Introduction

Germline inactivation of the DICER1 gene, either via truncating variants or deletions, is associated with familial DICER1 syndrome, a pleiotropic tumor predisposition syndrome with increased risk of malignant and nonmalignant neoplasms, with onset most frequently in childhood [1]. Sporadic tumors, in which two somatic DICER1 variants are identified, have been described in the absence of an identified germline alteration; these tumors typically resemble those seen among individuals with DICER1 syndrome [2,3,4,5]. Classic DICER1-associated tumors occurring outside the central nervous system (CNS) include pleuropulmonary blastoma (PPB), cystic nephroma, Wilms tumor, Sertoli–Leydig cell tumor, uterine cervical embryonal rhabdomyosarcoma, multinodular goiter, nasal chondromesenchymal hamartoma, and other rare entities [6, 7]. While metastatic PPB is thought to be the most commonly encountered DICER1-associated intracranial tumor, primary DICER1-associated CNS tumors include pineoblastoma, pituitary blastoma, ciliary body medulloepithelioma (CBME), embryonal tumor with multilayered rosettes (ETMR)-like infantile cerebellar tumor, and the recently described “primary DICER1-associated CNS sarcoma” (DCS) [8,9,10,11,12,13,14,15,16].

The first published report of DCS may have been as early as 1996, in a PPB family pedigree containing a proband’s sibling reported to carry a diagnosis of brain sarcoma “histologically indistinguishable from PPB” in the setting of a normal chest X-ray [17]. Over 20 years later, the description of primary DCS has emerged, including one case from Boston Children’s Hospital [8, 14] and one case from McGill University [13], both of which are described in further detail in this series. A series of 22 intracranial sarcomas followed, designated “spindle cell sarcoma with rhabdomyosarcoma-like features, DICER1 mutant” [11]. Additional cases have since been reported with DICER1 mutations, as “primary intracranial sarcoma”, or “spindle cell sarcoma with rhabdomyosarcoma-like features” [12, 16, 18, 19]. Histologically, these lesions were described as variably cellular neoplasms of spindled cells accompanied by rhabdomyoblastic differentiation, and occasionally cartilaginous differentiation [11, 12, 16, 18, 19].

The DICER1 gene encodes an RNase III endoribonuclease that facilitates the activation of the RNA-induced silencing complex essential for RNA interference through cleavage of double-stranded RNA and pre-microRNA into small-interfering RNA and microRNA [6]. Disruption of this pathway results in alterations in protein expression, and in regulatory functions downstream of DICER1, including proliferation, differentiation, and DNA repair [6, 20]. The biallelic pattern of oncogenic DICER1 alterations is distinctive, most often with an inactivating germline or somatic alteration accompanied by a somatic missense hotspot variant within the DICER1 RNase IIIb domain on the other allele [2, 4].

A study examining PPB found that DICER1 variants were frequently accompanied by activating NRAS variants and biallelic inactivation of TP53; copy-number alterations were also described [2]. In initial reports describing this entity, DCS also demonstrates biallelic TP53 inactivation; additional altered genes include KRAS, NRAS, NF1, FGFR4, EGFR, and PDGFR [11,12,13,14].

Knowledge about the clinicopathologic spectrum of DCS is evolving. A precise understanding of the features of affected patients and their tumors, including characteristic molecular findings of DCS, may help to distinguish this clinical entity, and provide insight into its pathobiology and treatment. We report the clinical, radiographic, histologic, and molecular features of six cases of DCS, and compare their genomic profiles to those of other DICER1-associated tumors, with the aims of improving recognition and multidisciplinary evaluation of this new entity, and informing avenues of investigation for future treatment.

