Identification of novel therapeutic targets for blocking acantholysis in pemphigus

Background and Purpose Pemphigus is caused by autoantibodies against desmoglein (Dsg) 1, Dsg3, and/or non‐Dsg antigens. Pemphigus vulgaris (PV) is the most common manifestation of pemphigus, with painful erosions on mucous membranes. In most cases, blistering also occurs on the skin, leading to areas of extensive denudation. Despite improvements in pemphigus treatment, time to achieve remission is long, severe adverse events are frequent and 20% of patients do not respond adequately. Current clinical developments focus exclusively on modulating B cell function or autoantibody half‐life. However, topical modulation of PV autoantibody‐induced blistering is an attractive target because it could promptly relieve symptoms. Experimental Approach To address this issue, we performed an unbiased screening in a complex biological system using 141 low MW inhibitors from a chemical library. Specifically, we evaluated PV IgG‐induced Dsg3 internalization in HaCaT keratinocytes. Validation of the 20 identified compounds was performed using keratinocyte fragmentation assays, as well as a human skin organ culture (HSOC) model. key Results Overall, this approach led to the identification of four molecules involved in PV IgG‐induced skin pathology: MEK1, TrkA, PI3Kα, and VEGFR2. Conclusion and Implications This unbiased screening revealed novel mechanisms by which PV autoantibodies induce blistering in keratinocytes and identified new treatment targets for this severe and potentially life‐threatening skin disease.


| INTRODUCTION
Pemphigus is a group of prototypical, organ-specific autoimmune diseases characterized and caused by autoantibodies targeting desmosomal adhesion molecules (Hammers & Stanley, 2016;Kasperkiewicz et al., 2017;Ludwig et al., 2017;Pollmann, Schmidt, Eming, & Hertl, 2018). The clinical disease manifestation largely depends on the autoantibody specificities. In the presence of antidesmoglein (Dsg) 1 autoantibodies, intraepidermal blistering occurs solely on the skin, causing pemphigus foliaceus (PF). If Dsg3 autoantibodies are present, lesions are restricted to mucosal sites, causing mucosal pemphigus vulgaris (PV). If both Dsg1 and Dsg3 autoantibodies are present, the skin and mucosal sites are affected, which is termed mucocutaneous PV. The particular disease manifestation on the skin and/or mucous membranes is thought to be caused by the differential expression patterns of Dsg1 and Dsg3 in these tissues (Mahoney et al., 1999;Payne, Hanakawa, Amagai, & Stanley, 2004). In addition to Dsg1 and/or Dsg3 autoantibodies, several other autoantibodies have been detected in pemphigus patients and functionally validated, for example desmocollin 3 (Chernyavsky, Amber, Agnoletti, Wang, & Grando, 2019;Lotti et al., 2019;. These non-Dsg1/3 autoantibodies are, however, less frequent compared to Dsg1/3 autoantibodies. Antigens of potentially pathogenic non-Dsg autoantibodies include desmocollin 3, M 3 muscarinic receptors and the secretory pathway Ca 2+ /Mn 2+ -ATPase isoform 1 (Amber, Valdebran, & Grando, 2018;Chernyavsky et al., 2019;. Currently, the treatment of pemphigus is based on non-specific, long-term systemic immunosuppression, including high doses of corticosteroids. Despite recent significant improvements in pemphigus treatment (Joly et al., 2017;Schmidt et al., 2017), the rate of adverse events in patients is still high. Hence, novel treatment options for pemphigus are still needed. Based on the current understanding of pemphigus pathogenesis, three different treatment targets have emerged: (a) elimination of autoreactive B cells; (b) modulation of autoantibody (IgG) half-life; and (c) inhibition of the blister-inducing activity of autoantibodies. Currently, the elimination of autoreactive B cells, as well as all other B cells, is achieved by the anti-CD20 antibody rituximab (Joly et al., 2017). Furthermore, low MW inhibitors of Bruton tyrosine kinase (BTK), which is essential for B cell activation and maturation (Corneth, Klein Wolterink, & Hendriks, 2016), have been successfully evaluated in a phase I clinical trial in pemphigus patients (Murrell, 2018). In pemphigus, blockade of the B cell activating factor of the tumour necrosis factor family receptor (BAFF receptor) and infusion of ex vivo expanded autologous regulatory T cells are also currently being evaluated in clinical trials (Izumi, Bieber, & Ludwig, 2019). In preclinical settings, autoreactive B cells were selectively depleted by chimeric autoantigen receptor T cells (Ellebrecht et al., 2016). The half-life of the circulating IgG antibodies directed against Dsg1/3 or the pathogenic non-Dsg autoantigens in pemphigus can be reduced by blockade of the neonatal Fc receptor, and phase I studies have shown promising results in pemphigus (Lee et al., 2018).
In line with these developments, adjuvant immunoadsorption (IA) has been successfully applied in severe and/or refractory cases, and Dsg1/3-specific adsorbers are currently being developed (Hofrichter et al., 2018;Langenhan et al., 2014). However, B cell targeting treatments need time until the clinical manifestation improves (Joly et al., 2017). Hence, agents that improve symptoms in the acute phase of the disease would fill a so far unmet medical need.
Finally, pemphigus may also be treated by inhibiting the blisterinducing activity of the autoantibodies. There is still ongoing debate concerning which precise molecular mechanisms contribute to blister formation induced by autoantibodies (Spindler & Waschke, 2018).
Furthermore, and in contrast to other autoimmune skin blistering diseases (Joly et al., 2002), topical treatment is not considered helpful in PV patients . Nonetheless, local blockade of signalling events would in principle be a great addition to the treatment of pemphigus because of the relatively long time needed to induce a clinical response with B cell depletion strategies. In addition, cutaneous application of low MW inhibitory molecules may be more tolerable than systemic administration. To address this issue, we first obtained a comprehensive and functional overview of signalling events driving pemphigus pathogenesis in keratinocytes after incubation with pemphigus IgG. Subsequently, based on this understanding, we evaluated the potential therapeutic use of the identified therapeutic targets.

