Development and characterization of a human Th17‐driven ex vivo skin inflammation model

Abstract Skin models mimicking features of psoriasis‐related inflammation are needed to support the development of new drugs in dermatology. Reconstructed skin models lack tissue complexity, including a fully competent skin barrier, and presence and/or diversity of immune cells. Here, we describe InflammaSkin®, a novel human Th17‐driven ex vivo skin inflammation model. In this model, skin‐resident T cells are in situ activated by intradermal injection of anti‐CD3 and anti‐CD28 antibodies and Th17 cell polarization is sustained by culture in a chemically defined medium supplemented with IL‐1β, IL‐23 and TGF‐β for seven days. The acquired Th17 signature is demonstrated by the sustained secretion of IL‐17A, IL‐17AF, IL‐17F, IL‐22, IFN‐γ, and to some degree IL‐15 and TNF‐α observed in the activated ex vivo skin inflammation model compared with the non‐activated skin model control. Furthermore, expression of S100A7 and Keratin‐16 by keratinocytes and loss of epidermal structure integrity occur subsequently to in situ Th17cell activation, demonstrating cellular crosstalk between Th17 cells and keratinocytes. Finally, we demonstrate the use of this model to investigate the modulation of the IL‐23/IL‐17 immune axis by topically applied anti‐inflammatory compounds. Taken together, we show that by in situ activation of skin‐resident Th17 cells, the InflammaSkin® model reproduces aspects of inflammatory responses observed in psoriatic lesions and could be used as a translational tool to assess efficacy of test compounds.

TGF-β as well as IL-23 to sustain production of IL-17 and IL-22 from Th17 cells. IL-17 induces expression of other pro-inflammatory mediators in keratinocytes, such as IL-17C, IL-19 and IL-36 that, together with IL-17 and IL-22, contribute to keratinocyte activation and epidermal hyperplasia, associated with expression of Keratin-16 and S100A7. [1] In addition, increased levels of IFN-γ in psoriatic skin by activation of skin-resident Th1 cells contribute to the inflammatory activation of keratinocytes. [1] Several skin models have been published which mimic some aspects of psoriasis-related inflammation, using keratinocyte cultures, reconstructed skin models (derived from healthy or psoriatic tissue), or skin explants, which are stimulated with cytokines to induce inflammatory responses in keratinocytes. [2][3][4][5][6][7][8][9] Such models exclude investigation of cellular interplay as activated immune cells/Th17 cells are not present. In a publication by van den Bogaard et al, [10] a human reconstructed skin model with incorporation of polarized CD4 + T cells was described allowing crosstalk between keratinocytes and T cells. [10] Reconstructed skin models, although containing a stratified epidermis, lack a fully competent skin barrier and the exact cellular architecture and compartments, such as the different immune cell types, which are all present in human skin, imposing limitations for pharmacological studies of topically applied drugs or therapies targeting specific immune cell types in psoriasis.
Here, we describe a novel human Th17-driven skin inflammation model (InflammaSkin®) based on in situ activation of skin-resident Th17 cells in the NativeSkin® model, a full thickness ex vivo skin model that retains physiological skin biology, cell viability, skin barrier function and metabolism over 7 days of culture. [11] This skin model includes crosstalk between Th17 cells and keratinocytes and reproduces aspects of the inflammatory responses observed in psoriatic lesions. Furthermore, we demonstrate the use of this model to investigate modulation of the IL-23/IL-17 immune axis by topically applied anti-inflammatory compounds.

| Test reagents
0.5 mg/g betamethasone dipropionate gel and 2.5 mg/g LEO 29102 (PDE4 inhibitor) cream with corresponding placebo formulations were manufactured by LEO Pharma.

| Production and culture of the InflammaSkin® model
Anonymized human skin samples were obtained from donors that underwent abdominoplasty procedures and had given their written informed consent. Donors did not have any record of allergies or dermatological disorders and did not use corticosteroids. Full ethical approval for the study protocol was obtained from the French ethical research committee (Comité de Protection des Personnes) and authorization was given from the French ministry of Research.
All studies were conducted according to the Declaration of Helsinki protocols. Performed studies were conducted on female Caucasian donors aged between 32 and 64.
Immediately following surgery, skin samples were transported at room temperature before being processed. Subcutaneous adipose tissue was removed from the skin sample. 8-mm-diameter punch biopsies were excised and intradermally injected with an activation cocktail containing 10 ng/mL recombinant human (rh) IL-2 (202-IL, Bio-Techne), 50 µg/mL free anti-CD3 antibody and 50 µg/mL free anti-CD28 antibody (130-091-441, Miltenyi Biotec) in order to activate skin-resident T cells. Biopsies were embedded in a proprietary biological matrix in transwells (Millicell) according to the patented NativeSkin® procedure developed by Genoskin.
The epidermal surface of skin biopsies was left in contact with the air, and the dermal compartment was immersed in the matrix. The skin models were cultured up to 7 days in 12-well plates in a proprietary and chemically defined hydrocortisone-and serum-free medium in the presence of 100 µg/mL penicillin and 100 µg/mL streptomycin, and supplemented with a Th17 polarization cocktail containing 10 ng/mL rhIL-1β protein (201-LB, Bio-Techne), 50 ng/ mL rhIL-23 protein (1290-IL, Bio-Techne) and 10 ng/mL rhTGF-β protein (240-B, Bio-Techne) in a humidified atmosphere of 5% For treatment studies, 10 µL of each formulation was topically applied once daily either from the start of in situ activation (prophylactic setting; from day 0 to day 6 of culture) or after inflammation was induced (therapeutic setting; from day 4 -day 6 of culture).

