Alterations in ALK/ROS1/NTRK/MET drive a group of infantile hemispheric gliomas

Infant gliomas have paradoxical clinical behavior compared to those in children and adults: low-grade tumors have a higher mortality rate, while high-grade tumors have a better outcome. However, we have little understanding of their biology and therefore cannot explain this behavior nor what constitutes optimal clinical management. Here we report a comprehensive genetic analysis of an international cohort of clinically annotated infant gliomas, revealing 3 clinical subgroups. Group 1 tumors arise in the cerebral hemispheres and harbor alterations in the receptor tyrosine kinases ALK, ROS1, NTRK and MET. These are typically single-events and confer an intermediate outcome. Groups 2 and 3 gliomas harbor RAS/MAPK pathway mutations and arise in the hemispheres and midline, respectively. Group 2 tumors have excellent long-term survival, while group 3 tumors progress rapidly and do not respond well to chemoradiation. We conclude that infant gliomas comprise 3 subgroups, justifying the need for specialized therapeutic strategies.

G liomas are the most common primary central nervous system (CNS) neoplasm and result in the highest tumorassociated morbidity and mortality in children and adults 1,2 . Traditionally, gliomas are divided into low grade (LGG, WHO grades I-II) and high grade (HGG, WHO grades III-IV) based on their histological characteristics 3 . Molecularly, adult lower grade gliomas commonly harbor alterations in IDH1/2 in association with TP53 and ATRX mutations or TERT mutations and 1p/19q co-deletions 4 . In comparison, most childhood LGG are driven by RAS/MAPK activation-predominantly in the form of fusions or mutations involving the BRAF gene-and rarely undergo malignant transformation [5][6][7] . In contrast, adult LGG rarely contain RAS/MAPK alterations 8 and invariably transform to HGG over time 9 . Pediatric HGG are usually not the result of transformation from LGG and, in contrast to adult HGG, most commonly harbor recurrent mutations in the genes encoding histone H3.3 and H3.1 10,11 .
In contrast to the abundance of genetic and clinical information now available for pediatric glioma, far less is known about the infant demographic (under 1 year of age), despite the incidence of CNS tumors being highest in this group 1 . Although steady improvements in the overall outcome of childhood cancer have been observed globally, infants with brain tumors remain at high risk for early death after diagnosis, are less likely to be enrolled in clinical trials and are critically under-studied 12 . Further, the association between tumor grade and outcome is less predictable in infants; infant LGG (iLGG) show a more aggressive course [13][14][15] , while infant HGG (iHGG) have a better outcome 16,17 when compared with older children and adolescents. As such, the classic treatment approaches for pediatric LGG (low dose chemotherapy) and HGG (surgery, radiation and alkylator-based chemotherapy) are often either ineffective or excessive, respectively. Therefore, clinicians caring for infants with gliomas are faced with the challenging task of treating an exceptionally vulnerable population of patients where the best treatment options remain ambiguous and data are scarce.
To address the lack of knowledge regarding the genetic underpinnings of infant gliomas, we assemble a multi-institutional, international collaborative taskforce to comprehensively characterize a large, clinically well-annotated cohort with followup data spanning three decades. We find that infant gliomas comprise three main subgroups: (1) hemispheric receptor tyrosine kinase (RTK)-driven tumors, including ALK, ROS1, NTRK, and MET fusions, which are enriched for HGG and have an intermediate clinical outcome, (2) hemispheric RAS/MAPK-driven tumors, which show excellent long-term survival with minimal clinical intervention post-surgery, and (3) midline RAS/ MAPK-driven tumors, which are enriched for LGG with BRAF alterations and have a relatively poor outcome even after conventional chemotherapeutic approaches. Together the clinical and molecular features of each subgroup indicate age-specific mechanisms underlying tumor initiation. This suggests that updated clinical approaches are required to modernize treatment and improve the outcome of these infants.
