Molecular Profiling of Hard-to-Treat Childhood and Adolescent Cancers

Key Points Question Can genome sequencing facilitate the molecular profiling of the patient's tumor to identify actionable and targetable alterations? Findings In this diagnostic study of 62 consecutive pediatric patients with hard-to-treat cancer who were enrolled in the TRICEPS study, incorporating multimodal genomic sequencing, including RNA sequencing, into the management of refractory or relapsed childhood and adolescent cancers identified potentially actionable alterations in 54 (87%) of patients. Meaning Molecular profiling may enable the identification of potentially actionable alterations with clinical implications for most patients tested, including targeted therapy and clinically relevant information of diagnostic, prognostic, and monitoring significance.

This supplementary material has been provided by the authors to give readers additional information about their work.

Study design and participants
Children and adolescents with refractory or recurrent cancer and who were less than 18 years at initial diagnosis were considered eligible for the study. Exclusion criteria included estimated life expectancy of less than 3 months, as assessed by the treating oncologist, and insufficient or low-quality tumor samples. From April 2014 to February 2018, 85 consecutive patients with either a solid, brain or hematological neoplasms with poor prognosis were eligible. The institutional review board approved the research protocol and written informed consent was obtained from all participants and their parents or legal guardians. It is important to note that such studies, in patients with a very reserved prognosis, raise many ethical issues, including the risk of giving "false hope", the risk related to the biospy that would not result in any change in treatment, delaying the decision of comfort care, spending more time in the hospital, and the risk of incidental findings. A companion ethical study was designed and conducted to explore these issues in depth (Janvier et al, in preparation). Clinical and demographic data of patients, including age, sex, disease status at enrolment, were collected at time of enrolment. Biological material as well as patient data were stored in our institutional biobank and database.

Sample and clinical data collection
Following consent, normal and tumor patient samples were obtained. For solid and brain tumors, 3 ml of peripheral blood (Vacutainer blood EDTA tubes) or Saliva (ORAGENE OG-SOO kits) was collected as normal material. Since the protocol did not mandate biopsy for research purposes only, tumoral tissue was obtained following biopsy or resection following routine clinical procedures. When fresh or frozen tissue was not available, Formalin-Fixed Paraffin-Embedded (FFPE) blocks were used instead. Clinical pathologists reviewed all tumor specimens to determine tumor cell content and overall quality of the specimen. Genomic profiling required at least 5 mg of tumor tissue and >25% tumor content. Decalcified specimens were considered inadequate for tumor profiling. For hematologic malignancies, saliva was collected using the ORAGENE OG-5OO or SC-2 kits as normal material. Cancer cells were obtained from either bone marrow, pleural fluid or peripheral blood (Vacutainer blood EDTA tubes). Leukemia samples with more than 25% blasts and at least 100 000 cells were considered suitable for molecular profiling. Genomic DNA and total RNA were extracted from the patient's tumor and normal cells using mini or micro AllPrep DNA/RNA kits from Qiagen, or ORAGENE OG-250/500 or CS-2 protocols for saliva specimens.

Whole exome sequencing (WES)
Bioinformatic analysis was performed as described elsewhere 1 . Details of pipelines used for bioinformatics analysis are given in Figure S1 (supplementary materials). Briefly, the resulting exome reads were aligned to the hg19 (GRch37) reference genome using BWA (version 0.7.7) 2 . Picard (http://picard.sourcefourge.net) was used to remove duplicate mappings, calculate metrics and manipulate SAM/BAM files. Base quality score recalibration and local realignment of reads around small insertions/deletions (InDels) were performed using the Genome Analysis ToolKit (GATK Version 3.3) 3 . SNVs and InDels were called using Varscan2 4 and MuTect (https://www.nature.com/articles/nbt.2514). The sequencing information from the corresponding germline genome was used to confirm the somatic status of the mutations. The tumor specific SNVs and indels were considered validated if detected by both WES and transcriptome analysis, otherwise they were confirmed by targeted sequencing (> 1000x coverage) on a MiSeq Illumina system (at the McGill University and Genome Quebec Innovation Center). CNAs were detected, by selecting off-target reads to simulate a low coverage WGS (Sinnett, unpublished results) and then using the R package QDNAseq 5 .
Validation of CNAs was done by qPCR. Tumor mutation burden (TMB), defined as the rate of SNVs per megabase, was determined for all tumors.

Whole Transcriptome sequencing (RNA-seq)
Alignment to the hg19 (GRCh37) genome reference was performed using STAR aligner 6 .