Keratinocyte signaling in atopic dermatitis: Investigations in organotypic skin models toward clinical application

Atopic dermatitis (AD) is a common complex chronic inflammatory skin disease. Notwithstanding the key contribution of immune-mediated disease factors, ever since the identification of filaggrin (FLG) loss-of-function mutations as a major predisposing factor for AD, epidermal barrier defects have been one important initiating factor of the disease based on the inside-out and outside-in barrier hypothesis. Most therapeutic agents target immune cells, cytokines, or related signaling pathways. The importance of more direct skin barrier–supporting therapies has recently been tested in multiple studies aimed at reducing the risk or delaying the onset of AD by using emollient application in early infancy. Although pilot work and some subsequent studies have suggested that emollients may protect against AD, Kelleher et al (2021, DOI: 10.1111/cea.13847) performed a meta-analysis of larger cohorts and did not detect a preventive effect. Instead, emollients appear to increase skin infections and may also increase allergic sensitization to food. For future studies, an evidence-based consensus opinion regarding the type of emollient to be used (eg, petrolatum-based or defined ceramide-based) and the appropriate timing of the start of intervention are required. Investigative studies toward better understanding of epidermal signaling events that contribute to atopic inflammation may fill the current knowledge gap regarding keratinocyte-specific druggable targets per disease endotype. Epidermal keratinocytes contribute to the physical skin barrier function by controlling skin permeation and water evaporation via networks of structural proteins, including cross-linked FLG and tight junctions of claudins and occludins and the biosynthesis of a myriad of lipids required for proper stratum corneum (SC) formation. Multi-omics (single-cell) analysis of skin biopsy samples or SC tapes have elucidated genotype (FLG)-specific transcriptome changes in lesional and nonlesional skin. Olah et al (2022, DOI: 10.1016/j.jdermsci.2022.04.007) found more deregulated immune-related genes in patients with wild-type FLG and less transcriptomic deregulation and skin barrier–related differential gene expression in patients with FLG mutations. FLG mutations could therefore lower the skin's threshold for developing AD. Analyzing the first keratinocyte-specific transcriptome by laser dissection allowed for pinpointing deregulated transcription in AD keratinocytes. A complex immune signature of not only TH2 cell-driven but also strong TH17 and TH22 cell-driven marker genes, as well as a clear deregulation of specific tight junction genes, has been found in the epidermis.1Esaki H. Ewald D.A. Ungar B. Rozenblit M. Zheng X. Xu H. et al.Identification of novel immune and barrier genes in atopic dermatitis by means of laser capture microdissection.J Allergy Clin Immunol. 2015; 135: 153-163Abstract Full Text Full Text PDF PubMed Scopus (167) Google Scholar Building on the first single-cell transcriptomic analysis of AD lesional skin,2Reynolds G. Vegh P. Fletcher J. Poyner E.F.M. Stephenson E. Goh I. et al.Developmental cell programs are co-opted in inflammatory skin disease.Science. 2021; 371eaba6500Crossref PubMed Scopus (191) Google Scholar emerging bioinformatic tools and disease maps help future scrutinization of AD-related keratinocyte signaling and the key factors involved. Keratinocytes function as crucial innate immune cells in the first line of defense and regulate adaptive immune responses. Through cytokine and chemokine production, keratinocytes draw effector and regulatory immune cells into the skin and influence immune cell plasticity and polarization. In addition, keratinocytes drive the chemical and microbial barrier function with key roles for FLG: (1) FLG peptides and their breakdown products regulate skin pH; (2) FLG (and its paralogs) may possess antimicrobial activity; and (3) FLG deficiency is associated with lower abundance of commensal gram-positive anaerobic cocci (GPACs), which are postulated to depend on FLG-derived amino acids for their growth.3Zeeuwen P.L. Ederveen T.H. van der Krieken D.A. Niehues H. Boekhorst J. Kezic S. et al.Gram-positive anaerobe cocci are underrepresented in the microbiome of filaggrin-deficient human skin.J Allergy Clin Immunol. 