Written and maintained by CASRAI Editorial Board
Last updated
Immunohistochemistry (IHC) uses antibodies to show where a protein sits inside a tissue section. Done well it is one of the most informative techniques in biology, because it preserves spatial context that a Western blot or a lysate assay destroys. Done badly it produces confident-looking brown staining that means nothing — and IHC is unusually easy to do badly, because a failed experiment and a real result can look identical.
This guide covers the workflow end to end, the controls that separate a trustworthy result from a decorative one, how to troubleshoot by observed symptom, and the antibody-validation problem that sits underneath the whole technique.
The workflow
Every IHC protocol is a variation on the same sequence. Understanding what each step is for is what lets you fix it when it fails.
1. Fixation
Fixation cross-links proteins to stop degradation and hold structure in place. Formalin (formaldehyde) is the standard. The tension to understand: fixation preserves morphology and destroys epitopes at the same time. Under-fix and the tissue degrades; over-fix and the antibody can no longer reach or recognise its target. Fixation time is therefore a real experimental variable, not a housekeeping detail — and it is the most common uncontrolled difference between a batch that works and one that does not, especially with archival clinical material whose fixation history you do not know.
2. Processing, embedding and sectioning
Tissue is dehydrated, cleared, and embedded in paraffin (FFPE) or frozen (cryosections). FFPE gives far better morphology and stores indefinitely; frozen sections preserve epitopes better and are faster, at the cost of structure. Which you have constrains everything downstream.
3. Antigen retrieval
This step exists to undo what fixation did — breaking cross-links so the epitope becomes accessible again. Two families:
- Heat-induced epitope retrieval (HIER) — heating in a buffer, usually citrate at acidic pH or EDTA/Tris at basic pH. The buffer and pH are antibody-specific and often the single variable that decides whether staining works at all.
- Proteolytic-induced epitope retrieval (PIER) — enzymatic digestion. Harsher on morphology, and easy to overdo.
If an antibody works on Western blot but not on FFPE tissue, retrieval is the first thing to interrogate.
4. Blocking
Two distinct blocking problems, often confused:
- Endogenous enzyme activity — tissue peroxidase (blocked with hydrogen peroxide) or alkaline phosphatase, which would otherwise produce signal wherever the enzyme is, independent of your antibody.
- Non-specific protein binding — blocked with serum, ideally from the species the secondary antibody was raised in, or with BSA.
Also relevant: endogenous biotin, if you are using a biotin-based detection system.
5. Primary antibody
Concentration and incubation time and temperature are all titratable, and all interact. Over-concentrated primary is a leading cause of background that gets misread as signal.
6. Detection
Direct detection (labelled primary) is simpler but less sensitive. Indirect detection (labelled secondary against the primary) amplifies signal and is standard. Polymer-based systems give further amplification without the endogenous-biotin problem that avidin-biotin systems carry.
7. Chromogen and counterstain
DAB gives the familiar permanent brown precipitate; AEC gives red but is alcohol-soluble. Haematoxylin counterstain provides the nuclear context you need to interpret localisation at all.
The controls, and why they are not optional
This is the section that separates publishable IHC from decoration. An IHC image without controls is uninterpretable, and reviewers increasingly say so.
| Control | What it rules out |
|---|---|
| No-primary control | Signal generated by the secondary antibody or the detection system alone. |
| Isotype control | Non-specific binding attributable to the antibody class rather than its specificity. Must match the primary’s isotype AND concentration. |
| Positive-control tissue | A false negative. A tissue known to express the target proves the protocol works at all. |
| Negative-control tissue | A false positive. A tissue known not to express the target should be blank. |
| Absorption / peptide-block control | That the signal depends on the specific epitope. |
| Knockout or knockdown tissue | Everything above, definitively. This is the gold standard and is discussed below. |
One caution worth stating plainly: the isotype control is weaker than it is usually treated as being. A matched isotype at matched concentration tells you about class-driven background; it does not tell you your antibody binds the protein you think it binds. Many published “negative controls” are isotype controls doing work they cannot do.
The antibody-validation problem
Underneath every IHC result is an assumption: that the antibody binds the intended target and not something else. That assumption fails often enough to have become a recognised contributor to the reproducibility crisis in biomedical research. Commercial antibodies are frequently sold with validation data that does not cover the application you are using them for — an antibody validated on Western blot is not thereby validated for IHC on FFPE tissue, because the epitope is presented completely differently.
An International Working Group for Antibody Validation was convened specifically to address this and published proposed standards in Nature Methods (Uhlén et al., 2016). Their framework sets out validation strategies that are application-specific rather than generic, including:
- Genetic strategies — knockout or knockdown of the target, then confirming signal loss. The strongest available evidence.
- Orthogonal strategies — comparing antibody-derived signal against an antibody-independent measure of the same protein’s abundance.
