Abstract

BackgroundAccurate lung cancer classification is crucial to guide therapeutic decisions. However, histological subtyping by pathologists requires tumor tissue—a necessity that is often intrinsically associated with procedural difficulties. The analysis of circulating tumor DNA present in minimal-invasive blood samples, referred to as liquid biopsies, could therefore emerge as an attractive alternative.MethodsConcerning adenocarcinoma, squamous cell carcinoma, and small cell carcinoma, our proof of concept study investigates the potential of liquid biopsy-derived copy number alterations, derived from single-end shallow whole-genome sequencing (coverage 0.1–0.5×), across 51 advanced stage lung cancer patients.ResultsGenomic abnormality testing reveals anomalies in 86.3% of the liquid biopsies (16/20 for adenocarcinoma, 13/16 for squamous cell, and 15/15 for small cell carcinoma). We demonstrate that copy number profiles from formalin-fixed paraffin-embedded tumor biopsies are well represented by their liquid equivalent. This is especially valid within the small cell carcinoma group, where paired profiles have an average Pearson correlation of 0.86 (95% CI 0.79–0.93). A predictive model trained with public data, derived from 843 tissue biopsies, shows that liquid biopsies exhibit multiple deviations that reflect histological classification. Most notably, distinguishing small from non-small cell lung cancer is characterized by an area under the curve of 0.98 during receiver operating characteristic analysis. Additionally, we investigated how deeper paired-end sequencing, which will eventually become feasible for routine diagnosis, empowers tumor read enrichment by insert size filtering: for all of the 29 resequenced liquid biopsies, the tumor fraction could be increased in silico, thereby “rescuing” three out of five cases with previously undetectable alterations.ConclusionsCopy number profiling of cell-free DNA enables histological classification. Since shallow whole-genome sequencing is inexpensive and often fully operational at routine molecular laboratories, this finding has current diagnostic potential, especially for patients with lesions that are difficult to reach.

Highlights

  • Accurate lung cancer classification is crucial to guide therapeutic decisions

  • Since approximately 70–75% of all lung cancer cases are diagnosed as advanced stage diseases, and plasma genomic abnormality increases with tumor stage, we focused on patients with advanced stage tumors during recruitment [10, 15]

  • Negative controls included Liquid biopsy (LB) from healthy subjects (females from routine non-invasive prenatal testing (NIPT) and healthy males; n = 60) and formalin-fixed paraffinembedded (FFPE) samples from benign tissue (n = 9). Other than these inhouse cases, public segmental copy number data, derived from Single nucleotide polymorphism (SNP) array 6.0 (Affymetrix, Santa Clara, CA) experiments, complemented with clinical information and a list of significantly aberrant loci per histological subtype, were collected from the supplement of the study of Seidel et al, which presents the collective effort from the consortia “Clinical Lung Cancer Genome Project” (CLCGP) and “Network Genomic Medicine” (NGM) [16]

Read more

Summary

Introduction

Accurate lung cancer classification is crucial to guide therapeutic decisions. In order to administer the most appropriate therapy, accurate histological classification is essential to guide individual decisions. The subcategorization of non-small cell lung cancer (NSCLC), representing approximately 85% of all lung cancers, in inter alia adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC), has long been clinically relevant, especially for targeted therapy [3]. Further subclassifying small cell lung cancer (SCLC) has fewer diagnostic consequence, as it is sufficient to correctly determine the small cell histology in order to initiate chemotherapeutic treatment [5]. Ongoing clinical trials are evaluating targeted and immunotherapies for molecularly characterized SCLCs, yet none of these are routinely implemented at present [6]

Methods
Results
Discussion
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call