneys, as well as our lineage-specific β-catenin mutants. This was accomplished by performing RNA-Seq on both Six2cre;Catnbex3/+ and Foxd1cre;Catnbex3/+ mutant kidneys (see Table S1). We next used BETA to integrate the differentially expressed genes from these mutant mouse models with beta-catenin CHIP-seq data (Table S2), thus generating a list of genes that are likely to be directly activated by β-catenin in each lineage.
To identify genes that are likely to be mis-regulated in WT directly due to activation of β-catenin, we then analyzed RNA-seq from WT samples in the TARGET database, comparing gene expression profiles of human WTs with and without CTNNB1-activating mutations and cross-referenced these data with the list of likely direct β-catenin target genes from BETA (Table S3) and previously published targets. We next qualitatively compared this list with the lineage-specific mutant mouse models to determine whether one model more closely matched the expression pattern of WT with CTNNB1-activating mutations. There was no discernible alignment using this method or gene-set enrichment analyses. In fact, somewhat unexpectedly, both mouse mutants showed upregulation of several of the same WT-enriched target genes. This is surprising, given that many of these target genes, including Nkd1, Gap43, Axin2, Notum, Mmp11 and Apcdd1 are normally expressed specifically in the papillary stroma of wild-type kidneys (Fig. S2 and Fig. 6). Characterization of the expression of several of these genes in Six2cre;Catnbex3/+ and Foxd1cre;Catnbex3/+ kidneys showed that all β-catenin target genes assessed were precociously and ectopically expressed in both mutants (Fig. 6). However, in the Foxd1cre mutants, target expression was observed in all Foxd1cre lineage positive stroma (rather than being restricted to the medullary interstitium); in Six2cre mutants, expression was relatively normal in the papillary interstitium but was also observed ectopically in Six2cre-derived cells.
We next developed a more comprehensive method to compare β-catenin targets in human WT with our mutant mouse models. Given the considerable complexity and size of the TARGET dataset, we used a deep learning classification technique. Using a trained neural network classifier, the expression profiles for 124 Wilms' tumor samples were mapped to the expression data generated from kidneys of each of the three mouse genotypes (wild type, Six2cre;Catnbex3/+ and Foxd1cre;Catnbex3/+), resulting in a score ranging from 0 (no match) to 1.0 (representing a perfect match). As shown in Fig. 7, nearly all the human WT samples, including six tumors with known CTNNB1-activating mutations, showed almost exclusive mapping to the Foxd1cre;Catnbex3/+ transcriptome (scores ranging from 0.55 to 0.99), with only two samples showing an appreciable degree of similarity to wild-type kidneys (scores of 0.28 and 0.44), and none of the samples mapping to the Six2cre;Catnbex3/+ expression profile (Table S4). This unbiased, quantitative approach, along with our histological studies, suggests that activation of β-catenin in the NPC lineage does not appear to transcriptionally recapitulate WT. Interestingly, despite human WT histologically resembling normal development, this analysis suggests that it is transcriptionally more similar to the mutant kidneys with stromal activation of β-catenin than to normal, wild-type embryonic kidney.
Activation of β-catenin simultaneously within both the NPC and stromal lineages results in ectopic bone formation
Although our data suggest WTs with activating mutations in β-catenin are most similar at the molecular level to mouse kidneys with an activating mutation in the stromal lineage, neither Foxd1cre;Catnbex3/+ or Six2cre;Catnbex3/+ mutants precisely mimic WT. As it is likely that WTs acquire mutations in β-catenin early in their history, we next sought to determine whether activation of β-catenin in both lineages was sufficient to lead to WT. We first sought to activate β-catenin in cells that represented a common progenitor to both the stroma and NPCs using T (brachyury)-creERT2. Using a Rosa-LSL-YFP lineage tracer, we observe recombination within the kidney when tamoxifen was administered to the TcreERT2;Catnbex3/+ embryos early in gestation at E7.5, E8.5 or E9.5; however, unexpectedly the kidneys appeared completely normal at E18.5 (Fig. 8) and at 4 months of age even though phenotypes outside the kidneys (such as curly tails) were visible (data not shown). Further analysis showed that lineage-positive cells within the kidney lacked detectable nuclear β-catenin (Fig. 8B). This was quite unexpected given the strong detectable nuclear β-catenin staining observed in the Six2cre;Catnbex3/+ and Foxd1cre;Catnbex3/+ mutants (Fig. 8C,D). This lack of nuclear β-catenin in the TcreERT2 mutants suggests that early renal precursor cells expressing increased levels β-catenin are preferentially selected against or subsequently downregulate β-catenin to allow continued development. Similar results were observed with another line driving Cre recombination in the intermediate mesoderm (Osr1CreERT2, data not shown).
