Cultures of a single cell type, either immune cells or keratinocytes, may be consider as the simplest in vitro models available for analysis of cellular phenomena involved etiological processes of epidermal inflammatory diseases. Of course, such models do not, however, take into account complex structures and interactions within tissues, but present the enormous advantage to allow precise evaluation of a specific cell response toward any given challenge.

In order to understand parameters regulating roles and abundance of lymphocytes, mast cells, eosinophils, and other immune cell types in AD-patients skin, it can be of interest to study each cell type, one after the other, to characterize their reaction to different stimuli and environmental conditions. Immune cells implicated in the pathogenesis of AD can be isolated from peripheral blood of AD patients or normal subjects. For instance, purified lymphocytes can be studied in vitro, allowing thereby to analyze gene expression of surface receptors or investigating cell responses toward specialized cytokines or any other potentially regulating ligand. For example, it was observed using this type of investigation that the frequency of lymphocytes expressing the TSLP receptor was correlated with severity of AD disease (63).

Mast cells can also be investigated in vitro after isolation and culture in serum-free methylcellulose media containing stem cell factor, IL-6, and IL-3. Initially, this type of study has been conducted in order to compare proliferation of mast cells isolated from donors with normal IgE serum levels, to those isolated from AD patients with high serum IgE levels. Both mast cell populations exhibited no highlighted difference in proliferation and similarly released histamine. Furthermore, significantly enhanced histamine release was observed for both types of mast cell donors treated with IL-4, suggesting that regulation of mast cell function rather happens through environmental stimulation by interleukins, than by genetic predisposition in the case of AD (64). Conversely, while the number of colonies made of mast cells was unchanged, eosinophils and basophils from AD donors did exhibit enhanced growth potentials (64).

Basophils or eosinophils, both recognized as actual effectors for allergic inflammation, have also been analyzed through in vitro techniques by other investigators. For instance, fibroblasts were investigated as monolayers in the presence of culture medium containing basophils or eosinophils. Such studies interestingly highlighted the requirement for direct intercellular contacts during interactions between basophils and fibroblasts, while soluble mediators were found able to mediate cross talk between eosinophils and fibroblasts. These types of interaction in presence of toll-like receptor 2 ligands (e.g., Staphylococcus aureus) could thus be considered responsible for secretion of pro-inflammatory cytokines, possibly participating thereby in inflammatory responses that are typically exacerbated during Staphylococcus aureus infections (65).

Using culture of cell monolayers, some epidermal differentiation of keratinocytes can be achieved by either culture confluence (66) or increasing the calcium medium concentration (67), or by both. Such cultures are easily performed and remain very useful for studying simple basic molecular mechanisms. Incubation of either HaCaT cells (49) with IL-4 and IL-13 (10 ng/ml each cytokine) or primary human keratinocytes (20) with IL-4 and IL-13 (50 ng/ml each), two Th2 cytokines overexpressed in AD skin, have been performed to identify their effects on such cells. Indeed, during incubation with these cytokines, epidermal cells exhibit a notably reduced gene expression for FLG. Further, Omori-Miyake and colleagues demonstrated a downregulation of expression of keratin 1 and 10, desmoglein 1 and desmocollin-1 via STAT6-dependent mechanisms transduced via IL4RA. In addition, this type of analysis allows the demonstration for instance that silencing FLG expression using shRNA interference can simultaneously induce alterations in the amounts of synthesized cornified envelop-related proteins [expression of keratin 5, 10, and 14, IVL, and transglutaminase-1 is decreased, while the one of loricrin (LOR) is upregulated]. Interestingly, FLG silencing does also result in increased release of Th2 cytokines IL-4, IL-5, and IL-13, while production of IFNγ is decreased (68).

