Optical genome mapping identifies structural variants in potentially new cancer predisposition candidate genes in pediatric cancer patients

Genetic predisposition is one of the major risk factors for pediatric cancer, with ~10% of children being carriers of a predisposing germline alteration. It is likely that this is the tip of the iceberg and many children are underdiagnosed, as most of the analysis focuses on single or short nucleotide variants, not considering the full spectrum of DNA alterations. Hence, we applied optical genome mapping (OGM) to our cohort of 34 pediatric cancer patients to perform an unbiased germline screening and analyze the frequency of structural variants (SVs) and their impact on cancer predisposition. All children were clinically highly suspicious for germline alterations (concomitant conditions or congenital anomalies, positive family cancer history, particular cancer type, synchronous or metachronous tumors), but whole exome sequencing (WES) had failed to detect pathogenic variants in cancer predisposing genes. OGM detected a median of 49 rare SVs (range 27‐149) per patient. By analysis of 18 patient‐parent trios, we identified three de novo SVs. Moreover, we discovered a likely pathogenic deletion of exon 3 in the known cancer predisposition gene BRCA2, and identified a duplication in RPA1, which might represent a new cancer predisposition gene. We conclude that optical genome mapping is a suitable tool for detecting potentially predisposing SVs in addition to WES in pediatric cancer patients.

18 patient-parent trios, we identified three de novo SVs.Moreover, we discovered a likely pathogenic deletion of exon 3 in the known cancer predisposition gene BRCA2, and identified a duplication in RPA1, which might represent a new cancer predisposition gene.We conclude that optical genome mapping is a suitable tool for detecting potentially predisposing SVs in addition to WES in pediatric cancer patients.
cancer predisposition, optical genome mapping, pediatric cancer, structural variants What's new?
Around 10% of patients with pediatric cancer have been found to harbor a cancer predisposing germline alteration.However, the entire spectrum of genetic alterations has yet to be considered.Here, the authors analyzed 34 pediatric patients with a suspected underlying cancer predisposition syndrome by using the new optical genome mapping technique.They identified de novo or inherited genetic structural variants in BRCA2 and in the potential novel cancer predisposition gene RPA1.The study demonstrates that optical genome mapping is a suitable tool for detecting potentially predisposing structural variants in addition to whole exome sequencing in pediatric cancer patients.

| INTRODUCTION
2][3][4][5] However, most of these studies focused on single nucleotide variants (SNVs) and small insertions and deletions (indels) ≤20 bp.In all of these studies, the whole exome sequencing (WES) analysis was mainly concentrated on a predefined gene panel of known cancer predisposition genes (CPG).Therefore, there is some question as to whether the remaining $90% of patients carry any predisposing germline alteration, in particular as not the entire spectrum of genetic alterations was considered.In line with this, no causative genetic alteration was detectable in $80% of pediatric cancer patients with clinical signs indicative of an underlying cancer predisposition syndrome (CPS). 4,6though SNVs outnumber other types of DNA sequence variants by 7:1, structural variants (SVs) are 30-fold more prone to alter the gene expression.Moreover, 34% of all disease-causing variants are larger than SNVs, with 1/3 being >20 bp in size. 7Based on previous studies, 1.5% of adult cancer patients harbor a cancer predisposing SV. 8 By contrast, the frequency seems to be more variable in pediatric cancer patients, with reports of 0.4%-3.2% of patients harboring a pathogenic germline SV. 1,5 These studies are mostly based on short-read next generation sequencing approaches that have limitations with respect to accurate alignment in challenging genomic regions, which are prone to breakpoints causing SVs. 9 Hence, the exact prevalence of SVs in the germline of cancer patients might be substantially underestimated.
Recent proof-of-principle studies have shown that optical genome mapping (OGM) is a new method of reliably detecting constitutional SVs that can be used as an alternative to karyotyping, fluorescence in situ hybridization and DNA microarrays. 10,11 here report on the applicability of OGM as a complementary instrument to the existing sequencing methods for unbiased germline screening of pediatric cancer patients regarding cancer predisposing SVs.