To circumvent possible negative selection/cell competition, we simultaneously and uniformly activated β-catenin in both the NPC and stromal lineages by creating Six2cre;Foxd1cre;Catnbex3/+ mutant kidneys. As shown in Fig. 2, Six2cre;Foxd1cre;Catnbex3/+ mutant kidneys show no distinguishable features of normal kidney development at E15.5, with progenitor cells appearing to surround a primitive UB/nephric duct. Interestingly, by E18.5, these cells have a ‘bone-like’ appearance, with histology showing a non-calcified, bone-like matrix and strong alkaline phosphatase staining within the lineage-positive cells (Fig. 9A-C).
Although β-catenin targets including Lef-1, a transcription factor involved in canonical Wnt/β-catenin signaling that has also been shown to be a direct target of β-catenin, are upregulated in lineage-positive cells in all three mutant lines (Fig. 9G-K), it is notable that the ‘bone-like’ phenotype, which is observed in some WTs, is observed only in the Six2cre;Foxd1cre;Catnbex3/+ mutants. This observation clearly supports the hypothesis that activation of β-catenin in the stromal lineage of WTs is not inert. We propose that activation of β-catenin in the stroma contributes in multiple ways to WT pathology.
WT, the most common type of kidney cancer in children, is thought to arise from transformed cells originating within the developing kidney. While WT1 and β-catenin mutations were some of the first known pathways proposed in WT pathogenesis, advances in sequencing technology have revealed WT to be genetically heterogenous, with a number of different genetic perturbations commonly resulting in a preserved nephron progenitor state and/or interrupted normal development (Gadd et al., 2017; Treger et al., 2019).
Wnt/β-catenin signaling plays a crucial role in multiple aspects of normal kidney development and is upregulated in a significant proportion of human WTs. However, how β-catenin drives tumorigenesis remains unclear. By examining the transcriptomic effects of β-catenin-activating mutations in multiple cell lineages of the developing kidney, we have further defined how abnormal activation of this signaling pathway exerts both cell-type specific and lineage-independent effects in renal development. Although it has been proposed that activation of β-catenin within the nephron progenitors or renal vesicles leads to tumor formation through inhibition of MET or activation of EMT, respectively, our findings suggest that activation of this gene within the mouse nephron progenitors and their derivatives does not lead to a WT-like phenotype at the molecular level. However, activation of β-catenin in the stromal progenitors results in impaired nephrogenesis, leading to nephrogenic rest-like structures that closely resemble human WTs at the transcriptional level. We show that Foxd1cre;Catnbex3/+ mutants maintain expression of Six2 and Ncam in the developing NPCs, similar to human WT blastema, whereas Six2cre;Catnbex3/+ mutants do not recapitulate this phenotype. Additionally, we show that the patterning of the renal interstitium is severely disrupted in Foxd1cre;Catnbex3/+ kidneys, with an early loss of Foxd1 progenitor cells and cortico-medullary stroma and a pronounced expansion of papillary stromal markers. Although distinctions between cortical and medullary interstitial cells have been previously recognized, recent work has revealed a surprising degree of heterogeneity in the embryonic renal interstitium (England et al., 2020), suggesting that unique subpopulations of interstitial cells regulate adjacent cell types in normal kidney development. In support of this, inactivation of β-catenin in the stromal progenitor population (using Foxd1Cre) blocks development of the papillary interstitium as well as adjacent epithelial cells of the loop of Henle (England et al., 2020; Yu et al., 2009). Conversely, we show here that activation of β-catenin in the stromal progenitor population drives expression of papillary interstitial cells. We hypothesize that the abnormal interstitial patterning in the Foxd1cre;Catnbex3/+ kidneys disturbs the normal stromal microenvironment present in developing kidneys, leading to the altered gene expression and lack of differentiation in the NPC population. We hypothesize that a similar disruption to the stromal microenvironment contributes to Wilms' tumorigenesis, as has been suggested in numerous other tumors (Clark and Vignjevic; 2015; Mao et al., 2013; Bremnes et al., 2011; Valkenburg et al., 2018; Li et al., 2007).
