The germline variants in DNA repair genes in pediatric medulloblastoma: a challenge for current therapeutic strategies

The defects in DNA repair genes are potentially linked to development and response to therapy in medulloblastoma. Therefore the purpose of this study was to establish the spectrum and frequency of germline variants in selected DNA repair genes and their impact on response to chemotherapy in medulloblastoma patients. The following genes were investigated in 102 paediatric patients: MSH2 and RAD50 using targeted gene panel sequencing and NBN variants (p.I171V and p.K219fs*19) by Sanger sequencing. In three patients with presence of rare life-threatening adverse events (AE) and no detected variants in the analyzed genes, whole exome sequencing was performed. Based on combination of molecular and immunohistochemical evaluations tumors were divided into molecular subgroups. Presence of variants was tested for potential association with the occurrence of rare life-threatening AE and other clinical features. We have identified altogether six new potentially pathogenic variants in MSH2 (p.A733T and p.V606I), RAD50 (p.R1093*), FANCM (p.L694*), ERCC2 (p.R695C) and EXO1 (p.V738L), in addition to two known NBN variants. Five out of twelve patients with defects in either of MSH2, RAD50 and NBN genes suffered from rare life-threatening AE, more frequently than in control group (p = 0.0005). When all detected variants were taken into account, the majority of patients (8 out of 15) suffered from life-threatening toxicity during chemotherapy. Our results, based on the largest systematic study performed in a clinical setting, provide preliminary evidence for a link between defects in DNA repair genes and treatment related toxicity in children with medulloblastoma. The data suggest that patients with DNA repair gene variants could need special vigilance during and after courses of chemotherapy.


Background
Brain tumors represent the leading cause of childhood cancer mortality. The most common malignant brain tumor among them is medulloblastoma [1]. Although multimodality treatment regimens have substantially improved survival in this disease, up to 30-40% of patients with medulloblastoma still die of the disease. Detrimental effect of current treatment on long-term survivors is also observed [2]. Our understanding of the molecular background of pediatric brain tumors has expanded significantly over the past few years. The vast amount of genomic and molecular data generated recently has proved that medulloblastoma is not a single entity but is composed of at least four subtypes: Wingless (WNT), Sonic Hedgehog (SHH), Group 3 and Group 4 (non-WNT/SHH types), with distinct genetic and biological profiles as well as different course of disease requiring adequate therapeutic approaches [2][3][4][5][6]. Despite of improved understanding of the molecular basis of medulloblastoma, many cases still lack an obvious genetic driver [4,7,8]. The further research focused on additional potential mechanisms responsible for the development of this tumor may lead to identification of new susceptibility factors as well as new markers that predict response to therapeutic agents and provide prognostic information. So far, majority of driver mutations detected in medulloblastoma are of somatic character. Impact of germline genetic variability that may affect clinicopathologic presentation of this tumor have not been in-depth investigated yet [5,7,[9][10][11][12].
In our study we focused on evaluation of germline defects in genes that play a role in DNA repair pathway because of the following reasons. Firstly, DNA-repair deficiency is associated with cancer development and the key role of germline alterations in promoting tumorigenesis is highlighted by several cancer predisposition syndromes e.g. Li-Fraumeni, Fanconi anemia or Turcot syndrome, where occurrence of medulloblastoma has been recorded. Secondly, it is well known that germline defects may modulate the response to treatment since DNA-repair mechanisms make cells prone to the effects of DNA-damaging chemotherapy [13][14][15]. It is important to notice that majority of evidence about the impact of DNA-repair genes defects on toxicity in medulloblastoma comes from either description of single cases [16,17] or from mouse models and cell lines experiments [13,18] but not from systemic clinical based investigation. Therefore, all these data indicate that DNA repair genes are a promising targets possibly linked both to development of tumor and response to therapy in medulloblastoma.
Biallelic defects in NBN gene result in Nijmegen Breakage Syndrome (NBS; OMIM:251,260), while homozygous defects in MSH6, PMS2 or MLH1 genes are molecular cause of Constitutional Mismatch Repair Deficiency Syndrome (CMRDS; OMIM:276,300), hereditary disorder associated with increased risk of cancers including medulloblastoma [25]. Among other genes responsible for CMRDS is also MSH2 (ID:4436, MIM:609,309), one of the key factor of DNA mismatch repair system which recognizes and repairs mispaired or unpaired nucleotides resulted from DNA replication errors [25]. There is an evidence that germline MSH2 defects may predispose to primary early-onset CNS tumors, especially glioblastoma [26]. In addition, De Rosa et al. suggest that in some families with Turcot syndrome the coexistence of colorectal and childhood brain tumors may result from a complete MMR deficiency [27]. However, association between MSH2 defects and medulloblastoma was not evaluated yet.
A very similar phenotype to NBS was seen in patients with Nijmegen Breakage Syndrome-like Disorder (NBSLD -OMIM:613,078) caused by defects in the RAD50 gene (ID:10,111, MIM:604,040). This gene encodes the protein involved in DNA double-strand break repair, cell cycle checkpoint activation, telomere maintenance and meiotic recombination suggesting that molecular variants disrupting its function may lead to genome instability and carcinogenesis [28]. Furthermore, inactivation of proteins like RAD50 required for the homologous recombination machinery leads to defects in the nervous system development indicating that components of this system can play crucial role in development and progression of various neuro-oncological diseases [29]. The frequency of the molecular variants in RAD50 gene was, similarly to MSH2, not determined in medulloblastoma patients up to now.
Therefore the first purpose of this study was to establish the spectrum of germline defects in MSH2 and RAD50 genes, as well as frequency of two known NBN variants in 102 patients with medulloblastoma. In the next step we have evaluated the hypothesis that DNA repair genes may affect a response to therapy in medulloblastoma patients. We have found that alterations in a range of DNA repair genes are associated with occurrence of rare severe adverse effects during chemotherapy in patients. in Warsaw, Poland were investigated in this study. Based on a combination of genomic and immunohistochemical (IHC) analyses, patients were divided into molecular subgroups (see methods). Presence of metastases at diagnosis was classified according to Chang et al. [30]. The clinical characteristics of the study cohort is outlined in Table 1.

