Rare germline structural variants increase risk for pediatric solid tumors

Abstract

Pediatric solid tumors are rare malignancies that represent a leading cause of death by disease among children in developed countries. The early age-of-onset of these tumors suggests that germline genetic factors are involved, yet conventional germline testing for short coding variants in established predisposition genes only identifies pathogenic events in 10-15% of patients. Here, we examined the role of germline structural variants (SVs)-an underexplored form of germline variation-in pediatric extracranial solid tumors using germline genome sequencing of 1,766 affected children, their 943 unaffected relatives, and 6,665 adult controls. We discovered a sex-biased association between very large (>1 megabase) germline chromosomal abnormalities and a four-fold increased risk of solid tumors in male children. The overall impact of germline SVs was greatest in neuroblastoma, where we revealed burdens of ultra-rare SVs that cause loss-of-function of highly expressed, mutationally intolerant, neurodevelopmental genes, as well as noncoding SVs predicted to disrupt three-dimensional chromatin domains in neural crest-derived tissues. Collectively, our results implicate rare germline SVs as a predisposing factor to pediatric solid tumors that may guide future studies and clinical practice.

Fig. 1 |. The landscape of germline SVs in 9,374 pediatric cancer patients and controls

(A) Roadmap of key datasets and analyses in this study. (B) Summary of study cohort structure after excluding low-quality WGS samples. (C) A principal component analysis of common, high-quality SVs stratified individuals based on their genetic ancestry. See also Fig. S2. (D) We detected an average of 7,276 high-quality SVs per genome, which was correlated with sample ancestry as expected based on human demographic history. (E) The allele frequencies (AFs) of SVs detected in this study were strongly correlated with AFs of SVs reported from WGS of 63,046 unrelated individuals in the Genome Aggregation Database (gnomAD). European samples shown here; see Fig. S3 for other ancestries. (F-G) Most SVs were small (<1kb) and rare (AF<1%). Del.: deletion. Dup.: duplication. mCNV: multiallelic CNV. Ins.: insertion. Tloc.: reciprocal translocation.

Fig. 2 |. Very large germline CNVs increase risk for pediatric solid tumors in a sex-specific manner.

(A) We identified 143 rare (AF<1%), large (>1Mb) germline SVs in a set of 1,745 pediatric cancer cases and 4,983 ancestry-matched adult controls. Auto.: autosomal. Allo.: allosomal. Neuro.: neuroblastoma. Osteo.: osteosarcoma. (B) WGS delineated sample sex based on ploidy (i.e., copy number) estimates of X and Y chromosomes. (C) We discovered an association between pediatric solid tumors and the 84 unbalanced SVs from (A). We did not observe any significant associations with comparably large but balanced rare SVs. Error bars indicate 95% confidence intervals (CI); P values derived from logistic regression adjusted for sex, cohort, and ancestry. (D) Results from (C) restricted to autosomes and stratified by male (karyotypic XY) vs. female (karyotypic XX) samples. (E) Proportion of cases and controls who carried at least one singleton unbalanced germline SV larger than the size specified on the X-axis. See also Fig. S4F. (F) Relationship between the size of unbalanced singleton SV size and the corresponding odds ratio for pediatric cancer based on the data from (D) presented as a cubic smoothing spline. Shaded area is 95% CI for a pooled model of both sexes from all histologies. (G) The total sum of nucleotides altered by autosomal, rare, unbalanced SVs did not significantly differ between cases & controls.

Fig. 3 |. Pediatric cancer patients carry an excess of rare germline SVs that impact disease-relevant genes

(A) On average, the number of protein-coding genes disrupted by all germline SVs was significantly greater among pediatric cancer cases relative to adult controls. (B) Rare gene-disruptive SVs were enriched in cases versus controls, and this enrichment was inversely correlated with SV frequency. Shaded area indicates 95% CI for all histologies. (C) We found comparable enrichments in two independent subsets of singleton SVs with opposing predicted consequences on gene function: loss-of-function (LoF) and whole-gene copy gain (CG). (D) We carried out a category-wide association study (CWAS) in neuroblastoma and Ewing sarcoma, combining six layers of filters to categorize types of coding SVs for burden testing.^, We evaluated 679 and 714 categories of gene-disruptive germline SVs in Ewing sarcoma and neuroblastoma, respectively. (E) In neuroblastoma, 27 categories of SVs exceeded Bonferroni significance, including singleton LoF SVs impacting mutationally constrained genes and genes expressed in adult adrenal gland. (F) In Ewing sarcoma, no single category of SVs was enriched in cases or controls at Bonferroni significance, although multiple categories of potential biological interest were nominally significant.