It has long been assumed that driving mutations in WT develop in the NPC/blastemal component. However, more recently, studies using laser microcapture techniques to analyze different components of human WT, including data from three individuals in our study, demonstrate identical mutations in blastemal, stromal and epithelial components consistent with β-catenin mutations occurring in an early common precursor cell (Duhme et al., 2019; Uschkereit et al., 2007). Interestingly, our examination of mice with mosaic activation of β-catenin in early metanephric precursor lineages using TcreERT2 reveals that mutant cells were either selected against or downregulated the forced expression of β-catenin and underwent grossly normal development. In contrast, simultaneous activation of β-catenin in the nephron progenitor and stromal lineages showed severely perturbed development, with the formation of bone-like tissue. Interestingly, bone as well as other heterologous elements, including cartilage and skeletal muscle have been reported in human WTs. Although these observations are consistent with a model in which activating mutations must occur in both NPC and stromal lineages (or a common progenitor for both), we cannot rule out the possibility that the triphasic morphology of WTs is due to aberrant tumor cell differentiation (through either multi-lineage potential or an ability to transdifferentiate) or that activating mutations in epithelial structures lead to EMT. Although we feel that the molecular phenotype of tumor stroma along with the histology of mouse mutants makes these possibilities unlikely (specifically, lineage-traced NPC cells carrying an activated allele of β-catenin do not take on a stromal appearance), it is clear that activation of β-catenin alone is not sufficient to drive WT formation in mice and we cannot rule out contributions by other mutant genes.
β-Catenin plays multiple, cell type-specific roles during kidney development by activating different target genes (Pan et al., 2017). Surprisingly, stabilization of β-catenin in either the stroma or the NPCs results in activation of the same stromal target genes. Previous studies have shown that different targets of β-catenin are regulated in a dose-specific manner (Ramalingam et al., 2018). It is possible that the stromal targets represent genes activated by the highest levels of β-catenin, independent of cell type. Another possibility is that the expression of co-regulators determines target gene expression. It is interesting to note that Lef1 was over-expressed in all gain-of-function models (Six2cre, Foxd1cre and dual Six2cre;Foxd1cre) examined. The ‘stromal targets’ of β-catenin may actually be specific targets of a Lef1/β-catenin complex. These two hypotheses (which are not mutually exclusive) will need to be further explored.
Surprisingly, we observed no gross abnormalities in adult kidneys upon mosaic activation of β-catenin in the NPC, stromal or intermediate mesoderm lineages even though we did detect lineage positive cells. This is in contrast to a previous publication that found that activation of β-catenin with Six2cre but not Foxd1cre resulted in tumor like structures in adult mice (Huang et al., 2016). We cannot explain the discrepancy; however, it is important to note that Huang et. al performed limited molecular characterization of their dysplastic tissues. Whether these lesions truly represent models of WT is still unclear. As we are unable to detect cells with nuclear β-catenin in E18.5 or adult kidneys after mosaic activation of β-catenin using standard immunofluorescence techniques, we hypothesize that, under otherwise normal conditions, strong β-catenin activity either leads to precocious differentiation or is detrimental to cell survival (or both). In that case, additional mutations in genes affecting cell survival or differentiation may help to maintain these cells postnatally. Indeed, mutations in the WT1 gene, which affects NPC differentiation, as well as in TP53, which impacts cell survival, are frequently found in tumors with β-catenin mutation (Huff, 2011; Maiti et al., 2000). It will be interesting to see whether mutation of either of these interacts with β-catenin activation, especially in the TcreERT2 context.
