Abstract
DNA methylation is an essential molecular assay for central nervous system tumor diagnostics. While some fusions define specific brain tumors, others occur across many different diagnoses. We performed a retrospective analysis of 219 primary CNS tumors with whole genome DNA methylation and RNA NGS. DNA methylation profiling results were compared with RNAseq detected gene fusions. We detected 105 rare fusions involving 31 driver genes, including 23 fusions previously not implicated in brain tumors. In addition, we identified 6 multi-fusion tumors. Rare fusions and multi-fusion events can impact the diagnostic accuracy of DNA methylation by decreasing confidence in the result, such as BRAF, RAF or FGFR1 fusions, or result in a complete mismatch, such as NTRK, EWSR1, FGFR, and ALK fusions.
Introduction
Central nervous system (CNS) tumors are biologically diverse with the 5th edition of the WHO classification recognizing more than 100 entities with numerous molecular drivers, which have been incorporated into diagnostic criteria [1–4].
DNA methylation has emerged as a pan-CNS molecular diagnostic tool that can improve the diagnostic accuracy of brain tumor entities [5–7]. DNA methylation reflects cell lineage and methylation changes due to somatic mutations and gene fusions, and is well suited for tumor classification [8–10]. The DNA methylation classifier was developed from almost 3,000 reference samples and 91 methylation classes [6]. The resulting classes reflecting the molecular makeup of their respective tumor types, including tumors defined by specific fusions.
Gene fusions have become increasingly recognized as important molecular drivers in brain tumors. After the discovery of the BRAF-KIAA1549 fusion in pilocytic astrocytoma, multiple tumor-defining gene fusions were identified [11]. Embryonal tumors previously diagnosed as primitive neuroectodermal tumor (PNET), were reclassified into molecular subtypes defined by gene fusions [12]. Infantile gliomas are driven by a variety of gene fusions that likely resulting in similar downstream signaling [13]. We and others have shown that fusions, such as NTRK and ROS1 fusions, can occur in many different brain tumor histological entities and grades. Fusions for which the classifier was not trained can result in a low diagnostic confidence of DNA methylation or even misleading results [14, 15].
Furthermore, while most tumors are driven by a single gene fusion, rare reports of multiple concurrent gene fusions exist [16]. How multiple concurrent gene fusions impact the DNA methylation signature of brain tumors has not been explored.
In this study we sought to examine the distribution and impact of rare gene fusions and multiple concurrent gene fusions on DNA methylation classifier results.
Materials and Methods
Study design
We performed a retrospective analysis of 219 primary CNS concurrently profiled by whole genome DNA methylation profiling and RNA next generation sequencing (NGS) upon Institutional Review Board approval. Tumors with disease defining gene fusions, on which the DNA methylation classifier was trained, for example BRAF-KIAA1549, were excluded. Histologic diagnoses were updated according to the 5th edition of the CNS WHO. Tumor cell content was estimated from the H&E slide.
DNA methylation
DNA methylation testing was performed as part of clinical care using the Illumina EPIC array, as described previously [17], and analyzed using the clinically validated DNA methylation classifier (V12.5) [6]. Per clinically validated protocol, DNA methylation results were stratified into three groups based on the calibrated score. Score >/= 0.9: represents positive for DNA methylation class; 0.3–0.9 is considered indeterminate, and <0.3 is considered not matching for DNA methylation class [6]. Cases that failed profiling due to low tumor cell content and poor DNA or RNA quality were excluded from further analysis.
RNA NGS
Fusion detection was performed using RNA NGS assays clinically validated at each institution’s CLIA laboratory as part of clinical testing and results retrieved from clinical reports. Fusions were detected using NYU Fusion SEQer, performed as described previously [18], Oncomine Comprehensive Assay V2 and TruSight Oncology 500 at Weil Cornell, Illumina TruSeq RNA Exome Kit at the NIH, and Archer FusionPlex at MSKCC (Supplemental table 1).
Integration of pathology and molecular data
DNA methylation profiling results were compared with RNA NGS detected gene fusions. Cases were sorted into groups based on the driver gene in the fusion and by the classifier score. The fusions were given the distinction of novel or known based on literature review. Concordance was defined as a match between DNA methylation class and histology diagnosis as we described previously [7].