Patients and controls
To estimate the population frequency of detected MSH2 and RAD50 variants (independently of the data deposited in the public databases) the populationspecific control group was assembled. DNA samples from 300 healthy donors with negative cancer family history and sex matched to the patients' groups were collected. To exclude potential bias between adult and childhood population the control group consisted of donors age matched to the study group.

Methods
A total DNA was extracted from peripheral blood and tumors samples by use of the automatic magnetic bead-based (MagnaPure, Roche) and phenol/chloroform methods, respectively [31].
To evaluate the sequence of MSH2 and RAD50 genes in 102 patients with medulloblastoma targeted gene panel sequencing was used. The NBN c.511A>G and c.657_661del5 variant status was determined upon previously described conditions [19]. In three patients with severe adverse events after the chemotherapy but with no variants detected in MSH2, RAD50, and NBN genes whole exome sequencing (WES) was carried out to explore possible defects in other DNA repair genes. Sequencing Kit according to the manufacturer's instructions. The sequence reads were mapped to the haploid human reference genome (hg19) with Novoalign (Novocraft Technologies). SNVs and short insertions and deletions (indels) were called using GATK version 2.8. [32] Variant annotation was performed with Jannovar [33].

Targeted gene panel sequencing
Whole exome sequencing WES Library preparation was performed using Nextera Rapid Capture Exome kit (Illumina). The samples were run on ¼ of lane each on HiSeq 1500 using 2 × 75 bp pairedend reads. After initial processing by the CASAVA, the generated reads were aligned to the hg19 reference genome with Burrows-Wheeler Alignment Tool and further processed by Genome Analysis Toolkit [32]. Base quality score recalibration, indel realignment, duplicate removal and the SNP/INDEL calling were done as described [34]. The detected variants were annotated using Annovar5 [35].

