Background

Malignant rhabdoid tumor (MRT), including atypical teratoid/rhabdoid tumor (AT/RT), MRT of the kidney, and extracranial MRT, is a rare and highly aggressive malignancy. These three subtypes have a common genetic hallmark, namely, loss-of-function variants or deletions in the SMARCB1 (also known as the hSNF5/INI1/BAF47) gene [1]. Most MRT cells show bi-allelic alteration in SMARCB1 [1] and are a typical example of Knudson’s two-hit hypothesis [2,3,4].

Variants in the SMARCB1 gene are not always somatically acquired. Previous studies demonstrated that approximately one-third of patients had a germline SMARCB1 abnormality resulting in the first hit of Rhabdoid Tumor Predisposition Syndrome (RTPS) [5,6,7]. Actually, most germline SMARCB1 abnormalities are presumed to be sporadic, but a few familial cases have been reported [5, 6, 8]. Although some studies reported that individuals with RTPS showed a high rate of multiple tumor development at an early age, the clinical characteristics and prognosis remain controversial [6, 7, 9].

In a study by Acuna-Hidalgo et al., intensive genetic analysis revealed that some variants which were previously considered to be de novo actually were post-zygotic mosaic variants with a proportion of 6.5% [10], and it has been suggested that similar mosaic variants may be present in a certain fraction of childhood cancer cases [11]. However, no studies to date have focused on the frequency of post-zygotic mosaicism in MRT. We investigated herein the prevalence of germline variants in the SMARCB1 gene by intensive analysis, including quantitative assessment of copy numbers and a variant detection assay using droplet digital PCR (ddPCR) capable of detecting low-frequency germline alterations.

Methods

Patients

This study was approved by the ethics committee at the National Center for Child Health and Development (#1035) and written informed consent was obtained from the patients and/or guardians. The enrolled patients received the diagnosis of MRT at the National Center for Child Health and Development, the Shizuoka Children’s Hospital or St. Marianna University School of Medicine Hospital between 2006 and 2018. Histological diagnosis was confirmed by the negativity of SMARCB1 by immunohistochemistry. In total, 16 patients were included. Both tumor-derived DNA and germline DNA were obtained from all patients (Table 1).

Table 1 SMARCB1 alterations in 16 patients with germline and/or somatic abnormalities.
Full size table

DNA isolation and PCR analyses for SMARCB1

DNA was extracted using a QIAamp DNA Mini kit or a QIAamp DNA FFPE tissue kit (Quiagen, Hilden, Germany) following the manufacturer’s instructions. Standard PCR reactions were carried out using AmpliTaq Gold 360 Master mix (Applied Biosystems, Carlsbad, CA, USA). The primer sequences for each exon of SMARCB1 are shown in Supplementary Table 1.

First, screening for SMARCB1 alterations in the tumor specimens was done by direct sequencing and ddPCR in all the samples. SNP array analysis was additionally performed for 12 cases with a sufficient quantity and quality of DNA. Then, analysis of germline DNA samples focusing on the genomic abnormalities detected in the paired tumors in each case was performed.

Droplet digital PCR

The ddPCR analysis was performed using QX-200 (Bio-rad, Hercules, CA, USA) to determine the copy number of each exon of SMARCB1 (exons are numbered like in NM_003073.3) [12]. Specific primers and probes were designed for each exon (Supplementary Table 2). An RPP30 gene primer/probe mix (Bio-Rad, Hercules, CA, USA) was used as an internal control [13]. Sensitivity testing of copy number analysis for exon 4 of SMARCB1 showed that the detection limit for deletions was as low as 0.75% (Supplementary Fig. 1). Primer/probe sets were also designed for single nucleotide variants and indels detected in tumors by direct sequencing and were used in analyzing for the presence of low-frequency germline abnormalities (Supplementary Table 3).

SNP array analysis

Genome-wide analysis for copy number alterations and allelic imbalances was performed by single nucleotide polymorphism (SNP) array analysis. DNA extracted from the samples was analyzed using GeneChip Human Mapping 250 K NspI array (Affymetrix, Santa Clara, CA, USA). The obtained data were processed using CNAG/AsCNAR software according to the previously described method [14].

Statistical analyses

The difference in the median age between patient groups was analyzed by variant status using the Mann–Whitney U test. Overall survival and progression-free survival (PFS) were estimated using the Kaplan–Meier method and compared using the log-rank test. Cox regression models were used to examined the prognostic factors affecting survival time.

Results

The 16 patients included those with AT/RT (n = 8), MRT of the kidney (n = 2), extracranial MRT (n = 4), or both intracranial and extracranial MRT (n = 2). The median age at diagnosis was 16.5 months (range, 0–154 months). Bi-allelic alterations of the SMARCB1 gene were found in the tumor cells of all 16 patients (Table 1). Of the 32 genomic alterations, ten were nonsynonymous nucleotide substitutions and frameshift variants, ten were deletions of one or more exons within SMARCB1, and 12 were copy number neutral losses of heterozygosity identified mainly by SNP array analysis. All the detected nonsynonymous variants were nonsense variants or frameshift insertions/deletions leading to the early truncation of the SMARCB1 protein (Table 1 and Supplementary Fig. 2). No patients had any malformations or neurodevelopmental disorders.