Data Availability
The data generated in this study, DNA methylation classes, calibrated scores, and gene fusions, are available within the article and its supplementary data files.
Results
Cohort description
We identified 219 primary CNS tumors, which included 41 different histologic diagnoses and 42 DNA methylation classes. We detected 105 different fusions that do not define DNA methylation class involving 31 genes, including 23 gene fusions not previously described in brain tumors. List of genes, fusion partners and diagnoses for each category is listed in Table 1. In addition, we identified 6 cases with multiple concurrent gene fusions (Figure 2). A complete list of cases, DNA methylation results and their fusions is available in Supplemental Table 2.
Table 1.
Overview of the single fusion cohort
| Driver | Histologic diagnosis (N) | Methylation class (N) | Partner (N) (novel partners are bolded) | % calibrated scores <0.9 |
|---|---|---|---|---|
|
| ||||
| ALK (N=4) | Meningioma (1) Glioma NOS (1) Neoplasm NOS (1) Ganglioglioma (1) |
MNG (1) IHG (1) No Match (2) |
LOC101929418 (1) PPP1CB (1) PPFIBP1 (1) SFPQ (1) |
60% |
|
| ||||
| BCOR (N=5) | Glioma NOS (2) Neuroepithelial tumor (3) |
HGNET_BCOR (3) PLEX_PED_B (1) No match (1) |
EP300 (5) | 40% |
|
| ||||
| BEND2 (N=4) | Epithelial/glial neoplasm (1) Diffuse glioma (1) Ependymoma (1) Neoplasm NOS (1) |
HGNET_MN1 (3) No match (1) |
EWSR1 (2) MN1 (1) MAMLD1 (1) |
25% |
|
| ||||
|
BRAF
(N=20) |
Ganglioglioma (1) Glioblastoma (3) Glioma NOS (5) Glioneuronal tumor (1) Low grade glioma (1) Pilocytic astrocytoma (8) PXA (1) |
A_IDH_HG (1) CONTR_INFLAM (1) GBM_MES (1) GBM_RTK_I (1) HGAP (1) LGG_PA_GG_ST (7) LGG_PA_PF (4) MGNT (1) No match (1) PXA (2) |
AGK (5) ARGLU1 (1) BCAS1 (1) CADPS2 (1) DENND2A (1) EXOC4 (1) GTF2I (3) KIAA1549 (2) PAG1 (1) PRKAR2B (2) RIN2 (1) TMEM178B (1) |
57% |
|
| ||||
| EGFR (N=10) | Glioblastoma (10) | GBM_RTK_II (8) GBM_MES (1) GBM_RTK_I (1) |
SEPT14 (6) DYNC1I1 (1) PKD1L1 (1) PSPH (2) |
36% |
|
| ||||
| EWSR1 (N=8) | Ependymoma (1) Glioma NOS (2) Spindle cell neoplasm (1) Neoplasm NOS (3) Nerve sheath tumor (1) |
MNG (1) NET_PATZ1 (1) NET_PLAGL1 (1) No match (4) SCHW (1) |
ATF1 (2) CREM (1) PATZ1 (3) PLAGL1 (1) ZNF444 (1) |
75% |
|
| ||||
| FGFR1 (N=13) | DNET (2) Glioblastoma (1) Glioma NOS (4) Glioneuronal tumor (1) HGAP (2) Neoplasm NOS (1) Pilocytic astrocytoma (2) |
ANA_PA (3) CONTR_CEBM (1) GBM_MID (1) LGG_DNT (3) LGG_GG (1) LGG_PA_GG_ST (1) LFF_PA_PF (2) LGG_RGNT (1) No match (1) |
TACC1 (13) | 64% |
|
| ||||
| FGFR2 (N=2) | Diffuse glioma (1) Glioma NOS (1) |
PLNTY (1) IHG (1) |
VPS35 (1) ERC1 (1) |
100% |
|
| ||||
| FGFR3 (N=57) | DNET (1) Glioblastoma (55) Glioneuronal tumor (1) |
GBM_MES (29) GBM_RTK_II (21) LGG_DNT (1) LGG_GG (1) No match (5) |
KLF15 (2) TACC3 (55) |
30% |
|
| ||||
| MET (N=15) | Astrocytoma (1) Astro IDH mut (3) DNET (1) Glioblastoma (10) |
