Embryonal precursors of Wilms tumor

See allHide authors and affiliations

Science  06 Dec 2019:
Vol. 366, Issue 6470, pp. 1247-1251
DOI: 10.1126/science.aax1323

A childhood tumor—from the beginning

Many adult cancers arise from clonal expansions of mutant cells in normal tissue. These premalignant expansions are defined by somatic mutations shared by the cancers. Whether pediatric cancers originate in a similar way is unknown. Coorens et al. studied Wilms tumor, a childhood kidney cancer. Phylogenetic analyses revealed large clones of mutant cells in histologically and functionally normal kidney tissue long before tumor development. Thus, like adult tumors, Wilms tumor appears to arise from a premalignant tissue bed.

Science, this issue p. 1247


Adult cancers often arise from premalignant clonal expansions. Whether the same is true of childhood tumors has been unclear. To investigate whether Wilms tumor (nephroblastoma; a childhood kidney cancer) develops from a premalignant background, we examined the phylogenetic relationship between tumors and corresponding normal tissues. In 14 of 23 cases studied (61%), we found premalignant clonal expansions in morphologically normal kidney tissues that preceded tumor development. These clonal expansions were defined by somatic mutations shared between tumor and normal tissues but absent from blood cells. We also found hypermethylation of the H19 locus, a known driver of Wilms tumor development, in 58% of the expansions. Phylogenetic analyses of bilateral tumors indicated that clonal expansions can evolve before the divergence of left and right kidney primordia. These findings reveal embryonal precursors from which unilateral and multifocal cancers develop.

Adult cancers typically arise as a consequence of aging and mutagen exposure, at times through the generation of precancerous clonal expansions. Examples of these precancerous states include Barrett’s esophagus, clonal hematopoiesis, and colonic polyps. It is unknown whether childhood tumors, which are thought to result from aberrant fetal development, likewise arise from precancerous clonal expansions (Fig. 1A). To address this question, we studied Wilms tumor (nephroblastoma), the most common kidney cancer of childhood. Wilms tumor is a prototypical embryonal malignancy of infants and young children (1). It arises from abnormal fetal nephrogenesis, which it resembles morphologically (1) and transcriptionally (2). It occurs sporadically or in the context of bilateral tumors, multifocal lesions, urogenital developmental disorders, or overgrowth syndromes (1).

Fig. 1 Precursor clonal expansions in normal human kidneys.

(A) Wilms tumor can arise during embryogenesis on the background of an otherwise normal kidney (the traditional view) or from precursor lesions residing in normal tissues, as found here. (B) Overview of tissue sampling in the kidney of patient PD37272. (C) Somatic mutations can be timed using their VAF across the corresponding normal tissues, which is higher for earlier mutations. If the mutation is present in tumor, kidney, and blood, it is an early embryonic mutation (mutations 1 and 2). If it is present in kidney samples and tumor only, it is clonal nephrogenic (mutations 3 to 5, marked by asterisk). If it is only in the tumor, it is labeled as such. White and black circles indicate whether the observed VAF is insignificant (white) or significant (black), P < 0.001 (test of presence using beta-binomial overdispersion). (D) The VAF for the last embryonic mutation in kidney samples and tumor compared with blood.

To identify potential precursors of Wilms tumor, we used somatic mutations to infer the phylogenetic relationship between cancers and corresponding normal tissues (kidney and blood). We analyzed 229 whole-genome sequences of 54 individuals: 23 children with Wilms tumor, 16 parents of affected children, three children with other types of kidney cancer (congenital mesoblastic nephroma, malignant rhabdoid tumor), 10 adults with clear cell renal cell carcinoma (ccRCC), and two adults without renal tumors (one kidney transplant patient and one kidney obtained at autopsy; table S1). We called base substitutions against the reference human genome and extracted mosaic mutations from each set of donor-related tissues. We validated the method for calling mosaic mutations by sequencing parental germline DNA, resequencing tissues, and inspecting raw data (fig. S1). On the basis of the variant allele frequency (VAF) and distribution of mutations across related tissues, we built phylogenetic trees of tumor development. We supplemented DNA data with analyses of RNA sequences and genome-wide methylation patterns (table S1).

