The study was approved by the responsible Ethics Committees of Hannover Medical School and University Medicine Greifswald (IIIUV 23/06a, BB007/17, MHH7565). All human study participants gave their written informed consent. Serum samples were obtained from 50 atopic dermatitis (AD) patients at the Department of Dermatology and Allergy, Hannover Medical School, Germany. The median age of AD patients was 31 y; 24 patients (48%) were male, 26 (52%) were female ( Supplementary Table S1 ). Thirty serum samples from healthy volunteers were collected in-house at the Department of Immunology, University Medicine Greifswald, Germany. The healthy subjects’ median age was 23 y; 13 (28.2%) were male and 33 (71.8%) female. The samples were kept at -80°C until analysis and were used to determine antibody binding (IgG1, IgG4, and IgE) to the target proteins of S. epidermidis.

For T cell stimulation assays, EDTA blood samples were taken from 8 AD patients ( Supplementary Table S2 ) and 16 healthy individuals. The AD blood samples were obtained from the Department of Dermatology and Allergy, Hannover Medical School, Germany. At the same time, the healthy subjects were recruited in both Hannover and Greifswald, Germany. The median age of AD patients was 27 y; 5 patients (62.5%) were male, and 3 (37.5%) were female. The median age of healthy T cell donors was 30.5 y; 7 individuals (43.75%) were male and 9 (56.25%) were female.

The S. epidermidis reference strain RP62A was used in this study. The bacteria were plated on a Columbia agar blood base containing 5% (v/v) sheep blood and grown at 37°C. One single colony was grown in 10 mL of Tryptic Soy Broth (TSB) at 37°C and 200 rpm overnight and then inoculated into sterile TSB medium to an OD₅₉₅ of 0.05. The culture was incubated at 37°C until 3 hours after the stationary phase was reached.

Extracellular proteins (ECPs) of S. epidermidis were prepared as previously described (41). Briefly, the bacterial culture supernatants were collected by centrifugation at 8000 g for 10 min at 4°C and filtered through a 0.22-μm membrane to remove residual bacteria. The supernatants containing the ECPs were precipitated using trichloroacetic acid solution (100% TCA). The filtrates were mixed 9:1 (v/v) with pre-chilled 100% TCA solution and incubated on ice overnight at 4°C. After centrifugation at 8000 g for 1 hour at 4°C, the pellets were resuspended in 70% ethanol and washed four times and at the end washed once in 96% ethanol. The resulting pellets were air-dried. Finally, the precipitated proteins were dissolved in rehydration buffer (8M urea, 2M thiourea, 2% CHAPS). The protein concentration was determined by the Bradford method, and the extracts were stored at -80°C until further analysis.

One-dimensional (1D) immunoblotting was performed with an automated capillary-based blotting system (PeggySue Simple Western Assay). ECPs of S. epidermidis strain RP62A were used at a concentration of 1 mg/mL. Human serum (1:50) and horseradish peroxidase (HRP)-conjugated anti-human IgG4 secondary antibody (Thermofisher, A-10654) at a concentration of 10 μg/mL were diluted in the provided blocking diluent (ProteinSimple). The resulting signals were analyzed with Compass software 2.6.5 (ProteinSimple).

Two-dimensional (2D) immunoblotting was performed to visualize the binding of serum IgG4 to the ECPs of S. epidermidis. The procedure was performed as previously described in (42, 43), with the following adaptations. ECPs were treated with the 2D Clean-up kit (GE Healthcare) following the manufacturer’s instructions. Proteins were separated by 2D electrophoresis. In the first dimension, proteins were separated according to their isoelectric point using 11-cm Immobiline DryStrips (IPG, Immobilized pH Gradient, pH range 4-7 (Cat no. 18-1016-60); and 6-11 (Cat no. 17-6001-95; GE Healthcare). IPG strips were rehydrated for 16 hours in rehydration sample buffer (7M urea, 2M thiourea, 4% CHAPS, 20mM DTT, 1.6% ampholytes) containing 160 μg of the protein samples in a total volume of 200 μL. Isoelectric focusing was performed using a Bio-Rad PROTEAN IEF cell (Bio-Rad). The focusing was conducted at 20°C by a stepwise increase of the voltage as follows: 300 V for 0.5 h, 3500 V for 2 hours, and increases in steps of 3500 V until 15,000 V or 12,000 V was finally reached, for 4-7 IPG strips or 6-11 IPG strips, respectively. Then each IPG strip was incubated in 5 mL of equilibration buffer 1 (50mM Tris−HCl pH 8.8; 6M urea, 2%SDS, 20% v/v glycerol, 2% DTT) for 15 min with shaking and then in 5 mL of equilibration buffer 2 (50mM Tris−HCl pH 8.8; 6M urea, 2% SDS, 20% v/v glycerol, 135mM iodoacetamide) for another 15 minutes with shaking.

In the second dimension, proteins were resolved using SDS-PAGE (12,5% polyacrylamide gel) according to molecular mass, and 0.5 W per gel was applied overnight until the tracking dye reached the gel bottom.

The separated proteins were transferred onto a polyvinylidene difluoride (PVDF) membrane (Merck Millipore, IPVH00005). Electrotransfer time was 1 hour and 40 minutes at 80 mA per gel. After blocking with 5% skim milk powder in Tris-buffered saline/Tween buffer (20 mmol/L Tris-HCl, 137 mmol/L NaCl, and 0.1% [vol/vol] Tween 20 [pH 7.6]) for 1 hour at room temperature, membranes were incubated with pooled AD serum, diluted 1:500 in blocking buffer. IgG4 binding was visualized by incubation with HRP-conjugated mouse anti-human IgG4 (Thermofisher, A-10654, 0.1 μg/mL; 1 h at room temperature). Membranes were developed using the SuperSignal™ West Femto Maximum Sensitivity Substrate (Thermo Scientific™, 34095) and then visualized with the high sensitivity Chemiluminescence Imager Octoplus QPLEX (Dyeagnostics, Germany).