2017; 139: 1368-1371Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar Furthermore, the characteristic alteration in keratinocyte terminal differentiation by the AD inflammatory milieu not only affects SC formation but also putatively influences skin host defense mechanisms, as these major epidermal proteins (eg, late cornified envelope proteins, FLG, hornerin) were recently coined as cationic intrinsically disordered antimicrobial peptides.4Latendorf T. Gerstel U. Wu Z. Bartels J. Becker A. Tholey A. et al.Cationic intrinsically disordered antimicrobial peptides (CIDAMPs) represent a new paradigm of innate defense with a potential for novel anti-infectives.Sci Rep. 2019; 9: 3331Crossref PubMed Scopus (38) Google Scholar To dissect cell- and disease-specific signaling events and pinpoint targets for drug development and screening, preclinical human skin models that recapitulate AD disease mechanisms are of utmost importance. Although in vivo animal models may recapitulate the skin's complexity and encompass sequential immunologic responses, organotypic in vitro epidermal or full-thickness skin models enable focused studies into disease-related keratinocyte signaling, including those on AD-specific genetics, immunologic responses, skin microbiota, and environmental cues, which will be further discussed later in this article and are illustrated in Figs 1 and 2.Fig 2In vitro modeling of AD disease factors and in-depth molecular analysis of cause-effect relationships. A, In organotypic skin models (either epidermal equivalents or full-thickness skin models), a wide range of environmental cues and microbiome-derived inflammatory factors can be applied to study keratinocyte responses, including intracellular signal transduction. With use of keratinocytes with disease-associated genetic profiles (eg, FLG mutations) and addition of proinflammatory cytokines or immune cells, specific AD endotypes can be mimicked. Interaction of this inflammatory milieu with vascular endothelial cells can further fine-tune the molecular disease phenotype. Cutaneous neuroimmune interactions that are important in AD have been modeled by reinnervation of organotypic models or ex vivo skin using rat dorsal root ganglions or induced pluripotent stem cell–derived peripheral neurons; however, incorporation of all these disease-related cellular components in 1 complex tissue-engineered model has not yet been successful, as indicated by the separation via dotted lines. B, A comprehensive omics-driven signaling analysis toolbox has been generated over the past decade, with the included tools ranging from pretranscriptional epigenomic regulation to measurements of keratinocyte-derived metabolites and end-point functional analytics. As these sophisticated technologies are becoming more affordable by the day, data-driven model optimization by molecular comparison with patient skin can enable researchers to leverage the full potential of organotypic skin models as true alternatives to patients and minimize the need for animal experimentation in experimental and translational dermatology research. The meanings of the numerals are as follows: 1 indicates nerve innervation, 2 indicates fibroblast, and 3 indicates vasculature. ATAC, Assay for transposase-accessible chromatin, ChIP, chromatin immunoprecipitation; (CID)AMP, (cationic intrinsically disordered) antimicrobial peptide; H3K27ac, acetylation of the lysine residue at N-terminal position 27 of histone protein H3; ID-seq, immunodetection by sequencing (van Buggenum et el. Nat Comm 2018, DOI: 10.1038/s41467-018-04761-0); LC/MS, liquid chromatography–mass spectrometry; RAID, RNA and immunodetection; SB, stratum basale; sc, single cell; seq, sequencing; SG, stratum granulosum; SS, stratum spinosum; TEER, transepithelial electric resistance; TEWL, transepidermal water loss.View Large Image Figure ViewerDownload Hi-res image Download (PPT) At the genomic level, FLG loss-of-function mutations pose the strongest risk for development of AD, but many other risk loci have been discovered by genome-wide association studies (as reviewed by Liang [2016, DOI: 10.1007/s12016-015-8508-5] and Brown et al [2021, DOI: 10.1016/j.jid.2020.05.100]). At the epigenetic level, AD-keratinocytes display changes in DNA methylation status, microRNA expression, and histone acetylation (reviewed in Schmidt et al [2021, DOI: 10.1111/exd.14392]). To model genetic risk factors in human epidermal equivalents, human skin equivalents, or skin organoids, various keratinocyte sources can be utilized, including (1) primary keratinocytes from patients with known FLG null mutations; (2) transient or stable gene knockdown by small interfering RNA or short hairpin RNA, respectively; (3) CRISPR-Cas9 genome editing in primary or immortalized (eg, N/TERT) keratinocyte cell lines; or (4) patient-derived induced pluripotent stem cells. Studies on the consequences of FLG deficiency have yielded mostly conflicting results, likely owing to the lack in protocol standardization between research groups and the use of cells with different genetic backgrounds. To identify cause-effect relationships on how (epi)genetic risk factors lead to abnormal keratinocyte signaling and dysfunction, multi-omics technologies are advocated with a focus on the integration of transcription regulation with proteomic and/or lipidomic analyses for functional end products. Proteomic analysis of FLG knockdown organotypic models5Elias M.S. Wright S.C. Nicholson W.V. Morrison K.D. Prescott