Mainly used to answer mostly basic questions regarding the biology of particular cell types, cultures of monolayers are broadly investigated because they are easy to perform, allow simple analytical probing, and are thus efficient at providing results with potential significance. Such cultures of human keratinocytes have indeed brought methods to test new ideas, as well as data suggestive for explanation of still obscure phenomena. For instance, while trying to understand in monolayer cultures of keratinocytes how cell signaling produced by cholesterol depletion was affecting their phenotypes, transcriptomic homologies were found between such keratinocytes and keratinocytes analyzed inside AD lesions (69). In other words, a study dealing with plasma membrane lipid microdomains opened up questions about their potential relevance and/or relevance of consequences linked to their signaling properties in order to understand and investigate AD.

Still, epidermal keratinocytes isolated from AD patients can also be cultured and compared to cells of healthy donors. In such cultures of keratinocytes, the profile for chemokine production inside cells from non-lesional areas of AD patients can be compared to normal keratinocytes (70, 71). It was reported this way that keratinocytes from AD patients spontaneously express elevated levels of granulocyte macrophage colony-stimulating factor (71). Keratinocytes from AD patients can further be used to evaluate how cytokines affect their phenotype. For example, IL-4, IFN-γ, and TNF-α were reported to trigger the production of other chemokines, like IL-8, C–C motif chemokine ligand 5, RANTES, or C–X–C motif chemokine ligand 10 in keratinocytes from non-lesional AD skin (70).

Although keratinocytes cultured as immerged monolayers are simple and useful research materials, they nonetheless do not stratify and neither produce an efficient barrier. Therefore, the development of epidermal in vitro models that produce a functional barrier has become crucial to screen treatments affecting this barrier in preclinical studies.

Several different cytokines sets have been chosen to create barrier alterations inside cutaneous in vitro models. Depending on the cytokines, cell regulations are activated in keratinocytes and demonstrate that this kind of treatments can be able to induce phenotypic features that recall epidermal lesions seen in AD lesions, like widening of intercellular spaces (spongiosis), alteration of expression and localization of differentiation markers, and/or modified lipid organization.

For instance, HSE generated on de-epidermized dermis were stimulated with IL-4 and IL-13 from day 10 until day 13 of reconstruction of the epidermis, using a concentration of 30 ng/ml for each interleukin (83). IL-4 and IL-13 are two Th2-type cytokines that exhibit elevated expression in AD lesional skin (20, 23). The treatment induced spongiosis, apoptosis, and increased expression of genes that become specifically expressed in AD epidermis, like carbonic anhydrase II (CA2) and neuron-specific Nel-like protein 2 (NELL2) (10, 84). Conversely, the treatment was unable to trigger the expression of psoriasis-associated genes, like human beta defensin 2 (hBD2) and elafin. Several other studies as well have investigated the epidermal consequences produced by incubation with these two Th2 interleukins, eventually in combination with other inflammatory molecules. For example, IL-25 was added to IL-4 and IL-13 because this cytokine can act on both immune cells (85) and keratinocytes, where it can reduce expression of FLG (19), explaining thereby the systematic link observed between inflammation and barrier disruption in AD lesions (22, 86). Consequently, IL-4 and IL-13 (50 ng/ml each) were also combined with IL-25 (20 ng/ml) in a study performed on RHE model during 48 h (87). Interestingly, the epidermal consequences produced by these cytokines were enhanced when cholesterol was depleted from the plasma membrane of keratinocytes ahead of the interleukin treatment, suggesting that plasma membrane lipid microdomains disruption would render keratinocytes more sensitive to the Th2 interleukins. In such model, spongiosis and hypogranulosis were observed together with alterations in the expression of specific AD-markers [FLG, LOR, CA2, NELL2, TSLP and hyaluronic acid synthase 3 (HAS3)], as well as in barrier functions, as highlighted by transepithelial electrical resistance measurements and lucifer yellow permeation tests (87). In other studies, Th2 cytokines (IL-4 or IL-13 at 100 ng/ml each) combined with pro-inflammatory cytokines like TNF-α (20 ng/ml) or IL-1α (100 ng/ml) for 48 h have been shown to act synergistically in order to induce production of TSLP in human skin explants (88). In a RHE model, IL-4, IL-13 (30 ng/ml each), and TNF-α (3.5 ng/ml) were further combined to IL-31 (15 ng/ml), a pruritus-related cytokine, and this cocktail induced AD-like features such as decreased expression of epidermal differentiation proteins like FLG and LOR, spongiosis, increased secretion of TSLP, and alterations of barrier properties that concern lipids (89). Indeed, lipid organization was affected as the treatment induced a decrease in the level of long chain free fatty acids and ester linked ω-hydroxy ceramides. Still in another study, IL-22 was combined to TNF-α, IL-4, and IL-13 and applied together as an “AD-mix” on RHE at day 10 of epidermis reconstruction for 48 h at a concentration of 3 ng/ml (90) in order to compare this set of interleukins with another “psoriasis-mix” set of interleukins containing IL-17, a psoriasis-related cytokine, instead of Th2 cytokines IL-4 and IL-13. Using the “AD-mix,” one can observe epidermal features of AD including decreased expression of FLG, small proline rich proteins (SPRR2A) and increased expression of IL-13RA2 (one of the IL-13 receptor subunit), together with a weak increase of S100 calcium-binding protein A7 expression. Whereas using the “psoriasis-mix,” lighter decrease in the expression of FLG is observed, together with a stronger increase of S100 calcium-binding protein A7 expression, and a weaker increase of SPRR2A and IL-13RA2 are reported.

