Immunohistochemistry

In the laboratory, six steps take a slide from section to image.

1. Tissue preparation. FFPE sections go on charged slides. Antigen retrieval, by heat or by enzyme, unmasks the target epitopes.
2. Primary antibody. An antibody specific to the target protein goes on at a validated concentration and incubation time.
3. Secondary detection. An enzyme-conjugated secondary antibody (HRP-anti-mouse, HRP-anti-rabbit) binds the primary. Polymer-based detection systems, which use no biotin, reduce background.
4. Chromogen development. DAB substrate is applied, and HRP oxidizes it into an insoluble brown precipitate at the antibody binding site. Development time controls the intensity.
5. Counterstain. HematoxylinLoading... stains the nuclei blue-purple for morphological context.
6. Mounting and imaging. The section is mounted permanently under a coverslip and imaged in brightfield on a whole-slide scanner or standard microscope.

StrataQuest picks up from the image. Color SeparationLoading..., available only for brightfield projects, turns the color image into one grayscale image per marker. In those images marker quantity is shown as brightness, so the separated hematoxylin image shows nuclei as bright structures on a dark background, the form the NucleiLoading... engine requires. The Add IHC 2/3 quick action wires exactly this: Color Separation with two or three markers, Nuclei detection on the first marker, a Cellular mask for the others and measurements for all markers.

Source: StrataQuest 8.1 Engines Reference, ch. 7 (pp. 361–362), ch. 10 (pp. 382–395), ch. 26 (pp. 449–450), ch. 29 (pp. 465–466); User Manual §1 (pp. 11–12), Quick Actions (p. 154), App markers (p. 251).

IHC contrast comes from absorption. DAB absorbs mostly blue light (about 400 to 500 nm), so red and green get through and the deposit looks brown. Hematoxylin absorbs red and green and looks blue-purple. Classical color deconvolutionLoading... (Ruifrok and Johnston, 2001) uses these known absorption profiles to pull the stains apart, converting RGB values to optical densityLoading... and solving for at most three stain vectors.

StrataQuest's Color Separation is described in its own terms. Single Reference Shades unmixes each pixel relative to a reference color you pick, and Intensity sets the gray value that stands for that reference amount. Multiple Reference Shades instead classifies pixels by their color distance to reference shades. Neither is described as optical density, so treat the gray values as relative to your reference, not as calibrated absorbance; for brightfield Apps the User Manual allows 2 to 10 channels.

In optical density space, an ideal absorber follows the Beer–Lambert lawLoading...: OD = ε × c × l, with c the chromogen concentration and l the path length. DAB is widely described as scattering as well as absorbing light, so it follows the law only approximately, worst in dark deposits. The step from antigen to deposit is less tidy still: HRP saturates at high antigen density, and longer development deposits more chromogen but not in proportion.

IHC is only semi-quantitative. Fluorescence intensity tracks fluorophore concentration more directly than an enzyme-amplified stain can, and a cell with twice the antigen does not necessarily produce twice the DAB. That is why manual IHC scoring uses the ordinal 0/1+/2+/3+ scale, and why consistent digital measurement improves reproducibility without turning the stain into a linear assay.

Here is PD-L1 analysis in lung adenocarcinoma stained with the 22C3 anti-PD-L1 antibody.

1. Stain an FFPE section with the 22C3 anti-PD-L1 antibody, DAB detection and a hematoxylin counterstain, and scan it in brightfield.
2. Run the Add IHC 2/3 quick action with two markers, hematoxylin first and DAB second. Pick reference shades for each; Autodetect works when exactly two markers are defined.
3. NucleiLoading... runs directly on the separated hematoxylin image, where nuclei are already bright on dark. No inversion is needed.
4. For the membrane, use a Cellular ring mask around each nucleus, or Grow (global)Loading..., and measure DAB inside it. The MembraneLoading... engine can also detect the membrane structures as a mask.
5. Identify the tumor cells. A single PD-L1 stain has no tumor marker, so use morphology: tumor regions annotated by an expert reader, or a ClassifierLoading... trained to detect tumor areas.
6. Plot the tumor cells' DAB measurement on a histogramLoading..., set a cutoff, and build positive ÷ total × 100 in the Statistics Report.

The result approximates the Tumor Proportion Score, which the published convention scores on partial or complete membrane staining at any intensity, so validate the cutoff against manual scoring of the same slides. TissueGnostics states that the system is for research use only and that its results must be reviewed by a qualified professional.