Recurrent ACVR1 mutations in posterior fossa ependymoma

Molecular profiling of ependymal tumors has resolved a previously histologically defined entity into clinically meaningful tumor types driven by gene fusions, copy number variants, and epigenetic alterations. Moreover, the distinct relationship between neuroanatomic compartment (i.e., supratentorial, posterior fossa, spinal cord), molecular alteration, and clinical outcome [10] now forms the basis for their classification in the 2021 WHO Classification [6]. In contrast to other central nervous system (CNS) tumors, recurrent point mutations are uncommon in ependymal tumors, which instead most commonly harbor specific gene fusions (e.g., ZFTA, YAP1) or copy number alterations (MYCN amplification, chromosome 6 loss). In posterior fossa (PF) ependymomas, recurrent somatic mutations are particularly uncommon [7, 8, 10]. However, a recent study demonstrated that a small subset of posterior fossa group A (PFA) ependymomas harbor recurrent missense mutations in the EZHIP/CXorf67 gene [9]. Histone H3 K27M mutations, seen in the majority of diffuse midline gliomas (DMG), have also been detected in a small proportion of PFA ependymomas [4, 9]. TERT promoter mutations combined with monosomy of chromosome 6 were recently shown to identify a group of clinically aggressive posterior fossa ependymal tumors with hybrid histologic and epigenetic features of ependymoma and subependymoma (EPN/SE) [13]. Exceptionally, ZFTA (C11orf95) fusions have also been detected in PF ependymomas [5]. The lack of recurrent mutations in PF ependymomas has hindered the development of effective targeted therapies. Here, we report the occurrence of somatic heterozygous coding point mutations in the activin receptor type I (ACVR1) gene in a small subset of PF ependymomas. Furthermore, we show that, in the context of retained histone H3 lysine 27 trimethylation (H3K27me3), ACVR1-mutant PF ependymomas exhibit a DNA methylation signature distinct from other PF ependymomas.

We initially identified an uncharacterized group of ependymal tumors (n = 7 unique samples) through unsupervised clustering of genome-wide DNA methylation array data from approximately 16,000 CNS tumors. Next-generation sequencing of tumors from this group revealed ACVR1 missense mutations in all cases for which sufficient material was available (n = 6) (Fig. 1a, herein referred to as PF-ACVR1). Evaluation with the Heidelberg Methylation CNS Tumor Classifier (v11b4/12.5) failed to match these tumors to a known DNA methylation class (calibrated score < 0.9). We also identified two epigenetically-defined PFA ependymomas harboring oncogenic ACVR1 mutations (Fig. 1a, PFA-ACVR1). Clustering analysis of PF-ACVR1 among established PF ependymal tumor types (PFA, PFB, EPN/SE) revealed a distinct DNA methylation signature associated with these tumors (Fig. 1b), confirming findings on non-linear dimensionality reduction. In contrast to ACVR1-mutant DMG, ACVR1-mutant PF ependymomas lacked histone H3 mutation and co-occurring alterations in PIK3CA, PIK3R1, and PPM1D (Fig. 1c). One case harbored deleterious mutations in TP53 and RB1 and another case contained a TERT promoter hotspot mutation. Copy number analysis revealed broad chromosomal gains and losses in the majority of PF-ACVR1 (6/7), resembling profiles that are often observed in PFB (Supplementary Fig. 1, online resource). The median age at diagnosis (32 years) for patients with PF-ACVR1 ependymoma was similar to PFB (29 years), but significantly older than PFA (p < 0.01) and younger than EPN/SE (p < 0.05) (Fig. 1b; Supplementary Fig. 2, online resource). All patients with PF-ACVR1 ependymoma were alive at last follow-up with a median overall survival of 153 months, although most patients suffered from local recurrence over time (Fig. 1d; Supplemental Tables 1 and 2; Supplemental Fig. 4, online resource).

Fig. 1.

