Multiple Imprinted and Stemness Genes Provide a Link between Normal and Tumor Progenitor Cells of the Developing Human Kidney

Wilms' tumor (WT), the embryonic kidney malignancy, is suggested to evolve from a progenitor cell population of uninduced metanephric blastema, which typically gives rise to nephrons. However, apart from blastema, WT specimens frequently contain cells that have differentiated into renal tubular or stromal phenotypes, complicating their analysis. We aimed to define tumor-progenitor genes that function in normal kidney development using WT xenografts (WISH-WT), in which the blastema accumulates with serial passages at the expense of differentiated cells. Herein, we did transcriptional profiling using oligonucleotide microarrays of WISH-WT, WT source, human fetal and adult kidneys, and primary and metastatic renal cell carcinoma. Among the most significantly up-regulated genes in WISH-WT, we identified a surprising number of paternally expressed genes (PEG1/MEST, PEG3, PEG5/NNAT, PEG10, IGF2, and DLK1), as well as Meis homeobox genes [myeloid ecotropic viral integration site 1 homologue 1 (MEIS1) and MEIS2], which suppress cell differentiation and maintain self-renewal. A comparison between independent WISH-WT and WT samples by real-time PCR showed most of these genes to be highly overexpressed in the xenografts. Concomitantly, they were significantly induced in human fetal kidneys, strictly developmentally regulated throughout mouse nephrogenesis and overexpressed in the normal rat metanephric blastema. Furthermore, in vitro differentiation of the uninduced blastema leads to rapid down-regulation of PEG3, DLK1, and MEIS1. Interestingly, ischemic/reperfusion injury to adult mouse kidneys reinduced the expression of PEG3, PEG10, DLK1, and MEIS1, hence simulating embryogenesis. Thus, multiple imprinted and stemness genes that function to expand the renal progenitor cell population may lead to evolution and maintenance of WT. (Cancer Res 2006; 66(12): 6040-9)

Metanephroi are the primordia of adult mammalian kidneys (1). The metanephric blastema contains multipotential precursors that give rise to all cell types of the nephron (2). Errors in nephrogenesis that lead to human disease are common (1). They include congenital abnormalities (e.g., renal dysplasia) and cancer. Wilms' tumor (WT) or nephroblastoma is a pediatric kidney cancer believed to arise from multipotent embryonic renal precursors of the metanephric blastema (3), which fail to terminally differentiate into epithelium and continue to proliferate, thus forming blastemal elements in the tumor. Nevertheless, WTs contain structures at different stages of maturation that mimic those present in the nephrogenic zone of the growing fetal kidney (tubular epithelia and stromal elements in addition to uninduced blastema), and also other mesoderm elements (rhabdomyoblats, cartilage), suggesting that blastemal cells have differentiated at least in part (3). Derived from primitive embryonic tissue and being strongly associated with various congenital syndromes, WT is an attractive model to study developmental pathways leading to cancer. In fact, WT had already provided significant information regarding the genetic and epigenetic events leading to the development of cancer in general (4–7). Global gene profiling during kidney development has provided novel insights into the genetic program that control murine and human nephrogenesis (8–11). Furthermore, molecular signatures of normal human kidney development set a reference for abnormal one (12). Accordingly, we could also show that the overall gene expression profile of a WT specimen was most similar to that observed for an 8-week human gestation kidney (11). Li et al. (13) have recently done more detailed experiments involving gene expression profiling in WT. They identified 357 genes as differentially expressed between favorable histology WTs and fetal kidneys (16-22 weeks of gestation). To determine stage-specific expression of these genes, they compared their data set to that previously obtained for normal rat nephrogenesis (8). One hundred twenty-four matches to genes on the microarray used by Stuart et al. (8) were found. Mapping between the two data sets showed that WTs systematically overexpressed genes corresponding to the earliest stage of metanephric development. Nevertheless, profiling gene expression in whole heterogeneous tissues, such as WT, is complicated by mixed populations of cells and is therefore less powerful for discovering genes involved in specific developmental processes. In addition, it is becoming clear that many, if not most, malignancies arise from a rare population of cells that exclusively maintain the ability to self-renew and sustain the tumor via the expression of tumor-progenitor genes (14, 15). Moreover, these “cancer stem cells” are often biologically distinct from the bulk of differentiated cancer cells that characterize the disease. In that regard, the analysis of WT xenografts established and propagated in immunodeficient mice, which pressure for metanephric blastemal maintenance/proliferation and the disappearance of differentiated tubular and stromal structures, is advantageous. The selection of a more homogenous subset of early metanephric cancer cells is of paramount importance as these are the cellular characteristics that might be clinically linked to morbidity.

