Immunohistochemistry (IHC)

The fundamental mechanism of IHC relies on the paratope-epitope interaction between a primary antibody and the target protein antigen. This binding is governed by non-covalent forces—hydrogen bonds, hydrophobic interactions, and van der Waals forces—and is validated through affinity constants (Kd).

• Monoclonal antibodies: Bind a single epitope, offering high specificity and low cross-reactivity.
• Polyclonal antibodies: Recognize multiple epitopes, increasing signal intensity but potentially reducing specificity.
• Cross-reactivity must be systematically ruled out using isotype controls and tissue known to lack the target protein.

IHC visualizes the antibody-antigen complex through an enzyme-substrate reaction that deposits a colored precipitate at the binding site. The most common system uses horseradish peroxidase (HRP) or alkaline phosphatase (AP) conjugated to a secondary antibody.

• HRP + DAB (3,3'-Diaminobenzidine): Produces a brown, alcohol-insoluble precipitate that is permanent and compatible with hematoxylin counterstaining.
• Polymer-based detection: Dextran backbones carrying multiple enzyme molecules amplify signal without the background noise of older avidin-biotin complex (ABC) methods.
• Tyramide signal amplification (TSA): HRP catalyzes the deposition of fluorescently labeled tyramide radicals, enabling detection of low-abundance targets.

Formalin fixation cross-links proteins via methylene bridges, masking epitopes and preventing antibody binding. Antigen retrieval reverses this masking through controlled physicochemical treatments.

• Heat-induced epitope retrieval (HIER): Uses citrate (pH 6.0) or Tris-EDTA (pH 9.0) buffers at 95-100°C to hydrolyze cross-links.
• Enzymatic retrieval: Proteinase K or trypsin digestion physically cleaves cross-linked proteins.
• Over-fixation and under-fixation are both sources of false-negative results, making standardized fixation protocols critical for reproducible IHC.

IHC interpretation has evolved from qualitative visual estimation to continuous digital quantification using whole-slide image analysis algorithms.

• H-score: A weighted score calculated as Σ(Pi × i), where Pi is the percentage of cells at each staining intensity (0, 1+, 2+, 3+), yielding a range of 0-300.
• Allred score: Combines proportion score (0-5) and intensity score (0-3) for a total of 0-8, commonly used for estrogen receptor (ER) assessment.
• Computational pixel classification using color deconvolution separates DAB signal from hematoxylin counterstain, enabling objective, reproducible quantification of biomarker expression.

Traditional single-marker IHC is being augmented by multiplex immunohistochemistry (mIHC) techniques that simultaneously detect multiple proteins on a single tissue section, preserving spatial relationships.

• Sequential staining and stripping: Cycles of antibody incubation, imaging, and antibody removal allow detection of 5-7 markers on a single slide.
• Spatial proximity analysis: Quantifies the distance between different cell phenotypes, such as PD-1+ T cells and PD-L1+ tumor cells, which is a stronger predictor of immunotherapy response than cell density alone.
• Tissue segmentation into tumor, stroma, and immune compartments enables compartment-specific biomarker quantification.

Clinical IHC assays require rigorous validation to ensure reproducibility across laboratories, operators, and reagent lots.

• Positive tissue controls: Sections known to express the target antigen at defined levels (high, low, negative) are included in every staining run.
• Negative reagent controls: Replacing the primary antibody with an isotype-matched immunoglobulin confirms that staining is not due to non-specific Fc receptor binding.
• External quality assessment (EQA): Programs like NordiQC and UK NEQAS distribute standardized tissue microarrays to participating laboratories, measuring inter-laboratory concordance using Cohen's Kappa statistics.

A computational pre-processing technique that standardizes the color appearance of IHC and H&E images to reduce variability caused by different staining protocols, scanner models, and reagent batches. Critical for robust downstream quantification.

• Methods include histogram matching, Macenko, and Vahadane decomposition
• Generative adversarial networks (GANs) enable style-transfer-based normalization
• Essential for multi-institutional studies and federated learning pipelines

A standardized immunohistochemical assessment of human epidermal growth factor receptor 2 overexpression on breast cancer cell membranes. Scores range from 0 to 3+ based on staining intensity and completeness of membrane staining.

• Score 0/1+: Negative, no anti-HER2 therapy indicated
• Score 2+: Equivocal, reflexed to FISH for confirmation
• Score 3+: Positive, strong complete circumferential membrane staining in >10% of tumor cells
• Deep learning models now automate scoring with concordance rates exceeding 90% against pathologists

A proliferation biomarker calculated as the percentage of tumor cells staining positive for the Ki-67 nuclear protein. Used to assess tumor aggressiveness and guide treatment decisions in breast cancer and neuroendocrine tumors.

• Manual counting suffers from high inter-observer variability
• Computational methods use nuclear segmentation followed by DAB/hematoxylin color deconvolution
• Hotspot vs. global averaging remains a debated quantification strategy

An advanced imaging technique that simultaneously labels multiple protein markers on a single tissue section using distinct fluorophores, enabling spatial profiling of the tumor microenvironment beyond single-plex IHC.

• Typical panels include 6–8 markers (e.g., CD8, PD-L1, CK, CD68)
• Requires spectral unmixing to separate overlapping fluorophore signals
• Enables spatial analysis of cell-cell interactions and immune exclusion patterns

A visualization technique that highlights the image regions most influential to a deep learning model's IHC scoring decision. Provides spatial interpretability by mapping attention weights back onto the original tissue image.

• Enables pathologists to verify that models focus on biologically relevant regions
• Critical for regulatory submissions requiring explainability
• Can reveal if models rely on spurious correlations like pen marks or stromal artifacts