| Experiments with human biomaterial
All experiments with human samples were reviewed by the Ethical Committee of the Medical Faculty of the University of Lübeck (reference numbers: 12-178 and 06-109) and were performed in accordance with the Declaration of Helsinki. All patients gave their written informed consent for the use of the serum or skin for research purposes.
Sera (n = 10) or immunadsorption materials (n = 2) from PV patients were taken from the remaining volume after routine diagnosis or from the discarded immunadsorption material. IgG from patient serum was isolated using protein G (cat. # 19459, Sigma-Aldrich, Taufkirchen, Germany), following established protocols (Kasprick, Bieber, & Ludwig, 2019). Each isolated IgG from PV patient serum or immunadsorption material was tested to induce Dsg3 internalization (see below) before use in our work. Human skin was obtained from elective plastic surgery. Diagnosis of PV was based on the clinical presentation and the presence of anti-Dsg1 and anti-Dsg3 autoantibodies by ELISA (Euroimmun AG, Lübeck, Germany) (Schmidt et al., 2015). The IgG concentration of each serum/IA material was determined by NanoDrop.

| Selleckchem Target Selective Inhibitory Library
The Target Selective Inhibitory Library (Selleckchem, Munich, Germany) purchased at the time of the screening contains 141 different substances that inhibit signalling pathways (Table S1). All of the substances were diluted in 100% DMSO (Sigma-Aldrich) and stored at −80°C until use.

| Experimental design and group sizes
Experimental design and analysis and their reporting follow the relevant editorial in the BJP (Curtis et al., 2018). For an initial screening, the effects of compounds from the Target Selective Inhibitory Library on Dsg3 internalization were determined. To balance between target discovery and experimental workload, we decided to perform the screening with an n of 2, which is similar (or even higher) to other drug screenings (Zhang, Wu, & Sills, 2005). The keratinocyte dissociation assay with HaCaT (CVCL_0038) cells was performed with nine independent replicates per group. If less than nine replicates were included (never less than seven), this was due to technical or procedural failures, that is, number of cells not sufficient for all groups. For validation of results obtained in the keratinocyte dissociation assay in HaCaT cells, as well as excluding the possibility that the effects observed were due to any mutations specific for HaCaT cells, we repeated these experiments using NHEK cells from single adult donors (PromoCell, Heidelberg, Germany), which are primary keratinocytes derived from different individual donors. Experiments with NHEK were performed with nine replicates per group. If less than nine replicates were included (never less than eight), this was due to technical or procedural failures. Each of the experiments using PV IgG was performed with at least three different PV IgG preparations to introduce variability. For each experiment, 3 of the 10 PV IgG preparations were randomly selected. Once a given PV IgG preparation was selected, it was again used for randomization if all other preparations had been used. Next, the human skin organ culture (HSOC) model was used to validate these findings in an ex vivo model of pemphigus. In order to have a sufficient number of controls, negative and positive control skin specimens were always performed in duplicates, and data from both (if applicable) were included. In each experiment, inhibitors were investigated in one replicate per dose.
Hence, the n for the controls is 16 (negative control [NC]) and 19 (positive control), while the n for the inhibitors ranges from 7 to 11.
For experiments using selumetinib in DAC base cream, six experiments per condition were performed. For technical reasons, one replicate of the "selumetinib intradermal" group was excluded, which explains the unequal sample size in these experiments. Non-inclusion of experiments was due to technical or procedural failures, that is, false positive/negative controls. Finally, a literature search was performed for the compounds successfully validated in the HSOC model.

