Skin Inflammation Induced by the Synergistic Action of IL-17A, IL-22, Oncostatin M, IL-1α, and TNF-α Recapitulates Some Features of Psoriasis Free

Keratinocytes play an important role in the regulation of skin inflammation, responding to environmental and immune cells stimuli. Over the past few years, an increasing number of reports has demonstrated that keratinocytes are direct targets for a specific set of cytokines, leading to the regulation of their biological properties, such as secretion of cytokines, chemokines, and antimicrobial peptides, their differentiation and migration capacities (1–6). The comparison of the biological profile induced in vitro by cytokines on normal human epidermal keratinocyte (NHEK) cultures with those of human biopsies from inflammatory cutaneous diseases, especially psoriasis, revealed a number of common features. Indeed, IL-1α, IL-22, TNF-α, and oncostatin M (OSM) induce a “psoriasiform” profile on NHEKs in vitro (1–4, 7). These in vitro studies identified cytokines able to induce specific expression patterns related to innate immune response, such as IL-1α, TNF-α, or IL-17A, or related to the keratinocyte differentiation program, such as IL-22 or OSM. Some of these cytokines, or related ones, such as IL-1α, IL-22, or IL-23, can induce skin inflammation in animal models (7–10). Moderate to full redundancy within each cytokine family has also been described for IL-1, IL-6, IL-10, IL-17, and TNF families (11–17). It appears that single cytokine stimulation generates a rather limited effect on keratinocytes, namely, a limited number and/or a limited modulated expression of targeted genes, reflecting only partial features with psoriasis. Recent data showing that immune cells infiltrating psoriatic skin secreted large amounts of several inflammatory cytokines, including IL-22, IL-17A, and IL-23, led to classify psoriasis as the paradigm of an inflammatory disease involving the proinflammatory Th17 subset, challenging the Th1 theory previously admitted (4, 18–21). We previously reported that supernatants from T cells isolated from psoriatic skin were able to induce a “psoriasis-like” phenotype in NHEKs (18, 21), which would result from synergistic effects among the cytokines produced by these T cells. For example combination of IL-17A and IFN-γ or IL-17A and TNF-α results in a synergistic effect on CXCL8 production (22, 23). IL-17A and IL-22 synergize in the upregulation of β-defensin 2 (BD2) and S100A9 production (24). Our goal was to identify an optimal and relevant cytokine combination able to synergize to generate in vitro an inflammatory keratinocyte model recapitulating some features of lesional psoriatic skin.

The use of skin samples for this study was approved by the Ethical Committee of the Poitiers Hospital. After informed consent, lesional skin biopsies were obtained from eight different patients with moderate-to-severe plaque psoriasis (mean age = 47 y; skin involvement 30–90% of body surface area) that did not receive any therapy for >4 wk. Normal skin biopsies were obtained from surgical samples of healthy breast skin.

NHEK were obtained as previously described, from surgical samples of healthy breast skin (1). NHEK were cultured to 80% of confluence and then starved for 24 h in Keratinocyte serum-free medium without addition of growth factors before stimulation. Cells were stimulated with or without 10 ng/ml recombinant IL-17A, OSM, TNF-α, IL-22, and IL-1α alone or in combination (R&D Systems Europe, Lille, France) during 24 h for mRNA quantification or 48–96 h for protein quantification.

Outbred OF1 mice were purchased from Charles River Laboratories (Chatillon, France). Ear intradermally injections were realized at day 0 under brief isoflurane (Forene, Abott France, Rungis, France) gas anesthesia. The 250 ng carrier-free IL-17A, OSM, TNF-α, IL-22, and IL-1α (R&D Systems Europe) or PBS or 10 μg LPS (Sigma-Aldrich, Saint Quentin Fallavier, France) were injected in a total volume of 20 μl. After 24 or 48 h, ears were collected and frozen immediately in liquid nitrogen for H&E staining, immunohistochemistry analysis, or RNA quantification.

Pieces (2 × 2 cm) of healthy breast skin were washed in PBS and intradermally injected with 10 ng each cytokines (IL-17A, OSM, TNF-α, IL-22, and IL-1α) or with PBS 0.1% BSA in a total volume of 50 μl. Each sample was placed individually in a 6-well plate and incubated up to 72 h at 37°C, 5% CO₂, in SkinEthic maintenance medium (SkinEthic Laboratories, Nice, France). Punch biopsy of 4 mm was taken after 24 h of treatment at the site of injection and immediately frozen in liquid nitrogen for RNA quantification. Culture supernatants were collected for cytokine ELISA determination.

