Molecular Landscape of Pediatric Low‐Grade Gliomas: Insights From RNA‐NGS and Bioinformatic Analysis

ABSTRACT

Pediatric low‐grade gliomas (pLGG) are the most common group of childhood brain tumors. Genetic alterations in the RAS–RAF–mitogen‐activated protein kinase (MAPK) pathway are the molecular drivers in the vast majority of pLGG. A large proportion of pediatric pLGG are characterized by the presence of fusion genes. An institutional molecular analysis together with an RNA‐NGS study was performed to reveal LGG‐associated molecular alterations. In our cohort of pLGG patients, molecular alterations were identified in 318 out of 342 cases (92.9%) through a combination of RT‐PCR, Sanger sequencing, and NGS methodologies. Fusion events were independently called using three fusion callers: Archer Analysis 6.0 and/or 7.0, Arriba version 2.4, and STAR‐Fusion 24. Among these, STAR‐Fusion had the lowest sensitivity, detecting rearrangements in only 67% of fusion‐positive cases. In contrast, Arriba detected rearrangements in 97.77% of cases, while Archer detected rearrangements in 88.6% of cases. These findings highlight differences in detection efficiency among fusion callers, emphasizing the importance of tool selection in molecular diagnostics. The detection of fusion genes is very important for correct diagnosis, prognosis, and adequate targeted treatment.

1. Introduction

Pediatric central nervous system (CNS) tumors are the most common (20%–25%) solid tumors in children. Within this group, pediatric low‐grade gliomas (pLGG) are the most common, accounting for more than 40% of all childhood CNS tumors [1]. They represent a heterogeneous group of tumors with different locations, ages at presentation, histologic subtypes, and clinical behavior.

Pediatric CNS tumors are mostly characterized by oncogenic fusions rather than multiple mutated genes and invariably involve activation of the mitogen‐activated protein kinase (MAPK) cascade [2, 3]. During development, the rat sarcoma‐virus family proteins/the mitogen‐activated protein kinase (RAS/MAPK) pathway are important in the formation of the cortex, midbrain, and cerebellum [4, 5, 6]. Its role in neurogenesis is particularly interesting with respect to glial pathogenesis, as the cell of origin for gliomas is now proposed to be a neural stem cell or neural progenitor rather than a post‐mitotic glial cell [6, 7, 8]. The B‐Raf protooncogene (BRAF) is the most commonly altered gene in pLGG with two predominant alterations: the KIAA1549::BRAF fusion and the BRAF V600E pathogenic variant, both resulting in activation of the BRAF kinase [1, 2, 3]. In addition to common pLGG alterations, rarer alterations affecting RAS/MAPK signaling were detected, including those involving FGFR1–3, NTRK1–3, RAF1, ALK, and ROS1 [3, 4, 5, 6].

Only a small proportion of pLGG is characterized by some aberrations that indirectly affect the RAS/MAPK pathway, such as MYB/MYBL1 alterations. MYB is involved in the control of the proliferation and differentiation of hematopoietic and other progenitor cells and is associated with proto‐oncogenic functions in both human leukemia and solid tumors [7]. MYBL1 is a member of the MYB family of proteins. MYBL1 is not a known oncogene but is closely related to the proto‐oncogene MYB [8]. MYBL1 is involved in the positive regulation of transcription by RNA polymerase II.

Less commonly, alterations in Isocitrate dehydrogenase—IDH1,2 are observed in pLGG [9]. The IDH1 R132H mutation was found in only 0.8% of pLGG, with those identified occurring in older children/adolescents [10].

A large proportion of pediatric LGGs are characterized by the presence of fusion genes. Gene fusions are known to result in the formation of a hybrid protein that is either constitutively active or exhibits altered function. The most common example is the KIAA1549::BRAF fusion, first described in 2008 by Jones et al. [11]. The KIAA1549::BRAF fusion results in the tandem duplication of the BRAF gene, leading to the loss of the auto‐inhibitory domain of BRAF and the constitutive activation of its kinase domain [11, 12, 13]. The KIAA1549::BRAF fusion represents the molecular hallmark of pilocytic astrocytoma (PA). Since then, the omics revolution has enabled us to find many other driver‐gene transcript fusions with various fusion partners and genomic breakpoints, including but not limited to BRAF, RAF1, NTRK1/2/3, FGFR 1/2/3, ROS1, EGFR, ALK, and PDGFRA [14, 15].

