10.2 Adverse Outcome Pathways (AOP): Molecular Initiating Events, Key Events & Organ-Level Outcomes

Why AOPs are a Domain II.4 companion to MOA

Section 10.1 evaluated a chemical-specific MOA. Handbook II.4 also expects you to follow a perturbation from a molecular interaction to organ-level and organism outcomes. The Organisation for Economic Co-operation and Development (OECD) adverse outcome pathway (AOP) is the shared vocabulary for that sequence when you are not tying it to one substance’s metabolism. Independent OpenExamPrep teaching in this section covers OECD AOP architecture, AOP-Wiki as a knowledge base, modularity, and how AOPs structure integrated approaches to testing and assessment (IATA). It is not an OECD, AOP-Wiki, or ABT product and does not claim official approval, review, or partnership with those bodies.

An AOP starts with a molecular initiating event (MIE): the initial chemical/stressor interaction with a biomolecule that perturbs biology (covalent protein haptenation; receptor agonism; acetylcholinesterase serine phosphorylation; DNA adduct formation). It proceeds through intermediate key events (KEs)—measurable, essential changes in biological state—and ends in an adverse outcome (AO) that matters for risk assessment or a regulatory decision (allergic contact dermatitis; hepatocellular carcinoma; population-level reproductive failure in fish). Key event relationships (KERs) are the causal, predictive links: if the upstream KE is in a given state, what can you infer about the downstream KE?

The AOP Developers’ Handbook (version 2.8, March 2026, succeeding the older “Users’ Handbook”) sits beside OECD ENV/JM/MONO(2013)6. You are not asked to author a wiki page on the examination. You are asked to use the nouns correctly and to know that an AOP is not a complete PBPK model, not a full chemical risk assessment, and not a claim that every chemical hitting the MIE will reach the AO. Severity, duration, ADME, repair, and modulating factors decide whether the cascade runs.

MIE, KEs, KERs, and AO

MIE. Specialized KE at the molecular interaction. It is still a KE: it must be measurable in principle (peptide reactivity, receptor occupancy, enzyme inhibition). “Oxidative stress” as a vague cloud is a poor MIE; “covalent binding to cysteine thiols on skin proteins” is a usable MIE.

KE. A change that is essential to progression toward that AO. Essentiality is tested when you can block the KE (knockout, antagonist, recovery) and the downstream events weaken or disappear. A correlative omics node that can be bypassed is a modulating factor, not a KE. Chapter 8’s warning still holds: an Nrf2 transcript without dysfunction is not an AO.

KER. Captures biological plausibility (does physiology allow upstream to drive downstream?) and empirical support (dose–response, incidence, temporal concordance—the same Hill-type ideas as MOA, now applied to a pair of events). Quantitative KERs (how much KE1 produces how much KE2) are the prize for predictive toxicology; many AOPs are still qualitative.

AO. An apical effect used in a decision: organism disease, death, or an ecological endpoint such as impaired reproduction or population decline. Organ-level outcomes (follicular-cell hyperplasia; centrilobular necrosis) often sit as late KEs immediately before the AO. Do not stop the pathway at a convenient in vitro readout and call that the AO unless the decision problem truly is that readout.

The simple teaching chain is MIE → KE1 → KE2 → AO. Real AOPs may have more KEs. The OECD skin sensitization AOP, the cleanest regulatory example, is often listed as protein haptenation (MIE), keratinocyte inflammatory responses, dendritic-cell activation, T-cell priming/proliferation, then allergic contact dermatitis. The diagram below compresses the cellular immune KEs so you can see the four-box shape the examination stem will draw.

Modularity and AOP networks

KEs and KERs are modular. The same “Sustained AhR activation” KE can feed more than one AO. The same “hepatocyte death” KE can sit downstream of different MIEs (reactive metabolite versus BSEP inhibition). AOP networks are those shared nodes drawn together. Modularity is why you do not rewrite biology from scratch for every CAS number: you reuse KE descriptions and point a new chemical at the MIE it actually hits.

Modularity is also a trap. Sharing a KE does not mean two chemicals share potency, ADME, or even the same AO mix. A chemical can launch several AOPs at once (a sensitizer that is also a mutagen). An AOP does not encode first-pass extraction; that still lives in the chemical-specific MOA and in PBPK.

