Resolving the Pathogenesis of Anaplastic Wilms Tumors through Spatial Mapping of Cancer Cell Evolution

This article is featured in Highlights of This Issue, p. 2559

Wilms tumor is the most common pediatric renal neoplasm (1). Its histology is similar to that of the developing kidney and is typically triphasic, containing undifferentiated blastemal cell populations along with elements maturing along epithelial and stromal lineages. Biphasic tumors and monophasic tumors also occur, but are rarer (2, 3). The histologic classification of Wilms tumor is determined by the predominance (>66%) of one histologic component; if there is no predominance the tumor is annotated as mixed type (4). In Wilms tumor, there are three risk groups: (1) low risk, which is represented by totally necrotic Wilms tumor and cystic, partially differentiated Wilms tumor (2), intermediate-risk (IR), represented by the group with epithelial, stromal, mixed morphology, and focal anaplasia (3), and high-risk, represented by blastemal type (BT) and diffuse anaplastic Wilms tumor (5). Diffuse anaplasia makes up 5%–10% of Wilms tumors, and is characterized by multiple foci containing pleomorphic tumor cells with hyperchromatic, giant nuclei, and large multipolar mitoses (6).

Most patients with Wilms tumor in Europe are treated with preoperative chemotherapy within prospective clinical trials conducted by the International Society of Paediatric Oncology – Renal Tumor Study Group (SIOP–RTSG, Europe). For localized disease, this entails actinomycin D and vincristine. For metastatic disease at diagnosis, the treatment regimen consists of the two aforementioned drugs with the addition of doxorubicin. The long-term survival rate of Wilms tumor is now more than 90%, all subtypes combined (7). However, patients with diffuse anaplasia have an overall survival (OS) of only around 50% (8). This is caused by an inferior response to chemotherapy, radiotherapy, and a lack of available targeted therapies or immuno-oncologic treatments. With the significant achievements in treatment of favorable histology Wilms tumors, we have here chosen to focus on Wilms tumor diffuse anaplasia tumors with their clinical challenges. Previous studies have shown that anaplastic tumors have mutations in the tumor suppressor gene TP53 much more frequently than other Wilms tumors (9). However, there is remarkable intratumor genetic heterogeneity in Wilms tumor diffuse anaplasia, with TP53 mutations typically observed only in specific parts of diffuse anaplasia tumors, indicating that they are late events in tumor evolution (10–15). In previous reports, around 50% of Wilms tumors with anaplasia have been shown to harbor loss of and/or mutations in TP53 (7), with a close correlation between TP53 mutations and anaplastic morphology (14, 15). However, also non-anaplastic areas, particularly blastemal areas, have been described as having TP53 mutations suggesting that mutations in this gene may precede the development of full-scale anaplasia (14). We, and others, have previously pointed to the importance of multiregional tumor sampling to get a representative overview of genetic aberrations in Wilms tumor and comprehensively resolve the genetic pathways leading up to anaplasia (11, 12).

In this study, we mapped in detail the genetic events leading up to the state of anaplasia in individual patients, along with correlations to morphologic parameters. To this purpose, we used a unique type of histologic specimen: a tumor whole-mount section (WMS). A WMS is produced by embedding in paraffin an approximately 0.5-cm thick section through a tumor, covering a surface area of up to a decimeter in diameter, allowing mapping of histologic features in a larger anatomical context. Using WMSs from a retrospective Wilms tumor cohort, we characterized the landscape of TP53 mutations and segmental copy-number alterations along with anaplastic features, proliferation rates, and histologic elements across 189 tumor areas from 20 Wilms tumors.

