Translational Relevance of Mouse Models of Atopic Dermatitis

The complexity of atopic dermatitis (AD) continues to present a challenge in the appropriate selection of a mouse model because no single murine model completely recapitulates all aspects of human AD. This has been further complicated by recent evidence of the distinct AD endotypes that are dictated by unique patterns of inflammation involving Th1, Th2, Th17, and Th22 axes. A review of currently used mouse models demonstrates that while all AD mouse models consistently exhibit Th2 inflammation, only some demonstrate concomitant Th17 and/or Th22 induction. As the current understanding of the pathogenic contributions of these unique endotypes and their potential therapeutic roles expands, ongoing efforts to maximize a given mouse model’s homology with human AD necessitates a close evaluation of its distinct immunological signature.


Introduction
Atopic dermatitis (AD) is a common, relapsing inflammatory skin condition characterized by pruritic, erythematous plaques and papules typically affecting the body's flexural surfaces. While AD is known to emerge due to barrier dysfunction, aberrant immune activation, and genetic predisposition, a clear understanding of the pathogenesis of its varying clinical presentations remains under investigation. Current knowledge of AD's multifaceted pathogenesis has been predicated on a diverse array of murine models that have played a pivotal role in delineating the functions of various susceptibility genes and exogenous triggers in the disease process.
However, the heterogeneity of AD disease in humans continues to present a challenge in selecting an appropriate mouse model for preclinical studies, given that no single model fully recapitulates all aspects of human AD. This has been further complicated by the recent identification of immunologically distinct human AD subtypes that occur due to differential inflammatory axis activation [1]. As the roles of these unique inflammatory patterns and their potential therapeutic implications in AD are further clarified, the selection of appropriate mouse models based on downstream immune pathways that modulate these clinically distinct subtypes is especially important in drug validation studies.
Thus, this review seeks to evaluate commonly used mouse models for AD and to highlight the immune pathways that are affected in mice. We hope to aid investigators in the selection of appropriate models that carefully balance immune factors alongside the underlying genetics, phenotype, and transcriptomic similarities to human AD in order to optimize their translational relevance in future studies.

An Overview of Mouse Models for Atopic Dermatitis
The current repository of AD murine models reflects a broad range of mechanisms used to induce eczematous dermatitis, including the use of exogenous agents, transgenic mice, and inbred mice. Several of these mechanisms, such as mucosa-associated lymphoid tissue lymphoma translocation protein 1 (MALT1) deficiency, fibroblast-specific inhibitor of nuclear factor kappa-beta subunit beta (Ikk2) deficiency, and Matt deficiency, have been only loosely correlated to human AD, and a clear understanding of their pathogenic contributions resulting in AD has yet to be fully delineated [2][3][4]. Nevertheless, the cutaneous inflammation observed in most models demonstrates significant overlap with key features found in human AD lesions, including elevated serum IgE, inflammatory infiltrate consisting of eosinophils, mast cells, and lymphocytes, increased epidermal thickness, hyperkeratosis, parakeratosis, acanthosis, and spongiosis [4][5][6].