More than half of the reported unique fusions are intrachromosomal (54%), such as the KIAA1549::BRAF fusion mentioned above, or large deletions, such as those causing the recurrent FAM131B::BRAF fusion. The duplication of the kinase domain of FGFR1 (KDD FGFR1) may be another intrachromosomal rearrangement typical of LGG, consisting of a ~4.5 kb internal tandem duplication of the portion of the gene encoding the kinase domain [5]. However, due to the difficulties in detecting intrachromosomal fusions, their incidence number might be underrepresented and might therefore increase with improved detection algorithms [16].

Molecular analysis, particularly next‐generation sequencing (NGS), is essential to: confirm diagnosis, guide treatment, provide prognostic information, and enable patient recruitment to basket trials based on a tumor's molecular signature. Nevertheless, bioinformatic analysis of raw data obtained from NGS can be challenging and necessitates a range of approaches.

Therefore, an institutional molecular analysis together with an RNA‐NGS study was performed to reveal LGG‐associated molecular alterations and their therapeutic implications. Another goal was to test and compare different callers for the detection of fusion genes, kinase domain duplications, and splice variants.

2. Material and Methods

2.1. Patient Cohort

To determine the molecular alterations in pLGG, we analyzed 362 non‐NF1 tumors (tumors from children without known neurofibromatosis type 1 syndrome). Overall, 94.47% (n = 342/362) had sufficient material for molecular profiling (with positive amplification).

The present study's cohort of pLGG consisted of 362 non‐NF1 patients who were followed and treated at Motol University Hospital from 2000 to 2024. The median age at diagnosis was 8.16 years (range 0–20.2). Patients were followed on an outpatient or inpatient basis with regular MRI imaging. The institutional review board approved the study, and all patients gave informed consent according to our routine procedure.

2.2. Histopathology and Immunohistochemistry

The most representative FFPE block of tumor tissue was selected, and consecutive sections were immunohistochemically examined. The LGG cases were immunohistochemically studied using the following panel of primary antibodies: GFAP (1:1000, rabbit polyclonal, Dako, (Agilent Cat# N1506, RRID:AB_10013482)), NeuN (1:1000, mouse monoclonal, clone A60, Chemicon (Millipore Cat# MAB377, RRID:AB_2298772)), neurofilament protein (1:50, mouse monoclonal, clone 2F11, Dako, (Agilent Cat# M0762, RRID:AB_2314899)), CD34 (1:40, mouse monoclonal, clone QBEnd‐10, Dako, Agilent Cat# M716501‐2, RRID:AB_2750581), and Ki‐67 (1:150, mouse monoclonal, clone MIB‐1, Dako, Agilent Cat# M7240, RRID:AB_2142367). All immunohistochemical reactions were visualized by the PolyDetector DAB HRP Brown (BioSB) immunohistochemistry detection system.

2.3. Molecular Examination

Tumor RNAs were purified from FFPE blocks using the High Pure FFPET RNA Isolation Kit (Roche Diagnostics) or from cryosections using Trizol (Invitrogen). To detect the most common alterations, we have used cost‐effective algorithms that include the initial use of RT‐PCR diagnostics or Sanger sequencing (Figure 1).

FIGURE 1.

Scheme of a cost‐effective algorithm for the detection of molecular alterations in pLGG.

PCR and Sanger sequencing were conducted to examine hotspot mutations at codon 600 of BRAF ex15, codons 546 and 656 of FGFR1 ex12, and of FGFR1 ex14, respectively, using the previously described primer pairs [17, 18]. Amplification was performed using 2× PCRBIO HS Taq Mix Red (PCR Biosystems Ltd., London, UK). Direct Sanger sequencing was performed using BigDye Terminator v 3.1 chemistry (Life Technologies) and an ABI PRISM 3500Dx genetic analyzer (Applied Biosystems). The results were analyzed using Chromaslite 2.01 (Technelysium, Pty Ltd., Brisbane, Australia).

RT‐PCR was used to detect the expected KIAA1549::BRAF fusion because it was more cost‐effective and less time‐consuming. For the RT‐PCR, we used the primers published by Badiali et al. [19]. A similar approach was used to detect the V600E hot‐spot variant of the BRAF gene using mutation‐specific PCR and Sanger sequencing. This approach was used for posterior fossa, spinal cord, midline, and optic pathway tumors. For hemispheric LGGs, in most cases, we proceeded directly to diagnosis by NGS, as for the above localizations with negative findings by PCR techniques.

Molecular detection of the kinase domain duplication of FGFR1 (KDD FGFR1) in the histologic diagnosis of DNET was also performed by RT‐PCR in the majority of cases [3].

