Macrophage Migration Inhibitory Factor Triggers Chemotaxis of CD74⁺CXCR2⁺ NKT Cells in Chemically Induced IFN-γ–Mediated Skin Inflammation

The chemical agent 12-O-tetradecanoylphorbol 13-acetate (TPA) has been used to create an experimental murine model of Th1 cytokine IFN-γ–mediated, psoriasis-like skin inflammation. The acute model is characterized by the presence of leukocyte infiltration and edema, and the chronic model is characterized by epidermal cell proliferation and dermal angiogenesis (1–3). Psoriasis is a chronic, immune-mediated skin disease that has shown increased occurrence in recent years (4, 5). The development of psoriasis is speculated to be trigged by genetic and environmental factors, such as infection, trauma, drugs, stress, and chemical agents (4–6). The hypothesis of immunopathogenesis in human psoriasis is that environmental factors stimulate skin cells, including keratinocytes, NKT cells, macrophages, and dendritic cells, to activate dendritic cells and to promote Th1 and Th17 immune responses (5, 7, 8). Proinflammatory cytokines, such as TNF-α, IFN-γ, IL-17, and IL-22, induce psoriatic skin inflammation (5, 7, 8). Psoriasis is highly recurrent; however, the factors released from epidermal keratinocytes that determine the initial progression remain undefined (4–6).

Macrophage migration inhibitory factor (MIF) is a proinflammatory cytokine that induces skin inflammation (9–11) and may be secreted from damaged cells together with the release of damage-associated molecular patterns, such as high-mobility group protein B1 (12); therefore, MIF is possibly released by damaged cells. In skin inflammation, endotoxins and UV light induce MIF⁺ cells, such as T cells, macrophages, keratinocytes, and fibroblasts, to release their MIF and increase their expression of TNF-α, IL-1β, IL-6, and growth factors, resulting in skin inflammatory responses (9, 10). However, the mechanism of MIF release, the identity of its target cells, and its physiopathological role remain unclear. Additionally, MIF has multiple receptors, including CD74, CD44, CXCR 2 (CD182), and CXCR4 (CD184) (10, 13, 14), and it displays a chemotaxis-like function and directly mediates immune cell recruitment via these MIF receptors (13, 14); furthermore, animal studies indicated that MIF may increase T cell activation and infiltration (15, 16). The targeting of MIF with a neutralizing Ab or by the pharmacological MIF antagonist (S,R)-3-(4-hydroxyphenyl)-4,5-dihydro-5-isoxazole acetic acid methyl ester (ISO-1) has been proposed as a treatment strategy for inflammatory disorders (17, 18).

We (3) and other investigators (1, 2) used TPA to induce IFN-γ-mediated, psoriasis-like skin inflammation in murine ears to mimic human psoriatic lesions. In TPA-stimulated mice, the level of IFN-γ is increased, most likely via CD3⁺ IFN-γ–producing cells, including CD4⁺ T, CD8⁺ T, and NKT cells (2, 3), through an unknown mechanism. In IFN-γR1–knockout mice, the acute ear swelling and chronic epidermal hyperproliferation and dermal angiogenesis induced by TPA are totally abolished; however, the infiltration of leukocytes, particularly CD3⁺ leukocytes, remains (3). Given the possibility that TPA treatment directly results in cellular damage to keratinocytes (19) and the proinflammatory role of MIF (9, 10), we hypothesized that MIF is released from TPA-damaged keratinocytes and promotes skin inflammation by facilitating the infiltration and activation of IFN-γ–producing cells. In this study, we investigated the expression of MIF upon TPA treatment and its role in triggering the chemotaxis of IFN-γ–producing cells in the initial step of ear skin inflammation. Considering that NKT cells may be involved in the pathogenesis of psoriasis (20–22), we also identified a novel population of CD44⁺CD74⁺CXCR2⁺CXCR4⁺ NKT cells that belongs to the previously identified subpopulation of CD56^brightCD3⁺ NKT cells (23, 24); these cells were particularly recruited by MIF stimulation, particularly through CD74- and CXCR2-mediated pathways.

Values are expressed as the mean ± SD. The groups were compared using the Student two-tailed unpaired t test or one-way ANOVA in GraphPad Prism version 5 (La Jolla, CA). A p value < 0.05 was considered statistically significant.

DMSO, TPA, ionomycin, and DAPI were purchased from Sigma-Aldrich (St. Louis, MO). MIF antagonist ISO-1 was obtained from Calbiochem (La Jolla, CA). Rabbit anti-mouse MIF, mouse anti-human MIF, rabbit polyclonal anti-CXCR2, and Ki-67 Abs were obtained from Abcam (Cambridge, MA). Alex Fluor 488–conjugated anti-mouse CD3 (clone 17A2), Gr-1 (clone RB6-8C5), CD31 (clone 390), CD54 (clone YN1/1.7.4), and PerCP-conjugated anti-mouse NK1.1 (clone PK136) were obtained from BioLegend (San Diego, CA). PE mouse anti-human CD3 (clone UCHT1), Alexa Fluor 488–conjugated mouse anti-human CD3 (clone UCHT1), PE-conjugated mouse anti-human CD4 (clone RPA-T4), FITC-conjugated mouse anti-human CD8a (clone HIT8a), PE-conjugated mouse anti-human CD14 (clone M5E2), PE-conjugated mouse anti-human CD44 (clone 515), purified NA/LE rat anti-mouse CD44 blockade (clone IM7), Alexa Fluor 647–conjugated mouse anti-human NCAM-1 (CD56; clone R19-760), purified mouse anti-human CD74 blockade (clone M-B741) and rat anti-mouse CD74 blockade (clone In-1), FITC-conjugated mouse anti-human CD74 (clone M-741), PE-conjugated mouse anti-human CD182 (CXCR2; IL-8RB; clone 6C6), PE-conjugated mouse anti-human CD184 (CXCR4; clone 12G5), allophycocyanin-conjugated mouse anti-mouse NK-1.1 (clone PK136), anti-mouse NK1.1 (clone PK136), and rat IgG isotype control Abs were from BD Pharmingen (San Diego, CA). Alexa Fluor 488– or 594-conjugated goat anti-mouse and goat anti-rabbit secondary Abs were from Invitrogen (Carlsbad, CA). The rabbit polyclonal anti-CD74 was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). PerCP-Cy5.5–conjugated anti-mouse IFN-γ (clone XMG1.2), rat anti-mouse CXCR4 blockade (clone 2B11), and mouse IgG isotype control were obtained from eBioscience (San Diego, CA). Rabbit monoclonal lactate dehydrogenase (LDH) Ab (clone EP1563Y) was from GeneTex (Irvine, CA). Monoclonal anti-CXCR2 (IL-8 RB) blockade (clone 48311.211) was from Sigma-Aldrich. Rat anti-mouse CXCR2/IL-8 RB Ab (clone 242216) was from R&D Systems (Minneapolis, MN). Recombinant human MIF and murine MIF were prepared as described (25).

