Skin and blood collection from healthy volunteers or patients was performed after written informed consent was obtained. All experiments with human samples were approved by the ethical committee of the Medical Faculty of the University of Lübeck and performed in accordance with the Declaration of Helsinki.

C57BL/6J (B6) and B6.SJL-H2^s C3^c/1CyJ (B6.s) mice were obtained from colonies held at the animal facility at the University of Lübeck. Animals were fed acidified drinking water and standard chow ad libitum and were held on a 12-hour light-dark cycle. Sex-matched mice aged 6-10 weeks were used for the experiments. All experiments were approved by the Animal Care and Use Committee (Kiel, Germany) and performed by certified personnel. All clinical examinations, biopsies, and bleedings were performed under anesthesia using i.p. administration of a mixture of ketamine (100 mg/g, Sigma-Aldrich) and xylazine (15 mg/g, Sigma-Aldrich). The percentage of the affected body surface area was determined in a blinded manner.

Parsaclisib (INCB050465), a potent, selective, and orally active inhibitor of PI3Kδ, was synthesized as described (27). For in vitro studies, the drug was initially diluted in 100% DMSO (Sigma-Aldrich, Hamburg, Germany). Of this, different concentrations (500, 200, 20, 5, 2.5, 1, 0.5, and 0.1 nM) were prepared in 0.1% DMSO. For in vivo administration 0.3, 1, or 3, mg/kg of parsaclisib was freshly prepared in 95% (v/v) of methylcellulose solution (0.5% (w/v)) and 5% (v/v) N,N-Dimethylacetamide (DMA; Sigma-Aldrich). Parsaclisib was dosed twice daily by oral gavage. Methylprednisolone (MP) (Urbason™) was purchased from Sanofi-Aventis (Frankfurt, Germany). MP was used as a reference treatment and injected intraperitoneally (i.p.) once daily at a concentration of 20 mg/kg.

Immobilized immune complexes were formed using recombinant proteins of h-COL7EF and anti-h-COL7 IgG1, as previously described (28). In brief, a 96-well white plate (Greiner BioOne, Frickenhausen, Germany) was coated with 10 µg/mL of h-COL7EF in 50 mM carbonate/bicarbonate buffer (pH 9.6) for 3 h. Afterwards, wells were blocked and subsequently incubated for 2 h with 2 μg/mL of anti-h-COL7 IgG1 or PBS. In parallel, PMNs were isolated from healthy donors using PolymorphPrep™ (Axis-Shield GmbH, Heidelberg, Germany) according to the manufacturer’s instructions. After purification, PMNs were resuspended in chemiluminescence medium (color-free RPMI-1640 (Millipore Sigma) supplemented with 1% heat-inactivated fetal calf serum (FCS), 2 g/l glucose, and 25 mM HEPES) containing 1 mM luminol (5-amino-2,3-dihydro-1,4-phthalazindione; Sigma-Aldrich). 2×10⁵ PMNs were seeded into each well. In case of murine PMNs, bone marrow cells were harvested from C57BL/6J mice by flushing femurs and tibias with cold HBSS buffer supplemented with 0.5% FCS and 20 mM HEPES. Cells were collected by centrifugation and filtered through a 70 μm strainer. Neutrophils were isolated from the cell suspension using the Neutrophil Isolation Kit, mouse, an LS Column, and a MidiMACS Separator (Miltenyi Biotec GmbH, Teterow, Germany) following the manufacturer’s protocol. A total of 2×10⁵ PMNs were added to each well in the presence of 0.1% DMSO or parsaclisib (at the concentrations of 500, 200, 20, 5, 2.5, 1, 0.5, and 0.1 nM). 10 ng/mL of phorbol-myristate-acetate (PMA; Sigma-Aldrich) served as positive control. The chemiluminescence reaction was analyzed at 37°C for 60 repeats using a GloMax^® Discover System microplate reader (Promega GmbH, Mannheim, Germany).