Materials and methods

Case ascertainment

Two cases of DCS (DCS_1 and DCS_2) were identified by integrating clinicopathologic features with molecular data from the Dana-Farber Cancer Institute PROFILE study [21]. One additional case (DCS_3) was subsequently identified via the Genomic Assessment Informs Novel (GAIN) therapy consortium [ClinicalTrials.gov Identifier: NCT02520713]. Archival cases with anatomic location in the brain, spinal cord, or meninges, and histologic descriptions that could represent DCS were sought via keyword searches in the Boston Children’s Hospital Department of Pathology database from 1993 through 2018. Selected cases were reviewed microscopically, and 11 bearing resemblance to DICER1-associated malignancies were selected for molecular analysis. Sufficient DNA was extracted for sequencing in 8 of the 11 cases.

Two cases (DCS_5 and DCS_6 [13]) were identified via interinstitutional collaboration. In one of these cases (DCS_5), primary tumor tissue was unavailable for testing (a subset of genes were evaluated at an outside laboratory); instead, tissue from a presumed lung metastasis was provided for molecular analysis using the OncoPanel assay, described below. A comparison group of 14 other DICER1-associated tumors was identified via database inquiries of the DFCI PROFILE and GAIN consortium studies. Of note, our series includes the following previously published cases, with new genomic data here: DCS_2 [8, 14], DCS_6 [13], and the ETMR-L_1 [8] case included in the non-DCS comparison series.

Clinical and radiologic study

Clinical data were abstracted from the medical record. Available imaging reports and/or studies, including radiographs, computed tomography (CT), magnetic resonance imaging, and nuclear medicine imaging of the chest/torso and the brain/neural axis, were reviewed.

Pathologic analysis

Microscopic examination of hematoxylin and eosin-stained slides was conducted in all cases. For a subset of cases, immunohistochemical stains for desmin (Leica, Buffalo Grove, IL), myogenin (Cell Marque, Rocklin, CA), H3K27me3 (EMD Millipore, Burlington, MA), alpha-1 antitrypsin (Roche Ventana), and p53 (Leica) were evaluated. Three cases were stained for periodic acid–Schiff (PAS) with diastase digestion.

Sequencing methods

Tumor sequencing

For both the DCS cases and the comparison group, molecular profiling of tumor DNA was achieved via massively parallel sequencing (OncoPanel) followed by tiered reporting of identified gene alterations as previously described [21, 22]. In brief, DNA extracted from formalin-fixed paraffin-embedded tissues (Qiagen, Valencia, California) was quantified using PicoGreen dsDNA detection (Life Technologies, Carlsbad, California). Targeted next-generation sequencing assessed 447 genes, using the TruSeq LT library preparation kit (Illumina, San Diego, California), a custom RNA bait set (Agilent SureSelect), and the Illumina HiSeq2500, as previously described [22]. Read analysis was conducted using Picard tools, the Genome Analysis Toolkit, Riker, MuTect, and Oncotator. Sequence variants were filtered to exclude those occurring at a population frequency of >0.1% in the gnomAD genome database (Broad Institute). Each variant was reviewed manually by a molecular genetic pathologist. Mutational burden, microsatellite stability, and mutational signatures were calculated as previously described [23].

Germline sequencing

Germline sequencing from peripheral blood was performed via clinical testing or in the context of the GAIN consortium study (ClinicalTrials.gov NCT02520713). For the GAIN study, germline sequencing was done with germline OncoPanel. In brief, DNA extracted from peripheral blood (Qiagen, Valencia, California) was quantified using PicoGreen dsDNA detection (Life Technologies, Carlsbad, California). Targeted next-generation sequencing assessed 147 genes, using the TruSeq LT library preparation kit (Illumina, San Diego, California), a custom RNA bait set (Agilent SureSelect), and the Illumina HiSeq2500, as previously described [24]. A stand-alone germline bioinformatic pipeline utilizing the GATK HaplotypeCaller, or the internally developed tools RobustCNV and BreaKmer, were used to detect SNVs, indels, and CNVs. Sequence variants were filtered to exclude those occurring at a population frequency of >1% in the gnomAD genome database (Broad Institute). Each variant was reviewed manually by a molecular genetic pathologist, and interpreted according to the American College of Medical Genetics guidelines [25].