Eight- to twelve-week-old progeny of wild-type C57BL/6J mice from The Jackson Laboratory (Bar Harbor, ME) were used in our experiments. They were fed standard laboratory food and water ad libitum in the Laboratory Animal Center of the National Cheng Kung University. The animals were raised and cared for according to the guidelines set by the National Science Council of Taiwan. The experimental protocols adhered to the rules of the Animal Protection Act of Taiwan and were approved by the Laboratory Animal Care and Use Committee of National Cheng Kung University (Approval No. 99013). As shown previously (3), to establish the acute model of TPA-induced acute skin inflammation, 1% DMSO was diluted in PBS as a solvent control, and 50 μg/ml TPA (Sigma-Aldrich) was dissolved in the solvent control (3 μg/ear), with or without the optimal dosage of 25, 50, or 100 μg/ml the MIF antagonist ISO-1 (Calbiochem) dissolved in the solvent control (1.5, 3, or 6 μg/ear). The solutions were dropped onto 1-cm² pieces of 100% spunlace rayon that was 0.5 mm thick. Two pieces of rayon were placed against the inner and outer surface of each ear for 1 h as one treatment. This chronic skin inflammation model was modified from the model developed by De Vry et al. and Park et al. (1–3). For the chronic model of TPA-induced cutaneous inflammation, we treated the mice ears once following the procedure of the acute model every 3 d; thus, six treatments were applied over 15 d. On day 16, the mice were administered a lethal overdose of pentobarbital (200 mg/kg i.p.; Sigma-Aldrich), and their ear tissue was harvested at the indicated times postinjection.

Human keratinocytic HaCaT cells were provided by Dr. Chia-Yu Chi (Division of Clinical Research, National Health Research Institutes, Tainan, Taiwan). The cells were routinely grown on plastic in RPMI 1640 (Invitrogen Life Technologies, Rockville, MD) containing 10% heat-inactivated FBS (Invitrogen Life Technologies), 50 U/ml penicillin, and 50 μg/ml streptomycin and were maintained at 37°C in a humidified atmosphere containing 5% CO₂ and 95% air.

Human keratinocytic HaCaT cells (provided by Dr. Chia-Yu Chi, Division of Clinical Research, National Health Research Institutes, Tainan, Taiwan) were seeded at 2 × 10⁵ cells/ml/well in a 24-well plate and incubated at 37°C overnight. The cells were challenged with 1% DMSO (solvent control) or 25 or 50 μg/ml TPA for 8 h, as previously described (19). The amount of TPA used was similar to the concentration of TPA applied in the animal model. The supernatants were collected individually after 2, 4, or 8 h of stimulation for ELISA and LDH activity determination. In addition, the cells were fixed using 4% paraformaldehyde for immunostaining. For apoptosis detection and confocal imaging, the cells were challenged with 1% DMSO or 50 μg/ml TPA for 4 h and then stained with Annexin V-FITC/propidium iodide (PI; eBioscience) or SYTO 16 Green/PI (Invitrogen).

To evaluate cell damage, LDH activity was assayed using a colorimetric assay (Cytotoxicity Detection Kit; Roche Diagnostics, Lewes, U.K.), according to the manufacturer's instructions. Aliquots of culture media were transferred to 96-well microplates. A microplate reader (SpectraMax 340PC; Molecular Devices, Sunnyvale, CA) was used to measure the absorbance at 620 nm with a reference wavelength of 450 nm, and the data were analyzed with SoftMax Pro software. Apoptosis was analyzed by staining live cells with PI and apoptotic cells with Annexin V-FITC (eBioscience), followed by flow cytometry, according to the manufacturer's instructions. The samples were acquired with a FACSCalibur instrument (BD Biosciences), with excitation at 488 and 633 nm and emission at 515–545 nm (FL-1 channel) and 565–610 nm (FL-2 channel). The samples were analyzed using CellQuest Pro 4.0.2 software, and quantification was performed using FlowJo 7.6.1 flow cytometry software (Miltenyi Biotec, Bergisch Gladbach, Germany). Small cellular debris was excluded by gating on a forward scatter plot. Time-lapse confocal images were captured continuously using a linear sequential C1Si laser scanning head, an Eclipse TE2000E inverted microscope, and EZ-C1 software (Nikon, Tokyo, Japan) after the addition of SYTO 16 green fluorescent dye (Invitrogen) and PI.

For histopathological observations, portions of ear tissue were fixed in a 10% neutral-buffered formalin solution and then dehydrated in a graded alcohol series. The fixed tissues were embedded in paraffin and sliced into 4 μm–thick sections. The tissue sections were mounted on regular glass slides, deparaffinized in xylene, rehydrated in decreasing concentrations of ethanol, and stained with H&E.

