The genomic landscape of dysembryoplastic neuroepithelial tumours and a comprehensive analysis of recurrent cases

Abstract Aims Dysembryoplastic neuroepithelial tumour (DNT) is a glioneuronal tumour that is challenging to diagnose, with a wide spectrum of histological features. Three histopathological patterns have been described: specific DNTs (both the simple form and the complex form) comprising the specific glioneuronal element, and also the non‐specific/diffuse form which lacks it, and has unclear phenotype–genotype correlations with numerous differential diagnoses. Methods We used targeted methods (immunohistochemistry, fluorescence in situ hybridisation and targeted sequencing) and large‐scale genomic methodologies including DNA methylation profiling to perform an integrative analysis to better characterise a large retrospective cohort of 82 DNTs, enriched for tumours that showed progression on imaging. Results We confirmed that specific DNTs are characterised by a single driver event with a high frequency of FGFR1 variants. However, a subset of DNA methylation‐confirmed DNTs harbour alternative genomic alterations to FGFR1 duplication/mutation. We also demonstrated that a subset of DNTs sharing the same FGFR1 alterations can show in situ progression. In contrast to the specific forms, “non‐specific/diffuse DNTs” corresponded to a heterogeneous molecular group encompassing diverse, newly‐described, molecularly distinct entities. Conclusions Specific DNT is a homogeneous group of tumours sharing characteristics of paediatric low‐grade gliomas: a quiet genome with a recurrent genomic alteration in the RAS‐MAPK signalling pathway, a distinct DNA methylation profile and a good prognosis but showing progression in some cases. The “non‐specific/diffuse DNTs” subgroup encompasses various recently described histomolecular entities, such as PLNTY and diffuse astrocytoma, MYB or MYBL1 altered.


INTRODUCTION
Paediatric diffuse low-grade gliomas (LGG) and low-grade glioneuronal tumours (LGGNT) account for 25-30% of all central nervous system (CNS) tumours of childhood [1,2]. This large group of tumours is highly heterogeneous, causing major diagnostic challenges, and also encompasses newly recognised types in the fifth edition of the World Health Organisation (WHO) CNS tumours, such as diffuse astrocytoma, MYB/MYBL1 altered, polymorphous low-grade neuro-epithelial tumour of the young and diffuse low-grade, MAPK pathway-altered.
LGGNTs, including dysembryoplastic neuroepithelial tumours (DNTs), are particularly challenging to diagnose since this group includes a large spectrum of tumours that are difficult to discriminate based on their histopathological features. DNTs are characterised by a cortical location and their association with early-onset drug-resistant focal epilepsy and account for 5 to 20% of histopathological diagnoses in epilepsy surgery depending on the histopathological criteria used [3][4][5]. The histopathological hallmarks are a multinodular growth pattern and the specific glioneuronal element, which is observed in specific DNTs. The simple form consists of the unique glioneuronal element, and the complex one consists of the glioneuronal element in

Key Points
• DNTs are characterised by FGFR1 alterations, including single nucleotide variations and gene rearrangements, but the lack of FGFR1 disruption is not sufficient to rebut a diagnosis of DNT.
• As occasionally observed in glioneuronal tumours, pilocytic astrocytomas or paediatric-type diffuse low-grade gliomas, DNTs can progress (over many years) through iterative surgeries and require a long-term follow-up. Such progression does not signify malignant transformation and should not necessarily question the diagnosis of DNT.
• DNA methylation profiling can help in diagnosing challenging DNT cases and may convey prognostic information at diagnosis on the progression risk.
• The terms "non-specific" and "diffuse DNT" mentioned in the WHO CNS tumours classification correspond to a heterogeneous molecular group encompassing diverse and newly described distinct glioneuronal tumours and paediatric-type diffuse low-grade gliomas. combination with glial nodules. Non-specific DNTs lack the specific glioneuronal element, rendering them difficult to diagnose, and although this entity was described in the 2016 WHO classification, it remains under debate [6].
The genomic profile of DNTs has been shown to be stable, with only a few copy number alterations. Mutation or duplication of the tyrosine kinase domain of FGFR1 have been found in approximately 70% of DNTs, whilst recurrent BRAF V600E mutations have been identified in approximately 30% of the DNTs, including all subtypes [7][8][9][10]. However, most of these studies were performed before the era of DNA methylation profiling.
DNTs are WHO grade I tumours and classically depicted as stable, even in the event of partial surgical resection. Their prognosis depends primarily on epilepsy-related morbidity [11][12][13][14]. Nevertheless, some cases of progressive tumours or those showing malignant transformation have been described in the literature [15][16][17][18][19][20][21][22][23][24][25][26]. However, the immunohistochemical features of these cases were succinct, and molecular data were extremely scarce or non-existent, so that the diagnosis of DNT could legitimately be challenged. Indeed, very limited molecular information is available on progressive DNTs, and so far, there is no series that included molecular data characterising these rare circumstances.
We used targeted methods (immunohistochemistry (IHC), fluorescence in situ hybridisation (FISH) and targeted sequencing), and largescale genomic and epigenetic methodologies to perform integrative analyses with histology and imaging data to further characterise specific and "non-specific" DNTs, including specific DNTs that went on to progress ("progressive DNTs").

