Cell-Free DNA Fragmentomics for Early Detection of Malignant Transformation in NF1 Patients Harboring Plexiform Neurofibromas

Abstract

Malignant peripheral nerve sheath tumors (MPNST) are aggressive soft tissue sarcomas that are resistant to radiotherapy and chemotherapy, and are associated with poor survival. MPNST typically arise from benign or atypical plexiform neurofibromas (PN) in the setting of neurofibromatosis type 1 (NF1). Unfortunately, early detection by imaging and clinical examination remains very challenging. We hypothesize that cell-free DNA fragmentomic analysis can be used to precisely detect the risk for developing MPNST in NF1 patients, which could significantly improve clinical outcomes in the future through early cancer detection. We performed low-pass whole genome sequencing of plasma cell-free DNA samples from healthy controls (n = 21), patients with PN (n = 113), AN (n = 39) and MPNST (n = 71). Cell-free DNA was analyzed at a per-fragment level by tracking 4-nucleotide end motifs, and using a bin-wise fragmentomic machine learning approach called DELFI, which integrates across ratios of short (100-150 bp) versus long (151-220 bp) fragments in 5 megabase bins across the genome. End motif frequencies were compared between MPNST and PN patients, and receiver operating characteristic (ROC) analysis was used to classify disease states using DELFI fragmentomic scores. Using DELFI for genome-wide fragmentomics, followed by ROC analysis, we obtained the following accuracies: MPNST vs. AN (acc 0.62), MPNST vs. PN (acc 0.83), MPNST vs. healthy (acc 0.86), AN vs. PN (acc 0.87), AN vs. healthy (acc 0.72) and PN vs. healthy (acc 0.88). DELFI-based cell-free DNA fragmentomics was superior to the ichor CNA-based copy number alteration classifier that we published previously (Szymanski et al. PLOS Medicine, 2021) for all the aforementioned group comparisons (p < 0.001 by McNemar's test). Strikingly, the two AN cases with DELFI scores most closely resembling MPNST were separately identified by independent clinical care teams to have very high-risk features and recent history warranting short-interval follow up. Specifically, one patient had a paraspinal AN exhibiting new onset pain and rapid growth on imaging, while the other had increased FDG uptake of multiple AN lesions in a supraclavicular field treated years earlier with radiotherapy. We also measured fragment end motifs across different patient groups, and found that the CTCA end motif was more commonly present in cell-free DNA from MPNST and PN patients compared to healthy donors (Bonferroni-corrected p-value < 0.001). This study suggests that fragmentomic analysis of cell-free DNA can accurately risk-stratify NF1 patients with plexiform neurofibromas. Cell-free DNA fragmentomics outperformed our previously published copy number alteration method for distinguishing closely related patient subgroups with different malignancy risks. Atypical neurofibroma patients with cell-free DNA fragmentomic scores suggestive of MPNST were deemed clinically very high-risk for malignant transformation.