Challenges in reporting pathogenic/potentially pathogenic variants in 94 cancer predisposing genes - in pediatric patients screened with NGS panels

The benefit of reporting unsolicited findings in Next Generation Sequencing (NGS) related to cancer genes in children may have implications for family members, nevertheless, could also cause distress. We aimed to retrospectively investigate germline variants in 94 genes implicated in oncogenesis, in patients referred to NGS testing for various rare genetic diseases and reevaluate the utility of reporting different classes of pathogenicity. We used in silico prediction software to classify variants and conducted manual review to examine unsolicited findings frequencies in 145 children with rare diseases, that underwent sequencing - using a 4813 gene panel. The anonymized reanalysis revealed 18250 variants, of which 126 were considered after filtering. Six pathogenic variants (in BRCA1,BMPR1A,FANCA,FANCC,NBN genes) with cancer related phenotype and three unsolicited variants (in BRCA2,PALB2,RAD50 genes) were reported to patients. Additionally, three unsolicited variants in ATR, BLM (in two individuals), and FANCB genes presented potential cancer susceptibility, were not reported to patients. In retrospect, 4.8% (7/145) of individuals in our cohort had unsolicited NGS findings related to cancer. More efforts are needed to create an updatable consensus in reporting variants in cancer predisposing genes, especially for children. Consent process is crucial to inform of both value and risk of additional genetic information.

guideline for diagnostic NGS in 2016 stated that laboratories should have a clearly defined protocol for addressing unsolicited and secondary findings 6 .
Interpretation of the ACMG recommendation vary between different laboratories in United States 7,8 , while in Europe there is similar lack of consistency 5 . A recent study showed that some laboratories limit their reporting to findings that are relevant to the clinical question, while others report unsolicited findings to varying degrees 9 .
The benefit of reporting secondary findings for children are even more sensible, as they may have implications for the parents and other family members, nevertheless, they could cause distress 10 . Therefore, strategies are needed to ensure that the consent process provides appropriate information on both value and risk of additional genetic information.
In the recommended minimum list of genes selected by the ACMG committee 2 , on the basis of their medical actionability, 23 out of 56 are highly penetrant cancer susceptibility genes. The 2015 update, added two more genes related to cancer to the minimum list: SMAD4 (MIM 600993) and BMPR1A (MIM 601299) 1 . The disadvantage of a published minimum list is that it needs constant update, considering that cancer is one of the fields of medicine with the fastest development 11 .
In this context, we aimed to retrospectively investigate germline variants in 94 genes, causally implicated in oncogenesis, in patients referred for NGS testing for various rare genetic diseases, reevaluate them, and discuss the utility of reporting different classes of pathogenicity to the referring physician, families and patients.

Material and Method
The referrals for genetic testing were children presenting disorders with intellectual disability with or without congenital anomalies, neuro-muscular diseases, inborn errors of metabolism, hematologic diseases, immunodeficiencies, genodermatoses, skeletal dysplasia or endocrine diseases. The request for constitutional TruSightOne panel (Illumina) testing was made at the discretion of the referring clinical geneticist in the Timis Regional Centre of Medical Genetics, affiliated with "Louis Turcanu" Emergency Hospital for Children.
Sequencing analysis. Libraries were generated according to manufacturer's protocols using TruSightOne kits (Illumina Inc., San Diego, CA, USA) in the Center of Genomic Medicine from the University of Medicine and Pharmacy "Victor Babes" Timisoara. Targeted DNA sequencing was performed on TruSightOne library, targeting 4813 genes, and sequenced on a MiSeq(Illumina).
Only SNVs that were reported to patients and relevant for their referral were confirmed by Sanger sequencing, including all low-quality variants (GATK quality score below 500). Some variants were confirmed among familial segregation analysis or parental carrier.