Our discovery cohort consisted of three children with unilateral Wilms tumor. We sampled tumors, blood, and histologically normal kidney tissues from the same individuals (table S1 and fig. S2). As expected, whole-genome sequences revealed mosaic mutations attributable to the first cell divisions of the fertilized egg (fig. S3 and table S2) (3, 4). In two cases, we also detected mosaic mutations in normal kidneys that were present in the corresponding cancer but absent from blood (Fig. 1, B to D, and table S2), indicating that the tumors had arisen from that particular normal kidney tissue.

Several features of these mutations showed that they defined clonal expansions in normal kidney tissue, as illustrated by patient PD37272 (Fig. 1, B to D). The VAFs of mutations in the normal tissue of this kidney, variants 3 to 5 (Fig. 1C), were as high as 44%, which suggested that the mutation was present in 88% of all cells in the biopsy. Mutations 3 to 5 were present in the two parenchymal biopsies (i.e., cortex and medulla) but were absent from blood cell DNA, deeply sequenced to 106× genome-wide (fig. S4). Similarly, mutations 3 to 5 were undetectable in renal pelvis, which is embryologically derived from a different lineage than kidney parenchyma (5). Furthermore, the VAF of early embryonic mutations, variants 1 and 2, was inflated in parenchyma and in tumors (Fig. 1D). Such inflation of early embryonic mutations is a feature of tissues that contain a clonal expansion of a single cell (fig. S3). By contrast, in tissues devoid of a major clone, such as renal pelvis and blood, the VAFs of early embryonic mutations were not inflated (Fig. 1D). Thus, these normal tissue variants 3 to 5 demonstrate the presence of clonal expansions within kidney parenchyma, which we termed clonal nephrogenesis, accounting for up to 88% of cells sampled in the cortex.

To further study and validate our discovery of clonal nephrogenesis as an antecedent of Wilms tumor, we studied another 20 cases: 15 unilateral tumors with normal tissue biopsies curated through a British childhood renal tumor study (IMPORT), four cases of bilateral Wilms tumor, and one tumor with 10 normal tissue biopsies (table S1 and fig. S5). Within the entire group of 23 children (discovery and validation cohorts), we found evidence of clonal nephrogenesis in 10 of 19 children with unilateral disease (53%) and in all four children with bilateral cancers (Fig. 2, A and B). The presence of clonal nephrogenesis was further substantiated by the significant (P < 0.01, Wilcoxon signed-rank test) inflation of VAFs of early embryonic variants (fig. S6). There were no copy number changes detected in normal tissues by three different methods.

Fig. 2 Clonal nephrogenesis and H19 hypermethylation.

(A) Sizes of nephrogenic clones, as predicted by twice the VAF of the most prominent nephrogenic mutation, alongside corresponding methylation values of the H19 locus. Green symbols indicate significant deviation (P < 0.05, Wilcoxon rank-sum test) from the background methylation distribution (gray shaded area) as obtained from normal kidney samples without clonal nephrogenesis. N/A, no methylation data available. (B) Mutations present in samples obtained from normal kidneys but absent in matched blood. Only in Wilms tumor were some of these mutations shared with the corresponding tumor. In the presence of clonal nephrogenesis, the VAF distribution of these mutations was significantly elevated (***P < 0.001, Wilcoxon rank-sum test). (C) Histology images showing components (arrowheads) of the human nephron excised by laser capture microscopy. Scale bar, 250 μm. (D) VAF simulations to derive expected distributions depending on clonality of a tissue: monoclonal origin (~Bin(n~Pois(cov), P = 0.5)), oligoclonal origin (~Bin(n~Pois(cov), P = 0.3)), or polyclonal origin (~Bin(n~Pois(cov), P = 0.1)), where cov = coverage = 40×. (E) VAF distributions for 22 microdissected samples (10 proximal tubules, five distal tubules, seven glomeruli) from three patients, one rapid autopsy donor, and two ccRCC patients. Color indicates the underlying maximum likelihood VAF as predicted by a truncated binomial mixture model. (F) Group-level methylation beta values of H19 (*P < 0.05, Wilcoxon rank-sum test). (G) Relationship between predicted clone sizes from nephrogenic mutation (see Fig. 2A) and the methylation level of H19. The black dot represents patient PD40738g, who is affected by germline H19 hypermethylation (omitted from correlation and linear regression).