Genetic and epigenetic characterization of ACVR1-mutant PF ependymomas. Unsupervised embedding of DNA methylation array data from ependymal tumors across all neuroanatomic compartments (a); also included are H3 K27-altered diffuse midline gliomas (DMG) with and without ACVR1 mutations. Notably, PF ependymomas with ACVR1 mutations and retained H3K27me3 expression (labeled me3 +) form a distinct epigenetic group. Heatmap of DNA methylation array data from PF ependymal tumors confirming four distinct tumor types with hierarchical clustering and their corresponding age distributions (note: two matched primary/recurrent tumors are included) (b). Oncoplot of co-occurring oncogenic or likely oncogenic somatic alterations from ACVR1-mutant PF ependymomas (left) and diffuse midline gliomas (DMG, right) (c). Lollipop plot of the distribution of ACVR1 mutations in PF ependymomas (top) and DMG (bottom) using RefSeq transcript NM_001105 (d); novel missense mutations are labeled with arrows. Kaplan–Meier plot of overall survival and progression-free survival for PF-ACVR1 (e). DMGK27 diffuse midline glioma, H3 K27-altered; ZFTA supratentorial ependymoma, ZFTA fusion-positive; EPN SPINE spinal ependymoma; YAP1 supratentorial ependymoma, YAP1 fusion-positive; MPE myxopapillary ependymoma; EPN/SE PF posterior fossa ependymoma/subependymoma [13]; SUBEPN SPINE spinal ependymoma; SUBEPN ST supratentorial subependymoma; PFA posterior fossa ependymoma, group A; PFB posterior fossa ependymoma, group B; MYCN spinal ependymoma, MYCN-amplified

Histologic review of PF-ACVR1 tumors showed characteristic morphologic features of ependymoma (e.g., perivascular pseudorosettes) and a compatible immunohistochemical profile (i.e., GFAP + , OLIG2-, perinuclear dot-like EMA staining). Interestingly, we observed focal papillary or pseudopapillary features in most cases (Supplemental Fig. 3, online resource). Immunohistochemical staining of PF-ACVR1 tumors showed retained H3K27me3 and an absence of EZHIP/CXorf67 expression (Supplemental Fig. 3, online resource). The two tumors clustering with PFA (PFA-ACVR1) both occurred in young children and demonstrated loss of H3K27me3 with concurrent EZHIP expression in tumor cells (Supplementary Fig. 3, online resource).

To date, ACVR1 mutations in gliomas have been restricted to specific amino acid residues within the TGF-β glycine-serine-rich (GS) (codon 206), protein kinase catalytic (PKc; codons 258, 328), and protein kinase catalytic-like (PKc-like; codon 356) domains [1, 3, 12, 14]. While five of our eight PF ependymomas had known oncogenic ACVR1 missense variants at these mutational hotspots that have been recurrently observed in pediatric gliomas, the other three PF ependymomas harbored p.G328Q (c.982G > T, n = 2) and p.R375C (c.982G > T, n = 1) substitutions in ACVR1 that have not been previously reported in CNS tumors (Fig. 1e; Supplemental table 3, online resource). Germline mutations in ACVR1 are associated with fibrodysplasia ossificans progressiva (FOP) and result in ligand-independent upregulation of bone morphogenic protein (BMP) pathway signaling [3, 11]. Interestingly, the p.G328Q variant reported here has not been reported in a somatic malignancy or in FOP (https://cancer.sanger.ac.uk/cosmic, queried Feb. 22, 2022), but is likely to be activating/oncogenic in this clinical context, nonetheless. In DMG, ACVR1 mutations are associated with distinct clinical and molecular features: they frequently co-occur with H3 K27M mutation in the less common histone H3.1 isoform [encoded by H3C2 (previously annotated as HIST1H3B)], are more common in females, and have been associated with longer overall survival (OS) [14].

Despite a relative lack of recurrent somatic mutations, recent evidence suggests that leveraging ‘super-enhancer’ dependency may be an effective therapeutic approach in ependymal tumors [7]. This approach has identified subtype-specific transcriptional dependencies in ependymomas that may be responsive to small molecule inhibitors. Thus, the identification of lineage- or molecular-based groups will be increasingly important in directing future therapeutic approaches in ependymoma. A recent study also revealed clinical benefit of repurposing a receptor tyrosine kinase inhibitor (vandetanib) in combination with an mTOR inhibitor in patients with ACVR1-mutant DMG [2]. Vandetanib, a multi-RTK inhibitor, has demonstrated a synergistic effect with everolimus in inhibiting ACVR1 downstream effector signaling (i.e., SMAD), as well as improved CNS penetration in this context [2], thus raising the possibility of this approach in ACVR1-mutant PF ependymoma. Here, we show that posterior fossa ependymomas harboring ACVR1 mutation with retained H3K27me3 have overlapping clinicopathologic features with PFB ependymomas but resolve into a unique epigenetic subgroup distinct from other currently recognized ependymoma/subependymoma DNA methylation classes. Further study is needed to delineate the biologic and clinical implications of ACVR1 mutations in PF ependymomas.

Supplementary Material

Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00401-022-02435-2.

Data availability

The raw methylation array data (IDAT format) have been made available for download at the Gene Expression Omnibus (GEO) repository under the accession number GSE196013 (https://www.ncbi.nlm.nih.gov/geo/).