To exclude that parsaclisib acts as a ROS scavenger, a cell-free ROS release assay was employed as previously reported (29). In brief, human myeloperoxidase (MPO; Enzo Life Sciences, Lörrach, Germany), catalase (CAT; Sigma-Aldrich), and glucose oxidase (GOX; Sigma-Aldrich) were mixed at final concentrations of 10 mU/mL, 2 U/mL, and 0.5 U/mL, respectively. The mixed substances were added to HBSS buffer containing 0.1% v/v Triton X-100, 5 mM glucose, and 2 mM luminol. Subsequently, the medium was supplemented with 0.1% DMSO or parsaclisib at concentrations of 500, 200, 20, 5, 2.5, 1, 0.5, and 0.1 nM. 10 mM N-acetyl-L-cysteine (NAC; Sigma-Aldrich) was used as a positive control. ROS generation was monitored as stated above.

In order to evaluate a potential drug-induced toxicity, a flow cytometric assay was used: PMNs, isolated from healthy donors by PolymorphPrep™, were seeded into the plate and incubated with parsaclisib (500, 200, 20, 5, 2.5, 1, 0.5, and 0.1 nM) for 2 h at 37°C. After centrifugation, the pellet was resuspended in Annexin V Binding Buffer (BioLegend GmbH, Fell, Germany). FITC Annexin V (BioLegend GmbH) was then added to the cell suspension at a final concentration of 1 μg/mL. Following 20 min incubation at 4°C, cells were washed, and the pellet was resuspended in PBS with 1% BSA. Propidium iodide (PI; Miltenyi Biotec GmbH) was added to the samples (dilution, 1:300) and cytotoxicity was measured using a MACSQuant^® Analyzer 10 (Miltenyi Biotec GmbH).

Following published protocols (30, 31), cryosections of human skin, that was obtained from elective plastic surgery, were incubated with immunoapheresis materials or IgG isolated from serum of two EBA patients using protein G columns, of EBA patients for 1 h at 37°C, followed by addition of human PMNs isolated by dextran sedimentation (Carl Roth, Karlsruhe, Germany) of freshly collected, heparinized blood from healthy volunteers (n=3). Afterwards, 0.5 ml of PMNs (2-2.5 × 107 cells), pre-incubated with solvent or parsaclisib at 2, 20, and 200 nM. PMNs, pre-incubated with solvent or parsaclisib at varying concentrations (2, 20, and 200 nM/mL) for 15 min at 37°C, were added onto the skin sections. Slides were further incubated at 37°C for 3 h. After washing, slides were fixed in formalin and H&E-stained. Using ImageJ software (https://imagej.nih.gov/ij/), DES split formation was quantified. The percentage of dermal-epidermal separation (DES) was calculated as length of separation divided by total length of the dermal-epidermal junction (DEJ) on skin section measured on a Keyence microscope (BZ-9000 series, Keyence GmbH, Neu-Isenburg, Germany).

Human PMNs were freshly isolated from healthy donors using a standard Ficoll-PaqueTM Plus (GE Healthcare Europe GmbH, Freiburg, Germany) density gradient method. The effect of parsaclisib on PMN chemotaxis was assessed by Boyden chamber (20). In brief, interleukin (IL)-8 (6 nM; Peprotech, Hamburg, Germany) was diluted in PBS with Ca²⁺, Mg²⁺/0.1% BSA and added to the bottom wells of the chamber. After covering the bottom wells with a polycarbonate membrane (pore size: 5 µm; Costar Nucleopore GmbH, Tübingen, Germany), the top wells were filled with freshly prepared PMNs (1 × 10⁵ cells/well) in PBS with Ca²⁺, Mg²⁺/1% BSA that were pre-incubated for 10 min at 37°C with parsaclisib (500, 200, 20, 5, 2.5, 1, 0.5, and 0.1 nM). After 1-hour incubation at 37°C, the chamber was disassembled and migrated cells were transferred from the bottom wells to a microtiter plate. Afterwards, migrated neutrophils were lysed by 0.1% (w/v) hexadecyltrimethylammonium bromide solution (Millipore Sigma). Number of migrated neutrophils was determined via endogenous MPO activity by adding TMB substrate solution (Thermo Fisher Scientific, Dreieich, Germany). The redox reaction was stopped by 1 M H₂SO₄ and oxidized tetramethylbenzidine was measured at OD450 nm using a GloMax^® microplate reader. The number of migrated neutrophils was calculated from a standard curve of cell lysates run in parallel.