A.R. Ten Have S. et al.Functional and proteomic analysis of a full thickness filaggrin-deficient skin organoid model.Wellcome Open Res. 2019; 4: 134PubMed Google Scholar revealed quantitative changes in networks that are consistent with transcriptome analysis of skin biopsy samples stratified by FLG genotype. Similar comparative analyses between publicly available omics data from patient cohorts and experimental organotypic models will help in improvement of model systems and functional validation of identified biomarkers. To study keratinocyte-expressed receptor activation, signaling transduction pathways and downstream effects, human epidermal equivalents, or human skin equivalents can be manipulated by the addition of relevant inflammatory molecules to the culture medium. Traditionally, the TH2 cytokines IL-4 and IL-13 have been used to model AD and supplemented with IL-25, IL-31, IL-31 plus TNF-α, or IL-22 plus TNF-α (reviewed by Das et al [2022, DOI: 10.3390/cells11101683]). In principle, the epidermal response to any disease-related cytokine can be dissected in vitro, and many signal transduction pathways have thereby been identified (eg, Janus kinase/signal transducer and activator of transcription [JAK/STAT], MAPK, PI3K/Akt, as recently reviewed by Humeau et al [2022, DOI: 10.3389/fimmu.2022.801579]). With use of this approach, epidermal AD hallmarks (including hyperproliferation, spongiosis, impaired differentiation, and SC lipid changes) have been successfully modeled. These are targeted by therapeutics that are indicated for AD (eg, dupilumab, tofacitinib, coal tar, tapinarof). More complex immunocompetent, vascular, or even innervated organotypic skin models have attempted to more closely mimic the intercellular cross talk in AD in a 3-dimensional microenvironment. However, this increasing complexity brings additional challenges for reproducibility and cell viability and function, and it precludes high-throughput screening. Although the extent to which microbiome dysbiosis is a cause or consequence of AD remains unclear, the involvement of the skin microbiome in AD pathophysiology is generally accepted. Microbiome-mediated receptor signaling in keratinocytes has been studied mostly in the context of pattern recognition receptors. More recently, commensal skin microbiota have been found to hijack a different signaling route, namely, the aryl hydrocarbon receptor (AHR) pathway. The AHR is a ligand-activated transcription factor that is known mostly for interacting with microbiome- and diet-derived metabolites in the gut. Intriguingly, gut microbiome dysbiosis may steer disease in peripheral organs and also in the skin (better known as the gut-skin axis). We recently postulated that keratinocyte proliferation, differentiation, antimicrobial peptide expression, and skin barrier function may be under the control of the skin microbiota through AHR-driven cellular signaling events6van den Bogaard E.H. Esser C. Perdew G.H. The aryl hydrocarbon receptor at the forefront of host-microbe interactions in the skin: a perspective on current knowledge gaps and directions for future research and therapeutic applications.Exp Dermatol. 2021; 30: 1477-1483Crossref PubMed Scopus (15) Google Scholar; however, this concept may further extend to the gut microbiome as well. Regarding AD, we found that commensal GPACs also target the AHR (Van der Krieken et al [2023]) and are underrepresented on FLG-deficient skin.3Zeeuwen P.L. Ederveen T.H. van der Krieken D.A. Niehues H. Boekhorst J. Kezic S. et al.Gram-positive anaerobe cocci are underrepresented in the microbiome of filaggrin-deficient human skin.J Allergy Clin Immunol. 2017; 139: 1368-1371Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar Barrier dysfunction in AD may thus arise because of aberrant gene (FLG)-microbe (GPAC) interactions. The adverse host-microbe interactions in AD could be targeted by future microbiome therapies based on prebiotics, probiotics, postbiotics, or species-specific targeting of pathogenic subpopulations within the microbiome via phages, plasmids, or transposable elements. For host-microbe interaction studies, organotypic skin models offer a biologically relevant growth substrate (the SC) that mimics natural skin factors (eg, physical and chemical barrier) defining communication between microbial metabolites and keratinocyte receptors in vivo. The choice of specific bacterial strains (eg, disease-related, pathogenic, or healthy commensal) for cocultures will fundamentally affect the experimental outcome, and no single strain can represent the diversity within a bacterial species. Furthermore, a mixed community of microbiota influences skin processes more than a single species alone,7Loomis K.H. Wu S.K. Ernlund A. Zudock K. Reno A. Blount K. et al.A mixed community of skin microbiome representatives influences cutaneous processes more than individual members.Microbiome. 