Five-micrometer biopsied tissue sections were deparaffinized, rehydrated, and incubated with 3% H₂O₂ in methanol for 15 min. The sections were then subjected to heat-induced Ag retrieval by autoclaving them for 7 min in 1 mM EDTA buffer (pH 8; J.T. Backer, Center Valley, PA). Following two washes with PBS (pH 7.2–7.4), the tissue sections for immunohistochemistry (IHC) and immunofluorescence (IF) were blocked with Ab Diluent (Dako, Carpinteria, CA) for 20 min and 1 h, respectively, at room temperature. For IHC, the sections were incubated with primary Ab in Dako Ab Diluent at 4°C overnight. For IF, the samples were washed twice with PBS and then incubated with primary Ab in PBS at 4°C overnight. The following day, the samples were washed twice with PBS and then incubated with Dako REAL Envision HRP mouse/rabbit solution or the appropriate fluorophore-labeled secondary Ab and DAPI (1 μg/ml; for nuclear counter-staining; Sigma-Aldrich) diluted in PBS at room temperature for 1 h. For IHC, the sections were washed with PBS, developed using AEC substrate (Dako), counterstained with hematoxylin (Sigma-Aldrich), mounted with glycerol-gelatin (Sigma-Aldrich), and visualized using an inverted microscope (IX71; Olympus, Tokyo, Japan). For IF, the sections were washed with PBS and mounted with Fluorescence Mounting Medium (Dako). The sections were laid flat, dried at 4°C overnight, and visualized using a fluorescent BX51 right microscope (Olympus) connected to a DP70 camera or a linear sequential C1Si laser scanning head and an Eclipse TE2000E inverted microscope (Nikon). The confocal images were captured in a single x–y scan of the cells (100× magnification) or tissue slices (40× magnification). Photo-bleaching and cross-talk in the triple or quaternary stainings, including blue, green, red, and pseudo-color purple, were excluded individually using the line-λ scanning mode or separated using a spectral detector combined with sequencing laser excitation at 647, 543, 488, and 405 nm. The confocal images were generated using EZ-C1 software (Nikon). The blocking reagents used for the IF of cells were different from those used for the biopsied tissues. A solution of 0.5% BSA (Sigma-Aldrich) or 10% human serum in PBS was used for blocking in human keratinocytic HaCaT cells or primary human blood cells.

The levels of human MIF, IL-1β, IL-6, and TNF-α were measured with DuoSet ELISA Development System Kits (R&D Systems), and the level of human IFN-γ was measured with a CytoSet ELISA Kit, according to the manufacturer’s instructions. All measurements were acquired at least in duplicate. After the reactions, the plates were washed, and 100 μl tetramethylbenzine (TAB) substrate solution (eBioscience) was added to each well and incubated for 30 min at room temperature, after which 50 μl 2 N sulfuric acid was added per well. The plates were read at 450 nm on a microplate reader (SpectraMax 340PC; Molecular Devices), and the data were analyzed using SoftMax Pro software (Molecular Devices).

The collection and analysis of the human whole blood samples followed the protocols and procedures of the institutional review board (ER-98-167) of National Cheng Kung University Hospital (Tainan, Taiwan). First, the whole blood samples were diluted in PBS containing 0.5% BSA and 2 mM EDTA with two to four volumes of buffer. The diluted samples (30 ml) were carefully layered over 15 ml HISTOPAQUE-1077 (Sigma-Aldrich) and centrifuged at 400 × g for 40 min at 20°C in a swinging bucket without braking. The upper layer was aspirated, leaving the mononuclear cell layer undisturbed at the interphase, and the cellular layer was transferred to a new tube. RPMI 1640 was added to the cells, and the samples were centrifuged at 500 × g for 40 min at 20°C. The supernatants were carefully removed, and the cellular pellets were resuspended in RPMI 1640. Following two washes with cell culture medium, the human PBMCs were transferred into 10-cm culture dishes. Most of the platelets were attached to the bottom of the culture dishes after incubation at 37°C for 2 h, and the human PBMCs were transferred into a new tube. The number of living cells was calculated using a counting chamber (Marienfeld-Superior, Lauda-Königshofen, Germany). The human PBMCs, from which most of the platelets had been removed, were placed in an appropriate amount of buffer and used for fluorophore-conjugated Ab labeling, magnetic labeling, or stimulation. All buffers were adjusted to a pH of 7.2–7.4, filtered with a 0.22 μm–pore membrane, and stored at 4°C.

Mouse whole blood was collected by cardiac puncture following the rules of the Animal Protection Act of Taiwan (Institutional Animal Care and Use Committee Approval No. 99013). The mouse blood was diluted in PBS containing 0.5% BSA and 2 mM EDTA with three volumes of buffer. The diluted mouse blood was carefully layered over 5 ml HISTOPAQUE-1083 (Sigma-Aldrich) and centrifuged at 400 × g for 40 min at 20°C in a swinging bucket without braking. The upper layer was aspirated, leaving the mononuclear cell layer undisturbed at the interphase, and the cellular layer was transferred to a new tube. RPMI 1640 was added to the cells, and the samples were centrifuged at 500 × g for 40 min at 20°C. The supernatants were carefully removed, and the cellular pellets were resuspended in RPMI 1640. Following two washes with cell culture medium, the mouse PBMCs were transferred into the new tube, and these cells were stained by fluorophore-conjugated anti-mouse CD3 and NK1.1 Abs. Finally, the percentages of murine NK cells (mNK cells) and murine NKT cells (mNKT cells) in the mouse PBMCs in the samples were analyzed by flow cytometry.

Human PBMCs were blocked with 10% human serum at 4°C for 15 min and then fluorescently stained with PE–anti-human CD3 and Alexa Fluor 647–anti-human CD56 Abs at 4°C for 30 min before sorting with a FACSAria instrument (BD Biosciences). The human NKT cells (hNKT cells; CD56⁺CD3⁺) and human NK cells (hNK cells; CD56⁺CD3⁻) were gated in dot blots for CD3⁺ (FL-2) and CD56⁺ (FL-4) to separate the two populations. Then the isolated cells were processed for stimulation and the transmigration assay. For separation of murine NK1.1⁺CD3⁺ cells, murine PBMCs were blocked with 10% murine serum at 4°C for 15 min and then fluorescently stained with Alex Fluor 488–conjugated anti-mouse CD3 (clone 17A2) and allophycocyanin-conjugated mouse anti-mouse NK-1.1 (clone PK136) Abs at 4°C for 30 min before sorting with a FACSAria instrument (BD Biosciences). The mouse NK1.1⁺CD3⁺ cells (mNKT cells) were gated in the dot blots for CD3⁺ (FL-1) and NK1.1⁺ (FL-4) to separate these populations.

