Adult male (8-10 weeks) Par2^-/- and wild type C57BL/6J mice were used in this study (Par2^-/- mice were purchased from Jackson Laboratory). The mice were housed at 23°C with a 12:12 light/dark circle and 50 ± 10% humidity under specific pathogen-free conditions with food and water available ad libitum. All experimental procedures were approval by the Animal Studies Committee at Peking University First Hospital.

To induce the HDM-allergic AD mice model, calcipotriol (MC903; 6 nmol/60 μL; Tocris Bioscience, UK) dissolved in ethanol was topically applied to a shaved area on mouse neck (2.5 cm × 2.5 cm) daily for seven days. Then, 100 mg of HDM ointment (Biostir AD, Japan) was topically applied to the same area, followed by MC903 application three times weekly. After six HDM treatments, the mice were euthanized, and back skin specimens from the treated areas were collected for analysis.

Scratching behavior was observed for 60 minutes at 5-minute intervals, as described previously (13). The AD severity index (clinical score) assessed the intensity of skin symptoms, redness, bleeding, eruptions, and scaling, scored as follows: 0, none; 1, mild; 2, moderate; 3, severe. This index was evaluated weekly throughout the experiment.

Freshly collected skin tissues from mice were placed in MACS^® Tissue Storage Solution (Miltenyi Biotec, Germany). Tissue samples were digested using the Tissue Digestion Kit (Miltenyi Biotec) according to the instructions. Briefly, tissues were cut up and placed in a mixed digestive enzyme solution and incubated at 37°C for 1.5 hours. Single cell suspensions were obtained after filtration using SmartStrainer filters (Miltenyi Biotec). ScRNA-seq was conducted using the 10× Genomics Chromium platform. Raw expression data were processed with the Seurat (v3.1.5) toolkit in R (v4.0.0) for quality control and downstream analysis. Unsupervised clustering was performed, followed by visualization using UMAP. The Seurat ‘findmarker’ function identified differentially expressed genes for each cluster. Potential interactions between cell types were identified using the CellChat function.

Primary keratinocyte cultures were prepared as previously described (13). Briefly, the back skin of newborn mouse pups (P0-P3) was digested with 0.25% Dispase II (Roche, Switzerland) at 4°C overnight. The epidermis was then separated to obtain a single-cell suspension. Keratinocytes were cultured in CnT-07 medium (Advanced Cell Systems, Switzerland) with 1% Penicillin-Streptomycin (Gibco, USA) at 37°C and 5% CO₂.

Mice back skin tissues or primary keratinocytes were incubated with primary antibodies: anti-MIF antibody (1:100, Cell Signaling Technology), anti-P115 antibody (1:300, Proteintech), or anti-KIF13B antibody (1:100, Invitrogen) at 4°C overnight. Slides were then incubated with secondary antibodies: Alexa Fluor488 goat anti-rabbit IgG (1:200, Invitrogen) or Alexa Fluor594 goat anti-mouse IgG (1:200, Invitrogen) for 2 hours at room temperature. Sections were counterstained with DAPI and visualized using a confocal microscope (Leica).

Total RNA was extracted from mouse back skin tissues using TRIzol reagent (Invitrogen) according to standard protocol followed by reverse transcription to generate cDNA by TransScript IV One-Step gDNA Removal and cDNA Synthesis SuperMix (TransGen Biotech). Real-time PCR was performed by using Power SYBR Green PCR Master Mix (Applied Biosystems, UK) according to the manufacturer’s instructions. The primers used are listed in Supplementary Table S1.

The cultured medium of mouse primary keratinocytes was collected after stimulated with Der f1 (10 μg/mL) or SLIGRL (100 μM) over a time course. MIF in the medium was measured using a mouse MIF ELISA kit (Elabscience) following the manufacturer’s instruction. This ELISA kit has the sensitivity of 9.38pg/mL and the detection range of 15.63-1000 pg/mL. The cultured media were diluted to 1:10 before the measurement.

Co-immunoprecipitation was conducted using a magnetic IP/co-IP kit (Invitrogen) according to the standard protocol. Protein A-G-magnetic beads were incubated with anti-KIF13B antibody (3 μg, Invitrogen) or anti-IgG antibody (3 μg, Invitrogen) at 4°C overnight. After adding cell lysates (1 mg of protein), each reaction was incubated at room temperature for 2 hours. The beads were washed with IP elution buffer. Cell lysates were separated on a 4-12% Bis-Tris NuPage gel (Invitrogen) and transferred onto a nitrocellulose filter membrane. The interaction was visualized using anti-P115 antibody (1:1000, Proteintech).