- Independent antibody strategies — two antibodies raised against different epitopes of the same protein giving concordant results.
- Tagged-protein expression — comparing against a tagged version of the target.
- Immunocapture followed by mass spectrometry — identifying what the antibody actually pulled down.
Open-science efforts such as YCharOS now characterise commercial antibodies against knockout controls and publish the results, and their findings have repeatedly shown that a substantial share of widely-used commercial antibodies do not perform as advertised.
Practical implication: record the antibody’s vendor, catalogue number, lot number and RRID in your methods section. RRIDs (Research Resource Identifiers) exist precisely so that an antibody can be unambiguously identified years later, and many journals now require them. “Anti-X antibody (Abcam)” is not a reproducible methods statement.
Troubleshooting by symptom
No staining at all
Check the positive control first — if it also failed, the problem is protocol-level, not sample-level. Then: antigen retrieval (wrong buffer or pH is the most common cause), primary antibody too dilute or degraded, wrong secondary species, detection reagents expired, or the target genuinely absent. Over-fixation of the specific block is a frequent culprit with archival material.
Weak staining
Increase primary concentration or incubation time, switch to a more sensitive polymer detection system, optimise retrieval, or extend chromogen development — but develop the control slides alongside so you can tell amplified signal from amplified background.
High background everywhere
Primary too concentrated, insufficient blocking, inadequate washing, endogenous peroxidase not quenched, or sections dried out at some point (drying causes irreversible non-specific binding). Increase wash stringency before increasing block.
Edge or rim staining
Classically a fixation-penetration artefact — the tissue edge fixed properly and the centre did not, or vice versa. Also caused by sections drying at the edges.
Nuclear staining where you expect cytoplasmic (or vice versa)
Treat this as a specificity alarm, not a curiosity. Confirm with a second antibody against a different epitope before interpreting it as biology.
Uneven staining across the section
Incomplete reagent coverage, uneven heating during retrieval, or section thickness variation from the microtome.
Quantification, and its limits
DAB chromogen is not stoichiometric — the amount of brown is not linearly proportional to the amount of antigen, because DAB deposition is an enzymatic amplification that saturates. This matters because it makes densitometry on DAB-stained IHC unreliable for anything beyond coarse comparison.
If you need genuine quantification, the honest options are semi-quantitative scoring by a blinded observer using a defined scale (H-score, Allred), digital image analysis with carefully validated thresholds, or a switch to immunofluorescence with proper controls, which behaves better quantitatively. Whichever you choose, blinding matters — IHC scoring is subjective and unblinded scoring is a documented source of bias.
What research administrators and lab managers should know
- Human tissue triggers governance obligations. IHC on human specimens generally requires IRB review or a documented determination that it is not human-subjects research, plus attention to consent scope for archival material and to any biobank material transfer agreement. Do not assume “the samples were already collected” resolves this.
- Antibodies are a real budget line, and validated antibodies cost more than unvalidated ones. That difference is far cheaper than an unreproducible result.
- Histology cores typically handle processing, embedding, sectioning and often staining on a recharge basis. Budget core time explicitly on grant applications, and settle core-staff authorship expectations early.
- Image integrity rules apply. Adjustments must be applied to the whole image, must be disclosed, and must never selectively alter a region. Retain unadjusted originals — journals increasingly request them, and inappropriate image manipulation is a documented cause of retraction.
Frequently asked questions
What is the difference between IHC and immunofluorescence?
Both use antibodies to localise proteins. IHC typically uses an enzyme and a chromogen to produce a coloured precipitate viewable on a light microscope; immunofluorescence uses fluorophore-labelled antibodies read on a fluorescence or confocal microscope. IF multiplexes more easily and quantifies better; IHC gives permanent slides and better morphological context on standard equipment.
Why is antigen retrieval necessary?
Formalin fixation cross-links proteins, which masks epitopes so the antibody cannot bind. Retrieval reverses enough of that cross-linking to make the epitope accessible again.
Which retrieval buffer should I use?
It is antibody-specific. Citrate (acidic) and EDTA/Tris (basic) are the common starting points, and the antibody datasheet is the first place to look. If staining fails, testing both is a standard early optimisation.
Is an isotype control enough?
No. It controls for class-driven non-specific binding but says nothing about whether your antibody binds the intended target. Genetic controls — knockout or knockdown — are the strong form of that evidence.
Can I quantify IHC by measuring DAB intensity?
Only coarsely. DAB deposition is enzymatically amplified and saturates, so it is not linearly proportional to antigen amount. Use blinded semi-quantitative scoring, validated digital analysis, or immunofluorescence if you need real quantification.
What should I report in my methods section?
Antibody vendor, catalogue number, lot, RRID and dilution; fixation type and duration; retrieval method, buffer and pH; detection system; and every control you ran. That is what makes the experiment repeatable by someone else.
