Abstract 2144: Holistic sampling of clonal dynamics using cfDNA in lung TRACERx

Abstract

Abstract Background: Minimal residual disease (MRD) detection using liquid biopsy has the potential to improve patient outcomes in non-small cell lung cancer (NSCLC). Liquid biopsy may also provide representative clonal sampling through the disease course, but current clonal deconvolution methods are ineffective in plasma samples with <1% tumor content, common in the localised or MRD setting. Methods: We analysed 1071 plasma samples from 198 TNM I-III NSCLC patients in TRACERx who underwent multiregion exome sequencing of primary tumor and relapse tissue, with 416 standard of care surveillance scans. Seventy-four patients suffered a relapse. We targeted a median of 200 tumor-specific mutations per patient consisting of clonal and subclonal variants and sequenced cell-free DNA (cfDNA) using anchored multiplex PCR to a median unique depth of 2230X with 5 supporting duplicates. We used library-specific trinucleotide background models to call MRD. We developed ECLIPSE (Extraction of CLonality from LIquid bioPSiEs), to perform formal clonal deconvolution in <1% purity plasma samples by leveraging copy number and mutation clone identities from tumor tissue. Results: Median MRD lead time was 119 days (range 0-1137) in patients with pre-operative circulating tumor DNA (ctDNA) detection. In the 13% of relapse patients lacking pre-operative ctDNA detection the median MRD lead time was 0 days (range 0-589). MRD lead times positively correlated with clonal ctDNA fraction doubling times (DTs). Cancer cell fractions (CCFs) of subclones estimated in plasma with ECLIPSE and tissue collected concurrently were proportional (P < 0.001, R2 = 0.6). Subclonal mutations that would appear clonal in single biopsies could be separated from true clonal mutations using their CCF in plasma (P < 0.001, OR = 0.44), distinguishing clonal from subclonal mutations, where the latter are unlikely to represent effective therapeutic targets. Where cfDNA and tissue was available at relapse we detected 28/29 metastatic tissue subclones in cfDNA with an additional 8 cfDNA-unique subclones. These subclones were more frequently estimated as present in only a subset of metastatic cells using cfDNA (P = 0.008, OR = 5.5) consistent with localisation to unsampled metastatic sites. Metastatic competent subclones had higher CCFs in pre-operative plasma (P < 0.001, OR = 4.5). Shifts in clonal dynamics were concurrent with treatment. Finally, patients with cfDNA-detected polyphyletic metastatic seeding had shorter disease-free survival than those with monophyletic seeding (HR = 2.89, 95% CIs 1.46-5.73). Conclusions: Tumor-informed anchored multiplex PCR most commonly detected MRD before clinical relapse and allowed determination of clonal ctDNA DT. Using ECLIPSE, plasma samples of <1% purity allow formal measurements of clonal dynamics from diagnosis to relapse, which impacts patient outcome and has the potential to guide personalised medicine. Citation Format: Alexander Mark Frankell, Christopher Abbosh, Aaron Garnett, Judit Kisistok, Thomas Harrison, Morgan Weichert, Abel Licon, Selvaraju Veeriah, Bob Daber, Mike Moreau, Aamir Shahpurwalla, Aaron Odell, Adrian Chesh, Kevin Litchfield, Emilia Lim, Daniel E. Cook, Clare Puttick, Maise Al-Bakir, Fabio Gomes, Akshay Patel, Lizi Manzano, Paula Roberts, Ariana Huebner, Nicolas Carey, Joan Riley, Todd Druley, Jacqui A. Shaw, Nicholas McGranahan, Mariam Jamal-Hanjani, Josh Stahl, Nicolai Birkbak, the Lung TRACERx consortium, Charles Swanton. Holistic sampling of clonal dynamics using cfDNA in lung TRACERx [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2144.