Abstract 4528: RNAseq RNAaccess is the preferred method for expression profiling of low quality FFPE gastric cancer samples

Abstract

Abstract Clinical trial tissue samples are a valuable resource for biomarker discovery and are frequently collected as Formalin-Fixed, Paraffin-Embedded (FFPE) blocks. FFPE tissues pose challenges for next generation sequencing (NGS)-based expression analysis due to poor yield from highly fragmented RNA. The objective of the current study was to evaluate and compare RNAseq RNAaccess (RR) and Nanostring (NS) platforms for molecular profiling of highly fragmented gastric cancer FFPE tissue. RR (TruSeq RNAaccess, Illumina) and NS (PanCancer immune) were performed according to manufacturer protocol using RNA extracted from sourced FFPE gastric cancer tissues (N=6) with a range of DV200 scores (6-34%) and case-matched fresh-frozen (FF) tissue (N=6). For NS, 200ng and 400ng RNA input levels were tested. Data quality from the 400ng samples proved to be superior, resulting in two more samples passing QC. A final sample size of 4 matched pairs was utilized for cross-platform comparisons. For RR, 100ng RNA was used. Pearson correlation coefficients were examined for the following comparisons: 1) FFPE vs. FF samples from RR, 2) FFPE from NS vs. FF from RR, and 3) FFPE NS vs. FFPE RR. Analyses were performed on log2 transformed expression data, for all available genes (N= 770 common genes from RR and NS) and an 18-gene IFNγ signature (IFNγ). All correlations were plotted against a range of DV200 scores from the FFPE samples in order to evaluate a potential tissue quality cutoff for RR, specifically at DV200 levels below Illumina guideline of 30. All RR data samples yielded >80% exonic rate in uniquely mapped reads. Conversely, samples with DV200<14 analyzed by NS failed to pass QC due to low fraction of genes detected above background signal. Comparing FFPE to FF tissue from RR, samples with DV200>10 were highly correlated (r≥0.92), globally or for IFNγ. Comparisons of FF RR to FFPE NS were also well correlated (global median r=0.8; IFNγ median r=0.86). Similarly, expression of FFPE RR was well correlated with FFPE NS (global median r=0.81; IFNγ median r=0.89). Altogether, NS and RR data were highly comparable and minimally impacted by DV200 scores below 30. RR generates gene expression profiles from FFPE tissue that are highly concordant with case-matched FF tissues. This study supports using RR for transcriptional profiling of poor quality gastric cancer FFPE samples. Samples below the Illumina suggested DV200 cutoff of 30 generated good quality RR and NS expression data. IFNγ scores derived from NS and RR from the same tissue were highly concordant. However, NS required higher RNA input level to compensate for low DV200 score and no acceptable expression data was generated on samples below DV200 14. This study will be expanded with data on FFPE samples with DV200<10 to identify a lower limit of RNA quality for RR. Citation Format: Emon Elboudwarej, Marianna Zavodovskaya, Xi Zhao, Luting Zhuo, Carrie Baker Brachmann, Jinfeng Liu. RNAseq RNAaccess is the preferred method for expression profiling of low quality FFPE gastric cancer samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 4528.