Main

Sarcomas are a rare group of histologically and genetically heterogeneous tumours of mesenchymal origin (Fletcher et al, 2013). Most sarcomas arise sporadically. However, a small number of cases manifest in individuals with germline mutations in genes associated with cancer predisposition syndromes, such as TP53, NF1, RB1, APC, RECQL4, and HRAS (Fletcher et al, 2013; Ballinger et al, 2016; Thomas and Ballinger, 2016). The age of onset of sarcomas is often lower than that observed for most epithelial cancers and, as such, the heritable proportion of sarcomas is likely to be higher than is currently documented (Fletcher et al, 2013; Thomas and Ballinger, 2016). Along with translocations, intra-exonic somatic mutations may also contribute to sarcoma development. In a heterogeneous series of 811 next-generation-sequenced sarcomas, the Cancer Genome Atlas Research Network identifies TP53, PIK3CA, ATRX, PCLO, and LRP1B to be the five most frequently somatically mutated genes (Supplementary Tables S1a and S1b) (cBioPortal for Cancer Genomics).

There are rare reports of sarcomas arising in the context of the DICER1 syndrome (Foulkes et al, 2011; Rio Frio et al, 2011; Kim et al, 2013; Doros et al, 2014; Schultz et al, 2016), a rare paediatric tumour predisposition syndrome caused by germline mutations in DICER1 (OMIM 601200). Priest et al (1996) noted the occurrence of paediatric-onset sarcomas co-occurring with pleuropulmonary blastoma, a tumour now known to be prototypic of the syndrome. Hill et al (2009) further substantiated the association by reporting sarcomas in germline DICER1 mutation carriers. Subsequent reports of sarcomas in DICER1 germline-mutated patients include a para-spinal rhabdomyosarcoma in a 20-year-old (Rio Frio et al, 2011) and a pleomorphic sarcoma of the thigh (consistent with a leiomyosarcoma) in a 26-year-old (Foulkes et al, 2011). A cervical primitive neuroectodermal tumour (Ewing/cPNET) was also reported in a germline DICER1-mutated patient (Foulkes et al, 2011). However, as testing for characteristic second somatic DICER1 RNase IIIb mutations (Foulkes et al, 2014) was not performed, it is not possible to discern whether the lesions are manifestations of the syndrome or co-incidental occurrences. In contrast, an Askin/Ewing family tumour that arose in a 13-year-old germline DICER1 mutation carrier (for more details, see de Kock et al, 2014b) was not found to harbour a characteristic RNase IIIb hotspot mutation (Foulkes, unpublished data). There are also several reports of somatic DICER1 RNase IIIb hotspot mutations in uterine carcinosarcoma (Table 1 and Supplementary Table S1c).

Table 1 Literature review—sarcomas with somatic DICER1 mutations

More recently, DICER1 mutations have been strongly implicated in the pathogenesis of embryonal rhabdomyosarcoma (ERMS) of the uterine cervix (cERMS) (Tomiak et al, 2014; de Kock et al, 2016) the ovary (de Kock et al, 2015), and anaplastic sarcoma of the kidney (D1ASK) (Doros et al, 2014; Wu et al, 2016). Characteristic hotspot DICER1 RNase IIIb mutations were identified in the three aforementioned lesions. Biallelic somatic DICER1 mutations were similarly detected in a case of adult-onset cERMS (de Kock et al, 2016).

Despite the above evidence, the true contribution of DICER1 mutations to sarcomas is not yet known. In this study, we aimed to uncover the contribution of DICER1 mutations to a convenience sample of 61 predominantly adult-onset sarcomas of various subtypes. We recruited an additional 12 Ewing sarcomas consequent to the observation of a cPNET/Ewing and Askin/Ewing family tumour in DICER1 kindred, as described above, for a total of 73 sarcomas.

Materials and methods

Patients and samples

We collected 73 sarcomas of 24 different subtypes, as detailed in the Supplementary Materials and Methods. Age of diagnosis ranged from ages 3 months to 87.4 years (median age 45.7 years), and 38 of the patients were female and 35 were male. This study was approved by the Institutional Review Board (IRB) of the Faculty of Medicine of McGill University, Montreal, Quebec, Canada, number A12-M117-11A, and patients signed consent forms in accordance with the IRB approval.

DICER1 screening

Fluidigm access array

We screened the full DICER1 coding region and exon–intron boundaries in tumour gDNA from 67 (of 73) sarcomas (Supplementary Tables S2a and S2b) using a custom Fluidigm Access Array, which targets all exons and exon–intron boundaries of DICER1, followed by next-generation sequencing on an Illumina (San Diego, CA, USA) MiSeq, as previously described (de Kock et al, 2014a). All identified mutations were validated by Sanger sequencing and matched-normal gDNA, if available, was used to determine whether mutations were germline or somatic in origin.

