10.1 Mode of Action (MoA) Framework & Human Relevance Evaluation

Why mode of action is a Domain II.4 skill

Handbook II.4 asks you to evaluate mode of action (MOA) and human relevance, including sensitive subpopulations. Independent OpenExamPrep teaching in this section covers the International Programme on Chemical Safety (IPCS) and International Life Sciences Institute (ILSI) frameworks that regulators actually cite, three species-specific MOAs that often fail the human-relevance test, and a preview of why a mutagenic versus non-mutagenic cancer MOA changes the low-dose family that chapter 12 will quantify. This material is not an ABT, IPCS, ILSI, IARC, or EPA product and does not claim official approval, review, or partnership with those bodies.

MOA is a biologically plausible sequence of key events from a chemical interaction to a defined adverse effect. Mechanism of action is the finer molecular description (which residue, which reactive orbital). EPA cancer guidelines and DABT items usually ask for MOA: can you name necessary, measurable steps and test them? A complete crystal structure is not required. Chapter 8 gave organ-level sketches (zone-3 CYP2E1 necrosis; male-rat α2u droplets). This section is the structured evaluation: postulated MOA, concordance, alternatives, uncertainties, then the human-relevance questions. Adverse outcome pathways (AOPs) in section 10.2 are the chemical-agnostic cousins of that work.

Postulated MOA and key events

Write the postulated MOA as a theory of the case before you hunt papers to decorate it. Example: “Chemical X binds male-rat α2u-globulin → lysosomal overload in proximal tubules → cytotoxicity → sustained regeneration → male-rat renal tubule tumors.” Each arrow should be a key event: empirically observable, and argued as necessary for the adverse outcome in that species. “The liver looks busy” is not a key event. “Centrilobular necrosis with a four-fold ALT rise at doses that later produce regenerative nodules” can be.

IPCS built the cancer MOA framework in Sonich-Mullin et al. (2001) using modified Bradford Hill considerations originally developed for causality. Meek et al. (2003) and Boobis et al. (2006) added human relevance for cancer. Boobis et al. (2008) extended the same logic to non-cancer effects. Chapter 9 used Bradford Hill on epidemiologic chemical–disease associations. Here the Hill-type questions are applied to key events inside an animal MOA. Mixing those jobs is a common item trap: a strong occupational cohort does not by itself prove a rodent key-event sequence, and a tidy rodent cartoon does not by itself prove human disease.

Concordance, plausibility, alternatives, uncertainties

Dose–response concordance. Key events should appear at doses that can produce the adverse outcome, with earlier events at similar or lower doses than later ones. If tumors appear at 5 mg/kg/day but the “necessary” cytotoxicity is only seen at 500 mg/kg/day, the postulated MOA is not concordant. Incidence or severity of early events should rise with dose in a way that can support the later event.

Temporal concordance. Key events must precede the adverse outcome. A protein adduct measured after tumors are already present does not establish initiation. For cancer, early key events should be visible in subchronic windows if the MOA claims they drive a two-year tumor. For non-cancer, a TSH rise should precede follicular-cell hyperplasia if TSH is the proposed driver.

Strength, consistency, and specificity. The association of key events with the outcome should repeat across studies, sexes (when the biology predicts both sexes), and related chemicals that share the interaction. Specificity helps when only one organ and one sex show the package (α2u is the teaching case). Lack of specificity does not kill a MOA by itself—many genuine MOAs hit more than one organ.

Biological plausibility and coherence. The sequence must fit known physiology. Lysosomal overload causing regeneration is plausible. Claiming that a nuclear receptor in liver directly mutates DNA without adducts or clastogenicity needs extra evidence. Coherence asks whether the story contradicts what is already known of that organ’s natural history (for example, treating male-rat hyaline droplets as a glomerular immune-complex disease).

Alternative MOAs. Ask what else could produce the same apical effect. A male-rat kidney tumor chemical that is also Ames-positive and produces tumors in mice is not “α2u only.” A peroxisome proliferator that also activates constitutive androstane receptor (CAR) or causes cytotoxicity independent of peroxisome proliferator-activated receptor alpha (PPARα) needs those alternatives on the table. Failure to name alternatives is not the same as having excluded them.

Uncertainties, inconsistencies, and data gaps. List them. Missing immunohistochemistry for α2u, missing TSH time course, missing genotoxicity battery, or conflicting PPARα-null results are uncertainties. Uncertainty is not a free pass to declare non-relevance. If the animal MOA is not established, the human-relevance framework does not let you drop the finding. The default is that an unexplained animal adverse effect remains potentially human-relevant until a MOA that fails human relevance is actually supported.

Human relevance: three questions

ILSI’s Risk Science Institute and IPCS framed human relevance as a concordance analysis of key events, not a popularity vote of committees.