A_IDH_HG (6) GBM_MES (2) GBM_MID (1) GBM_RTK_I (3) GBM_RTK_II (2) LGG_DNT (1) |
AGK (1) CAPZA2 (3) CLIP2 (1) PTPRZ1 (9) ST7 (1) |
20% |
|
| ||||
|
NTRK1 (N=11)
NTRK2 (N=22) NTRK3 (N=6) |
Ana PA (1) Astro IDH mut (1) Diffuse glioma (1) DNET (1) Ganglioglioma (3) Glioblastoma (12) Glioma NOS (2) High grade glioma (7) Infiltrating glioma (1) Low grade glioma (2) Neoplasm NOS (2) Neuroepithelial tumor (2) Pilocytic astrocytoma (2) PXA (1) Neuroepithelial tumor (1) |
A_IDH_HG (1) ANA_PA (2) ANT_CON (1) CONTR_INFLAM (1) DLGNT (1) GBM_MES (1) GBM_MID (4) GBM_RTK_I (2) GBM_RTK_II (3) HGAP (1) HMB (1) IHG (2) LGG_DNT (2) LGG_PA_GG_ST (2) LGG_PA_MID (1) LGG_PA_PF (2) No match (8) PLEX_PED_B (1) PXA (3) |
AFAP1 (1) ARHGEF2 (1) BCAN (1) BCR (1) CD247 (1) DST (1) EML1 (1) ETV6 (6) FMN2 (2) FRY (1) GKAP1 (1) KANK1 (2) KANK2 (1) KTCD16 (1) KIF21B (1) LMNA (1) MEF2D (1) MYO5A (1) NACC2 (2) NBPF20 (1) NTRK1 (1) SLMAP (1) SPECC1 (1) SPECC1L (1) TPM3 (2) TPR (1) TRIM24 (2) TSN3 (1) |
60% |
|
| ||||
| RAF1 (N=4) | Glioblastoma (1) Pilocytic astrocytoma (2) PXA (1) |
GBM_MES (1) LGG_PA_PF (2) PXA (1) |
ATG7 (2) QKI (1) TBL1XR1 (1) |
50% |
|
| ||||
| ROS1 (N=10) | Ana PA (1) Glioblastoma (6) Glioma NOS (1) Glioneuronal tumor (1) Neuroepithelial tumor (1) |
CONTR_INFLAM (1) GBM_MES (1) GBM_RTK_I (1) GBM_RTK_II (3) HGAP (1) IHG (1) LGG_PA_MID (1) No match (1) |
ANKS1A (1) GOPC (8) ZCCHC8 (1) |
30% |
|
| ||||
| YAP1 (N=2) | Meningioma (2) | MNG (2) | FAM118B (2) | 0% |
|
| ||||
| CIC (N=2) | CIC sarcoma (1) Undiff SRBCT (1) |
EFT_CIC (2) | NUTM1 (1) LEUTX (1) |
0% |
|
| ||||
| ESR1 (N=2) | Myxopap epen (1) Ependymoma (1) |
EPN_MPE (1) EPN_PF_B (1) |
CCDC170 (2) | 0% |
|
| ||||
| ETV6 (N=1) | Glioblastoma (1) | GBM_RTK_II (1) | RIMBP2 (1) | 0% |
|
| ||||
| GLI1 (N=2) | Glioblastoma (2) | GBM_RTK_II (1) No match (1) |
MDM4 (1)
CPM (1) |
50% |
|
| ||||
| MAML2 (N=1) | DIPG (1) | DMG_K27 (1) | RRAGB (1) | 0% |
|
| ||||
| MYB/MYBL1 (N=2) | Glioblastoma (1) Diffuse astrocytoma MYB/MYBL1 altered (1) |
GBM_MYCN (1) LGG_MYB (1) |
QKI (1) CA10 (1) |
50% |
|
| ||||
|
PDGFRA
(N=1) |
Glioblastoma (1) | GBM_RTK_II (1) | CCT2 (1) | 0% |
|
| ||||
| PIK3CA (N=1) | Glioblastoma (1) | GBM_MES (1) | SOX2 (1) | 100% |
|
| ||||
| PPARG (N=1) | Medulloblastoma (1) | MB_G3 (1) | SYN2 (1) | 0% |
|
| ||||
| PRKCA (N=1) | Glioneuronal tumor (1) | LGG_DNT (1) | GPRC5B (1) | 100% |
|
| ||||
| PTEN (N=1) | Glioblastoma (1) | GBM_RTK_II (1) | LIPJ (1) | 0% |
|
| ||||
|
RAD51B
(N=1) |
AT/RT (1) | ATRT_MYC (1) | CEP170 (1) | 0% |
|
| ||||
| RET (N=2) | Glioblastoma (2) | GBM_MID (1) GBM_RTK_II (1) |
CCDC6 (1) PCM1 (1) |
0% |
|
| ||||
| ZFTA (N=1) | Glioma NOS (1) | EPN_RELA (1) | MAML2 (1) | 100% |
PXA: pleomorphic xanthoastrocytoma; AnaPA: anaplastic pilocytic astrocytoma; Undiff SRBCT: undifferentiated small round blue cell tumor, Myxopap epen: myxopapillary ependymoma, DIPG: diffuse intrinsic pontine glioma, AT/RT: atypical teratoid and rhabdoid tumor
Figure 2.
