Abstract PS6-60: A clustering and mutual information based analysis of the ELIMA study results: The additive value of multi-parametric liquid biopsies, including CTCs, EVs and cfDNA, in metastatic breast cancer

Abstract

Abstract Background: To gain comprehensive insights into the aspects of genomic and transcriptomic complexity which could be beneficial for therapy management in metastatic breast cancer (MBC), we established the isolation and analysis of mRNA and gDNA from circulating tumor cells (CTCs), mRNA from extracellular vesicles (EVs), and cell-free DNA (cfDNA) from a minimized blood volume. Here, we aimed to assemble the results of all four analytes and elucidated the relevance of the diverse parameters within a multimodal data set. Methods: EDTA blood (2x 9 ml) was drawn from 26 MBC patients with hormone receptor-positive and HER2 negative primary tumors at the time of disease progression. CTCs and their mRNA were isolated using the AdnaTest EMT2/StemCell Select/Detect. Plasma of CTC-depleted blood was used for cfDNA isolation, while mRNA from EVs was isolated by exoRNeasy using the remaining blood. The mRNA purified from CTCs and EVs was analyzed by a multimarker qPCR panel. gDNA from CTCs was isolated from the mRNA-depleted CTC lysates using the AllPrep DNA/RNA Nano Kit prototype. CTC gDNA and cfDNA were analyzed with a customized QIAseq Targeted DNA Panel for Illumina with unique molecular indices. Consumables: QIAGEN, Germany. The statistical tools for evaluating the results included: Hierarchical clustering according to Ward’s method with Euclidean distance, singular value decomposition, mutual information calculation, and k-means clustering. Results: Isolation of mRNA and gDNA from CTCs, mRNA from EVs, and cfDNA was successfully established in a condensed workflow. 88% of the patients showed at least one variant in CTC gDNA or one overexpression signal in the CTC mRNA fraction. The mean number of variants/signals was also higher in CTC gDNA/mRNA when compared to cfDNA/EV mRNA. The analysis of individual analytes identified a similar number of patients (50%-73%) with actionable markers, but a multi-parametric evaluation of all four analytes identified actionable markers in 96% of the patients. After hierarchical clustering of the results of each individual analyte into four clusters, combining the two patient clusters with the worst overall survival resulted in prognostic value for CTC gDNA, cfDNA, and EV mRNA. Combination of the information above analyte borders showed additive value and resulted in a prognostic factor defined here as the ‘ELIMA score’. A calculation of the mutual information showed that CTC gDNA has the highest potential to describe the other three analytes. However, an Eigenvector analysis based on singular value decomposition revealed that the 10 most influential vectors contain parameters from all four analytes. This indicates that each of the analytes add valuable information not conveyed by the other analytes. K-means clustering was used to generate clusters based solely on CTC gDNA and based on all four analytes. A comparison of the clusters based on these two criteria showed that clustering based on all four analytes resulted in a division of patients according to their tumor histology type and ERBB2 variants in CTCs; thereby, underscoring the importance of taking all four analytes into consideration. Conclusion: We established a workflow for parallel isolation of multiple liquid biopsy analytes from a minimized blood volume. Though the mutual information calculation showed that CTC gDNA has a relatively greater ability to describe the other three analytes, further statistical analysis showed that each analyte carried information of additive value. Thus, a comprehensive picture of the genomic and transcriptomic complexity obtained by a multi-parametric liquid biopsy might enable easier identification of the most suitable therapy regimen for each individual patient in the future. Citation Format: Sabine Kasimir-Bauer, Vinay Suryaprakash, Markus Storbeck, Peter Hahn, Siegfried Hauch, Markus Sprenger-Haussels, Oliver Hoffmann, Rainer Kimmig, Corinna Keup. A clustering and mutual information based analysis of the ELIMA study results: The additive value of multi-parametric liquid biopsies, including CTCs, EVs and cfDNA, in metastatic breast cancer [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS6-60.