Infantile gliomas comprise three subgroups. Analysis of the clinical features associated with each class of molecular alterations suggested that infant gliomas represent three distinct clinical/molecular groups: (1) Hemispheric, RTK-driven,  (Table 1).  (Fig. 4b, c). Five-year OS was 53.8, 25.0, and 42.9% for ALK, ROS1, and NTRK fused tumors respectively, although the numbers in each group were small (12, 8, 7, respectively). Interestingly, when compared with ALK-driven HGG, low-grade ALK gliomas tended to be diagnosed at an older age (median = 5.0 versus 1.6 months) and showed a better clinical outcome; all patients with ALK-fused LGG (n = 5) were alive at a median follow-up of 5 years (range, 1.4-7.2 years), whereas 42.9% (3/7) patients with ALK-fused HGG were deceased at a median follow-up of 3 years (range 0.01-8.55 years). Interestingly, in two patients with NTRK-fused HGG that underwent a second resection post-chemotherapy, tumor from the second resection had lower grade histology, suggesting that Group 1 tumors may comprise an LGG/HGG continuum and/or have the potential to differentiate and slow their growth over time (Fig. 4d).

Discussion
In this study we comprehensively characterize the landscape of genetic drivers and their clinical impact, revealing 3 subgroups of infant glioma (Fig. 6). Group 1 tumors are enriched for ALK/ ROS1/NTRK/MET fusions, alterations analogous to those detected in adult carcinomas such as non-small cell lung cancer 20  and colorectal cancer 22 . Despite similar conservation of the tyrosine kinase domain and region of breakpoints, for most cases the binding partners identified in infant gliomas differ from those in other malignancies. Interestingly, ETV6-NTRK3 can also be detected in other congenital tumors (congenital mesoblastic nephroma and congenital fibrosarcoma), suggesting a common age-specific mechanism. With the exception of NTRK fusions, which were previously shown to be enriched in non-brainstem infant HGG 23 , these alterations have been rarely reported in gliomas and this study provides a comprehensive explanation for the isolated case reports of ALK 24,25 and ROS1 26 fusions in pediatric glial tumors. Indeed, these alterations are recurrent and define, together with NTRK, Group 1 hemispheric infant gliomas. Examination of the clinical data in these cases reveals several interesting facts: (1) their overall survival is good compared with that of older children with HGG and if they survive past two years, almost none progress further; (2) cases where a second surgery was done post-chemotherapy show differentiation and decreased proliferation of the tumor; and (3) cases with LGG histology tend to occur in older infants. Taken together, these observations suggest the capacity for differentiation over time in Group 1 tumors, perhaps, as in other pediatric gliomas, through oncogene induced senescence [27][28][29][30] . Alternatively, as seen in neuroblastoma, a common infant tumor that harbors ALK alterations, inherent maturation (also a part of normal development) may explain the morphological and clinical "maturation" of some iHGG into iLGG [31][32][33] . This has important implications for our therapeutic approach as it suggests that if we can use nonmorbid treatment options, which may include targeted kinase inhibitors, to get them through the rapid growth phase of their tumor, their long-term outlook may be positive. Since infantile gliomas are mostly single-driver tumors, unlike adult lung and colorectal cancers, they are particularly suitable for precisionmedicine treatment approaches. Several ALK inhibitors have either already shown efficacy or are in clinical trials for ALKdriven tumors in children, including Crizotinib 34 and Ceritinib, and the newer generation inhibitors with enhanced blood-brain barrier penetration Lorlatinib and Ensartinib. The NTRK inhibitor Larotrectinib has also shown antitumor activity in pediatric patients with NRTK-fused tumors regardless of age or histology [35][36][37] . For example, in the NAVIGATE Phase 2 trial, Larotrectinib treatment resulted in a significant decrease in tumor volume in a 35-year-old woman with glioblastoma 38 . In the STARTRK1 trial, Drilon et al. 39 report a pontine astrocytoma harboring an NTRK fusion that showed tumor volume reduction upon treatment with Entrectinib, a tyrosine kinase inhibitor known to target NTRK, ALK and ROS1. These encouraging results have led to a current phase I/Ib study being conducted in pediatrics to evaluate Entrectinib in primary CNS tumors (NCT02650401), which includes NTRK, ROS1, and ALK fused tumors. Results thus far are promising 40 .