| Patient cohort
We included 34 patients of our germline study and, if available, their parents. 12Those patients were positive for clinical criteria suggesting the presence of an underlying CPS, 13,14 but lacking pathogenic cancer predisposing variants in WES. 4 Refer to Supplementary Material and Methods for a detailed description of the cohort.

| Extraction and labeling of high-molecular weight DNA
We extracted high-molecular weight (HMW) DNA peripheral blood or bone marrow mononuclear cells from remission or, if available, from fibroblasts of the children and peripheral blood of the parents.Sufficient material was available from 18 child-parent trios, six child-parent duos and 10 child-singletons.The HMW DNA (750ng) was labeled, loaded on a Saphyr G2.3 chip and molecules were imaged by the Saphyr instrument (Gen 2, Bionano Genomics).Refer to Table S1 for an overview of the run statistics and quality metrics and Supplementary Material and Methods for a detailed description.

| De novo assembly and variant calling
The de novo assemblies against the human reference genome GRCh38, structural variant calling and annotation were performed using Bionano Solve (version 3.5.1,Bionano Genomics).
The Bionano Access software (version 1.5.2,Bionano Genomics) was used to visualize the data.For OGM data from the patient and one or both parents, we performed "Dual" or "Trio" variant annotation, respectively, to differentiate de novo and transmitted SVs.Refer to the Supplementary Material and Methods for a detailed description of the performed analysis and SV filtering.

| Variant annotation and interpretation
The final list of SVs was manually annotated using the UCSC genome browser tracks to identify whether an SV targets a gene (NCBI RefSeq, Ensembl: MANE Project [v1.0],GENCODE V39) and whether the SV region overlaps with a common SV (≥1% of the population) using the NCBI dbVAR Curated Common SVs database. 15We also checked whether the SVs overlap with known regulatory regions using GeneHancer (Double Elite). 16We annotated the overlap of an SV with a CPG using a published list. 4Genes targeted by an SV that potentially hits a coding region were explored with regard to clinical relevance for the cancerogenesis of the respective patient.

| Validation and functional testing
We performed validation experiments using DNA microarray, PCR-based assays and/or RNA sequencing for selected SVs and performed functional testing on P28 fibroblasts and age/gender-matched controls.Refer to the Supplementary Material and Methods for a detailed description of the performed analysis.

| Description of patient cohort
We analyzed 34 pediatric cancer patients with a median age of 5.4 years (range 0-18.5) at initial tumor diagnosis and a male to female ratio of 1.6:1 (Table S2).All patients harbored clinical signs indicative of an underlying predisposition such as congenital anomalies, concomitant conditions, multiple tumors or excessive toxicity to the chemotherapy (Figure 1A).Fifty-five percent of the patients harbored congenital anomalies, mainly of a mild form and not corresponding to a specific syndrome.Thirty-eight percent of the patients harbored positive family histories of pediatric and/or adult cancer.Nine patients were diagnosed with types of cancer strongly associated with genetic germline predisposition, such as sonic hedgehog-activated medulloblastoma.Moreover, three patients harbored more than two tumors (synchronous or metachronous) or showed excessive toxicity after chemotherapy.

| Identification of structural variants using optical genome mapping
We generated an average of 422.5 Gb of data per sample resulting in average 106Â effective coverage of the de novo assemblies (Table S1).SV calling yielded a median of 6455 (range 5696-6685) SVs per sample.Applying filter steps to exclude artifacts and to detect rare SVs, yielded 1831 SVs, with a median of 49 (range 27-149) SVs per patient.Deletions and insertions were the most frequent SV types (median of 30 [range 15-93] and 18 [range 10-53], respectively; Figure 1B, Table S3).95.6% of unbalanced SVs (considering deletions, duplications and insertions that can also constitute duplications) had a length of less than 50 kbp (Figure 1C).The majority of SVs were either intergenic (55%) or were located in an intron (19%) (Table S3).However, 26% of the SVs potentially affected coding sequences of one or more genes.
Due to our unique OGM dataset of 18 child-parent trios, we were able to identify de novo SVs (two deletions and one insertion) in 17% (3/18) of the children, affecting the DDX10 or FLG1 gene or an intergenic locus, respectively (Figure 1D, Table S4).This is in line with estimates based on WES published by Kloosterman et al, who reported de novo SVs >20 bp or >500 bp in 16% or 8% of children, respectively. 17