Filtering strategy. Variants were subject to filtering, which excluded variants with allele frequency in gno-mAD(all) higher than 1% 17 , or higher than 5% in house frequency. Also, variants were excluded if reported as benign and likely benign in ClinVar or InterVar, and those with CADDphred score below 20. Higher CADD scores indicate that a variant is more likely to have deleterious effects. A scaled score of 10 or greater indicates a raw score in the top 10% of all possible reference genome SNVs, and a score of 20 or greater indicates a raw score in the top 1% 16 . Synonyms variants and those with a CADDphred score lower than 20 were excluded, unless reported as pathogenic or likely pathogenic in ClinVar. Data was checked for compound heterozygosity. All DNA sequencing results were manually reviewed by two clinical geneticists to prioritize variants and subsequent reporting of consensus variants. Ethics. Written informed consent was obtained from all participant's legal guardians, after the risks and benefits had been explained to the parent, caregiver or patient. The consent offered the possibility to opt yes or no for disclosure of secondary findings, unrelated to the referral condition. Patients could choose if they wished to have their samples and/or data stored for future research, both anonymously or not. Along the 94 selected genes, only, the variants that were reported to families had phenotype data available, all other variants were irreversibly anonymized. All experiments were performed in accordance with relevant guidelines and regulations. University of Medicine and Pharmacy Timisoara Ethics Committee approved this retrospective analysis of partially anonymized genomics data (no 02/22.01.2019).

Results
From 1034 individuals evaluated between 2015-2018, a number of 145 individuals (12.1%) had NGS solo analysis using the TruSightOne Illumina Panel. Of them, two caregivers opted out of being informed about unsolicited findings in informed consent process. By retrospectively reanalyzing anonymized sequencing data in 94 genes related to cancer, we identified 18250 variants that passed the quality control threshold. After filtering, a total of 126 variants were considered (shown in The heterozygous variant c.7273 C > T(p.Arg2425Ter) in PALB2 gene, was reported as likely pathogenic as a risk for breast cancer, in one child presenting with dystonia. Family history was negative in this case. Both parents were invited for genetic counseling, however, only the mother attended the meeting. She did not request family screening for the variant for the time being, but mentioned she will consider it.
The variant c.8331 + 1 G > A in BRCA2 was reported as likely pathogenic as a risk for breast cancer, in a boy presenting with thrombocytopenia. Family history was also negative. After genetic counseling with both parents, family screening was desired and carried out, demonstrating the variant's presence in the father and sister of the patient. The potential risks for people with BRCA2 pathogenic variants, considering the male gender and young age of the girl, were discussed with the family.
The variant c.3050 G > A (p.Trp1017Ter) in RAD50 in a boy presenting for myopathy. The patient had a sister with Hodgkin Lymphoma that deceased. The family did not desire carrier testing in the family at the time of genetic counseling.
Pathogenic and likely pathogenic variants (secondary findings) NOT reported to patients associated with susceptibility to cancer (shown in Table 2 and in Supplementary Table 1 in pink-with full annotation). One heterozygous variant c.93dupA p.(Leu32Thrfs) in autosomal dominant ATR gene (in one patient). One heterozygous variant c.3532 G > A p.(Gly1178Arg) in X linked recessive FANCB gene (in one patient). One heterozygous variant c.1642C > T p.(Gln548Ter) in autosomal recessive BLM gene (in two patients). Heterozygous variants in BLM and FANCB (in females) have the potential to be associated with increased susceptibility to cancer.
Additionally, two heterozygous variants in autosomal recessive genes associated with cancer predisposition, were identified in MUTYH gene and XPC gene (each in 1 patient). Matching compound heterozygous variants were not identified in these patients. Currently there is insufficient evidence of causality for heterozygous variants in these genes in relation to cancer.