In case of a negative result or hemispheric localization, we used NGS. FusionPlex Lung V1 (14 genes) or V2 (17 genes) panels, and in some cases the FusionPlex Pan Solid Tumor V2 (137 genes), Oncology Research (75 genes), and custom panel EMEA (57 genes) (ArcherDX) were used to prepare the NGS libraries. The final amplicons were subsequently sequenced on an MiSeq (Illumina) instrument. Archer FusionPlex also offers robust performance even for FFPE samples. RNA extraction, library preparation, and parallel sequencing were performed as per the manufacturer's recommendations. Fusion events were independently called using three fusion callers: Archer Analysis 6.0 and/or 7.0, Arriba, and STAR‐Fusion 24. Splicing Isoform was called using in‐house workflow. We then designed primers for the newly detected rearrangements and confirmed the new fusion genes by Sanger sequencing.

2.4. Bioinformatics Methodology

The bioinformatics analysis for pLGG was performed using a well‐established pipeline aimed at detecting gene fusions, kinase domain duplications, and alternative splicing events associated with tumorigenesis.

2.5. Reference Genome and Data Preparation

For the transcriptome analysis, we utilized the Gencode GTF file version 45 (GRCh38.p14) [20]. Raw sequencing data were first processed with Fastp, a fast and efficient tool for quality control and preprocessing of high‐throughput sequencing data [21]. To assess the global quality of the trimmed data, we used FastQC [22]. The complete bioinformatics pipeline is described in the Supporting Information.

2.6. Alignment and Fusion Detection

For sequence alignment, we used STAR (Spliced Transcripts Alignment to a Reference) version 2.7.9a, a widely used RNA‐seq aligner, to map the filtered FASTQ files to the GRCh38 reference genome. The alignment process was conducted with default parameters, and the junction file was generated inside the BAM file with hard clipping. One modification was made to the standard STAR settings by adjusting the –chimScoreDropMax parameter to 30, which refines the alignment for chimeric reads—those that map to multiple locations in the genome or have fusion‐gene signatures. This adjustment is intended to enhance the detection of gene fusions, a hallmark of pLGGs [23].

Fusion gene detection was performed using Arriba (version 2.4) [24] as well as STAR‐Fusion [24], both of which are popular tools for detecting fusion events in RNA‐seq data. Arriba was run with a modified parameter ‐E 0.8, which specifies the minimum expected number of supporting reads for a fusion event to be considered. This threshold ensures a higher specificity by minimizing the detection of false positives, particularly in cases where low‐abundance fusion transcripts are present. The use of a stringent threshold like 0.8 is critical in rare fusion detection, as it reduces noise from non‐specific alignments while maintaining sensitivity. Both fusion tools are known for their sensitivity in detecting clinically relevant fusions that drive tumorigenesis [16].

Additionally, we used STAR‐Fusion with default parameters for fusion gene detection. While both tools identify similar fusion events, the use of multiple fusion callers allows for cross‐validation and ensures the robustness of the results [25].

2.7. Detection of Splicing Events and Isoforms

In addition to gene fusions, we sought to identify smaller genetic events, such as exon skipping and alternative splicing patterns, which are often associated with cancer‐specific isoforms. These alterations are often difficult to detect with conventional exon‐based approaches but can be identified by analyzing chimeric junctions.

For this purpose, we developed an in‐house workflow designed to detect these splicing events by identifying chimeric junctions between exons, which represent the skipped regions of the genome.

3. Results

To determine the true frequency of molecular alterations in our pLGG cohort, we analyzed 342 tumors where material quality was sufficient for molecular analysis. In our group of LGG patients, we were able to detect molecular alterations in 318/342 patients (92.9%) using a combination of RT‐PCR, Sanger sequencing, and NGS techniques: KIAA1549::BRAF (n = 143/342, 41.81%), BRAF p.V600E (n = 59/342, 17.25%) together accounted for 59% of non‐NF1 pLGGs. Rare BRAF alterations, non‐canonical BRAF fusions (n = 14/342, 4.09%), and BRAF SNVs other than V600E (n = 7/342, 2%) accounted for an additional 6.14%. The next most common alterations were those affecting receptor tyrosine kinases—mainly FGFR1–3, including FGFR1::TACC1 (n = 9), KDD FGFR1 (n = 17), hot spot mutations in FGFR1 (n = 13), FGFR2 fusions (n = 12), and finally the rare FGFR3::TACC3 fusion (n = 3). The FGFR‐altered LGG subgroup comprised 16.08% of the pLGGs studied. Other tyrosine kinase rearrangements like ALK, ROS1, NTRK1–3, and RAF1 were rare (n = 18, 5.26%). SNVs in the KRAS gene were detected in four patients and a mutation in the PDGFRA gene in one patient. Non‐RAS/MAPK alterations such as MYB (n = 3) or MYBL1 (n = 1) and IDH1 SNVs (n = 11/342, 3.2%) were detected in a small proportion of our cohort.