Statistical analysis

Comparison of the tumor mutational burden in the DCS series and the other DICER1-associated tumors used a Wilcoxon rank-sum test for unpaired data. Comparison of the rate of TP53 and KRAS variants between the DCS and other DICER1-associated tumors was done using Fisher’s exact test. Statistical support was provided by the ICCTR Biostatistics and Research Design Center at Harvard Medical School.

Results

Clinical highlights

The demographic and clinical history of this series, comprising two males and four females diagnosed between the ages of 3 and 15 years (median 6.5 years), is summarized in Table 1.

Table 1 Clinical and genetic information for six cases of DICER1-associated CNS sarcoma.
Full size table

Three of the six patients had previous neoplasms (DCS_3, DCS_4, and DCS_6). Two individuals had a germline DICER1 aberration (DCS_5, DCS_6). For DCS_5, a child with a germline DICER1 mutation, family history revealed no DICER1-related neoplasms other than thyroid nodules in the patient’s mother (mother’s germline DICER1 status not available). DCS_6, a child with a germline chr14q32 deletion including DICER1, had two DICER1-related tumors prior to his diagnosis of DCS: cystic nephroma (age 1) and malignant teratoid CBME (age 4) [13]. The CBME had a previously described second DICER1 variant, different from that identified in the DCS [13]. In addition, a presumed Type I or Ir PPB was identified radiologically shortly after DCS resection [13].

Three cases without germline DICER1 variants were presumed to represent sporadic disease (DCS_1, DCS_2, DCS_3). DCS_1 and DCS_2 occurred in two unrelated children of Peruvian descent. DCS_2 was also found to carry a germline variant of uncertain significance in DICER1 (p.P375R), which was not felt to be clinically significant. DCS_3 had germline testing on the OncoPanel platform. No pathogenic or likely pathogenic variants were identified. Germline material was unavailable for DCS_4.

The results of lung imaging at or around the time of DCS diagnosis are summarized in Table 1. Four of six cases had negative lung imaging at diagnosis. DCS_3 was diagnosed 14 years after resection of a paraspinal “Ewing sarcoma” (fusion testing was unavailable), and 10 years after a lung lesion presumed to be metastatic Ewing sarcoma (both unavailable for histopathologic review). Of note, at the time of diagnosis, DCS_3 had a stable subcentimeter nodule in the right lower lobe that was long-standing and present on imaging prior to DCS diagnosis. DCS_6 with germline DICER1 deletion was noted to have a 1–2-cm lung cyst, felt to represent a Type 1 or 1r PPB as above.

Treatment of DCS for all patients involved surgery, generally aimed at gross total resection. Four were treated with adjuvant radiation and three with multi-agent systemic chemotherapy, detailed in Table 1.

Subsequent to the DCS diagnoses, three patients developed additional neoplasms, including two new lung lesions, although only one (DCS_5) had a known pathogenic germline DICER1 variant. DCS_2 was diagnosed with a presumed low-grade glioma, which was slow-growing and monitored nonoperatively. DCS_3 developed progressive disease in the lungs with multiple nodules. Interestingly, this patient additionally developed a progressive dominant pulmonary lesion close to a previous surgical resection suture site in the left upper zone, first seen with the second DCS recurrence. This lesion had a growth pattern suggestive of another dominant lesion; it is unclear whether this represented a separate primary. In addition to this dominant lesion, DCS_3 also developed multiple pleural-based and pulmonary lesions with appearance characteristic of metastatic disease. DCS_5 developed a single site of disease in the chest 5 months after initial DCS diagnosis. Importantly, sequencing of the lung lesion revealed the same biallelic variants in DICER1, the same TP53 variant (p.S240fs), and the same KRAS variant (p.G12R) found in the brain primary.

We have chosen DCS_2 as an illustrative example of the imaging appearance of DCS (Fig. 1). Images demonstrate the typical location at the gray–white matter junction, a propensity for intralesional hemorrhage, and negative imaging of the chest at the time of diagnosis (supporting the diagnosis of de novo neoplasm rather than metastatic PPB).