The comparison of the effects of 36 different cytokines on the expression of 154 genes of potential interest for skin physiology was performed using home-made cDNA macroarrays analysis (Supplemental Table I). After cytokine stimulation, total RNA was extracted using TRIzol Reagent (Invitrogen Life Technologies, Cergy Pontoise, France) and conventional 33P-cDNA target synthesis and hybridization were performed (4). Genes were considered regulated if expression levels differed >3-fold relative to untreated control and 3-fold relative to mean background noise.

NHEK and skin total RNA were isolated and reverse transcribed as previously described (1). Quantitative real-time PCR (QRT-PCR) was carried using the LightCycler-FastStart DNA Master^Plus SYBR Green I kit on LightCycler 480 (Roche Diagnostics, Meylan, France). The reaction components were 1× DNA Master Mix, and 0.5 μM HPLC purified sense and antisense oligonucleotides purchased from Eurogentec (Eurogentec France, Angers, France), designed using Primer3 software. Samples were normalized to three independent control housekeeping genes (G3PDH, RPL13A, or ACTB for human samples and G3PDH, HMBS, or β₂-microglobulin for mouse samples) and reported according to the ∆∆C_T method as RNA fold increase: 2^∆∆CT= 2^{∆CT sample − ∆CT reference}. For comparison of normal skin and psoriatic skin the REST 2008 software was used (25).

Levels of BD2, BD3, and CXCL8 were determined using Human ELISA development kit (PeproTech, Neuilly sur Seine, France), and CXCL1 and CXCL5 with DuoSet reagents (R&D Systems Europe).

After 96 h of stimulation, NHEK lysis was performed as previously described (1). After separation by SDS-PAGE, proteins were transferred to nitrocellulose membranes (Amersham Pharmacia Biotech, Orsay, France) by electroblotting. S100A7 was detected with mouse anti-human S100A7 mAb (Imgenex, San Diego, CA) and sheep anti-mouse IgG peroxidase-conjugated polyclonal Ab (Amersham Biosciences, Orsay, France). Bound Ab were detected by chemiluminescence (ECL Hyperfilm and ECL Plus Reagen, Amersham Biosciences).

Chemotaxis assay was performed using 24-well transwell inserts (transparent polyethylene terephthalate membrane, 3-μm pore; Becton Dickinson Biosciences, Le Pont de Claix, France). Human neutrophils obtained from peripheral blood of healthy volunteers were labeled with 5 μM Calcein AM (Molecular Probes, Invitrogen Life Technologies). NHEK culture supernatants were incubated or not with anti-CXCL8 (10 μg /ml), anti-CXCL1/2/3, and/or anti-CXCL5 mAb (25 μg/ml) (all from R&D Systems) for 30 min at 37°C. The 400 μl NHEK supernatants were added to the lower chamber of a Transwell plate and 200 μl calcein-labeled neutrophils were added to the upper chamber. After incubation 2 h at 37°C, 5% CO₂, the number of migrating cells in the lower chamber was determined by measuring calcein fluorescence. Results are expressed as percentage of migrating neutrophils per well.

Supernatants from NHEK treated or not for 96 h with cytokines were 50-fold concentrated by centrifugation using Amicon Ultra 3000 Da (Millipore, Saint-Quentin en-Yveline, France) and dialyzed against sodium phosphate buffer (10 mM, pH 7.4). Escherichia coli (ATCC 29325) was grown to exponential phase, bacterial concentration was adjusted to 2.10⁵ bacteria/ml in 10 mM sodium phosphate buffer and mixed (ratio 2:1) with concentrated NHEK supernatants or with human recombinant BD2 (PeproTech). After incubation at 37°C for 1 h, serial dilutions of bacterial suspensions were plated onto Brain Heart Infusion agar plates and cultured for 24 h at 37°C for determination of bacterial CFU. Results are expressed as percentage of CFU in control condition with the mean and SEM of three independents experiments.

The 6-μm sections of mouse ear were fixed in 10% formalin in PBS. Ear thickness was measured after H&E coloration at three different points in the injection area. SD from the mean is shown for three separate experiments.

Sections of mouse ear were stained for granulocytes by using rat IgG2b anti-mouse Ly-6G mAb (Gr-1, Becton Dickinson Biosciences) or with isotype control (IgG2b, Caltag, Invitrogen Life Technologies) associated with a donkey anti-rat IgG Alexa Fluor 488-conjugated secondary Ab (Invitrogen Life Technologies). Confocal microscopy was carried out on a Olympus FV1000 confocal. Pictures are representative of three experiments.