Sanger sequencing

For the six remaining sarcomas (all FFPE-derived) (Supplementary Tables S2a and S2b), we focused our investigation on the RNase domains of DICER1 to identify known hotspot mutations (Foulkes et al, 2014). The regions encoding the RNase III domains were PCR amplified and Sanger sequenced (de Kock et al, 2014b). Other regions of DICER1 were not sequenced in these six samples.

MLPA assay

We screened for deletions or duplications of DICER1 in the germline of 53 patients from whom good quality non-tumour DNA was available (cases 1–52 and 56) using an in-house multiplex ligation-dependent probe amplification (MLPA) assay, as described previously (Sabbaghian et al, 2014).

Details of bioinformatics methods, cloning experiments, mosaicism experiments, TruSight Tumour 15 panel sequencing of case 1 (including TP53 gene, Illumina), and DICER1 copy number variation (CNV) experiments are provided in the supplement (Materials and Methods section).

Results

We identified multiple DICER1 variants in an ultimately fatal case of abdominal ERMS that arose in a 23-year-old female following a short history of abdominal pain (case 1) (Figure 1 and Supplementary Tables S2a and S3). Two of these variants are likely to be pathogenic (discussed below). The ERMS was detected on ultrasound as a mixed solid and cystic pelvic mass in the broad ligament, measuring 20 cm in its longest diameter with a 10–11 cm solid component (Figure 1). The ERMS, obtained following chemo- and radiotherapy (see Figure 1), harboured a DICER1 RNase IIIb hotspot mutation in exon 25 (c.5439G>T; p.E1813D), which co-occurred with a predicted-truncating DICER1 mutation in exon 11 (c.1785_1786insA; p.T596Nfs*3), both of which were not detected by regular sequencing techniques in the patient’s germline. The patient carried an additional germline insertion (c.2040+53_2040+54insT) in intron 12 of DICER1 (Figure 1C). Experiments to investigate a potential mosaic origin of the exon 25 and exon 11 mutations suggest that neither are likely to be mosaic in nature (Supplementary Table S4). Given the young age of sarcoma onset, we also screened the patient’s germline and tumour samples for TP53 mutations and did not identify any pathogenic TP53 alterations (Supplementary Table S5). Further characterisation of the DICER1 mutations revealed that the exon 11 mutation was in trans with both the intron 12 and exon 25 mutations. The latter two were therefore present in cis (Figure 2 and Supplementary Figure S1). Cloning of a cDNA fragment encompassing all three mutations revealed that the transcript bearing the exon 11, c.1785_1786insA insertion was almost always degraded by nonsense-mediated decay as only 3 of 48 sequenced clones expressed the mutation. No cDNA clones were found to exhibit aberrant splicing as a consequence of the intron 12, c.2040+53_2040+54insT variant, indicating that this variant is most likely to be non-contributory (Supplementary Figure S1).

Figure 1
figure 1

Diagnostic images and mutations for case 1. (A) Axial computed tomography (CT) of pelvis demonstrating a solid pre-sacral soft tissue mass (arrows) with low attenuation signal suggesting central cystic/necrotic change (bottom right arrow). (B) Fused positron emission tomographic and CT image of pre-sacral mass demonstrating high metabolic activity as reflected by F-18 flourodeoxyglucose avidity. Central area of reduced activity coincides with area of central tumor necrosis (arrow). Following surgical resection of the tumour, the patient underwent chemotherapy (vincristine, doxorubicin, and cyclophosphamide) for 4 months and radiotherapy of the abdomen and pelvis (24 Gy). Recurrent pelvic disease was detected after an 18-month disease-free interval. Surgical resection was attempted, but complications were incurred. Three months later, recurrent disease was again noted on positron emission tomography imaging. Two cycles of irinotecan/temozolamide chemotherapy were administered. Fifty-two months after initial diagnosis, the patient succumbed to her disease. (C) The exon 25 c.5439G>T somatic mutation (Panel I), exon 11 c.1785_1786insA somatic mutation (Panel II) and intron 12 c.2040+53_2040+54insT germline mutation (Panel III) as seen in Fluidigm-derived data (left) and chromatogram (right). The mutations are indicated by an asterisk and the wild-type sequence is provided below each chromatogram. A full color version of this figure is available at the British Journal of Cancer journal online.