Human PBMCs were resuspended in 1× BD IMag buffer, which contained PBS (pH 7.2–7.4), 0.5% BSA, 2 mM EDTA, and 0.01% sodium azide, following aseptic filtration, and adjusted to a concentration of 1–2 × 10⁷ cells/ml. CD56⁺ cells and CD14⁺ monocytes were isolated individually using the BD IMag Anti-Human CD56 and CD14 Magnetic Particles-DM kits (BD Biosciences), according to the manufacturer’s instructions.

For detection of cellular surface markers, primary human blood cells were fixed with 4% paraformaldehyde, blocked, and incubated with the indicated fluorophore-conjugated primary Abs at 4°C for 30 min. The samples were acquired with a FACSCalibur flow cytometer (BD Biosciences), with excitation at 488 and 633 nm and emission at 515–545 nm (FL-1 channel), 565–610 nm (FL-2 channel), and >675 nm (FL-4 channel). The samples were analyzed using CellQuest Pro 4.0.2 software, and quantification was performed using FlowJo 7.6.1 flow cytometry software (Miltenyi Biotec). Small cellular debris was excluded by gating on a forward scatter plot.

Isolated primary human blood cells, including PBMCs, CD56⁺CD3⁺ cells (hNKT cells), CD56⁺CD3⁻ (hNK cells), CD4⁺ T cells, and CD8⁺ T cells, and the mixed and separated culture systems of CD4⁺ T/CD14⁺ monocytes and CD8⁺ T/CD14⁺ monocytes were treated individually with DMSO (1%; solvent control), TPA (50 ng/ml), MIF (0.1 μg/ml), MIF (1 μg/ml), ionomycin (500 ng/ml), TPA plus MIF (0.1 μg/ml), TPA plus MIF (1 μg/ml), MIF (0.1 μg/ml) plus ionomycin, MIF (1 μg/ml) plus ionomycin, or TPA plus ionomycin (positive control) in each well of a 24-well plate at 37°C for 6 h. The supernatants were collected after 6 h of stimulation and processed for subsequent assays. The human PBMCs, hNKT cells, hNK cells, CD4⁺ T cells, and CD8⁺ T cells were individually seeded at 1 × 10⁶, 2.5 × 10⁴, 5 × 10⁴, 1 × 10⁵, and 1 × 10⁵ cells/ml/well, respectively, in a 24-well plate. In the separated culture system, the upper chamber was seeded with 2 × 10⁴ monocytes/200 μl/well and the lower chamber was seeded with 1 × 10⁵ T cells/800 μl/well in a 24-well plate containing Costar Transwell Permeable Supports (0.4 μm polycarbonate membrane; Costar, Corning, NY). In the mixed culture system of T cells and monocytes, the cells were seeded at the same concentrations.

Sorted human and mouse CD56⁺CD3⁺ cells (NKT cells) were seeded at 5 × 10⁴ cells/200 μl/well in the upper Transwell chamber (5 μm polycarbonate membrane; Costar) of a 24-well plate in serum-free RPMI 1640. The appropriate concentration of MIF (0, 50, 100, or 200 ng/ml in serum-free RPMI 1640) was added, and the cells were incubated at 37°C for 1 h. The cells in the upper chamber were removed with a sterile cotton swab, and the cells in the lower chamber were carefully preserved and fixed with a 10% neutralized formalin solution at 4°C. To block the MIF receptors, the NKT cells were pretreated with Abs against human CD74 (10 μg/ml; clone M-B741; BD Pharmingen), CD44 (10 μg/ml; clone 515; BD Pharmingen), CXCR2 (5 μg/ml; CD182; clone 48311.211; Sigma-Aldrich), or CXCR4 (5 μg/ml; CD184; clone 12G5; eBioscience), or with IgG (10 μg/ml; eBioscience) as a control, at 37°C for 30 min. The sorted mouse NK1.1⁺CD3⁺ cells were subjected to the same transmigration assay used for mNKT cells with additional recombinant murine MIF (200 ng/ml) and anti-mouse CD74 Ab (10 μg/ml; clone In-1; BD Pharmingen), anti-mouse CD44 Ab (10 μg/ml; clone IM7; BD Pharmingen), anti-mouse CXCR2/IL-8 RB Ab (50 μg/ml; clone 242216; R&D Systems), anti-mouse CXCR4 Ab (10 μg/ml; clone 2B11; eBioscience), or rat IgG isotype control (50 μg/ml; BD Pharmingen). The transmigrated cells in the lower chamber were stained with DAPI (1 μg/ml; Sigma-Aldrich), and images were captured with an Evolution QEi cooled CCD camera (Media Cybernetics, Rockville, MD) interfaced with a TE2000E inverted fluorescence microscope (Nikon). The cell numbers were calculated with NIS-Elements Br imaging software (Nikon).

To deplete NK1.1⁺ cells, including NK and NKT cells, each mouse received an i.p. injection of 150 μg purified NA/LE mouse anti-mouse NK1.1 Ab (clone PK136; BD Pharmingen) (26). On the third day after the injection of the NK1.1 Ab, we individually applied DMSO or TPA to the left or right ears of the mice.

To evaluate the pathological parameters, including ear thickness, number of infiltrating cells, and protein expression levels, images of two fields/ear section were quantitatively analyzed using Image-Pro Plus version 6.0 software (Media Cybernetics). The parameter analyzed in each image was calibrated and quantified as described below. For ear thickness and epidermis thickness, we directly measured the average ear thickness in each 10×- or 20×-objective low-power field. To quantify the level of mouse MIF and LDH in the 40×-objective images, we first determined the optical red, green, and blue lights (RGB) intensity range using the images from the TPA group, established the appropriate standard RGB intensity range, and calculated the integrated OD as the level of MIF or LDH. We used the area-of-interest mode to separate the signals for epidermal and dermal expression within each field. To determine the number of infiltrating cells in each tissue, we used the same protocol to establish the standard RGB intensity range for each image and each cell marker. We then used the appropriate standard RGB intensity range for each cell marker to count the number of cells that were stained as CD54⁺, Gr-1⁺, CD31⁺, or Ki-67⁺ for the 20×-objective general fluorescent images or as IFN-γ⁺, NK1.1⁺CD3⁺, NK1.1⁺CD3⁻, NK1.1⁺CD3⁺ IFN-γ⁺, NK1.1⁺CD3⁺CD74⁺, NK1.1⁺CD3⁻CD74⁺, NK1.1⁻CD3⁻CD74⁺, NK1.1⁺CD3⁺CXCR2⁺, NK1.1⁺CD3⁻CXCR2⁺, NK1.1⁺CD3⁻CXCR2⁻, or NK1.1⁻CD3⁻CXCR2⁺ for the 40×-objective confocal images. We alternatively used TissueQuest software (TissueGnostics, Vienna, Austria) to analyze the immunofluorescent staining of lesion tissues.