Study design
All samples were first subjected to histopathological review (MP, PV). Complex DNTs and non-specific DNTs were screened for IDH1 R132H and H3K27M substitutions by IHC. All samples were then subjected to DNA methylation profiling and targeted analyses for the most common alterations described in DNT: FGFR1 mutations in exons 12 and 14, FGFR1 internal-tandem duplication (ITD) and BRAF V600E mutations, using IHC and droplet digital polymerase chain reaction (ddPCR, supporting information Figure S1). Since gene rearrangements are a frequent event in paediatric LGGNT, negative cases with frozen tissue available were submitted for RNA sequencing analyses to screen for gene fusions. When no frozen tissue was available, samples were subjected to FISH analysis using break-apart probes targeting the four most frequent gene rearrangements in paediatric LGNT (BRAF, FGFR2, MYB and MYBL1). Finally, samples without an identified potential driver alteration were submitted for more extensive sequencing: whole-exome sequencing (WES) for samples with frozen tissue and targeted sequencing of a panel of 22 cancer-relevant genes for samples with only formalinfixed-paraffin embedded (FFPE) material (supporting information Figure S1).

Patients and tumour samples
We screened the GHU-Paris-Sainte-Anne hospital neuropathology longitudinal database for patients diagnosed with DNT from January 1993 to December 2016 allowing us to retrieve 217 tumours. After patient consent collection as well as assessment for suitable tissue availability (FFPE or frozen tissue), 172 tumours were eligible. Subsequently excluding the tumours with no available clinical or magnetic resonance imaging (MRI) data and those for which the retrieved material did not allow further histomolecular investigation (poor tissue quality and failed DNA/RNA extraction), we obtained a cohort of 112 tumours. We selected 82 tumours that met criteria for the diagnosis of specific and non-specific DNT according to the current WHO CNS tumour classification. We limited the inclusion of "classic" DNT, which have been previously well characterised in the literature (30 "classic" DNT were excluded from this analysis), with the aim to focus our efforts on characterising recurrent DNTs. Therefore, the cohort was enriched with DNTs that went on to progress after surgery (called "progressive DNTs") and was not composed of consecutive cases. Clinical data were collected from the patient records, including sex, age at diagnosis, tumour location and seizure history.
Imaging review was performed under the supervision of a senior paediatric neuroradiologist (NB). Pre-and post-operative MRIs were compared to evaluate the presence of a post-operative residue, and to evaluate tumour-size increase and contrast enhancement occurrence.
Progressive DNT was defined as a DNT showing at least one of the following three criteria during the follow-up: an increased size of hypoT1/hyperT2 signal and/or an increased size of the contrast enhancement and/or the occurrence of contrast enhancement.
Sections for genomic analyses and IHC were prepared from zinc formalin-fixed paraffin-embedded tissue specimens (5% formalin,

Droplet digital PCR
An area representative of tumour was selected from haematoxylineosin-saffron (HES) stained sections, and the tumour cell content was estimated for each sample. Tumour DNA was extracted from 4-μm thick sections of zinc formalin-fixed paraffin-embedded tissue.

Fluorescence in situ hybridisation
FISH analysis was performed on interphase nuclei on paraffinembedded tissue (4 μm), following standard procedures as previously described [30,31] and using break-apart probes targeting BRAF, FGFR2, MYB, MYBL1 and FGFR3. A case was considered positive when the scored nuclei displayed a break-apart signal in at least 20% of the counted nuclei. Hybridisations were considered noninformative if the FISH signals were either lacking or too weak to be interpreted. The results were recorded using a DM6000 imaging fluorescence microscope (Leica Biosystems, Nanterre, France) fitted with appropriate filters, a CCD camera and digital imaging software (CytoVision, v7.4).

Targeted sequencing
An area representative of the tumour was selected from HES sections, and the tumour cell content was estimated for each sample. Tumour instructions and as previously reported [32]. All samples were checked for expected and unexpected genotype matches.
The .idat files were uploaded to the online CNS tumour DNA methylation classifier at https://www.molecularneuropathology.org (v11b4) and a report for every tumour was generated, providing prediction scores for methylation classes and chromosomal copy-number plots. The calibrated scores were integrated in the histopathological findings according to the recommendations from Capper et al. [33] and as previously reported [34].
Additional analyses were performed in R studio (v4.0.2). Raw signal intensities were obtained from .idat files using the minfi Bio-

Clinical and histopathological data of the cohort
We retrieved 82 paediatric and young adult tumours diagnosed as DNT from the neuropathological archives of GHU-Paris-Sainte-Anne hospital, including 51 males and 31 females, with a median age at diagnosis of 10 years (range 2-29) ( Table 1 and supporting information Tables S1 and S2).
After histopathological review, a specific glioneuronal component was observed in 58 cases (specific DNTs), and 13 and 45 cases were classified as simple and complex DNTs according to the WHO classification criteria. Other cases (n = 24) were classified as "non-specific" DNT according to the WHO classification and further analysed separately from specific DNTs (Table 1 and supporting information   Tables S1 and S2).