3.1. BRAF Fusions

The most common BRAF fusion in pLGG was KIAA1549::BRAF. In our cohort, we detected nine different KIAA1549::BRAF exon‐exon junctions, including 16–9, 15–9, 16–11, 15–11, 13–11, 13–9, 10–9, 17–11, and 19–9, all resulting in the loss of the regulatory domain of the BRAF gene. The most common fusion was 16–9 (n = 71), the second was 15–9 (n = 43), and the third was 16–11 (n = 10). Other fusions were rare: 13–11 (n = 6), 15–11 (n = 6), 10–9 (n = 4), 13–9 (n = 1), 17–11 (n = 1), and finally 19–9 (n = 1). All 10–9 fusions were detected in spinal LGGs, and similarly, most KIAA1549 exon 13 fusions were also detected in spinal LGG patients (n = 4). The next two KIAA1549::BRAF fusions, 13–11, were seen in cerebellar LGGs, and one in an optic pathway tumor. The majority of the typical 16–9 fusions was seen in the posterior fossa region, similar to the 16–11 fusion. KIAA1549::BRAF–positive tumors occurred in patients at a younger age (median age of 5 years), were predominantly located in the posterior fossa, and they were predominantly pilocytic astrocytoma histologically.

We detected 13 different N‐terminal partners of the BRAF gene other than KIAA1549 (BCAS1, FAM131B, GNAI1, BCAS1, GTF21, MKRN1, NRF1, NUDCD3, PAG1, PRKAR2B, RNF130, TAX1BP1, and a novel EPHB2). Most of the fusion breakpoints occur on exon 9 of BRAF, and, in all fusions, the inhibitory regulatory domain located in the first six exons is disrupted by the fusion. Half of the non‐canonical BRAF fusions were seen in hemispheric LGGs (Figure 2).

FIGURE 2.

Pie charts showing the frequency of molecular alterations in pLGG without NF1 diagnosed between 2000 and 2024 according to tumor location.

3.2. BRAF SNVs

The second most frequent alteration in the LGG group was the BRAF V600E mutation (n = 59/342, 17.25%). BRAF V600E tumors were more common in older patients (median age of 10.14 years) and were more likely to be located in hemispheric regions. However, in contrast to KIAA1549::BRAF, they were rare in the cerebellum, and they included a spectrum of histologic subtypes. SNVs in the BRAF gene other than V600E were rare. The mutation BRAF T599dup (n = 3) indicates the insertion of a duplicate amino acid, threonine, into the protein kinase of the BRAF protein. The next rare BRAF SNV was G469A, which is a hotspot mutation within the protein kinase domain of BRAF.

3.3. Other Direct Members of RAS/MAPK

A further 2.34% (n = 8) of cases contained alterations in other direct members of the RAS/MAPK pathway, including four KRAS SNVs and four RAF1 fusions (SRGAP3::RAF1, QKI::RAF1, and FCHCD2::RAF1). KRAS is an upstream molecule in the RAS/MAPK pathway. Reports on the frequency of KRAS mutations in the pLGG range from 1%–5% and primarily arise in pilocytic astrocytoma [5, 26, 27, 28, 29]. Two of four patients with the KRAS mutation had hemispheric ganglioglioma, one had hemispheric PA, and the last one had spinal DLGNT (diffuse leptomeningeal glioneuronal tumor).

RAF1 fusions contain an intact RAF1 kinase domain (encoded by exons 10–17) and are functionally similar to BRAF fusions [30]. Three out of four patients with a RAF1 fusion had a PA, while one patient had a glioneuronal tumor.

3.4. FGFR1‐3 Alterations

FGFR1‐3 are receptor tyrosine kinases (RTKs) that play a key role in signal transduction via activation of their intra‐membranous tyrosine kinase domain (TKD) [26]. FGFR1 is the second most commonly altered gene in pLGG. FGFR1 alterations in pLGG arise via three mechanisms: FGFR1 mutations, FGFR1::TACC1 fusions, and FGFR1 kinase domain duplications [3, 5, 31, 32].

FGFR1‐3 alterations were detected in 54 patients (16.8%), with a median age of 8.69 years. Among the FGFR1 alterations, KDD FGFR1, which is typical for the diagnosis of DNET, was the most common (n = 17), with a median age of 6.78 years. The fusion gene FGFR1::TACC1 was detected in nine patients, with a median age of 10.4 years. FGFR1 hot spot mutations were observed in 13 patients (FGFR1 K656E or D, N546 D or K) with a median age of 10 years. In most cases, one causal change is typical for pLGG, except for FGFR1 SNPs, where we found either an additional FGFR1 SNV or a combination with a PIK3CA gene SNV (E545K or H1047R) in 3 of 13 patients.