| Culture of HaCaT cells
HaCaT cells were cultured in T-175 cm 2 cell culture flasks at 37°C and 5% CO 2 with the cell culture medium Keratinocyte Growth Medium 2 (KGM2; PromoCell) supplemented with Supplement Mix (PromoCell), 0.06 mM CaCl 2 (PromoCell), and 1% penicillin/streptomycin (PAN Biotech, Aidenbach, Germany). For passaging HaCaT cells, the culture medium was discarded, and the cells were detached from the flask with prewarmed trypsin/EDTA (0.05%/0.02%; PromoCell) at 37°C. The trypsin activity was stopped with FBS (Bio&Sell, Feucht, Germany), and detached cells were collected in a tube and washed twice with prewarmed KGM2. After discarding the supernatant, cells were cultured with KGM2 and seeded in 8-well chambers, with each well containing a concentration of 5 × 10 5 cells·ml −1 in preparation for the Dsg3 internalization assay.

| Dsg internalization assay
All of the following steps were performed within 8-well chambered cell culture slides (Fisher Scientific, Schwerte, Germany) that were attached to an object slide. The keratinocytes (HaCaT) were cultured until they reached~95% confluence in KGM2 with a concentration of 0.6 mM Ca 2+ . At this point, the Ca 2+ concentration was increased to 1.5 mM to induce the formation of cell contacts and the differentiation of the keratinocytes. After 24 h, fresh KGM2 and a substance from the Target Selective Inhibitory Library at a concentration of either 0.1, 1, or 10 μM, diluted in 0.1% DMSO and Dulbecco's PBS (DPBS, Gibco), were added to each well of the 8-well chamber, according to a predefined pipetting order, which was maintained throughout the experiment. After 2 h of preincubation with the substances, the medium was changed again, and the keratinocytes were exposed to KGM2 with PV IgG from one patient, with an IgG concentration of 160 μg·ml −1 . Additionally, dilution of the substances in DPBS was performed again and added to the culture wells, and the samples were incubated for another 24 h at 37°C. The supernatant of each well was removed after 24 h, and the wells were washed carefully 3 times with 500 μl of DPBS. For fixation, the cells were incubated with formaldehyde for 10 min at room temperature (RT). The keratinocyte layer was blocked after washing with a blocking buffer containing 10% goat serum, 1% BSA (Roth, Karlsruhe, Germany), and 0.05% sodium azide (Sigma-Aldrich, Hamburg, Germany) for 20 min at room temperature. The supernatant was removed again, and the keratinocytes were prepared for staining with an anti-Dsg3 antibody (mouse, monoclonal [6G2C11], IgG1; Acris Antibodies, Herford, Germany) diluted 1:100 in DPBS for 1 h. The washing step with DPBS was repeated 3 times to stain the keratinocytes with a Cy3-conjugated anti-mouse secondary antibody (goat, polyclonal; Jackson ImmunoResearch, West Grove, PA, USA) diluted 1:100 in DPBS for 1 h in the dark at room temperature. Following three washing steps with DPBS, the cells were fixed with 100% ethanol for 10 min in the dark. The 8-well chambers were carefully removed. The object slide was air-dried briefly, then mounted with 1 μl of Mowiol (Sigma-Aldrich), and covered with cover glass. The experiment was performed at 5 different time points using different passages of HaCaT cells. All of the stained slides were stored at 4°C in the dark until examination 2-3 days later under a fluorescence microscope (Keyence, Neu-Isenburg, Germany). Images at 100× magnification were photographed for later analysis.

| Evaluation of Dsg internalization using ImageJ
The slides were examined, and photos were taken from each chamber.
The settings for each photo were identical for each series of experiments. The photos were taken from five randomly chosen fields of view on each slide. The photos were taken from spots that were not located close to the edges of the field to prevent high keratinocyte density resulting in higher fluorescence intensity (FI). For each experimental series that usually consisted of~3-4 slides with eight chamber fields, one field in which only the PV IgG was added served as a positive control, and one field in which normal human IgG (NH IgG) was added served as a NC. To quantify total Dsg3 internalization, the FI of each photo was measured with ImageJ (RRID:SCR_001935, version 1.48, https://imagej.nih.gov/ij/). With the mean FI of all five photos, the mean intensity for one chamber field was calculated. This calculated mean for each inhibitor from the Target Selective Inhibitory Library (cFI) was taken in proportion to the NC, which was set at 100%. Next, the difference in FI between the positive (pcFI) and NCs (set at 100) was calculated (ΔFI). If cFI/ (pcFI + ΔFI/2) was 0.95 or higher, the compound was considered a potential inhibitor of PV IgGinduced Dsg3 internalization. As the read-out for Dsg3 internalization was an objective measurement, the investigator performing the analysis was not blinded. For each experiment, cells were incubated with the indicated drugs for 24 h. Experiments were performed in three biological replicates with two technical replicates per experiment whereby the average of the two technical replicates was used as the data point (n = 3 per group). We used this exploratory screening of potential toxicity to exclude any major effects.

