High resolution clonal architecture of hypomutated Wilms tumours

Abstract

A paradigm of childhood cancers is that they have a low mutation burden, with some ostensibly bearing fewer mutations than the normal tissues from which they derive. We set out to resolve this paradox by examining paediatric renal cancers with exceptionally few mutations using high resolution, high depth sequencing approaches. We find that apparent hypomutation is the result of unusual clonal architecture due to a normal tissue-like mode of tumour evolution, raising the possibility that the mutation burden of some cancers has been systematically misjudged.

Fig. 1. Mutation burden of Wilms tumours.

a The total number of substitutions called from whole genome sequencing (WGS) of Wilms tumours from two previously published cohorts^, is shown (log scale). Where there is more than one sample per tumour, the median burden across samples is displayed. A horizontal solid red line represents the mutation burden of normal neonatal kidney established by duplex sequencing (nanoseq) in this study, (mean of one sample from each of three children), while the dashed red line shows that for normal kidneys of school-age children with Wilms tumours (mean of one sample from each of two children). b For a cohort of 21 adult renal cell carcinomas, the substitution burden derived from nanoseq is plotted against that derived by bulk WGS. c–h Comparison of substitution burden per diploid genome (log scale) of bulk whole genome sequencing and nanoseq for each patient’s tumour. The bulk estimate has been corrected to 30× coverage, except for PD37585 (f) for which this was not possible (Methods, Supplementary Fig. 10). Each point represents one biopsy from a tumour, and, where the same biopsy has been sequenced using both methodologies, points are linked by grey lines. The fold change between the median burden from nanoseq and the median burden from bulk whole genome sequencing along with the absolute difference in substitutions (subs) is shown within each plot. For each case, a solid (infants) or dashed (children) red horizontal line represents a point estimate of the mutation burden derived from duplex sequencing of the normal kidney of the same patient as the tumour sample. c PD49348, a neonatal Wilms tumour (four biopsies studied by both bulk sequencing and nanoseq); d PD54846, a neonatal Wilms tumour (four biopsies studied by bulk sequencing alone and one biopsy studied by nanoseq alone); e PD63760, a Wilms tumour from a 4-month-old (four biopsies studied by both bulk sequencing and nanoseq); f PD37585, a Wilms tumour from a 6-month-old (three biopsies studied by both bulk sequencing and nanoseq); g PD52209, a Wilms tumour from a school-age child (two biopsies studied by both bulk sequencing and nanoseq); and h PD52230 a Wilms tumour from a school-age child (one biopsy studied by both bulk sequencing and nanoseq, and one biopsy studied by nanoseq alone). Source data are provided as a Source Data file.

Fig. 2. Clonal architecture of Wilms tumours.

a A schematic illustrates that with a conventional clonal architecture (left-hand phylogeny), a high proportion of the per-tumour-cell mutation burden may be captured by bulk sequencing. In contrast, with extensive polyclonal diversification (right-hand phylogeny) only a small proportion of the per-cell mutation burden is captured. The colour key applies to all the phylogenies shown below. b–f Phylogenies of Wilms tumours, newly generated using either single-cell-derived organoids laser capture microdissections. Every tip represents either a single-cell-derived organoid or a microbiopsy. g Previously published phylogeny of an adult colorectal cancer, constructed using whole genome sequencing of single-cell-derived organoids. h Previously published phylogeny of normal colonic crypts from an adult, constructed using whole genome sequencing of crypts isolated by laser capture microdissection. Source data are provided as a Source Data file.

Fig. 3. FOXR2-mutant Wilms tumours.

a–c For each of three infant Wilms tumours, the structural variant placing FOXR2 under the control of a different promoter is shown. d For an example neonatal Wilms tumour with FOXR2 rearrangement, FOXR2 and MYCN immunohistochemistry showed strong protein expression compared to normal kidney. Further cases are shown in the Supplementary Note. Immunohistochemistry for each antibody was performed three times with similar results. e Whole transcriptomes were generated for 486 samples from a cohort of 264 Wilms tumours and ranked by FOXR2 expression (TPM, transcripts per million). All cases save one (highlighted in red) did not express FOXR2 at all. The case with elevated FOXR2 expression had a complex rearrangement placed FOXR2 under the control of the promoter of ZNF529, shown beneath the RNAseq. f–h Haematoxylin and eosin sections showing fibroadenomatoid morphological features of FOXR2-rearranged infant Wilms tumours (see Supplementary Note for a further discussion). For each case, the image shown was chosen as representative of fibroadenomatoid architecture based on the examination of a single section of the tumour. i Hierarchical clustering of bulk RNAseq for a cohort of Wilms tumours including infant FOXR2-rearranged Wilms cases, other infant Wilms tumours, and an older FOXR2-rearranged case. Infant FOXR2-rearranged cases cluster together and separately from other cases. j Comparison of substitution burden and number of annotated driver mutations for a cohort of Wilms tumours by patient age. Only tumours from patients without a known predisposition are included. If there is more than one sample per tumour, the median across all samples is used. For each boxplot the black bar represents the median, the box the interquartile range, and the whiskers extend to the most extreme data point which is no more than the interquartile range from the box. Source data are provided as a Source Data file.