In our pLGG cohort, we observed a relatively high frequency of FGFR2 fusion genes (n = 12), namely FGFR2::INA (n = 5), FGFR2::KIAA1598 (n = 3), FGFR2::ZCCHC24 (n = 2), FGFR2::CTNNA3 (n = 1), and FGFR2::PASD1 (n = 1). The median age of the patients with the FGFR2 fusion was 6.84 years. All FGFR2 rearranged patients had tumors located in the hemisphere (Figure 2).

In three patients, the FGFR3::TACC3 fusion gene was detected, which is generally rare in the pLGG population. All were hemispherically located.

3.5. Other RTK Fusions

Fusions in other RTK were rare in pLGG (5.26%), with a median age of 5.45 years. ALK, ROS1, NTRK, and MET fusions are typical in infantile HGG [33] and are also found in a variety of pediatric gliomas [16]. NTRK1‐3 genes (n = 9) had the highest abundance among RTK fusions: NACC::NTRK2, CLIP2::NTRK2, SPTAN1::NTRK2, KANK1::NTRK2, PDE4DIP::NTRK1, KIF21B::NTRK1, ETV6::NTRK3, and finally BCR::NTRK3. ALK fusions included ALK::GIGYF2, PPP1CB::ALK, and EML4::ALK. All ALK‐positive pLGGs were located in the hemisphere, in contrast to the other RTK fusions, which were located in different CNS compartments. Paradoxically, all but one ALK‐fused LGG were not infantile LGGs. The other two ALK fusion‐positive patients were older than 10 years. ROS1 fusions were represented by GOPC1::ROS1 and GIT2::ROS1. The first case was a congenital hemispheric tumor and the second a midline tumor in a 3‐year‐old girl.

3.6. IDH SNVs

IDH1 mutations are common in adult low‐grade gliomas. In pLGG, IDH1 mutations were rare, accounting for only 3.2% of cases. The IDH1 hot spot mutation was observed in 11 patients (R132H, R132C, R132G, R132S). Patients with IDH SNVs were diagnosed in late childhood (median age of 17.13 years), with the youngest patient diagnosed at 12.6 years. There was no difference in median age between IDH1 variants.

3.7. RNA‐NGS Analysis

The rearrangement was identified in 88 of 115 pLGG cases analyzed by RNA NGS (74.1%). Fusion‐gene detection was performed using three bioinformatic tools: Archer, Arriba, and STAR‐Fusion. Among these, STAR‐Fusion demonstrated the lowest sensitivity, detecting rearrangements in only 59 of 88 fusion‐positive cases (67%). In contrast, Arriba identified 86 of 88 rearrangements (97.77%), while Archer detected 78 of 88 rearrangements (88.6%). These findings highlight differences in detection efficiency among fusion callers, emphasizing the importance of tool selection in molecular diagnostics (Table 1).

TABLE 1.

Discordant results from three callers for the detection of fusion genes in the pediatric LGG population.

