The developmental programme for genesis of the entire kidney is recapitulated in Wilms tumour

Wilms tumour (WT) is an embryonal tumour that recapitulates kidney development. The normal kidney is formed from two distinct embryological origins: the metanephric mesenchyme (MM) and the ureteric bud (UB). It is generally accepted that WT arises from precursor cells in the MM; however whether UB-equivalent structures participate in tumorigenesis is uncertain. To address the question of the involvement of UB, we assessed 55 Wilms tumours for the molecular features of MM and UB using gene expression profiling, immunohistochemsitry and immunofluorescence. Expression profiling primarily based on the Genitourinary Molecular Anatomy Project data identified molecular signatures of the UB and collecting duct as well as those of the proximal and distal tubules in the triphasic histology group. We performed immunolabeling for fetal kidneys and WTs. We focused on a central epithelial blastema pattern which is the characteristic of triphasic histology characterized by UB-like epithelial structures surrounded by MM and MM-derived epithelial structures, evoking the induction/aggregation phase of the developing kidney. The UB-like epithelial structures and surrounding MM and epithelial structures resembling early glomerular epithelium, proximal and distal tubules showed similar expression patterns to those of the developing kidney. These observations indicate WTs can arise from a precursor cell capable of generating the entire kidney, such as the cells of the intermediate mesoderm from which both the MM and UB are derived. Moreover, this provides an explanation for the variable histological features of mesenchymal to epithelial differentiation seen in WT.


Introduction
The kidney develops from two distinct lineages involving a mutual inductive interaction between the metanephric mesenchyme (MM) and ureteric bud (UB) that emanates from the Wolffian duct [1]. This reciprocal inductive effect leads both lineages to develop into different PLOS  types of epithelial structures [1]. The MM differentiates into glomerular epithelia, proximal tubules, and distal tubules, while the UB differentiates into the collecting ducts. Wilms tumour (WT) is an embryonal tumour whose histology can consist of three major components (epithelial, blastemal and stromal) [2] which recapitulates the developmental phases of nephrogenesis [3], but in a disorganised manner. Some WTs additionally show myogenic and other mesodermal differentiation. In the WT that contain epithelial elements, we have noted that there are at least two different types of epithelial structures based on the morphology and WT1 immunohistochemistry [4]. One of these is a not fully differentiated epithelial structure, which is positive for WT1 immunostaining, and which might be equivalent to early glomerular epithelia of the developing kidney [4]. The other has the appearance of a fully formed epithelial structure, which is negative for WT1 immunostaining, suggesting that this type of epithelial structure is equivalent to the UB and its derivative, the collecting duct [4]. However, in spite of other supporting evidence including morphology [2,3], lectin histochemistry [5] and ROBO1 and SLT2 immunohistochemistry [6], the presence of UB-like structures in WT remains unverified. This is possibly because a series of studies consistently demonstrated that WTs are monoclonal [7][8][9] which can be seen as inconsistent with the involvement of two different tissue origins. Consequently, it has been generally acknowledged, albeit with some uncertainty, that WT originates from precursor cells in the MM [10][11][12].
Nephrogenic rests (NR) are precursor lesions found within WT-bearing and WT predisposition syndrome-associated kidneys, which are composed of remnants of embryonal cells [2,13]. NRs are categorized into two major types: intralobar nephrogenic rests (ILNRs) and perilobar nephrogenic rests (PLNRs) [2,13]. ILNRs occur at deeper sites within the renal lobe and ILNR-derived tumours normally show mesenchymal (rhabdomyogenic) differentiation. PLNRs are located at the periphery of the renal lobe and PLNR-derived tumours typically have blastemal-or epithelial-predominant histology lacking myogenic differentiation. The localization of NRs and histologic elements of their associated tumours suggests the origins of precursor cells.
This study focuses on the identification of UB-like structures in WT using gene-expression profiling validated by immunofluorescence and immunohistochemistry. The results support the hypothesis that WT can arise from a precursor cell with the capacity to differentiate into both the MM and UB epithelial lineages and thereby recapitulate all phases of kidney development [14].

