In the clinical management of paediatric solid tumours, histological examination of tumour tissue obtained by a biopsy remains the gold standard to establish a conclusive pathological diagnosis. The DNA methylation pattern of a tumour is known to correlate with the histopathological diagnosis across cancer types and is showing promise in the diagnostic workup of tumour samples. This methylation pattern can be detected in the cell-free DNA. Here, we provide proof-of-concept of histopathologic classification of paediatric tumours using cell-free reduced representation bisulphite sequencing (cf-RRBS) from retrospectively collected plasma and cerebrospinal fluid samples. We determined the correct tumour type in 49 out of 60 (81.6%) samples starting from minute amounts (less than 10 ng) of cell-free DNA. We demonstrate that the majority of misclassifications were associated with sample quality and not with the extent of disease. Our approach has the potential to help tackle some of the remaining diagnostic challenges in paediatric oncology in a cost-effective and minimally invasive manner.

Background: Paediatric tumours are often characterised by the presence of recurrent DNA copy number alterations (CNAs). These DNA copy number profiles, obtained from a tissue biopsy, can aid in the correct prognostic classification and therapeutic stratification of several paediatric cancer entities (e.g. MYCN amplification in neuroblastoma) and are part of the routine diagnostic practice. Liquid biopsies (LQBs) offer a potentially safer alternative for such invasive tumour tissue biopsies and can provide deeper insight into tumour heterogeneity. Procedure: The robustness and reliability of LQB CNA analyses was evaluated. We performed retrospective CNA profiling using shallow whole-genome sequencing (sWGS) on paired plasma circulating cell-free DNA (cfDNA) and tissue DNA samples from routinely collected samples from paediatric patients (n Z 128) representing different tumour entities, including osteosarcoma, Ewing sarcoma, rhabdomyosarcoma, Wilms tumour, brain tumours and neuroblastoma.Results: Overall, we observed a good concordance between CNAs in tissue DNA and cfDNA. The main cause of CNA discordance was found to be low cfDNA sample quality (i.e. the ratio of cfDNA (<700 bp) and high molecular weight DNA (>700 bp)). Furthermore, CNAs were observed that were present in cfDNA and not in tissue DNA, or vice-versa. In neuroblastoma samples, no false-positives or false-negatives were identified for the detection of the prognostic marker MYCN amplification. Conclusion: In future prospective studies, CNA analysis on LQBs that are of sufficient quality can serve as a complementary assay for CNA analysis on tissue biopsies, as either cfDNA or tissue DNA can contain CNAs that cannot be identified in the other biomaterial.

The methylation profile of circulating cell-free DNA (cfDNA) in blood can be exploited to detect and diagnose cancer and other tissue pathologies and is therefore of great diagnostic interest. There is an urgent need for a cost-effective genome-wide methylation profiling method that is simple, robust and automatable and that works on highly fragmented cfDNA. We report on a novel sample preparation method for reduced representation bisulfite sequencing (RRBS), rigorously designed and customized for minute amounts of highly fragmented DNA. Our method works in particular on cfDNA from blood plasma. It is a performant and cost-effective methodology (termed cf-RRBS) which enables clinical cfDNA epigenomics studies.The methylation profile of DNA can be exploited to detect and diagnose cancer and other tissue pathologies and is therefore of great diagnostic interest. To analyze the genome-wide DNA methylation status in high-molecular weight genomic DNA that is typically obtained from tissues, many profiling methods have been developed 1 . Reduced representation bisulfite sequencing (RRBS) strikes a particularly good balance between genome-wide coverage and accurate quantification of the methylation status and an affordable cost. It is thus ideally suited for epigenomics studies involving large numbers of samples 2 . However, none of the RRBS methods reported to date are suited for analyzing the minute quantities of highly fragmented circulating cell-free DNA (cfDNA) that can be purified from blood plasma 3,4 . We report on a novel biotechnological sample preparation method for RRBS, rigorously designed and customized for minute amounts of highly fragmented DNA. This robust and automatable methodology (termed cf-RRBS) is highly cost-effective and enables clinical cfDNA epigenomics studies.