Fig. 1: Intracranial mass in patient DCS_2.
figure 1

a Coronal T2-weighted magnetic resonance image (MRI) shows a heterogeneous hyperintense lesion in the anterior right temporal lobe with the surrounding edema. b Axial susceptibility-weighted MRI shows low signal “blooming” in the region of the mass consistent with intralesional hemorrhage. c Axial post-contrast T1-weighted MRI shows enhancement of the mass and the adjacent dura. d Coronal computed tomography (CT) chest and (e) axial positron emission tomography (PET)–CT chest showing no evidence or pleural or pulmonary disease.

Full size image

At the last follow-up, three patients were alive without DCS disease at 46, 30, and 21 months from initial DCS diagnosis. Three patients were dead of disease.

Pathologic features

Microscopically (Figs. 2 and 3), tumors were solid to focally cystic cellular malignant neoplasms with at least a component of fascicular fibrosarcoma-like spindle cells showing brisk mitotic activity and focal or diffuse anaplasia. A striking finding in five of the six cases was areas of differentiation reminiscent of embryonic-type tissues; in areas, cells coalesced into more densely cellular “organoid” formations (including cartilaginous islands in one case) that appeared as light and dark areas at low magnification. Three cases showed prominent round rhabdomyoblastic cells with clear-to-eosinophilic cytoplasm; these cells were often admixed among poorly differentiated primitive cells that sometimes appeared to form a halo surrounding the rhabdomyoblasts. In areas, cellular spindle cell areas predominated, resembling fibrosarcoma or high-grade malignant peripheral nerve sheath tumor (MPNST). Cytoplasmic eosinophilic globules were seen in five cases, most prominent in areas where the cytoplasm was otherwise clear and abundant. Palisading necrosis was seen in one. A florid vascular proliferation at the interface with the underlying brain was present in two.

Fig. 2: Microscopic features of DCS.
figure 2

“Light and dark” areas, with interspersed islands of more densely cellular tumor (a DCS_2; b DCS_4; c DCS_6). Whorled-to-fascicular homogeneous fibrosarcoma-like areas (d DCS_2) with frequent mitoses and powdery chromatin (e DCS_2). Florid microvascular proliferation in a checkerboard pattern at the tumor’s interface with cortex (f DCS_2).

Full size image
Fig. 3: Microscopic features of DCS.
figure 3

Round rhabdomyoblastic cells resembling “spider cells” of rhabdomyoma (a DCS_4) were myogenin-positive, often surrounded by a halo of myogenin-negative smaller cells (b). Hyaline droplets (c DCS_4) staining with PAS (d) and alpha-1-antitrypsin (e) were a frequent but focal finding. Anaplasia (f DCS_1) was seen in all cases. Patchy (g DCS_2) or complete loss of H3K27me3 staining was seen in all cases stained.

Full size image

Immunohistochemical staining showed patchy desmin staining in all cases, patchy myogenin staining in a subset, and negative-to-strongly positive p53. Patchy-to-complete loss of H3K27me3 was observed in all (five of five) cases tested. Eosinophilic globules were positive for PAS (diastase-resistant) and alpha-1-antitrypsin.

Initial pathologic diagnosis (reported before the identification of DICER1 abnormalities in all but one case) was descriptive and included the term “high-grade;” some pathologists applied the term “sarcoma,” and some favored MPNST.

DCS_4 had a prior atypical teratoid/rhabdoid tumor (ATRT). Microscopic examination of the ATRT showed primitive cells with large pleomorphic eccentric nuclei, prominent nucleoli, and loss of INI1 expression, whereas INI1 was retained in the subsequent DCS.