Given the heterogeneous nature of WT, understanding the effects of β-catenin activation in different cell lineages of the developing kidney in comparison with the molecular changes of human WT will aid in unraveling the pathogenesis of this embryonal tumor. Although our data suggest that activation of β-catenin in the stromal lineage of a mouse kidney is sufficient to alter the microenvironment in which nephrogenesis occurs and can lead to WT-like phenotypes, they certainly do not prove that this occurs in WTs, as the lineage of the different cell types observed in WTs is still not known. Application of single cell transcriptomic techniques to WTs of known genetic background will certainly be enlightening. No matter what, our findings provide additional insights into the genetic programs driven by β-catenin in the developing kidney and suggest further studies are necessary to understand the role of stromal signaling in the development of WT.
MATERIALS AND METHODS
Catnbex3/+, Six2cre, Foxd1cre and TcreERT2 mouse lines in Mus musculus were used as previously described (Harada et al., 1999; Kobayashi et al., 2014; Kobayashi et al., 2008; Imuta et al., 2013). All mice were bred on a mixed genetic background. For experimental assays described below, Catnbex3/ex3 females were crossed with male Cre-line mice, with day of plug counted as E0.5. Pregnant females were sacrificed at various gestational time points. Lineage-tracing experiments were performed by crossing Rosa26YFP (JAX Stock #006148) or Rosa26Tomato (JAX Stock #007909) reporter mice with the above mouse lines. Mice with the desired genotype were randomly selected regardless of sex with Cre-negative littermates used as controls. Tamoxifen (Sigma, T5648) was administered by gavage at a dose of 2 mg per 40 g body weight. All animals were housed, maintained and used according to National Institutes of Health (NIH) and Institutional Animal Care and Use Committees (IACUC) approved protocols at the University of Texas Southwestern Medical Center (OLAW Assurance Number D16-00296).
Kidney sample preparation, immunostaining and in situ hybridization assays
Embryonic tissue was fixed in 4% paraformaldehyde, embedded in paraffin, sectioned into 5 μm slices, and subjected to Hematoxylin and Eosin staining or immunofluorescence (IF). Slides for IF were immersed and boiled with either 10 mM sodium citrate or TE antigen retrieval buffer, and blocked with a solution of 5% FBS/PBS for 1 h at room temperature followed by the application of primary antibodies diluted in blocking solution. The following antibodies were used: anti-GFP (Aves, GFP-1020, 1:200), anti-RFP (Rockland, 600-401-379, 1:200), anti-Ncam (Sigma, C9672, 1:200), anti-CK (DSHB, TROMA-I-s, 1:50), anti-Six2 (Proteintech, 11562-1-AP, 1:200; Abnova, H000110736-M01, 1:200), anti-β-catenin (Sigma, C7207, 1:200), anti-Lhx1 (DSHB, 4F2-c, 1:200), anti-Pax8 (Proteintech, 10336-1-AP, 1:200) and anti-Lef1 (Proteintech, 2230S, 1:200). For in situ hybridization assays, tissue was fixed with 4% paraformaldehyde, cryoprotected with 30% sucrose, embedded in OCT medium (TissueTek), sectioned into 10 μm slices, rehydrated with PBS before being treated with 15 μg/ml proteinase K for 10 min and fixed in 4% PFA followed by an acetylation step. Slides were then washed and incubated with pre-hybridization buffer for 1 h at room temperature before being hybridized with the specific probe overnight at 65°C. Slides were then washed in 0.2×SSC then transferred to NTT before blocking with 2% blocking solution (Roche) for at least 1 h at room temperature. Slides were then incubated with anti-Dig alkaline phosphatase-conjugated antibody (Roche, 1:4000) overnight at 4°C. The next day, slides were washed in 3×NTT and 3×NTTML before incubating with BM purple (Roche) for color reaction. After color reaction, slides were fixed with 4% PFA and mounted using Permount. E18.5 Six2cre;Foxd1cre;Catnbex3/+ bone-like tissue was fixed in 4% paraformaldehyde, decalcified with EDTA, sectioned into 5 μm slices, and subjected to Hematoxylin and Eosin staining, immunofluorescence and alkaline phosphatase staining performed by our institutional histology core.