Fig. 4 |. Ultra-rare germline SVs in pediatric cancer patients dysregulate gene expression in premalignant tissues and in tumors

(A) Diagrams of CWAS category relationships. Singleton SV categories are depicted as nodes and filters as edges. The x-axis represents the number of filters applied to each category, while the vertical space between categories is proportional to their similarity (Jaccard index). Example filter paths containing at least nominally significant categories are shown in black. (B) Gene set enrichment results for risk-carrying singleton SVs and GO biological processes. Odds ratios were computed by a Fisher’s exact test of an SV-affected gene being in the category and also in the gene set. (C) Gene set enrichment results for select gene sets and categories in cases and controls. The three categories shown for each gene set represent singleton SVs, singleton deletions, and singleton deletions affecting genes expressed in the tissue-of-origin (i.e., successively applied filters). Bars represent 95% CI. (D) Expression in GTEx v8 for genes affected by singleton SVs in controls, unique to controls, in cases, and unique to cases. P values from Wilcoxon test. (E) Expression heatmap of genes expressed in adrenal tissue that were affected by singleton SVs in the subset of neuroblastoma patients with tumor RNA data available. Each row represents a gene and each column represents a sample; samples within each row are ordered by expression z-score. Samples with SVs (one per row) are connected by a purple line. Diagonal gray line indicates expectation under a uniform null. (F) Effect on tumor expression of SVs in significant neuroblastoma categories. Y-axis represents the P value of comparing to the null uniform distribution (i.e., no effect on expression). Larger points represent higher-level groupings of SVs. Indicated P value is from bootstrap-comparing the category from (E) to the background of all coding SVs.

Fig. 5 |. Gene-disruptive germline SVs in COSMIC and cancer predisposition genes (CPGs) in pediatric patients with solid tumors

(A) We found germline LoF deletions in DNA damage repair genes, such as PALB2 in Ewing sarcoma and BARD1 in neuroblastoma. (B) We also observed germline LoF SVs of known CPGs, like PHOX2B in neuroblastoma and FANCA in Ewing sarcoma, that were carried by affected children but were inherited from unaffected parents. (C) RAS-MAPK genes were impacted by germline SVs, including a singleton de novo complex SV resulting in a two-exon deletion of BRAF in one neuroblastoma case and an ultra-rare polymorphic duplication over RAF1 in three unrelated cases. (D) The germline RAF1 duplication from (C) was associated with high RAF1 expression in a neuroblastoma tumor, and increased expression of TMEM40, the gene upstream of the RAF1 promoter, in neuroblastoma and osteosarcoma tumors. (E) A rare polymorphic duplication predicted to result in CG of ERCC2, a DNA damage repair gene, was found at a three-fold higher frequency in Ewing sarcoma cases than in controls. (F) We discovered a de novo germline MYCN duplication and a likely post-zygotic MYCN duplication in two unrelated neuroblastoma cases. (G) The overall rates of gene-disruptive rare SVs in CPGs and COSMIC cancer genes were not significantly higher in cases relative to controls.

Fig. 6 |. Noncoding germline SVs impacting TAD boundaries in disease-relevant tissues increase risk for neuroblastoma

(A) We performed a CWAS for rare, noncoding SVs in neuroblastoma cases vs. controls, finding that singleton germline SVs overlapping adrenal gland-derived TAD boundaries were significantly enriched in cases after correcting for the estimated number of effective CWAS tests (“Bonferroni”). (B) No noncoding SV categories reached the threshold of Bonferroni significance for enrichment in Ewing sarcoma cases relative to controls. (C) Quantile-quantile plots for all noncoding CWAS categories for neuroblastoma (top) and Ewing sarcoma (bottom) demonstrated good calibration of CWAS test statistics. Dashed horizontal line corresponds to Bonferroni significance. (D) We observed a significantly stronger effect size for noncoding singleton SVs intersecting TAD boundaries defined in the putative tissue-of-origin (adrenal gland) relative to all noncoding singletons in neuroblastoma (P=3.6×10⁻⁴, Welch’s t-test), whereas this difference was not seen in Ewing sarcoma (P=0.42), consistent with the lack of a TAD-related association in the Ewing sarcoma CWAS. (E) The enrichment in neuroblastoma cases for singleton germline SVs intersecting TAD boundaries was significant in the St. Jude and GMKF cohorts when analyzed separately. (F) Effect sizes for singleton SVs intersecting TAD boundaries in neuroblastoma increased monotonically when subsetting to deletions in proximity to genes expressed in adrenal tissue.