All images of each group were individually entered into the TissueQuest software, and the software auto-detected background threshold and DAPI+ cells, but we were still able to change the parameters of the nuclei size, discrimination by area, and discrimination by gray to fit an optimal label of nucleated cells. We individually selected the optimal parameters for other three-color channels to recognize the nuclear membrane (interior radius) and cell membrane (exterior radius) as the cellular cytoplasm. Following the setting of cellular cytoplasm and nucleus, we individually set the suitable parameter of each color channel to get the threshold of the fluorescent intensity by repeatedly checking whether the counterpart from dot to cell of images represented each positive-expressed nucleated cell. Finally, we separately gated IFN-γ-expressing CD3+NK1.1+ (mNKT) cells, CD74+ mNKT, and CXCR2+ mNKT of each group, and calculated the cell percentage and number.

To identify the factors required for chemotaxis of IFN-γ–producing cells to sites of inflammation, we studied the role of MIF with respect to its potential effects in facilitating immune cell recruitment (13, 14). We investigated whether MIF is released from epidermal keratinocytes upon TPA challenge, because TPA was proposed to cause cytotoxicity in keratinocytes (19). Our IHC staining of MIF and LDH showed that both are constitutively expressed in mouse ear keratinocytes and that TPA caused a significant (p < 0.01) and early (at 2 h posttreatment) reduction in their cytosolic expression compared with the solvent control DMSO (Fig. 1A). To confirm the induction of cytotoxicity by TPA, we used the same concentration of TPA to treat human keratinocytic HaCaT cells in vitro as we did for our in vivo test. The levels of MIF and LDH in the supernatants of TPA-treated cells showed a concentration response and were positively correlated with the extracellular release of MIF and LDH (Fig. 1B), indicating that the release of MIF was accompanied by cytotoxicity. Furthermore, immunostaining confirmed the constitutive intracellular expression of MIF, which might be released upon TPA treatment (Fig. 1C). However, the levels of other cytokines, such as IL-1β, IL-6, and TNF-α, in the supernatants did not change or were below the level of detection (Supplemental Fig. 1A). These results show that TPA induces the release of MIF from injured keratinocytes.

We next examined the type of cell injury that is caused by TPA. Staining with Annexin V, a marker of apoptosis, and PI, a marker of necrosis, followed by flow cytometry, indicated that TPA caused nonapoptotic cell death, which was characterized by Annexin V⁺PI⁺ double-positive staining (Supplemental Fig. 1B). We also performed staining with SYTO Green, which stains the nuclei of living cells, and PI, which stains the nuclei of dead cells. Time-lapse confocal microscopy revealed that the number of PI⁺ cells increased over time upon TPA stimulation (Supplemental Fig. 1C). These results demonstrate that a high dose of TPA may directly cause cytotoxicity in keratinocytes by inducing nonapoptotic cell death.

To verify that the release of MIF leads to an inflammatory response, the MIF antagonist ISO-1 was used to investigate whether inhibition of MIF reduces TPA-induced skin inflammation in the acute and chronic models. In the acute model, following the procedure to induce acute skin inflammation with TPA, the ears of the mice were treated with the MIF antagonist 1 h before TPA challenge, and samples were harvested at 8 h after TPA application (Fig. 2A). Notably, H&E staining demonstrated that the inhibition of MIF attenuated TPA-induced ear swelling in a dose-dependent manner (Fig. 2B). Furthermore, the ears treated with ISO-1 displayed an attenuation of the TPA-induced increases in ear thickness (p < 0.01; Fig. 2C), CD54 expression, and granulocyte infiltration (p < 0.01; Fig. 2D). These results demonstrate that MIF is pathogenic in various processes in TPA-induced acute skin inflammation, including edema, CD54 expression, and granulocyte infiltration.

To further clarify the role of MIF in skin inflammation, we also assessed the protective effects of MIF inhibition in a model of chronic TPA treatment (Fig. 2E). ISO-1 treatment strikingly attenuated TPA-induced edema and epidermal swelling (Fig. 2F), ear thickness (p < 0.001; Fig. 2G), and epidermal thickness (p < 0.001; Fig. 2H). To investigate the role of MIF in IFN-γ–regulated chronic TPA-induced cutaneous inflammation, immunostaining of Ki-67, a proliferation marker, and CD31, an endothelial cell marker, was performed. The results showed that inhibition of MIF significantly suppressed TPA-induced epidermal cell proliferation (p < 0.01; Fig. 2I, Supplemental Fig. 2A) and dermal angiogenesis (p < 0.01; Fig. 2J, Supplemental Fig. 2A). Recombinant murine MIF (1 μg) was microinjected s.c. into the mice ear, with or without ISO-1 (10 μg). The results revealed that administering MIF did not cause ear edema (Supplemental Fig. 3A, 3B), but it induced the recruitment of CD74⁺NK1.1⁺CD3⁺ or CXCR2⁺NK1.1⁺CD3⁺ NKT cells in mice ear; ISO-1 cotreatment reduced MIF-triggered infiltration of CD74⁺ and CXCR2⁺ NKT cells (Supplemental Fig. 3C–E). According to the results, it is speculated that the presence of TPA is needed to initiate inflammatory activation in MIF-targeting NKT cells in vivo. Basically, our in vitro study demonstrated that TPA can cause IFN-γ production in NKT cells. These results demonstrate that inhibition of MIF attenuates processes involved in chronic TPA-induced cutaneous inflammation, including edema, epidermal proliferation, and angiogenesis.