Selection and validation of candidate variants
A way of prioritizing variants was based on three main filtration steps, as follows:  For all selected variants the amino acid position in functional domains and posttranslational modifications were verified using NCBI Protein [36] and Alamut-2.4-6 Software (Interactive Biosoftware; Rouen, France). Their contribution in carcinogenesis was verified in the Catalogue of Somatic Mutations (COSMIC database) [37], Human Gene Mutation Database (HGMD) [38], ClinVar database and Online Mendelian Inheritance in Man (OMIM). Correlation with the clinical features and the course of disease was assessed for each candidate variant.

Variant validation
Next generation sequencing results were validated using Sanger sequencing. Specific primers for PCR reactions are available upon request. PCR products were sequenced in 3130 Genetic Analyzer and evaluated with Sequencing Analysis Software v.5.4 (Applied Biosystems/ Life Technologies; Foster City, CA). The frequency of validated variants were evaluated in the matched control group in the same conditions. Additionally, prioritized variants in MSH2 and RAD50 were cross-checked with the parents' sequence data to identify inherited versus de novo changes. In four out of five carriers of MSH2, RAD50, FANCM or EXO1 gene variants tumour tissues were available and the presence of identified changes (p.V606I, p.R1093*, p.L694* and p.V738L) were verified.

Determination of molecular subgroups in medulloblastoma patients
Tumors included in the study were divided into the following molecular groups:  (Table 2). Due to the lack or low frequency of candidate variants in control groups, the estimation of cancer risk associated with the presence of identified variants was not possible. DNA testing of probands' parents confirmed parental origin of all selected variants. The presence of identified changes in tumor tissue had been proven for MSH2 p.V606I and RAD50 p.R1093* variants. For MSH2 p.A733T variant this analysis was not possible because of the lack of tumor tissue.
Within the prioritized variants only p.R1093* in RAD50 gene (rs121912628) was reported as pathogenic in ClinVar database (OMIM: 604,040.0001) and in the COSMIC list of variants that have previously been associated with cancer predisposition (COSM1060699) [37].

NBN c.511A>G and c.657_661del5 variants analysis
In 102 patients with medulloblastoma six carriers of NBN germline likely pathogenic variants (c.511A>G or c.657_661del5) were identified. Additional four NBN carriers reported in our previous study [19] were included in the analysis to increase a number of patients for assessment of clinical relevance of detected NBN variants.

Whole exome sequencing analysis
In three patients with severe treatment complications but no presence of MSH2, RAD50 and NBN candidate variants the exome sequence was analyzed. In all cases rare candidate variants in genes essential for DNA repair pathway were detected, including p.L694* in FANCM (ID 57697, MIM:609,644), p.R695C in ERCC2 (ID 2068, MIM:126,340) and p.V738L in EXO1 (ID 9156, MIM:606,063). None of identified variants was reported in 1000 Genomes Database and in matched control group. Two variants p.R695C in ERCC2 and p.V738L in EXO1 were found in ExAC Database with frequency 0.000000012 and 0.00000086, respectively. All detected variant were present in heterozygous state ( Table 2).