Conceivably these findings could be due to tumor infiltration into normal tissue not visible histologically (fig. S2) or to cross-contamination of DNAs. This explanation is implausible, as contamination would manifest as shared variants at a low VAF, rather than select mutations at a high VAF. We statistically excluded the possible contribution of tumor infiltration and contamination by using a binomial mixture model on the observed base counts of tumor mutations in the normal samples (fig. S7).

Next, we investigated whether clonal nephrogenesis represents the normal clonal architecture of human nephrons by three approaches. First, using laser capture microscopy (LCM), we excised glomeruli (n = 7) and proximal and distal tubules (n = 15) from the kidneys of three individuals (Fig. 2, C and D). We subjected these LCM cuts to whole-genome sequencing using an established method for generating low-input DNA libraries (6, 7). Kidney tissues were obtained at autopsy (one case) or from normal portions of kidneys affected elsewhere by ccRCC (two cases). Our analysis of somatic mutations across these LCM cuts revealed a VAF distribution (Fig. 2E) that is inconsistent with the monoclonal organization seen, for example, in endometrial glands or colonic crypts (6, 7). Second, we assessed whether mutations were commonly shared between renal tumors and surrounding normal kidney tissue. We studied childhood congenital mesoblastic nephroma (two tumors and six normal kidney samples), childhood malignant rhabdoid tumor (one cancer and one normal kidney sample), and adult ccRCC (eight cancers, including one bilateral case, and 15 normal tissues). Applying the same analysis pipeline, we identified early embryonic mutations shared between tumor, normal kidney tissues, and blood (fig. S8). However, we did not find mutations shared only between tumor and normal tissue (Fig. 2B); this finding shows that such mutations were specific to, and enriched in, Wilms tumor (P < 0.001, Fisher exact test). Of particular relevance were renal cell carcinomas, which, like Wilms tumor, are derived from nephrons. If normal embryological clonal dynamics typically had generated clonal expansions, we would have expected to find clonal nephrogenesis in ccRCC cases. Third, we examined all developmental mutations of normal kidney tissues listed thus far, supplemented by an additional 18 biopsies obtained from bilateral kidneys that had been declined for transplantation (Fig. 2B). We analyzed somatic mutations present in kidney tissue and absent from non-renal tissue, irrespective of whether they were shared with tumors. Collectively, these analyses of 77 normal kidney biopsies revealed that variants of tissues without clonal nephrogenesis have a significantly lower VAF distribution than clonal nephrogenesis mutations (P < 0.001, Wilcoxon rank-sum test; Fig. 2B). Taken together, these results indicate that clonal nephrogenesis represents aberrant kidney development.