Variants of unknown significance with potential for pathogenicity (secondary findings), NOT reported to patients (shown in Table 2 and in Supplementary Table 1 in blue -with full annotation). Eleven variants that met three of the ACMG criteria to classify pathogenic variants were identified in heterozygous state in BLM, BRCA2, CHEK2, DICER1, ERCC4, MLH1, RET, SDHB, TP53 genes (each in 1 patient), and in BRCA1 gene (in 2 patients). Except for ERCC4, all these genes have autosomal dominant transmission.
Nine patients had an association of VOUS in two different genes while three patients had associations of VOUS in 3 different genes. Two compound heterozygote variants c.679 G > A(p.Ala227Thr) and c.1248 C > A(p. Leu420Met) in MUTYH gene, were identified in one patient. Variants were not reported, as they were considered VOUS. Thirty-two VOUS were positioned in intronic, UTR3 and UTR5 regions.

Discussion
The ACMG 2013 guideline regarding incidental findings in children state that: "results from genetic testing of a child may have implications for the parents and other family members. Health care providers have an obligation to inform parents and the child, when appropriate, about these potential implications" 2 . The reasons provided by the Working Group for these highly debated recommendations were that: "at this moment in the evolution of clinical sequencing, an incidental finding relevant to adult disease that is discovered and reported through clinical (2020) 10 www.nature.com/scientificreports www.nature.com/scientificreports/ sequencing of a child may be the only way in which that variant will come to light for the parent…Tailoring the reporting of such information according to the age of the patient could place an unrealistic burden upon laboratories facing increasing volumes of clinical sequencing" 2 . www.nature.com/scientificreports www.nature.com/scientificreports/ Although of great medical interest, the discovery of cancer predisposing genes, can be overwhelming and might affect life quality for the people involved, affecting decisions related to family planning. Cancer development is complex and several factors influence their development: incomplete penetrance and variability of allele expression in cancer predisposing genes, copy number variants, effect of modifier genes, digenic or oligogenic inheritance pattern, age and gender related penetrance, epigenetic alterations, and environmental exposures (lifestyle) 20 . It is extremely important to counsel patients and families, so they understand that penetrance, expressivity and severity can vary tremendously, in and between families.
Granting we can now generate large amounts of sequence data, our ability to accurately interpret this information, is still limited, creating a significant increase in the numbers of VOUS 19 . Possibly the greatest worry is the likelihood of reporting a false positive unsolicited finding to a patient, due to its potential negative impact. A recent study showed that half of the laboratories did not report any unsolicited findings, while others reported only once the variant had been discussed within a board formed by an independent doctor, ethicist and lawyer 5 .
Likely pathogenic variants reported to patients in our cohort. BRCA2 variant identified in the patient was previously published and considered predisposing for breast cancer. The gene is included in the minimum gene list of ACMG recommendation. The parents understood this finding might imply an increased susceptibility to cancer and opted to screen the family members for this variant. The finding created some distress, especially in the waiting time before family screening results. After results of tests the family felt relieved, to some extent, that they have a prognosis and a course of action.
The heterozygous variant in PALB2 identified in our patient was previously published as deleterious in one patient with breast cancer 23 . PALB2 gene is considered to have an intermediate risk for breast cancer, conferring a 2-to 3-fold increased risk of breast cancer 24 and is not in the minimum gene list of ACMG recommendation. Nonetheless, is included in many breast cancer panels. Biallelic PALB2 pathogenic variants were showed to cause Fanconi anemia, similar to BRCA1 and BRCA2 25 . Family history is an important aspect to evaluate in this context, however in a recent study only 5 of 21 PALB2 mutation carriers had a family history of breast or ovarian cancer. Thus, many patients with pathogenic variants in predisposition genes may not be identified by a family history of cancer 23 . For this patient, only the mother came for counselling, although both parents were invited. The option to test the variant was offered and the mother mentioned it will be considered after discussion with her partner. Notably her main concern was the condition of the child, granting less importance to a cancer predisposing gene.