We next examined the effects of MIF inhibition on the infiltration of IFN-γ–producing cells, because MIF was shown to induce chemotaxis in dermal IFN-γ⁺ T cells (13). Ear biopsies from the acute model showed that ISO-1 pretreatment significantly (p < 0.01) decreased TPA-induced IFN-γ⁺ cell infiltration (Fig. 3A). To distinguish the types of cells that produce IFN-γ in TPA-induced acute skin inflammation, confocal image analysis of NK1.1⁺ and CD3⁺ immunostaining was used to detect dermal mNK cells (NK1.1⁺CD3⁻), mNKT cells (NK1.1⁺CD3⁺), and T cells (NK1.1⁻CD3⁺); all of these cells regularly produce IFN-γ (24). The dermal mNKT cells (NK1.1⁺CD3⁺) had the highest level of infiltration at 8 h after TPA challenge (p < 0.001), although the dermal NK (NK1.1⁺CD3⁻) cells also were significantly recruited in the ears with a significant change (p < 0.05); however, the T cell infiltration did not show a significant change (Fig. 3B, 3D, Supplemental Fig. 2B). Triple immunostaining for IFN-γ in dermal NK1.1⁺CD3⁺ cells revealed that TPA significantly (p < 0.001) induced the recruitment of dermal NK1.1⁺CD3⁺IFN-γ⁺ cells and that ISO-1 significantly (p < 0.01) decreased cell infiltration (Fig. 3C, 3D, Supplemental Fig. 2B). These data indicate that TPA induces a chemotactic effect on dermal IFN-γ⁺ mNKT cells in a MIF-regulated manner.

To investigate the potential effects of TPA and MIF on specific hNKT cell types, subpopulations of hNK cells and hNKT cells were isolated from human PBMCs by sorting for cells expressing the cell markers CD56 and CD3 (Supplemental Fig. 4A). After stimulation for 6 h and compared with the positive-control treatment (TPA plus ionomycin), the ELISA data showed that TPA, but not MIF, significantly induced IFN-γ production in PBMCs (p < 0.001; Fig. 4A) and hNKT cells (p < 0.001; Fig. 4B) but not in hNK cells (Supplemental Fig. 4B). These results demonstrate that TPA, but not MIF, is able to trigger IFN-γ production in hNKT cells.

In this study, the number of hNKT cells in the sorted NKT cell group was adjusted to equal the calculated number of hNKT cells in the tested human PBMC group. However, the level of IFN-γ production in hNKT cells was considerably lower than that in PBMCs. Thus, we next tested the effects of TPA on magnetically isolated CD4⁺ T and CD8⁺ T cells in the absence or presence of CD14⁺ monocytes using a mixed culture system. However, with or without monocytes, TPA did not induce production of IFN-γ in the CD4⁺ (Supplemental Fig. 4C) and CD8⁺ (Supplemental Fig. 4D) T cells. These results indicate that TPA only induces IFN-γ production in hNKT cells.

Given that our data showed a chemotactic effect of MIF on NKT cells in vivo and a stimulatory effect of TPA on IFN-γ production in NKT cells in vitro, it is speculated that the mechanism for TPA skin inflammation involves MIF-induced NKT recruitment, followed by TPA stimulation. However, it is still unclear which MIF receptors, including CD44, CD74, CXCR2, and CXCR4 (10, 13, 14), are expressed on CD56⁺CD3⁺ hNKT cells. Therefore, magnetically isolated CD56⁺ cells were stained individually for CD3 and CD74, CD44, CXCR2, and CXCR4. We found that 41.8% of the CD56⁺ cells isolated from the PBMC population were CD56⁺CD3⁺ (hNKT cells) (Fig. 5A). In the CD56⁺CD3⁺ hNKT cell population, only 13.4 and 14.1% expressed CD74 and CXCR2, respectively, and most of the hNKT cells expressed CD44 (99.7%) and CXCR4 (89.4%) (Fig. 5B). It is notable that, consistent with previous studies (23, 24), two subpopulations of CD56⁺ cells, including CD56^bright and CD56^dim, were detected by flow cytometry (Fig. 5C). Further analysis of these two subpopulations showed that most of the CD56^brightCD3⁺ hNKT cells (3.5/41.8 = 8.4%) expressed CD44 (100%) and CXCR4 (96.5%), as well as CD74 (96.1%), and some expressed CXCR2 (25.2%). In contrast, most of the CD56^dimCD3⁺ hNKT cells (38.3/41.8 = 91.6%) expressed CD44 (99.9%) and CXCR4 (86.1%) (Fig. 5C). Notably, MIF induced the transmigration of the CD56⁺CD3⁺ hNKT cells in a concentration-dependent manner (Fig. 5D). The blockade of CD74 and CXCR2, but not CD44 and CXCR4, with neutralizing Abs significantly retarded the MIF-induced transmigration of human and mouse NKT cells (p < 0.05; Fig. 5E). MIF caused nearly 1% of the NKT cells to transmigrate through the Transwell membrane in our assay. Importantly, neither CD74⁺CXCR2⁻CD56^brightCD3⁺ NKT cells (8.4-2.1 = 6.3%) nor CD74⁻CXCR2⁺CD56^dimCD3⁺ NKT cells (91.6 × 9.4/100-2.1 = 6.5%) were activated by MIF for transmigration. However, all of the NKT cells showed CD44 and CXCR4 expression, and only a unique population of NKT cells with a CD74⁺CXCR2⁺CD56^brightCD3⁺ phenotype (8.4/25.2 = 2.1% of NKT cells) was able to be activated by MIF for transmigration (Fig. 5F). These results identify a novel subpopulation of NKT cells within the PBMC population that is indispensable for MIF-triggered chemotaxis.

