Perilobar Nephrogenic Rests Are Nonobligate Molecular Genetic Precursor Lesions of Insulin-Like Growth Factor-II-Associated Wilms Tumors

Purpose: Perilobar nephrogenic rests (PLNRs) are abnormally persistent foci of embryonal immature blastema that have been associated with dysregulation at the 11p15 locus by genetic/epigenetic means and are thought to be precursor lesions of Wilms tumor. The precise genomic events are, however, largely unknown.

Experimental Design: We used array comparative genomic hybridization to analyze a series of 50 PLNRs and 25 corresponding Wilms tumors characterized for 11p15 genetic/epigenetic alterations and insulin-like growth factor-II expression.

Results: The genomic profiles of PLNRs could be subdivided into three categories: those with no copy number changes (22 of 50, 44%); those with single, whole chromosome alterations (8 of 50, 16%); and those with multiple gains/losses (20 of 50, 40%). The most frequent aberrations included 1p- (7 of 50, 14%) +18 (6 of 50, 12%), +13 (5 of 50, 10%), and +12 (3 of 50, 6%). For the majority (19 of 25, 76%) of cases, the rest harbored a subset of the copy number changes in the associated Wilms tumor. We identified a temporal order of genomic changes, which occur during the insulin-like growth factor-II/PLNR pathway of Wilms tumorigenesis, with large-scale chromosomal alterations such as 1p-, +12, +13, and +18 regarded as “early” events. In some of the cases (24%), the PLNRs harbored large-scale copy number changes not observed in the concurrent Wilms tumor, including +10p, +14q, and +18.

Conclusions: These data suggest that although the evidence for PLNRs as precursors is compelling, not all lesions must necessarily undergo malignant transformation.

The cell of origin of Wilms tumor (nephroblastoma) is thought to be the metanephric blastemal cell, which forms a pluripotent renal stem cell population during organogenesis and, in response to penetration by the ureteric bud, induces branching morphogenesis and subsequent kidney development (1, 2). In a paradigm of the links between abrogated organ development and tumorigenesis, these pluripotent cells may persist beyond 36 weeks of gestation and appear in up to 1% of routine infant postmortem as nephrogenic rests (3). These lesions are thought to be precursors of Wilms tumors, as they are reported to be found in 30% to 40% of kidneys containing a sporadic tumor and in close to 100% of bilateral cases (3). Their natural history is, however, uncertain: most appear to spontaneously involute, as they are rarely observed in normal kidneys after age 1 year (3).

Nephrogenic rests are histologically classified as either perilobar (PLNR) or intralobar (ILNR), and both types may be further described as dormant, sclerosing, adenomatous, or hyperplastic (3). PLNRs are usually multifocal, are found at the periphery of the renal lobe, are sharply demarcated, and usually consist of blastema and tubules (3). In contrast, ILNRs are often unifocal, are randomly situated within the renal lobe, show poorly defined, irregular, and intermixed margins, and are more commonly composed of stroma (usually predominant), blastema, and tubules (3). Clues to the genetic components of their persistence come from their links to developmental syndromes: PLNRs are associated with overgrowth syndromes including hemihypertrophy and Beckwith-Wiedemann-syndrome, whereas ILNRs are commonly found in WAGR (Wilms tumor-aniridia-genital anomaly-mental retardation) and Denys-Drash syndromes (4).

Direct evidence for the genetic link of rests to Wilms tumor evolution is limited but compelling. Most striking is the identification of heterozygous mutations of the WT1 gene in PLNRs and homozygous mutations in a few hyperplastic ILNRs (5), although this has not been explored further. In an allelic imbalance study, Charles et al. observed a more frequent loss of heterozygosity (LOH) at 11p13 (WT1 locus) in ILNRs, with a stronger association of alterations at 11p15 (WT2 locus) with PLNRs (6). It is notable that the LOH on chromosome 11p preferentially involves the maternal chromosome and results in a uniparental disomy of the paternal allele and biallelic insulin-like growth factor-II (IGF-II) expression (7). These cases with LOH harbored the same abnormalities in the adjacent Wilms tumor, suggesting that aberrations at WT1/WT2 in nephrogenic rests may represent early genetic events in Wilms tumorigenesis.