Materials and methods WT samples
Forty-nine frozen tissues were available for RNA extraction. Total RNA was isolated from frozen samples using Tri Reagent (Molecular Research Center, OH, USA) and RNeasy purification columns (Qiagen, Hilden, Germany). These 49 New Zealand tumours (41 WT1 wild-type, 8 WT1-mutant) and additional six Japanese tumours (4 WT1 wild-type, 2 WT1-mutant) and three fetal kidneys (21,22, and 37 weeks gestation) were prepared for formalin-fixed paraffinembedded tissue sections. The New Zealand tumours were molecularly characterized and previously reported [4,14,15]. Ethics approval for this study was obtained from North Health Ethics Committee, Auckland, New Zealand and Tokyo Metropolitan Children's Medical Center Ethics Committee. The Ethics Committees waived the need for consent from all patients on condition that the confidentiality of personal information was protected. Fetal kidneys were obtained at autopsy at Tokyo Metropolitan Children's Medical Center after obtaining written comprehensive consent for utilizing all organ specimens for medical research and education.

Microarray analysis
Affymetrix Human Genome U133 Plus 2.0 microarray data from Fukuzawa et al. [14] was used in this study, and 13 additional tumours were analysed on the same platform following identical protocols. Microarray data are available in the ArrayExpress database (www.ebi.ac. uk/arrayexpress) under accession number E-MTAB-3516. Microarray pre-processing was performed as described previously [14]. Differential expression analysis was performed with the ANOVA tool in Partek Genomics Suite (Partek, St. Louis, MO) with a false discovery rate cutoff of 5%.

Immunohistochemistry and immunofluorescence
Immunohistochemistry (IHC) and immunofluorescence (IF) were performed according to standard protocols. The primary antibodies used in this study are listed in

Literature search
We primarily used a database of mouse kidney development known as Genitourinary Molecular Anatomy Project (GUDMAP, www.gudmap.org) because no database for human kidney development is currently available. The GUDMAP database describes anchor genes whose expression is restricted to one temporospatial anatomical compartment and marker genes which display a regional or enriched expression pattern [16,17]. Pubmed was also used to search for additional candidate genes for the developing kidney structures based on previous review articles [1,[18][19][20][21] and a kidney development database (http://golgi.ana.ed.ac.uk/ kidhome.html).

Identification of ureteric bud genes and ureteric bud-like structures
We undertook the following approaches to identify ureteric bud (UB)-like structures in WTs: (i) We compared the gene expression profile of triphasic tumours with that of blastemal predominant tumours. (ii) We searched the differentially expressed genes to identify candidate genes of UB and collecting duct (CD) primarily from the GUDMAP database and Pubmed (Tables 2 and 3, Fig 1). (iii) The localization of the candidate genes for UB and CD was validated in the nephrogenic zone of fetal kidneys by IHC and IF and compared with WTs. (iv) To differentiate UB-derived epithelial structures from MM-derived ones, we characterized the structures around the UB-like structures. (v) We similarly searched candidate genes for pretubular aggregate (PA), condensing mesenchyme (CM), renal vesicle (RV), comma-shaped body (CSB), S-shaped body (SSB), early proximal tubule (ePT) and distal tubule (eDT), immature Henle loop (imHL) listed from the differentially expressed genes (Tables 2 and 3, Fig 1).
Likewise, their locations were validated by IHC and IF in the fetal kidney and compared with WTs.

Wilms tumour histology
Tumours were classified into four subtypes according to the predominant histological elements. Triphasic tumours were composed of blastemal, epithelial and stromal elements. Triphasic tumours predominantly had a central epithelial blastema pattern with aggregations of blastemal cells around the UB-like epithelial structure, forming a nodular condensed structure (Fig 2A). The central epithelial blastema pattern is commonly seen in triphasic histology (Dr Beckwith personal communication, Fig 2A) and resembles the induction/aggregation phase of the developing kidney. Consequently, this structure represents a morphological landmark for tumours containing UB-equivalent structures. The representative features of blastemal-predominant tumours, epithelial-predominant tumours, and stromal-predominant tumours are described in the Fig 2 legend (Fig 2B-2D). Of ten tumours with WT1 mutations, nine had a stromal-predominant histology. Tumours with WT1 mutations mimic the entire kidney development with divergent mesenchymal differentiation. UB-like structures typically found in the central epithelial blastema pattern are also a histological characteristic of WT1-mutant WTs ( Fig 2D).