| Culture of normal human epidermal keratinocytes
Analogous to HaCaT cells, normal human epidermal keratinocytes (NHEKs) were cultured in 175 cm 2 cell culture flasks at 37°C and 5% CO 2 with the cell culture medium KGM2 supplemented with Supplement Mix (PromoCell) and 0.06 mM CaCl 2 . For passaging NHEKs, the supernatant was discarded, and the cells were washed twice with prewarmed DPBS (Gibco) before adding prewarmed 0.05% trypsin/0.02% EDTA (PromoCell) to detach the cells at 37°C. Trypsin activity was reduced by adding DPBS; afterwards, the solution containing the cells was transferred immediately to a tube and washed twice by centrifugation (3 min, 220 x g, room temperature ) with prewarmed KGM2. Subsequently, the supernatant was discarded, and the cells were cultured with KGM2 at a density of~3,500-5,000 cells·cm −2 .

| Keratinocyte dissociation assay
Fifty thousand HaCaT cells or 40,000-60,000 NHEKs were seeded per well on a 12-well cell culture plate. The cells were cultivated with low Ca 2 + KGM2 (as described above) until they reached 95-100% confluency.
The medium was then changed to high Ca 2+ KGM2 with 1.5 mM CaCl 2 .
After culturing the cells for 22 h, the compounds were added (at a 0.1, 1, or 10 μM concentration) in 0.1% DMSO/high Ca 2+ medium, according to a predefined pipetting order, which was maintained throughout the experiment. After 2 h of incubation at 37°C and 5% CO 2 , the supernatant was discarded, and the cells were incubated with immunoapharesis material from a PV patient (2 mg·ml −1 ) in the presence of the inhibitors (same concentrations). For each 12-well plate, patient material in 0.1% DMSO served as a positive control, and either human IgG (2 mg·ml −1 in high Ca 2+ medium) or high Ca 2+ medium was prepared as an NC. In experiments using HaCaT cells, immunoapharesis material from different patients was used. In experiments, using NHEKs, each experiment was performed using a unique combination of either patient material and different NHEK cells. The plates were incubated again for 24 h at 37°C in 5% CO 2 . Subsequently, the supernatant was discarded, and the cells were detached as a monolayer by incubating the wells with a 1.25 U·per well dispase-HBSS solution (dispase from Stemcell Technologies Inc., Vancouver, British Columbia, Canada; HBSS from Gibco) for 15-30 min. Afterwards, the cells were washed twice with prewarmed HBSS, 1 ml of HBSS per well was added, and the cell monolayers were then exposed to mechanical stress by pipetting up and down 5 times using a 1-ml pipette.
The tips were coated in 1% BSA for 15 min to prevent the cells from sticking to the pipette tip. Finally, the cell fragments were fixed by adding 4% Roti-Histofix (Roth) and stained with crystal violet (Sigma). After 1-2 days, photos of the wells were taken with a Gel Documentation System, and cell fragments were counted manually by an investigator unaware of the applied treatments.

| Western blotting
HaCaT cells were seeded in 6-well plates and cultured using KGM2 medium (see above). For the experiment, the medium was changed to KGM2 containing 1.5 mM CaCl 2 and cultured for 24 h. Next, inhibitors at 1 μM or solvent was added for 2 h. Thereafter, the medium was changed again, and cells were exposed to PV IgG or NH IgG for 24 h.
Next, cells were lysed with 1% Triton X-100 in PBS containing a protease inhibitor cocktail (Thermo Fisher Scientific, Waltham, MA, #1862209), followed by centrifugation at 5,000 x g for 20 min. The supernatant was used as the cytosol/membrane fraction. Triton X-100-insoluble pellets were solubilized in 2% SDS sample buffer and stored as the cytoskeleton fraction, containing desmosomal components. For the immunoblotting, the samples were applied to the 8% SDS gels and SDS-

