P4‐005: Quantitation of beta‐amyloid in transgenic mice using whole slide digital imaging and image analysis software

Defining the effects of biological manipulation on beta-amyloid in mouse models requires quantifying treatment-induced changes. The gold standard remains microscopic quantitation of beta-amyloid in histologic sections; however, there is no consistency how this is performed. This presentation describes methods for defining the area of vascular and parenchymal beta-amyloid in transgenic mice. Methods were established on Tg2576 mice. All sections used formalin-fixed, paraffin embedded tissues sectioned at 5um. Sectioning focused on capturing both the cortex and the hippocampus and for each brain, 300 consecutive sections were produced. An immunohistochemical (6E10) and histochemical stain (thioflavin-S) defined the presence of beta-amyloid. Staining with 6E10 antibody used diaminobenzidine and Fast Red as chromogens. Slides were scanned on an Aperio XT. Thioflavin S staining was performed by hand and slides were scanned on an Aperio FL. Images were hand-annotated to define the cortex and hippocampus and analyzed using algorithms to separate tissue from non tissue, and vascular from parenchymal plaques. Critical analysis of the methods to quantify beta-amyloid demonstrated four major areas of variability:1) variation in tissue processing; 2) variation in pipetting; 3) chromogen (use of Fast Red is preferred to distinguish melanosis from brown DAB staining of beta amyloid); and 4) poor focus of scanned image. In order to use the same algorithm throughout large studies, it was necessary to critically review all slides, scanned images and processed images. In general, bright field images were easier to define the area of interest but more difficult to differentiate vascular amyloid from parenchymal plaques. Inter-animal variability requires that large numbers of animals be used to define a biological effect in amyloid reduction studies. Thus, whole slide digital imaging and sophisticated image analysis software represent a major improvement in efficient and consistent quantification of beta amyloid in tissues. However, this is not a“set it and forget it process” and requires constant review during all stages to produce consistent results. In our experience, the quality of the scanners and the software to quantify morphologic changes on these digital images are associated with much less variability than the slides being evaluated.