Here, we present molecular analysis of these authentic WT xenografts, concomitant with fetal and adult kidneys, and renal cell carcinoma. By comparing the gene expression profiles and analyzing models of murine kidney development and regeneration, we have identified multiple imprinted and stemness-related genes that participate in early nephrogenesis and also signify maintenance and expansion of WT blastema. Thus, epigenetic changes, which regulate imprinted genes, may serve as a driving force for the human renal progenitor cell population affecting both organogenesis and tumorigenesis.

Animals and surgical procedures. Severe combined immunodeficient (SCID) (c.b-17/Icr beige or nonobese diabetic), nude (BALB/c nu/nu), and C57BL/6 mouse strains were obtained from the Weizmann Institute Animal Breeding Center (Rehovot, Israel). Animals used were 6- to 10-week-old. In all the surgical procedures, mice were anesthetized with i.p. injections of 100 mg/kg ketamine and 10 mg/kg xylazine.

Establishment of WT xenografts. WT xenografts were established from WT stage I with favorable histology. The original surgical samples were placed on ice, minced into 3 to 5 mm pieces, and implanted s.c. into SCID mice. After an initial latency period of 2 months, tumor growth was noted in 90% of the mice. Thereafter, the tumor was serially passaged s.c. with Matrigel (Becton Dickinson, Bedford, MA) as minced tumor pieces or by direct injection of single cell suspension.

Maintenance of WT xenografts. The xenografts were maintained by serial passages in SCID and nude mice by harvesting the tumors under sterile conditions and placing them immediately in cooled HBSS (Sigma-Aldrich Co., Ltd., St. Louis, MO). Tumor cells were dissociated under sterile conditions, first by mincing the tissue with scissors to small fragments and then by gentle mechanical homogenization through a stainless steel mesh. Viable cells were separated from debris by layering over Ficoll-Paque 400 (Pharmacia Biotech AB, Uppsala, Sweden) and centrifugation at 500 × g for 20 minutes. Viable cells at the interface were collected, counted, and resuspended in cooled HBSS in a concentration of 2 × 10⁷/mL. For orthotopic implantation, 100 μL tumor cell suspension (2 × 10⁶ cells) were directly injected into several spots of the mouse kidney using a 27-gauge needle. The musculature layer of the abdominal wall and the skin were separately closed with 4-0 absorbable Vycril sutures. Tumor growth was determined by caliper measurements of length, width, and depth, and the tumor volume (mm³) was approximated using the formula length × width × depth × 05236 (16). Doubling time of the tumor growth was calculated during the logarithmic growth of s.c. growing tumors.

Dissection of rat metanephric mesenchymes. Metanephric mesenchymes were dissected from E_13.5 rats after trypsinization and mechanical disruption of embryonic kidneys using minutien pins. Cross-contamination between ureteric buds and metanephric mesenchymes was ruled out by visual inspection and staining with dolichos bifloris lectin, which selectively stains the ureteric bud at this stage (10). For organ culture, rat metanephric mesenchymes were placed on filters (Corning Transwell, Corning, NY; collagen coated, 0.4-μm pore size) and grown in DMEM/F12 with insulin (5 μg/mL), transferrin (5 μg/mL), selenium (5 ng/mL), dexamethasone (5 μg/mL), prostaglandin (5 μg/mL), T3 (5 ng/mL; Sigma), fibroblast growth factor-2 (FGF-2, 3 nmol/L), transforming growth factor-α (TGF-α, 3 nmol/L), and leukemia inhibitory factor (LIF; 50 ng/mL; all cytokines from R&D Systems, Minneapolis, MN).