Dg.	Archer	Arriba	Star fusion	RT‐PCR
Astr. Gr2	KIAA1549::BRAF 16–11	KIAA1549::BRAF 16–11	X	KIAA1549::BRAF 16–11
PA	KIAA1549::BRAF 15–9	KIAA1549::BRAF 15–9	X	KIAA1549::BRAF 15–9
DA	KIAA1549::BRAF 15–9	KIAA1549::BRAF 15–9	X	KIAA1549::BRAF 15–9
PA	GNAI1::BRAF 1–10	GNAI1::BRAF 1–10	X	GNAI1::BRAF 1–10
PA	KIAA1549::BRAF 13–11	KIAA1549::BRAF 13–11	X	KIAA1549::BRAF 13–11
PA	X	KIAA1549::BRAF 15–9	X	KIAA1549::BRAF 15–9
PA	KIAA1549::BRAF 10–9	KIAA1549::BRAF 10–9	X	KIAA1549::BRAF 10–9
PA	KIAA1549::BRAF 16–9	KIAA1549::BRAF 16–9	X	KIAA1549::BRAF 16–9
Astr. Gr2	KIAA1549::BRAF 13–9	KIAA1549::BRAF 13–9	X	ND
GG	X	EPHB2::BRAF 9–10	EPHB2::BRAF	EPHB2::BRAF 9–10
PA	X	KIAA1549::BRAF 16–9	X	KIAA1549::BRAF 16–9
GG	X	NRF1::BRAF 10–10	X	ND
PA	KIAA1549::BRAF 15–9	X	X	KIAA1549::BRAF 15–9
PA	ETV6::NTRK3 5–15	ETV6::NTRK3 5–15	X	ND
PA	PDE4DIP::NTRK1 16–12	PDE4DIP::NTRK1 16–12	X	PDE4DIP::NTRK1 16–12
PA	NACC2::NTRK2 5–13	NACC2::NTRK2 5–13	X	NACC2::NTRK2 5–13
LGG NOS	X	PAG1::BRAF 5–9	PAG1::BRAF	PAG1::BRAF 5–9
GG	PPP1CB::ALK 5–20	PPP1CB::ALK 5–20	X	PPP1CB::ALK 5–20
GG	FGFR2::KIAA1598 17–7	FGFR2::KIAA1598 17–7	X	FGFR2::KIAA1598 17–7
PA	TAX1BP1::BRAF 5–9	TAX1BP1::BRAF 5–9	X	TAX1BP1::BRAF 5–9
DNET	X	truncat. MYBL1	X	ND
DNET	FGFR3::TACC3 17–10	FGFR3::TACC3 17–10	X	FGFR3::TACC3 17–10
GG	FGFR2::KIAA1598 17–7	FGFR2::KIAA1598 17–7	X	FGFR2::KIAA1598 17–7
DNET	FGFR2::ZCCHC24 17–2	FGFR2::ZCCHC24 17–2	X	FGFR2::ZCCHC24 17–2
PA	KIAA1549::BRAF 13–11	KIAA1549::BRAF 13–11	X	KIAA1549::BRAF 13–11
DNET	X	KDD FGFR1	X	KDD FGFR1
PA	X	KDD FGFR1	X	KDD FGFR1
PA	X	KDD FGFR1	X	KDD FGFR1
LGG	MKRN1::BRAF 4–11	X	X	MKRN1::BRAF 4–11
DNET	X	KDD FGFR1	X	KDD FGFR1
infant. glioma	GIT2::ROS1 16–36	GIT2::ROS1 16–36	X	GIT2::ROS1 16–36

The duplication of the kinase domain of the FGFR1 gene (KDD FGFR1) proved to be the most challenging and could only be detected using the Arriba caller.

In addition, the use of multiple callers was successful in detecting BRAF rearrangements, both KIAA1549::BRAF and non‐KIAA BRAF fusions. For the two LGGs with low‐quality RNA, Archer software failed to detect the KIAA1549::BRAF fusion, but the Arriba caller was successful. Similarly, for the rare non‐KIAA BRAF fusion genes, we were able to detect EPHB2::BRAF and PAG1::BRAF fusions using Arriba and the STAR‐Fusion caller. Another NRF1::BRAF fusion was again only detected with the Arriba caller. On the other hand, the MKRN1::BRAF fusion gene was not detected by either the Arriba caller or STAR‐Fusion, but it was detected by Archer. The result of Archer analysis was confirmed by RT‐PCR and subsequent Sanger sequencing (Figure 3).

FIGURE 3.

(A) Schematic visualization of the detected fusion transcript MKRN1::BRAF using the Archer software. (B) RT‐PCR and sequencing analysis of MKRN1::BRAF positive LGG.

Archer FusionPlex kits work on the principle of amplicon NGS and can be successfully used on archived FFPE material of lower quality. We were able to demonstrate a KIAA1549::BRAF fusion with exon 16–9 in a patient with recurrent PA 33 years after diagnosis. Using Archer FusionPlex Lung, we were subsequently able to demonstrate the same fusion in 33‐year‐old archived FFPE material, but, due to a very low QC score, the fusion was only detected by the Arriba caller. In this case, it is again well documented that even in low‐quality material, a combination of callers is needed to give us a better chance of detecting a causal fusion gene. Truncation of several genes, e.g., a MYBL1 truncating rearrangement, was also detected only by Arriba. In contrast, oncogenic variants of genes, such as EGFRvIII and MET exon 14 skipping, were more often missed by Arriba and STAR‐Fusion compared to Archer (data not shown). For the detection of these changes, it is advisable to use an additional caller for the detection of splice variants. The Venn diagram (Figure 4) schematically summarizes the success rate of the individual callers used.

FIGURE 4.

Venn diagram. The Venn diagram schematically summarizes the success rate of the individual callers used.

All discordant rearrangements were confirmed by RT‐PCR followed by Sanger sequencing. Similarly, novel or rare fusions were confirmed when we first designed primers. After RT‐PCR, the final product was verified by Sanger sequencing (Figure 5).

FIGURE 5.