Still another way to induce in vitro AD-like features in RHE was found while combining Poly I:C (10 µg/ml), a toll-like receptor 3 ligand mimicking viral double-stranded RNA, with TNF-α (10 ng/ml), IL-4, and IL-13 (50 ng/ml each) at day 14 of epidermis reconstruction and for 48 h (91). In this case, spongiosis, alterations of differentiation markers, increase in TSLP expression and IL-8 secretion was observed in the RHE, while its transcriptomic profile was reminiscent of the one observed in AD keratinocytes.

In vivo, AD epidermis is bathed by a complex mixture of cytokines that can produce overlapping, redundant, additive, or even opposite effects. Indeed, IL-4, IL-13, IL-22, and IL-25 are all able to downregulate expression of differentiation markers such as FLG (19, 22, 24, 86). Conversely, TNF-α exhibits only weak effects on the expression of keratinocyte differentiation markers. Nonetheless, TNF-α appears necessary, in combination with Th2 cytokines, in order to induce epidermal expression of TSLP. TSLP is released in response to a combination of different molecules like Th2 cytokines (IL-4 and -13) and pro-inflammatory cytokine (TNFα) or poly I:C combined to these cytokines. However, when present, IL-17 suppresses this upregulation (92). Regarding other opposite effects, both TNF-α and IL-22 induce expression of antimicrobial peptides such as hBD2, while Th2 cytokines IL-4 and IL-13 conversely decrease the levels of beta defensin 2 (93). Still more complicated, concentrations and timings of treatment with chosen cytokines seems to likely favor different types of tissue response. For now, the precise concentrations of these interleukins in skin of AD patients and in healthy skin are unknown yet. Thus, one must admit that currently available research data on in vitro models have been evaluated in regard of dose-dependent effects produced by those cytokines on the expression levels of proteins like FLG or TSLP inside the tissue.

Also interesting to mention, because pruritus is an important symptom of AD, other factors have been studied in vitro in order to better understand basic mechanisms regulating itch. An innervated skin model was set up for this purpose since increased innervation is another feature reported about AD skin (94, 95). Porcine dorsal root ganglia neurons were seeded in a collagen gel in top of which collagen gel comprising fibroblasts was added. Keratinocytes were then seeded on top of this scaffold in order to reconstruct an epidermis for 12 days at the air–liquid interface. That model has allowed to identify that neurons release a molecule called calcitonin gene-related peptide able to induce epidermal thickening (96). Besides increased innervation, elevated numbers of mast cells releasing histamine, among other molecules, are also part of AD skin. HSE cultured for 14 days were, therefore, incubated with histamine (10 µM) for the whole duration of reconstruction or only during the three first days. Decreased expression of differentiation markers such as keratin 1 and 10, FLG, LOR, tight junction proteins, and desmosomal proteins was found, likely responsible for alterations in the epidermal barrier function (97).