| The pemphigus HSOC model
The protocol for the HSOC model has been described in detail elsewhere (Burmester et al., 2019). In brief, human skin was cut into 5 × 5 mm sections and stored in a sterile Petri dish containing William's E medium (Gibco) on ice until further use (the skin was used within 24 h after surgery). Prewarmed defined keratinocyte serumfree medium (D-K-SFM; Gibco) was added to the wells of a Transwell cell culture insert plate (Sigma-Aldrich) to a height that created an airliquid interface between the epidermis and the air. The inhibitors were prepared at three different concentrations (0.01, 0.1, or 1.0 mM). As an NC, human IgG or DPBS was used. As a positive control, an antibody phage display-derived human bispecific monoclonal single-chain variable fragment (scFv) against Dsg3 and Dsg1 was used (Hammers et al., 2015;Payne et al., 2005). For each HSOC, at least two skin specimens were required as positive control or NC, as well as three skin pieces for testing each compound at different concentrations. Using ImageJ, the percentage of epidermal split formation was quantified by an investigator unaware of the applied treatments.

| scFv production and purification
The HA-tagged anti-Dsg1/3 scFv was produced in Escherichia coli and purified as detailed elsewhere (Payne et al., 2005).

| Data and statistical analysis
If not otherwise indicated, data are presented as mean ± SEM.

| PV IgG-induced Dsg3 internalization identifies 20 pemphigus treatment targets
In the initial screening, HaCaT cells were incubated with immunadsorption material from two PV patients in the absence or presence of compounds from the Selleckchem Target Selective Inhibitor Library (Table S1). The intraday variability of the FI was low (not shown), whereas the day-to-day variability was relatively high. For example, the average SD for NH IgG amounted to 23% and to 28% for PV IgG in the experiments of the PV10 dataset. Of the 141 compounds, 20 prevented pemphigus IgG-induced Dsg3 internalization from either one or both pemphigus patient IgG preparations by the predefined cut-off at a concentration of either 0.1 or 10 μM (Figure 1). Inhibitors of p38 MAPK and PKC were among the compounds that prevented PV IgG-induced Dsg3 internalization. These findings validated the screening approach because these compounds are known to block pathogenicity induced by PV autoantibodies in keratinocytes (Spindler & Waschke, 2018;Walter et al., 2017).

| The HaCaT cell dissociation assay validates seven pemphigus treatment targets
To validate the 20 potential therapeutic targets identified, we next

| NHEK cell dissociation validates six pemphigus treatment targets
To introduce variability, we next repeated the cell dissociation assay using primary keratinocytes originating from three different donors. All six compounds that had been selected in the cell dissociation assay using HaCaT cells also inhibited PV IgG-induced fragmentation of NHEK cells. Specifically, dose-dependent inhibition was observed for vandetanib, BX-795, and BIRB 796. Inhibition independent of the inhibitor dose was detected for selumetinib, A66, and GW441756 (Figure 3).

