Examination of Genetic Susceptibility in Radiation-Associated Meningioma

Abstract

Previous epidemiological studies have demonstrated elevated susceptibility to ionizing radiation in some families, thus suggesting the presence of genetic components that conferred increased rate of radiation-associated meningioma (RAM). In this study, we exome-sequenced and investigated the segregation pattern of rare deleterious variants in 11 RAM pedigrees. In addition, we performed a rare-variant association analysis in 92 unrelated familial cases of RAM that were ancestry-matched with 88 meningioma-free controls. In the pedigree analysis, we found that each family carried mostly a unique set of rare deleterious variants. A follow-up pathway analysis of the union of the genes that segregated within each of the 11 pedigrees identified a single statistically significant (q value = 7.90E-04) “ECM receptor interaction” set. In the case-control association analysis, we observed no statistically significant variants or genes after multiple testing correction; however, examination of ontological categories of the genes that associated with RAM at nominal P values <0.01 identified biologically relevant pathways such as DNA repair, cell cycle and apoptosis. These results suggest that it is unlikely that a small number of highly penetrant genes are involved in the pathogenesis of RAM. Substantially larger studies are needed to identify genetic risk variants and genes in RAM.

Editor’s note.

The online version of this article (DOI: https://doi.org/10.1667/RADE-21-00035.1) contains supplementary information that is available to all authorized users.

INTRODUCTION

Meningiomas are tumors of the central nervous system (CNS). They are relatively rare compared to the neoplasms of other organs and tissues; however, in the CNS, these neoplasms are the most prevalent, constituting ~1/3 of all brain and spinal cord tumors (1). The majority of meningiomas are slowly growing WHO grade I benign neoplasms, but a small proportion of them occasionally progress to grade II and III tumors (1). Familial aggregation of meningiomas including cases of neurofibromatosis type 2 (OMIM #101000) and individuals with SMARCB1 (2), SUFU (3) or SMARCE1 (4, 5) germline pathogenic variants cases is rarely observed. The etiology of meningioma is largely unknown; however, ionizing radiation is a well-established risk factor for its formation (6). Effects of radiation on meningioma tumorigenesis have been primarily investigated in studies of children undergoing radiotherapy for primary brain tumors; in evaluation of diagnostic procedures involving exposure of head and neck to X rays; as well as in clinical examinations of atomic bomb survivors in Japan (6).

The tinea capitis study (TCS), from which the study population of the current analysis was derived, was among the first ones that demonstrated an association between radiation and brain tumors (7). The TCS was initiated in Israel in the 1960s and followed for >40 years (7-9). Between the late 1940s and early 1960s, >20,000 Israeli citizens, mostly children of North African and Middle Eastern descent, who immigrated to Israel with their families, were subjected to a series of X-ray exposures to the head and neck areas as a treatment for tinea capitis, a fungal infection of the skin. Dosimetry estimations in various organs have shown that the mean dose absorbed by the brain was 1.5 Gy (range 0.98–6.00) (8). Tinea capitis is considered highly contagious, and if a single person in a family was diagnosed with the disease, many of their family members were also infected or considered to be infected and, therefore, were subjected to radiation treatment as well. Since the TCS population was distinguished by pedigrees with relatively high number of siblings (mean number of siblings in the families was 3.9 ± 1.5), this practice resulted in numerous multiplex pedigrees with both irradiated and unirradiated family members (10). In the early 1970s, it was found that patients who received X-ray treatment had a higher rate of head and neck tumors, specifically of the brain and thyroid gland, compared to untreated population and sibling controls (7). In 1988, it was observed that after irradiation the relative risk for meningioma increased 9.5-fold (95% CI = 3.5–25.7) (8), and in 2005 it was quantified that excessive relative risk per Gy was 4.63 (95% CI = 2.43–9.12) (9).

