Quantitative Molecular Microscopy

Credit: COURTESY OF MARISA DOLLED-FILHAST, HISTO Rx" /> Credit: COURTESY OF MARISA DOLLED-FILHAST, HISTO Rx Traditional histopathology analysis has two basic problems. First, it isn't granular enough: Pathologists typically grade overall marker-staining intensity using a four-point scale. The other problem is that these measurements don't account for the sometimes-subtle changes in subcellular localization that can indicate disease. Beta-catenin, for instance, is a biomarker for colon can

Jeffrey M. Perkel
Mar 31, 2006
<figcaption> Credit: COURTESY OF MARISA DOLLED-FILHAST, HISTO Rx</figcaption>
Credit: COURTESY OF MARISA DOLLED-FILHAST, HISTO Rx

Traditional histopathology analysis has two basic problems. First, it isn't granular enough: Pathologists typically grade overall marker-staining intensity using a four-point scale. The other problem is that these measurements don't account for the sometimes-subtle changes in subcellular localization that can indicate disease. Beta-catenin, for instance, is a biomarker for colon cancer, but only when localized to the nucleus. Yale University pathologists Bob Camp and David Rimm developed a series of algorithms called AQUA (Automated Quantitative Analysis) to overcome these shortcomings, providing compartment-specific staining-intensity data over the range of 0 to 255, which has been shown to be directly proportional to absolute protein concentrations.

In this example, estrogen receptor staining is being quantified in cancerous breast tissue. The cells are stained with three tags: cytokeratin (green, A), which differentiates the tumor from surrounding tissue; DAPI (blue, B), which stains nuclei; and an anti-estrogen (ER) receptor antibody (red, C).

<figcaption> Credit: COURTESY OF MARISA DOLLED-FILHAST, HISTO Rx</figcaption>
Credit: COURTESY OF MARISA DOLLED-FILHAST, HISTO Rx

Cytokeratin staining is used to create a tumor "mask," which defines the tumor location D. Next, the nuclear compartment is identified using two algorithms: RESA (rapid exponential subtraction algorithm) and PLACE (pixel-based locale assignment for compartmentalization of expression). RESA subtracts a fraction of light from the image determined by intensity in another focal plane. Thus it converts B to E and C to G. PLACE defines the cellular compartments and computes signal intensity for the target (in this case, ER). The nuclear image is overlaid onto the cytokeratin image A, and overlapping pixels are eliminated. At this point, each pixel is identified as either part of the cytoplasm or part of the nucleus F.

The estrogen receptor-stained image is then subjected to RESA/PLACE (C to G) and overlaid onto image F H. Finally, the software calculates how much of the ER stain is present in the nuclear and cytoplasmic compartments, and it reports the results as the ratio of staining intensity over compartment size. This histogram I, plotting AQUA score (X-axis) versus case number, illustrates the distribution of nuclear ER expression in almost 600 breast cancer cases.