| The HSOC model validates four novel targets for treating pemphigus
Next, we evaluated the effects of selumetinib, GW441756, A66, vandetanib, or BX-795 on PV IgG-induced Dsg3 internalization F I G U R E 1 Modulation of PV IgGinduced Dsg3 internalization in HaCaT cells by targetselective, low MW inhibitors. (a) HaCaT keratinocytes were incubated with one of the two pemphigus vulgaris (PV) immunadsorption materials, and Dsg3 expression on the cellular surface was determined by immunohistochemistry. A compound was considered to block PV IgG induction if cFI/(pcFI+FI/2) was 0.95 or higher. The red line indicates the predefined cutoff. All data were normalized to NH IgG. Collectively, 10 compounds inhibited PV10 IgGinduced Dsg3 internalization (upper left corner), and seven compounds were identified with PV2 and three compounds inhibited Dsg3 internalization for both patients IgG. had no effect on cell fragmentation. The PI3K inhibitor TGX221 is shown as one example of a compound that had no effect on PV IgGinduced cell fragmentation. All data were normalized to the fragmentation induced by PV IgG. Data are shown as individual values, with means SD from seven to nine independent replicates per group. *P<0.05, significantly different, as indicated, from PV IgG; KruskalWallis H test followed by Dunns post hoc test. (b) All compounds, with the exception of BX795, had no toxic effects at the indicated doses. Data are shown as individual values, with medians SD from three replicates per group and is an exploratory screening of the potential toxicity F I G U R E 3 Inhibitors of MEK1, VEGFR2, PDK1, p38 MAPK, PI3K, and TrkA impair PV IgGinduced cell fragmentation in NHEK keratinocytes. (a) NHEK cell sheaths from three donors were exposed to defined mechanical stress in the presence of PV IgG from three different patients. Normal human (NH) IgG served as a negative control. Inhibitors of MEK1 (selumetinib), VEGFR2 (vandetanib), PDK1 (BX795), p38 MAPK (BIRB 796), PI3K (A66) and TrkA (GW441756) impaired the PV IgGinduced fragmentation of the cell sheaths. All data were normalized to the fragmentation induced by PV IgG. Data are shown as individual values, with means SD from eight to nine independent replicates per group. *P<0.05, significantly different, as indicated, from PV IgG; KruskalWallis H test followed by Dunns post hoc test. (b) Representative images from one experiment are shown in panel (a). The brightness of all images was increased to the same extent using GIMP (www.gimp.org) to better contrast the sheaths from the medium using Triton extraction to distinguish between the cytoskeletalanchored from the non-cytoskeletal-anchored Dsg3 pool. We refrained from testing the p38 MAPK inhibitor BIRB 796 in this assay, because intraepidermal blistering was unaltered in epidermisspecific p38α-deficient mice (Mao, Sano, Park, & Payne, 2011), and pemphigus patients treated with p38 MAPK inhibitors experienced no clinical benefit (Sayar et al., 2014). As reported (Yamamoto et al., 2007), we also demonstrate that PV IgG depletes both cytoskeletal-anchored and non-cytoskeletal-anchored Dsg3 from keratinocytes ( Figure 4). Treatment with either of the five compounds prevented Dsg3 internalization from both compartments ( Figure 4).
To further validate our findings in preclinical PV models, we selected a skin organ culture model. Experimental PV can also be induced in mice by transfer of PV IgG into neonatal or adult mice (Anhalt, Labib, Voorhees, Beals, & Diaz, 1982;Schulze et al., 2012).
However, we refrained from using this model because the mouse model does not fully reflect the situation encountered in patients (Spindler & Waschke, 2018). To standardize the skin organ culture model, we injected a well-characterized, pathogenic bispecific F I G U R E 4 Inhibitors of MEK1, VEGFR2, PDK1, PI3K, and TrkA impair PV IgGinduced internalization of the cytoskeletalanchored as well as from noncytoskeletalanchored Dsg3 pool in HaCaT keratinocytes. PV IgG was added to HaCaT keratinocytes for 24h and Dsg3 internalization from the cytoskeletalanchored and the noncytoskeletalanchored Dsg3 was assayed using western blotting of TritonX lysed cells.  (Hammers et al., 2015;Payne et al., 2005). This scFv consistently leads to the induction of intraepidermal blister formation when injected into normal human skin ( Figure 5). In the presence of the anti-Dsg1/3 scFv, intraepidermal splits were present in the skin specimen, while injection of NH IgG (NC) left the skin intact. If the skin was co-injected with the MEK1 inhibitor selumetinib, dose-dependent inhibition of anti-Dsg1/3 scFv-induced intraepidermal split formation was observed. Similar observations were made for the p38 MAPK inhibitor BIRB 796. In contrast, inhibition of anti-Dsg1/3 scFv-induced skin pathology was independent of the concentration of GW441756 and A66. Despite having demonstrated therapeutic effects in all of the above assays, the PDK1 inhibitor BX-795 did not modulate anti-Dsg1/3-induced acantholysis ( Figure 6).
As topical use of these inhibitors may be associated with less adverse events as opposed to systemic application, we evaluated the effects of topical application of one of the inhibitors in the HSOC model. We selected selumetinib because it was among the most potent inhibitors when co-injected with the anti-Dsg1/3 scFv ( Figure 6a). For this purpose, selumetinib was emulsified in base cream. As a control, the effect of 1 μM selumetinib was investigated in parallel. Topical selumetinib treatment reduced the median split formation from 44.6% to 13.2% (~70% reduction), while co-injection of the compound reduced median split formation from 63.3% to 6.1% (~90% reduction; Figure 6c-d).