Ureteric bud and metanephrogenic epithelial development gene expression signatures in WT
We investigated whether WTs contained molecular features of nephrogenesis that correspond to the two distinct epithelial cell lineages: epithelial structures derived from the MM and those derived from the UB.  Table 2). We performed a genome-wide differential expression analysis between six tumours with triphasic histology and 16 tumours with blastemal-predominant histology, and then searched for GUDMAP anchor and marker genes in the resulting gene list. Differentially expressed anchor and marker genes for UB, CD, CM, PA, RV, CSB, SSB, ePT, imHL, and eDT identified from the GUDMAP database are listed in Table 2. There was only one UB anchor gene (SLCO4C1) in the GUDMAP database, which was differentially expressed in the triphasic tumours relative to the blastemal tumours. Of six anchor genes for CD, two genes (CLMN, FAM129A) were found to be differentially expressed. A marker gene for CD, KITLG was also identified. There were no anchor genes for CM, PA, CSB, and eDT in the GUDMAP database. The number of anchor genes for RV, CSB, ePT, and imHL in the data was one, one, 25, and one respectively, of which five anchor genes for ePT (CRYL1, FBP1, SLC27A2, TCN2, SLC3A1) were included. The following marker genes were differentially expressed in the triphasic tumours relative to the blastemal tumours: PA (LHX1), RV/CSB/SSB (BMP2, CDH6, JAG1, PAPSS2), ePT and/or imHL (CIDEB, CLIC6, UNC5CL, VDR), and early DT/imHL (KCNJ1). Marker genes differentially expressed in the blastemal tumours relative to triphasic tumours included CM (EYA1, HOXA10, NR2F2) and PA (WT1) genes. These data suggested that WTs contain structural elements of the whole kidney.
Many genes up-regulated in blastemal-predominant tumours compared to triphasic tumours (Table 3) are known to be involved in early kidney development (stromagenic and  Red and green bars show average log2 normalized microarray expression levels of WT1-mutant and triphasic tumours respectively. Error bars represent standard error of the mean. GUDMAP marker genes are labeled on the x-axis in blue, anchor genes in black. Some genes are shown multiple times as multiple array probe sets were identified as differentially expressed. Gene names enclosed in black squares are those with expression patterns determined by IHC/IF in this study. Genes from each structure were analysed by two-way ANOVA with Bonferroni post-tests, except for DT / KCNJ1, which was analysed by two-sided unpaired t-test. Significant interaction in the ANOVA model between tumour type and genes were identified in following structures, suggesting that genes marking these structure were induction/aggregation phase), including the HOX genes (HOXA 10, 11,HOXC 4,5,6,9,HOXD 3,9,10,11) [30], the sine oculis homeobox homologues (SIX1 [31], SIX2 [32]), CITED1 [11], and ITGA8 [33,34] (Table 3). FOXC2, which is the first gene expressed from the podocyte-committed region [20,35], was highly expressed in the epithelial-predominant tumours containing abundant renal vesicle (RV)-like structures ( Table 3,

Immunohistochemistry and immunofluorescence for molecules relevant to kidney development in fetal kidneys
We performed IHC and IF for developing kidneys to validate the localization of the representative proteins whose genes were differentially expressed in triphasic tumours relative to blastemal tumours: Expression of GUDMAP anchor genes (SLCO4C1, CLMN, and FAM129A) and marker gene (KITLG) and putative marker proteins from the literature (MUC-1, CDH1, PKHD1, KRT7) in the UB/CD was ascertained. Additionally, proteins not listed in the profile of differentially expressed genes but known to be expressed in UB/CD, SOX9 (GUDMAP), AQP-2 [37], and WNT9B [38] were assessed. MUC-1, E-cadherin, CLMN, KRT7 encoding Cytokeratin 7 (CK7), and AQP-2 showed a differential localization pattern in the UB/CD of the nephrogenic zone of fetal kidneys (Fig 4A, 4C, 4E, 4F, 4M and 4O). Expression of SLCO4C1 encoding OATP-H was detected in the UB, CM, and RV. While its localization in the UB was observed in the cell membrane, that in RV was predominantly in the apical membrane (Fig 4I and 4K). Therefore, we judged that its localization pattern in the UB is different from that in RV although it is not as discrete as E-cadherin and CLMN. Expression of FAM129A was low in the UB and CD (S1A Fig). The expression of KITLG, PKHD1 encoding polycystin, SOX9, and Wnt-9b was observed in the UB/CD but they did not show differential localizations (S1C Fig and S2A, S2C and S2E Fig).