A central question raised by our findings is whether cancer-causing (driver) events underpin clonal nephrogenesis. Notably, the clonal nephrogenesis mutations (n = 66) (table S2 and fig. S8) were noncoding (64 of 66) or did not generate plausible oncogenic events. We searched further for driver events among germline and somatic DNA mutations, in transcriptomes, and, where available, in methylation patterns. We found hypermethylation of the H19 locus in 7 of 12 (58%) normal kidney tissues with clonal nephrogenesis. In contrast, we did not detect it in normal kidney tissues without clonal nephrogenesis or in blood (Fig. 2, A and F), except for the blood of a child (PD40738) with Beckwith-Wiedemann syndrome (fig. S9). H19 hypermethylation is an established driver event in the pathogenesis of Wilms tumor and is thought to operate by disrupting the epigenetic regulation of growth-promoting genes that reside on chromosome 11p15 (810). The degree of hypermethylation of H19 correlated with the VAF of clonal nephrogenesis, indicating that hypermethylation pervaded the entire clone (Fig. 2G). In the five wild-type samples, hypermethylation of H19 may therefore have evaded detection because of a small clone size (Fig. 2A). We also studied the KvDMR1 locus on chromosome 11p15.5 and determined that its methylation state was unchanged in clonal nephrogenesis. Hypomethylation of KvDMR1 underlies overgrowth syndromes that confer only a minimal predisposition to Wilms tumor (11). We did not find any further driver events accounting for clonal nephrogenesis, despite re-interrogation by exome sequencing of 15 of 17 tissues with clonal nephrogenesis. Gene expression profiles, including fetal transcripts, did not differ between normal renal tissues that displayed or did not display clonal nephrogenesis (fig. S9). Similarly, global methylation patterns did not differ between these two groups (fig. S9).

The timing of the emergence of clonal nephrogenesis during development could be defined in three children from whom we obtained bilateral tumors. In patients PD40735 and PD36159, left and right tumors were derived from the same trunk of clonal nephrogenesis (Fig. 3, A to D). In the third patient, PD40378, all five left tumors, but not the right neoplasms, were related to clonal nephrogenesis on the left (Fig. 3E). These findings indicate that in two children with bilateral cancers, clonal nephrogenesis must have arisen before left and right kidney primordia diverged, early in embryogenesis (5). For unilateral tumors, the timing of the occurrence of clonal nephrogenesis remains unclear. It may have evolved before the kidney was formed or thereafter, followed by a “clonal sweep” of clonal nephrogenesis replacing kidney tissue.

Fig. 3 Phylogenies of bilateral and multifocal Wilms tumor.

(A, C, E, and G) For each tumor, the phylogeny of shared mutations is shown including de novo germline mutations, embryonic mutations, mutations demarcating clonal nephrogenesis, and tumor mutations. Numbers refer to the number of substitutions defining each developmental trunk. Truncal driver events are detailed. (B) Heat map showing the contribution of a mutation to the sample PD40735 shown in (A). The pattern of shared mutations reveals a split between left and right kidney, which is obeyed by both tumor and normal samples. (D) As revealed by the pattern of shared mutations, in patient PD36159 the left tumor is more closely related to the right branch of clonal nephrogenesis than to the left branch. (F) Two mutations indicate the independent emergence of tumors at different time points from the nephrogenic clone in patient PD40641. As shown in (G), tumor and nephrogenic rest in patient PD36165 both originated from clonal nephrogenesis despite being situated at opposing kidney poles.

In five cases, we sampled multiple neoplasms of the same kidney, which revealed two configurations of tumor development (Fig. 3). Tumors either were derived from a shared trunk that had emerged from clonal nephrogenesis (Fig. 3, A to E) or arose independently and successively from clonal nephrogenesis, which suggests a sustained potential for the latter to spawn tumors (Fig. 3, F and G). For example, patient PD36165 presented with two lesions at opposing poles of the kidney: a nephrogenic rest and a Wilms tumor. Cancer and the nephrogenic rest had emerged from the same ancestral clone at different time points, followed by clonal diversification within each (Fig. 3G).

Finally, we compared the genetic alterations in Wilms tumor with and without clonal nephrogenesis. This revealed a near–mutual exclusivity of loss of heterozygosity (LOH) of chromosome 11p15 and clonal nephrogenesis. In tumors with clonal nephrogenesis, LOH of 11p15 was mostly absent (P = 0.009, Fisher exact test) (fig. S10). This suggests that there may be two distinct pathways for the pathogenesis of Wilms tumor that both involve dysregulation of 11p15 genes as a driver (Fig. 1A). Accordingly, cancers may arise either directly in isolation through LOH of 11p15, or indirectly via clonal nephrogenesis with perturbation of 11p15 by hypermethylation of H19.