The occurrence of epigenetic events on chromosome 11p provides the best studied lines of evidence. Loss of imprinting (LOI) at the WT2 locus on 11p15 is found in 33% to 50% of Wilms tumors (4), shows a marked ethnic variation (8), and has been observed in PLNRs (9). Aberrant imprinting control at this locus appears restricted to the genes IGF-II and H19 (10), with hypermethylation of the H19 differentially methylated region on the maternal allele preventing binding of the CTCF chromatin insulator, resulting in IGF-II LOI and biallelic expression. It has been suggested that sporadic Wilms tumor with IGF-II LOI is part of a spectrum of prenatal and postnatal overgrowth disorders associated with epigenetic mosaicism at the IGF-II locus (8) and that PLNRs are a morphologic consequence of the renal IGF-II mosaicism (9).

The Wilms tumors themselves, which are thought to arise from PLNRs or ILNRs, appear to have distinct morphologic features. “ILNR-like” tumors tend to show a spectrum of mesenchymal differentiation with the stromal-predominant histology associated with the underlying genetics (WT1/β-catenin mutation; ref. 11). By contrast, the “PLNR-like” tumors tend to show a limited nephrogenic differentiation with a marked blastemal and, to a lesser extent, epithelial cell predominance (12). These tumors harbor either LOH and subsequent uniparental disomy or LOI of the IGF-II/H19 locus resulting in a marked overexpression of IGF-II in the tumor cells. Coupled with our recent identification of blastemal cell copy number gain and overexpression of the IGF-I receptor in chemoresistant Wilms tumors (13), these tumors may be considered to be driven by IGF signaling in an autocrine/paracrine fashion.

We have investigated the genomic events associated with the earliest stages of the development of these IGF-II-associated Wilms tumors by examining a series of sporadic PLNRs, characterized for genetic and epigenetic aberrations at 11p15, by applied array comparative genomic hybridization (aCGH), and provide the first genomic evidence of the nonobligate precursor role of PLNRs in Wilms tumorigenesis.

Samples and DNA extraction. Archival pathology specimens of PLNRs and corresponding Wilms tumors from the same patients were obtained from the Royal Marsden and Great Ormond Street Hospitals with full ethical committee approval. All cases were reviewed by three pathologists (N.J.S., G.M.V., and J.S.R-F.). In total, 50 formalin-fixed, paraffin-embedded (FFPE) PLNRs were obtained from 42 nonsyndromic patients, with matched Wilms tumors from 25 cases. The vast majority (46 of 50, 92%) exhibited a high degree of WT1 protein expression by immunohistochemistry. There was an overrepresentation of patients with bilateral disease (17 of 50, 34%), as rests are known to be more prevalent in these cases. Lesions were subjected to either laser capture microdissection (PixCell LCM System; Arcturus) or manual microdissection according to the relative size and location of the rests. The vast majority of cases were treated with preoperative chemotherapy, which we have shown previously not to harbor significant genomic differences to those taken at immediate nephrectomy (14). Sections were stained with nuclear fast red to aid in morphologic assessment without interfering with downstream amplification protocols. The DNeasy Tissue Kit (Qiagen) was used for DNA extraction according to the manufacturer's protocol. For the laser-captured samples, up to three subsequent rounds of additional proteinase K digestions at 55°C overnight were carried out. DNA quantitation was determined by spectrophotometry (NanoDrop). We have reported previously the GenomePlex whole-genome amplification technique to be a highly accurate and reproducible method to amplify DNA samples from FFPE material for aCGH analysis using as little as 5 ng starting DNA (15). For the whole-genome amplification of the microscopic, laser-captured nephrogenic rests (up to 3,000 cells), an improved version of the kit (“WGA4”; Sigma) was used according to the manufacturer's protocol.

