In the largest DSRCT series using CGP to date, multiple, recurrent secondary GAs were identified in the majority of clinical samples. The most frequently identified GAs can be broadly classified as: (1) potential oncogenes (i.e., FGFR4); (2) tumor suppressor genes (i.e., TP53 and ARID1A); (3) GAs of unknown clinical significance (MSH3 and MLL3). Less frequently identified GAs were classified as (4) DDR pathway genes and (5) all others. With the exception of TP53 and ARID1A, the remainder of the GAs have not previously been noted [5,6,7,8, 12], likely due to limited gene sequencing. The results suggest the possibility of heretofore unidentified actionable mutations that may have significant implications upon DSRCT oncogenesis and the discovery of potential therapeutic targets.
FGFR4 is a member of the FGFR family (FGFR1, FGFR2, FGFR3 and FGFR4); however, the tyrosine kinase domain structurally different enough from FGFR1–3 that small molecule FGFR1–3 inhibitors are generally incapable of suppressing FGFR4 activation at similarly effective nanomolar concentrations [20]. Relatively frequent mutations in FGFR2 (10% of endometrial cancer) and FGFR3 (20% of urothelial cancer), gene fusion in FGFR2 (45% of intrahepatic cholangiocarcioma) and gene amplification in FGFR1 (19% of ER-positive breast and 17% of squamous cell lung cancer) and FGFR2 (< 10% of gastric cancer) have been described [20]. GAs in FGFR4 have been described in 6.1% of rhabdomyosarcomas (RMS) overall, but is enriched (9.6% in the PAX gene fusion negative subset (generally embryonal subtype) [21]. In a separate analysis, 7.5% of primary RMS contained a tyrosine kinase domain mutation, and the particular mutants K535 and E550 increased tumor proliferation, metastatic potential, autophosphorylation, and Stat3 signaling when expressed in a murine RMS cell line [22]. These results confirm the oncogenic potential of activated FGFR4.
The FGFR4 G388R SNP was originally discovered by Bange et al. and colleagues to be associated with tumor progression in breast and colon cancer patients [23]. Subsequently, this SNP has been reported to be associated with advanced stage and poor prognosis in patients with carcinomas of the lung, prostate, and head and neck; melanomas; and soft-tissue sarcomas [14,15,16,17,18]. This association was confirmed in two large meta-analyses and a pooled analysis of 2537 cancer cases [24, 25]. A causative relationship for this SNP to cancer progression was demonstrated when pMEFs from homologous FGFR4Arg385 knock-in mice were shown by Seitzer et al. to accelerate cell transformation with greater motility and invasive behavior [19]. In vivo, transforming growth factor (TGF)α-induced mammary carcinogenesis, tumor development and progression, and onset of pulmonary metastases were significantly advanced [19]. Later, Ulaganathan et al. established the underlying pathobiology of the SNP; substitution of the conserved human Gly 388 residue to a charged Arg residue modified the transmembrane spanning segment and exposed a membrane-proximal cytoplasmic signal transducer and activator of transcription 3 (STAT3) binding site Y390-(P)XXQ393 [26]. Such STAT3 binding motifs in the germline of type 1 membrane receptors enhance STAT3 activation by recruiting STAT3 proteins to the inner cell membrane. Enhanced STAT3 signaling induced by FGFR4 G388R was confirmed in vivo with the FGFR4Arg385 knock-in mice and transgenic mouse models for breast and lung cancers [26]. These results confirm the oncogenic potential of FGFR4 G388R. Interestingly, when the current DSRCT series was queried for the frequency of this SNP within each genomic alteration cohort, its frequency, albeit with a small sample size, approximated the normal population (32.1%) in the FGFR4 cohort (29%), but was overrepresented within all the other cohorts (44–50%) (Table 2). The significance of this finding suggests the hypothesis that FGFR4 genomic alterations (activating mutations, amplification, or SNP) are sufficient as the “second hit” in translocation-positive cells, whereas there may be an additional requirement for a “third hit” with the FGFR4 SNP in up to half of the other genomic alterations.
TP53 is a tumor suppressor gene, and its inactivation is a frequent event in tumorigenesis [27]. We detected inactivating mutations in TP53 at greater frequency in DSRCT (10%) than previously reported. Jiang et al. reported 0/10 DSRCT samples with TP53 mutations, whereas Bulbul et al. reported 1/15 (7%) DSRCT samples with TP53 mutations [7, 9]. The greater frequency in the current series may simply be related to the larger sample size. Nevertheless, both the present frequency and the other reported frequencies are significantly lower than reported for other soft-tissue sarcomas (32%) [10]. The biology underlying the low frequency of TP53 mutations in DSCRCT compared to other soft-tissue sarcomas is unclear.
