Image Analysis and Next Generation Sequencing: Strange Bedfellows for Quality Assurance of KRAS Mutational Status in Colorectal Cancer
Jeffrey P Baliff, Zi-Xuan Wang, Stephen C Peiper. Thomas Jefferson University, Philadelphia, PA
Background: The interface between anatomic pathologists and the molecular pathology laboratory is increasingly important with the development of oncologic therapies with companion diagnostics for solid tumors. One critical step in mutational analysis for KRAS, BRAF, and EGFR is the identification of tumor on a H&E slide, with tumor burden meeting analytical sensitivity to avoid false negatives. This process can be difficult when tumor content is low (e.g. biopsy specimens). Deep sequencing using Next Generation sequencing (NGS) is very accurate in quantifying the relative content of mutated and normal alleles in amplicons from tumor specimens. Using NGS as the gold standard, we assess if image analysis of tumor slides can be an objective and reliable method to provide quality assurance (QA) for mutation analysis.
Design: Specimens were selected from archives based on diagnosis of colonic adenocarcinoma and known positive KRAS mutation status by Sanger sequencing. A H&E stained slide with encircled tumor was scanned into an Aperio ScanScope digital scanner (Vista, CA) and image analysis to identify tumor cells, mucin, necrosis, and stroma performed within the circled region by both surface area and nuclear count. DNA was extracted from within the identical encircled area and exon 2 of KRAS was amplified by PCR and sequenced using a 454 Junior pyrosequencer (Roche, Branford, CT). 5 barcoded samples were sequenced in each run yielding 10,000-20,000 reads/case, equally distributed between forward and reverse reads. Correlation between the percentage of tumor calculated by image analysis compared to that from NGS was determined.
Results: 7 cases of KRAS mutated colorectal cancer were selected for analysis. NGS detected identical KRAS mutations as previously found by Sanger sequencing. In tumors containing large areas of necrosis, image analysis of surface area consisting of tumor and necrosis best correlated with mutated DNA by NGS. When minimal necrosis existed, image analysis by nuclear count was equivalent to surface area in the correlation with mutated DNA by NGS.
Conclusions: The amount of tumor present on a slide by image analysis serves as a QA verification of the relative content of mutations derived by NGS. Virtual microscopy may be incorporated into the workflow of the molecular laboratory to correlate the content of tumor cells with the relative content of specific mutations. This approach will be of clinical utility in the analysis of small specimens as well as those with limited numbers of tumor cells.
Monday, March 19, 2012 1:00 PM
Poster Session II # 285, Monday Afternoon