Assessment of consistency of multiplex fluorescent immunohistochemistry data across multiple users utilizing different quantitative analysis strategies

Shawn Jensen, Robert W Franz Cancer Center, Earle A Chiles Research Institute, Portland Providence Medical Center, Portland, OR, USA
Carmen Ballesteros-Merino, Robert W Franz Cancer Center, Earle A Chiles Research Institute, Portland Providence Medical Center, Portland, OR, USA
Sebastian Marwitz, Robert W Franz Cancer Center, Earle A Chiles Research Institute, Portland Providence Medical Center, Portland, OR, USA
Nikhil Lonberg, Robert W Franz Cancer Center, Earle A Chiles Research Institute, Portland Providence Medical Center, Portland, OR, USA
Bernard A Fox, Robert W Franz Cancer Center, Earle A Chiles Research Institute, Portland Providence Medical Center, Portland, OR, USA

Poster presented at Society for Immunotherapy of Cancer Annual Meeting, Washington, D.C., November 7 – 11, 2018.

Description

Background: Tools that facilitate examination of the tumor microenvironment in cancer patients who either respond or do not respond to treatment are informative to the future design of immunotherapeutic strategies. Multiplex fluorescent immunohistochemistry (mIHC) is a technique enabling examination of the number and location of cells within the tumor microenvironment. Recently, we described the Cumulative Suppression Index (CSI), which examines the number of CD8+ T cells in the invasive margin of tumor combined with the number of FoxP3+ or PD-L1+ cells within 30 um of CD8+ T cells[1]. This CSI correlated with overall survival in a cohort of 119 patients with HPV- oral squamous cell cancer. Future application of the CSI will require reliable analysis methods with minimal variation across studies and users to enable comparative analysis. In this current study, we systematically compared the reliability of three different methods of enumerating cellular phenotypes of mIHC images across multiple users.

Methods: Primary tumors obtained from oral squamous cell cancer patients were sectioned and stained with antibodies to CD3, CD8, FoxP3, CD163, and PD-L1. Nine representative images were collected from one patient and analyzed using either commercial phenotyping software based on machine-learning to phenotype cells (C), commercial phenotyping software coupled with a Thresholding method (C+T), or the Thresholding method alone (T). Multiple independent users analyzed the same nine images determining the number of CD3+ PD-L1-, CD3+PD-L1+, CD8+PD- L1-, CD8+PD-L1+, FoxP3+, CD163+PD-L1-, CD163+PD-L1+, or PD-L1+ cells using each of the three methods.

Results: Analysis of the variation within each user across the three different analysis methods demonstrated tight correlation for the principle phenotypes of the CSI, namely CD8+PD-L1-, CD8+PD-L1+, FoxP3+, and PD-L1+ cells (Spearman Rank Correlation p

Conclusions: Comparative analysis of mIHC data between multiple users requires confidence in a reproducible and consistent method for data analysis. These data demonstrate that C+T or T methods of analyzing data minimize inter-user variation when using the CSI mIHC panel tested in this study.

References: 1. Feng Z, et al. Multiparametric immune profiling in HPV- oral squamous cell cancer. JCI Insight. 2017;2(14):e93652.