Next, the immuno-localizations of CDH6, JAG1 and beta-catenin and PAX8 [39] in RV, CSB, SSB, and SYNPO in early GL, SLC3A1 encoding NBAT in early PT, and SCNN1A encoding ENaCA and KCNJ1 in early DT and immature HL were examined. The localizations of beta-catenin, PAX8, and JAG1 ( We carried out IHC and IF for developing kidneys to examine the localization of the representative proteins whose genes were differentially expressed in blastemal tumours relative to triphasic tumours: HOXD11 was ascertained as a maker for renal stroma and MM (Fig 5A), which were localized in the uninduced mesenchyme and CM. Expression of SIX2, CITED1, and ITGA8 in CM was examined. SIX2 and CITED1were localized in CM (Fig 5C and 5E). ITGA8 was expressed in CM, RV, and eGE ( Fig 5M). WT1 is a GUDMAP marker gene for PA, which was expressed in CM, PA, RV, CSB, SBB, and glomeruli (Fig 4A, 4C, 4E, 4I and Fig 5A, 5C, 5G, 5I, 5K, 5M and 5R). These fetal kidney IHC and IF results provide a basis for comparison with WT.

Identification and validation of ureteric bud/collecting duct-equivalent epithelial structures in WTs
We examined several protein markers that distinguish UB from MM in normal developing kidneys by IHC in the series of 55 WTs. We then selected representative tumours from consistently expressed at higher levels in triphasic tumours: RV, CSB, SSB p = 0.0003; PT p = 0.0008. * = p < 0.05, ** = p < 0.01, *** = p < 0.001. https://doi.org/10.1371/journal.pone.0186333.g003 Nephrogenic origin of Wilms tumour WT1 is highly expressed in early metanephrogenic epithelial structures in fetal kidneys, while its expression is absent in the UB of the developing kidney [4]. Meanwhile, MUC-1 is known to be expressed in the UB of the developing kidney [22,40], which was differentially expressed in the triphasic tumours in our microarray data. We examined whether the WT1-negative and MUC-1 positive (WT1 -/MUC-1 + ) epithelial structures (Fig 4B, S5A and  S5B Fig) in WTs are equivalent to the UB of the developing kidney. Expression of either WT1 or MUC-1 was observed in two different mutually exclusive epithelial structures RV-like tubules and UB-like structures respectively in the 39 WTs with detectable WT1 protein ( Fig  4B). Of 10 tumours with WT1 mutations, WT1 protein was detectable in two tumours both of which had constitutional missense WT1 mutations. Although the remaining eight WT1-mutant tumours lacked WT1 protein expression, MUC-1 was consistently positive in UB-like structures. Therefore, UB-equivalent structures are also present in all WT1-mutant tumours. Consistent with the reciprocal expression of WT1 and MUC-1, the remaining six blastemalpredominant tumours that all lacked WT1-negative epithelial structures were devoid of immuno-reactivity for MUC-1 in the examined sections, indicating the absence of UB-like structures.
Since E-cadherin shows a differential localization pattern in the UB compared to the early glomerular epithelium in developing kidneys (Fig 4C) [20], its expression was also examined in WTs. High levels of membranous E-cadherin expression in the WT1 -/MUC-1 + epithelial structures highlighted the UB-like structures, while its expression was weak and restricted to the cell membrane in WT1 + /MUC-1 -RV-like structures (Fig 4D, S5C and S5D Fig).
Although Fibrocystin, SOX-9, and Wnt-9b did not show differential localization patterns in the fetal kidney (S2A, S2C and S2E Fig), these proteins were expressed in the MUC-1 + UB/ CD-like structures in WTs (S2B, S2D and S2F Fig). Unexpectedly, there were some inconsistencies between our results and the GUDMAP data. FAM129A and KITLG did not show differential localization patterns in UB/CD of the developing kidneys and WTs (S1A-S1D Fig). This might be due to differences in the species (mouse and human), detection methods and products (RNA and protein), and developmental stages: we used the nephrogenic zone of predominantly second trimester fetal kidneys.