aCGH. All raw and processed data have been deposited in Array Express⁷

(E-TABM-436). The aCGH platforms used in this study were constructed at the Breakthrough Breast Cancer Research Centre. Hybridizations were carried out as described previously (14) onto a 5.8K, 0.9 Mb-spaced (E-MEXP-213) and/or a 16K, 100 kb-spaced (E-MEXP-1040) BAC array. Slides were scanned using an Axon 4000B scanner (Axon Instruments) and images were analyzed using Genepix Pro 4.1 software (Axon Instruments). The median localized background slide signal for each clone was subtracted and each clone Cy5/Cy3 ratio normalized by local regression (loess) against fluorescence intensity, print-tip subarray, and within-block spatial location. Data were further scaled by dividing each clone by its median absolute deviation. BAC clone replicate spots were averaged, and clones with poor reproducibility between replicates were excluded (SD > 0.2). In addition, clones with missing/poor values in >70% samples were also excluded as were those with no mapping information (March 2006 build of the human genome sequence, hg18). To allow integration of the two platforms, the 16K-derived data were reduced to the resolution of the 5.8K platform by taking the median value of the nearest clones within 1 Mb of the corresponding lower-resolution landmark. In this way, we were left with a final data set comprising 75 samples with 3,342 data points each.

Statistical analysis. All data transformation and statistical analysis were carried out in R 2.0.1⁸

and Bioconductor 1.5,⁹ making extensive use of modified versions of the package aCGH in particular (14). For identification of DNA copy number alterations, data were smoothed using a local polynomial adaptive weights procedure for regression problems with additive errors, with thresholds for assigning “gain” and “loss” derived from the overall median absolute deviation of each sample. For visualization purposes, the processed log₂ ratios were colored green (gain) or red (loss) after segmentation and copy number determination. Thresholded data for each clone were used for categorical analysis using a Fisher's exact test and paired statistics in significance analysis of microarrays, with a correction for multiple testing using the step-down permutation procedure maxT, providing strong control of the family-wise type I error rate.

The copy number data set was converted to wiggle track (WIG) format by a custom Perl script and displayed on the University of California-Santa Cruz genome browser¹⁰

in parallel with the positions of known copy number variations in the HapMap populations (16). Copy number changes in the current data that overlapped with these copy number variations were excluded from further analysis.

Statistical analysis of categorical data between different subgroups was carried out by Fisher's exact test and between continuous variables using the Mann-Whitney U test. P < 0.05 was considered significant. Concordance and similarity scores were calculated as described previously (17). Briefly, the percent concordance equates to the number of changes in common between a pair of samples divided by the number in common plus half the sum of the number of changes in only one or other of the pair. The similarity score was calculated as a weighted sum of all observed alterations based on the frequency of the changes in a larger Wilms tumor population (14). Such similarity scores will be increasingly positive when pairs are more alike.

Methylation analysis of the H19 differentially methylated region. Genomic DNA (300 ng) was bisulfite converted using the EZ DNA Methylation-Gold kit (Zymo Research) according to the manufacturer's instructions. Converted DNA was amplified by PCR using JumpStart REDTaq DNA polymerase (Sigma-Aldrich) for 45 cycles with H19pyroF 5′-GATTTTGATGGGGTTAGGATGT and H19pyroR 3′-CTCCTACTCCAAACATTATAAAA oligonucleotide primers (Eurogentec). Pyrosequencing was carried out using the PyroGold SQA reagent kit (Pyrosequencing AB) according to the manufacturer's instructions with the sequencing primer H19pyroS 5′-TATAAACAAATTCACCTCTC.

Loss of heterozygosity. LOH analysis was done as described previously (18), with the forward primer labeled fluorescently with either tetrachlorofluorescein or hexachlorofluorescein dyes (Invitrogen). Sequences of primers for markers at 11p15 were as follows: D11S4465 forward TAGCCCAAATATCATGTGCA and reverse TGACTGTCACTTAGTAGATGCTCC and TH forward CCTCCCCAGGGCGGCCCGGGGCCCA and reverse CACTTACTTACCCTTGGGGTGGGGG. PLNR, Wilms tumor, and corresponding normal DNA were amplified using touchdown 68-50 program for 35 cycles as follows: denaturing at 94°C for 10 min followed by 95°C for 30 s, 68°C for 45 s, decreasing by 1°C with every cycle followed by incubation at 72°C for 1 min for 17 more times. This was then followed by 94°C for 30 s, 50°C for 45 s, and 72°C 1 min and repeated for 34 more times, with a subsequent final extension at 72°C for 5 min. Each PCR was done under standard conditions in a 15 μL reaction volume containing 250 ng template DNA, 0.8 μmol/L of each primer, 0.2 μmol/L deoxynucleotide triphosphate, 1.5 mmol/L MgCl₂, 0.2 units Taq polymerase (ABGene), and 1× PCR buffer. Denatured reaction products were analyzed on an ABI 7700 capillary sequencer (Applied Biosystems) and visualized with Genotyper software (Applied Biosystems). Allele ratios for PLNR or Wilms tumor compared with normal DNA were calculated as (A1/A2)T/(A1/A2)N and LOH was scored if this ratio was <0.5 and >2.0 in one or both markers.