Twenty patients diagnosed with Wilms tumor were included in the study, all derived from a retrospective consecutive cohort that underwent preoperative chemotherapy based on the SIOP guidelines at Skåne University Hospital in Lund (Lund, Sweden; Supplementary Tables S1 and S2). The study was conducted in accordance with the Declaration of Helsinki and approved by the Regional Ethics Review Board (Reference no. L2011/289 with update 2017). Written informed consent was obtained from all the patients’ legal guardians. All patients with tissue sent to the regional biobank during the years 1992 to 2020 with the diagnosis Wilms tumor were screened for inclusion through a histopathologic review process. The following criteria had to be fulfilled for inclusion: diagnosis confirmed by two pathologists (primary and SIOP reference pathologist), availability of formalin-fixed, paraffin embedded (FFPE) blocks covering viable tumor components, and documented location of these samples within the primary tumor. Care was taken to include all tumor regions, and to include all regions where anaplasia was denoted by the pathologists. Primarily, tumors subjected to whole-mount paraffin embedding, that is, WMS, were selected for the present study. IR Wilms tumor WMS were selected on the basis of the high tumor cell-content, whereas all blastemal and anaplastic cases available were included in the cohort. When a WMS was not available, conventional paraffin blocks were sampled instead (BT_2, DA_1, DA_3, and DA_5). Some cases of WMS were complemented with samples from conventional paraffin blocks to provide an as extensive anatomic survey of the tumor parenchyma as possible. The selection process resulted in five patients with diffuse anaplasia with sufficient material for analysis; for comparison, five patients with blastemal-type tumors and 10 patients with IR histology were included. Paired normal FFPE samples from nearby non-tumor tissue were available for all patients. In total, 189 FFPE samples were included, where 20 of these samples were from normal kidney tissue (Supplementary Table S3).

To systematically evaluate spatial tumor heterogeneity, grids composed of 1×1 cm squares were marked on hematoxylin and eosin (H&E)-stained tumor slides to be used as tumor coordinates (Fig. 1). Large areas of necrosis, regression, or other debris were excluded from sampling. Grids were transferred to FFPE blocks to enable sampling. During multiregional sampling, as many tumor areas/coordinates as possible were sampled. To obtain a sufficient amount of DNA, two cores of 1 mm in diameter were taken from within each 1×1 cm square of the paraffin blocks (tools from Tissue-Tek Quick-Ray, Tissue Microarray System). The two 1 mm cores were combined to one sample for DNA preparation. Tumors within the cohort varied in size and tumor cell content and thus the number of samples taken from each tumor varied in number. DNA was extracted using the AllPrep DNA/RNA FFPE kit, from Qiagen, according to standard protocols.

DNA from FFPE samples was subjected to whole-genome copy-number analysis using the SNP array OncoScan CNV assay (Affymetrix). This platform is specialized for analyses of FFPE material. The analyses were performed at the Swegene Centre for Integrative Biology (SCIBLU) at Lund University and at the Array and Analysis Facility, Department of Medical Sciences, at Uppsala University (Uppsala, Sweden) and at Eurofins Genomics. SNP array raw data were analyzed as previously described (10, 16). In brief, the generated OSCHP-TuScan files were analyzed with Nexus Copy-Number 10.0 (BioDiscovery) from which segment files were produced where gains, losses, and copy number neutral imbalances (CNNI) were included if ≥5 Mbp, covering ≥20 SNPs and manifested as more than one data point in the Tumor Aberration Prediction Suite (TAPS) scatterplot (17). Allelic imbalances (AI) were assessed by inspecting TAPS plots. For some cases, the ploidy level was changed in Chromosome Analysis Suite (ChAs) and then converted to OSCHP-TuScan files to be analyzed in Nexus.

DNA samples with OncoScan SNP array profiles of adequate quality and sufficient DNA available were selected for deep sequencing of TP53. TP53 was targeted by using the Agilent SureSelectXT capture kit. The libraries were sequenced on an Illumina NovaSeq 6000 with a S4 XP flow cell using 150 bp paired-end reads (Eurofins Genomics). The data were base-called and demultiplexed by standard Illumina software. Reads were mapped to the human reference genome (GRCh37) using BWA MEM (arXiv:1303.3997v1; q-bio.GN.; BWA, RRID:SCR_010910), followed by sorting and deduplication with Samtools (SAMTOOLS, RRID:SCR_002105; ref. 18). Somatic small variants were called using Mutect2 (https://www.biorxiv.org/content/10.1101/861054v1) in multisample mode (i.e., jointly calling variants using all samples from the same patient) utilizing a panel based on all normal samples sequenced as part of this project. Raw variant calls were filtered using FilterMutectCalls from the GATK suite of tools, with standard settings except –max-events-in-region which was set to 15 to not filter out subclonal mutations located in close proximity to each other (19). It was required that a variant was present in 5% of all reads and that the coverage was ≥50 in one unique/single sample not to be filtered away. By this, low-frequency mutations that are likely to be noise due to FFPE samples were omitted. Mutations which remained after filtering were manually curated with the Integrative Genomics Viewer (Broad Institute, Cambridge, MA; ref. 20).