ARID1A is one of two mutually exclusive ARID1 subunits of the adenosine triphosphate-dependent chromatin modeling complex switch/sucrose-nonfermentable (SWI/SNF), which acts to mobilize nucleosomes and regulates gene expression and chromatin dynamics [28]. ARID1A is thought to provide specificity to this complex [28]. ARID1A mutations were originally described at high frequency in ovarian clear cell carcinoma (OCCC), an uncommon but aggressive subtype of ovarian cancer. Subsequently, genomic alterations in ARID1A have been described in a broad array of tumor types with the notable exception of sarcomas [29]. ARID1A participates in directing at least 3 processes relevant to tumor suppression: proliferation, differentiation, and apoptosis [28]. Accordingly, it has been labeled an epigenetic tumor suppressor [28]. A single ARID1A nonsense mutation was detected in 1/7 DSRCTs by Devecchi et al. [12], whereas in the current series ARID1A inactivating mutations (truncation or indels) were the third most frequent GA detected in 11% of DSRCT samples underscoring the significance of a larger sample size.
VUS in MSH3 and MLL3 were detected in 14 and 16% of DSRCT samples respectively, accounting for the most frequent genomically-defined subgroups. MSH3 forms a heterodimer with MSH2 to form MutS-β, which comprises part of the post-replicative DNA mismatch repair system. Inactivating mutations of MSH3 is considered a low-risk allele that contributes to development of hereditary nonpolyposis colorectal cancer (HNPCC), or Lynch syndrome [30]. Patients with HNPCC have an increased lifetime risk of developing colorectal cancer, as well as cancers of the endometrium, liver and biliary tract, stomach, small intestine, ovary, ureters, renal pelvis, and brain [30]. MLL3 is a member of the myeloid/lymphoid or mixed-lineage leukemia (MLL) family comprising a nuclear protein with an AT hook DNA-binding domain, a SET domain, a post-SET domain, a DHHC-type zinc finger, six PHD-type zinc fingers, and a RING-type zinc finger [31]. It is a member of the ASC-2/NCOA6 complex (ASCOM), and is involved in transcriptional co-activation through regulation of histone methylation [31]. MLL3 was recently shown to act as a haploinsufficient tumor suppressor gene in − 7/del(7q) acute myeloid leukemia [31]. As recurrent mutations in MSH3 and MLL3 have not been described for sarcomas, their exact role in the pathobiology of DSRCT remains unclear. Regardless, given their roles in other cancers, we suspect the GAs detected are inactivating in DSRCT.
Genes associated with the DDR pathway formed a fourth subgroup of GAs. However, the number of individual cases for each DDR-related gene were limited. Devecchi et al. reported 26 unique somatic mutations in genes involved in the DDR network in 6 of 7 DSRCT cases, including one each of ATR, TP53, and ARID1A [12]. GAs in these genes were also detected in the current CGP, however the remainder of the DDR genes reported here are unique. It is unclear whether the current DDR GAs are driver mutations, passenger mutations, or a result of therapy-induced alterations. Further research into the significance of these DDR genes in DSRCT oncogenesis is necessary.
The role of immunotherapy for sarcomas remains investigational. To date, the limited efficacy of anti-PD1 blockade in other soft-tissue and bone sarcomas has not been reported for DSRCT [32]. In the course of standard clinical care for DSRCT, genomic analysis including TMB and MSI analysis were performed as part of the FoundationOne® Heme panel as TMB High and MSI High have been positively correlated with response to anti-PD1 blockade therapy in other cancers [32]. The results demonstrated the tumors were neither TMB High nor MSI High. These results are consistent with recent reports that DSRCT had low TMB consistent with low immunogenicity, and carry a miRNA signature of immunological ignorance that is not responsive to PD-L1 blockade [9, 33].
The ongoing status of the patients regarding their clinical course were not provided to Foundation Medicine; therefore correlation of the identified genomic alterations to patient outcomes were not available. Certainly, it would be of significant interest should any of these genomic alterations have prognostic value. However, given the limited clinical information in the current series, it is impossible to evaluate. Specifically designed retrospective studies or future prospective studies will be able to determine the clinical significance of these findings.