Multifusion driven tumors. Gene fusions are generally considered to be unique driver events; however, we identified 6 tumors with multiple concurrent fusions. Case SG222 was diagnosed histologically descriptively as a low grade neuroepithelial tumor and classified poorly as a LGG_DNT with a score of 0.43. This case had 3 different gene fusions with all 6 genes involving chromosome 7 (A). Case SG183 was diagnosed histologically and by DNA methylation as a pilocytic astrocytoma with a calibrated score of 0.89 and also had 3 different gene fusions with all 6 genes involving chromosome 7 (B).
Distribution of gene fusions and impact on DNA methylation results
The most frequent gene in our cohort was FGFR3 (26% of all cases) and it almost exclusively partnered with TACC3 (98% of cases), with the exception of two cases in which the partner was KLF15. FGFR3-KLF15 fusion has been previously reported in cancer of the biliary tract but not in the brain [19]. FGFR3-TACC3 gene fusion is a well-recognized molecular driver of gliomas; however, there is no DNA methylation class of FGFR3 driven tumors [20]. The vast majority of FGFR3-TACC3 tumors were histologically glioblastoma (GBM) (96%) and classified by methylation with high scores as either GBM MES or GBM RTK II. Five tumors histologically diagnosed as GBM did not classify by DNA methylation.
While FGFR3-TACC3 fusion was almost exclusively detected in GBM, FGFR1-TACC1 fusions (6% of the cohort) were detected in low grade glioneuronal tumors and infiltrating gliomas ranging from CNS WHO Grade 1 to 4. Interestingly, in 7 of the 14 FGFR1-TACC1 fusion driven tumors, histology was not definitive for any WHO entities and cases carried descriptive diagnoses, and frequently classified poorly by DNA methylation. DNET, pilocytic astrocytoma, and GBM were concordant by DNA methylation but with a lower-than-expected score despite the high tumor cell content, suggesting a possible impact of the FGFR1-TACC1 fusion on the DNA methylation classifier.
FGFR2 fusions were only found in two cases, with VPS35 and ERC1 partners, with descriptive histologic diagnoses of diffuse glioma and glioma NOS and classified poorly by DNA methylation.
In BRAF gene fusions (9%), we identified 14 different gene partners (Figure 1). Novel BRAF fusion partners included ARGLU1, DENND2A, and PAG1. Histologically, BRAF fusion driven tumors were diagnosed as a variety of low or high-grade gliomas or glioneuronal tumors. Poorly classified tumors were diagnosed descriptively as glioma NOS or glioneuronal tumor and fusion partners included AGK, ARGLU1 , EXOC4, PAG1, and DENND2A. The remaining three fusion partners were part of multi-fusion cases (Figure 2). These fusions resulted either in concordant diagnoses between DNA methylation and histology, but with a low score or a complete mismatch between histological diagnosis and DNA methylation.
Figure 1.