| Structural variants targeting known cancer predisposition genes
In our list of high-confidence SVs, we checked whether they targeted known CPGs. 4 Twenty-one SVs overlapped with a CPG, of which five potentially affected a coding region (Table S5).
Patients P6 and P10 both carried the same 1.4 kbp deletion in 17p13.1,potentially targeting the TP53 gene (Figure S1A,B).Further analysis revealed that the deletion maps 5.5 kbp upstream of the TP53 gene, overlapping with a described common deletion (nssv16197110) (Figure S1C,D).
After literature mining, we classified a deletion in BRCA2, which occurred in P26 as likely pathogenic (Table 1, Figure 2A).A maternally transmitted 4.8 kbp deletion was detected that overlapped with the exons 2-9 of the BRCA2 gene.Performing PCR-based Sanger sequencing, we could confirm the presence of a 4.7 kbp deletion in chr13:32318724-32 323 440 bp, leading to deletion of the third exon of the BRCA2 gene (BRCA2 Δex3 , ENST00000380152) (Figure 2B).
Variants leading to BRCA2 exon 3 deletions have been reported in individuals with breast and ovarian cancer. 18,19Although this deletion does not alter the reading frame, a part of the highly conserved N-terminal domain is lost that is important for the interaction with PALB2 20 and DNA repair.Next, we performed a semiquantitative PCR to check whether both the BRCA2 Δex3 and BRCA2 wt transcripts were expressed.While material was available from the child and the father, unfortunately, no material was available of the mother or the initially diagnosed leukemia of the patient to be examined for down-stream analysis.Interestingly, the child exhibited a higher expression of BRCA2 Δex3 than BRCA2 wt compared to the father and the control samples with predominant expression of BRCA2 wt transcript (Figure 2C).There has been some debate about the pathogenic effect of exon 3 deletion, as the BRCA2 Δex3 transcript is constitutively expressed in tissue samples from apparently healthy individuals.However, Muller et al reported an allelic imbalance in favor of the BRCA2 Δex3 mutant allele 18 and Caputo et al proposed that only variants producing transcripts lacking exon 3 should be classified as pathogenic. 19In line with this, we observed an allelic imbalance toward the BRCA2 Δex3 transcript in our patient.The maternally transmitted deletion segregated within the cancer history in the maternal lineage.
The grandmother (carrier status unknown) was diagnosed with breast cancer at the age of 32, and the patient developed an ETV6::RUNX1 rearranged BCP-ALL at the age of 12, whereas the patients' mother T A B L E 1 Structural variants and affected genes potentially playing a role in cancer predisposition of two pediatric patients.

Case P26 P28
Tumor Note: Overlap common dbVAR: Indicating whether the region in which the SV was observed overlaps with a known copy number alteration using the common dbVar database (>1% in the population) as reference.had not yet developed any cancer at the age of 40.Although, BRCA2 alterations have not been described to increase the susceptibility to ALL as yet, the internal BRCA2 deletion might be a possible risk factor.