| Sampling
The culture supernatants were collected and snap frozen at −80°C for cytokine analysis. Skin biopsies were split into two parts; one part was fixed in 10% buffered formalin solution (HT501128-4, Sigma-Aldrich) at room temperature and processed for paraffin wax embedding. 5-µm-thick skin cross-sections were prepared using a microtome for histological analysis. The second part was snap frozen at −80°C for gene expression analysis.
Transmitted-light images were acquired with an optical microscope (Leica DMi1) and a Leica MC170HD camera connected to a computer with Leica Application Suite (LAS®) for image capture.
Anti-Ki67 (M724029, Dako) and anti-active Caspase-3 (ab2302, Abcam) immunofluorescence stainings were performed to evaluate proliferation and apoptosis in the model, respectively. Anti-Loricrin (PRB-145P-100, Covance), anti-Involucrin (J64013, CliniSciences) and anti-(pro)-Filaggrin (AHF3 clone) immunofluorescence stainings were performed to evaluate changes in epidermal terminal differentiation. Cross-sections were stored at 60°C for 1 hour prior to antigen retrieval step, saturation and incubation with antibodies for overnight at 37°C. A specific signal was detected using secondary antibodies conjugated to Alexa Fluor 555 or Alexa 647 dye (Invitrogen Life Technologies). DAPI (D9542, Sigma-Aldrich) was used to counterstain skin sections for the immunofluorescence analyses. Images were obtained with a widefield microscope (Leica DM5000B) and a CoolSNAP EZ CCD camera (Photometrics) controlled with the MetaVue software (Molecular devices) with strictly the same parameters and duration of exposure. Images were processed with ImageJ software.

| Cytokine analysis
Levels of cytokines secreted in the culture supernatants were quan-

| Statistical analysis
Statistical analysis was performed using the GraphPad Prism 8.

| Characterization of the InflammaSkin® model
In order to develop a human ex vivo T cell-driven skin inflammation model that reproduces inflammatory responses associated with psoriasis, our approach relied on activation of skin-resident Th17 cells present in normal skin, using NativeSkin® skin explants.

F I G U R E 1
Schematic representation of InflammaSkin® model production. InflammaSkin® models were created from NativeSkin® technology.Biopsies were injected with rhIL-2, anti-human CD3 and anti-human CD28 antibodies to activate skin-resident T cells (A). Injected biopsies were then embedded in a proprietary biological matrix in transwells (B). InflammaSkin® models were cultured at a humidified atmosphere with 5% CO2 at 37°C in a chemically defined culture medium supplemented with a cocktail containing recombinant human IL-1β, TGF-β and IL-23 to sustain a Th17 phenotype (C) The InflammaSkin® model was produced in three different steps ( Figure 1). Activation of resident T cells in the skin biopsies was induced by intradermal injection of an activation cocktail containing rhIL-2 and anti-CD3 and anti-CD28 antibodies ( Figure 1A). Skin biopsies were then embedded into the patented NativeSkin® proprietary matrix and culture system that has been previously demonstrated to enable maintenance of normal skin viability and histological features for 7 days of ex vivo culture ( Figure 1B). Finally, the NativeSkin® chemically defined and serum-free culture medium was supplemented with a polarization cocktail containing pro-inflammatory cytokines IL-1β, IL-23 and TGF-β in order to sustain the Th17 phenotype ( Figure 1C).
A time course study was performed to assess the optimal window for the induction of T cell-mediated inflammation. Therefore, T cell-activated models were cultured in the absence or presence of the polarization cocktail for either 3, 5 or 7 days. Expression of the Th17 cytokines IL-17A and IL-22 was induced at day 3 with highest expression at day 7 ( Figure S1A). However, expression of IFN-γ was also induced at day 3 but peaked at day 5 ( Figure  In order to assess epidermal responses consecutive to the activation of Th17 cells, skin structure integrity and expression of epidermal markers were investigated by immunofluorescent staining. Representative results of one donor from three independent experiments with a total of three donors are shown in Figure 3 and Expression of (pro)-Filaggrin did not seem to be dysregulated neither in the InflammaSkin® model nor in the negative controls ( Figure S3, panel f, g and h) compared with the NativeSkin® model ( Figure S3, panel e).