Ischemia/reflow experiments. For unilateral ischemia/reflow, a flank incision was made and the left renal pedicle was clamped for 40 minutes using a vascular clamp (Fine Science Tools, Inc., Foster City, CA). The abdomen was covered with gauze moistened in PBS, and the mice were maintained at 37°C using a warming pad. After 40 minutes, the clamp was removed and reperfusion was confirmed visually. To determine the extent of acute injury, control mice were sacrificed 24 hours after ischemia/reflow, and kidneys were collected and processed for histology using H&E and sirius red staining.

RNA isolation. Total RNA from human and mouse samples was isolated from each sample using TRIZOL (Life Technologies, Invitrogen, Carlsbad, CA).⁸

Total RNA from rat metanephric mesenchymes was extracted using RNeasy mini kit (Qiagen, Valencia, CA) with on-column DNase digestion according to the instructions of the manufacturer. An Agilent Bioanalyzer was used to confirm RNA integrity.

Microarray analysis. Malignant adult and fetal kidney specimens were obtained from kidneys removed for stage I clear cell carcinoma or WT. Adult kidney specimens were obtained from the normal kidneys removed for stage I clear cell carcinoma. Metastatic renal cell carcinoma was obtained from surgical remnants of biopsies. Fetal kidney tissue was obtained following curettage. Studies with human embryonic kidney tissue were approved by the Helsinki Ethical Committee. Total RNA was extracted and used as a template to generate double-stranded cDNA and biotin-labeled cRNA, as recommended by the manufacturer of the arrays and as previously described (17). Hybridization to a Genechip Human Genome HU95A oligonucleotide arrays containing 9,632 probe sets was done as described in the Affymetrix human_datasheet.pdf⁹

(Affymetrix, Santa Clara, CA).¹⁰ Data files were imported into a microarray database and then median scaled. Based on our previous experience, all expression levels <0.01 were brought to 0.01. For statistical and cluster analysis, we used the Cluster, Gene Cluster, Treeview programs, and Scoregene gene expression package.¹¹ A detailed description of the scoring methods and our approach to analysis of microarray data have been published (17). Genes were classified into functional groups using Go Annotation tools (DAVID: Database for Annotation, Visualization, and Integrated Discovery).¹² Overrepresentation calculations were done using Ease (17). Functional classifications with an Ease score <0.05 were marked as overrepresented. The complete set of gene array data can be found online.¹³

Quantitative reverse transcription-PCR. cDNA was synthesized using Omniscript Reverse Transcriptase (Qiagen) on total RNA. Real-time PCR of human and mouse samples was done using an ABI7900HT sequence detection system (Perkin-Elmer/Applied Biosystems, Foster City, CA) in the presence of SYBR green (SYBR green PCR kit; Qiagen, Hilden, Germany). This fluorochrome incorporates stoichiometrically into the amplification product, providing real-time quantification of double-stranded DNA PCR product. Primers were designed to amplify an 80 to 120 bp fragment with 50°C to 65°C annealing temperature (Supplementary Table S1). The relative initial amount of mRNA of a particular gene was extrapolated from a standard curve. For standard curve determination, we used a pool of all the samples, serially diluted in four log₂ steps and run in parallel to the samples. The total volume of each reaction was 20 μL, containing 300 nmol/L of each forward and reverse primer and 125 ng cDNA. Appropriate negative controls were run for each reaction. All of the reactions were done in triplicate. Optimization of the real-time PCR reaction was done according to the instructions of the manufacturer. For each analysis, transcription of the gene of interest was compared with transcription of the housekeeping gene β-actin, whose level of expression was not changed significantly according to the microarray data (data not shown) and which was amplified in parallel.