| Identification of novel candidate cancer predisposition genes
98.8% of all SVs identified by OGM in our study did not overlap with a known CPG.Performing a comprehensive literature search and further analysis, we identified three interesting SVs potentially targeting novel CPGs.
In patient P16, a girl who synchronously developed a medulloblastoma and an astrocytoma at the age of 3, we identified a maternally transmitted 180 kbp tandem duplication in 8p11.22,verified by SNP array and coverage-based WES data (Figure S2A-C).The duplicated region encompasses the ADAM9 and ADAM32 genes, which belong to the "disintegrin and metalloprotease" (ADAM) family.
ADAM9 is overexpressed in glioblastoma, where it promotes invasiveness and proliferation. 21,22To test the effect on ADAM9 and ADAM32 expression or if the duplication leads to a potential fusion transcript, we performed whole transcriptome analysis of the patient and the noncarrier father.We did not detect an ADAM32::ADAM9 fusion transcript or altered expression of both genes in the affected child (Figure S2D).In addition, we identified compound heterozygous SVs in the RAP1B gene, in P10, a boy with B-ALL.The patient carried a maternally transmitted 9.3 kbp insertion and a paternally transmitted 10.8 kbp deletion, both affecting regulatory regions of RAP1B gene locus (Figure S3A,B).RAP1B encodes for a protein of the RAS family of small GTP binding proteins with mitogenic and oncogenic properties. 23,24Activating missense variants in RAP1B have been described in three patients with syndromic thrombocytopenia and multiple malformations, including microcephaly and learning difficulties. 25,26In line with these findings, P10 had mild global developmental delay and a growth disorder in addition to his B-ALL.However, the analysis of RAP1B mRNA expression in the PBMCs of the patient revealed no altered expression in comparison to the father or the controls (Figure S3C).  1, Figure 2D).The duplication was present in the germline and the corresponding tumor of P28 (Figure S4A).We verified the copy number gain in the child using coverage-based information from WES and could show that this gain was not present in the WES data of the parents, indicating that the SV occurred de novo (Figure S4B).Using SNP array, we could show that the duplication is located in a region of copy number neutral loss of heterozygosity in 17p13.3-p13.1 (Figure S4C).Due to the consanguinity of the parents (first-degree cousins), 11.2% of the patient's autosomal genome is homozygous (Figure S4D).The duplication affected five genes, including RPA1, which has essential functions in DNA replication, repair, and recombination. 27It is associated with autosomal dominant inherited telomere-related pulmonary fibrosis and/or bone marrow failure-6 syndrome (PFBMFT6, MIM:619767).The PFBMFT6 syndrome belongs to the short telomere syndromes (STS) with variable symptoms, genetic heterogeneity and variable expressivity.Analysis of the RPA1 mRNA level in PBMCs and fibroblasts of the child revealed an increased expression in comparison to parents and controls, which points toward the duplication having a gain-of-function (GoF) effect (Figure 2E).This is in line with reports by Sharma et al, who described GoF RPA1 missense variants in patients with short telomeres and varying clinical features of STS. 28 further assessed RPA1 protein expression in the patient's fibroblasts in comparison to two age/gender-matched fibroblast samples from children without RPA1 alterations, which revealed similar RPA1 expression (Figure S5A).We screened for a potential telomeropathy by performing TeloTAGGG Assay on fibroblasts of P28 carrying the Additionally, it has been reported that RPA1 duplication induces elevated sensitivity to DNA damage. 29Therefore, we also checked for an underlying DNA repair defect in P28 by performing cell viability assays to analyze sensitivity to genotoxic agents (Figure S5D,E).We used the chemotherapeutic agent cisplatin, inducing DNA crosslinking and DNA double-strand breaks (DSBs), and the ribonucleotide reductase inhibitor hydroxyurea (HU) that depletes nucleotides and induces replication stress leading to subsequent DSBs.Fibroblasts of P28 carrying the RPA1 duplication exhibited a significant increase in sensitivity to cisplatin und HU compared to wild-type age/gendermatched controls.Thus, in line with the published findings, 29 P28 fibroblasts with RPA1 duplication showed elevated sensitivity to DNA damage.The extent to which the described genotype might be associated with the origin of the KMT2A+ infant ALL and if the patient, who remains very young, may develop shortened telomeres needs to be monitored in the future.