Histomolecular characteristics of the specific DNTs cohort
By IHC, among the 58 cases of specific DNTs, no simple DNT showed extravascular CD34 expression and only 6 (13%) complex DNTs did (Table 1 and supporting information Table S1). None of the samples LGG_DNT and two cases with a calibrated score <0.3) (supporting information Tables S1 and S3). A two-dimensional tstochastic neighbour embedding (t-SNE) projection alongside 244 LGGs and LGGNTs from the Heidelberg reference cohort showed that all samples clustered with DNTs from the reference cohort and separately from other LGGs and LGGNTs, including those that were associated with a low calibrated score, with the exception of one case that clustered with the LGG PA/GG ST methylation class ( Figure 1A).
Mutations or structural variations (SVs) involving FGFR1 were found in 37/51 of the DNA methylation profiled tumours (73%), including 23 ITDs, 9 mutations, 2 fusions and 3 other SVs ( Figure 1B and supporting information Table S1). All FGFR1-mutated DNTs harboured the hotspot mutation at residue K656 with the exception of one harbouring a mutation at residue N546, another hotspot. In addition, two cases harboured a second point mutation in FGFR1, H649R and D652G. No FGFR1 status was available for two patients for whom ddPCR failed, and no frozen tissue was available for further analyses (old samples, >10 years). Among the seven tumours without available methylation profiling, FGFR1 status was obtained for six (two ITDs, one mutation and three wild type) ( Figure 1B and supporting information Table S1).
All 58 specific DNTs of the cohort have been screened for BRAF V600E by IHC and/or ddPCR. No BRAF V600E mutation was observed.
We performed RNA sequencing in 14 cases and whole exome sequencing in 6 cases lacking FGFR1 disruption and for which frozen tissue was available (supporting information Tables S1 and S4), as well as targeted DNA sequencing for those with no frozen tissue available. We detected one fusion, and mutations in various genes across this panel of tumours, but none of these alterations were recurrent. However, several genes are known to be frequently involved in genomic events in other cancers, including paediatric LGGs, and even have been shown to play driver roles. A fusion involving BRAF was detected in one tumour (DNA methylation score of 0.58 for LGG-DNT) and represented the unique BRAF disruption identified in this cohort of specific DNTs ( Figure 1B). The fusion partner was RNF130, previously reported in ganglioglioma, pilocytic astrocytoma and DNT [35,36]. One case harboured a mutation in the proto-oncogene YES1, encoding Src family tyrosine kinase. A missense mutation leading to K1465N substitution in MTOR was identified in one case. One case was characterised by the presence of a copy gain on chr22q (1 Mb). Interestingly, two cases harboured mutations in IDH1/2 genes; IDH1 A51T of undetermined significance has not been previously described whilst IDH2 R140W is a hotspot mutation. One case stood out by the presence of two germline mutations, involving BRCA2 (DNA repair) and FGFR3 (tyrosine-protein kinase receptor). The nonsense BRCA2 R2318* mutation has been highly reported and is associated with very strong evidence of pathogenicity. We detected a CIC S1552R substitution in one case, previously reported in one oligodendroglioma III (COSS2375605); however, IHC did not show loss of expression of CIC. We indicate which platform was used for each tumour in supporting information Tables S1 and S4. Supporting information Table S4 and Figure S2 summarise the molecular findings.

