Abstract 4665: Quantitative in situ biomarker analysis via ultrasensitive RNA in situ hybridization and automated image analysis

Abstract

Abstract Background and objectives: Accurate biomarker analysis and interpretation require tissue and cellular context information, which is lost when DNA/RNA is extracted for molecular analysis. This is especially challenging in cancer biomarker analysis due to significant inter- and intra-tumor heterogeneity. To address this, we have recently developed an ultrasensitive RNA In situ hybridization (ISH) technique (RNAscope) that enables quantitative RNA detection in the presence of full morphological and cellular context that can be visualized under standard bright-field microscopy. The resulting stained slides can be scored by enumerating the punctate signal dots in each cell. Although feasible, manual scoring can be tedious, time-consuming and subjective for routine use. Therefore, we have developed dedicated image analysis software (SpotStudio) to automate this task. The purpose of this study was to validate the performance of this software. Materials and Methods: A melanoma tissue microarray (TMA) containing formalin-fixed paraffin-embedded (FFPE) specimens from 159 cases were used in this study. Two TMA sections were stained for the reference gene UBC and a long non-coding RNA (lncRNA) using RNAscope 2.0 Red assay (Advanced Cell Diagnostics, Hayward, CA). UBC staining was used to assess RNA quality. LncRNA expression was scored manually by two independent observers by scoring both the RNA signals in each cell using the following scale: 0 (0 dot/cell), 1+ (1-3 dots/cell), 2+ (4-10 dots/cell), 3+ (>10 dots/cell) and the percentage of cells in each signal category. Scanned TMA images were analyzed using SpotStudio software to estimate the number of signal dots in each cell in a region of interest selected by the user, and the single-cell level results were binned according to the same categories as manual scoring. An H-score was calculated for each case by summing the product of the signal category score and the percentage cells in that category for all scoring categories. Results: Among the 159 cases of TMA, 18 were excluded due to tissue detachment, less than 5% tumor, or excessive melanin obscuring ISH signals. Three cases were excluded from automated image analysis due to poor focusing of scanned images. There was good agreement between the two manual H-scores by the two independent observers (Spearman rho = 0.86, p<0.001). The automated H-scores also demonstrated excellent agreement with the manual H-scores (Spearman rho = 0.91, p<0.001). Conclusion: Automated quantitation of RNA ISH slides is highly desirable as manual scoring can be time consuming and a significant source of variability. The SpotStudio software demonstrated excellent agreement with manual scoring. In addition, the software can provide quantitative data on a cell-by-cell basis, which would be impractical manually. Quantitative analysis of biomarkers at the single-cell level will be invaluable in many areas of cancer research and diagnostics. Citation Format: Hongwei Wang, Laurent Lessard, Nan Su, Yuling Luo, Dave Hoon, Xiao-Jun Ma. Quantitative in situ biomarker analysis via ultrasensitive RNA in situ hybridization and automated image analysis. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 4665. doi:10.1158/1538-7445.AM2014-4665