Figure 2
figure 2

Graphic depiction of biallelic nature of DICER1 mutations identified in the recurrent ERMS of case 1. The exon 11 c.1785_1786insA mutation is in trans (Allele A, top panel) with the intron 12 c.2040+53_2040+54insT and exon 25 c.5439G>T mutations (Allele B, middle panel). The mutations are indicated by a triangle. Only 3 of 48 clones were found to express the exon 11 mutation, suggesting that the mutated transcript is almost always degraded by nonsense mediated decay (NMD) and thus, no protein is likely to be produced from this allele (p.0). No clones were found to exhibit aberrant splicing as a consequence of the intronic c.2040+53_2040_54insT mutation. As such, the resulting protein is predicted to be normal, except for the single amino acid substitution at position p.E1813 (asterisk, Allele B, middle panel). The wild-type (WT) scenario is depicted in the bottom panel. A full color version of this figure is available at the British Journal of Cancer journal online.

Because of DICER1’s involvement in the above-mentioned ERMS, we sequenced a further 72 sarcomas (60 sarcomas of various subtypes and 12 Ewing sarcomas; Supplementary Tables S2a and S2b) and an additional 9 DICER1 variants were identified (Supplementary Table S3). Of the nine variants, six were established to be germline in origin, two were somatic, and for one variant, the germline vs somatic origin remains undetermined (no germline DNA sample available). One somatic variant, c.3208C>G (p.L1070V), identified in a pleomorphic sarcoma with giant cells (case 46), is predicted to be damaging by both PolyPhen2 and SIFT with a score of 1 and 0.01, respectively. However, no additional characteristic RNase IIIb hotspot mutation was found within this sarcoma and therefore its causal role remains speculative. An intronic DICER1 variant, c.2257-7A>G, had previously been identified in the germline of patient 73. However, no RNase IIIb mutation was identified in the Ewing sarcoma. Based on mutation frequency data and in silico effect predictions, the remaining seven variants are unlikely to be involved in the pathogenesis of the sarcomas in question. Germline deletions in DICER1 have also been found to predispose to the DICER1 syndrome (Sabbaghian et al, 2014). We therefore screened for deletions or duplications in the germline of 53 patients from whom good quality non-tumour DNA was available (cases 1–52 and 56) and no such alterations were identified (Supplementary Figure S3). Copy number alterations of DICER1 have been identified in various cancers including breast cancer, ovarian cancer and melanoma (Zhang et al, 2006; Pugh et al, 2014). We screened for CNVs of DICER1 in 59 sarcomas using a ddPCR experiment (chosen due to low DNA input requirement) and detected copy number changes involving the DICER1 locus in 5 cases (8.5%), each of which was a unique subtype (Supplementary Table S6). However, the extent of the CNVs is not accurately definable using the ddPCR system.

Discussion

DICER1 is an RNase III endoribonuclease responsible for processing hairpin precursor microRNAs (miRNAs) into mature miRNAs, which in turn, regulate the expression of messenger RNAs (Foulkes et al, 2014). Germline mutations in DICER1 predispose to several early childhood or adolescent-onset phenotypes, including pleuroplumonary blastoma, Sertoli-Leydig cell tumour and paediatric cystic nephroma (Foulkes et al, 2014). Genetically, DICER1 syndrome-associated tumours are most often characterised by a predisposing germline DICER1 mutation that inactivates one allele, coupled with a highly distinctive second somatic missense mutation affecting one of the RNase IIIb metal ion-binding sites on the other allele (Foulkes et al, 2014). The biallelically mutated recurrent ERMS in our study (case 1) demonstrates that such mutations may contribute to the development of ERMS, even if both mutations are acquired somatically. Although most DICER1-related lesions manifest in early childhood (Foulkes et al, 2014), it is becoming increasingly evident that the acquisition of two somatic DICER1 mutations can lead to a later-onset of neoplasia (de Kock et al, 2016), as was observed in case 1.

Most ERMS that arise in the context of DICER1 mutations involve the urogenital system and interestingly, the biallelically mutated ERMS from our study arose in the broad ligament, which is the peritoneal fold that attaches the uterus, fallopian tubes and ovaries to the pelvis. Although a limited number of other sarcoma subtypes have been found to carry both truncating and/or RNase IIIb hotspot somatic DICER1 mutations (Table 1 and Supplementary Table S1c) (de Kock and Foulkes, 2016), ERMS appears to be the subtype that is most commonly DICER1 mutated. Clinicians should be mindful of the association between ERMS and DICER1 syndrome. Genetic testing should be performed particularly if ERMS are seen to arise in constellation with one or more known DICER1 syndrome phenotypes, as the identification of germline DICER1 mutations has important implications for the screening and counselling of patients and their families.

In summary, our study demonstrates that likely-pathogenic DICER1 mutations underlie the genetic basis of only a small fraction of sarcomas, with ERMS the most likely sarcoma subtype to harbour such mutations. Conversely, the occurrence of a sarcoma at a site other than the genito-urinary system and of a type other than an ERMS (with the exception of anaplastic sarcoma of the kidney) is not suggestive of the DICER1 syndrome.