Comparison series

Fourteen additional tumors sequenced with the same assay were determined to be DICER1-associated. Six PPBs included one type I (PPB_2), three type II (PPB_1, PPB_3, and PPB_6), and two type III (PPB_4 and PPB_5). Three uterine sarcomas included one described as a pelvic round-cell sarcoma with “probable” origin in the uterine cervix (US_1), one cervical embryonal rhabdomyosarcoma (US_2), and one diagnosed as a Mullerian adenosarcoma (US_3). Anaplasia was not described in any of the three uterine sarcoma cases. Four Sertoli–Leydig cell tumors included one case with “heterologous stromal differentiation in the form of cartilage and skeletal muscle” (SLCT_4). One thalamic/cerebellar ETMR-like tumor was also included (ETMR-L_1) [8]. The comparison series included 2 males and 12 females, with ages at tissue collection ranging from 3 months to 24 years. All 14 patients had been evaluated for germline DICER1 mutations, and the following had pathogenic (n = 9) or likely pathogenic (n = 1) variants: one of one ETMR-like, three of three uterine sarcoma, two of four Sertoli–Leydig cell tumor, and five of six PPB. Selected genetic alterations, including germline variants, are illustrated in Fig. 3. Complete data about genetic variation are provided in the Supplementary data.

Molecular genetic alterations

Each of the six DCS cases had one inactivating DICER1 alteration and one hotspot variant in the RNase IIIb domain. The inactivating alteration was either a splice site variant (DCS_1, DCS_2, and DCS_3), nonsense variant (DCS_5), or deletion (DCS_4 and DCS_6). The comparison group of DICER1-associated tumors also had both one inactivating and one hotspot variant identified in each tumor in 12 of 14 cases (86%). Cases SLCT_1 and SLCT_4, each had only one hotspot variant identified. A nonsense variant was noted in ten cases (in ETMR-L_1, all three uterine sarcomas and all six PPBs), splice site variants in SLCT_2 and SLCT_3. Case PPB_5 has a low copy gain of DICER1, and none of the comparator group had deletions of DICER1. TP53 inactivation, either by single-nucleotide variation and/or gene deletion, was identified in 83% (5 of 6) of DCS cases, compared with 14% (2 of 14) of other DICER1-associated tumors in our series (p = 0.007). The missense variants reported include known hotspot-inactivating variants such as p.R282W and p.P152L. Similarly, activating KRAS variants were more common in DCS cases (67%) compared with their DICER1-associated tumor counterparts (7%) in our series (p = 0.011). In addition, homozygous NF1 deletion was identified in one DCS case (DCS_3). Alterations in other pathways, including PI3K/Akt, Jak/Stat, chromatin remodeling, and cell cycle arrest, were noted in both series. Of note, many of these variants are variants of uncertain significance (VUS) or possible rare population variants; copy-number variants include low-level gain or loss (often as part of a larger segmental change) of unclear significance.

The DCS cases had a significantly higher tumor mutational burden, with a mean of 12.9 mutations/Mb compared with 6.8 mutations/Mb in the other DICER1-associated tumors (p = 0.035) assessed by the same assay. All 6 DCS cases and all 14 other DICER1-associated tumors were microsatellite stable. Of note, two DCS cases demonstrated findings consistent with previously described mutational signatures, although neither appeared to match known mutagenic exposures. DCS_1, naive to prior chemotherapy, had a mutational signature enriched for (T)C>T alterations (11 single-nucleotide variants), which has previously been associated with exposure to alkylating agents (including temozolomide) [26], while DCS_3, previously treated with radiation and chemotherapy, had a mutational signature enriched for (C)C>T alterations (17 single-nucleotide variants), which has been associated with UVA exposure [26]. None of the other DICER1-associated tumors demonstrated a significant mutational signature.

Figure 4 provides a summary of these molecular alterations. Details of sequence and copy-number variants are provided in Supplemental Tables 1 and 2. Data are reported in GRCh37 (hg19) coordinates.