The first evidence of the involvement of the genetic component in these families was reported by Flint-Richter and Sadetzki (10). The authors investigated 525 families in relationship to their radiation exposure status and the development of meningiomas. These families included: 1. 160 pedigrees with radiation-associated meningiomas (RAM); 2. 85 pedigrees with meningiomas and no history of radiation exposure; 3. 145 pedigrees without meningiomas and with X-ray exposure; and 4. 135 pedigrees without either meningiomas or a history of radiation exposure. In the RAM group, they found that 11% of families had additional first-degree relatives with meningiomas, while among the three other groups they observed only 1% of families that included additional first-degree members with meningiomas (P < 0.0001). The authors concluded that there were likely genetic components shared by the families that conferred increased susceptibility to radiation-induced meningioma (10).

The same group later attempted to identify the causative genetic elements by genome-wide single nucleotide polymorphism (SNP) linkage disequilibrium mapping approach. They genotyped members of 15 pedigrees affected with RAM with high-density SNP-arrays and then tested SNPs and haplotypes for the association with RAM (11). After adjustment for multiple testing, none of the SNPs or haplotypes were found significantly associated with RAM; however, several chromosomal loci, e.g., 18q21.1 and 10q21.3, comprising several biologically relevant genes (PIAS2, KATNAL2, TCEB3C, TCEB3CL and CTNNA3) were associated with RAM at the nominal P values in the range of 10⁻⁴ to 10⁻⁵. The authors concluded that the genetic susceptibility to RAM was likely determined by multiple risk loci (11).

Interestingly, a recent study that investigated risks of secondary neoplasms in association with pathogenic germline variants in 127 DNA repair genes among survivors of childhood cancers who received radiation and chemotherapies, found no significant association for subsequent meningioma development (adjusted P values range 0.48–0.82) (12). Another recent study that examined association of rare germline variants in 476 DNA damage response and radiation sensitivity syndrome genes with subsequent neoplasm risk in the Childhood Cancer Survivor Study, found virtually identical rates of meningioma formation (the difference was not statistically significant) in the radiotherapy patients with or without variants in these genes (13).

We hypothesized that inherited genetic factors may confer enhanced susceptibility to ionizing radiation in select individuals, which would result in elevated frequency of meningioma occurrence among these individuals. In this exploratory study, we exome-sequenced 11 RAM pedigrees and a set of unrelated 92 RAM-affected individuals matched with 88 unaffected controls to examine the relative burden of rare deleterious variants in RAM-affected patients and to identify genetic risk factors predisposing to the development of meningioma after irradiation (Fig. 1).

FIG. 1.

Schematic representation of analyses and sample sets used in the study.

METHODS

Recruitment of the Study Participants

The tinea capitis studies were approved by the Chaim Sheba Medical Center Review Board Committee. Collection of blood samples and clinicopathological information from subjects was undertaken with informed consent and relevant ethical review board approval in accordance with the tenets of the Declaration of Helsinki. We analyzed exome-wide germline DNA variation in 11 radiation-associated meningioma pedigrees, and in a set of unrelated 92 RAM-affected individuals matched with 88 unaffected controls. The set of 92 RAM-affected individuals included one affected person randomly selected from each of 11 pedigrees. The set of 88 controls included ethnicity-matched individuals who were irradiated with a similar dose of X rays, and who did not develop meningiomas.

Exome Sequencing

Exome sequencing (ES) was performed at the NCI Cancer Genomics Research Laboratory as previously described (14).

Population Structure and Principal Component Analysis

For ancestry estimation, the SNP weights software (15) with a reference panel of Ashkenazi Jews, South European and Northwest European populations was used (15). To limit population stratification, we used the ancestry estimation output from both affected and unaffected participants that were subjected to principal component analysis (PCA) using PLINK v1.90b3.32 (https://www.cog-genomics.org/plink/).