Group 2 hemispheric RAS/MAPK tumors have an excellent long-term survival and often require only surgery, suggesting that a safe resection and a careful "watch and wait" postsurgical strategy is appropriate for these patients. Group 3 represents midline LGG enriched for RAS/MAPK alterations. The lack of HGG histology, such as that observed in the pons or thalami of older children, and the lack of histone mutations in this age group suggest distinct tumor-and/or host-related factors underlying tumor development. In older children, BRAF fused-tumors tend to have favorable outcome 5,41,42 and a good response to conventional therapy. Strikingly, most Group 3 tumors, especially OPHG, progressed regardless of BRAF fusion or mutation status. The poor outcome of BRAF-fused midline tumors in infants is surprising and in stark contrast to the biological behavior of similar tumors in older children. This disparity may be related to age-specific genetic, tumor or microenvironment factors that are, at this point, poorly understood. As such, there is little or no room for "watch and wait" and a biopsy should be performed upfront to ascertain BRAF status and systemic therapy initiated readily thereafter. Given the multiple progressions typically observed with conventional chemotherapy and the encouraging results of targeted BRAF/MEK inhibitors in pLGG 43,44 , these patients should be prioritized for targeted therapies early after initial diagnosis.
Whereas future studies will certainly further characterize infant gliomas, our study broadens our understanding of cancers early   3. PrimePCR ddPCR mutation assay H3F3A WT/G35R for p.G35R, Human (unique assay ID: dHsaMDS720957813). 4. Prime PCR ddPCR copy number assay CDKN2A, Human unique assay ID: dHsaCP1000581) and reference prime PCR ddPCR copy number assay APB31 (unique assay ID: dHsaCP2500348). A known homozygous deleted cell line was used as a zero-copy control, whereas an Ontario Population Genomics Platform healthy control sample (ID: 85751) obtained from The Center of Applied Genomics at SickKids was used as a two-copy control. Samples that showed < 1.2 copy number value as calculated from the total target and reference event number were considered deleted. 5. FGFR1 TKD is a custom assay design 46

34.1%
LGG LGG HGG 0-3 3-6 6-9 9-12 0-3 3-6 6-9 9-12 0-3 3-6 6-9 9-12 LGG HGG/ LGG the negative control spikes included in each run. This was followed by a technical normalization using the four housekeeping transcripts included in each run (ABCF1, ALAS1, CLTC, and HPRT1). Data is viewed using a box plot and the extreme statistical outlier (3X the interquartile range (IQR)) method was used to detect the presence of an expressed fusion. Panel 2: To account for an evolving knowledge of fusions described in gliomas, a second NanoString fusion panel was designed. The fusion targets included on this panel are listed in Supplementary Table 3. In addition to fusion targets, three reporter targeting systems were also included targeting ALK, ROS1, and NTRK2. These reporter systems work by adding multiple sequence tags prior to and after the exons of well-defined breakpoint hotspots. In the event of a breakpoint, the reads from the nCounter appear significantly different between adjacent sequence tags, allowing for the identification of a likely fusion event with an unknown partner. Samples were tested for fusion gene expression with the NanoString nCounter Low Grade Glioma Panel 2 as described above. CodeSet probe sequences for Panel 1 and 2 are proprietary, but available from NanoString Technologies (Seattle, WA) and the Hospital for Sick Children upon request.
Fluorescent in situ hybridization. Fluorescent in situ hybridization (FISH) analysis was performed on formalin fixed paraffin embedded 4-μm tumor sections using a dual color breakapart probe for the ALK gene (Empire Genomics, Buffalo, NY). Slides were baked overnight to fix the section to the slide and were pretreated by using a paraffin pretreatment kit (Abbott, Chicago, IL). Sections were dehydrated before slide/probe co-denaturation on thermobrite (Intermedico, Markham, ON). Denaturation conditions used for paraffin-embedded slides/probes were as follows: 1. 83°C for 7 min 2. 37°C overnight Slides were washed in 0.4x Saline-sodium citrate(SSC)/0.3% NP-40 at 65°C for 30 s, followed by 2x SSC/0.1% NP-40 at room temperature for 30s. Slides were counterstained with DAPI. Nuclei were analyzed by using an Axioplan2 epifluorescence microscope (Zeiss, Jena, Germany). Images were captured by an Axiocam MRm Camera (Imaging Associates, Bicester, United Kingdom) and analyzed by using an imaging system with Isis Software (Version 5.1.110; MetaSystems, Boston, MA).