Where |${N_A}$| represents the number of A-alleles and |${N_B}$| represents the number of B-alleles, with |${N_B} > \;{N_A}$|⁠, at a heterozygous locus.

On the basis of the MCFs, subclonal deconvolution was performed using the DEVOLUTION algorithm (v.1.1; ref. 21). The input file for DEVOLUTION was a segment file made for each patient, encompassing all genetic changes, specified by chromosomal position, type (gain, loss, or CNNI), and MCF for each genetic alteration in each sample. Clones and subclones were defined as in (10), as cell populations with unique genetic profiles present in |$ \ge$|90% or <90% of tumor cells, respectively, in a sample. This was followed by phylogenetic reconstruction. Phylogenies were generated using both the maximum parsimony (MP) and maximum likelihood (ML) method as well as a modified maximum parsimony method (MMP). Each phylogeny was rooted in a normal cell having no genetic alterations. Trees were visualized using ggplot2 (v.3.3.5; ggplot2, RRID: SCR_014601). The code for MMP is freely available at github (https://github.com/NatalieKAndersson/MMP). It tries to minimize the number of genetic alterations in the phylogeny needed to explain the data (similar to MP), while also taking into account the proportion of cells in each sample that harbor a certain alteration, to not contradict the pigeonhole principle. When there were no contradictions in the dataset (such as MCF-crossover) the three methods resulted in identical phylogenies. When there were contradictions, the MMP-method generated trees that were more biologically plausible, not contradicting the pigeonhole principle in any sample. Four cases did not yield a phylogenetic tree, due to either an absence of detectable chromosomal aberrations (IR_1) or the same chromosomal aberrations being detected throughout the tumor (IR_2, IR_3, and BT_1). In cases BT_2, BT_3, DA_1, IR_4, IR_7, DA_2, and DA_3 CNAs were inferred in the segment file to avoid paradoxes such as loss of LOH events (21).

To quantify differences in phylogenetic parameters between clinical subgroups, we assessed phylogenetic species richness (PSR), divergence and irregularity for each phylogenetic tree (Fig. 2A). PSR is defined as the sum of all phylogenetic distances between all detected subclones (ref. 22; Fig. 2B). Divergence represents the mean phylogenetic relatedness within the phylogenetic tree, that is, total PSR/number of PSR comparisons (Fig. 2C). Irregularity characterizes how the phylogenetic tree varies from a star phylogeny. This was calculated as the variance of phylogenetic distances to all subclones from the stem (Fig. 2D). How different phylogenetic metrics relate to each other and how they can describe a phylogenetic tree, and reflect its complexity, are visualized in Fig. 2E and F. For comparisons between tree parameters, P values were calculated by Mann–Whitney U test with continuity correction, adjusted according to Bonferroni for multiple comparisons.

Potential correlations between the variables CNAs, mitoses, and anaplasia were assessed by Pearson correlation test (R version 4.0.4). Regression was scored as percentage of tumor surface consisting of hemosiderophages, fibrosis, lymphocyte infiltration, and micronecrosis. Proliferation was evaluated by the number of observed mitoses/10 high-power field (HPF). Anaplasia scoring was based on hyperchromasia, enlarged nuclei, and multipolar mitoses. Each feature contributed with 1 score point, with 0 corresponding to a complete absence of anaplastic features and 3 to full-scale anaplasia.

We applied this model to anaplastic and non-anaplastic cases separately. The model was fit using the lme4 package (https://doi.org/10.18637/jss.v067.i01) using the R statistical software R (v4.2.2; R Core Team 2022) with standard settings.