1. Is the weight of evidence sufficient to establish the MOA in animals? If no, stop. You cannot exclude human relevance. Use default risk methods for that endpoint (including, for cancer of unknown MOA, the protective linear default previewed below).
2. Are key events in the animal MOA qualitatively plausible in humans? If a key event cannot occur in humans as a matter of biology (no comparable androgen-driven α2u load; no rodent forestomach), the MOA may be not relevant for that apical effect.
3. Taking kinetic and dynamic factors into account, are the key events quantitatively plausible in humans at relevant exposures? Humans may have the same receptor but far less mitogenic response, or far slower clearance. Large quantitative differences can support “not relevant at plausible exposures” even when the qualitative hardware exists. Those same differences later inform chemical-specific adjustment factors (CSAFs) in chapter 13 when the MOA is relevant.

Qualitative “no” (question 2) is a fundamental species block. Quantitative “no” (question 3) is a magnitude-and-exposure argument and is easier to over-claim. Write both answers, then a confidence statement and implications for hazard identification versus dose–response.

Case 1: male-rat α2u-globulin nephropathy

Adult male rats synthesize large amounts of hepatic α2u-globulin under androgen control. Some hydrocarbons and metabolites (d-limonene 1,2-epoxide, 2,2,4-trimethylpentane metabolites, decalin, some gasoline streams) bind that protein. The complex resists lysosomal degradation, appears as hyaline droplets in proximal tubules, and can drive single-cell necrosis, granular casts, papillary mineralization, regeneration, and, chronically, male-rat renal tubule tumors. Female rats, mice, and humans do not run this androgen-driven protein load. Humans have related lipocalins at much lower abundance; they do not reproduce the syndrome. For this MOA, human relevance of the male-rat renal tumors and the associated male-only nephropathy is limited. EPA’s 1991 Risk Assessment Forum advice still shapes examination language: those male-rat kidney tumors do not contribute to the qualitative human cancer hazard when the α2u package is complete, and that nephropathy is not used as the RfD critical effect.

The package is not a slogan. Typical expectations include increased hyaline droplets, identification of α2u in those droplets, male-only occurrence, cytotoxicity/regeneration, and lack of mutagenicity as the driver. If female rats or mice have kidney tumors, if the chemical is DNA-reactive, or if other organs are neoplastic, you evaluate case by case—do not launder the whole file through α2u. tert-Butanol teaching in IRIS discussions is the cautionary mixed file: α2u evidence can be present and chronic progressive nephropathy (CPN) or female-rat kidney changes can keep a human-relevant renal story alive. Chapter 8 taught the droplets as an organ mechanism; here you must document the MOA and then apply question 2.

Case 2: PPARα rodent liver tumors (with caveats)

PPARα agonists (teaching chemicals include WY-14,643, fibrate drugs such as clofibrate, fenofibrate, and gemfibrozil, and some phthalates) produce rodent peroxisome proliferation, hepatocellular hypertrophy, altered growth signaling, increased S-phase / decreased apoptosis, clonal expansion of foci, and hepatocellular adenomas/carcinomas. Key events used in consensus reviews (Corton et al. 2014 and related workgroups) are receptor activation, perturbation of hepatocyte growth pathways, altered growth and survival, and selective clonal expansion.

Human hepatocytes and primate livers show the lipid-metabolism arm of PPARα (why fibrates lower triglycerides) but little peroxisome proliferation and little mitogenic burst. Humanized-PPARα mice can retain lipid-gene induction without the rodent proliferative tumor pathway. Workgroups have concluded that the rodent liver-tumor MOA is not relevant or unlikely to be relevant to humans, differing mainly in how absolute they are willing to be. That is a question 3 (and partly question 2 for downstream KEs) result for that apical tumor, not a claim that PPARα is absent in people.

Caveats the examination rewards. (1) Receptor activation in humans is real; hepatic enzyme induction and lipid changes can still be non-cancer findings. (2) CAR/PXR phenobarbital-type rodent liver tumors are a different nuclear-receptor MOA—do not merge them with PPARα. (3) Some chemicals, di(2-ethylhexyl) phthalate (DEHP) among them, have been argued to carry additional inflammatory, oxidative, or CAR-related pathways; PPARα-null or mixed-tumor datasets are why you do not rubber-stamp every phthalate liver tumor as “fibrates, therefore ignore.” (4) Even Corton-type panels split between “not relevant” and “unlikely,” which is a confidence statement, not a license to skip alternative MOAs.

Case 3: thyroid disruption via hepatic enzyme induction in rodents

Phenobarbital-type CAR inducers and some other microsomal-enzyme inducers increase hepatic UDP-glucuronosyltransferase (UGT) (and often sulfation and uptake transporters) toward thyroid hormone. In rats, thyroxine (T4) clearance rises, serum T4 falls, pituitary thyroid-stimulating hormone (TSH) rises, and thyroid follicular cells hypertrophy, proliferate, and, with chronic TSH drive, can form tumors. Supplemental T4 can block phenobarbital’s tumor-promoting effect in classic initiation–promotion designs, tying the apical tumor to the TSH loop rather than to a thyroid mutagen.