Our findings show that clonal expansions in histologically normal tissue are an atypical outcome of renal development that commonly antedates Wilms tumor. We demonstrated a direct phylogenetic link between clonal expansions, H19 hypermethylation, and the formation of cancer, thus identifying clonal nephrogenesis as an epigenetic progenitor of cancer comprising “neoplasia-ready cells” (12). In contrast to precursors of adult cancer, clonal nephrogenesis generated histologically and functionally normal tissues, which in the most pronounced cases occupied the bulk of renal tissues. If future work shows that the extent of clonal nephrogenesis is a marker of malignant potential and risk of cancer recurrence, this information potentially could be used to guide treatment and surveillance of patients with Wilms tumor. Moreover, if we were able to manipulate the neoplastic potential of clonal nephrogenesis, prevention of Wilms tumor could become feasible. Collectively, our findings portray Wilms tumor as an insurrection on the background of a premalignant tissue bed, rather than a clearly demarcated neoplasm in an otherwise normal polyclonal kidney. We speculate that embryonal clonal expansions, perhaps also driven by epigenetic mechanisms, may be a common phenomenon in childhood cancer.

Supplementary Materials

Materials and Methods

Figs. S1 to S12

Tables S1 to S8

References (1330)

References and Notes

Acknowledgments: We thank the research nursing staff at Cambridge University Hospitals, the Royal Hospital for Sick Children (Edinburgh), and the Royal Hospital for Children (Glasgow) as well as all IMPORT investigators. We thank M. Gerstung and A. J. Lawson for critical review of the manuscript. We are indebted to our little and older patients and their families for participating in our research. Funding: This project was principally funded by the Little Princess Trust, the St. Baldrick’s Foundation (Robert J. Arceci International Award to S.B.), and Wellcome (fellowship to S.B.; Sanger core funding). Additional funding was received from CRUK (IMPORT study; fellowship to T.J.M.; Cambridge Centre), NIHR (Biomedical Research Centre Great Ormond Street; Cambridge Human Research Tissue Bank; Oxford Biomedical Research Centre; fellowship to T.R.W.O.), the Royal College of Surgeons of England (fellowship to T.J.M.), Wellcome (fellowship to T.H.H.C. and K.S.), Great Ormond Street Hospital Children’s Charity (R.A.-S., K.P.-J.), and Li Ka Shing foundation (D.C.W.). J.D. acknowledges funding from the Alpe d’HuZes foundation/KWF Dutch Cancer Society Bas Mulder Award (#10218) and the Oncode Institute. Author contributions: S.B. conceived of the experiment. T.H.H.C., T.D.T., and S.B. analyzed data. Statistical expertise was provided by M.D.Y., D.C.W., and I.Ma. M.D.Y., T.J.M., G.C., P.S.T., P.J.C., M.R.S., I.Ma., and K.P.-J. contributed to discussion. L.M. and M.G.B.T. performed laser capture microscopy experiments. Samples were curated and/or experiments were performed by R.A.-S., C.T., S.T., M.O., Y.H., M.B., B.C.R., G.B., J.A., M.J., G.A.A.B., J.V., J.C.N., N.S., K.S.-P., G.D.S., K.S., T.C., I.Mu., J.D., and K.P.-J. Pathological expertise was provided by L.H., T.R.W.O., D.R., A.Y.W., N.C., and N.S. A.C. created kidney illustrations. T.H.H.C., T.D.T., and S.B. wrote the manuscript. S.B. directed the study. Competing interests: G.D.S. is a paid consultant for Pfizer, Merck, EUSA Pharma, and Cambridge Medical Robotics. All other authors declare no competing interests. Data and materials availability: Raw sequencing data have been deposited in the European Genome-phenome Archive (EGA) under study ID EGAD00001004774. The bespoke code used for the analyses is available on GitHub (

Stay Connected to Science


Navigate This Article