Discussion
Ongoing efforts to demystify the complexity of AD have led to the identification of immunologically distinct phenotypes that vary based on permutations of Th1, Th17, and Th22 inflammation [1]. AD in Asian patients is thought to harbor heightened Th17/Th22 bias, AD in African Americans have Th2/Th22 skewing with Th17 attenuation, and European American lesions exhibit Th1/Th17 suppression [50]. Similarly, pediatric AD is notable for an unaltered Th1 profile against a background of Th17/Th22 upregulation [51], while adult AD is characterized by broad inflammatory induction of all axes [52]. Studies have also observed immune heterogeneity across adult subpopulations when stratified by age group, noting more robust Th22 upregulation in younger adults compared to older adults [52]. The potential to tailor treatments based on a patient's unique endotype represents an exciting new frontier in clinical drug discovery as the treatment paradigm for AD shifts from systemic interventions to targeted therapies [52].
Thus, the ability to reproduce these endotypes in mice represents a critical step toward optimizing the selection of translationally relevant mouse models, particularly in an era driven by the rising importance of personalized medicine [52]. Kim et al. (2019) previously outlined several features that warrant consideration in the selection of a mouse model, including gross phenotype/histology, serum profile, transcriptomic similarities, and immunophenotype [6]. More recently, Gilhar et al. (2020) proposed a list of criteria that animal models of human AD should meet, which included the consideration of inflammatory patterns [29]. An in-depth assessment of a model's unique immunological signature and the ability to individually modulate specific immune axes may be key for future preclinical studies seeking to assess how novel biologics blockade the inflammatory upregulation that defines specific AD endotypes, as selection based on genetics, phenotype, or gene expression profile may not adequately capture the immunologic complexity of AD. For instance, while Ikk2 ∆NES mice are notable for their transcriptomic homology with human AD, their lack of Th22 induction potentially limits their utility for the study of Asian, pediatric, and young adult endotypes that are notable for Th22 inflammation [3]. Similarly, NC/Nga mice characterized by high transcriptomic homology with the human AD transcriptome may be of limited translational relevance for the European American endotype given its marked Th17 induction [30,31,50].
Looking to the future, newer techniques such as the bioengineered humanized skin model [17] and the transplantation of stimulated peripheral blood mononuclear cells (PBMC) [29] hold great promise in their potential ability reproduce population-specific immune signatures in mice. While these techniques are further explored and characterized, a more practical approach to improving existing models' translational relevance can be guided by efforts to increase their versatility by exploring modifications that selectively modulate inflammatory responses, such as the use of CMIT/MIT prior to OVA sensitization to bolster Th17 response [36], or by characterizing the inflammatory response demonstrated by different mouse strains to aid in the selection of a model that most closely mimics a desired immune signature [28,37]. For instance, researchers seeking to model the European American AD endotype, noted for its Th17 attenuation, may use C57BL/6 mice instead of BALB/c mice within the ft model in order to abrogate Th17 response [37]. Similarly, the Asian or pediatric endotypes may be best reproduced with HDM-induced C57BL/6 mice to bolster Th22 inflammation [37].
In both cases, a concern lies in the task of capturing the complex features of AD in their entirety, which is further complicated by the lack of consensus regarding the features that define such criteria [29]. Thus, an additional measure to optimize the utility of existing models may involve validating therapies across a heterogenous panel of animal models, taking advantage of the common mediators that underlie cutaneous inflammation in different models. Defining an agent's pattern of anti-inflammatory activity across multiple models that encompass all relevant features of AD may overcome their individual limitations and provide a broader picture of therapeutic response.
There are several additional limitations to selecting a model based on its endotype. Many models have not been completely evaluated for their immunologic signatures, while existing drug validation studies provide only limited insight into translational utility of tested models given their incomplete profiling of inflammatory suppression. Further studies are warranted to detail the inflammatory changes in all AD mouse models, and to characterize the degree of similarity in therapeutic response between murine AD and human AD. Moreover, generalizability is bound by interspecies differences in immune systems and their relationship to relevant biological pathways [6,53]. For instance, IL-17 is mostly produced by Th17 cells in humans and by γδ T cells in mice. While the contribution of IL-17-producing γδ T cells has been assessed in human psoriasis, their role in human AD remains uncharacterized [53,54]. Similarly, although humanized skin and PBMC models represent promising strategies to recreate specific inflammatory signatures in mice, xenografts are still subject to interference caused by interactions with host physiology [17].
In conclusion, as our understanding of the specific roles of the inflammatory axes crystallizes, future studies may shift toward optimizing the efficacy of targeted therapies based on their specific effects on the immune factors that determine specific AD phenotypes. Thus, in an effort to overcome the challenge of selecting a mouse model that broadly captures the intricacy of human AD, immunophenotypic considerations should play a more central role in the selection of AD mouse models in future preclinical studies.