Why adult human thyroid tumors often fail relevance for this MOA. Rats have a short T4 half-life (hours), limited high-affinity thyroxine-binding globulin (TBG) compared with humans, and a follicular epithelium that proliferates readily in response to TSH. Adult humans have TBG, a T4 half-life on the order of a week, and a much quieter proliferative response to modest TSH changes. Decades of human anticonvulsant use have not reproduced a rodent-like thyroid-cancer epidemic from this secondary TSH mechanism.

Caveats. UGT is not the only clearance path: Gunn-rat and transporter work shows T4 can fall even when UGT1A is crippled, so “UGT induction” is a teaching headline, not a complete kinetic inventory. Enzyme inducers that do not raise TSH do not earn the follicular-tumor MOA. Developmental hypothyroxinemia can still be human-relevant even when adult rodent thyroid tumors are not: the fetal and neonatal brain needs thyroid hormone on a timetable rats and humans share more closely than they share adult TSH-driven mitogenesis. Iodine-deficient populations, premature infants, and people on replacement levothyroxine are sensitive subpopulations for circulating-hormone effects. Do not use a “rodent thyroid tumors are not relevant” sentence to dismiss a T4 drop in pregnancy.

Sensitive subpopulations

Human-relevance question 3 is where II.4 and II.3 meet. Life stage (fetus, neonate, elderly), sex, genetic variants that change activation or clearance, pre-existing liver or kidney disease, iodine status, and combined exposures can make a quantitatively “implausible” average-adult story plausible in a defined group. A MOA that fails for male-rat kidney tumors can still leave a developmental or second-organ MOA untouched. Name the subpopulation and the key event they distort (clearance, receptor expression, repair reserve), not a generic “children are sensitive.”

Mutagenic versus non-mutagenic cancer MOAs (preview of chapter 12)

Once a cancer MOA is human-relevant (or cannot be excluded), EPA’s 2005 Guidelines for Carcinogen Risk Assessment change the low-dose family:

• Mutagenic / DNA-reactive MOA. Default linear extrapolation from a point of departure through the origin. The public-health rationale is that a mutational event can contribute to cancer without a practical threshold you can defend for regulation.
• Non-mutagenic MOA with well-supported key events that have a practical threshold (cytotoxicity with regenerative proliferation; some receptor-mediated mitogenic stories that survive human relevance). A nonlinear (reference-dose-like or margin-of-exposure) approach may be supported below the point of departure.
• Unknown or mixed MOA. Linear methods are often retained as a health-protective default when mutagenicity cannot be reasonably ruled out.

Chapter 9 taught that IARC groups do not choose EPA math. This section teaches that MOA evaluation does. Chapter 12 will attach cancer slope factors, unit risks, BMD/BMDL, and first-in-human math to that choice. Do not linearize a fully documented, human-non-relevant α2u kidney tumor, and do not invent a threshold for a DNA-reactive human leukemia hazard because the milligrams look small.

Scenario

A gasoline-range hydrocarbon produces male-rat hyaline droplets with α2u staining, granular casts, no female-rat or mouse kidney tumors, and a negative Ames/micronucleus package. Question 1 can be “yes” for the α2u MOA; question 2 is “no” for that renal-tumor sequence in humans. The same dossier still needs liver, inhalation portal-of-entry, and any mutagenic impurities treated on their own MOAs.

A fibrate-like agonist produces rodent liver tumors, peroxisome proliferation, and S-phase induction; human hepatocyte cultures show lipid-gene changes without proliferation; epidemiology of therapeutic fibrates does not show a liver-cancer signal. The rodent liver-tumor MOA is a strong candidate for limited human relevance, but you still read CAR assays, genotoxicity, and any non-liver tumors. A DEHP-like phthalate with messy PPARα-null or extra-hepatic tumors is not automatically the fibrate file.

A CAR inducer drops rat T4, raises TSH, and produces follicular tumors, with T4 supplementation blocking promotion. Adult human thyroid-tumor relevance for that TSH loop is weak. A concurrent gestational T4 decrement is a different endpoint and a different subpopulation—do not file it under “thyroid tumors, ignore.”

Traps

• Declaring non-relevance before the animal MOA is established (skipping question 1).
• Using α2u to dismiss female-rat or mouse kidney tumors or a mutagenic battery.
• Equating “PPARα is a human drug target” with “rodent liver tumors are default human liver carcinogens,” or the opposite error of discarding all phthalate tumors as PPARα-only.
• Treating rodent TSH-driven follicular tumors and developmental hypothyroxinemia as the same human-relevance call.
• Letting an IARC group pick linear versus nonlinear math, or inventing a threshold for a DNA-reactive MOA because chapter 12’s numbers look kinder that way.