Based on the results of our in vitro study in which we tested the effects of MIF on isolated NKT cells, we next checked for the presence of dermal CD74⁺CXCR2⁺ mNKT cells in TPA-stimulated mouse ears. Immunostaining of CD74⁺NK1.1⁺CD3⁺ and CXCR2⁺NK1.1⁺CD3⁺ mNKT cells showed a significant increase in these two subpopulations in mouse ear tissues with TPA challenge (p < 0.01; Fig. 6A, Supplemental Fig. 4E, 4F). However, we also found increases in NK1.1⁺CD3⁻CD74⁻ (mNK) cells, NK1.1⁻CD3⁻CD74⁺ cells (most likely including monocytes, macrophages, and dendritic cells), NK1.1⁺CD3⁻CXCR2⁺ cells (most likely indicating mNK cells), and NK1.1⁻CD3⁻CXCR2⁺ cells (including neutrophils) in the dermis of TPA-challenged mice ears compared with the ears in the DMSO group, and our results were particularly similar to previous reports from Park et al. (2) and Bernhagen et al. (13). We next used NK1.1 depletion, in which each mouse received an i.p. injection of 150 μg of purified NA/LE mouse anti-mouse NK1.1 Ab (clone PK136; BD Pharmingen) (26), to verify the role of mNKT cells. The NK1.1⁺ cells of mouse PBMCs treated with control IgG and NK1.1-depletion Ab individually were present at 18.2 and 2.12% of the original abundance. Therefore, the depletion efficiency for NK1.1⁺ cells and mNKT cells (NK1.1⁺ and CD3⁺) in the mice treated with the NK1.1-depletion Ab was 88.4% ([18.2 − 2.1]/18.2) and 73.6% ([7.2 − 1.9]/7.2) (Supplemental Fig. 4G). Thus, NK1.1 depletion significantly repressed TPA-induced edema and epidermal swelling (Fig. 6B), ear thickness (p < 0.05; Fig. 6C), CD54 expression, and granulocyte infiltration (p < 0.05; Fig. 6D). These results confirm the importance of mNKT cells in facilitating TPA skin inflammation.

MIF is aberrantly expressed during skin inflammation and was suggested as a therapeutic target for dermatological disorders (9–11). No reports have shown pathogenic activity of MIF in psoriasis, although MIF has been identified as an aberrant inflammatory biomarker in psoriatic patients (27, 28). In this study, we used a murine model of TPA-induced skin inflammation to demonstrate that the release of MIF from injured keratinocytes not only facilitates the recruitment of IFN-γ–producing cells but also regulates TPA-induced acute edema, granulocyte infiltration, CD54 expression, chronic epidermal proliferation, and dermal angiogenesis. This study provides novel evidence suggesting that a unique population of NKT cells (CD44⁺CD74⁺CXCR2⁺CXCR4⁺CD56^brightCD3⁺) is targeted by MIF for recruitment via CD74 and CXCR2 and by TPA for IFN-γ production via an unknown mechanism. These findings are summarized briefly in Fig. 6E. Thus, the targeting of MIF may be potentially useful for the development of therapeutic strategies against IFN-γ–mediated skin inflammation.

Some unresolved questions were raised in this study and require further investigation. Chemically induced, psoriasis-like skin inflammation is limited, because it entails specific stimulation by a chemical. Although TPA induces cytotoxicity, as demonstrated previously (19), the induction of keratinocytic injury, which is important for the progression of psoriasis, can be examined using the currently known in vivo animal models for studying psoriatic pathogenesis (29). An increase in MIF was demonstrated in patients with psoriasis through stimulation by an unknown factor or genetic involvement (27, 28, 30, 31). Consistent with the finding that TPA induces cytotoxicity in vitro (19), MIF was released from the cells injured by TPA treatment in our study. We also used in vivo and in vitro experiments to confirm the cytotoxic effect of TPA treatment. Some factors released from keratinocytes may facilitate the initial progression of skin inflammation (4–6). It is speculated that, similar to the immune modulation of damage-associated molecular patterns, a proinflammatory role of MIF prompts the early activation of the targeted cells, which express MIF receptors at high levels, to produce inflammatory mediators and undergo chemotaxis during skin inflammation. Furthermore, following the cell injury–related release of MIF, TPA may act synergistically with MIF to induce the initial step of the inflammatory response, particularly with respect to IFN-γ production. Although the molecular mechanism underlying MIF release remains unknown, our findings suggest that MIF seems to be ready to release and that keratinocytic cell death results from TPA treatment.

The targeting of factors that act downstream of MIF signaling by dimethylfumarate, a classical antipsoriasis agent, has beneficial effects against MIF-induced proliferation in keratinocytes (32). The effects of MIF inhibition on skin inflammation have not been investigated in vivo using an animal model of psoriasis. The pathological changes in psoriatic inflammatory lesions include a thickening of the epidermis, parakeratosis, elongated rete ridges, and the infiltration of several cell types, including T cells, dendritic cells, neutrophils, and other immune cells (4, 5, 29). To our knowledge, this study provides the first evidence that pharmacological inhibition of MIF with the MIF antagonist ISO-1 effectively retards TPA-induced ear edema, the infiltration of multiple immune cell types (IFN-γ–producing cells and granulocytes), epidermal keratinocyte proliferation, and angiogenesis. In addition to TPA-induced skin inflammation, further suitable models of psoriasis are needed to investigate the pathogenic role of MIF and the potential therapeutic implications of targeting MIF.

The immunopathology of psoriasis is mediated by cytokines in both Th1 (TNF-α and IFN-γ) and Th17 (IL-17A, IL-17F, and IL-22) responses, resulting in keratinocyte activation and proliferation and granulocyte infiltration in skin inflammation (4, 5). Aberrant IFN-γ production and STAT1 activation are typically found in psoriatic lesions (33–36). It is speculated that the IFN-γ response is derived from Th1 cells. In TPA-induced skin inflammation, we (3) and other investigators (1, 2) demonstrated induction of IFN-γ after TPA challenge, and we further used IFN-γR–knockout mice to show the pathological effects of IFN-γ. The current study further demonstrates the requirement of MIF for the recruitment of IFN-γ–producing cells. Therefore, we hypothesize that the early steps in skin inflammation occur via a mechanism that involves MIF release and MIF stimulation. However, in the search for IFN-γ–producing cells that are present in TPA-stimulated ears, NKT cells, but not Th1 cells, were identified, and the infiltration of IFN-γ⁺ NKT cells was predominantly regulated by MIF. The mechanism underlying TPA-induced NKT cell activation and subsequent IFN-γ production requires further investigation, which should particularly focus on the activation of protein kinase C signaling. Additionally, compared with NKT cells, it is unclear why TPA did not promote IFN-γ production in NK and T cells in vitro.