Quantitative reverse transcription-PCR. RNA extraction for all FFPE samples was carried out by using RNeasy FFPE kit (Qiagen) according to the manufacturer's instructions. cDNA was prepared from 10 μg RNA by random primed reverse transcription using a High-Capacity cDNA Reverse Transcription kit from Applied Biosystems. IGF-II Assay-on-Demand (Hs00171254_m1) was obtained from Applied Biosystems. PCRs were done in a 10 μL reaction volume containing 5 μL of 2× buffer/enzyme mix, 1 μL of 20× assay mix, 1 μL of 20× GAPDH (4326317E), or HPRT1 (4326321E) endogenous control assay mix and 1 μL input cDNA. Assays were run on an Applied Biosystems 7900 Sequence Detection System and results were analyzed by the standard curve method. Data were normalized to Universal Human Reference RNA (Stratagene).

Immunohistochemistry. Immunohistochemistry was done on 5 μm FFPE sections using a mouse monoclonal antibody to either human Ki-67 clone MIB-1 (M7240; DAKO), WT1 clone 6FH2 (M3561 DAKO), or IGF-II (ab9574; Abcam) using the Envision horseradish peroxidase system (K4006 DAKO) at a dilution of 1:150 for Ki-67/WT1 and 1:200 for IGF-II according to the manufacturer's instructions. Antigen retrieval for WT1 was carried out by pepsin digestion in 0.2 mol/L HCl at 37°C in a humidified chamber; the slides were boiled in 10 mmol/L Tris/NaOH, 1 mmol/L EDTA (pH 9) in a pressure cooker for 2 min for Ki-67 and in 10 mmol/L citric acid (pH 6) in the microwave for 9 min for IGF-II. Ki-67 proliferation index was determined by manually counting 1,000 cells per slide, of which the positively stained cells determined the percentage of proliferation. Samples with 5% to 20% positive cells were classified as actively proliferating, with cases >20% positive regarded as highly proliferative. IGF-II expression was classified as negative, weakly positive, or strongly positive relative to normal kidney.

PLNRs overexpress IGF-II due to genetic and epigenetic mechanisms. We examined genetic and epigenetic alterations at the 11p15 locus in our series of PLNRs by means of LOH and methylation analysis. We identified LOH in 10 of 39 (26%) informative cases (Fig. 1A). In only one case (PLNR06) was LOH observed in the rest and not the tumor. H19 hypermethylation by pyrosequencing was present in 37 of 40 (93%) assessable PLNRs. IGF-II LOI, identified by means of H19 hypermethylation in the absence of allelic imbalance, was observed in 23 of 33 (70%) cases where data from both assays were available (Fig. 1B). We investigated IGF-II expression by quantitative reverse transcription-PCR (RT-PCR) and immunohistochemistry and observed in all but 2 cases (40 of 42, 95%) high levels of the transcript (21 of 34) and/or the protein (30 of 42; Table 1).

PLNRs can harbor large-scale changes in DNA copy number. To determine whether additional genomic events were associated with the earliest stages of IGF-driven Wilms tumorigenesis, we carried out aCGH on our series of 50 PLNRs (Table 1). The genomic profiles of PLNRs could be subdivided into three categories: those with no copy number changes; those with single, whole chromosome alterations; and those with multiple gains and losses. In 22 of 50 (44%) of our cases, there were no alterations in DNA copy number detectable by our platform. These include relatively large, hyperplastic PLNRs with WT1 protein expression (Fig. 2A). It is of note that, in an earlier aCGH study of Wilms tumors, we similarly observed 24% of cases to harbor no such copy number changes (14).