Characteristics of patients and tumors with molecular variants in DNA repair genes
Distribution of clinical and biological features in patients with molecular variants in essential for DNA repair pathway genes is presented in Table 3.
There were no differences between groups of patients with presence vs. absence of either of candidate variants in MSH2, RAD50 and NBN genes in terms of age (<3 years of age vs. ≥ 3 years, not significant, n.s), gender (n.s), LCA pathology (n.s) and presence of metastases (M2 M3 vs. M0 M1, n.s). Tumours belonged to WNT Group (one tumour), Group 3 (one tumour) and Group 4 (8 tumours). Three tumours were not analyzed by NanoString method due to the lack of RNA but they belonged to non-WNT/SHH type (Group 3 or 4). In the remaining three cases molecular type was not determined due to the lack of tumor material. It has been noticed that none of the patients with SHH tumors had MSH2, RAD50 or NBN variants (9 patients analyzed) but these data did not reach statistical significance when compared to other groups (n.s).
Four out of 12 patients with MSH2, RAD50 and NBN molecular variants did not complete treatment protocol because of reduction of the dose of drugs or delays above 100 days due to presence of various degrees of adverse events. These included 4 out of 6 patients who died, therefore we refrained from examination of survival rates. Among them one NBN c.657_661del5 carrier died due to secondary leukemia 48 months after diagnosis.
To assess the potential impact of molecular defects on the course of treatment we recorded that three out of 9 patients with presence of NBN variants and available clinical data, suffered from rare grade 4 adverse events during chemotherapy after the first course of treatment. These included central nervous system toxicity, pneumonia and colitis with gastrointestinal bleeding. For comparison, among 89 patients without molecular variants in none of NBN, MSH2 and RAD50 genes only three patients suffered from similar complications (enterocolitis grade 4, gastrointestinal bleeding grade 4 and central nervous system toxicity grade 4) despite an application of the same PPNG protocol (Additional file 2: Figure S1 A and B) and the difference was significant (p = 0.01). Moreover, two out of three patients with identified variants in MSH2 or RAD50 also displayed similar grade 4 complications, namely central nervous system toxicity and pneumonia. Therefore, 5 out of 12 patients with defects in either of MSH2, RAD50 and NBN genes suffered from rare grade 4 adverse events during chemotherapy and these combined results were even more significant when compared to the control group (p = 0.0005) than for NBN gene alone. Among patients with detected defects in MSH2, RAD50 and NBN genes High Risk arm of the PPNG protocol (HR, Additional file 3: Table S2, Additional file 2: Figure S1 A and B) which included Cisplatin and Ifosfamide was not applied more frequently than in the series of patients from control group without detected defects in those genes (p = 0.13). When patients were subdivided according to treatment arm, the results pointing toward more frequent toxicity in patients with identified variants were also statistically significant (p = 0.01 for HR arm only, and p = 0.01 for protocol for children <3 years old only).   HR-PPNG High Risk protocol, SR-PPNG Standard Risk protocol (Additional file 2: Figure S1A), <3 yrs.

LCA -large cell/anaplastic
Because the above results indicate association between defects in analyzed DNA repair genes and presence of adverse effects during therapy we performed WES analyses in three patients who suffered from rare grade 4 adverse events but had no abnormalities in MSH2, RAD50 and NBN genes. In all three patients candidate variants were detected in FANCM, ERCC2 and EXO1 genes which are presented in Tables 2 and 3. In summary, taking into account the WES results, among patients with detected variants in DNA repair genes majority of them (8 out of 15; 53%) suffered from rare life-threatening grade 4 toxicity during the course of treatment.