Fig. 4: Summary matrix of selected recurrent and related alterations for 6 DICER1-associated central nervous system sarcomas (DCS), and 14 other DICER1-associated tumors [(embryonal tumor with multilayered rosettes (ETMR)-like) tumor (1), uterine sarcoma (3), Sertoli–Leydig cell tumor (4), and pleuropulmonary blastoma (6)], assessed with the same assay.
figure 4

Cases are in rows and genetic alterations are in columns. Copy-number alterations are represented as single genes, but often comprise segmental changes. Tumor mutational burden is displayed, measured in mutations/Mb. Note that for DCS_5, a lung lesion (presumed to represent a DCS metastasis) was sequenced and shown here. The DCS cases demonstrate a significantly higher mutational burden (four of six cases “hypermutated” at >10 mutations/Mb) than their counterparts, and are enriched for alterations in oncogenic pathways. Complete genetic data are available in the supplementary material.

Full size image

Discussion

We describe six children with DCS, expanding the clinical, pathologic, and molecular features of this rare disease, and highlighting similarities between this and other primary DICER1-associated tumors.

In all 6 DCS tumors, biallelic DICER1 variants were identified, including an inactivating variant and one of two hotspot mutations D1709N and G1809R in the RNase IIIb domain. These hotspot variants are known to impact cleavage of the 5p mature miRNA from the hairpin loop with significant reduction in expression of 5p-derived miRNAs [2, 27]. The incidence of biallelic alterations (100%) was higher in our study than in a previously published series [11]; notably, we describe three of six cases with splice site variants.

Germline DICER1 testing showed inactivating alterations in two of five patients for whom material was available. Of interest, two patients in our series were noted to be of Peruvian descent. The DCS reported by Koelsche and colleagues also included samples from the Peruvian National Cancer Center [11]. While a single founder mutation in DICER1 could not explain the two Peruvian cases we observed, as both had negative germline testing, the shared ancestry raises the interesting possibility of a shared haplotype or another predisposing gene.

Our study showed that the histologic findings in DCS overlap greatly with those in other DICER1-associated tumors. As with other DICER1-associated tumors, rhabdomyoblastic differentiation is common, and cartilaginous differentiation, when present, is helpful in raising the level of diagnostic suspicion. It is notable that DCS has thus far been described solely as a high-grade neoplasm, without a benign or low-grade counterpart such as that seen in PPB or renal sarcoma.

Our study describes the histologic characteristics of DCS, important for helping pathologists recognize the entity, and providing insight into the pathobiology of DCS. We conclude that immunohistochemical loss of H3K27me3 is not specific to MPNST, as previously reported [28, 29], and DCS may thus be included in the differential diagnosis of high-grade cellular malignant spindle cell neoplasms with loss of H3K27me3. Eosinophilic cytoplasmic globules have been recently reported in DCS [12, 16], and were almost always present in our series of DCS cases. Their positive staining for PAS-diastase and alpha-1 antitrypsin indicates that they are composed of alpha-1 antitrypsin, as is typically observed in areas surrounding the immature neuroepithelium in ovarian immature teratoma.

Our series of DCS differs from others in its breadth of clinicopathologic investigation to identify and exclude the differential diagnostic consideration of secondary brain metastases. Interestingly, both DCS_3 and DCS_4, with histologic and molecular similarities to the other DCS cases, developed in the context of previous malignancies. Case DCS_3 had a remote diagnosis of “Ewing sarcoma” of the thoracic spine and a subsequent large pleural mass, as well as later development of a dominant pulmonary lesion, first appreciated at the time of the second DCS recurrence. These tumors were not available for histologic review, which would have been helpful in elucidating the relationship between these lesions, for example, whether the initial spinal lesion might represent DCS or another tumor. DCS_4 occurred 6 years after an ATRT of the left temporal lobe, which was diagnosed and treated. Due to the distinct histologic and immunohistochemical presentations, the ATRT and DCS were felt to be separate malignancies. The history of two distinct neoplasms in this case raises the possibility of a tumor predisposition, although germline material was not available for testing.