The real-time PCR reaction for rat samples contained iQ SYBR green super mix (Bio-Rad, Hercules, CA), 200 nmol/L of each primer, and 0.2 μL template in a 25 μL reaction volume. The following primers were used after exclusion of primer dimer formation and nonspecific amplicons for each primer set. Rat E-cadherin (Cdh1; Genbank accession no. NM_031334): sense, ACAACGCTCCCATCTTCAAC; antisense, TGTGGAAGGGACAAGAGACC. Rat DLK1 (Genbank accession no. NM_053744): sense, GCAGTGTGTCTGCAAGGAAG; antisense, ATCGTTCTCGCATGGGTTAG. Rat Meis1 (Genbank accession no. XM_223643): sense, ACCAACCTCAAGCCATTCAC; antisense, GTCCACTCATTGTCGGGTCT. Rat Peg3 (Genbank accession no. XM_218226): sense, ACGTTGAAGAGCCAGAAGGA; antisense, GAGAGGCGGTCATTGAAGAG. Rat β-actin (Genbank accession no. BC063166.1): sense, CTAAGGCCAACCGTGAAAAG; antisense, TCTCAGCTGTGGTGGTGAAG. Specific fragments were verified by sequencing. Amplification was carried out using the MyiQ Single-Color Real-time PCR Detection System (Bio-Rad) with incubation times of 2 minutes at 95°C, followed by 50 cycles of 95°C/30 seconds and 60°C/30 seconds. Specificity of the amplification was checked by melting curve analysis and agarose gel electrophoresis. Relative levels of mRNA expression were calculated according to the ΔΔC_T method. Individual expression values were normalized by comparison to β-actin mRNA expression. To estimate the statistical significance of the results, one-way ANOVA followed by Dunnett's test was used.

Immunofluorescence. Whole-mount metanephric mesenchymes were stained with antibodies to E-cadherin (R&D Systems) and WT-1 (Santa Cruz Biotechnology, Santa Cruz, CA) and Cy2- or Cy3-labeled secondary antibodies (Jackson ImmunoResearch, West Grove, PA). Confocal images were obtained using a Zeiss LSM 510 META scanning confocal microscope (Zeiss, Thornwood, NY).

Immunohistochemistry. Tissue microarray multi–tissue block was prepared as previously described (18). The tissue microarray block contained in addition to WT xenografts and primary WT, samples of normal human lung, liver, renal cortex, and renal medulla tissues used as internal positive and negative controls. Primary and xenografted WT samples were also embedded in separate paraffin blocks. All primary WT tumor samples were triphasic, including tubular, blastemal, and stromal components, which were present in different proportions. Sections, 4-μm thick, were cut from whole and tissue microarray blocks for immunohistochemistry and pro cessed within 1 week to avoid oxidation of antigens. Before immunostaining, sections were treated with buffer citrate (pH 6.0) in a microwave oven for antigen retrieval. The slides were subsequently stained by an automated immunostainer (NexES, Ventana, Tuscon, AZ) using an avidin-biotin complex staining procedure. Anti-KI-67, CD34, vimentin, and P53 mouse mono clonal antibodies (Zymed Laboratories, Inc., San Francisco, CA), at a dilution of 1:50, were used. Controls were prepared by omitting the primary antibodies or by substituting the primary antibodies with goat IgG isotype.

Xenograft characterization. WITH-WT originates from a donor representing the common variant of WT with favorable histology and typical clinical behavior. WITH-WT has rapid growth rate with a doubling time of 3.5 days (Fig. 1A). Not infrequently, metastasis can be detected in the retroperitoneal lymph nodes and in the liver following orthotopic implantation to the mouse kidney (Fig. 1B). The histologic features of fifth-generation xenografts show blastemal expansion and major loss of differentiated tubular and stromal structures in comparison with original WT donor and other primary favorable histology WTs (Fig. 1C). Moreover, immunohistochemistry of a tissue array, which included WISH-WT and primary WT and control sections, showed overexpression of the cell proliferation marker KI-67, the stem cell marker CD34, vimentin, and P53 in WISH-WT, indicating dedifferentiation in situ and a more aggressive disease phenotype (Fig. 2; refs. 19, 20).

