Experimental Models of Autoimmune Hepatitis: Disease Fidelity and Translational Relevance

Autoimmune hepatitis (AIH) remains difficult to study mechanistically in patients because disease initiation is rarely observed directly, the relevant autoantigens differ across subsets, and clinically meaningful outcomes such as chronic inflammation, fibrosis, relapse, and treatment response evolve over time. Animal models therefore remain indispensable. At the same time, the field has become increasingly heterogeneous. Acute immune-mediated hepatitis systems, especially concanavalin A (ConA), dominate the recent literature because they are rapid, inexpensive, and experimentally tractable. However, ConA-induced hepatitis is not an antigen-driven autoimmune response and is better interpreted as acute bystander immune-mediated liver injury than as a stand-alone model of chronic AIH. In contrast, antigen-driven adenoviral models based on cytochrome P450 2D6 (CYP2D6) or formiminotransferase cyclodeaminase (FTCD), as well as genetically predisposed or spontaneous tolerance-defect models, provide stronger insight into loss of hepatic tolerance, chronicity, fibrosis, and the interaction between antigenic context and host susceptibility. This review proposes a pragmatic framework for evaluating AIH models on the basis of face validity, construct validity, predictive validity, chronicity, host susceptibility, and mechanistic fitness for a specific biological question. Using that framework, we classify current models into acute surrogate models, immunization- and xenoantigen-based systems, adenoviral antigen-driven chronic models, spontaneous and genetically predisposed models, transgenic or neoantigen-driven tolerance models, and humanized or microbiota-sensitive hybrid systems. We then synthesize what these systems have taught the field about central and peripheral tolerance, MHC and non-MHC genetic susceptibility, sex- and age-related disease context, CD4⁺ and CD8⁺ T-cell biology, B-cell function, regulatory T-cell instability, impaired suppressive function, defective IL-2-dependent regulation, macrophage and innate lymphocyte participation, cell-death programmes, and gut-liver or liver-microbiome interactions. A central conclusion emerges: AIH models are complementary tools with markedly different levels of disease fidelity, and they should not be treated as interchangeable. Acute systems remain useful for effector-phase biology and first-pass intervention studies, but the strongest translational inferences for chronic AIH come from antigen-defined chronic models and selected tolerance-defect systems. Future progress will depend on better benchmarking across models, stronger alignment with human immune profiling and ex vivo validation platforms, and wider use of tolerance-restoring rather than purely anti-inflammatory therapeutic strategies.