Intratumour heterogeneity is a major cause of treatment failure in cancer. We present in-depth analyses combining transcriptomic and genomic profiling with ultra-deep targeted sequencing of multiregional biopsies in 10 patients with neuroblastoma, a devastating childhood tumour. We observe high spatial and temporal heterogeneity in somatic mutations and somatic copy-number alterations which are reflected on the transcriptomic level. Mutations in some druggable target genes including ALK and FGFR1 are heterogeneous at diagnosis and/or relapse, raising the issue whether current target prioritization and molecular risk stratification procedures in single biopsies are sufficiently reliable for therapy decisions. The genetic heterogeneity in gene mutations and chromosome aberrations observed in deep analyses from patient courses suggest clonal evolution before treatment and under treatment pressure, and support early emergence of metastatic clones and ongoing chromosomal instability during disease evolution. We report continuous clonal evolution on mutational and copy number levels in neuroblastoma, and detail its implications for therapy selection, risk stratification and therapy resistance.

Liquid biopsy strategies in pediatric patients are challenging due to low body weight. This study investigated cfDNA size distribution and concentration in blood, bone marrow, cerebrospinal fluid, and urine from 84 patients with neuroblastoma classified as low (n = 28), intermediate (n = 6), or high risk (n = 50) to provide key data for liquid biopsy biobanking strategies. The average volume of blood and bone marrow plasma provided ranged between 1 and 2 mL. Analysis of 637 DNA electropherograms obtained by Agilent TapeStation measurement revealed five different major profiles and characteristic DNA size distribution patterns for each of the biofluids. The proportion of samples containing primarily cfDNA was, at 85.5%, the highest for blood plasma. The median cfDNA concentration amounted to 6.28 ng/mL (blood plasma), 58.2 ng/mL (bone marrow plasma), 0.08 ng/mL (cerebrospinal fluid), and 0.49 ng/mL (urine) in samples. Meta-analysis of the dataset demonstrated that multiple cfDNA-based assays employing the same biofluid sample optimally require sampling volumes of 1 mL for blood and bone marrow plasma, 2 mL for cerebrospinal fluid, and as large as possible for urine samples. A favorable response to treatment was associated with a rapid decrease in blood-based cfDNA concentration in patients with high-risk neuroblastoma. Blood-based cfDNA concentration was not sufficient as a single parameter to indicate high-risk disease recurrence. We provide proof of concept that monitoring neuroblastoma-specific markers in very small blood volumes from infants is feasible.

Chimeric antigen receptor (CAR) T cell therapy has emerged as a promising treatment strategy, however, therapeutic success against solid tumors such as neuroblastoma remains modest. Recurrence of antigen-poor tumor variants often ultimately results in treatment failure. Using antigen-independent killing mechanisms such as the FAS receptor (FAS)-FAS ligand (FASL) axis through epigenetic manipulation may be a way to counteract the escape achieved by antigen downregulation. Analysis of public RNA-sequencing data from primary neuroblastomas revealed that a particular epigenetic modifier, the histone lysine demethylase 1A (KDM1A), correlated negatively with FAS expression. KDM1A is known to interact with TP53 to repress TP53-mediated transcriptional activation of genes, including FAS. We showed that pharmacologically blocking KDM1A activity in neuroblastoma cells with the small molecule inhibitor, SP-2509, increased FAS cell-surface expression in a strictly TP53-dependent manner. FAS upregulation sensitized neuroblastoma cells to FAS-FASL-dependent killing and augmented L1CAM-directed CAR T cell therapy against antigen-poor or even antigen-negative tumor cells in vitro. The improved therapeutic response was abrogated when the FAS-FASL interaction was abolished with an antagonistic FAS antibody. Our results show that KDM1A inhibition unleashes an antigen-independent killing mechanism via the FAS-FASL axis to make tumor cell variants that partially or totally suppress antigen expression susceptible to CAR T cell therapy.

Circular RNAs (circRNAs) are a regulatory RNA class. While cancer-driving functions have been identified for single circRNAs, how they modulate gene expression in cancer is not well understood. We investigate circRNA expression in the pediatric malignancy, neuroblastoma, through deep whole-transcriptome sequencing in 104 primary neuroblastomas covering all risk groups. We demonstrate that MYCN amplification, which defines a subset of high-risk cases, causes globally suppressed circRNA biogenesis directly dependent on the DHX9 RNA helicase. We detect similar mechanisms in shaping circRNA expression in the pediatric cancer medulloblastoma implying a general MYCN effect. Comparisons to other cancers identify 25 circRNAs that are specifically upregulated in neuroblastoma, including circARID1A. Transcribed from the ARID1A tumor suppressor gene, circARID1A promotes cell growth and survival, mediated by direct interaction with the KHSRP RNA-binding protein. Our study highlights the importance of MYCN regulating circRNAs in cancer and identifies molecular mechanisms, which explain their contribution to neuroblastoma pathogenesis.

The detection and characterization of cell-free DNA (cfDNA) in peripheral blood from neuroblastoma patients may serve as a minimally invasive approach to liquid biopsy. Major challenges in the analysis of cfDNA purified from blood samples are small sample volumes and low cfDNA concentrations. Droplet digital PCR (ddPCR) is a technology suitable for analyzing low levels of cfDNA. Reported here are two quadruplexed ddPCR assay protocols that reliably quantify MYCN and ALK copy numbers in a single reaction together with the two reference genes, NAGK and AFF3, and accurately estimate ALK^F1174L (exon 23 position 3522, C>A) and ALK^R1275Q (exon 25 position 3824, G>A) mutant allele fractions using cfDNA as input. The separation of positive and negative droplets was optimized for detecting two targets in each ddPCR fluorescence channel by the adjustment of the probe and primer concentrations of each target molecule. The quadruplexed assays were validated using a panel of 10 neuroblastoma cell lines and paired blood plasma and primary neuroblastoma samples from nine patients. Accuracy and sensitivity thresholds in quadruplexed assays corresponded well with those from the respective duplexed assays. Presented are two robust quadruplexed ddPCR protocols applicable in the routine clinical setting and that require only minimal plasma volumes for the assessment of MYCN and ALK oncogene status.

Only half of patients with relapsed B-cell precursor (BCP) acute lymphoblastic leukemia (ALL) currently survive with standard treatment protocols. Predicting individual patient responses to defined drugs prior to application would help therapy stratification and could improve survival. With the purpose to aid personalized targeted treatment approaches, we developed a human–zebrafish xenograft (ALL-ZeFiX) assay to predict drug response in a patient in 5 days. Leukemia blast cells were pericardially engrafted into transiently immunosuppressed Danio rerio embryos, and engrafted embryos treated for the test case, venetoclax, before single-cell dissolution for quantitative whole blast cell analysis. Bone marrow blasts from patients with newly diagnosed or relapsed BCP-ALL were successfully expanded in 60% of transplants in immunosuppressed zebrafish embryos. The response of BCP-ALL cell lines to venetoclax in ALL-ZeFiX assays mirrored responses in 2D cultures. Venetoclax produced varied responses in patient-derived BCP-ALL grafts, including two results mirroring treatment responses in two refractory BCP-ALL patients treated with venetoclax. Here we demonstrate proof-of-concept for our 5-day ALL-ZeFiX assay with primary patient blasts and the test case, venetoclax, which after expanded testing for further targeted drugs could support personalized treatment decisions within the clinical time window for decision-making.