Kinase activity profiles were determined using PamChip^® 4 microarray system (PamGene International B.V., ‘s-Hertogenbosch, The Netherlands) (36). Thirty 10 μm-thick frozen slices from lesional ear biopsies were cut and lysed using M-PER (Mammalian Protein Extraction Reagent; Thermo Fisher Scientific) containing 1% (v/v) Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific). After centrifugation (15 min, 4°C, 10,000xg), the supernatants were snap frozen and stored at -80°C. Protein concentration was determined using BCA assay kit (Thermo Fischer Scientific) according to the manufacturer’s instructions. The Serine-Threonine Kinase (STK) and Protein Tyrosine Kinase (PTK) microarray assays were performed according to the manufacturer’s instructions. A FITC-conjugated anti-phosphotyrosine antibody was used for visualization during and after the pumping of lysates through the three-dimensional surface of the array. The capture of substrate phosphorylation signals was enabled by a computer-controlled CCD camera and measured repeatedly during a 1-hour kinetic protocol using the Evolve software (PamGene International B.V.). The analysis of the images was performed using the Bionavigator Software (Ver. 6.3). A kinase was considered to be modulated (either activated or inhibited) if it had a mean specificity score (negative decadic logarithm of the likelihood of obtaining a higher difference between the groups when assigning peptides to kinases randomly) of 1 (p=0.1) and a significance score (likelihood of obtaining a higher difference for random assignment of values to treatment- and control groups) of 0.5 (p=0.32). Hence, the data obtained indicates the differences in kinase activities, always comparing control to inflamed skin. To detect changes caused by the different disease models, samples from diseased mice were compared to the respective negative controls (NR IgG vs. Solvent for pEBA and pMMP, Titermax vs. Solvent for aEBA). The mean kinase statistic (calculated by averaging the difference between the signal intensities of a sample and its corresponding control sample, normalized against a pooled estimate of the standard deviation in each sample, for each peptide assigned to a specific kinase) was used for further analysis: Venn diagram analyses were performed in Venny2.0 (https://bioinfogp.cnb.csic.es/tools/venny/). The pathway ontology was performed using the STRING database (37).

Biopsies from mice, taken from perilesional skin or mucosa, were fixed in 4% Histofix Solution (Carl Roth, Karlsruhe, Germany). The 6μm-thick sections from paraffin-embedded tissues were stained with hematoxylin and eosin (H&E) according to the standard protocols. The extent of dermal cell infiltration at the microscopic level was semi-quantified by an investigator unaware of the applied treatments. To detect tissue-bound IgG depositions at the DEJ, direct IF microscopy was performed on perilesional skin biopsies as detailed elsewhere (26).

GraphPad Prism (Version 8, GraphPad Software, San Diego, CA, USA) was used for statistical analysis of the data. All animal data (EBA score over time) are presented as mean ± standard error of the mean (SEM), all other data are presented as Tukey Box and Whisker blots (showing median and the 25-75 percentiles and all “outlying” points individually). Shapiro-Wilk test was performed to examine normal distribution of data. For comparison of means and P-value determination of two groups, unpaired t test or Mann-Whitney-U test was used. For comparison of means and P-value determination of more groups, Kruskal Wallis test with Dunn´s post test was conducted. A two-way ANOVA test was performed for comparison of means of two groups in a time frame with taking the treatment into account as independent variable. All tests were followed by Sidak’s or Dunn’s multiple comparison test and adjusted P-value <0.05 was considered statistically significant. We used the STRING database (38, 39) to build connectivity networks between kinases identified herein using PamGene and kinases known to have an impact on clinical disease manifestation in experimental pemphigoid disease, specifically EBA (18–20). Here, Indeed, most interactions were observed with PI3Kδ across all models (18). Furthermore, previous work demonstrated that PI3Kδ inhibition has therapeutic effects in experimental pemphigoid (20). Thus, we used a PI3Kδ inhibitor to address if the kinome activation signature in pemphigoid diseases correlates with the treatment outcome using a PI3Kδ-selective inhibitor. To determine a possible overlap between kinase activation, we used VENNY 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/).

