19P - Single sample predictor of non-small cell lung cancer histology based on gene expression analysis of archival tissue
- Anna Karlsson (SE)
- Helena Cirenajwis (SE)
- Kajsa Ericson-Lindqvist (SE)
- Christel Reuterswärd (SE)
- Mats Jönsson (SE)
- Annika Patthey (SE)
- Annelie F. Behndig (SE)
- Mikael Johansson (SE)
- Maria Planck (SE)
- Johan Staaf (SE)
- Anna Karlsson (SE)
Abstract
Background
In non-small cell lung cancer (NSCLC), histological classification dictates choice of patient therapy. In this study, we aimed to establish a gene expression based single-sample predictor (SSP) for histological classification of NSCLC tumors using archival tissue that may be used in parallel with e.g. gene fusion detection as a multicomponent single assay.
Methods
A NanoString probe set was designed to target 12 genes routinely used as IHC markers for histological subtyping as well as fusion genes known to be frequently active in NSCLC. Gene expression data was derived from NanoString analysis of 78 formalin-fixed paraffin-embedded (FFPE) NSCLCs with known histological subtypes (development cohort). A SSP was trained using AIMS (1) in the development cohort for prediction of adenocarcinoma (AC), squamous cell carcinoma (SqCC), or neuroendocrine tumors (NE). The AIMS model was applied to 11 FFPE tumors classified as large cell carcinomas (LCC) according to the WHO2004 classification (2), and 199 fresh frozen NSCLC tumors analyzed by RNA sequencing (GSE81089)(3). Finally, the SSP will be applied to 11 NSCLC-not otherwise specified (NOS) cases (4) subjected to in-depth pathological re-evaluation.
Results
The SSP was successfully applied to NanoString data from 11 LCCs re-classified as AC, SqCC and LCC according to the revised WHO2015 guidelines (2). Of reclassified LCC tumors, 100% of AC cases and 75% (3/4) of SqCC tumors were correctly identified. In GSE81089, the SSP was erratically successful depending on histology of the tumor classified, with 97.4% concordance for AC, 97.1% for SqCC, but mismatch for 3 out of 5 NE tumors. In summary, the SSP could successfully classify tumors of AC and SqCC histology in both validation cohorts but could less successfully classify non-AC and non-SqCC tumors respectively.
Conclusions
Gene expression based SSP can accurately classify AC and SqCC histology. Expanded transcriptional profiling may be required to capture all aspects of lung cancer biology for precise and possibly refined histological subtyping of individual cases. Gene expression-based analysis could serve as a promising complement to existing techniques, providing a useful multicomponent assay for lung cancer diagnostics.
Legal entity responsible for the study
Lund University
Funding
The Swedish Cancer Society, the Mrs Berta Kamprad Foundation, the Gunnar Nilsson Cancer Foundation, the Crafoord Foundation, BioCARE a Strategic Research Program at Lund University, the Gustav V:s Jubilee Foundation, Skane University Hospital Foundation, and governmental funding (ALF)
Disclosure
All authors have declared no conflicts of interest.