2021; 9: 22Crossref PubMed Scopus (23) Google Scholar and specific culture conditions (eg, temperature, humidity, oxygen levels, pH, log growth vs stationary phase) will influence bacterial growth and metabolite production through quorum sensing mechanisms.8Williams M.R. Costa S.K. Zaramela L.S. Khalil S. Todd D.A. Winter H.L. et al.Quorum sensing between bacterial species on the skin protects against epidermal injury in atopic dermatitis.Sci Transl Med. 2019; 11eaat8329Crossref Scopus (147) Google Scholar Also, for the analysis of microbial communities in vitro, several hurdles in distinguishing viable bacteria and estimating actual bacterial number (by culture-based methods versus by capturing the complete microbiota composition by 16S rRNA gene or shotgun sequencing) arise, and to date, they have precluded the wide implementation of organotypic models for skin microbiome research. Specific genetic defects within keratinocytes cause disorders of cornification with local and systemic atopic features, illustrating that epithelial cells in skin and other barrier organs can contribute to the outcome of an allergen encounter. However, the specific mechanisms of gene-environment interaction remain to be defined. Of relevance to AD, there is an increased prevalence of atopic disease in individuals with FLG null mutations who are exposed to allergens and irritants such as pet dander and surfactants. Exposure to cats in childhood does not affect AD prevalence except in those with FLG null mutations; conversely, early exposure to dog allergens may actually provide protection against AD.9Pelucchi C. Galeone C. Bach J.F. La Vecchia C. Chatenoud L. Pet exposure and risk of atopic dermatitis at the pediatric age: a meta-analysis of birth cohort studies.J Allergy Clin Immunol. 2013; 132 (e7): 616-622Abstract Full Text Full Text PDF PubMed Scopus (86) Google Scholar Although allergens may be inhaled, the interaction with FLG expression raises the possibility of a direct effect of pet allergens on keratinocyte function, which can be modeled in vitro (unpublished data). Sodium lauryl sulphate (SLS) is a surfactant present in soap and skin care products, ranging in concentrations from 0.01% to 50%. SLS is known to drive skin irritation through (1) the removal of natural moisturizing factor; (2) changes in the microbiota composition; (3) altered keratinocyte differentiation, including downregulation of profilaggrin; and (4) an increase in keratinocyte-derived inflammatory cytokine expression. Greater barrier dysfunction occurs after SLS exposure in individuals with FLG null mutations.10Danby S.G. Brown K. Wigley A.M. Chittock J. Pyae P.K. Flohr C. et al.The effect of water hardness on surfactant deposition after washing and subsequent skin irritation in atopic dermatitis patients and healthy control subjects.J Invest Dermatol. 2018; 138: 68-77Abstract Full Text Full Text PDF PubMed Scopus (50) Google Scholar Environmental temperature and humidity affect SC hydration, having a well-recognized impact on AD prevalence and severity. The molecular mechanisms underpinning this effect are likely mediated via transcriptional regulation of FLG and other key barrier proteins within the epidermis, again highlighting the potential for gene-environment interactions in AD that may be dissected with the organotypic epidermal models described in this article. To date, neither the complexity of native skin nor the magnitude of intercellular interactions in AD can be captured in organotypic skin models. Studies are focusing instead on single epidermal and/or dermal signaling cues, circumventing the challenges in creating a favorable culture condition for many different cell types and the lack of perfusion in the overall static organotypic models. In addition, the skin tissue architecture and its plasticity are difficult to recreate in 3-dimensional models. Lack of rete ridges and hair follicles deprive these models from bearing various anatomic tissue niches important for keratinocyte biology. Current advances in bioprinting and biomaterial engineering may provide solutions for creating tailor-made and disease-specific skin tissue microenvironments. At the same time, improvements in genetic and tissue engineering methodologies and rapid innovations in omics technologies provide ample opportunities for precise modeling of disease parameters and investigation of cause-effect relationships in the controlled laboratory environment. Epidermal keratinocytes should be considered as bona fide cellular targets for primary or secondary prevention of atopic disease and for breaching the vicious cycle of chronic inflammation and alleviating major AD symptoms. We envision that in the near future, the collective efforts regarding integrative dermatology will pay off by combining longitudinal clinical trial data with high-tech organotypic human skin models, yielding important human skin screening platforms for the discovery and development of drugs to modulate keratinocyte signaling.