We first obtained skin from mice with antibody transfer-induced MMP, antibody transfer-induced EBA, and immunization-induced EBA. Samples were taken at the end of the experiments; specifically, day 12 from the antibody transfer-induced models and at week 4 after randomization from immunization-induced EBA. Skin from control mice injected with either normal rabbit IgG (antibody transfer-induced models) or mock-immunized mice (immunization-induced EBA) was taken from corresponding anatomical sites at the same time points. As PI3K activation does not translate into substrate phosphorylation (40), its activation cannot be detected using multiplex kinase activity profiling based on the detection of substrate phosphorylation by kinases. We, hence, added PI3Kδ (Pik3cd) to the list of activated kinases identified in each model and used STRING (39) to build connectivity networks of kinase activation in the skin of mice with experimental pemphigoid disease.

In all pemphigoid disease models, differential kinase activation following disease induction was noted. The greatest number of differently activated kinases was observed in the skin of mice with immunization-induced EBA (n=17), followed by the antibody transfer-induced MMP (n=13). Whereas only three kinases were differently activated in the skin of mice with antibody transfer-induced EBA ( Figure 1 , Raw Data Table in the supplement to this article). In immunization-induced EBA, cGMP-/cAMP- dependent protein kinases dominated the network. PI3Kδ was integrated in the network through interactions with hepatocyte growth factor receptor (Met, a receptor tyrosine kinase, Figure 1A ), which are both activated after FcgR stimulation in neutrophils. The kinase activation signature in antibody transfer-induced MMP centered around PI3Kδ, Syk and the Src kinase family ( Figure 1B ). By contrast, PI3Kδ only interacted with pyruvate dehydrogenase kinase isozyme 1 (Pdpk1) in the connectivity network of activated kinases in antibody transfer-induced EBA ( Figure 1C ). When investigating for an overlap among the three models, we found that the kinase activation signatures are distinct across the three mouse models of pemphigoid diseases ( Figure 1D ): More specifically, only 2 kinases showed an overlap: Calcium/calmodulin-dependent protein kinase type IV (Camk4) between antibody transfer-induced MMP and immunization-induced EBA, and serine/threonine-protein kinase ATR (Atr) shared between antibody transfer-induced and immunization-induced EBA.

We next evaluated the impact of parsaclisib on PMNs activated by immune complexes. For this purpose, human or mouse PMNs were isolated and activated by immune complexes in the presence or absence of parsaclisib at an 8-fold concentration range. In human PMNs, a significant reduction of ROS release, as a sugorrate for PMN activation, was observed at 2.5nM, with an almost complete inhibition at 0.5 μM. Compared to human PMNs, immune complex activation of murine PMNs was less sensitive to PI3Kδ inhibition. More specifically, a significant reduction of immune complex-induced ROS release was observed at the 2x10^-7 M dose ( Figures 2A-C ). To exclude that the observed reduction of ROS is not due to any anti-oxidative properties of the compound, ROS were generated enzymatically. Here, in contrast to the treatment control, the antioxidant N-acetylcysteine (NAC), parsaclisib showed no anti-oxidative properties ( Figure 2D ). Interestingly, and differently when compared to other PI3Kδ inhibitors (20), parsaclisib had no effect on IL-8 induced PMN migration ( Figure 2E ). All effects were observed at non-toxic concentrations of parsaclisib within the same concentration range ( Figure 2F ). To validate our findings in an independent ex vivo model, we incubated IgG isolated from EBA patients on cryosections of human skin, followed by the addition of PMNs, isolated from healthy donors. In this cryosection assay, PMNs bind to the immune complexes located at the dermal-epidermal junction, become activated in a FcγR-mediated fashion (41), leading to ROS- and protease-dependent dermal-epidermal separation (42, 43). In cryosections incubated with EBA-IgG, PMN-dependent induction of dermal-epidermal separation was blocked by the addition of parsaclisib ( Figure 2G ).