To summarize, IL-4 and IL-13 seem to be the most crucial, probably unavoidable, cytokines which can induce an AD-like phenotype in epidermis, in vitro. However, variable interleukin cocktails can be added, together with other inflammatory molecules or neuropeptides, depending on research interests and on which target genes treatments or analysis are focused.

In both lesional and non-lesional skin of AD patients, FLG expression is decreased, due to loss-of-function mutations, or irrespective of any particular FLG genotype (19, 21). Organotypic skin models containing knockdown of FLG in keratinocytes have been set up. Two studies have analyzed FLG knockdown in HSE by means of siRNAs (98, 99). Kuchler et al. (87) showed that FLG silencing performed in keratinocytes used to produce HSE grown at the air–liquid interface for 14 days, disturbed the development of the stratum corneum and also induced spongiosis. An increased susceptibility to irritating exposure, as well as altered transepidermal absorption of testosterone, but not of caffeine, was also reported. Using siRNA as well but in HSE analyzed after 7 days of epidermis reconstitution, Mildner et al. (99) have described increased permeability of the epidermal barrier to the lucifer yellow dye, hypogranulosis, enhanced activation of caspase-3 after UVB irradiation, but no alteration of lipid composition or of keratinocyte differentiation in the knockdown tissue. While using an shRNA procedure to also suppress FLG expression in a RHE model grown for 11 days, Pendaries and collaborators (100) reported hypogranulosis and increased barrier permeability of the lucifer yellow dye, together with a reduced number of keratinocyte layers, reduced cornified layer thickness, and decreased levels of the NMF, potentially responsible for the increased UV sensitivity already reported by Kuchler and colleagues (98). However, unlike Mildner and co-workers (99), alterations were reported in the mRNA and protein expression of components deeply involved in the epidermal differentiation process (100). Indeed, filaggrin-2 (FLG2), LOR, caspase 14 and bleomycin hydrolase were altogether reported as reduced conversely to the increased expression of corneodesmosin (100). Knockdown of FLG using shRNA was also performed in HSE generated using N/TERT keratinocytes (101). In this model, no alteration is reported neither regarding epidermal morphology (keratin 10, LOR, and proliferation marker ki67 were analyzed), neither lipid composition nor epidermal permeability (butyl-PABA permeation studies were performed). The discrepancies reported by the different groups could probably be attributed to the fact that different targets/issues are analyzed, therefore, further analyses would be interesting regarding this domain. A summary of the different studies regarding FLG knockdown in 3D models is available in Niehues et al. (102).

Interestingly, FLG silencing by mean of siRNAs in skin equivalents, cultivated for 14 days at the air–liquid interface, has been shown to render HSE more sensitive to the Th2 cytokines IL-4 and IL-13 (15 ng/ml of each when combined), increasing their effects on the model (93). These consequences include histological changes such as epidermal thickening and alterations in values of the superficial cutaneous pH measured. Furthermore, several effects induced by the deficiency in FLG expression, such as the upregulated expression of IVL or of occludin, were hampered by the addition of IL-4 and IL-13 (93).

Filaggrin-2 knockdown in RHE cultured for 11 days was also created by the use of shRNA production (103). Silencing FLG2 in keratinocytes of RHE led them to produce a thinner epidermis exhibiting parakeratosis, a compact cornified layer, and several alterations in the keratinocyte differentiation program including reduced processing of FLG, hornerin, corneodesmosin and reduced levels of caspase 14 and bleomycin hydrolase. A less acidic pH and an increased sensitivity to UVB was also attributed to FLG2 knockdown in RHE (103).