Histologic and DNA methylation diagnoses in BRAF, NTRK, EGFR, and MET driven fusions with their fusion partners. BRAF driven fusions had a variety of different fusion partners and occurred in low grade gliomas, low grade glioneuronal tumors, and high grade gliomas (A). NTRK1, 2, and 3 genes represented 17% of our cohort (N=40). NTRK1 fusions, highlighted in red, had a variety of different fusion partners and were histologically mostly high grade gliomas that classified by DNA methylation as glioblastoma with scores <0.9 in 5 of 12 cases. NTRK2 fusions, highlighted in green, had a variety of different fusion partners and consisted of an even mix of high grade gliomas and low grade glioma/glioneuronal tumors. The histologic and DNA methylation classes were mostly concordant, 9 cases had a calibrated score <0.9, and 4 cases did not classify by DNA methylation. NTRK3 fusions, highlighted in blue, were primarily fused with ETV6 and spanned many different histologic and DNA methylation classes (B). EGFR fusions were exclusively seen in glioblastoma, diagnosed as such by histology and DNA methylation. The primary fusion partner was SEPT14 (N=6) (C). MET fusions had a variety of different fusion partners and consisted, almost exclusively, of high grade astrocytomas with 100% concordance between histology and DNA methylation and calibrated scores >0.9 in 80% (D).
NTRK1, 2, and 3 fusions, represented 17% of our cohort. NTRK1 fusions had a variety of different fusion partners (Figure 1) and were histologically mostly high-grade gliomas (75% of cases) that classified by DNA methylation as subtypes of glioblastoma. NTRK2 fusion were a mix of high- and low-grade gliomas and or glioneuronal tumors with a variety of fusion partners. Though the histologic and DNA methylation diagnoses were mostly concordant, many had low scores, or did not classify by DNA methylation. In addition, NTRK2-FMN2 and NTRK2-KANK2 were to our knowledge not previously detected in brain tumors. NTRK3 primarily partnered with ETV6, additional partners included, MYO5A and a novel partner, FRY. Interestingly, all tumors showed a complete discrepancy between DNA methylation and histologic diagnosis suggesting that NTRK3 fusions have a profound impact on the DNA methylation classifier.
MET fusions represented 7% of our cohort and had a variety of different fusion partner. Tumors were almost exclusively high grade astrocytomas with the exception of one DNET (Figure 1). The DNA methylation classes were concordant with high calibrated scores in most tumors. The three cases with lower calibrated scores included a DNET and an IDH mutant high-grade astrocytoma, both with PTPRZ1-MET fusion, and a glioblastoma, with a CAPZA2-MET fusion.
EGFR fusions were exclusively seen in glioblastoma diagnosed with largely high DNA methylation scores. The most common fusion partner was SEPT14.
As we have shown before, ROS1 primarily partners with GOPC [15]. Other ROS1 partners included ZCCHC8 and ANKS1A. ROS1 fusions occurred across multiple different entities and good correlation with DNA methylation.
EWSR1 gene fusions are rare in primary brain tumors [21]. In our cohort, fusions involving EWSR1 accounted for 4% of cases and most tumors classified poorly by DNA methylation. In addition, 6 of the 8 tumors carried descriptive histologic diagnosis, two of which classified well with a recently described methylation class that is defined by the fusion present in the tumor, in these cases, NET_PATZ1 and NET_PLAGL1 [22, 23].
ALK gene fusions were detected in five histologically different tumors each with a different partner. The methylation class correlated with the histologic diagnosis with a high calibrated score in only one case of a meningioma with an ALK-LOC101929418 fusion. Of the 4 remaining tumors, 2 tumors did not classify with any class, one tumor diagnosed descriptively as glioma NOS classified as an infantile hemispheric glioma with a score <0.9, and one was part of a multi-fusion case.
EP300-BCOR fusions (5 cases) were diagnosed descriptively and most classified well with HGNET_BCOR while one classified as a choroid plexus carcinoma, and one did not match with any class.
BEND2 had a variety of different fusion partners, and mostly included tumors that were histologically diagnosed descriptively and primarily classified well with the DNA methylation class HGNET_MN1.
RAF1 fusions, were present in low grade glioneuronal tumors or high-grade glioma and had variety of partners and while histologic diagnosis and DNA methylation class correlated well, low scores were seen in cases of RAF1-QKI and RAF1-ATG7.