Microarray analysis and hierarchical clustering. Gene expression profiles were determined in a sample of the WT donor tissue (WISH-WT₀), six WISH-WT tumor samples obtained from separate mice and derived from passages 3 (WISH-WT₃) and 5 (WISH-WT₅) of the xenografts, five human fetal kidney tissues (12-18 gestational weeks) and six adult kidney tissues, and six primary and two metastatic renal cell carcinomas. We subjected the expression profiles of all samples to a hierarchical clustering analysis to investigate similarities among them (17). Unsupervised hierarchical clustering of 9,362 valid genes done for all samples clearly distinguished embryonic-derived tissues (fetal kidney, WISH-WT_0,3,5) from adult-derived tissues (adult kidney, and primary and metastatic renal cell carcinoma) with close similarities of all WISH-WT samples and fetal kidney tissues (Fig. 3). In addition, cluster analysis of gene expression profiles showed embryonic and adult renal tumors (WISH-WT_0,3,5, and primary and metastatic renal cell carcinoma) to be more similar than normal kidney tissues (fetal and adult), especially WISH-WT₅ and metastatic renal cell carcinoma, suggesting universal nonspecific markers for advanced cancer. Data set containing expression levels of all genes and comparisons among the groups can be found online.¹³ The microarray expression profile of all WISH-WT samples (including WT₀) was examined by comparison with all renal cell carcinoma samples. This way, the “universal” cancer markers were excluded, whereas active developmental pathways were determined. In addition, microarray experiments and data analysis were extremely robust when all embryonic and adult tumor samples were included in single groups. A total of 779 genes were found to be differentially expressed (P < 0.01, >2.0-fold ratio), with 279 genes significantly overexpressed in WISH-WT and 500 genes down-regulated. Functional annotation of the differentially expressed genes according to DAVID, database for annotation, revealed that although most of those overexpressed in WISH-WT function in cell cycle, DNA-dependent transcription, and regulation of transcription (49.4% of the genes), renal cell carcinoma shows a predominance of defense response, inflammatory, and immune response genes (38.3%), with a decreased proportion of regulatory genes (17.5%; Fig. 3). As expected, WISH-WT overexpressed the nephrogenic patterning genes, PAX2, LIM1, EYA1, SIX1, SALL1, FOXC1, and WT1. In addition, we identified genes of the Wnt/β-catenin signaling pathway (FZD2, FZD7, SFRP1, and CTNNBIP1), which has recently emerged as a critical regulator of self-renewal signals of stem and cancer cells (21). These genes have been thoroughly described in kidney development (22, 23).

Fifty genes that were the most significantly elevated in WISH-WT compared with renal cell carcinoma (average fold ratio >3) are listed in Table 1. Strikingly, six of the highest ranked elements are classified as paternally expressed, maternally imprinted genes (IGF2, DLK1, PEG1/MEST, PEG3, PEG5/NNAT, and PEG10), where only the paternal allele is activated, whereas the maternal one is silenced by promoter hypermethylation (6). Of these genes, only NNAT and IGF2 have been previously implicated in WT tumorigenesis (6, 13), where loss of imprinting of the IGF2 has been described in 70% of WTs (24). Moreover, we identified additional highly overexpressed genes yet to be described in WT carcinogenesis. For instance, the Meis homeobox genes, myeloid ecotropic viral integration site 1 homologue 1 (MEIS1) and MEIS2, which encode for homeoproteins, are shown to be crucial for the self-renewal of hematopoietic stem cells and the program of a cancer stem cell character as well as for the control of blastema cells during vertebrate limb regeneration (25–27). Interestingly, HOXA9, which has been shown to cooperatively maintain the stem cell character with MEIS1 (26), was also significantly elevated in WISH-WT (expression fold, 2.8).

Paternally expressed genes and MEIS1 are overexpressed in WISH-WT compared with favorable histology WT. The observation that IGF2, DLK1, PEG1/MEST, PEG3, PEG5/NNAT, PEG10, and MEIS1 are all highly overexpressed in the embryonic versus adult renal cancer microarrays lead us to examine whether these genes are associated with progression of WT. We therefore analyzed transcript levels of selected genes in independent samples of favorable histology WT and WISH-WT₅ (three samples of each) using real-time PCR (list of primer sets are summarized in Materials and Methods). Quantitative reverse transcription-PCR (qRT-PCR) showed that IGF2, DLK1, PEG1, PEG3, PEG10, and MEIS1 mRNA levels were all significantly elevated in WISH-WT₅ compared with favorable histology WT (Fig. 4A). We could not establish a reaction for PEG5/NNAT as we might have had a poorly working primer set or a splice variant in WT. Thus, positive selection of the metanephric blastema and WT progression in vivo is accompanied by up-regulation of these genes.