GDNF and its receptor Ret provide induction signals essential for the initiation of kidney development [18,41]. GDNF and Ret have been shown to be expressed in WTs [10,42]. We confirmed the expression of both genes in our cohort by QPCR (S4A Fig). We ascertained coexpression of GDNF and Ret in fetal kidneys and WTs using double-immunofluorescence staining. In the fetal kidneys, while Ret was localized in the renal tubular structures including UB/CD (S2H Fig), GDNF expression was faint around and in the UB (S2G and S2H Fig). On the other hand, GDNF-expressing blastemal cells tightly aggregated around the UB-equivalent epithelial structure whereas Ret was localized mostly to the membrane of the UB-equivalent structures (S2I Fig), suggesting the potential for blastemal and ureteric epithelial cells to signal to each other. The signalling that appears to occur in this central epithelial blastema pattern (Fig 2A) corresponds to the induction/aggregation stage of the developing kidney.

Molecular and morphological characterization of the epithelial structures derived from the metanephrogenic mesenchyme
Genes that were more highly expressed in specific tumour subtypes (Tables 2 and 3) are likely to be those that are required for different components of nephrogenesis. To complete the identification of developing kidney structures, genes that were expressed highly in WTs with blastemal-predominant and triphasic histology were selected for analysis by IHC and IF (Table 1). Further support for the presence of UB-like epithelial structures in WTs was sought by confirming the expression of markers of the metanephric mesenchyme lineage in the other surrounding epithelial structures. We additionally addressed each developmental stage in order to verify the presence of the full kidney developmental programme, that can be seen in WTs, including: a. metanephric specification and stromagenesis, b. condensation of metanephric mesenchyme, c. transition from metanephric mesenchyme to renal vesicle, d. segmentation of glomerulus-, proximal tubule-, and distal tubule-equivalent structures, and vascularization). The strategy for IHC/IF based on the nephrogenic programme below (a-d) is further summarized in Fig 6. (a) Metanephric specification and stromagenesis (Fig 6-1, 2). Embryonal cells that surround the UB-equivalent epithelial structure, called blastemal cells, are considered to be equivalent to the MM. HOXD11, a specific marker of the metanephric region including the MM and the renal stroma [43], was expressed in the nuclei of the blastemal cells and stromal cells of WTs (Fig 5B; cf. Fig 5A). This confirms that the blastemal and stromal components of WTs are truly derived from the metanephric lineage. Expression levels of FOXD1, a marker for stromagenesis, were ascertained by QPCR and shown to be highest in the blastemal-predominant tumours whose histology includes few or no UB-like structures (S4B Fig). FOXD1 is considered to be expressed in non-aggregated blastemal cells, which might share characteristics with the uninduced mesenchymal cells of the developing kidney. Therefore, expression in blastemal-predominant tumours would be expected to be higher than in triphasic or epithelial-predominant or stromal (WT1-mutant) tumours where the aggregating blastemal cells would express lower levels of FOXD1. (b) Condensation of metanephric mesenchyme (Fig 6-3). We next examined cap mesenchyme-equivalent structures in WTs, which are composed of densely packed MM cells closely associated with the tip of the UB in the developing kidney. Expression of WT1 overlapped with that of SIX2 (Fig 5D; cf. Fig 5C) in aggregating blastemal cells, which are considered to be equivalent to cap mesenchyme. The presence of both SIX2 and CITED1-positive cells within the cap mesenchyme of WTs (Fig 5F; cf. Fig 5E) indicates the ability to generate the entire range of epithelial structures derived from the MM, because SIX2 + /CITED1 + cap mesenchyme cells in the developing kidney are nephron-committed progenitor cells capable of forming all segments of the nephron from podocyte to connecting tubule [20].
