Abstract 1552: Image-based microchip sorting of pure, immuno-phenotypically defined subpopulations of tumor cells from tiny formalin-fixed paraffin embedded (FFPE) samples reveals their distinct genetic features

• Published in: Cancer research (Chicago, Ill.) Ill.), 2015-08, Vol.75 (15_Supplement), p.1552-1552
• Main Authors: Bolognesi, Chiara, Doffini, Anna, Buson, Genny, Lanzellotto, Rossana, Signorini, Giulio, Sero, Valeria, Calanca, Alex, Fontana, Francesca, Romano, Rita, Gianni, Stefano, Bregola, Giulia, Medoro, Gianni, Tanzi, Raimo, Giorgini, Giuseppe, Morreau, Hans, Barberis, Massimo, Corver, Willem E., Manaresi, Nicolo
• Format: Article
• Language: English
• ISSN: 0008-5472; 1538-7445
• DOI: 10.1158/1538-7445.AM2015-1552

Background: We provide a solution of pressing needs in preparation of FFPE samples for genomic analysis: small sample size, unwanted admixture of normal cells, analysis of tumor rare-cell subpopulations present at low percentages in the tumor fraction. Methods: We disaggregated into cell suspensions archival FFPE samples from 12 ovarian, pancreatic and lung cancer patients, staining for Vimentin, Keratin and DNA. We sorted by DEPArray™ precise numbers (mean = 107, median 58, range = 5-600) of pure homogenous cells from the major population of tumor cells, the contaminant diploid stromal cells, and other minority tumor cell types indicative of epithelial-to-mesenchymal transition (EMT). Using IonTorrent AmpliSeq CHPv2, we generated sequencing libraries, after lysis of the pure cells recovered by DEPArray™ (n = 54), or unsorted samples (either QIAmp DNA columns or disaggregated cells). Libraries were sequenced with IonTorrent PGM (mean depth>2,000x), and analyzed using IonTorrent software. Results: On several loci, we detected somatic mutations with 100% variant frequency, only observable as heterozygous in the unsorted samples and as wild-type in stromal cells of same patient, confirming 100% purity of sorted cells. Moreover, in the EMT-phenotype subpopulations we identified clear somatic mutations, different from tumor cells majority and undetectable in unsorted samples. Frequently, for loci harboring germ-line heterozygous SNPs with variant frequency around 50% for pure stromal cells, we readily detected loss-of-heterozygosis in tumor cells subpopulations as binary (0%/100%) variants. Quantitative traits such as copy number gains and losses were also reproducibly identified in tumor cell replicates as deviations from the 50% variant frequency of germline SNPs of pure stromal cells. Furthermore, we observed an excellent coverage uniformity (mean = 96%) for recoveries (n = 27) in the range of 81-600 cells, even higher than the uniformity obtained with (n = 2) QIAmp-purified DNA (92%). Mean uniformity gradually decreased to 89% for cell recoveries (n = 13) in the range 21-80, and further decreased to 70% for lower cell numbers (n = 14). Highlights: Sorting tumor rare-cell subpopulations reveals their genetic characteristics, undetectable in unsorted samples. Analyzing homogenous cell subpopulations boosts signal-to-noise ratio working around inherent sensitivity/specifitiy trade-offs of rare-variant calls. The proposed workflow further enables reliable detec