MYB / MYBL1 fusions define a specific subset of low-grade glioma, and while these were excluded from our cohort, we identified a case of a low-grade glioma that classified as LGG_MYB but with CA10, a fusion partner that, to our knowledge, has not been described previously. In addition, we identified a case of a GBM with a MYB fusion, which classified by DNA methylation as GBM_MYCN with low calibrated score.
Most of our tumors were interaxial, however in two meningiomas, which are tumors not driven by gene fusions, we detected FAM118B-YAP1 fusion and had high score DNA methylation.
We identified several gene fusions where one partner is a known driver in brain tumors and the other partner is novel in brain tumors (N=22). These fusions occurred in tumors ranging from low grade glioneuronal tumors to glioblastoma. While many of these tumors had concordant histologic and DNA methylation diagnoses some showed low DNA methylation score or no matching class (Table 1).
Multi-fusion tumors
Gene fusions are generally considered to be unique driver events with a single fusion driving growth of the tumor. However, we identified 6 tumors with multiple concurrent fusions (Table 2). These multi-fusion events seemed to span different tumor types and could involve multiple different chromosomes or within a single chromosome. Two cases with multiple fusions showed fragmentation and multiple breaks of chromosome 7. One was diagnosed as a low grade neuroepithelial tumor and classified poorly as a LGG_DNT by DNA methylation and had 3 fusions with all 6 fusion genes on chromosome 7 (Figure 2A). A second case with three concurrent gene fusion was diagnosed histologically and by DNA methylation as a pilocytic astrocytoma with a calibrated score of 0.89 also with all 6 genes residing on chromosome 7 (Figure 2B). A third multi-fusion case was histologically a classic pilocytic astrocytoma and even though it classified by DNA methylation as a pilocytic astrocytoma, it did so with a calibrated score <0.9. RNA NGS detected two fusions, a hallmark KIAA1549-BRAF fusion and an additional LMNA-NTRK1 fusion. Most surprisingly, one multi-fusion tumor histologically diagnosed as a meningioma was classified by DNA methylation as a solitary fibrous tumor (SFT) with a high score of 0.94. However, fusion analysis showed an ALK-DYNC1I2 fusion and an EGFR VIII fusion, but not the SFT defining, NABT2-STAT6 fusion.
Table 2.
Multifusion cases
| Case | Histologic diagnosis | Fusion | Methylation class | Calibrated score |
|---|---|---|---|---|
| SG027 | Meningioma WHO grade 1 | DYNC1I2-ALK EGFRvIII |
SFT_HMPC | 0.95 |
| SG056 | Pilocytic astrocytoma WHO grade 1 | LMNA-NTRK1 KIAA1549-BRAF |
LGG_PA_PF | 0.87 |
| SG127 | Glioma NOS | FGFR1-TACC1 PTPRZ1-ETV1 |
LGG_DNT | 0.96 |
| SG183 | Pilocytic astrocytoma WHO grade 1 | AKAP9-BRAF GTF2I-BRAF PTPRZ1-CUL1 |
LGG_PA_GG_ST | 0.89 |
| SG222 | Low grade neuroepithelial tumor | GNAI1-BRAF TNS3-ETV1 EGFR-IMMP2L |
LGG_DNT | 0.43 |
| SG226 | Glioblastoma WHO grade 4 | KLF15-FGFR3 SPTAN1-THADA EGFRvIII |
GBM_MES | 0.99 |
Discussion
The development of the DNA methylation classifier revolutionized our ability to classify brain tumors. The classifier was developed using orthogonally validated groups characterized by the combination of histological features and molecular drivers [12].
Gene fusions are important drivers of CNS tumors and play a critical role in both the diagnosis and therapy. While some gene fusions are disease defining (ZFTA), we and others have shown that some gene fusions such as NTRK and ROS1, can be present across a wide spectrum of histological entities [14, 15]. While the DNA methylation classifier continues to expand with novel entities[22–24], rare fusions and novel fusion partners can have a profound impact on the DNA methylation signature, potentially decreasing diagnostic accuracy.