As DCS and metastatic PPB may be histologically indistinguishable, we recommend a chest CT at diagnosis of DCS to confirm that the CNS tumor is the primary disease site. As the histologic differential diagnosis of DCS may also include metastasis from other DICER1-associated tumors (DICER1-associated renal sarcomas, DICER1-associated ovarian sarcomas, primary peritoneal PPBs, etc.), abdominal imaging is also suggested. Interestingly, two of six individuals in our series developed lung tumors subsequent to the diagnosis of DCS. In addition, a recently reported case noted bilateral pulmonary nodules that were presumed to be pulmonary metastases after a diagnosis of DCS [18]. Sequencing of the lung lesion from DCS_5 demonstrated clonal similarity to the DCS, strongly suggesting a metastasis rather than a second primary tumor.

Our findings therefore suggest a role for chest imaging, not only at diagnosis, but also in the follow-up surveillance of primary DCS.

Treatment of the DCS in this series included surgery generally aimed at gross total resection and radiation, with chemotherapy used in some instances that includes similar agents used in advanced PPB (including ifosfamide, doxorubicin, vincristine, and dactinomycin). Consideration for agents that cross the blood–brain barrier should be given.

Molecular profiling revealed enrichment for a pattern of genetic alterations in DCS compared with their DICER1-associated tumor counterparts: TP53 inactivation and activating alterations of genes in the Ras pathway, including variants in KRAS and NF1. While it is possible that the differences may be attributed in part to sample size, our data confirm and expand the previous DCS findings [11,12,13, 16]. The presence of Ras pathway gene activation in DICER1-associated lesions suggests possible therapeutic avenues currently under clinical investigation. Variants in other pathways identified by these studies include PI3K/Akt, Jak/Stat, chromatin remodeling, and cell cycle genes. Many of these variants observed have not been well described in the literature, and therefore represent VUS, or tier 3 variants according to the AMP/CAP/ASCO guidelines for somatic variant interpretation [30]. Until these variants are functionally characterized, we caution against overinterpreting the associated treatment potential.

The significantly higher tumor mutational burden (mean 12.9 mutations/Mb in the DCS cases compared with 6.8 mutations/Mb in the other DICER1 tumors included in our study), albeit among a relatively limited number of cases, may be of clinical interest due to the known association between mutational burden and response to immunotherapy [31,32,33]. However, one important limitation of our TMB analysis is that it is based on tumor-only sequencing, and our data almost certainly include germline variants, given the number of DCS cases occurring in populations not well-represented in population databases [34]. Notably, four of six (67%) DCS cases had mutational burdens >10 mutations per Mb, putting them into the proposed “hypermutant” category of tumors, as previously described [35], although this cutoff has not been formally established as a threshold on the Oncopanel platform used in this study, or standardized across sequencing assays. By comparison, the other DICER1-associated tumors in our series had only 2 of 14 (14%) “hypermutant” cases by this definition. Another potential caveat of these findings is that the comparison group includes lower-grade malignancies (e.g., PPB type I). In addition, while we know that two of the DCS patients had been exposed to prior therapies, the treatment history of our comparator group was not examined. Nevertheless, we note that a higher mutational burden among the DCS cases is statistically significant despite the small sample size. Further examination of a larger number of cases will be needed to corroborate this observation, and to determine whether the elevated TMB observed in DCS cases might predict therapeutic response to immunotherapy.

In summary, we present an expanded series of DCSs with novel clinical, radiologic, histopathologic, and genetic features. The histologic and radiologic features overlap considerably with other DICER1-associated tumors, notably PPB. Given this, distinguishing DCS from metastatic CNS disease is of primary importance. Compared with previously described DICER1-associated tumors, DCS has a higher mutational burden and an enrichment of variants in potentially targetable pathways. The diagnosis of DCS benefits from interdisciplinary care between treating physicians from oncology, radiology, neurosurgery, and surgical and molecular pathology. DCS should be on the differential diagnosis of primary intracranial tumors, particularly when features characteristic of other DICER1-associated tumors are seen. Further descriptions of this new entity will continue to broaden and advance our understanding of its biological features and potential molecular vulnerabilities.