The data generated in this study are publicly available in Zenodo at https://doi.org/10.5281/zenodo.7784749. For any additional data, please contact corresponding author.

Whole-genome profiling of CNAs was performed for multiple tumor regions (median 9/tumor) in 20 Wilms tumor: 10 of IR, five of BT and five with diffuse anaplasia. All tumors had been treated with actinomycin D and vincristine prior to resection. Case DA_5 was also treated with doxorubicin as it presented with metastatic disease (Supplementary Table S1). In total, 169 regions were successfully analyzed for CNAs, resulting in the identification of 144 genetically distinct tumor cell subpopulations, which were either clonal (constituting ≥90% of tumor cells in a region) or subclonal (constituting <90%; Supplementary Table S4: https://doi.org/10.5281/zenodo.7784749). There was no difference in the number of samples (analyzed regions) among the different histologic tumor types (P = 0.18; Mann–Whitney U test), nor was there a correlation between the number of samples from each tumor and the number of clones and subclones identified (r = 0.02; P = 0.92; Spearman rank correlation).

Following subclonal deconvolution, phylogenetic trees were reconstructed for 16 tumors on the basis of the CNA profiles and from TP53-sequencing data (Supplementary Fig. S1). One case did not show any AIs (IR_1, six regions sampled) and three cases (IR_2, IR_3, and BT_1) each had a single clonal population (Fig. 3A–C; Supplementary Fig. S1) precluding phylogenetic reconstruction. The phylogenetic trees showed a broad diversity in structure (Fig. 3D–R). A few trees showed only linear (n = 2, IR_7 and IR_10) or branching (n = 1, IR_6) evolution, while the majority (n = 13) exhibited a combination of linear and branching evolution across the tumor space, clearly confirming that intratumoral heterogeneity of CNAs is the norm in Wilms tumor.

Although the structures of the phylogenetic trees were highly variable among IR and BT tumors, diffuse anaplasia cases uniformly exhibited multigenerational trees with combined branching and linear evolution (Supplementary Fig. S1, DA_1-DA_5 and Fig. 3M–R). In total 4/5 diffuse anaplasia cases exhibited one or several TP53 mutations in combination with AIs leading to loss of the nonmutated allele (Supplementary Table S3). However, these events occurred in variable combinations across the space of each tumor. In DA_1, they were confined to one single nodular tumor compartment, with clonal CNNI of chromosome 17 combined with two TP53 missense mutations (p.C135F and p.C141Y). Interestingly, in a specific subcompartment defined by surrounding necrosis, this was followed by an additional missense mutation (p.R175H), while in a neighboring compartment there was a subclonal population with deletion of the TP53 region (Supplementary Fig. S1; DA_1: A, B, I, and J). Similarly, in DA_3 two neighboring regions shared the same clonal CNNI involving the TP53 locus in 17p accompanied by a missense mutation (p.R273H). This was followed by additional but distinct CNAs involving 17p in each of the regions (Supplementary Fig. S1; DA_3 I and J). In DA_4, a combined deletion and missense mutation (p.H179Y) of TP53 was confined to an island of viable tumor cells largely bordered by necrosis, while surrounding viable tumor areas either had only the aforementioned mutation or either of two distinct deletions of 17p (Supplementary Fig. S1; DA_4 I–J; Fig. 3,M–O). Finally, in DA_2 an intragenic deleterious deletion in TP53 (p.MFREL340L) was common to five distinct compartments surrounded by necrosis. These compartments also harbored distinct copy-number AIs rendering the aforementioned mutation homozygous (Supplementary Fig. S1 DA_2: A, B, I and J; Fig. 3P–R). The fifth case with diffuse anaplasia DA_5 had an atypical, monotonous pleomorphic spindle cell morphology with stromal features but was nevertheless classified as Wilms tumor diffuse anaplasia by the SIOP review panel. DA_5 harbored neither sequence mutations nor allelic loss of TP53 but a highly complex phylogeny resulting in a very high CNA burden distributed across multiple compartments separated by cystic necrosis (Supplementary Fig. S1; DA_5). In all cases of classical diffuse anaplasia (DA_1–4), full-scale anaplastic features (giant, hyperchromatic nuclei, and multipolar mitoses) were confined to areas with either hemi- or homozygous TP53 mutation. Only in one single area in one case (DA_4, an area without multipolar mitoses) was there a TP53 mutation or loss without all three features of diffuse anaplasia (Fig. 3,M–O; Supplementary Fig. S1, DA_4). In summary, all tumors with classical Wilms tumor diffuse anaplasia morphology exhibited TP53 inactivation events, with features of parallel evolution across different anatomic regions with anaplastic morphology. Notably in three of these cases, this parallel evolution was often preceded by an earlier disruption of TP53, either homo- or hemizygously.