WISH-WT overexpressed genes are developmentally regulated during mouse nephrogenesis. Having established that a cluster of paternally expressed genes and MEIS1 are significantly elevated in WISH-WT, we determined their expression in normal nephrogenesis. A microarray comparison between human fetal and adult kidney tissues showed 550 genes to be differentially expressed (expression fold, >2.0) of which 260 genes were up-regulated in fetal kidney tissues. Similar to WISH-WT, functional annotation revealed that the most abundant gene categories are related to cell cycle and regulation of transcription. IGF2, DLK1, PEG1/MEST, PEG3, PEG5/NNAT, PEG10, and MEIS1, were all significantly elevated, suggesting a role in human nephrogenesis (Supplementary Table S2). The fact that, for instance, HOXA9 was not similarly overexpressed in fetal kidney tissues indicated that it is not intrinsic to normal development but rather to WT progression.

Nevertheless, the microarray experiment did not include temporal expression throughout nephrogenesis, which can better define developmental expression. We approached this by analyzing quantitative gene expression in sequential samples of developing mouse kidneys (E₁₂, E₁₃, E₁₅, E₁₉, 2 weeks, adult kidney). Here again, we could not establish a qRT-PCR analysis for NNAT. The temporal expression patterns of IGF2, DLK1, PEG1, PEG3, and PEG10 during mouse nephrogenesis appear in Fig. 4B and show mostly discrete peaks of expression in early and midgestation and rapid down-regulation in the neonatal (2 weeks) and adult kidneys (PEG3 mRNA is down-regulated only in the adult kidney), suggesting that they are strictly developmentally regulated. Unexpectedly, in contrast to the chip data, MEIS1 was found only at very low levels (average C_T ∼35), and was therefore precluded from this analysis. These results indicated that the imprinted genes are expressed during mouse nephrogenesis in a developmentally regulated manner.

Regulation of WISH-WT overexpressed genes during the differentiation of the rat metanephric mesenchyme. To further suggest a role for the selected genes in normal nephrogenesis, we examined their expression levels in a recently published data set comparing genes in the rat metanephric mesenchyme and tips of ureteric buds (10). For PEG1/MEST and PEG10, there is no known rat sequence to date and both genes could not be located on the microarray. Analysis of PEG3, PEG5/NNAT, DLK1, IGF2, and MEIS1 gene levels showed all to be overexpressed in the metanephric mesenchyme with average expression ratios (metanephric mesenchyme/ureteric bud tip) of 3.1, 2.1, 5.6, 2.75, and 15.1, respectively. We then studied their expression levels during in vitro differentiation of the rat metanephric mesenchyme. As shown in Fig. 5, DLK1, MEIS1, and PEG3 were significantly down-regulated with differentiation of the metanephric blastema into tubular epithelia. Although, similar to DLK1, IGF2 peaked at 24 hours after induction, the fold changes thereafter were insignificant. We could not establish a reproducible qRT-PCR for rat PEG5/NNAT. Thus, most of the genes analyzed are overexpressed in the undifferentiated metanephric mesenchyme and silenced along its maturation.

Induction of WISH-WT overexpressed genes following ischemic kidney injury. Because developmental pathways can be activated during tissue regeneration (28, 29), we sought to determine whether these genes, which are silent in the normal adult kidney, participate in its repair. We therefore analyzed gene expression during mouse kidney regeneration by real-time PCR. IGF2 was excluded from this analysis as its role in kidney repair has been previously suggested (30). Mice were subjected to ischemia/reperfusion renal injury and total RNA was extracted from regenerating kidney tissues at times ranging from 24 hours to 4 weeks after ischemia. As shown in Fig. 4C, whereas levels of PEG1 mRNA did not change significantly over this time course, DLK1, PEG3, PEG10, and MEIS1 transcript levels were all temporally induced with significant elevation observed with time, especially at 2 and 4 weeks following ischemia. Thus, an acute ischemic insult, which triggers kidney regeneration, can induce up-regulation of these metanephric genes in the adult kidney, simulating early embryogenesis.