| DISCUSSION
Pemphigus is a potentially life-threatening disease. Treatment with high doses of corticosteroids in combination with the anti-CD20 antibody rituximab induces complete remission in the majority of patients (Joly et al., 2017;Kasperkiewicz et al., 2017). However, complete remission is only achieved after 120-180 days of treatment, severe adverse events occur in one third of patients, and 20% of patients do not reach complete remission (Joly et al., 2017). Hence, there is a clear need for the development of novel therapeutic options for pemphigus. Fortunately, several new treatment strategies are under clinical investigation. However, these treatments focus solely on the blockade of autoreactive B cells or enhanced clearance of autoantibodies (Hofrichter et al., 2018;Lee et al., 2018). Previously, application of a tandem peptide directed against Dsg3 was reported to inhibit p38 MAPK activation and keratin filament retraction both in vitro and in the neonatal mouse model of pemphigus. However, this project has not been translated into a clinical application (Heupel et al., 2009;Spindler et al., 2013). Thus, to the best of our knowledge, no treatment targeting autoantibody-induced intraepidermal blistering is under development.
Given its rapid therapeutic effect, modulation of autoantibodyinduced intraepidermal blistering would be an ideal adjuvant treatment modality to corticosteroid/rituximab treatment because it would induce healing of lesions at a time at which corticosteroid/rituximab treatment is not yet effective. To address this unmet medical need and based on the notion of aberrant keratinocyte signalling in pemphigus (Pollmann et al., 2018), we used an unbiased screening approach to identify novel treatment targets for pemphigus that specifically modulate pemphigus autoantibody-induced loss of keratinocyte adhesion. Through primary screening and multiple validation experiments, including an HSOC model, we identified four novel therapeutic targets that were able to modulate autoantibody-induced intraepidermal blistering in pemphigus (Figure 7). Specifically, these targets are MEK1, TrkA, PI3Kα, and VEGFR2, as the corresponding inhibitors impaired anti-Dsg1/3-induced pathology in four different pemphigus model systems. In addition, we identified the p38 MAPK pathway, one of the best described anti-Dsg3-induced signalling cascades, as an important cause of pathology in the assays employed here. While pharmacological inhibition of the p38 MAPK pathway almost completely abolished the blister-inducing activity of pemphigus patient IgG after injection into neonatal mice (Berkowitz et al., 2006), intraepidermal blistering was unaltered in epidermis-specific p38α-deficient mice (Mao et al., 2011), and pemphigus patients treated with p38 MAPK inhibitors experienced no clinical benefit (Sayar et al., 2014). Hence, p38 MAPK and its downstream signalling components, such as HSP27, are not likely to be therapeutic targets for pemphigus.
Based on the results presented here and elsewhere (Hariton et al., 2017;Pollmann et al., 2018;Spindler & Waschke, 2018), F I G U R E 5 AntiDsg1/3 scFv induces acantholysis in a full thickness human skin organ culture model. (a) Images of six skin specimens in culture. (b) H&Estained section from a skin section incubated with normal human (NH) IgG shows no intraepidermal blistering. (c) AntiDsg1/3 scFv (red) and DAPI (blue)stained sections from a skin section incubated with NH IgG show no autoantibody binding to keratinocytes. (d) H&Estained section from a skin section incubated with pemphigus vulgaris (PV) IgG, with extensive intraepidermal blistering. (e) AntiDsg1/3 scFv (red) and DAPI (blue)stained sections from a skin section incubated with antiDsg1/3 scFv show the presence of autoantibodies bound to keratinocytes, as well as intraepidermal blistering. Scale bars correspond to 100m F I G U R E 6 Inhibitors of MEK1, TrkA, PI3K, p38 MAPK, and VEGFR2 impair PV IgGinduced intraepidermal split formation in the human skin organ culture model. (a) The antiDsg1/3 scFv was injected into human skin samples, in the absence or presence of the indicated inhibitors. Data are shown as individual values, with medians SD from 819 sections per group from three organ cultures. *P<0.05, significantly different from antiDsg1/3 scFv alone; KruskalWallis H test followed by Dunns post hoc test. (b) Representative H&Estained images from the experiments are shown in panel (a). For better contrast, white correction was applied using GIMP (www.gimp.org). Scale bars correspond to 100m. (c) The same experimental setup as for panel (a), with the exception that selumetinib was emulsified in base cream (1%), which was applied on top of the organ culture. For control, skin specimens were treated with base cream alone. Data are shown as box (first and third quartiles) and whisker (range) plots with medians, and are based on 3940 sections from four to five organ cultures. * P<0.05, significantly different from base cream alone; ranksum test. (d) The same experimental setup as for panel (a), performed in parallel to the experiments from panel (c) as a positive treatment control. Data are shown as box ( first and third quartiles) and whisker ( ) plots with medians and based on 3744 sections from five to six organ cultures; * P<0.05, significantly different from solvent alone; ranksum test. (e) Representative H&Estained images from the experiments are shown in panel (c). For better contrast, white correction was applied using GIMP (www.gimp.org). Scale bars correspond to 100m F I G U R E 7 Summary of screening and validation. A total of 141 compounds from the Selleckchem Target Selective Inhibitor Library (TableS1) were used for the initial screening. Of these, 20 compounds were defined as