| CONCLUSION
By applying OGM, we comprehensively characterized the SV landscape of 34 pediatric cancer patients who were clinically highly suspicious for an underlying CPS and had a nonconclusive WES result.We identified three de novo SVs among 18 child-parent trios and two hot germline alterations affecting the known CPG BRCA2 and the potential new CPG candidate RPA1 in two respective leukemia patients.
Our study highlights the suitability of OGM in a clinical setting to detect potentially predisposing SVs in suspicious pediatric cases.
However, there remains the necessity for applying additional sequencing-based methods to determine the exact breakpoint boundaries when interpreting OGM reported SVs.We expect that extended germline screens, improvement of the OGM reference database and integration of the telomere-to-telomere reference genome will enhance SV detection and interpretation of disease-causing events.
Further studies will likely use a combination of OGM and long-read sequencing technologies to overcome the current limitations of OGM resolution (<500 bp) to cover the full spectrum of genetic predisposition in children with cancer.

F I G U R E 1
Structural variants detected by OGM in 34 pediatric patients suspicious for an underlying cancer predisposition syndrome.(A) Summary of patients' tumor diagnoses and status of clinical criteria.Each line represents a clinical feature and each column represents a patient.(B) Overview of structural variants identified using OGM.The dot plot represents the number of germline SVs per type, with every dot representing a patient.The red horizontal lines represent the median.(C) The violin plot depicts the length distribution of the identified SVs.The dashed line represents the 50 kbp threshold that is a lower limit of copy number variant length detectable by conventional DNA array-based methods.(D) Analysis of 18 child-parent trios using OGM revealed that 3/1007 detected SVs were de novo variants.

F I G U R E 2
Structural variants affecting the known CPG BRCA2 and the potential candidate CPG RPA1 in two pediatric cancer patients.(A) Germline deletion detected in the cancer predisposition gene BRCA2 in P26.Genome maps (green bar above reference and blue bar below patient) indicating the presence of a 4.8 kbp deletion in 13q13.1.Lower part of the figure depicts the location of the deleted region, highlighted in red, which is located within the BRCA2 gene.(B) Sanger sequencing of the mRNA of the patient confirms the deletion of exon 3 in BRCA2 (ENST00000380152).(C) Upper panel: Semiquantitative PCR of BRCA2 wt and BRCA2 Δex3 transcript RNA extracted from the PBMCs of P26 carrying the 4.8 kbp deletion in comparison to the father (P26_father) and four control samples (PBMC ctrl 1-4).Lower panel: GUSB was used as control.(D) Genome map of duplicated region affecting the RPA1 gene locus (17p13.3) in P28 (green bar represents reference map and blue bars the patient map) and the location within the genome using UCSC genome browser (duplication highlighted in blue).(E) Analysis of RPA1 mRNA expression levels reveals higher expression in PBMCs and fibroblasts of P28 in comparison to the parents and controls.The controls were separated based on tissue of origin, including fibroblasts (blue) and PBMC (gray) control samples.The dashed lines represent the mean expression levels of the fibroblast controls (blue) and PBMC controls (gray).

Conclusively, additional experiments
including corresponding tumor analysis are required to clarify whether the identified germline SVs affecting ADAM9 and RAP1B have a functional influence on the development of the patient's tumors.The most interesting SV was a 166 kbp duplication in 17p13.3 in P28, an infant diagnosed with KMT2A-rearranged B-cell-precursor acute lymphoblastic leukemia (BCP-ALL) at the age of 4 months, who relapsed 6 months after initial diagnosis (Table

RPA1
duplication in comparison to age/gender-matched control fibroblasts (FigureS5B,C).However, we even detected a mild increase of mean telomere restriction fragment length in the RPA1 duplicated cells compared to the controls (P28 = 13.3 kb vs Ctrl1 = 10.4 kb; Ctrl2 = 11.6 kb, respectively).