Variant Annotation

Variants were annotated using ANNOVAR for RefGene (https://bit.ly/3tSvY6E). In silico deleteriousness prediction was done by using dbNSFP (16) and REVEL (17). ClinVar calls were made based on clinical laboratories meeting minimum requirements for data sharing to support quality assurance (https://bit.ly/3Lqvs5v), archive update of January 2021. Variant annotation with InterVar (18) was performed by using Python v.0.1.7.

Variant Filtering and Tier Assignment

Variants were removed if flagged by CScoreFilter. Three variant callers were used: HaplotypeCaller, Freebayes and Unified Genotyper. Variants were accepted if called by the Haplotype caller and at least one of the remaining two, Freebayes or Unified Genotyper. The genotypes of variants with the read depth <2 and the genotype quality <20 were masked (e.g., converted to “./.” values). In addition, all non-coding variants (except for canonical splice-sites) and variants with population frequency (gnomAD_exome_non-TCGA, ESP, 1000 genomes) >1% were filtered out. All variants with read depth less than 2 in more than 50% of samples and ABHet values above 0.8 or below 0.2 were filtered out as well. Variants were classified as pathogenic (P), likely pathogenic (LP), variant of unknown significance (VUS), likely benign (LB), or benign (B) using ClinVar calls. Variants without ClinVar classification, were classified by InterVar into the same five classes. P/LP variants classified by ClinVar or InterVar constituted tier 1. All remaining loss-of-function (LOF) variants (frameshifting deletions or insertions, nonsense, start-loss and canonical splice-site disruptions) were grouped in tier 2. In-frame insertions/deletions, stop-loss and missense variants with CADD_phred-score>25, REVEL_score >0.5 and metaSVM_score >0.825 were grouped in tier 3. All remaining variants (non-tier 1-3) were analyzed for their potential involvement in aberrant splicing using spliceAI software (19). Variants with splice delta score>0.2 were classified as tier 4 variants. Tier 1-4 variants were considered deleterious, all remaining variants – non-deleterious.

Variant Segregation Pattern Analysis in Pedigrees

Eleven pedigrees that included 1–4 affected individuals were analyzed. Tier 1–4 variants present in all affected individuals within each pedigree were further considered.

Analyses in the Set of Unrelated Individuals

Analyses were done on a variant-level (Fisher’s exact tests, fisher.test function in R) and on a gene-level [association tests KBAC (20), SKAT (21) and SKAT-O (22)]. For the variant-level analysis tier 1–4 variants were used and for the gene-level analysis we used all rare variants regardless of their predicted deleteriousness.

Pathway and Ontological Analysis of Genes

Pathway and ontological analyses of genes harboring segregating variants was performed in Enrichr (https://maayanlab.cloud/Enrichr/). Gene lists were analyzed in BioPlanet 2019, WikiPathways 2021 Human, KEGG 2021 Human, MSigDB Hallmark 2020, and GO Biological Process 2021 databases.

RESULTS

Population Stratification and Matching

Principal component analysis demonstrated that cases and controls were well matched. Three main clusters observed on the plot correspond to three sub-populations of Iraqis, Moroccans and Tunisians (Fig. 2). A few Moroccan samples present in the mainly Tunisian cluster likely reflect close geographical proximity of these two countries on the North African Mediterranean coast.

FIG. 2.

Principal component analysis of 92 RAM cases and 88 controls. PCA was performed using ancestry estimation output of principal component 1 (PC1) and principal component 2 (PC2) to compare RAM cases and controls using PLINK. Cases are shown with red circles and controls are shown with green circles. The country of origin (Iraq, Morocco or Tunisia) is shown above each circle to confirm homogeneity of each cluster.