Pathogenic variants in RAD50 was shown to be associated with genomic instability assessed by cytogenetic analysis of peripheral blood T-lymphocytes 26 , suggesting an effect for RAD50 haploinsufficiency on genomic integrity and susceptibility to cancer. A larger, more recent study has characterized the gene with intermediate-risk breast cancer susceptibility 27 . In our patient the presence of this variant was considered possibly relevant for the sister's phenotype. The family did not wish to continue with variant screening. In counselling, it seemed that discussing the death of the sister was unsettling, suggesting this was a possible cause for their decision.

Likely pathogenic variants NOT reported to patients in our cohort. The heterozygous variant in
ATR gene in our cohort was not previously reported as deleterious, however it is a stopgain variant with deleterious in silico predictions, unreported in ExAc or GnomAD. ATR gene is not in the minimum gene list of ACMG recommendation, nonetheless, is involved in DNA-replication and repair. Pathogenic autosomal-recessive variants in ATR gene were reported in Seckel syndrome and recently it has translated into an autosomal-dominant inherited disease encompassing oropharyngeal cancer, skin telangiectases, and mild developmental anomalies of the hair, teeth, and nails 28 . In this case the benefit and risk of reporting are similar to that of PALB2 gene, however there is less evidence of pathogenicity and unidentified penetrance.
BLM gene was associated with autosomal recessive Bloom syndrome, while heterozygous status was associated to breast cancer susceptibility in several studies 29 . A metanalysis showed that BLM pathogenic variants were associated with a 2 to 5-fold increase in breast cancer 30 . However a longitudinal study showed that the standardized incidence rates for cancer were not higher than expected and thus heterozygous pathogenic variants carriers are not at increased risk for developing cancer 31 . The p.Gln548Ter variant in BLM gene was previously identified as a Slavic founder mutation 30 . The nonsense variant in our cohort was identified in 2 individuals. The benefit and risk of reporting to patient are similar to that of PALB2 gene.
The FANCB gene is not in the minimum ACMG recommended list, however, identifying X linked carries is relevant for future pregnancies. However, germline heterozygous variants in FANCB were associated with increased susceptibility for head and neck carcinoma 32,33 .
Reporting unsolicited variants to the family in NGS testing for a rare disease in children, might be the only opportunity to learn about a variant, which could become relevant at reproductive age and later into adult life. The parents could undergo cancer screening, if carrier status was demonstrated. However, the benefit cannot be fully estimated, nor the risk of creating distress when living under Damocles' sword 34 . Therefore, parents informed decision is crucial. Notably, the parent's understanding of the possibility to receive such unsettling news, could be difficult to comprehend 35 , considering the child is referred for genetic testing due to a rare disease that is usually already a major health issue. For the 3 cases where we reported unsolicited variants, the benefits and concerns of the families were different. This variability could be associated with educational status and familial context.
In our cohort, 12 individuals (8.2%) presented 11 variants classified as VOUS, however with a high likelihood for pathogenicity, gathering 3 ACMG criteria for pathogenic variants. Reporting such a variant is considered unethical due to the high risk of being false positive 2 . Nonetheless, the possibility that unsolicited variants classified as VOUS will be reanalyzed is unlikely due to the burden it would create for the laboratories.
The two compound heterozygote variants in MUTYH gene identified in one patient, could be missed in our current filtering strategy to identify variants causative for a rare disease. This finding raised the issue that a (2020) 10:223 | https://doi.org/10.1038/s41598-019-57080-9 www.nature.com/scientificreports www.nature.com/scientificreports/ secondary finding (especially in compound heterozygosity) has a high chance of being overlooked. As MUTYH gene is included in the minimum gene list of ACMG recommendation, a laboratory adhering to ACMG recommendations should have bioinformatic strategies to identify it.