Copy number analysis. The OncoScan FFPE Assay Kit (Affymetrix, Santa Clara, CA, USA) was used to assess copy number and loss of heterozygosity events in selected samples that remained uncharacterized by the targeted methods described above. Samples for this assay were sent to the Genome Quebec Innovation Centre for completion of the analysis. The OncoScan FFPE Assay Kit (Affymetrix, Santa Clara, CA, USA) was used according to manufacturer's specifications and sample preparation, including digestion, labelling, quality checks, hybridization, and scanning was performed at the Genome Quebec Innovation Centre. Data was analyzed using the Chromosome Analysis Suite (ChAS) (ThermoFisher Scientific, CA, USA) and copy number calls based on normalized data.
Targeted RNA sequencing. TruSight Sequencing Panel: Samples with sufficient RNA for sequencing had their total RNA constructed into RNA-sequencing libraries using the Illumina TruSight RNA Pan-Cancer Panel Kit (Illumina, San Diego, CA), following the manufacturer's guidelines. cDNA generation was completed by random priming during first and second strand synthesis, followed by 3′ end adenylation. Sequencing adapters were then ligated to the fragments to allow for amplification of the cDNA followed by a validation step to ensure proper adapter ligation. Samples were then hybridized to specific target probes used to enrich for cancer-associated genes outlined in the manufacturer's documentation. Paired-end RNA-sequencing was performed using the NextSeq 550 (Illumina, San Diego, CA), sequencing platform. Raw sequencing data was converted to fastq files and analyzed using the BaseSpace application (Illumina, San Diego, CA) with RNA-Seq Alignment V.1.0.0. Variant calling was completed in BaseSpace using the Isaac Variant Caller 48 while structural rearrangements were identified using Manta 49 and TopHat 50 .
Whole-transcriptome sequencing. Samples with sufficient RNA quality and quantity were sent for whole transcriptome sequencing at The Center for Applied Genomics (Hospital for Sick Children, Toronto, ON). Library preparation was completed using the TruSeq RNA Library Prep Kit v2 (Illumina, San Diego, CA) using the rRNA depletion kit RiboZero Gold (Illumina, San Diego, CA) according to the manufacturer's specifications. Paired-end sequencing was performed on the Illumina HiSeq 2500 platform. STAR 51 was used to align the raw sequencing data to genome reference "Homo sapiens UCSC hg19". Fusion events were called using four fusion callers: defuse 52 , tophat 50 , ericscript 53 , and fusionmap 54 .
DNA methylation analysis. Methylation profiling was completed at the microarray centre at the Centre for Applied Genomics at the Hospital for Sick Children (Toronto, Canada). Bisulphite conversion was completed using the EZ DNA Methylation kit (Zymo Research) according to the manufacturerʼs guidelines. Genome-wide DNA methylation patterns were analyzed using the HumanMethylation450 BeadChip platform according to manufacturer specifications (Illumina, San Diego, CA). Raw data underwent quality control and preprocessing using the R package "minifi" 55 and normalized using the R package "noob" 56 . Probes with a SNP at or near the CpG, plus those on the X and Y chromosomes were removed. t-SNE plots were completed using the R package "t-SNE" 57 . Raw.idat files are available at at the GEO wesbite under the ascension code GSE135017.
In vivo: all in vivo studies were reviewed and accepted by the Animal Care Committee at The Centre for Phenogenomics (Toronto, ON), an affiliate of the Hospital for Sick Children (Toronto, ON). For the intracranial orthotopic in vivo model, 200,000 iNHA mcherry EV, CCDC88A-ALK or PPP1CB-ALK cells were injected in the brain hemispheres of age (8-10 weeks) and sex-matched NOD/scid/ gamma (NSG) mice randomly assigned to either a control or experimental group. Animals were independently monitored by a third party and euthanized at humane endpoints when physiological signs of a brain tumor (hunched posture, scruffy appearance, weight loss, etc.) were detected or at 6 months post injection for the control group. CNS samples were collected at endpoint and evaluated histologically for tumors by The Centre for Phenogenomics (Toronto, ON.) histology core.