(c) Transition from metanephric mesenchyme to renal vesicle (Fig 6-3, 4). RV-like structures were characterized as follows. (i) It is thought that canonical Beta-catenin WNT signaling is activated through the secretion of WNT9B from the UB [18]. Beta-catenin is required for renal vesicle induction, which activates early stage nephrogenic markers such as PAX8, WNT4, and LHX1 [18,44]. Beta-catenin and Pax-8 were expressed in the RV-like structures where WT1 was expressed (Fig 5H and 5J; cf. Fig 5G and 5I). (ii) WNT4 also responds to WNT9B and prompts the expression of LHX1 [20,45], which is a distal nephron fate determinant that regulates the expression of another distal gene CDH6 encoding K-cadherin [20,46]. K-cadherin was located in the RV-like epithelial structures (S3B Fig; cf. S3A Fig). (iii) NOTCH2 functions downstream of LHX1 and determines a proximal nephron fate [20,46]. Jagged-1, a NOTCH signalling receptor that is also involved in the formation of the proximal tubule [20,46], was expressed in the RV-like epithelial structures (Fig 5L; cf. Fig 5K). Integrin alpha 8, which is required for mesenchymal to epithelial transition in developing kidneys [33], was highly expressed in the extracellular matrix of the tightly condensed MM and RV-like structures (Fig 5N; cf. Fig 5M). This indicates that the committed cap mesenchyme in WTs can also differentiate into renal vesicles, confirming that polarized epithelial structures expressing WT1 adjacent to UB-equivalent structures are equivalent to renal vesicles. The expression of WNT4, FOXC2, LHX1, and NOTCH2 was confirmed by microarray and/or QPCR (S4C Fig) because effective antibodies were not available.
(d) Segmentation of glomerulus-, proximal tubule-, and distal tubule-equivalent structures, and vascularization (Fig 6-5, 6). Proximal and distal tubule (PT, DT)-equivalent epithelial structures were identified in association with the presence of the glomerulus (GL)-like structure, supporting that the hypothesis that WTs recapitulate the kidney developmental programme (Fig 6): RV-equivalent epithelial formation is considered to be followed by genesis of GL-, PT-and DT-equivalent structures.
GL-like structures, which are occasionally found in WTs, are considered to be equivalent to presumptive podocytes in developing kidneys because of co-expression of podocyte markers, Synaptopodin and WT1 (Fig 5P and 5Q; cf. Fig 5O). Tubular structures equivalent to the PT were morphologically ill-defined in WTs; however SLC3A1, a functional marker for PT encoding Neutral and basic amino acid transport protein rBAT (NBAT) [28,47], showed overexpression in our microarray data ( Table 2). The localization of NBAT in the tubules connected to GL-equivalent structures confirmed the presence of PT-equivalent structures ( Fig  5S; cf. Fig 5R). Both KCNJ1 and ENaCA co-localized with MUC-1 in well-differentiated thin tubular structures, confirming the formation of DT-equivalent structures [27] (Fig 5U, V; cf. Fig 5T, S3D and S3E Fig; cf. S3C Fig).
Incorporation of vessels into the GL-equivalent structure, which is considered to correspond to the "vascularization phase" of WTs and developing kidneys was validated by endothelial markers (CD31, CD34) and alpha smooth muscle actin (SMA) [48](S6B, S6D and S6F Fig; cf. S6A, S6C and S6E Fig).

Discussion
We have verified that UB-like structures in WT have the protein expression characteristics of UB-derived structures in normal fetal kidney. The lack of established specific markers for UB in humans and occasional overlap of putative UB markers with other epithelial structures such as the distal nephron that occur during development represent a limitation for the pathological identification of UB-like structures in WTs. Here we utilized the histological central epithelial blastema pattern as a clue for distinguishing UB-like from RV-like structures, and furthermore used a validation of GUDMAP data on human fetal kidney tissues to better characterize lineage specific markers. Then, by examining the staining pattern and localization of proteins of putative UB markers and by differentiating them from surrounding epithelial structures, we concluded UB-like structures are involved in WTs.
The possibility that these UB-like structures are normal ureteric buds embedded in the WT can be excluded as ureteric buds do not exist in the normal juvenile kidney tissues resected with WT, having disappeared at around 34-36 weeks of gestation. Therefore, normal ureteric buds cannot be involved in Wilms tumour tissues. Ureteric buds can be seen in nephrogenic rests which are pre-neoplastic lesions, most of which have genetic and/or epigenetic alterations [49][50][51]. In addition, UB-like structures are morphologically obviously neoplastic because they have nuclear and structural atypia.