We demonstrate that rare fusions can have two major impacts on the DNA methylation classifier. First, the tumor diagnosis may remain concordant between histology and DNA methylation class but the calibrated score is below the commonly accepted calibrated score 0.9, which indicates high-confidence. This seems to be typical in tumors with BRAF, RAF1, FGFR1, and ROS1 fusions. Second, the tumor diagnosis can be discordant between histology and DNA methylation class with a low calibrated score, which seems to be common in tumors with NTRK, EWSR1, FGFR, and ALK fusions. Finally, the fusion may have no impact on the DNA methylation classifier as was seen with FGFR3, MET, and EGFR fusions.
Our findings highlight the importance of comprehensive molecular analysis for gene fusions in tumors with discordant diagnoses between histology and DNA methylation or concordant diagnoses but with a low calibrated score. Importantly, RNA NGS assays that can detect any fusion partner, are superior to RNA NGS assays that can only detect predefined fusion partners.
Comprehensive fusion analysis by RNA NGS also provides opportunity for discovery of novel drivers. Here we describe 23 novel gene fusions not previously reported in CNS tumors. Genes that are known drivers in CNS tumors, such as EGFR, ROS1, NTRK, BRAF, and ALK that are fused with genes that have no previously documented implications in CNS tumors such as RRAGB, which is involved in mTORC1 signaling, RIMBP2, a gene involved in neuromuscular synaptic transmission; ANKS1A, which is involved in ephrin receptor signaling pathway and neuron remodeling and PKD1L1, a polycystin family gene involved in Shh signaling. We also identified a glioblastoma with a MDM4-GLI1 fusion, MDM4 is known to bind the p53 tumor suppressor protein and inhibit its activity and GLI1 in Shh signaling.
Lastly, we identified CNS tumors with multiple concurrent fusions some as a result of chromosome shattering. The presence of multiple concurrent fusions can completely mislead the DNA methylation classifier, as in the case of meningioma with concurrent ALK and EGFR gene fusions, which classified as solitary fibrous tumor by DNA methylation. Furthermore, concurrent gene fusions may present a therapeutic dilemma such as the described case of a pilocytic astrocytoma with two targetable gene fusions, BRAF and NTRK1. There are currently no guidelines for the concurrent treatment of tumors with multiple gene fusions. This case also highlights the limitations of targeted assays, such as FISH for confirmatory studies.
In summary, we demonstrate that DNA methylation can be sensitive to rare or multiple concurrent gene fusions. Gene fusions that are not tumor type specific represent a particular caveat as DNA methylation cannot be trained to detect them as a defining driver. Therefore, any CNS tumors with discordant and poorly classifying DNA methylation results should undergo a comprehensive gene fusion analysis with NGS assays designed to detect a large spectrum of fusions and potential novel fusion partners.
Supplementary Material
Implications:
DNA methylation signatures need to be interpreted in the context of pathology and discordant results warrant testing for novel and rare gene fusions.
Acknowledgements
The study was supported by the Friedberg Charitable Foundation, Gray Family Foundation, Sohn Conference Foundation, Making Headway Foundation and by NIH grants R01-CA226527 and R01-NS122987. This work has been supported by the NIH/NCI Cancer Center Support Grant P30-CA016087-40 to the Laura and Isaac Perlmutter Cancer Center, NYU Langone Health.
M.S. is scientific advisor and shareholder of C2i Genomics. Heidelberg Epignostix and Halo Dx, and a scientific advisor of Arima Genomics, and InnoSIGN and received research funding from Lilly USA.
Footnotes
Conflict of interest:
Other authors declare no conflict of interest.