(A) Schematic visualization of PAG1::BRAF fusion transcript detection with Arriba software (https://github.com/suhrig/arriba/). (B) RT‐PCR and sequencing analysis of PAG1::BRAF positive LGG.

4. Discussion

With the recent 5th edition of the WHO classification for tumors of the CNS, which emphasizes the role of molecular diagnostics, the detection of fusion genes has become an important diagnostic marker in pediatric CNS neoplasms [34]. Based on our experience with LGG molecular diagnostics, we developed a cost‐effective algorithm that combines targeted testing (RT‐PCR or Sanger sequencing for the most common alterations) with NGS multicalling in negative cases (Figure 1).

Our findings confirm that the majority of genetic events in pLGG converge on the MAPK signaling pathway. The most prevalent were KIAA1549::BRAF fusions (41.8%), BRAF V600E mutations (17.25%), and FGFR1 alterations (16.8%). RNA sequencing revealed nine distinct KIAA1549::BRAF exon–exon junctions as well as 15 rare or novel BRAF fusions, including intrachromosomal rearrangements on chromosome 7 (NUCD3::BRAF, MKRN1::BRAF, FAM131B::BRAF, TAX1BP1::BRAF, PRKAR2B::BRAF, NRF1::BRAF) and translocation‐derived fusions such as RNF130::BRAF [t(5;7)], BCAS1::BRAF [t(7;20)], PAG1::BRAF [t(7;8)], and GTF21::BRAF [t(7;7)] [13, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44]. Notably, we report for the first time an EPHB2::BRAF fusion in a pediatric brain tumor. The EPHB2 gene encodes a member of the Eph‐receptor family of receptor tyrosine kinases, which are composed of an N‐terminal ligand‐binding domain, a transmembrane region, and an intracellular kinase domain. Fusion with EPHB2 has previously been described with other partners, such as NTRK1 [45, 46], PDZD4 [46] or NBPF3 [47], ASAP3 [45] or INTS11 [48]. Similar to other BRAF fusions, the resulting protein EPHB2::BRAF lacks the N‐terminal inhibitory domain, leading to constitutive BRAF kinase activity and enhanced MEK/ERK signaling.

Detection of BRAF fusions required the combined use of multiple bioinformatic callers, as lower transcript expression frequently led to false negatives. Arriba proved most reliable, missing only 2 of 47 fusion‐positive cases compared with 4 missed by Archer and 14 by STAR‐Fusion. Failures occurred across both intrachromosomal and translocation‐derived fusions, including canonical KIAA1549::BRAF, underscoring the importance of a multicaller strategy. Even canonical fusions like KIAA1549::BRAF can be difficult to detect, especially in suboptimal samples. Determining an optimal cut‐off for fusion‐gene detection is challenging, especially when dealing with suboptimal FFPE samples with degraded RNA. The QC is determined by Archer Analysis, so that the number of unique RNA start sites for the GSP2 control must be greater than 10. In our cohort of pLGG patients, we detected the fusion genes by Archer Analysis, even in 10 cases where the sample failed QC (seven cases of KIAA1549::BRAF, 1 case of GNAI1::BRAF, 1 case of NACC2::NTRK2, and 1 case of PPP1CB::ALK). This fact clearly indicates the importance of using an additional caller, since in all cases, the fusion gene was detected at least by the Arriba caller and RT‐PCR.

Beyond fusions, we also detected rare BRAF SNVs. In addition to V600E, approximately half of which occurred in hemispheric tumors, we observed T599dup (n = 3), which enhances MEK/ERK phosphorylation, and G469A, a known hotspot mutation that activates the MAPK pathway independently of RAS [49, 50, 51, 52].