Variant Segregation Pattern in Pedigrees

First, we analyzed segregation of deleterious variants (tiers 1–4) in 11 families. Three pedigrees (P2, P6 and P10) included a single RAM-affected individual each, six pedigrees (P3, P4, P5, P8, P9 and P11) included two affected individuals, and two pedigrees (P7 and P1) included three and four RAM-affected individuals, respectively. The number of deleterious variants found in all affected individuals within each family ranged from four (pedigree P11) to 57 (single individual pedigree P6). In total, there were 28 tier 1 (Table 1) and 304 tier 1–4 segregating deleterious variants in 11 families (Supplementary Table S1; https://doi.org/10.1667/RADE-21-00035.1.S1). We observed a modest overlap between pedigrees on a variant- and gene-level: there were 17 variants shared by two, three or four pedigrees (Supplementary Table S2 https://doi.org/10.1667/RADE-21-00035.1.S2). A frameshifting insertion in MTCH2 (rs138027870) was observed segregating among all affected individuals in four pedigrees and another MTCH2 frameshifting insertion was observed in three families; however, these variants were also found at the comparable frequency in RAM-free controls (data not shown) and, therefore, are unlikely to play any role in RAM etiology.

TABLE 1.

Pathogenic and Likely Pathogenic Variants Segregating in RAM Pedigrees

Gene symbol	Chr	Position (hg19)	ID	Reference allele	Variant allele	Variant class*	Classified by	Variant type	Allele frequency in population (AF_popmax gnomAD)	MetaSVM rank score	REVEL score	CADD phred score	Pedigree ID
GNRHR	4	68619737	rs104893836	T	C	P	ClinVar	nonsynonymous	0.0042	0.744	0.606	23.2	P1
CLIP1	12	122825299	rs144240398	C	A	P	InterVar	splicing	0.0007	-	-	27.4	P1
MEFV	16	3293407	rs61752717	T	C	P	ClinVar	nonsynonymous	0.0005	0.493	0.406	0.011	P2
PKLR	1	155261709	rs116100695	G	A	LP	ClinVar	nonsynonymous	0.0037	0.978	0.914	34	P2
FANCA	16	89837021	-	A	AC	LP	InterVar	frameshift	-	-	-	-	P2
PIGV	1	27121404	-	GC	G	LP	InterVar	frameshift	0.00002891	-	-	-	P2
STX7	6	132781989	rs773669589	G	A	LP	InterVar	nonsynonymous	0.00006168	0.648	0.253	25.1	P2
MUTYH	1	45797228	rs36053993	C	T	P	ClinVar	nonsynonymous	0.0049	0.85	0.954	29.4	P3
FDXR	17	72860471	-	T	C	P	InterVar	splicing	0.000009819	-	-	12.03	P4
ITPR2	12	26878620	-	C	G	LP	InterVar	nonsynonymous	-	0.926	0.565	17.87	P4
STS	X	7243435	rs371806193	G	T	LP	InterVar	nonsynonymous	0.00002446	0.992	0.907	25.8	P4
SYNE1	6	152551743	-	A	AGTCTCTT	LP	InterVar	frameshift	-	-	-	-	P4
CPT2	1	53675705	rs121918528	A	G	LP	ClinVar	nonsynonymous	0.0003	0.996	0.981	25.6	P5
C8B	1	57406638	rs41286844	G	A	P	ClinVar	stopgain	0.0019	-	-	41	P6
MYOCD	17	12647716	-	C	T	LP	InterVar	stopgain	-	-	-	37	P6
SLFN14	17	33879855	-	T	G	LP	InterVar	nonsynonymous	0.0002	0.921	0.732	27.1	P6
SLFN14	17	33879855	-	T	G	LP	InterVar	nonsynonymous	0.0002	0.921	0.732	27.1	P7
CLIP1	12	122825388	rs147792689	G	T	LP	InterVar	stopgain	0.00001761	-	-	36	P8
DYNC2H1	11	103128448	rs181011657	C	T	P	ClinVar	stopgain	0.0007	-	-	58	P9
GNRHR	4	68619737	rs104893836	T	C	P	ClinVar	nonsynonymous	0.0042	0.744	0.606	23.2	P9
NPHP3	3	132408107	rs751527253	CCT	C	P	ClinVar	synonymous	0.0005	-	-	-	P9
TYR	11	88911261	rs61753180	G	A	P	ClinVar	nonsynonymous	0.0011	0.962	0.958	27.3	P9
MEFV	16	3293407	rs61752717	T	C	P	ClinVar	nonsynonymous	0.0005	0.493	0.406	0.011	P10
TYR	11	88911770	rs63159160	C	T	LP	ClinVar	nonsynonymous	0.0004	0.921	0.588	25.1	P10
VWF	12	6143978	rs41276738	C	T	P	ClinVar	nonsynonymous	0.0054	0.722	0.487	35	P10
BBS9	7	33376156	-	C	T	LP	InterVar	stopgain	0.00001767	-	-	36	P10
NQO1	16	69748897	rs748374631	G	GT	LP	InterVar	stopgain	0.00005437	-	-	-	P10
STRC	15	43902548	rs576724182	G	A	P	InterVar	stopgain	0.0011	-	-	37	P10