An alternative to the preparation of knockdown epidermal 3-D models has been developed by the use of keratinocytes that carry mutation(s) in genes of interest. Such methods avoid knockdown-derived off-target effects, but depend on availability of patient biopsies. For example, FLG-null keratinocytes can be obtained from ichthyosis vulgaris patients and be used to create pathological HSE that can be compared to healthy HSE (102). In this case, no alteration in the expression of keratin 10, IVL and transglutaminase-1 differentiation markers was reported, whereas some decrease in occludin and claudin-4 expression could be observed at protein level, affecting tight junctions. Despite such a decrease, no alteration in the epidermal barrier function could be reported using lucifer yellow and biotin permeation assays. Such FLG-null HSE were further analyzed after incubation with IL-4 and IL-13 (50 ng/ml for each cytokine), a treatment which again produces alterations in the expression of several differentiation markers (102).

Finally, explants from healthy and non-lesional human AD skin, harboring or not FLG mutations, have been cultured on dermal equivalents (104). Expression profile for most proteins remains identical in culture to their counterpart established in vivo, regardless of the presence of FLG mutations. This kind of model has allowed assessing that FLG mutations neither alter expression of the kallikrein 5 protease involved in cutaneous desquamation or the expression of the lympho-epithelial Kazal type related protease inhibitor (LEKTI).

In order to understand potential cross talk between fibroblasts and keratinocytes in the context of AD, organotypic skin cultures consisting of fibroblasts derived from perilesional atopic skin and healthy keratinocytes, as well as healthy fibroblasts and atopic keratinocytes have generated for comparison with controls (105). That study demonstrated that atopic fibroblasts contribute to the particular pathological microenvironment of keratinocytes by releasing paracrine mediators which are responsible for hyperproliferation and reduced expression of differentiation markers in keratinocytes. Moreover, healthy fibroblasts have been shown able to rescue alterations in stratification and differentiation of keratinocytes as they are usually observed in AD epidermis. Interestingly, cocultures made of the two atopic cell types could not form any stratified epidermis in the model, the effects being partially dependent on expression of leukemia inhibitory factor by fibroblasts.

Many cytokines are being produced by lymphocytes. Thus, in order to better mimic the in vivo situation, as well as in order to create a model that is adequate to test drugs or other actives which could target together the epidermal and immune compartments, cocultures of keratinocytes and lymphocytes have been developed. Two different organotypic skin equivalents have been developed using HaCaT cells cocultured with lymphocytes (106). For both procedures, HSE were prepared using dermal matrix containing fibroblasts. In the first case, the epidermal tissue was removed from its dermal matrix and then cocultured in the presence of activated T cells. In the other procedure described by Engelhart and his colleagues (94), the whole HSE was transferred onto a collagen matrix containing the lymphocytes, coated with cell-free collagen mixture that acts as an adhesive for the HSE. Using this integrated organotypic skin model, spongiosis, keratinocyte apoptosis, reduced expression of E-cadherin, expression of intercellular adhesion molecule-1, upregulation of neurotrophin-4, as well as elevated levels of pro-inflammatory cytokines and chemokines were reported (94). In addition, this integrated epidermal model was used to show positive effects of anti-inflammatory drugs on the epidermal barrier through measurements of transepithelial electrical resistance.

Cross talk between T cells and keratinocytes has also been studied in a skin equivalent generated on de-epidermized dermis (107). In such model, CD4⁺ T cells were introduced between the transwell membrane and the dermal side of a fully developed skin equivalent. Th1 and Th17 polarization of the CD4⁺ T cells could be obtained and was able to induce psoriasis-like epidermal inflammation. However, induction of Th2 polarization of the CD4⁺ T cells was not successful, indicating that this model still requires refinements to adequately mimic features of AD.