In eight of 16 cases (50%), genomic alterations observed in the tumor-derived DNA were also detected in constitutional DNA. It is worth noting that three patients had germline mosaicism (UPN4, 5 and 8) (Fig. 1 and Supplementary Fig. 3). Two patients had mosaic deletion, including SMARCB1, and the average copy number of the deleted region in the SMARCB1 gene in the germline was 1.60 (UPN4) and 1.76 (UPN5), respectively, indicating that 40 and 24% of their cells had heterozygous deletions. For UPN8, the fraction of SMARCB1 variants in normal cells was 1.7%, indicating that 3.4% of cells had the SMARCB1 variant.

Fig. 1: Representative results of mosaic alterations in SMARCB1.
figure 1

a, b Shown are the copy number analyses for each exon of the SMARCB1 gene using digital droplet PCR. Homozygous deletion of exons 1–8 was observed in tumor specimens of UPN4 (a). The copy number of exons 1–8 was estimated at 1.60, indicating that 40% of cells in peripheral blood had a heterozygous deletion in this region (b). c A nonsense variant (c.157C>T) was observed in 76.5% of tumor cells in UPN8 and detected in peripheral blood at a mutant allele frequency of 1.71% indicating a mosaicism rate of 3.42%. Error bars represent 95% confidence intervals.

Full size image

The median age at diagnosis with or without germline abnormalities was 5 months (range, 0–75 months) and 23.5 months (range, 13–154 months), for the respective groups, and the age difference was not statistically significant (p = 0.082, Supplementary Fig. 4). Two cases with both intracranial and extracranial MRT showed constitutional genomic alterations. Recurrence with or without constitutional abnormalities was observed in four and three patients, respectively, in whom PFS at 2 years was 46.9% and 37.5% (p = 0.52, Supplementary Fig. 5).

In three patients (UPN1, 2, and 4) with a constitutional abnormality, genetic analysis of their parents was performed using peripheral blood samples following genetic counseling. None of the parents had any abnormality of the SMARCB1 gene observed in their offspring.

Discussion

We performed a detailed analysis of the prevalence of SMARCB1 germline/constitutional abnormalities using the ddPCR assay, and identified three new cases (37.5%) among the eight patients with a germline abnormality with mosaicism contributing to the pathogenesis of MRT. These cases had a small proportion of mutated cells indicative of mosaicism and were therefore possibly overlooked by conventional qualitative analyses. Our study revealed a prevalence of RTPS in almost half of the MRT cases exceeding previously reported figures [5,6,7] and underscoring the importance of constitutional alterations in the etiology of pediatric cancers, as shown in recent, large-scale studies [15, 16].

In our cohort, truncating SMARCB1 variants, which predispose to RTPS, were observed in all eight patients with constitutional abnormalities including mosaicism [17]. Germline variants in SMARCB1 are also known as causes of schwannomatosis and Coffin-Siris syndrome, thus some patients may have manifestations of these diseases, as well as of RTPS, although none of our cases had the typical phenotype of schwannomatosis and Coffin-Siris syndrome during follow-up period [12, 17].

Some patients with RTPS show MRT development at an older age. In our cohort, two (UPN5 and 8) of eight patients with RTPS had MRT onset at age 2 years or older, suggesting that genetic screening tests are needed in all patients regardless of age. Interestingly, the two cases showed a low frequency of mosaicism for SMARCB1 variants.

As in previous reports [5,6,7], most patients with multiple lesions had germline abnormalities. Bhatt et al. reported MRT development in the soft tissue of a patient with a germline SMARCB1 variant 8 years after the end of AT/RT treatment [18]. Germline carriers are presumably at risk of second MRT development, which is difficult to distinguish from a recurrence.

Considering the high frequency of patients with RTPS, genetic counseling for parents is necessary because familial cases have been reported despite their low frequency [5, 6, 8, 9, 19, 20]. Although genetic tests of the parents (UPN1, 2, and 4) were all negative, the possibility of gonadal mosaicism should be considered because there are patients with constitutional mosaicism and also previous reported cases [5, 6, 8, 9, 19, 20].

This study has several limitations. First, due to the small sample size, the precise prevalence of germline variants and the correlation between clinical features and variant origins could not be accurately assessed and require investigation in larger-scale studies. Second, although our ddPCR method was able to detect a frequency of mosaicism as low as 1%, the prevalence of constitutional variants may have been underestimated due to overlooking other, lower frequency mosaicisms.

In conclusion, approximately half the MRT cases in this study had SMARCB1 constitutional alterations, including low-frequency mosaicism of deletion or nonsynonymous variants. Considering the presence of low-frequency mosaicisms, inherited germline variants in predisposition genes are more important than previously assumed for the pathogenesis of pediatric cancers.