^*Abbreviations: P = pathogenic; LP = likely pathogenic.

Pathway analysis of the union of genes segregating in each of the 11 pedigrees (n = 278 genes) revealed a single significant association with “Extracellular Matrix (ECM) receptor interaction” pathway (KEGG 2021 human pathway database; nominal P = 3.73E-06; adjusted for multiple testing q value = 7.90E-04; odds ratio = 8.32) and included 9/88 genes (FRAS1, VWF, TNN, ITGA2B, ITGA1, LAMB4, ITGA11, COL6A5, HSPG2). Analysis of Gene Ontology Biological processes database did not identify any statistically significant associations; however, it demonstrated that tier 1–4 variants resided in genes involved in several biologically relevant processes (Supplementary Table S3; https://doi.org/10.1667/RADE-21-00035.1.S3). We found that at least one of DNA repair, recombination, and replication genes, such as FANCA, MSH6, MUTYH, PMS1, POLR2F, TEP1 and TIMELESS, was present in 4/11 pedigrees. Similarly, we found genes involved in apoptosis, such as ALOX15B, MTCH2, PRODH and SOX7 in 5/11 families, and genes involved in cell cycle and mitosis, such as ALOX15B, MYOCD, RNF167, SAPCD2 and SPTBN1, in 4/11 families (Supplementary Table S3; https://doi.org/10.1667/RADE-21-00035.1.S3).

Variant- and Gene-Level Association Analyses in the Set of Unrelated RAM Individuals and Matched Controls

We next analyzed a set that included 92 unrelated RAM cases with ethnicity-matched 88 tumor-free controls. For this analysis an input that contained all rare coding variants was used. The rationale for using all variants in the input was that it is difficult to accurately define/predict true pathogenicity/deleteriousness for most of the variants, and that many rare-variant association tests, including those used in this study, are more sensitive to the omission of true causative variants than to the presence of neutral ones.

Variant-level Association Analysis

None of the variants was significant at the q < 0.05 after multiple testing correction and only 16 variants were associated with RAM at the nominal P < 0.01 (Supplementary Table S4A; https://doi.org/10.1667/RADE-21-00035.1.S4). Most of the variants resided in genes with unknown biological function or which function is not highly relevant to the etiology of RAM; however, we identified a variant in VCPIP1 (nominal P = 0.007), a gene involved in DNA repair following phosphorylation by ATM or ATR (https://www.uniprot.org/uniprot/Q96JH7). We also identified a variant in CEP295 (nominal P = 0.009), a gene involved in centriole biogenesis and centriole-to-centrosome conversion during mitotic progression (https://www.uniprot.org/uniprot/Q9C0D2) The variant in CEP295 appears to be protective, i.e., we observed nine variant alleles in controls and only one variant in cases (Supplementary Table S4A; https://doi.org/10.1667/RADE-21-00035.1.S4).