The second most common alterations in our cohort were FGFR fusions/SNVs, with an overall prevalence of FGFR alterations of 16.8%—notably higher than the 6.1% reported by Ryall et al. [4]. This may reflect the higher proportion of patients from the epilepsy program, as epilepsy‐associated tumors are enriched for FGFR gene alterations. FGFR1 alterations occurred via three mechanisms: kinase domain duplication (KDD), oncogenic fusions with TACC partners, and hotspot mutations [3, 5, 31, 32]. FGFR1 KDD, the most frequent, results in constitutive kinase activation [3, 53], while FGFR::TACC rearrangements also promote ligand‐independent activation via dimerization domains and have been reported across CNS tumors, including extraventricular neurocytoma, glioblastoma, IDH‐wildtype gliomas, and pLGG [3, 54, 55, 56]. In our cohort, we identified nine FGFR1::TACC1 and three FGFR3::TACC3 fusions, as well as 12 cases with FGFR2 alterations, spanning polymorphous low‐grade neuroepithelial tumor of the young (PLNTY—7 cases), dysembryoplastic neuroepithelial tumor (DNT—2 cases), ganglioglioma (GG—2 cases), and multinodular and vacuolating tumor (MVNT, 1 case), confirming their heterogeneous histopathological spectrum and overall benign clinical behavior [28, 29, 57, 58]. We also detected five FGFR1‐mutant tumors, involving recurrent N546K and K656E variants, both of which increase kinase activity and confer oncogenic potential [59, 60, 61]. Prognostically, FGFR1 KDD and FGFR1/FGFR2 fusions are generally linked to favorable outcomes, whereas FGFR1 mutations portend intermediate risk and inferior survival [53, 62]. From a technical standpoint, some FGFR alterations, particularly KDD events, are missed by individual callers, reinforcing the need for multicaller pipelines and splice‐aware algorithms. Clinically, their accurate identification remains essential both for diagnosis and for potential eligibility for FGFR‐targeted therapies.

Rare alterations in our cohort further expanded the genetic spectrum of LGG. We identified four RAF1 fusions, including a novel FCHSD2::RAF1, in addition to SRGAP3::RAF1 and QKI::RAF1. The breakpoint in FCHSD2::RAF1 juxtaposed exon 8 of RAF1 with exon 13 of FCHSD2, an adaptor protein involved in clathrin‐mediated endocytosis, resulting in constitutive RAF1 kinase activity. Although rare, RAF1 fusions have been described in pilocytic astrocytoma and several extracranial tumors and consistently activate MAPK and PI3K pathways [63, 64, 65]. Their sensitivity to RAF and MEK inhibitors highlights their clinical significance.

We also detected nine LGGs with NTRK1–3 rearrangements, including two previously undescribed fusions: BCR::NTRK3 and SPTAN1::NTRK2. The BCR::NTRK3 fusion retained the oligomerization domain of BCR together with the tyrosine kinase domain of NTRK3, a structure consistent with known oncogenic NTRK fusions and analogous to the BCR::NTRK2 fusion described by Jones et al. [66]. The SPTAN1::NTRK2 fusion joined exon 55 of SPTAN1 with exon 14 of NTRK2 (Figure 6); SPTAN1 has previously been described as a fusion partner in T‐ALL with ABL1 [67]. Both novel fusions extend the spectrum of NTRK‐driven pLGG.

FIGURE 6.

Schematic visualization using Arriba software (https://github.com/suhrig/arriba/) of the detected SPTAN1::NTRK2 fusion transcript.

5. Conclusion

Our study underscores that the genetic landscape of pediatric LGG is dominated by alterations converging on the MAPK pathway, with BRAF, FGFR, RAF1, and NTRK events representing key diagnostic and therapeutic targets. Molecular stratification is no longer optional but essential, as effective inhibitors against BRAF p.V600E and MEK are already in clinical use, and agents directed at FGFR and other tyrosine kinases are rapidly emerging. Accurate detection of oncogenic fusions is therefore central to diagnosis, prognostication, and treatment selection. Implementing optimized diagnostic algorithms with robust multicaller strategies will be critical to reliably distinguish true driver events from benign fusions or technical artifacts, ultimately enabling precision medicine for all patients with LGG.

Author Contributions

Petr Brož: writing – original draft, investigation. Martina Strnadová: methodology‐NGS, editing. Denisa Olejníková: methodology‐NGS, editing. Johana Kotiš: methodology‐NGS, editing. Tereza Pospíšilíková: methodology‐NGS, editing. Adéla Mišove: PCR, RT‐PCR, editing. Miroslav Koblížek: morphology, review and editing. Josef Zámečník: morphology, review and editing. Michal Zápotocký: clinical data, review and editing. David Sumerauer: clinical data, review and editing. Aleš Vícha: investigation, review and editing. Martin Kynčl: imaging processes, review and editing. Petr Libý: neurosurgery, editing. Vladimír Beneš: neurosurgery, editing. Lenka Krsková: writing – original draft, investigation, validation, supervision. All authors read and approved the final manuscript.

Ethics Statement

The present study was approved by the ethics committee of the University Hospital Motol (reference no. EK‐97/25) and adhered to the tenets of the Declaration of Helsinki.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Primer table for verification of NGS results.

Data S1: Supporting Information.

Brož P., Strnadová M., Olejníková D., et al., “Molecular Landscape of Pediatric Low‐Grade Gliomas: Insights From RNA‐NGS and Bioinformatic Analysis,” Genes, Chromosomes and Cancer 64, no. 10 (2025): e70085, 10.1002/gcc.70085.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.