Eight of 50 (16%) PLNRs exhibited only a single chromosomal alteration. This was loss of 1p and 11p in two cases each and single examples of -11q, -12q24, -21q, and +X (Fig. 2B). The remaining rests harbored multiple chromosomal aberrations (Fig. 2C). Taken together, the most frequent aberrations included loss of 1p (7 of 50, 14%) and gain of chromosomes 18 (6 of 50, 12%), 13 (5 of 50, 10%), 10p15 (4 of 50, 8%), and 12 (3 of 50, 6%; Fig. 3). PLNRs with adenomatous features harbored more copy number changes than those without (3.3 versus 1.0; P < 0.001, Fisher's exact test). There were no significant differences in rests from bilateral versus unilateral patients.

There were no statistically significant associations of gross copy number changes in PLNRs with LOH versus LOI at 11p15. As allelic imbalances and copy number losses of 11q and 16q have been reported to associate with IGF-II LOI in Wilms tumors (19), we further analyzed LOH at these loci in our PLNRs. 11q LOH was observed in three cases (PLNR11, PLNR15, and PLNR22), all of which had LOI in our series. LOH at 16q was observed in only a single, LOI-positive rest (PLNR22), intriguingly in a case where the concurrent Wilms tumor was of anaplastic histology.

Evidence for a nonobligate malignant progression from PLNR to Wilms tumor. For a subset of our samples (n = 25), we were able to directly compare the aCGH profiles of the PLNRs and their concurrent Wilms tumors. For the majority (19 of 25, 76%) of cases, the Wilms tumors harbored the same copy number changes as those observed in the associated rest (Table 1). In three cases, the profiles were identical (Fig. 4A); however, for the most part, the tumors contained further changes in addition to those observed in the PLNR (Fig. 4B). There were quantitatively more alterations in Wilms tumors than rests (median, 4 versus 1; P < 0.001, Mann-Whitney U test), even when those cases with no copy number changes were excluded (median, 4 versus 3; P = 0.011, Mann-Whitney U test).

To probe more rigorously the differences and similarities between PLNRs and Wilms tumors, we took several statistical approaches. Firstly, we applied class comparison techniques between our complete series of 50 PLNRs and 25 Wilms tumors (all from FFPE) and our larger published series of 76 frozen Wilms tumor specimens (14) to look for widespread differences between the lesions. A Fisher's exact test, corrected for multiple testing, revealed few statistically significant differences between rests and Wilms tumors, reflecting the common genetic alterations present between the precursor and the cancer. There were, however, two significantly different loci, gain of 1q and loss of 16q, which were present in Wilms tumors but not observed in PLNRs in our series (Fig. 4C).

In addition, we were able to apply paired statistics for our 25 patient-matched cases to reduce between-subject variability and to track the molecular evolution of specific tumors. Using the paired significance analysis of microarrays algorithm on our segmented data again highlighted +1q and -16q as significant alterations “acquired” in the development from PLNR to Wilms tumor. This analysis showed that gains of chromosome 6 and losses at 10p and 14q were significantly associated with the later stages of Wilms tumorigenesis, as these were also not seen in rests (Fig. 4D).

Furthermore, we calculated both concordance and similarity scores for our PLNR/Wilms tumor pairs as described by Waldman et al. (17). For the majority of rests with detectable genomic alterations (9 of 15, 60%), the similarity score between a given rest and its associated Wilms tumor was considerably higher than for the remaining tumors (Fig. 5, blue filled circles), and the concordance between pairs was high (>50%). In two cases, there was only a single higher score for the PLNR in the series of unmatched tumors. Cases of PLNR with no copy number changes inevitably had a low score when compared with the tumors (Fig. 5, gray filled circles).

Although these data are consistent with the notion that PLNRs are the molecular genetic precursor of the associated Wilms tumors, there were a few samples in which this model did not fit. In six cases, PLNRs harbored large-scale copy number changes not observed in the concurrent Wilms tumor (Table 1). These included the samples that contained the lowest similarity scores between the rest and matched tumor (Fig. 5). There were three recurrent aberrations noted in these rests: gain of 18 (commonly seen in Wilms tumor) and 10p and 14q (rarely reported in Wilms tumors). All of these PLNRs had a low proliferative index by Ki-67 immunohistochemistry. It is apparent from these data, and from five cases consisting of 13 multifocal lesions with differing genomic profiles (Table 1), that although the evidence for PLNRs as precursors is compelling, not all lesions must necessarily undergo malignant transformation.