Gene-level Association Analyses

Finally, we performed rare-variant association analyses at the gene level using KBAC, SKAT and SKAT-O methods. We observed 174 genes associated with RAM at a nominal P < 0.01 (Supplementary Table S4B; https://doi.org/10.1667/RADE-21-00035.1.S4). None of these genes was significant at the q < 0.05 after correction for multiple testing. A follow-up pathway enrichment analysis revealed no pathways or ontological categories significantly associated with these genes; however, we observed a number of genes from DNA repair and damage response (e.g., ABL1, DNMT1, EP300, KDM2A, TP73, WDHD1), apoptosis (e.g., ABL1, ARHGEF26, CARD14, EP300, HOXA5, TP73, TNFRSF8), cell cycle (e.g., CNOT3, EP300, KIF18A, KIF23, TP73) and other relevant ontological categories (Supplementary Table S5; https://doi.org/10.1667/RADE-21-00035.1.S5). We note that several genes with versatile cellular roles (e.g., TP73, EP300, ABL1, CNOT3) are found in multiple signaling hubs and protein complexes in the cell and are likely to play an important role in RAM etiology.

DISCUSSION

In this exploratory study of association of rare genetic variants with RAM pathogenesis, we investigated the segregation pattern of rare deleterious variants in 11 RAM pedigrees (8/11 were multiplex) and analyzed a matched case-control dataset (N = 180) with rare-variant association tests. In the pedigree analysis, we found that each family carried mostly a unique set of rare deleterious variants. In the case-control analysis of 92 unrelated RAM patients that were ethnicity matched with 88 meningioma-free controls we observed no variants or genes that were statistically significant after multiple testing correction. A follow-up pathway analysis of the union of genes that segregated within each of the 11 pedigrees, identified a single statistically significant (q value = 7.90E-04) “ECM receptor interaction” cellular process. Examination of ontological categories of the genes that associated with RAM in the case-control dataset at nominal P < 0.01 identified several genes with biologically relevant functions such as DNA repair, cell cycle and apoptosis. Larger discovery and validation studies are required for identification of RAM genetic risk factors.

In the KEGG 2021 human database of pathways and cellular processes, the ECM receptor interaction pathway includes 88 genes, of which nine were present in the union set of 278 genes segregating in each of the 11 RAM pedigrees. Deleterious variants in FRAS1, VWF, TNN, ITGA2B, ITGA1, LAMB4, ITGA11, COL6A5 and SDC2 segregated in RAM-affected individuals in 5/11 pedigrees. Various cellular activities such as adhesion, migration, differentiation, proliferation, and apoptosis are controlled by the interaction of ECM and transmembrane receptors (https://www.genome.jp/dbget-bin/www_bget?pathway:map04512). Interestingly, mouse homolog of FRAS1, a component of ECM, is transcribed in the early embryonal period in the animals’ basement membrane of the developing brain meninges and is hypothesized to interact with morphogens and growth factors in the specific regions of the meningeal basement membrane and to regulate cell fate decisions (23). Another RAM-segregating gene in the ECM-receptor interaction pathway, SDC2 (syndecan 2), is robustly expressed in human adult and fetal brain and is involved in dendritic spine development (24). Additionally, there is accumulating evidence that syndecan 2 is involved in regulating cell biology in a wide array of cancers (pancreatic, breast, colon, osteosarcoma, fibrosarcoma) via interaction with a complex network of pluripotent growth factors (25). Notwithstanding, currently there is no evidence that dysregulation of FRAS1, SDC2 or other ECM-receptor interacting macromolecules may be linked to RAM or other meningioma risk.