The potential effects of CD74⁺CXCR2⁺ NKT cells were not completely demonstrated in this study. CXCR4 and CCR7 are lymphoid-homing chemokine receptors used for directing NKT cells to lymph nodes; however, we did not assess the expression of CCR7, whereas CXCR4 is highly expressed in MIF-targeting CD74⁺CXCR2⁺ NKT cells. It is speculated that MIF-targeting CD74⁺CXCR2⁺ NKT cells also migrate into the skin-draining lymph node of skin-inflamed mice and participate in the activation of APCs and T cells for immunopathogenesis of skin inflammation. NKT cells are potent regulators of immune responses; however, they are not a unique population and may contribute to different cellular functions. In addition to conferring protection against autoimmune and allergic disorders, NKT cells have an active role in inflammatory skin responses, particularly in psoriatic lesions (37). However, increased levels of NKT cells are usually observed in psoriatic lesions (38, 39). Rapid activation of NKT cells allows the release of IFN-γ to aggravate certain immune-mediated skin inflammation.

An interesting finding of this study, which is consistent with previous studies (23, 24), is that two subsets of NKT cells exist: CD56^brightCD3⁺ cells (8.4% of NKT cells) and CD56^dimCD3⁺ cells (91.6% of NKT cells). Based on our results, the CD56^brightCD3⁺ NKT cells are the major cells that produce IFN-γ in response to TPA. An analysis of the MIF receptors in NKT cells showed differences in expression between these two subsets of NKT cells. Although CD44 (99.7%) and CXCR4 (89.4%) were constitutively expressed in all NKT cells, CD74 was limited to CD56^brightCD3⁺ cells (96.1%), and CXCR2 was selectively expressed in NKT cells. According to previous studies (23, 24), CD56^bright NK cells produce a high level of IFN-γ. Regarding the importance of NKT cells in the immunopathogenesis of psoriasis (20–22), our study further verifies the importance of NKT cell subsets in skin inflammation.

MIF was demonstrated to be a chemotactic factor in mononuclear cell migration (13). To the best of our knowledge, this study provides the first evidence of a chemotactic effect of MIF on NKT cells in vivo and in vitro. Because MIF did not induce IFN-γ production in vitro, but MIF inhibition attenuated the number of IFN-γ–producing cells in TPA-stimulated ears, we investigated the possible regulation of MIF-mediated NKT cell recruitment. The MIF receptors CD74, CXCR2, and CXCR4 are important for mononuclear cell migration (13). Our present study identified a novel subset of IFN-γ⁺ NKT cells that has a CD44⁺CD74⁺CXCR2⁺CXCR4⁺CD56^brightCD3⁺ phenotype (2.1% of NKT cells) and is specifically targeted by MIF for transmigration. It is notable that the inhibition of either CD74 or CXCR2 effectively abolished MIF-induced NKT cell transmigration, indicating that these two receptors are involved in chemotactic signal transduction. In the TPA-treated ears, immunostaining confirmed the presence of infiltrated CD74⁺ and CXCR2⁺ NKT cells.

Johnston et al. (40) showed that both of the NKT cells commonly express CXCR4 and could migrate in a manner that was mediated, in part, by stromal cell–derived factor-1 (CXCL12)-CXCR4. In addition to CXCL12-CXCR4, signaling of monokines induced by IFN-γ (MIG/CXCL9)-CXCR3 contributes to the migration of NKT cells isolated from the spleen, liver, bone marrow, and blood. Although the chemokines and chemokine receptors in psoriatic lesions do not include CXCL12 and CXCR4 (22), it will be interesting to investigate the chemoattractive effects of CXCL12-CXCR4 and CXCL9-CXCR3, which are highly expressed in psoriatic lesions, on NKT cell migration in inflamed skin in the TPA model in future experiments.

Apte et al. (41) demonstrated that MIF inhibits NK cell–mediated cytotoxicity and granule exocytosis. They did not examine the effects of MIF on cell migration, particularly that of NKT cells. Takagi et al. (42) reported that Ab-induced arthritis in MIF-transgenic mice is more aggravating than that in wild-type mice. In their study, α-galactosylceramide may have stimulated NKT cells in spleen to trigger immunomodulation against arthritis by promoting Th2-biased immune deviation. However, in the TPA-induced model that mimics properties of psoriasis, Th1 and Th17 responses are predominant for immunopathogenesis (2, 3). According to our findings, the regulation of MIF in NKT cell migration and activation was identified in this study. This finding is inconsistent with previous reports that NKT cells are protective against autoimmune diseases but harmful in psoriasis. Therefore, the effects of MIF on NKT cells and their roles in human inflammatory disorders are diverse.

In conclusion, this study demonstrates that TPA causes the release of MIF from injured epidermal keratinocytes. In our mice, MIF induced an initial step for the chemotaxis of CD44⁺CD74⁺CXCR2⁺CXCR4⁺CD56^brightCD3⁺ NKT cells via a CD74- and CXCR2-regulated pathway, and TPA stimulation produced IFN-γ in skin inflammation. Although the signaling pathways involved in the Th1 and Th17 immune response require investigation, our results suggest that the pharmacological inhibition of MIF is a potential therapeutic strategy for IFN-γ–mediated skin inflammation via the inhibition of IFN-γ production, CD54 expression, granulocyte infiltration, edema, epidermal hyperplasia, and angiogenesis. The findings of this study may provide insight for future preclinical or clinical studies using similar models of MIF-, NKT-, and IFN-γ–mediated human dermatological disorders.

We thank the Immunobiology Core, the Research Center of Clinical Medicine (Tainan, Taiwan), and the National Cheng Kung University Hospital (Tainan, Taiwan) for providing services that included training, technical support, and assistance with experimental design and data analysis. For image quantification, we gratefully thank Lin Trading (Taipei, Taiwan) for support regarding Image-Pro Plus software and quantification methods. We are grateful to Dr. Kuen-Jer Tsai and Ya-Chun Hsiao (FACS-like Tissue Cytometry in the Center of Clinical Medicine, National Cheng Kung University Hospital, Tainan, Taiwan) for image acquisition and analysis.

The authors have no financial conflicts of interest.