Integrative analysis of progressive specific DNT
We subsequently focussed our analysis on 25 specific DNTs, all with a DNA methylation profile and with post-surgical progression documented on MRI (called "progressive DNTs").
The median age at first surgery was 9 years old (range 2-16). The  shared the same histological pattern as described in isomorphic glioma, with regular cells scattered in a fine bubbly neuropil [37] (supporting information Figure S4). Although structural variants involving MYB have been described to be associated with angiocentric glioma, an extensive histopathological review of the MYB tumour did not detect any angiocentricity. F I G U R E 4 Methylation-based t-SNE distribution of 23 "nonspecific" DNTs. The 23 "non-specific" DNTs with DNA methylation data available were compared with 244 reference low-grade gliomas samples cohort belonging to 9 methylation classes and 74 control samples from the German Cancer Research Center (DKFZ) [32]. The 51 cases of this study are indicated as black dots RNA sequencing detected two fusions between FGFR2 and INA.
Using a break-apart probe FISH, we detected two additional tumours with an FGFR2 rearrangement, supported by the copy number profile analysis. Three of these tumours were classified as "LGG, GG", and one was not assigned to any class with certainty. As described in PLNTY, histology was heterogeneous [38]. Two shared the same oligo-like pattern, whereas one tumour presented a fibrillary architecture with astrocytic features. Calcification was observed in three cases. All four tumours were characterised by a remarkably strong extravascular CD34 positivity (supporting information Figure S6).
No recurrent alteration was detected in five tumours, but further analyses detected potential driver events. DNA methylation analysis was available for four of these cases. One was classified as "LGG, GG" and another one as "LGG, DNT". DNA-methylation profiling failed to classify with certainty two cases (supporting information Table S2 and S3).
In summary, the integrative analysis enables one to refine this tumour group formerly named "non-specific DNT" into more precise diagnoses of recently described entities ( Figure 5).

DISCUSSION
Most of the genomic alterations identified in paediatric glioneuronal tumours have been discovered in that last 5-10 years, resulting in recent and extensive restructuring of their classification, which has been updated in the sixth Consortium to Inform Molecular and Practical Approaches to CNS Tumour Taxonomy-Not Official WHO (cIM-PACT-NOW) update and integrated into the 2021 WHO CNS classification [39,40]. DNT is the second most common glioneuronal tumour and accounts for 5 to 20% of histopathological diagnoses in epilepsy surgery [3]. This range of reported incidence could be explained by a low inter-observer diagnostic concordance, even among experienced neuropathologists [41], due to sampling artefacts, heterogeneity of the cyto-architectural distribution of the glial and neuronal components, and also the recent description of various F I G U R E 5 Final data integration results in 24 "non-specific" DNTs. The integration of data from DNA methylation array to the histopathological diagnosis and molecular data helped establishing final diagnoses showing that the WHO tumour group "non-specific" DNTs encompassed several morphomolecular entities. LGG = low-grade glioma; GG = ganglioglioma; PLNTY = polymorphous low-grade neuroepithelial tumour of the young; NEC = not elsewhere classified; wt = wild type paediatric oligodendroglioma and diffuse glioneuronal tumour).
Consequently, due to terminology issues and the wide spectrum of histopathological features, there was significant heterogeneity in the published cohorts of DNTs and other epilepsy-associated tumours, which some of which (but not all) will have included non-specific DNTs, depending on preferences and convictions of each group.
No series with molecular data including non-specific DNT has been published to date, in such a way that makes clear whether certain genomic alterations distinguish them from specific DNTs or from other LGNT. We studied 24 non-specific DNTs. BRAF was the most commonly altered gene in the cohort (33%), followed by MYBL1/MYB (20%) and FGFR2 (16%). We also found that a subset of "non-specific" DNT had the classical DNT molecular hallmarks (FGFR1 disruption and DNA methylation profile). Although these cases were fully resected, no specific glioneuronal element has been found even after a careful review. It could be suggested that sampling artefacts and/or the loss of the semi-liquid mucoid specific element during the neurosurgical procedure explains this discrepancy. Our integrated analysis shows that non-specific DNTs encompass a large spectrum of tumours, including recently described histomolecular types. Thus, "non-specific DNT" could correspond to a generic term rather than to a distinct tumour entity, and greater efforts are necessary to harmonising the terminology of these tumours.
DNTs are WHO grade 1 tumours, classically depicted as stable, even in the event of partial surgical resection, and their prognosis depends primarily on the epilepsy-related morbidity [11][12][13][14].  [23]. These data support the idea that although DNTs are grade 1 tumours, they can go on to progress, as can be observed in other grade 1 paediatric brain tumours with MAP kinase alterations (ganglioglioma and pilocytic astrocytoma). The median follow-up of the cohort was 7 years, with a median of 5 years between the first and the second operation. This confirms that local recurrence can occur several years after the first surgery (up to 9 years in our study). These data point out the relevance of long-term clinical and imaging monitoring of these patients. Interestingly, our data tend to indicate that DNA methylation profiling could convey prognostic information and may represent a relevant biomarker in predicting a risk of recurrence.
Presently, there are few prognostic factors in glioneuronal tumours.
Of note, as in our study, two prognostic subtypes of diffuse leptomeningeal glioneuronal tumours (DLGNT) have been recognised based only on DNA methylation profiling (DLGNT-MC-1 and MC2).
DLGNT-MC-2 is enriched for 1q gain and is associated with a shorter survival [46]. In DNTs, the cluster 1 is enriched with progressive form of DNT, but we did not find any clinical, radiological, histological or molecular characteristics associated with this epigenetic signature, and further studies are needed to validate these new data.