Exposure to ionizing radiation is one of the few well-established risk factors predisposing to development of meningiomas (26). In this study, we hypothesized that inherited genetic factors confer enhanced susceptibility to ionizing radiation in select individuals and result in elevated frequency of meningioma occurrence. Understandably, germline variation in the DNA repair genes, which provide the main defense mechanism against this environmental insult, was the principal focus of this study. In pedigrees, we observed several genes involved in multiple DNA repair/replication/recombination processes, including mismatch repair (MSH6, PMS1, MUTYH), cellular response to DNA damage stimulus (MSH6, TIMELESS, FANCA, MUTYH), transcription-coupled nucleotide-excision repair (POLR2F), nucleotide-excision repair (AQR), mitotic recombination (TEP1), DNA replication (TIMELESS, ORC3) among others. Cumulatively, variants in these genes segregated in 5/11 families. Additionally, in the case-control analysis of 180 samples, we identified several DNA repair/replication/recombination genes that associated with RAM at the nominal p-value<0.01, including ABL1 (DNA damage induced protein phosphorylation); TP73, CNOT3, and EP300 (DNA damage response and signal transduction by p53 class mediator); WDHD1, ABL1, TP73 (DNA repair); and KDM2A (double-strand break repair via nonhomologous end joining). Except for EP300, which is associated with Rubinstein-Taybi syndrome (OMIM#613684) and may be associated with development of meningiomas, no other genes involved in inherited syndromes associated with meningioma formation, e.g., BAP1 (OMIM# 614327), NF2 (OMIM#101000), PTEN (OMIM#158350), RECQL2 (OMIM# 277700), or SMARCB1/SMARCE1/SUFU/MN1/PDGFB (OMIM#607174), were observed in this study.

Comparison of our study results to previous investigations of the association of common variants with RAM did not identify genes in common. MLLT10 (27, 28) and RIC8A (29) that were associated with meningioma risk in genome-wide association studies were not observed in our results. Similarly, the candidate loci at PIAS2, KATNAL2, TCEB3C, TCEB3CL and CTNNA3 that were identified in genome-wide SNP linkage disequilibrium mapping study (11) were not present in this study’s results either. It should be noted, however, that the beforementioned studies were investigating common polymorphisms, while in this study we considered only rare (<1% population frequency) variants.

We also examined our data for the presence of pathogenic variants in ATM. Gilad and co-authors (30) reported elevated frequency of a nonsense variant (NM_000051, exon3:c.103C>T, p.R35X) in the ATM gene among Jewish ataxia-telangiectasia (A-T) families coming from Morocco and Tunisia. They observed 3/488 alleles (0.62%) in their study, while currently the global frequency of this variant in gnomAD is 0.002%. The c.103C>T is classified as pathogenic in ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/variation/3025/); and heterozygous carriers of pathogenic variants in ATM are considered at moderately elevated risk of cancer (31). Sensitivity to radiation among A-T carriers is not completely understood (31), but several in vitro studies demonstrated that cells heterozygous for pathogenic variants in ATM showed an increased level of chromosomal radiosensitivity compared to controls (32, 33). Approximately 80% of the participants in our case-control study and 8/11 of our pedigrees are of Moroccan or Tunisian descent. We found the ATM c.103C>T variant in a single RAM-affected heterozygous individual among 180 cases and controls, thus, allowing us to estimate this allele’s frequency as ~0.25%, which was similar to that reported by Gilad et al. (30). Notwithstanding, the sample size and the power of our study were insufficient to adequately interrogate the association of this variant with RAM.

In conclusion, we investigated the burden of rare deleterious variants in the germline DNA samples of patients who developed cranial meningiomas after radiation exposure they received as a treatment for tinea capitis. In pedigrees, segregating deleterious variants were mostly unique to each family. The pathway analysis identified a single statistically significant pathway of the ECM-receptor interaction, which included genes segregating in 5/11 families. Variant- and gene-level analyses of the case-control sample set (N = 180) revealed several biologically plausible candidates at the nominal P < 0.01; however, none of them were significant after correction for multiple testing. These results suggest that it is unlikely that a few highly penetrant genes are involved in the pathogenesis of RAM. However, since this study was focused on the protein-coding portion of the genome only, we cannot exclude a possibility of involvement of non-coding genomic elements. Substantially larger studies are needed to identify genetic risk variants/genes in RAM.