Actively proliferating PLNRs represent a more advanced molecular genetic precursor. We investigated the genomic profiles of our entire series of PLNRs based on the proliferation index assessed by Ki-67 staining. Sixteen of 50 (32%) PLNRs were classified as actively proliferating (4 of 50, 8% were highly proliferative) in contrast with 24 of 25 (96%) Wilms tumors (14 of 25, 56% highly proliferative; Table 1). There were a greater number of copy number changes in proliferating versus nonproliferating PLNRs (median, 2 versus 0.5; P = 0.024, Mann-Whitney U test). In particular, the qualitative differences between these high and low proliferation index rests was such that the aCGH profiles of highly proliferating PLNRs more closely resembled their associated Wilms tumor than did those with no Ki-67 immunoreactivity [mean Spearman rank correlation coefficient, 0.583 (range, 0.489-0.641) versus 0.214 (range, 0.029-0.430); P < 0.001, Mann-Whitney U test]. In addition, actively proliferating PLNRs/wild-types had a significantly higher similarity score (median, 4.86 versus -6.015; P = 0.023, Mann-Whitney U test). Indeed, in some instances, the distinction cannot be made based on the genetic, epigenetic, or genomic data between actively proliferating, morphologically hyperplastic blastemal cells as part of a PLNR or an adjacent Wilms tumor (Fig. 4A).

Here, we present the first genome-wide profiling of copy number changes in PLNRs and associated Wilms tumors. The proposed precursor status of both PLNRs and ILNRs has been largely defined by epidemiologic (20, 21), morphologic (3, 22, 23), but only recently by molecular (6) evidence. Here, we show clear clonal progression based on copy number profiles of PLNRs and Wilms tumors.

The route by which Wilms tumors arise from ILNRs is believed to be accompanied by mutations and deletions of WT1 at 11p13 at an early developmental stage (5), followed by mutation and nuclear localization of β-catenin during the progression to Wilms tumor (24), with Wnt pathway activation and downstream transcription factor activation. This sequence of events is strongly associated with a stromal histology in the subsequent Wilms tumor (11).

By contrast, PLNRs and the Wilms tumors accompanying or arising from them have been associated with dysregulation at the 11p15 locus by genetic and epigenetic means, leading to an overexpression of IGF-II (12). The molecular events accompanying this IGF-II-driven pathway have not been elucidated, although it has been suggested previously that Wilms tumors with wild-type WT1 (presumably largely developed via the PLNR pathway) harbored more genetic alterations than those with WT1 mutations (ILNR pathway; ref. 25).

We have determined the temporal order of changes in DNA copy number, which occur during the IGF-II/PLNR pathway of Wilms tumorigenesis. Large-scale chromosomal alterations such as gain of 12, 13, and 18, and loss of 1p were frequently observed in PLNRs and can be regarded as “early” events. This is in contrast to other common Wilms tumor-associated alterations, gain of 1q and 6 and loss of 10p and 16q, which were not observed in our PLNRs and are therefore “later” events in the proposed multistep model of Wilms tumor development.

It is of interest to note the disparity in timing between the copy number changes most convincingly associated with treatment failure and tumor relapse. Although loss of 1p was frequently observed in PLNRs, gain of 1q and loss of 16q were restricted to Wilms tumors. These three events have shown concordance in previous Wilms tumor profiling experiments, suggesting a common mechanism (14). The data provided here suggest a sequence in which the copy number loss of 1p occurs before a coordinated +1q/-16q at least via the IGF-II/PLNR pathway. The implication of this is that we would therefore expect the formation of an isochromosome 1q to also be associated with the WT1/ILNR pathway, although this remains to be determined.