Discussion
Assuming that the fundamental feature of cancer is genomic instability, functional defects of proteins which are responsible for maintenance of genome integrity by correcting DNA replication errors, should be carcinogenic. It is therefore not surprising that a number of cancer susceptibility genes encode key factors of DNA repair pathways. Recent comprehensive analysis of germline mutations in pediatric cancers pointed to DNA repair genes as the most commonly mutated genes, including TP53 and BRCA2 [43]. It is also increasingly clear that defects in DNA repair genes may determine patient's response to radio and chemotherapy [13,16,17]. In view of that we evaluated the potential association between DNA repair defects and treatment related toxicity as well as their potential role as a susceptibility factor for medulloblastoma.
The sequence analysis of two well-known repair genes MSH2 and RAD50 conducted in large cohort of 102 medulloblastoma patients revealed three new germline variants MSH2 p.V606I and p.A733T as well as RAD50 p.R1093*. Both the localization and the character of detected variants allow for prediction of their probably pathogenic impact on the encoded proteins what was supported by the results of the in silico analysis ( Table 2). The p.V606I and p.A733T substitutions are localized in the crucial DNA mismatch repair protein V (MutSV -aa 619-854, pF00488) in highly (p.V606I-phyloP:4.40) and moderate (p.A733T-phyloP:2.55) conserved amino acid region. MutSV domain contains the dimerization interface and nucleotide-binding site with C-terminal helix-Uturn-helix motif that is critical for MutS function [44]. The RAD50 p.R1093* variant resulting in premature stop codon has severe consequences on the protein translation and predicts suppression of its protein. All identified variants were uncommon in our patients (1/102; 0.98%). This is consistent with the published data indicating that molecular variants in MSH2 and RAD50 in CNS tumors occurred very rarely (5/1637 -0.31% and 4/1743 -0.23%, respectively [37]. In support of that, in published recently study of germline mutations in pediatric cancers, including medulloblastoma, MSH2 and RAD50 variants were not reported [43]. Due to the lack or low frequency of candidate variants in control groups the estimation of cancer risk associated with their presence was not possible (Table 2). However, deleterious character of detected germline variants, the role of the encoded proteins in DNA repair system and their annotation with genetic syndromes, including NBSLD and CMRDS associated with medulloblastoma, make them the potential susceptibility variants for this kind of tumor. The RAD50 p.R1093* variant was reported as one of two known molecular defects (HGMD CMO92910) responsible for NBSLD. In medulloblastoma patients pathogenic variants in MLH1, MSH6 and PMS2 genes were detected previously [21][22][23][24]43]. Alterations in these genes together with MSH2 defects lead to CMRDS. Both MSH2 and RAD50 encode the crucial components of the DNA repair system. MSH2 belongs to mismatch repair genes (MMR) while RAD50 together with MRE11 and NBN constitute the MRN complex responsible for connecting DNA damage detection to DNA repair and cell cycle checkpoint function [44,45]. Biallelic deficiency in MMR genes had been referred as a molecular cause of increased predisposition to gastrointestinal and hematological malignances, as well as early-onset CNS tumors (especially glioblastoma; GBM) [22,26]. Additionally, the germline heterozygous variants in MMR gene were reported in patients with Turcot syndrome associated with medulloblastoma incidence. The molecular variants affecting genes of the MRN complex might also play a role in pediatric tumor development. The evidence that NBN heterozygous variants predispose to childhood acute lymphoblastic leukemia and medulloblastoma was already published [19,[46][47][48]. All these facts reinforce potential role of DNA repair genes, including MSH2 and RAD50 in susceptibility to medulloblastoma but detailed mechanistic studies are required to confirm this preliminary hypothesis.
Notwithstanding the role of DNA repair genes in pathogenesis of medulloblastoma, it is profoundly important from the clinical perspective that the presence of molecular defects in these genes may have an impact on the course of treatment.
The MRN complex genes, including MRE11, NBN and RAD50 are required for double-stand DNA break (DBS) repair via one of the DNA repair system, homologous recombination (HR). Defects in HR system lead to hypersensitivity to agents that produce DSB and topoisomerase inhibitors eg. etoposide [14,15,49]. MMR genes remove mispaired nucleotides by the cooperation in mismatch repair system whose defects are associated with hypersensitivity to DNA crosslinks and platinumbased chemotherapeutic agents (eg. mitomycin C and carboplatin) [14,15,49].
Medulloblastoma treatment protocol (Additional file 2: Figure S1 A and B) includes the drugs mentioned above, specifically platinum-based chemotherapeutic agents (carboplatin, cisplatin), topoisomerase inhibitor etoposide and, in addition, mitotic inhibitor vincristine. Therefore it is very likely that our patients with molecular variants in DNA repair genes may be more prone to complications in recovering from chemotherapy induced DNA damage. They include the patients with variants detected by WES analysis in ERCC2, FANCM or EXO1 genes, an essential components of DNA repair systems [14]. All three identified variants (p.R695C, p.L694* and p.V738L, respectively) were localized in highly conserved nucleotide position (phyloP: 0.89-0.99) in crucial for the encoded protein domains (Table 2) and the character of detected variants (nonsense and splice site) strengthens their pathogenic role. Previous functional studies had also linked these variants to increased sensitivity to therapeutic agents. Defects of ERCC2 protein were reported as a cause of faults in the nucleotide excision repair mechanism (NER) which is responsible for removal of variety of helix-distorting DNA lesions, as well as a hypersensitivity to platinum derivatives. The FANCM gene is one of the elements of the Fanconi Anemia (FANC) pathway responsible for DNA crosslinks repair, possibly through coordination of three main DNA repair systems: nonhomologus endjoing (NHEJ), homologus recombination (HR) and translesion DNA synthesis (TLS). Loss of function of this system results in sensitivity to DNA crosslinking agents and platinum derivatives [14,15,49]. Finally, EXO1 gene encode a nuclease which cooperates with MRN complex in DSB repair via HR pathway as well as interacts with MMR genes in repair of DNA mismatches [50,51].
Although we acknowledge that functional studies are necessary to explore the mechanism through which DNA repair gene defects influence the treatment related toxicity we have already found significant association between defects in NBN, MSH2, RAD50, FANCM, ERCC2 and EXO1 genes and clinical data. Indeed, more than half of patients with variants in DNA repair genes suffered from rare adverse grade 4 events after administration of chemotherapy (Table 3). We acknowledge that validation cohorts would be necessary for confirmation of our results. Unfortunately, recently published NSG 'discovery sets' of medulloblastoma ranged only from 39 to 92 samples and molecular defect in MSH2, RAD50 and NBN gene were not identified [7,[9][10][11][12]43]. Also, an information related to the therapy and accompanied side effects was not provided in these studies. However in two pediatric medulloblastoma patients with mutations in DNA repair genes (PALB2 and BRCA2) chemotherapy inducted grade 4 side effects were reported [16,17]. In addition, effect of other drugs being introduced to medulloblastoma treatment protocols e.g. temozolomide (TZM) may be dependent on the status of mismatch repair genes.
In melanoma one variant g.73170T>C in MSH2 gene (rs2303428) was associated with response and side effects and could be used as a molecular marker for TMZ treatment response [52].
On the other hand it is difficult to compare toxic effects caused by cancer treatment in adult patients harboring defects in DNA repair genes with toxicity observed in still developing and vulnerable tissues in children. Different spectrum of tumors in children and therefore different treatment protocols, including very high doses of drugs, may influence dissimilar reaction in the latter population.