Co-staining of CD86 (1:200, Cell Signaling Technology) and CD163 (1:300, Abcam) was performed using a three-color mIHC fluorescence kit (Hunan Aifang Biotechnology) based on tyramide signal amplification technology according to the manufacturer’s introduction.

Statistical comparisons were performed using GraphPad Prism (version 10.0) with Student’s t test or one-way ANOVA as indicated (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001). Results are presented as mean ± SEM.

HDM derived allergens are important causative factors in the development of AD (1). To establish a HDM-allergic AD model in C57BL/6J mice, we first applied MC903 on mice shaved neck to induce a skin barrier dysfunction and to make it easier for external allergens entering into skin, then HDM ointment containing Dermatophagoides farinae1 (Der f1) was topically applied to the same area ( Figure 1A ). The HDM group showed severe redness, swelling and eruption on the skin area compared to the control group ( Figures 1B, D ). Skin pathology showed significant hyperkeratosis and inflammatory cells infiltration in dermis of HDM group ( Figure 1C ). In addition, mice scratching bouts significantly increased in HDM-allergic AD mice ( Figure 1E ). We also set a control group applied with MC903 only, which showed slight pathological changes and pruritus sensation compared to MC903+HDM mice ( Supplementary Figures S1A–C ). Therefore, we have established a HDM-allergic AD mouse model with typical AD characteristics.

ScRNA-seq analysis was conducted for the skins from control mice (n=2), HDM-WT mice (n=2) and HDM-Par2^-/- mice (n=2), and obtained a total of 68,939 single-cell transcriptome data of high-quality. UMAP for dimension reduction to visualize clustering was used, and a total of 14 cell types could be defined in skin samples: fibroblast, keratinocyte, T cell, antigen-presenting cell (APC), pericyte, endothelial cells (EC), Schwann cell, lymphatic endothelial cell (LEC), neutrophil, mast cell, melanocyte, NKT cell, sweat gland cell and basophil ( Figure 1F ). Then we employed CellChat function of Seurat to investigate the complicate cellular communication network; the results showed that the communication between the ligand MIF from keratinocytes and the receptors of Cd74 and Cd44 on APCs was the strongest ( Figures 1G, H ). In other words, keratinocytes secrete MIF after they contact with the external antigens of HDM, and the MIF as a ligand binds to the receptors of Cd74 and Cd44 on the APCs nearby the keratinocytes to initiate the immunologic responses of AD.

To further exclude the role of the MC903 in HDM-allergic AD model, we also employed oxazolone, a hapten to induce contact dermatitis, combined with HDM to establish a HDM-allergic AD model ( Supplementary Figure S2A ). Oxazolone+HDM mice displayed severer AD-like symptoms, hyperkeratosis and lymphocyte infiltration, and higher MIF mRNA and stronger MIF immunofluorescence signals in skin, similar to MC903+HDM mice; while oxazolone without HDM mice showed slight AD-like presentations ( Supplementary Figures S2B–G ), indicating that Der f1 rather than MC903 is the main allergen to produce the AD model.

The protease activity of HDM may activate PAR2 on keratinocytes. We therefore investigated the effects of PAR2 on the HDM-allergic AD model. The HDM-allergic AD model was also established in Par2^-/- mice, and found that HDM-Par2^-/- mice displayed milder erythema, swelling and eruption, less hyperkeratosis and lymphocyte infiltration, and a decrease number of scratch bouts compared to HDM-WT mice ( Figures 1B–E ). Immunofluorescence and RT-PCR confirmed that MIF and its mRNA in skin were higher in HDM-WT mice and much lower in HDM-Par2^-/- mice ( Figures 2A–C ). Primary keratinocytes from WT and Par2^-/- mice were cultured and stimulated with Der f1, and SLIGRL, a PAR2 agonist. Der f1 and SLIGRL stimulation caused higher expression of MIF in mouse keratinocytes by immunofluorescence ( Figures 2D–F ). Moreover, Der f1 and SLIGRL stimulation induced more MIF released from WT keratinocytes than from Par2^-/- keratinocytes by ELISA ( Figures 2G, H ).