Microarray experiments, which efficiently survey thousands of genes, have opened new vistas for studying normal and abnormal nephrogenesis (12). However, they must be supplemented and integrated with additional techniques and experimental models to reveal the roles of various genes in these processes. Here, we applied a strategy based on the molecular global analysis of a progressive WT xenograft, human fetal kidneys, and their counterparts—human adult kidney and renal cell carcinoma. To our knowledge, this is the first time that expression profiles of normal and malignant embryonic and adult human kidney tissues have been concomitantly analyzed and catalogued. Given the intimate connection between WT and renal development (31), we hypothesized that novel genes appearing as most significantly overexpressed in the WISH-WT versus renal cell carcinoma microarray data will be important in both renal embryonic tumorigenesis and normal organogenesis. A comparison between WISH-WT and fetal kidney tissues would highlight genes important in WT carcinogenesis, but not necessarily normal nephrogenesis [STAT4 and PRAME, the most significantly induced genes in WISH-WT versus fetal kidney tissues (Supplementary Table S3), were not elevated in the human fetal kidney tissue data set]. Furthermore, because WISH-WT displays blastemal accumulation and dedifferentiation compared with favorable histology WT, which contain cells that have matured into tubular epithelia and stroma, such genes are anticipated to regulate progenitor abundance by suppressing differentiation and maintaining self-renewal of the metanephric blastema. Accordingly, we observed a surprising large number of paternally expressed genes (IGF2, DLK1, PEG1, PEG3, PEG5, and PEG10) in the highest-ranking gene list and, therefore, went on to examine this group. Although several other significantly elevated genes that are not known to be involved in kidney development or WT carcinogenesis could have been also chosen for further analysis (e.g., PTTG or MEOX1), new discoveries in stem cell biology that show a putative role for MEIS1 in self-renewal of stem cells and in programming cancer stem cell character (25, 26) lead to us to concomitantly study this gene. Indeed, these genes were verified to be significantly elevated in progressive WT xenografts compared with FH-WT and were also overexpressed in developing versus mature human kidneys.

To further address the question of biological relevance of these marker genes, we asked whether these mRNAs were differentially expressed in mouse and rat nephrogenesis. We could show that most of the genes (a) are expressed in a developmentally regulated manner during mouse kidney development (IGF2, DLK1, PEG1/MEST, PEG3, and PEG10), (b) are overexpressed in the rat metanephric mesenchyme by comparison to the ureteric bud (IGF2, DLK1, PEG3, PEG5/NNAT, and MEIS1), and (c) are down-regulated with differentiation of the rat metanephric mesenchyme (DLK1, PEG3, and MEIS1), suggesting a role, at least for these genes, in the uninduced metanephric mesenchyme (PEG3 and MEIS1) or early stages of metanephric mesenchyme commitment (DLK1, peaking 24 hours after induction followed by rapid down-regulation). We could further show that these genes (DLK1, PEG3, MEIS1, and PEG10) can be reactivated in adulthood following the introduction of an ischemic insult to the kidney, which triggers a regenerative response. It is noteworthy that during human mesenchymal stem cell differentiation, DLK1 maintains the size of the progenitor cell pool by inhibiting the formation of mature osteoblasts and adipocytes (32). Moreover, expression of DLK1 in hematopoietic cells results in inhibition of differentiation and proliferation (33). In addition, expression of DLK1 has been shown to be critical for glioma cell survival and proliferation (34). PEG3 is a zinc finger gene previously identified in a screen to isolate muscle stem cell regulators that has been implicated in the myogenic and neuronal lineages (35). It has been more recently characterized as a mediator of the p53 signaling pathway and necessary for the p53 apoptotic response (36). PEG10 has been shown to decrease the cell death mediated by SIAH1, a new member of the human Siah family that is induced in response to p53 (37). Thus, these observations suggest a role for these genes in determining cell death versus survival in the metanephric blastema and in maintaining multipotentiality of the cells (nonrenal lineages) along with the nephric lineage genes. From a practical point of view, multipotential progenitor cells of the metanephric blastema (normal and transformed) can be isolated on the basis of surface cell antigen expression. For instance, NCAM (CD56), previously localized to the mouse metanephric condensates (38) and overexpressed in both WISH-WT and fetal kidney tissues, is a potential marker. Similarly, DLK1 is a type I membrane protein and can therefore be useful for the isolation of DLK (+) cells from both progressive WISH-WT and early human fetal kidney tissues. Interestingly, embryonic liver progenitors that give rise to hepatocytes and biliary epithelial cells have been isolated on the basis of DLK expression by magnetic beads (39). Although our goal was to identify novel common pathways that would possibly characterize both normal and transformed blastemal progenitor population, it is clear that both of these cell types also differ. Accordingly, HOXA9, which is significantly elevated in WISH-WT but not fetal kidney tissues, can specifically join MEIS1, which is already induced in the normal blastema to establish a more complete differentiation block in WT blastemal cells, as these transcription factors target independent differentiation pathways (25). Recently, the BMI-1 oncogene–driven gene expression pathway has been shown to be essential for the self-renewal of cells from multiple types of human cancer (40). We found that although BMI-1 was not differentially expressed in either the developing versus mature human kidneys (similar to HOXA9) nor in the embryonic versus adult kidney cancer, there were high expression levels for BMI-1 in both WISH-WT and renal cell carcinoma microarrays (data not shown). Thus, although clearly not specific to WT, BMI-1 can also contribute to the “cooperative differentiation arrest” of WT progenitors.