To assess the complexity of phylogenetic trees, three different parameters were used: PSR, divergence, and irregularity (Fig. 2). By summing up all phylogenetic distances between subclones in a tree, PSR provides an overall measure of genetic diversity, while divergence reflects how genetically distant subclones are from each other on average; the irregularity parameter, in its turn, reflects to what degree genetic distances in a tree are uniform (a star phylogeny) or different from each other (Fig. 2). On the basis of these three parameters phylogenetic trees of Wilms tumor diffuse anaplasia displayed a significantly increased complexity compared with the phylogenetic trees from IR and BT tumors (Fig. 4A–C; Supplementary Fig. S2A–S2C). In addition, the total number of CNA was significantly higher in anaplastic tumors (P = 0.008; Mann–Whitney U test with continuity correction, adjusted according to Bonferroni for multiple comparisons; Fig. 4D; Supplementary Fig. S2D). The tumors with diffuse anaplasia showed a significantly higher number of genetically distinct tumor cell subpopulations (median 13) compared with IR tumors (median 3; P = 0.01; Mann–Whitney U test), with BT tumors midway between (median 7; Fig. 4E; Supplementary Fig. S2E). This difference was retained after normalization against the number of samples per tumor (P = 0.025; median 0.37, 1.0, and 1.3 subpopulations per sample for IR, BT, and diffuse anaplasia, respectively). Because of variations in tumor size and tumor cell-content, the number of analyzed samples differed between tumors. There was no association between the number of samples per tumor and the different complexity parameters, nor with the number of CNAs detected (Spearman correlation, PSR P = 0.7, divergence P = 0.6, irregularity P = 1.0, and CNA P = 0.3).

In summary, these findings indicate that the increased complexity of tumor cell evolution in diffuse anaplasia manifests in several ways, including a higher number of identified subpopulations (clones and subclones), increased PSR, and divergence as well as an increased asymmetry in the phylogeny.

To understand how CNA burden was connected to anaplasia, cell proliferation, and cell death across different Wilms tumor histologies we used a negative binomial hierarchical model as described in Material and Methods section. The results showed that there was an accumulation of CNAs as well as increased regressive changes as anaplastic features increased in Wilms tumor with diffuse anaplasia (P = 0.01 and P ≤ 0.001), whilst in non-anaplastic tumors, the CNA burden increased with proliferation (P = 0.001; Supplementary Fig. S3A–S3C).

Wilms tumors typically consist of one or several tumor nodules, separated by fibrous septae, necrosis, or other regressive phenomena [Supplementary Fig. S1A (IR1–10, BT 1–5, and DA_1–5)]. We sought to compare clonal evolution happening within compartments to events taking place at formation of new compartments — the latter manifested as genetic variation across compartments. To this end, we utilized the maps of CNAs across WMSs (phylogeographies) to compare shifts in the subclonal landscape within compartments to shifts occurring across compartments (Fig. 5A). Shifts in the subclonal landscape were classified into four evolutionary trajectories as previously described (ref. 10; Fig. 5B). Trajectories at transition from one tumor sample (area) to another included tumor cell twinning (TCT), that is, an identical and homogeneous clonal landscape, clonal coexistence (COEX), subclonal variation (VAR), and clonal sweeps (SWE), the latter implying the emergence of a new clone encompassing ≥90% of tumor cells with or without new daughter clones. A total of 18 tumors could be included in this analysis after excluding one case where no aberrations were detected and one where compartment architecture could not be evaluated because the tissue sections available did not cover a sufficient area. This selection allowed 170 comparisons between areas within compartments (intracompartmental) and 133 comparisons across compartments (intercompartmental).