| Response of the InflammaSkin® model to topically applied anti-inflammatory drugs
We next investigated the ability of the InflammaSkin® model to respond to topically applied anti-inflammatory drugs of therapeutic relevance for psoriasis. Two treatments were used in this study: a gel containing 0.5 mg/g betamethasone dipropionate (BDP) and a cream containing 2.5 mg/g LEO 29102, a PDE4 inhibitor which showed efficacy in a four week psoriasis plaque study (NCT00875277). Corresponding placebo gel and cream were used as controls for betamethasone gel and PDE4 inhibitor cream treatments, respectively. Untreated InflammaSkin® model was   Figure 4B, panels d to f, and j to l), except for S100A7 expression that seemed to be decreased after treatment with the placebo gel ( Figure 4B, E). Interestingly, treatment with the BDP gel resulted in normal epidermal structure and viability apart from a mild spongiosis observed on H&E staining ( Figure 4B, panels a and g), as well as a strong decrease in S100A7 sion was inhibited by the PDE4 inhibitor, but not by BDP ( Figure S4).
In this setting, the treatments were not able to restore normal skin structure and viability nor inhibit S100A7 and Keratin-16 protein expression by immunofluorescent staining likely due to the short treatment time. However, FLG gene expression was partially restored by treatment with the BDP gel but not by the PDE4 inhibitor cream ( Figures S5 and S6).