In vitro proliferation assay. Cells were seeded at 10,000 cells/well in a 96-well plate and allowed to adhere for 48 h. 20 µl of Alamar Blue (Thermo Fisher, CA., USA) was added to each well and the plates incubated at 37°C and 5% CO2 for 4 h. The fluorescence intensity was measured using a Spectramax Gemini plate reader (Molecular Devices, San Jose, CA, USA) using an excitation wavelength of 530 nm and an emission wavelength of 580 nm. Analysis was completed by normalizing intensity values against wells containing media alone. Data was represented as the mean of each condition. No detectable batch effect was observed.
In vitro drug dose assay. Ceritinib (LDK-378) and Crizotinib (PF-02341066) were purchased from Selleckchem.com and prepared according to the manufacturer's guidelines. Drugs were diluted in DMSO to the defined concentrations. Cells were seeded at 5,000 cells/well in a 96-well plate and allowed to adhere for 24 h. After 24 h, the appropriate drug concentration was added to each well and the plates incubated at 37°C and 5% CO2 for 48 h. The fluorescence intensity was measured using a Spectramax Gemini plate reader (Molecular Devices, San Jose, CA, USA) using an excitation wavelength of 530 nm and an emission wavelength of 580 nm. Analysis was completed by normalizing intensity values against wells containing media alone. Data was represented as mean of each condition. No detectable batch effect was observed.
Immunohistochemistry. ALK and FLAG immunostaining was performed using 10 μm-thick sections of the tumor samples post de-parafinization. Antigen retrieval was performed in a citrate buffer (pH 6.0) for 5 min prior to peroxidase quenching with 3% hydrogen peroxide (H2O2) in PBS for 10 min. The sections were then washed in water and pre-blocked with a normal goat or horse serum for 1 h. Next, the tissue sections were incubated overnight at 4°C in primary antibody: MIB-1, Synaptophysin and GFAP immunohistochemistry was performed on a Benchmark Ventana Machine (Tucson, AZ) using the Optiview detection kit (Tucson, AZ). CC1 was used for heat retrieval for 40 min. Tissue sections were incubated with primary antibody for thirty-six minutes: RTU anti-MIB-1 (mouse monoclonal primary antibody, GA626, ready-to-use, Dako Omnis, Santa Clara, CA, USA).
After washing the sections with PBS, they were incubated with secondary antibodies (1:100) for 1 h. The Mouse on Mouse Polymer IHC kit (Abcam, Cambridge, UK) was used via the manufacturer guidelines prior to image acquisition to mitigate cross-reactivity and improve sensitivity for antibodies raised in mice. Finally, the sections were developed with diaminobenzidine tetrahydrochloride substrate for 10 min, and counterstained with hematoxylin. Pictures were obtained using a Nikon E600 microscope (Nikon, Canada).
Statistics. Statistical analyses were performed using R version 3.5.0 and R Commander Version 2.4-4 with the plugins "Survival" (version 1.2-0), "KMggplot2" (version 0.2-5) and "Plot by Group" (version 0.1-0). PFS was defined as the time between diagnosis and tumor progression requiring a change in clinical management. OS was defined as the time from diagnosis until death or last follow up for the patients still alive. Estimations of survival were calculated using the Kaplan-Meier method and log rank test, p values below 0.05 were considered significant. 5 and 10 year survival is reported as a percentage with 95% confidence intervals. Univariate and multivariate analysis was performed using SPSS v25 (IBM Corporation). This was done using a univariate or multivariate Cox proportional hazards model and significance testing (α = 0.05) based on the Wald test.
Source data. Uncropped and unedited gels and blots are contained within Supplementary File Source Data. Raw clinical features used for survival plots and prognostic analysis are also included in this file.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The targeted and whole transcriptome sequencing data sets have been deposited in the European-Genome-phenome Archive under accession code EGAS00001003714. The methylation data is available from the GEO website under the accession code GSE135017. All the other data supporting the findings of this study are available within the article, its supplementary information files and from the corresponding author upon reasonable request. A reporting summary for this article is available as a Supplementary Information file.