Almost five percent (7/145) of individuals in our cohort had unsettling NGS findings (6 variants in 6 genes) related to cancer in retrospective analysis. Lower frequencies were reported by a recent study, where 1% of WES samples had reportable secondary findings in the cancer related genes recommended by ACMG 36 . Similarly, prevalence of pathogenic and likely pathogenic variants in 24 ACMG cancer genes in a family-based cancer research cohort was 1,2% and in cancer-free controls it was 0.8% 37 . However, both studies included only 24 gene recommended in the v.2 minimum list, whereas our study had a much-extended list. Consequently, the higher frequency in our study is due to variants in genes with more recent or less evidence for cancer predisposition. In Kim study 37 , the median review time estimated per-variant was 30 min. The authors highlighted how the analysis of secondary findings required database and literature review, which is a time-and labor-intensive process hindered by the difficulty of interpreting conflicting determinations 37 .
This work has led to change of practice in pre-test genetic counselling in our Center, including: (1) information about family history related to cancer and late onset disease to be enclosed in the details sent to the laboratory; and (2) extended explanation related to course of action after possible identification of unsolicited findings (example: the possibility for testing the variant in other members in the family and screening for the relevant conditions in family members identified at risk after family testing). Additionally, the consent process informs about the distress that these variants could create and the fact that cancer predisposing variants may imply a risk and are not equivalent to a diagnosis.
Considering that cancer is one of the fields of medicine with the fastest development 11 , extending the number of genes in secondary analysis, beyond the ones selected in the minimum recommended list, is needed. In this context, ClinGen (clinicalgenome.org) Hereditary Cancer Gene Curation Working Group focuses on curating cancer predisposition genes for their major associated syndromes. The work will provide a rapidly updatable  www.nature.com/scientificreports www.nature.com/scientificreports/ approach, compared to publishing guidelines. This ongoing effort will facilitate more informed utilization of genomic variants in clinical and cancer research.
Limitations. Our analysis does not include copy number variants (CNV) in cancer genes. In our cohort CNV were assessed. Although several patients also have SNP array, the results were not included in the manuscript. CNV from sequencing data was not assessed in our cohort. Although having many advantages, large sequencing panels still present some disadvantages compared to Sanger sequencing or smaller NGS panels, such as incomplete coverage of some genes or exons. These tests cannot exclude pathogenic variation. However, despite the limitations, NGS efficiently screen for most variants, supporting their clinical use.
As the authors suggest, the ACMG variant interpretation guideline is imperfect for classifying unsolicited findings 4 . It is important to consider the distinction between implicating a variant as pathogenic (causative for a disease) and a variant that may be predicted to be damaging to the protein but not necessarily implicated in a disease 4 . Authors also state that the use of the ACMG guideline may result in a larger proportion of variants being categorized as uncertain significance 4 . Despite its limitations the guideline is comprehensive and is used extensively for diagnostic and unsolicited findings.
One major limitation is lack of cancer family history information, which is a key component in identifying cancer-predisposition variants. However, Zang et al. 38 , showed in their cohort, that family history could not predict an underlying predisposition syndrome in most patients. Furthermore, some individuals with cancer, have de novo predisposing variants, whereas others inherit them with incomplete penetrance; where, the family history is likely to be negative.

conclusion
In this retrospective study we have identified 126 germline variants, in 94 genes causally implicated in oncogenesis, in patients referred for NGS testing for various rare genetic diseases. Seven individuals in our cohort (4.8%) had unsolicited findings related to cancer. Six pathogenic and likely pathogenic variants were identified in BRCA2, PALB2 and RAD50 genes were reported to families, while variants in BLM (in two individuals), FANCB, ATR genes were not reported. All consequently raised difficult ethical debate regarding their reporting. As only the BRCA2 gene was included in the 2015 ACMG minimum recommended list, we underline the need for constant update of this list. More efforts are needed to create an easily updatable consensus in reporting variants in cancer predisposing genes. Additionally, strategies are required to ensure that patients and physicians understand laboratories NGS reporting practices. Also, the consent process needs to inform of both value and risk of additional genetic information.