| D ISCUSS I ON
There are several human in vitro models addressing different aspects of psoriasis-related inflammation. However, to the best of our knowledge, most of these models lack tissue complexity, complete skin barrier function and/or diversity of skin-resident immune cells, since they are based on either keratinocyte cultures, reconstructed skin models or skin explants stimulated with cytokines to induce inflammatory responses in keratinocytes without activation of skin immune cells. [2][3][4][5][6][7][8][9] Therefore, there is a need for translational models that capture the interplay between activated resident T cells and keratinocytes to support pharmacological testing of novel anti-inflammatory drugs. Thus, the aim of our work was to develop a Th17-driven skin inflamma- to exercise their cellular function. [12] In particular, Th1 differentiation requires IL-12 and IFN-γ to sustain IFN-γ production by Th1 cells, while Th17 differentiation requires the concomitant stimulation by IL-1β, IL-6, TGF-β and IL-23 to sustain production of IL-17 and IL-22 from Th17 cells. [12] Indeed, the presence of IL-23 or IL-6 alone is insufficient to induce a sustained IL-17 expression and stabilization of a pathogenic Th17 phenotype. [13,14] Consequently, in our InflammaSkin® model, the in situ activation of skin-resident Th17 cells is achieved by intradermal injection of IL-2, anti-CD3 and anti-CD28 antibodies and sustained by addition of IL-1β, IL-23 and TGF-β to the culture medium. Using this strategy, various T-cell subsets (both CD4 + and CD8 + T cells and γδ T cells) will be activated into IL-17-secreting T cells. Furthermore, although IL-2 is an autocrine factor expressed by T cells upon activation, [15] injection of IL-2 was added to ensure an immediate availability of IL-2 to activated T cells, since lack of adequate amounts of IL-2 can induce complete arrest of T-cell activity. [16] As described above, IL-6 is an essential factor for Th17 differentiation. [14] IL-6 is not added in our Th17 polarization cocktail since IL-6 is endogenously induced in the skin due to tissue trauma caused by the process of skin biopsy punching (data not shown).
In Other skin models using in situ T-cell activation and Th17 polarization have been recently published and share some similarities to the InflammaSkin® model. [17,18] The model described by Smith SH et al 17 employs skin samples defatted and dermatomed to 750 µm, which could imply that some cellular architecture and dermal immune cells may not been fully represented. Importantly, the Th17 polarization cocktail used in this model does not contain IL-23, which is reported to be essential for a sustained Th17 differentiation. [14] Moreover, the use of animal-derived substances, such as 64% bovine collagen solution and 2% fetal bovine serum added to the tissue culture wells and to the culture medium, respectively, creates potential variability in the model described by Smith SH et al [17] compared with the model described by Garret SM et al [18] and to the InflammaSkin® model, which are both developed in 100% serumand animal-free settings. The model described by Garret SM et al [18] is based on an ex vivo culture of defatted but not dermatomed skin from healthy individuals, stimulated with different differentiation cocktails, but all containing anti-CD3 and anti-CD28 antibodies, IL-1β and IL-23 and thus has more similarity to the InflammaSkin® model. However, in their model, skin biopsies are cultured in an air-liquid interphase where the dermis is submerged directly in the culture medium, creating an artificial wet setting for the dermal compartment that may induce spongiosis due to continuous absorption of liquid. [18] Consequently, the feasibility to culture skin tissue in this condition is limited to a short period (48 hours), as described by Garret SM et al [18] Furthermore, culture conditions used in their setting (3-5 biopsies of 3 mm diameter per well in a semi-submerged condition) does not allow topical application of formulations, limiting F I G U R E 4 Pharmacological response of the InflammaSkin® model to prophylactic treatment with BDP gel and PDE4 inhibitor cream. InflammaSkin® models were cultured for 7 d either untreated or topically treated with placebo gel, Betamethasone dipropionate gel (BDP gel), placebo cream or PDE4 inhibitor cream (PDE4 inh. cream), applied every day from day 0 to day 6. At Day 7, skin models were fixed in 10% buffered formalin solution and processed for paraffin wax embedding prior specific histological staining and culture media were assessed for the presence of Th17/Th1-related inflammatory cytokines. A, Levels of cytokines released in the culture media were measured using specific MSD for IL-17A (orange graph), IFN-γ (green graph) and TNF-α (blue graph) and ELISA for IL-22 (red graph). The results expressed as percentage difference compared with untreated control are mean values ± SEM from three independent experiments (n = 3). One-way ANOVA statistical test was used to compare treatment effect by BDP and PDE4 inhibitor versus matching vehicle control. ***P < .001; **P < .01; *P < .05. B, Haematoxylin and eosin staining (panels a, d, g, j and m) was performed on 5-µm-thick skin cross-sections.
All images are representative of the whole sample. Scale bar is 100 µm. Anti-S100A7 (panels b, e, h, k and n) and anti-K16 (panels c, f, i, l and o) immunofluorescence staining were performed on 5-µm-thick skin cross-sections. Red and cyan signals indicate positive signal and DAPI-counterstained nuclei, respectively. Dotted white lines represent the epidermal-dermal junction. Epidermal and dermal compartments are localized above and below the dotted line, respectively. All images are representative of the whole sample. Scale bars are 100 µm. C, FLG gene expression was assessed by real-time quantitative PCR. The results are expressed as fold change relative to the NativeSkin® control ± upper and lower limits (n = 3). Two-way ANOVA statistical test was used to compare treatment effect by BDP and PDE4 inhibitor versus matching vehicle control. **** P < .0001 the use of this model to test therapeutic strategies. On the contrary, the InflammaSkin® model is based on the NativeSkin® explant technology, by which the skin biopsies are embedded in a proprietary biological matrix inside trans-well inserts that are placed in wells containing proprietary culture medium. In this system, the epidermal compartment of skin biopsies is left in contact with the air, while the dermal compartment is embedded in a moist, but not wet, environment with constant nutrients supplied by the adjacent culture media and thus resembling more closely a physiological cutaneous environment enabling maintenance of skin viability and integrity. [11] Due to the NativeSkin® explant technology, the skin explant models, including InflammaSkin®, can be cultured for at least 7 days.
Moreover, the presence of a silicon ring at the surface of the skin biopsy enables repeated topical applications of compounds in a reproducible manner by delimiting a defined treatment area and preventing any lateral diffusion.
We further characterized the InflammaSkin® model for its capacity to establish an interplay between activated Th17 cells and keratinocytes, as exemplified by the increased expression of the epidermal markers, Keratin-16 and S100A7, known to be highly upregulated in psoriasis lesions. [19] Importantly, we provide evidence that the psoria-

sis-like inflammatory interplay between Th17 cells and keratinocytes
can be inhibited by topical treatment with two classes of drugs of therapeutic relevance for psoriasis, a corticosteroid and a PDE4 inhibitor.
Treatment effect was observed for both prophylactic and therapeutic treatment regimes; however, less effect was seen in the therapeutic setting. Importantly, pharmacological studies employing systemic treatments delivered either in the culture media or by subcutaneous injection can also be performed, the latter using a variant of the InflammaSkin® model containing a defined thickness of adipose tissue to support reproducible injection of biologics without any leakage. [20] In conclusion, here we describe the InflammaSkin® model, a novel fully human Th17-driven skin inflammation model, based on in situ activation of skin-resident Th17 cells that promote psoriasis-like inflammatory responses in human skin. We provide evidence that (a) the InflammaSkin® model reliably reproduces the cellular interplay between Th17 cells and keratinocytes and features of skin inflammation as observed in psoriasis lesions and (b) the InflammaSkin® model can be used as a preclinical model system to assess efficacy of test compounds with different drug modalities and administration.

ACK N OWLED G EM ENTS
The authors acknowledge Hanne Rosendal and Trine Gejsing for technical assistance.

CO N FLI C T O F I NTE R E S T
The