Statistical analysis of significance was calculated using either Mann-Whitney U test or Kruskal Wallis one-way ANOVA by ranks with a Dunn’s test. The p values ≤ 0.05 were considered as significant, and all data are represented as mean and SEM. Comparison study used the Spearman rank correlation test.

To identify major skin inflammation inducers we screened the activity of 36 different cytokines previously described for their effect in the skin or for their involvement during regulation of the immune/inflammatory response. We compared their effects on the expression of 154 genes of potential interest for skin physiology (Supplemental Table I). Among the cytokines able to modify the expression of at least five genes (with at least a 3-fold increase or decrease), we identified IL-22, IL-24, IL-6, OSM, IL-1α, IL-1β, TNF-α, and IL-17A (Fig. 1, Supplemental Table II).

We further selected the most potent cytokine from each family, based on the number of genes regulated and on the fold increase or decrease gene expression (Supplemental Table II), namely, IL-1α, IL-17A, IL-22, OSM, and TNF-α (named M5 combination) and showed their strong synergistic activity on the production of BD2 (161900 pg/ml versus 85 pg/ml for control culture) and CXCL8 (41800 pg/ml versus 80 pg/ml) (Fig. 2A). By successive removal of each cytokine of the M5 combination and replacement by a cytokine of the same family, we identified major contributors of keratinocyte inflammation and analyzed cytokine redundancy. IL-17A and TNF-α were more critical to the activity of the M5 combination than IL-1α, OSM, or IL-22 on BD2 and CXCL8 production (Fig. 2B). Substitution by cytokines of the same family demonstrated a total redundancy between IL-1α and IL-1β, strong redundancy between TNF-α and TNF-β, limited redundancy between IL-17A and IL-17F, IL-22, and IL-24 and weak redundancy between OSM and IL-6 (Fig. 2B).

To evaluate the proinflammatory activities of the M5 cytokine combination in the skin, we setup human ex vivo and murine in vivo skin challenges. Human normal skin explants were injected with the M5 cytokine mix and cultured for 24 h before S100A7, CXCL8, and BD2 mRNA quantification. A strong increase of S100A7, CXCL8, and BD2 gene transcription was observed after injection of the M5 combination when compared with the saline-injected or noninjected skin explants (Fig. 3). BD2 quantification by ELISA in the skin explants culture supernatants confirmed the 10- to 100-fold increased in BD2 production over a 3-d culture period (Fig. 3).

To assess in vivo the effect of the proinflammatory mixture, the M5 cytokine combination was injected intradermally in mouse ears in parallel to LPS that was used as a positive control. After 48 h, redness and swelling were observed in M5-injected ears as well as a 2-fold increase in ear thickness in comparison with saline-injected ears (Fig. 4A, 4B). Interestingly, histological analysis revealed an important inflammatory cellular infiltrate in ears injected with M5, comparable to the one obtained in LPS-injected ears. Immunohistological analysis with anti-Gr1 mAb revealed abundant granulocytes in ear tissue from LPS- or M5-injected mice compared with saline-injected mice (Fig. 4A). This infiltrate was predominantly present in the dermis and associated with an increase in CXCL1, CXCL3, and to a lesser extent CXCL2 gene transcription. In addition, transcription of the antimicrobial peptide encoding genes S100A9 and BD3 was strongly induced by the M5 combination (Fig. 4B).

To further characterize the activity of the cytokine mix on innate immunity, the antimicrobial peptides and chemokines transcriptional profile was determined by QRT-PCR analysis. The transcription of 9 chemokines and 12 antimicrobial peptides encoding genes was upregulated in response to the M5 combination (Fig. 5). Among these 21 genes, expression of 19 was upregulated by IL-1α, 16 by IL-17A, 20 by TNF-α, 14 by IL-22, and 13 by OSM (Fig. 5). A strong transcriptional synergy with the M5 cytokine combination was observed for BD2, BD3, LL37, RNASE7, PI3, S100A7, S100A7A, S100A12, CXCL1, CXCL2, CXCL3, CXCL5, CXCL6, and CXCL8, whereas only an additive effect of M5 was observed for S100A8, S100A9, CCL5, CCL20, WFDC5, and WFDC12. Finally, the activity of the M5 combination on CCL27 gene transcription seems to be due to the redundant activity of IL-1α and TNF-α (Fig. 5).