Genetic loss at 16q has been associated previously with LOI at 11p15 (19, 26), with the gene encoding CTCF, a regulator of imprinting at the H19/IGF-II locus, found on 16q22. Our data add weight to the evidence (6, 27) that, despite this, 16q abnormalities are later events and preceded by genetic and epigenetic up-regulation of IGF-II. In the single PLNR in which we observed not a copy number change, but allelic loss, of 16q, the associated Wilms tumor was of anaplastic histology, a subtype strongly associated with mutations in TP53 (28). We were not able to determine the timing of the event in this case. Other correlates of IGF-II LOI in Wilms tumors have been reported as 11q loss and trisomy 12 (19). Although we observed copy number gain of chromosome 12 and both LOH and loss at 11q in our rests, suggesting that these are early events with the IGF-II/PLNR pathway, neither correlated directly with LOI.

In around half of the PLNRs we profiled, no large-scale changes in DNA copy number were observed. This is around twice the proportion of Wilms tumors that gave similarly copy neutral aCGH profiles (14) or normal karyotypes (29). Almost all of these cases, however, exhibited high levels of IGF-II expression due to either LOI or LOH (and assumed uniparental disomy). Although this suggests that, in some cases, the genetic/epigenetic targeting of IGF-II is sufficient in its own right to drive the tumorigenic process through to blastemal cell persistence, which additional abnormalities may be required for tumor progression are yet to be identified. Alterations at the genetic, epigenetic, transcript, small RNA, and/or post-translational modification levels may all play a role. In any case, the observation that those rests that were actively proliferating harbored more genomic alterations, and more closely resembled their associated Wilms tumors, suggests that, in a proportion of cases at least, there are additional alterations driving clonal expansion and selection for the malignant phenotype (30).

In a small number of cases, the genomic profiles of PLNRs did not match with either the associated Wilms tumor or a topographically distinct rest. Given the often multifocal nature of PLNRs, this is perhaps not surprising, although it has not been shown before. This evidence seems to suggest that PLNRs are nonobligate precursors, in that not all lesions necessarily develop into Wilms tumors, a fact clear from the seminal early studies by Beckwith, who reported that ∼1 in 100 kidneys at birth harbor nephrogenic rests, whereas the incidence of Wilms tumor is only 1 in 10,000 (3, 20). We provide a molecular underpinning of this observation in patients who did develop Wilms tumor, showing that certain lesions, often containing +10p and/or +14, may be genomic dead-ends, as their associated tumor, and Wilms tumors in general, did not harbor these alterations. It is possible that genomic gains at these loci may even actively promote regression or involution of the rest, an intriguing thought given that they are otherwise commonly lost in Wilms tumors themselves. By contrast, the frequent occurrence of +18 in these and other PLNRs and Wilms tumors suggest it to be in some instances a noncritical mutation in sporadic IGF-II/PLNR Wilms tumorigenesis.

Although Wilms tumors have been classified as “PLNR-like” and “ILNR-like” based on their associated precursor lesion and appear to evolve by different genetic and epigenetic pathways, the two routes may incorporate significant cross-talk. The novel X-chromosome tumor suppressor gene, WTX, acts as a negative regulator of β-catenin in the cytoplasm, and WTX mutations may also drive nuclear localization of β-catenin and subsequent up-regulation of its transcriptional targets (31). Unfortunately, we have not been able to address whether WTX deletion occurs early in development and is thus present in the PLNRs, as our array platform does not contain a suitable BAC clone at the Xq11 locus. Additionally, WT1 itself may act as a transcriptional repressor of both IGF-II (32) and its receptor, IGF-I receptor (33), which may in turn mediate the nuclear translocation of β-catenin (34). The IGF-II/IGF-I receptor axis has recently been identified as playing a direct and central role in the self-renewal of embryonic stem cells (35). Immature blastemal cells, stem cells of the developing kidney, are more likely to predominate in the IGF-II/PLNR Wilms tumors (12) and IGF-I receptor may also be up-regulated in Wilms tumor blastema by genomic means (13). Deconvoluting the critical initiating steps and functional endpoints of these two developmental pathways is a key challenge for Wilms tumor biology.

No potential conflicts of interest were disclosed.

We thank the Children's Cancer and Leukaemia Group Tumour Bank, which is funded by Cancer Research UK; the contributing pathologists, oncologists, and surgeons for access to samples; and Boo Messahel (Institute of Cancer Research), Lorna Tinworth (St George's Hospital Medical School), and Michelle Lazenby and Mandy Gilkes (Department of Haematology, Cardiff University) for technical assistance.