Genomic imprinting is parent-of-origin–specific allele silencing, or relative silencing of one parental allele compared with the other parental allele (6). It is maintained, in part, by differentially methylated regions within or near imprinted genes and is normally reprogrammed in the germ line (6, 7). Loss of IGF2 imprinting in the chromosome 11p15–imprinted gene cluster has been previously implicated in the evolution of WT (6, 7, 24). Loss of IGF2 imprinting can arise in the germ line or very early in development in Beckwith-Wiedemann syndrome, which causes prenatal overgrowth (including nephromegaly—overgrowth of the whole kidney), birth defects, and predisposition to various embryonal tumors of childhood, including WT. Here, loss of imprinting and a double dose of IGF2 expression affect each and every renal embryonic cell, leading to proliferation and kidney growth, and with the advent of a second genetic hit to multiple WTs (6). In addition, loss of imprinting has been shown to arise sporadically as a somatic mosaic epigenetic alteration in some kidney cells. Premalignant nephrogenic rests containing embryonic renal cells that have been maintained in the postnatal kidney are one such example. Because our results implicate growth-promoting imprinted genes in the expansion of the renal blastemal stem/progenitor population, it is likely that epigenetic changes and loss of imprinting might cause and also maintain WT by altering these and may be other stemness genes by specifically increasing the progenitor cell population. Validation of this claim requires formal studies of loss of imprinting for these genes, as well as patterns of DNA methylation in fetal kidney tissues, adult kidney tissues, WT, and WISH-WT. Nevertheless, aberrant epigenetic gene activation or silencing is apparent, as among the most highly overexpressed tumor progenitor genes in WISH-WT is the one encoding for the polycomb group protein enhancer of zeste homologue 2 (EZH2), very recently shown to directly control DNA methylation (41). In addition, loss of imprinting has been recently shown to be much more common in sporadic WTs, from which WISH-WT was established compared with WTs from syndromic cases (Denys-Drash and WT-aniridia syndromes; ref. 42). Overall, our findings in WT support the recently proposed epigenetic progenitor model of human cancer (7), where nonneoplastic but epigenetically disrupted stem/progenitor cells are suggested as a fundamentally common basis for malignancy. Accordingly, it remains to be determined whether treatment of embryonic kidney progenitor transplants established in immunodeficient mice (43) with the drug combination 5-aza-2′-deoxycytidine, a demethylating agent, and trichostatin A, an inhibitor of histone deacetylases, can result in altered growth and development or more dramatically in a malignant phenotype switch.

Grant support: “Talpiut” Sheba Career Development Award, Moriss Kahn Career Development Award, and the Israel Science Foundation Bat-Sheva De Rothschild Physician-Scientist Grant Award (B. Dekel).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the Kahn Family Foundation for supporting our research.