potential inhibitors of PV IgGinduced pathology in keratinocytes and were used in the first validation experiment. Next, in the HaCaT cell dissociation assay, seven compounds were selected for further validation. In the NHEK cell dissociation assay, 6/6 compounds displayed inhibitory effects on PV IgGinduced pathology. Ultimately, these remaining compounds were tested in the human skin organ culture model, and 5/6 of the compounds impaired the induction of intraepidermal blistering induced by a monoclonal bispecific antiDsg1/3 scFv. Based on the literature (Sayar et al.,2014), we excluded p38 MAPK from the list of potential therapeutic targets. In summary, the screening and subsequent validation identified four novel therapeutic targets for the modulation of PV IgGinduced intraepidermal blistering, namely, MEK1, TrkA, PI3K, and VEGFR2 EGFR (Sayar et al., 2014), VEGFR2, and TrkA, as well as their downstream targets (Figure 8), seem to be more promising therapeutic targets for modulating intraepidermal blistering induced by pemphigus autoantibodies. Modulation of these pathways may be achieved by either systemic or topical administration of compounds.
As for corticosteroids (Joly et al., 2002), topical application should have significantly fewer adverse events than systemic treatment.
Topical treatment using low MW compounds has already been described for a number of inflammatory skin diseases (Craiglow, Tavares, & King, 2016;Putterman & Castelo-Soccio, 2018). Albeit for other, mostly oncological indications, compounds targeting the identified pathways are approved or are currently being evaluated in clinical trials, such as the VEGFR2 inhibitor apatinib (Miao et al., 2018), the TrkA kinase inhibitor CT327, which lessens itching in psoriasis patients as a topical formulation (Roblin et al., 2015), the PI3Kα/δ inhibitor copanlisib (Dreyling et al., 2017), and several MEK inhibitors (Sarkisian & Davar, 2018). Of these inhibitors, the topically applicable TrkA inhibitor CT327 seems the best suited for testing in pemphigus patients.
While we believe that screening using a complex biological system has several advantages over molecule-based screening methods, the data obtained here should be interpreted within the limitations of the methodology. Due to the complexity of the Dsg3 internalization assay, only a limited number of experiments can be performed. The screening of 141 compounds with two replicates required 5-6 months of work by an experienced researcher. Furthermore, the complexity of the experiment induces variability. Of note, the results obtained with PV2 did not correlate with those obtained with PV10 (Pearson correlation).
Hence, we decided to include any experiment at any concentration of potential drug candidates if the given compound had the predefined inhibitory effect on PV IgG-induced Dsg3 internalization. Overall, this approach may introduce a bias towards not detecting true inhibitory compounds. When evaluating this possibility, the screening missed detecting EGFR (Sayar et al., 2014) because PD153035, a potent and specific EGFR inhibitor with an IC 50 in the picomolar range (Fry et al., 1994), is present in the library. PKC (Seishima et al., 1999) was identified in the screening because enzastaurin impaired PV IgG-induced Dsg3 internalization in both experiments. However, in the HaCaT cell dissociation assay, enzastaurin at 0.1 or 1.0 μM had no effect on PV IgGinduced fragmentation, and at 10 μM, the compound increased cell fragmentation over twofold (P < 0.05, Kruskal-Wallis H test followed by Dunn's post hoc test). Caspase, ERK, and Src inhibitors were not present in the library.
Thus, screening with PV patient IgG rather than defined autoantibodies better reflects the situation encountered in patients. In addition, the use of HaCaT cells, an immortalized cell line with potentially different signalling pathways compared to nonimmortalized cells, may have contributed to missing some compounds that would have inhibitory effects in other cell culture systems. Hence, taken together, our screening approach was biased towards false negative results.
We decided to use the anti-Dsg1/3 scFv in the HSOC model as opposed to pemphigus mouse models in neonatal mice or IgG preparations from pemphigus patients, because the neonatal mouse skin is quite different from human adult skin, and patient IgG does not consistently induce splitting in the skin organ culture model (not shown).
The ex vivo model used in this study is well-standardized. While, due to relatively low variability, this is an advantage in identifying potential new pathways in pemphigus, the model has to be interpreted within its limitations. Specifically, while we detect scFv binding in the model, the protein expression (determined by immunohistochemistry) of Dsg1/3 is not grossly altered (not shown). Hence, the precise mechanisms by which blistering occurs in this model, as well as in patients, need to be addressed in more detail in further experiments .
The advantage of using a complex system as opposed to an enzymic assay for a known target is the fact that no therapeutic target has to be known a priori. Hence, screening in biological systems can be performed in an unbiased fashion without prior insights into the molecular mechanisms of the disease. In addition, if a desired (inhibitory) effect is observed, the inhibited pathway is likely of functional relevance.
In summary, we provide novel insights into PV IgG-induced signalling events in keratinocytes, which contribute to skin pathology in pemphigus. Furthermore, by applying an unbiased screening approach, with subsequent validation in a standardized skin organ culture model, we were able to identify four novel potential therapeutic targets for the treatment of pemphigus.