AOP-Wiki, conceptually

AOP-Wiki (aopwiki.org) is the collaborative knowledge base that supports the OECD AOP development programme. Release 2.8 of the wiki and handbook landed in March 2026. Entries typically include prototypical stressors (examples that illustrate the MIE, not an exhaustive chemical list), KE and KER pages that can be reused, and a status that may be under development, internally reviewed, or further along toward OECD endorsement. Browsing a page is not the same as citing an OECD Test Guideline. Many living entries are hypotheses with incomplete essentiality data. Examination skill: use Wiki vocabulary; do not treat every page as adopted regulatory text.

Prototypical stressors help teaching (2,4-dinitrochlorobenzene for sensitization haptenation; a named AChE inhibitor for cholinergic AOPs). They do not convert the AOP into a chemical-specific MOA. If the stem names a chemical, you still need that chemical’s metabolism, route, and potency to say whether it will traverse the AOP.

MOA (chemical-specific) versus AOP (chemical-agnostic)

This contrast is high-yield.

A chemical’s MOA may instantiate one or more AOPs plus ADME. Phenobarbital-type thyroid tumors: the AOP is “increased hepatic TH clearance → ↓T4 → ↑TSH → follicular proliferation → tumors.” The MOA for phenobarbital adds CAR activation, which UGTs are induced, rat versus human TBG, and the human-relevance call from section 10.1. Do not call the AOP “not relevant to humans” as a property of the pathway diagram; relevance is a species and chemical mapping problem.

How AOPs inform testing strategies and IATA

IATA (OECD Series on Testing and Assessment, including No. 260 on using AOPs to develop IATA) is a pragmatic, problem-driven integration of existing information with new tests chosen because they query named KEs. You start with the decision (skin sensitizer: yes/no/potency; endocrine activity; aquatic reproductive hazard), map data gaps onto KEs, and run the smallest set of methods that can fill those gaps—in silico, in chemico, in vitro, then targeted in vivo if still needed. IATA still allows expert judgment in how streams are combined.

A defined approach (DA) is stricter: a fixed set of information sources and a fixed data interpretation procedure (DIP). OECD Test Guideline 497 defined approaches for skin sensitization are the working example. Non-animal methods map onto early KEs: OECD 442C (peptide/protein binding, MIE), 442D (keratinocyte activation), 442E (dendritic-cell activation). Individual assays are generally not stand-alone replacements for classification; the DA’s DIP (for example a 2-out-of-3 hazard call, or an integrated testing strategy that adds in silico potency) is what yields a prediction inside the stated applicability domain. Chapter 6 taught LLNA EC3, guinea-pig tests, and when pharmaceuticals run these packages. Here the point is architectural: the AOP tells you which KE each assay is for, so you do not treat a keratinocyte reporter as a full animal LLNA or as a mutagenicity test.

Other testing-strategy uses: grouping chemicals that share an MIE; choosing a fish vitellogenin or aromatase assay because an endocrine AOP says that KE sits on the path to impaired reproduction; skipping a two-year bioassay question that an established non-genotoxic AOP plus IATA already frames as a different study type. AOPs also support read-across when analogues share an MIE and you can argue they will reach it in vivo—without pretending read-across is automatic.

Scenario

A new acrylate for a dermal adhesive is peptide-reactive (442C positive), activates a keratinocyte reporter (442D positive), and is negative in a dendritic-cell assay (442E) at soluble concentrations. A 2-out-of-3 DA may still call a sensitizer hazard if two KEs are positive and the chemical sits in domain; a single negative KE does not erase the MIE. The AOP does not tell you workplace airborne concentration. That is exposure assessment (chapter 11).

A second file is an organophosphate. The AOP (AChE inhibition → acetylcholine excess → cholinergic AO) is chemical-agnostic. The MOA for that OP still needs aging versus reactivation, whether neuropathy-target esterase is also inhibited, and species cholinesterase differences. Treating every insecticide as this AOP was the chapter 9 class error; treating this AOP as a full chemical risk assessment is the chapter 10 error.

A third file cites an AOP-Wiki page “under development” with no essentiality data as if it were OECD 497. The pathway is a hypothesis. It can still guide a literature search; it cannot, by itself, replace a required apical study without an accepted IATA/DA and a defined decision context.

Traps

• Calling an AOP chemical-specific, or calling a MOA chemical-agnostic.
• Treating every AOP-Wiki page as an endorsed test guideline.
• Equating a single KE assay with the AO.
• Forgetting ADME when mapping a chemical onto an MIE (a potent in chemico hapten that is instantly hydrolyzed in blood may never reach skin proteins in vivo).
• Using IATA and defined approach as synonyms: DAs are fixed recipes; IATA can include expert integration.