Through detection of this UB-equivalent component, we therefore show that the majority of WTs possess the ability to differentiate into the two lineages of nephrogenesis: the metanephric mesenchyme (MM) and the ureteric bud (UB). The MM and the Wolffian ductderived UB are two independent tissue compartments that arise from the precursor embryonic tissue, the intermediate mesoderm [1]. All seven tumours carrying constitutional WT1 mutations had the protein-expression characteristics of UB-derived structures, supporting the hypothesis that stem cells with the initiating mutations develop into both tumorigenic MM and UB cells. Our observations, together with the well established understanding of the ontogenesis of the fetal kidney [1], strongly support an early stem cell origin of WT. Since, as expected, WTs are monoclonal [7][8][9], the presence of the two epithelial structures is best explained by an intermediate mesoderm-like cancer stem cell (nephrogenic stem cell) that has the potential to differentiate into both structures. In addition, it is conceivable that the interplay between the two lineages in WT drives the nephrogenic developmental programme.
It has been reported that cells within the MM can differentiate into both metanephrogenic and ureteric epithelia in vitro [52], suggesting the persistence of pluripotent renal stem cells (nephrogenic stem cells) within the MM tissue of the developing kidney. We and other groups have shown that ILNR-associated tumours have molecular features that are characterized by genes expressed in the paraxial mesoderm [14,15] and intermediate mesoderm [53]. The location of NRs within the kidney and the histological features (presence or absence of myogenesis) of the associated tumours indicate ILNRs occur early in kidney development, while PLNRs are formed late in nephrogenesis [2,13] (Fig 7). Therefore, ILNRs predominantly contain precursor cells committed to nephrogenic and divergent mesenchymal lineages, while PLNR is predominantly composed of precursor cells with more restricted potential committed to the nephrogenic lineage.
The sequential nature of nephrogenic rest-WT development, and our past [14,15] and current expression profiling data of WTs indicate that WT can arise from precursor cells of different maturational states (Fig 7). A large expression-profiling study also demonstrated that ILNR-derived tumours had two different classes of molecular characteristics and different proportions of myogenic elements [53], suggesting the existence of precursor cells of different maturational stages (developmental potential) within NRs. Interestingly, a group of tumours were not associated with NRs and showed post-induction nephron signatures. We speculate that those tumours might be derived from a metanephric precursor cell whose differentiation is limited to metanephrogenic epithelia (Fig 7).
The characterization of gene and protein expression of metanephric and ureteric epithelia enabled us to interpret the histogenesis of WT. It is conceivable that as the MM and UB-equivalent epithelial cells interact with each other, the efficiency of interactions between the components and/or the number of the ureteric epithelium could determine the shape and proportion of the blastemal component (Fig 6). Appropriate interactions between the MM and UB lead to metanephrogenic epithelial development, which typically forms RV-equivalent epithelium that sometimes gives rise to GL-, PT-, and DT-equivalent structures sequentially (Fig 6) according to the kidney developmental programme. Such metanephrogenic epithelial development cannot proceed if WT1 is mutated as we have shown in WT1-mutant WTs [4,15]. WTs with an epithelial-predominant histology represent the stage of metanephric epithelial development in which RV-like epithelial structures are prominent (Figs 2C and 6). If the ureteric bud component is absent or minor, a blastemal-predominant histology would result (Figs 2B and 6). A subset of blastemal-predominant tumours might be purely of metanephrogenic precursor cell origin (Fig  7), as those tumours lack UB-equivalent structures. Such tumours might have a restricted range of differentiation and only show metanephric blastema and its derived epithelia.