RNA-sequencing and analyses of gene expression data from the TARGET database
RNA-Seq was performed on mouse kidneys (E12.5 whole kidneys; n=3 Cre-negative controls, 3 Foxd1cre;Catnbex3/+ mutants and 3 Six2cre;Catnbex3/+ mutants). RNA was isolated from dissected kidneys stored in RNA later solution (Invitrogen, AM7020). RNA-Seq was performed using single-end 75 bp with a minimum of 20 million reads per sample. Transcript abundance was estimated without aligning reads using Salmon (Patro et al., 2017) against an index of coding sequences from the Ensembl GRCm38 assembly. Transcript-level abundance was imported, and count and offset matrices generated using the tximport R/Bioconductor package (Soneson et al., 2015). Differential expression analysis was performed using the DESeq2 R/Bioconductor package (Love et al., 2014). WT expression data was downloaded from the TARGET database using the TCGAbiolinks R/Bioconductor package (Colaprico et al., 2016). A variance-stabilizing transformation implemented in the vst function of DESeq2 was applied prior to neural network processing.
To elucidate the relationship between activation of β-catenin in either Six2 or Foxd1 lineage cells and Wilms' tumor, we performed the following analysis. First, we used BETA (Wang et al., 2013) to integrate the results of RNA-Seq of whole kidneys from Foxd1cre;Catnbex3/+, Six2cre;Catnbex3/+ and wild-type comparator mice with the previously published β-catenin ChIP-seq data (Park et al., 2012). This resulted in a list of 2806 likely direct targets of β-catenin. A neural network was then trained to classify the expression profiles of Foxd1cre;Catnbex3/+, Six2cre;Catnbex3/+ and wild-type kidneys based on the expression of these 2806 genes. Next, the expression profiles for 124 WT samples curated in the TARGET database were mapped from human genes to mouse orthologues and then input to the neural network classifier. A sequential neural network with two hidden layers each containing 512 nodes was trained with the Adam optimizer using sparse categorical cross entropy loss using the TensorFlow platform (‘TensorFlow White Papers|TensorFlow’, n.d.). Five-fold cross validation was used during training. Classification scores ranging from 0 to 1.0 were assigned to each human WT sample for each mouse genotype (wild type, Foxd1cre;Catnbex3/+ or Six2cre;Catnbex3/+). The scores sum to 1 for each tumor sample and so can be interpreted as a probability of identity. None of the samples showed significant similarity to Six2cre;Catnbex3/+ mutants.
Data presented in figures are representative examples from one of at least three different experiments on at least three different embryos/organs. No significant variability was noted in tissues of the same genotype; all animals with correct genotypes were included in the analysis. Bioinformatic statistics were carried about as described above; algorithms and software availability are provided in Table S5.
We thank Andy McMahon for providing the OsrCreERT2 and Josh Mendell for providing Wt1CreERT2 mice. We also thank members of the Carroll lab, Denise Marciano and members of the Marciano lab, James Amatruda and Kenneth Chen, and members of the (Re-)Building a Kidney consortium for useful discussion.
Conceptualization: T.J.C.; Methodology: K.A.D., A.D., T.J.C.; Software: C.P.C.; Formal analysis: K.A.D., C.P.C., A.D., P.R., C.S.K., T.J.C.; Investigation: T.J.C.; Resources: D.R., T.J.C.; Writing - original draft: K.A.D., C.S.K., T.J.C.; Writing - review & editing: K.A.D., C.P.C., T.J.C.; Supervision: T.J.C.; Project administration: K.A.D., T.J.C.; Funding acquisition: K.A.D., T.J.C.
This study was supported by funding from the National Institute of Diabetes and Digestive and Kidney Diseases to T.J.C. (RO1DK095057, RO1DK080004, R24DK106743 and RO1DK080004). K.A.D. was partially supported by a UTSW SPORE CEP Award (P50CA196516) and by a U.S. Department of Defense KCRP Physician Research Award (W81XWH-19-1-0751). This work was supported by the UT Southwestern George O'Brien Kidney Research Core (DK079328). Deposited in PMC for release after 12 months.
RNA-Seq data presented in this manuscript has been deposited in GEO under accession number GSE150074.
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.189597.reviewer-comments.pdf
The authors declare no competing or financial interests.