The data revealed striking differences in how clonal landscapes evolved within, as compared with across compartments (Fig. 5C). The most prevalent trajectory within compartments was TCT, implying that they often contain homogenous populations with respect to copy-number variation. Variation was limited to either shifting panoramas of subclones (VAR) or the presence of the same set of subclones in different areas of a compartment (COEX). Only very rarely (2% of comparisons) were there SWE within compartments. In contrast, the majority (76%) of comparisons across compartments in the same case exhibited SWE. Although there were TCT and rare instances of VAR, not a single comparison between compartments was consistent with COEX. Taken together, these data implied that new compartments budded off from those already present in a process that rarely if ever encompassed more than a single subclone (Fig. 5D). If this clone had a genotype different from that of the founder cell of the ancestral nodule, the result would be SWE, but if it did not show any new mutations, the result would be TCT, or VAR if subclones were born after the new nodule had been formed.

Previous studies have demonstrated a stepwise accumulation of somatic mutations in Wilms tumor, with uniparental isodisomy of 11p being a frequent early event, followed by extensive branching evolution where TP53 mutations are usually late events coupled to the emergence of anaplasia (10, 13, 14). There is evidence that this evolutionary process is even initiated before birth in premalignant precursor lesions resident in the embryonic kidney or kidney primordia (23). In this study, we compared IR and BT tumors to diffuse anaplasia, to search for key events in the pathogenesis of diffuse anaplasia by spatially mapping cancer cell evolution in 20 Wilms tumor. The clonal landscapes across tumor space were mapped by using copy number profiling and targeted mutation analysis of the TP53 gene, followed by subclonal deconvolution and phylogenetic reconstruction. A total of 189 regions were analyzed, which resulted in identification of 144 genetically distinct tumor cell subpopulations. Intratumor diversity of CNAs was seen in 16/20 cases. Phylogenetic analyzes showed that 13/20 tumors demonstrated a combination of linear and branching evolution across tumor space.

We found significant differences in phylogenetic tree complexity between non-anaplastic and diffuse anaplasia tumors, with higher PSR, irregularity, and divergence in the latter. Wilms tumor diffuse anaplasia also had a higher CNA burden and amount of subpoplutaions than IR and BT tumors. We demonstrated that in diffuse anaplasia, morphologic features of anaplasia correlated to a high CNA burden. Thus, extensive phylogenetic tree complexity appears to be a hallmark of Wilms tumor diffuse anaplasia. The starting point for tree complexity could be debated, although in the current study, TP53 mutation often is the starting point of such complexity. In total, 4/5 diffuse anaplasia cases showed one or several TP53 mutations. When analyzing TP53 mutation and allelic loss we found that these events occurred in variable combinations across tumor space in each tumor. Notably, TP53 mutation/loss in different combinations often occurred by parallel evolution and was linked to saltatory evolution by an accumulation of CNAs. We finally showed that clonal sweeps were strongly linked to the formation of macroscopically distinct tumor compartments and that new compartments likely emerged from a monoclonal rather than a polyclonal background. When analyzing the diffuse anaplasia tumors we noticed that anaplastic features were detected both within and across tumor compartments.

Our findings are in agreement with previous studies showing that not all diffuse anaplasia tumors have TP53 mutations. Former studies were performed on fewer samples from each tumor which effects the likelihood of detecting TP53 mutations. No TP53 mutations were detected in DA_5 despite extensive sampling (five tumor samples and one sample from normal kidney as control, were analyzed for TP53 mutations). DA_5 did diverge from DA_1 to _4 histology wise. DA_5 had a monotonous pleomorphic spindle cell morphology with stromal features but was nevertheless classified as Wilms tumor diffuse anaplasia by the SIOP review panel; it did not fulfil the criteria of anaplastic sarcoma of the kidney, as anaplasia was not widespread and it had neither cartilage nor chondroid material (24). In addition, for DA_5, the tumor progressed and the patient died, which is an unlikely outcome for patients with anaplastic sarcoma of the kidney, which in general have a good prognosis, which supports that DA_5 is a Wilms tumor. One scenario could be that, in DA_5, the p53 pathway is disturbed in other ways than by TP53 mutations (25).

