Abstract
Abstract Introduction: Cell-free DNA (cfDNA) is a robust analyte for non-invasively genotyping cancer patients and useful for both detection and disease monitoring. Recent studies have revealed signals in cfDNA fragmentation patterns useful for inferring epigenetic information by analyzing chromatin accessibility patterns (Snyder et al. 2016; Ulz et al. 2017). However, it remains unclear whether it is feasible to robustly predict RNA expression levels for individual genes by using these cfDNA patterns, as might inform clinically relevant challenges in cancer genomics including cancer classification. Methods: We introduce EPIC-Seq (Epigenetic Profile Inference from CfDNA Deep Sequencing), a sequencing assay targeting transcription start sites (TSSs) of selected genes of interest. We adapt this approach for three clinical problems using a panel of ~200 gene TSSs, including genes informative for cell-of-origin classification of diffuse large B-cell lymphoma (DLBCL: ABC vs GCB), for classification of non-small cell lung cancer histologies (NSCLC: LUAD vs LUSC), and for deconvolution of leukocyte composition by CIBERSORTx digital cytometry. We collected plasma samples from 89 cancer patients (45 DLBCL, 44 NSCLC). We profiled cfDNA from all samples by EPIC-Seq, and then assessed its classification performance against established clinical standards. Results: We analyzed cfDNA fragmentation spectra at TSSs genome-wide in relation to transcriptome-wide gene expression levels. In so doing, we identified a novel chromatin accessibility feature at TSSs in cfDNA robustly varying with expression levels. Combining this accessibility feature with other cfDNA features, we developed a predictive model to infer gene-level transcriptional activity using individual TSSs. When comparing the imputed expression levels using this predictive TSS model with ground truth measured by RNA-seq in peripheral blood, we observed surprisingly high transcriptome-wide correlation (R = 0.81; 10-fold CV). We then asked if inferred expression levels from cfDNA could accurately classify lymphoma subtypes and lung cancer tumor histologies. The resulting classifiers achieved AUC of 0.89 and 0.87 for DLBCL COO and NSLSC subtypes, respectively. We also inferred B-cell and lung epithelial fractions in cfDNA by applying CIBERSORTx to deconvolute EPIC-Seq imputed gene expressions in “bulk cfDNA”. The resulting B cell fractions in DLBCL patients were significantly correlated with tumor burden as measured by mean CAPP-Seq VAF levels (R=0.8, P=0.0001). Similarly, the epithelial cell fractions in NSCLC patients were significantly correlated with mean VAFs (R = 0.5, P= 0.01). Conclusions: We show that by analyzing cfDNA fragmentation patterns at transcription start sites, EPIC-Seq accurately predicts tissue and tumor-specific gene expressions in healthy and cancer plasma samples. EPIC-Seq demonstrates the power of cfDNA analyses to survey multiple measures in cancer such as gene expression and tissue-of-origin. Citation Format: Mohammad Shahrokh Esfahani, Mahya Mehrmohamadi, Chloe B. Steen, Emily G. Hamilton, Daniel A. King, Joanne Soo, Charles Macaulay, Michael Jin, David M. Kurtz, Barzin Nabet, Everett Moding, Jacob Chabon, Aaron Newman, Maximilian Diehn, Ash A. Alizadeh. Chromatin accessibility patterns in cell-free DNA reveal tumor heterogeneity [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 3388.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.