Conclusions
Our study was conducted in a single institution on the largest series of uniformly treated patients which provided primary data indicating possible link between defects in DNA repair genes and treatment related toxicity in children. Given the complexity of the data in relation to the rarity of medulloblastoma, our results needs to be confirmed in independent cohorts. If proven, the special vigilance during and after treatment of patients with pathogenic variants in DNA repair genes should be required. Also additional screening for the presence of molecular variants in patients with manifestation of severe adverse events is necessary to acquire more information about their impact on the course of treatment. Finally, the potential revision of the mode of treatment should be considered for patients with germline defects in DNA repair genes in the future.

Additional files
Additional file 1: Table S1. The list of MSH2 and RAD50 gene germline variants detected in cohort of 102 MB patients. (DOC 92 kb) Additional file 2: Figure S1A. Polish Pediatric Neurooncology Group (PPNG) treatment protocol for medulloblastoma patients (in children older than 3 years).

Availability of data and materials
Complete results of WES including FASTQ files are be available in Sequence Read Archive (SRA) repository, (http://www.ncbi.nlm.nih.gov/bioproject/ 374506), BioProject ID:PRJNA374506, SubmissionID:SUB2397997. The full list of MSH2 and RAD50 gene germline variants detected in cohort of 102 MB patients is presented in Additional file 1: Table S1.