We further examined the production of CXCL1, CXCL5, and CXCL8 proteins by ELISA. As shown in Fig. 6A, untreated NHEK secreted low levels of CXCL1, CXCL5, and CXCL8. IL-1α was the most effective, followed by IL-17A or TNF-α, and finally IL-22 or OSM with a discrete effect on chemokine production. However, in combination, IL-17A, OSM, TNF-α, IL-22, and IL-1α synergistically induced a massive secretion of CXCL1, CXCL5, and CXCL8 with induction factors of 190, 80, and 650, respectively, when compared with controls (Fig. 6A).

We next analyzed the chemotactic activity of supernatants from control or stimulated NHEK. Supernatants from M5-stimulated NHEK had a significantly increased chemotactic activity for neutrophils compared with supernatants from unstimulated cultures (Fig. 6A). Neutrophil chemotactic activity was significantly reduced when neutralizing the activities of CXCL1/2/3 or CXCL5, and totally blocked when inhibiting CXCL8 (Fig. 6A).

Because we showed that antimicrobial peptide gene expression was strongly induced after NHEK stimulation by the M5 combination, we further quantified BD2 and BD3 protein by ELISA in NHEK supernatants. As shown in Fig. 6B, BD2 was significantly induced by the five cytokines alone, among them IL-1α and IL-17A being the most potent inducers. A strong synergy was seen with the M5 combination with an induction factor of 24,000 when compared with unstimulated keratinocytes. A significant increase of BD3 production was seen for IL-1α, IL-22, or OSM stimulation. Again, the M5 combination resulted in a strong synergistic effect on BD3 production (Fig. 6B), in agreement with previously observed increased gene transcription.

Expression of S100A7 in NHEK was evaluated by Western blotting analysis. S100A7 was barely detected in unstimulated NHEK (Fig. 6B). Each of the five cytokines alone induced S100A7 production, IL-1α being the most potent inducer, and a stronger expression was observed when the M5 combination was used.

We next evaluated the antimicrobial activity of M5-stimulated NHEK culture supernatants against E. coli, and showed that they exhibited strong antibacterial activity compared with the unstimulated NHEK supernatants. A similar activity for recombinant BD2 was observed (Fig. 6B).

To evaluate the pathophysiological relevance of the inflammatory phenotype observed in vitro when stimulating NHEK with the M5 combination, we quantified the gene expression of several proinflammatory cytokines and their potential targets, in particular chemokines and antimicrobial peptides, in normal skin and psoriatic skin. We were able to detect overexpression of IL-23, IL-17A, IL-22, and OSM in psoriatic skin as compared with normal skin, as well as a small increased expression of IL-1β, whereas IL-1α and TNF-α expression were not different. Among the 12 antimicrobial peptides analyzed, the transcription of 10 of them was higher in psoriatic skin compared with normal skin, with >100-fold increases for BD2, S100A7A, S100A12, PI3, S100A7, S100A9, and S100A8 (Table I). Expression of CXCL8, CXCL1, CXCL6, CCL20, CXCL5, and CCL5 encoding genes was also strongly increased in psoriatic skin compared with normal skin, but CCL27 expression was lower (Table I). In summary, among the 21 genes overexpressed in our in vitro model, the expression of 18 was higher in psoriatic skin compared with normal skin, and a correlation was found between these two sets of data (Fig. 7).