Having established that PI3Kδ is integrated in the kinome activation signature of lesional skin from mice with antibody transfer-induced MMP and immunization-induced EBA and having demonstrated that parsaclisib impairs a key pathogenic pathway in autoantibody-induced tissue pathology in pemphigoid diseases, we next evaluated the impact of parsaclisib treatment on the clinical manifestation of experimental pemphigoid. In immunization-induced EBA, parsaclisib was administered in therapeutic settings: Mice were immunized with COL7 for induction of EBA. If, in an individual mouse, 2% or more of the body surface area were affected by EBA skin lesions, it was randomized to one of the treatments. In solvent-treated mice, clinical disease worsened over 2-fold during the 4-week treatment period. By contrast, parsaclisib significantly abolished disease progression. We also included methylprednisolone as an active comparator (44). The effects of methylprednisolone at 20 mg/kg and parsaclisib at 3 mg/kg were comparable ( Figure 3 ). In lesional skin from all groups, a similar degree of dermal leukocyte infiltration was observed. Likewise, a similar IgG- deposition at the dermal-epidermal junction, and comparable serum levels of COL7-specific and total IgG were noted in all treatment groups ( Figure 3 , Supplement Figure 1 ). We then determined kinase activity in lesional skin from mice treated with parsaclisib to solvent-treated mice at the end of the experiment. Like above ( Figure 1 ), we included PI3Kδ to the list of differently activated kinases between the 2 groups when analyzing the data using STRING database ( Figure 3D ). In antibody transfer-induced MMP ( Figure 4 ) and EBA ( Figure 5 ), treatments were started when pathogenic IgG was administered for the first time and was maintained throughout the 12-day observation period. In line with the findings from immunization-induced EBA, parsaclisib (3 mg/kg) or methylprednisolone (20 mg/kg) impaired the induction of skin lesions in antibody transfer-induced MMP ( Figure 4A ). By contrast, neither MP, nor parsaclisib had an impact on clinical disease manifestation in antibody transfer-induced EBA ( Figure 5A ). In experimental MMP, involvement of the oral mucosa is frequently observed. Interestingly, parsaclisib (3 mg/kg), but not methylprednisolone (20 mg/kg) significantly reduced oral blistering and erosions ( Figure 4B ). In antibody transfer-induced MMP ( Supplement Figure 2 ) or EBA ( Supplement Figure 3 ), no changes in circulating specific and total IgG were noted among the treatment groups. Other secondary endpoints in all models included analysis of cytokine and leukocyte subset expression in the skin (Cxcl1, IL17a, Tnf, Ly6g, Csf2, Cd3e, Il10, Itgam) by RT-PCR, and determination of cytokine concentrations in the serum (IL-1β, IL-4, IL-1α, IFN-γ, TNF-α, CXCL1, IL-10, IL-13, IL-17A, GM-CSF). Here, no major changes were observed, with the exception of serum cytokine expression in antibody transfer-induced MMP ( Supplement Tables 1–6 ). Here, compared to solvent treated mice, parsaclisib reduced serum concentrations of IL-1β, IL-17A and GM-CSF, as well as increasing the serum concentrations of IL-1α and CXCL1. Nothing is known about the functional relevance of these cytokines in MMP. However, if considering data from EBA and BP mouse models, IL1-β, IL-17A, GM-CSF and CXCL1 have been demonstrated to promote skin inflammation, with no contribution of IL-1α (12, 45).

In immunization-induced EBA and antibody transfer-induced MMP, PI3Kδ was linked to the kinase activation signatures when compared to healthy skin ( Figure 1 ), as well as when comparing lesional skin in mice treated with solvent to lesional skin of mice treated with parsaclisib ( Figures 3D , 4C ). In antibody transfer-induced EBA, PI3Kδ showed only few connections to the network when contrasting healthy versus lesional skin, and no interactions were observed comparing lesional skin of mice treated with solvent to those that were treated with parsaclisib ( Figure 5B ). Thus, the interaction of PI3Kδ with the kinase activation signature in inflamed skin, is associated with a favorable treatment outcome using a PI3Kδ-selective inhibitor.