Considering that the 5-year overall and event-free survival rates are comparable between Wilms tumor diffuse anaplasia with wild-type (WT) TP53, and Wilms tumor with favorable histology, it could be discussed whether diffuse anaplasia tumors with and without TP53 mutations should be regarded as two different risk groups (7). However, it should be noted that the similar survival rates are contingent on WT TP53 diffuse anaplasia cases receiving high-intensity treatment and it is highly unclear if the same result would be achieved with less treatment. Also, the atypical morphology of DA_5 also highlights the fact that WT TP53 diffuse anaplasia cases may contain many diverse biological entities. Whereas TP53 mutations at times arise in tumors that lack criteria for anaplasia, the opposite may also occur, that is, that some anaplastic regions lack TP53 sequence mutations: however, these might still show an abnormal loss of the TP53 locus in chromosome arm 17p (13). Our findings that TP53 mutation and loss were invariably confined to branches in the subclone phylogenies indicate that these genetic changes are a late phenomenon in Wilms tumor pathogenesis, which is well in accordance with earlier studies (13). It has previously been suggested that anaplasia appears as an early focal event and later on spreads within the tumor (6). The phylogenetic trees of Wilms tumor diffuse anaplasia created here showed that tumor evolution leading up to anaplasia typically started with a single hit to TP53 in support of a focal event. However, this was followed by branching and saltatory evolution often encompassing parallel chromosome 17 alterations and further TP53 mutations across different tumor compartments, indicating that secondary TP53 alterations were actually multifocal.

This study leveraged the detailed anatomic resolution provided by using WMS to map subclone landscapes. On the other hand, the fact that WMS are FFPE tissue brings with it the limitations of low-quality fragmented DNA, currently not possible to interrogate efficiently with whole -genome or whole-exome sequencing methods. Therefore, in this study, sequencing was limited to a targeted approach, where we chose to specifically analyze TP53 mutations. Another limitation of this study is that some tumors were extensively affected by regressive changes, providing less viable tumor cells for analysis. Although, the number of patients that were included was small, the strength is the many samples that were extracted from each tumor. The fact that we found a high regional concordance between TP53 alterations and anaplastic morphology, in this study, highlights the value of comprehensive multiregional sampling.

In summary, our findings identify specific evolutionary features in Wilms tumor diffuse anaplasia compared with other Wilms tumor subtypes, which indicate that the TP53 mutations leading up to anaplasia start focally and then trigger secondary losses and/or mutations as part of a parallel and saltatory evolution. Our findings suggest that macroscopic compartments are markers of clonal sweeps, which points to the value of cross-compartmental sampling to maximize the information value of genetic analyses of Wilms tumor.

No disclosures were reported.

B. Rastegar: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. N. Andersson: Software, methodology, writing–review and editing. A. Petersson: Methodology, writing–review and editing. J. Karlsson: Formal analysis, methodology, writing–review and editing. S. Chattopadhyay: Formal analysis, writing–review and editing. A. Valind: Data curation, formal analysis, writing–review and editing. C. Jansson: Methodology, writing–review and editing. G. Durand: Methodology, writing–review and editing. P. Romerius: Resources, writing–review and editing. K. Jirström: Resources, writing–review and editing. L. Holmquist Mengelbier: Conceptualization, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. D. Gisselsson: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This study was supported by the Swedish Childhood Cancer Fund, the Swedish Cancer Society, the Swedish Research Council and an EU Interreg Grant (iCOPE). B. Rastegar, N. Andersson, J. Karlsson, S. Chattopadhyay, A. Valind, C. Jansson, G. Durand, L. Holmquist Mengelbier, and D. Gisselsson were all supported by these funds.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.