Virtual tissue microarrays for validating digital biomarker analysis in colorectal carcinoma.

Abstract

Tissue microarrays (TMAs) are used for high-throughput biomarker discovery and validation. While TMAs have numerous advantages, they may not always be representative of the tissue heterogeneity present in whole tissue sections (WTS) leading to inadequate biomarker quantification. In this pilot study, we studied biomarker expression in 50 randomly selected colorectal cancers and 36 MSI cases with or without BRAF variant. We used virtual TMAs to determine the minimum number of tissue cores needed to quantify biomarkers with the same precision as when using WTS. Paraffin sections were immunohistochemically stained for markers of T-cells, B-cells, cancer associated fibroblasts, and macrophages. Digitized WTS were divided into tumor center (TC) and invasive margin (IM) regions. The minimum number of virtual TMA cores in each region was determined by Bland-Altman plots with 95% limits of agreement. Bland-Altman plots showed substantial disagreement between TMAs and WTS, being highest for 3 cores and decreasing with increasing core numbers. However, even when using 8 cores, the limits of agreement between TMA and WTS were wide, indicating a high degree of measuring uncertainty using TMAs. When using 3 or 4 cores, TMAs underestimated expression of all the biomarkers in the TC; similarly, levels of macrophage markers in the TC, and levels of B-cells in both the TC and the IM remained considerably underestimated, even when using the maximum number of cores possible. However, 3 cores were sufficient to adequately classify biomarkers into categorical low and high expression groups. Microsatellite unstable tumors were characterized by high heterogeneity, which was further increased in the presence of BRAF variant(s). The virtual TMA technique is a useful method to establish the minimum number of cores to be included when constructing tumor TMAs for biomarker analysis. Our results emphasize the importance of TMA validation for a specific biomarker prior to conducting larger clinical studies.