Recent studies on cancer stem cells (CSCs) from WT xenograft models have proposed a nephrogenic/stromal developmental lineage origin for WT [54,55]. A series of studies on a population of cells marked by NCAM1 and ALDH1 with CSC characteristics identified in 5 lines of blastemal-predominant phenotype WT xenografts derived from two triphasic WTs suggested that self-renewing human WT CSCs can constantly evoke blastemal and differentiated renal tubular elements that exist in the blastemal element of WT [54]. Furthermore, a study demonstrated that this ALDH1+ WT CSC population expressed markers suggestive of commitment to epithelial lineages, but could also give rise to cells expressing mesenchymal/ stromal lineage markers, suggesting the ability to dedifferentiate in vitro [55]. These data do not directly support our results identifying the presence of UB components. However while the models characterize a type of WT histology, they are not able to fully explain the characteristic features of classical WTs containing ectopic mesenchymal elements such as tumorigenic rhabdomyoblasts. Dedifferentiation of CSCs is one explanation for stromal and divergent mesenchymal differentiation. However the two precursor lesions of WTs, in which CSCs are considered to reside, might have a clue to understand the inconsistency between the currently described experimental CSC models and the observed spectrum of WT histology. PLNR is composed predominantly of an embryonal type of cells (blastemal cells). Tumours arising from PLNR are predominantly blastemal with occasional epithelial components and usually lack ectopic mesenchymal elements, which closely resemble the aggregation phase [condensing blastema (mesenchyme)] of the developing kidney. The CSCs within PLNR-derived WTs have restricted potential committed to the nephrogenic lineage, which is consistent with the published CSC models derived from blastemal-predominant WT xenografts. Meanwhile, ILNR are composed of spindle cells considered to be equivalent to uninduced mesenchymal cells, which correspond to the stromagenic phase (non-aggregated nephrogenic mesenchyme) of the developing kidney. Tumours derived from ILNR are composed of three components (blastemal, epithelial, and stromal), of which stromal components are abundant including ectopic mesenchymal elements, and often show the entire range of kidney epithelial structures. The term "stroma" in WT does not mean connective tissue stroma [3](J.B.Beckwith personal communication). Thus, the stromal component is not derived from the blastemal component, but the blastemal component is derived from the stromal component [3](J.B.Beckwith personal communication). Therefore, it is conceivable that the CSCs within ILNR-derived WTs are more primitive state than those within PLNR-derived WTs. The published CSCs were obtained from human WT blastemal components, which would be expected to contain committed renal progenitors. If CSCs could be obtained from ILNR-derived WTs and their differentiation potential assayed, our question of the involvement of UB and the origin of ectopic mesenchymal elements might be clarified. Supporting our model, a transgenic mouse model of WT based on induction of Lin28 into nephrogenic lineage (wt1/six2) cells of the intermediate mesoderm generated triphasic WT, but no tumours were formed with induction of the Lin28 transgene in more differentiated cell types such as the cap mesenchyme, ureteric bud, or stroma [56]. The histology of the mouse WT in that study further shows a triphasic histology containing a central epithelial blastema pattern, supporting the involvement of a ureteric bud equivalent epithelial component in WT derived from the intermediate mesoderm cells which give rise to the entire kidney [56].We suggest that WTs arise from different maturational stages of CSCs (Fig 7), determining their histopathological features (Fig 6). From our expression-profiling studies, we reaffirmed the molecular definition of embryonal tumours: embryonal tumours follow the developmental programme of the organ from which they arise and a cellular lineage-associated oncogenic programme. However, malignant embryonal tumours show a failure of complete differentiation of their original organs. Similarly, it is anticipated that cells with pluripotent properties such as Embryonic Stem cells and inducible Pluripotent Stem cells could become germ cell tumours and embryonal tumours according to their developmental stages, if an oncogenic programme is activated.  Fig. Expression of FOXC2, LHX1, WNT4, and NOTCH2 in WTs. A) Expression of GDNF and cRET in WTs. GDNF and its receptor cRET are expressed in all types of WTs, with higher expression of GDNF seen in WT1-mutant and blastemal-predominant tumours. Data shown are log2-transformed QPCR expression levels normalized to UBE2G2. B) Expression of FOXD1 in WTs. Expression of FOXD1 is highest in blastemal-predominant tumours whose histology consists of non-aggregated blastemal cells with few or no UB-like structures. Data shown are log2-transformed QPCR expression levels normalized to UBE2G2. C) Expression of FOXC2, LHX1, WNT4, and NOTCH2 in WTs. FOXC2 and LHX1 are highly expressed in tumours with epithelial-predominant histology, which are predicted to be expressed in RV-equivalent structures. NOTCH2 was expressed in all histological-subtypes, with the highest expression in epithelial-predominant tumours. The graph for WNT4 shows log2-transformed QPCR expression levels normalized to UBE2G2. The second graph shows microarray log2-transformed normalized expression levels. For all graphs expression levels were compared by two-way ANOVA with Bonferroni posttests, Ã = p < 0.05, ÃÃ = p < 0.01, ÃÃÃ = p < 0.001.