MIF has no N-terminal signal peptide for trafficking in endoplasmic reticulum (14). MIF is therefore secreted in a non-classical mechanism that does not depend on signal peptide cleavage (14). Golgi vesicle transporter protein P115 can serve as a cofactor to help MIF secretion (15). Immunofluorescence of primary keratinocytes and skin sections found that MIF and P115 expression increased and co-localized in the samples from HDM-WT mice, and decreased in the samples from HDM-Par2^-/- mice ( Figure 2D , Supplementary Figure S3 ).

To determine the specific subsets of keratinocytes for PAR2-related MIF release process, the scRNA-seq data from control, HDM-WT and HDM-Par2^-/- groups were further analyzed. Keratinocytes can be divided into 4 distinct clusters: undifferentiated keratinocyte (KC), differentiated KC, proliferating KC and inner root sheath cell ( Figure 3A , Supplementary Table S2 ). Basal_1, a subtype of undifferentiated KC, was characterized by the expression of TSLP and IGFPB3, and was important in inflammation in AD development ( Figure 3B ). We also calculated differentially expressed genes (DEGs) from scRNA-seq data to compare basal_1 subtype in HDM-WT group against that in HDM-Par2^-/- group and revealed the higher expression of Cd74, TSLP and IGFBP3 in HDM-WT group, indicating the participation of basal keratinocytes in the inflammation and immunomodulation of AD ( Figure 3C ).

Notably, DEGs also showed that KIF13B, a kinesin protein, was highly expressed in basal_1 subtype of HDM-WT group ( Figures 3C, D ). Using immunofluorescence assay, we confirmed the higher expression of KIF13B in keratinocytes in HDM-WT group than those in HDM-Par2^-/- group ( Figures 3E, F ). We also proved that KIF13B, P115 and MIF were co-localized and highly expressed in HDM-allergic WT model by mIHC assay ( Supplementary Figure S4 ). Protein docking assay predicted the existence of reciprocal interaction regions in KIF13B and P115 ( Figure 3G ). P115 and KIF13B were co-immunoprecipitated in SLIGRL treated primary keratinocytes of WT mouse, but the co-immunoprecipitation was weaker in those of Par2^-/- mouse ( Figures 3H, I ).

KIF13B has a motor domain anchored on the microtubules and a tail region binding to the cargo to mediate transport of cargos on microtubules (16). We also found that acetylation of α-tubulin increased after stimulation by SLIGRL in WT keratinocytes but not in Par2^-/- keratinocytes by immunofluorescence ( Figures 3J, K ) and western blotting ( Figures 3L–N ). In addition, acetylated α-tubulin protein increased in a time-dependent manner after SLIGRL stimulation in WT but not in Par2^-/- keratinocytes ( Figures 3L–N ). These findings indicate PAR2 regulates the expression of KIF13B and the acetylated α-tubulin in keratinocytes to affect the intracellular trafficking of MIF.

The key role of MIF in the pathogenesis of AD was verified by the stronger communication between ligand MIF from keratinocytes and receptors Cd74 and Cd44 on APCs from the results of scRNA-seq, the higher immunofluorescence signals of MIF in HDM-WT skin, and the increased MIF secretion from keratinocytes after Der f1 stimulation. MIF antagonist ISO-1 was then used to observe its effects on HDM-allergic AD model. We first tested the usefulness of ISO-1 by stimulation of MIF or MIF+ISO-1 to cultured mouse macrophage cell line Raw264.7 cells. Cellular polarization such as horns and adherence and the expression of Il-1B, IL-6 and IL-10 mRNAs by RT-PCR were increased in Raw264.7 cells stimulated by MIF, and these changes were significantly inhibited by MIF+ISO-1 ( Figures 4A–F ). Moreover, Arg1 mRNA, a biomarker of M2 macrophage, decreased markedly after ISO-1 treatment by RT-PCR (17) ( Figure 4C ). Intraperitoneal injection of ISO-1 (35mg/kg) was then used every other day to HDM-allergic AD mice ( Figure 4G ). After ISO-1 treatment, the mice displayed milder redness, swelling and eruption, improved hyperkeratosis in epidermis and inflammatory cells infiltration in dermis, decreased number of scratching bouts ( Figures 4H–J ), associated with the decreases of the type 2 related cytokine mRNAs, including IL-4, IL-13 and TSLP mRNAs by RT-PCR ( Figure 4K ), and the decrease of CD163⁺ M2 macrophage infiltration by immunofluorescence ( Figures 4L, M ) in the neck skin of HDM-WT mice. Therefore, treatment of ISO-1 to HDM-WT mice achieved significant improvement of AD symptoms.