To identify major skin inflammation inducers we confirmed that IL-22, IL-20, IL-6, OSM, IL-1α, IL-1β, TNF-α, and IL-17A represent potent keratinocytes modulators, and demonstrated a powerful synergy between IL-17A, IL-22, OSM, TNF-α, and IL-1α on the expression of CXCL8 and BD2. We showed a total redundancy between IL-1α and IL-1β, strong redundancy between TNF-α and TNF-β, limited redundancy between IL-17A and IL-17F, IL-22 and IL-24, and weak redundancy between OSM and IL-6. A such partial redundancy has been described on keratinocytes for IL-22 and IL-24 (11, 12), TNF-α and TNF-β (14), or IL-17A and IL-17F (13, 17), the former being always the most active. We extended the characterization of this synergic effect using QRT-PCR analysis, and observed increased expression of 12 antimicrobial peptides and 9 chemokines in M5-stimulated NHEK. IL-1α, IL-17A, and TNF-α were the most potent inducers; however, the addition of IL-22 and OSM increased the activity of the cytokine mixture. Interestingly, the presence of NFκB, STAT3, and MAPK signaling cytokines seems crucial for this synergy, that could be reinforced by C/EBP transcription factors activation and/or mRNA stabilization induced by IL-17A (26). Antimicrobial peptides expressed in this in vitro NHEK model, in particular BD2, S100A7, LL37, and RNAse7, are crucial for the skin innate defense as direct bacterial killing compounds (27–30). Because antimicrobial peptides can demonstrate synergy with each other (31), the antimicrobial peptides induced in our system should generate a broad spectrum protection against Gram-positive and Gram-negative bacteria, fungi, and viruses. Host-defense function may be completed by chemokines, which have also direct antimicrobial activity, in particular CCL20, CXCL1, CXCL2, and CXCL3, also induced in our model (32). Alternatively some antimicrobial peptides have developed the capacity to play important roles in adaptive immune responses against microbial invasion. Indeed BD2 may recruit dendritic cells and T cells to the site of microbial invasion through interaction with CCR6 (33). More recently S100A7 was identified as chemotactic for neutrophils, monocytes, and lymphocytes, by interacting with the receptor for advanced glycation end products (34). In addition, a role of S100A8 and S100A9 in the release of neutrophils from the bone marrow, and subsequent neutrophil chemotaxis and adhesion could be involved in the recruitment of this inflammatory infiltrates (35, 36). However, the strongest synergy of the M5 combination targeted, mainly, neutrophils attracting CXCL chemokines. Experiments using blocking mAbs demonstrated that the neutrophil chemotactic activity was mainly due to CXCL8 production. The M5 combination-induced CCL20 expression, predominantly due to activity of IL-1α, TNF-α, and IL-17A, may also contribute to the recruitment of CCR6-expressing Th17 cells, and generate a positive feedback loop as described previously (37).

The combined activity of IL-1α, IL-17A, IL-22, OSM, and TNF-α led to high levels of host defense proteins expression, generating an in vitro NHEK model of skin inflammation. The pathophysiological relevance of this model was assessed by comparison of its transcriptional profile with that of psoriatic skin. We confirmed increased expression of IL-17A, IL-22, IL-23, and OSM in psoriatic skin. We detected similar TNF-α mRNA levels in lesional psoriatic skin compared with normal skin, but TNF-α expression has been described to be regulated posttranscriptionally (38). A correlation was found between in vitro and in vivo transcriptional profiles, demonstrating that a part of the characteristic “signature” for psoriasis was obtained in the present in vitro model of keratinocyte inflammation (39). This model, although restricted to keratinocyte, could help to better characterize and understand the epidermic distinctive gene expression pattern in psoriasis as compared with other skin inflammatory diseases, such as atopic dermatitis (40). Our data indirectly show that Th17 or even more specialized T cells, such as T17 or T22, can generate the inflammatory environment required to strongly modify the keratinocyte biological program (41). It is interesting to note that culture supernatants of activated T cells isolated from psoriatic lesions induced the expression of comparable set of gene products associated with keratinocyte inflammation (21). A crucial role of proinflammatory cytokine expressed in psoriatic skin was illustrated by their downregulation after treatment with cyclosporine A (42). In addition, a central role for T cells in psoriasis was illustrated by therapeutic efficacy of CTLA4-Ig (43). More recently the efficiency of anti–IL-12p40 Ab provided further evidence for a role of the IL-12/23 p40 cytokines in the physiopathology of psoriasis (44, 45). Similarly, the use of TNF-blocking drugs resulted in a reversal of the epidermal hyperplasia and cutaneous inflammation characteristic of psoriatic plaques (46). This effect has been linked to the decrease of dendritic cell activation and maturation, subsequent T cell activation, and cytokines, growth factors, and chemokines production by multiple cell types, including lymphocytes, neutrophils, dendritic cells, and keratinocytes. Recently, reversal of the epidermal hyperplasia and cutaneous inflammation by Etanercept has been associated with a reduced Th17 response (47). These observations were completed by the use of animal model of psoriasis, highlighting the requirement for IL-22 and/or TNF-α in IL-23–induced psoriasis-like skin inflammation (8–10). Our data demonstrate the advantage to study in vitro the synergistic effect of cytokines to approach pathophysiological conditions, and confirm that the altered phenotype observed in lesional psoriatic skin is the result of combined activities of cytokines. Targeting more than one cytokine could be a valuable strategy to assure a complete and sustained clinical improvement. Such in vitro models of inflammatory epidermis induced by a specific set of cytokines could be extended to others skin pathologies and a useful tool to screen new drugs.

We thank Dr. Yann Hechard for contribution to developing the antibacterial assay.

Disclosures The authors have no financial conflicts of interest.