References
- 1.Board W.C.o.T.E., World Health Organization Classification of Tumours of the Central Nervous System. 5th ed. 2021, Lyon: International Agency for Research on Cancer. [Google Scholar]
- 2.Gröbner SN, et al. , The landscape of genomic alterations across childhood cancers. Nature, 2018. 555(7696): p. 321–327. [DOI] [PubMed] [Google Scholar]
- 3.Pinarbasi E. and Pratt D, The Evolving Molecular Landscape of High-Grade Gliomas. Cancer J, 2021. 27(5): p. 337–343. [DOI] [PubMed] [Google Scholar]
- 4.Ryall S, et al. , Integrated Molecular and Clinical Analysis of 1,000 Pediatric Low-Grade Gliomas. Cancer Cell, 2020. 37(4): p. 569–583.e5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wu Z, et al. , Impact of the methylation classifier and ancillary methods on CNS tumor diagnostics. Neuro Oncol, 2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Capper D, et al. , DNA methylation-based classification of central nervous system tumours. Nature, 2018. 555(7697): p. 469–474. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Galbraith K, et al. , Clinical utility of whole-genome DNA methylation profiling as a primary molecular diagnostic assay for central nervous system tumors-A prospective study and guidelines for clinical testing. Neurooncol Adv, 2023. 5(1): p. vdad076. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sturm D, et al. , Hotspot mutations in H3F3A and IDH1 define distinct epigenetic and biological subgroups of glioblastoma. Cancer Cell, 2012. 22(4): p. 425–37. [DOI] [PubMed] [Google Scholar]
- 9.Arslan AA, et al. , Genome-Wide DNA Methylation Profiles in Community Members Exposed to the World Trade Center Disaster. Int J Environ Res Public Health, 2020. 17(15). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Horvath S, DNA methylation age of human tissues and cell types. Genome Biol, 2013. 14(10): p. R115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Pajtler KW, et al. , Molecular Classification of Ependymal Tumors across All CNS Compartments, Histopathological Grades, and Age Groups. Cancer Cell, 2015. 27(5): p. 728–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Sturm D, et al. , New Brain Tumor Entities Emerge from Molecular Classification of CNS-PNETs. Cell, 2016. 164(5): p. 1060–1072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Clarke M, et al. , Infant High-Grade Gliomas Comprise Multiple Subgroups Characterized by Novel Targetable Gene Fusions and Favorable Outcomes. Cancer Discov, 2020. 10(7): p. 942–963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Torre M, et al. , Molecular and clinicopathologic features of gliomas harboring NTRK fusions. Acta Neuropathol Commun, 2020. 8(1): p. 107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Richardson TE, et al. , GOPC-ROS1 Fusion Due to Microdeletion at 6q22 Is an Oncogenic Driver in a Subset of Pediatric Gliomas and Glioneuronal Tumors. J Neuropathol Exp Neurol, 2019. 78(12): p. 1089–1099. [DOI] [PubMed] [Google Scholar]
- 16.Black M, et al. , Concurrent Identification of Novel EGFR-SEPT14 Fusion and ETV6-RET Fusion in Secretory Carcinoma of the Salivary Gland. Head Neck Pathol, 2020. 14(3): p. 817–821. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Serrano J. and Snuderl M, Whole Genome DNA Methylation Analysis of Human Glioblastoma Using Illumina BeadArrays. Methods Mol Biol, 2018. 1741: p. 31–51. [DOI] [PubMed] [Google Scholar]
- 18.Hindi I, et al. , Feasibility and clinical utility of a pan-solid tumor targeted RNA fusion panel: A single center experience. Exp Mol Pathol, 2020. 114: p. 104403. [DOI] [PubMed] [Google Scholar]
- 19.Rizzato M, et al. , Prognostic impact of FGFR2/3 alterations in patients with biliary tract cancers receiving systemic chemotherapy: the BITCOIN study. Eur J Cancer, 2022. 166: p. 165–175. [DOI] [PubMed] [Google Scholar]
- 20.Lasorella A, Sanson M, and Iavarone A, FGFR-TACC gene fusions in human glioma. Neuro Oncol, 2017. 19(4): p. 475–483. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Lopez-Nunez O, et al. , The spectrum of rare central nervous system (CNS) tumors with EWSR1-non-ETS fusions: experience from three pediatric institutions with review of the literature. Brain Pathol, 2021. 31(1): p. 70–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Sievers P, et al. , Recurrent fusions in PLAGL1 define a distinct subset of pediatric-type supratentorial neuroepithelial tumors. Acta Neuropathol, 2021. 142(5): p. 827–839. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Alhalabi KT, et al. , PATZ1 fusions define a novel molecularly distinct neuroepithelial tumor entity with a broad histological spectrum. Acta Neuropathol, 2021. 142(5): p. 841–857. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sievers P, et al. , Pediatric-type high-grade neuroepithelial tumors with CIC gene fusion share a common DNA methylation signature. NPJ Precis Oncol, 2023. 7(1): p. 30. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data generated in this study, DNA methylation classes, calibrated scores, and gene fusions, are available within the article and its supplementary data files.