Western blotting was also performed to confirm the mature Esp generation and to visualize the IL-33 cleavage pattern (see below).

Briefly, for Esp detection, 0.5 μg of proEsp or 0.5 μg of mature Esp were run on 12 SDS-PAGE gels. To visualize IL-33 cleavage-patterns, the required amounts Esp and IL-33 were loaded onto a 12% SDS-PAGE gel. The proteins were then blotted onto PVDF membranes. Next, the membranes were incubated with blocking buffer Roti-Block (Roth, A151.1) for one hour at room temperature. This blocking buffer is devoid of protease activity. The blots were then incubated with primary antibodies diluted in Roti-Block blocking buffer (for Esp detection: mouse anti-His antibody, Biolegend 652501, 1:2000 dilution; for IL-33 detection: mouse anti-human IL-33 monoclonal antibody, Cayman chemical company 10809, 1:4000 dilution). The blots were incubated overnight at 4°C, with slow shaking. On the next day, the membranes were washed three times with wash buffer [TBS-Tween20 (0.05% Tween20)] and incubated with the HRP-conjugated goat anti-mouse IgG (South Biotech, 1030-05) at 1:50,000 dilution in Roti-block buffer for 1 hour at room temperature. After three washing steps, the blots were incubated with SuperSignalTM West Femto maximum sensitivity substrate (Thermo Scientific, 34095) for 5 minutes, and the chemiluminescence signals were acquired with a high sensitivity Chemiluminescence Imager Octoplus QPLEX (Dyeagnostics).

Recombinant Esp and GehD were cloned from genomic DNA of S. epidermidis strain RP62A using Gibson assembly protocols. The proteins were expressed in E. coli BL21 as recombinant proteins with C-terminal His-tag. The following primers were used to amplify the DNA sequence of Esp (Forward: ATGGTAGGTCTCAAATGAAAACCGATACAGAAAGCCATAATC, Reverse: ATGGTAGGTCTCAGCGCTCTGAATATTTATATCAGGTATATTGTTT) and GehD (Forward: gaaggagatatacaaatgGCTGAAATGACACAATCATC, Reverse: cgatcctctagcgctCTTACGTGTAATACCATCTAAC) with overhanging sequences compatible with the MCS region of the pASK-IBA33plus vector (IBA Lifesciences).

In the case of Esp, the vector pASK-IBA33plus was cut with the restriction enzyme BsaI (NEB, R3733) at 50°C for 1 hour in NEBuffer, and the reaction was stopped at 65°C for 20 min. In the case of GehD, the vector was linearized by PCR using linearization primers provided by the manufacturer. Later, the products’ ligation was performed using the Fast-Link™ DNA Ligation Kit (Epicenter, LK0750H) according to the manufacturer’s instructions. 200 μL of ligation product was transferred into chemically competent DH5α E. coli cells using standard molecular biology protocols. Recombinant colonies were randomly selected and checked for the presence of the inserted sequences by PCR. Next, pASK-IBA33plus-Esp and pASK-IBA33plus-GehD were transfected into chemical competent E.coli BL21 using standard protocols.

Protein expression was induced with anhydrotetracycline (AHT, diluted 1:10000; BA Lifesciences, 2-0401-001). The cells were harvested by centrifugation and lysed by sonication. The mixture was filtered and the protein extract was loaded on HisTrap nickel chromatography columns (HisTrap™ High Performance; GE Healthcare, 17-5248-02) using the ÄKTA Protein Purification System (Cytiva) according to the manufacturer’s instructions. The proteins were eluted with elution buffer (20 mM sodium phosphate, 0,5M sodium chloride, 500 mM imidazole, pH 7.4) using an isocratic elution program. The protein preparations were more than 95% pure, and the identity of the resulting protein was confirmed by MALDI-TOF MS analysis. For the T cell stimulation assays, LPS was rigorously depleted from the Esp preparations using an LPS depletion kit (Hyglos, Germany 800008). The final LPS concentration in the assays was < 0.1 EU/ml.

IgE, IgG1, and IgG4 levels were measured using an indirect ELISA assay previously described (34). In short, ninety-six-well plates were coated with 100 μL of either ECPs or recombinant S. epidermidis antigens (Esp or GehD) at a concentration of 5 μg/mL in coating buffer (Candor Bioscience,121125) at 4°C overnight or 1 hour at 37°C.

The plates were washed three times with phosphate-buffered saline (PBS)-Tween buffer (0.1% [vol/vol] Tween 20), and free binding sites were saturated by addition of 150 μL per well of blocking buffer (10% FCS in PBS) for 1 hour at room temperature on a plate shaker at 100 rpm. After that, 50 μL of human serum samples diluted in blocking buffer, 1:50 for IgG1 and IgG4 and 1:5 for IgE, were added to each well and incubated for 1 hour at room temperature and 100 rpm. Detection of bound IgG1 or IgG4 was achieved by incubation with HRP-conjugated secondary antibody, mouse anti-human IgG1 (Thermofisher, A-10648), or -IgG4 (Thermofisher, A-10654) at a concentration of 1 μg/mL in blocking buffer for 1 hour at room temperature and 100 rpm. To measure the IgE levels, a biotin-conjugated secondary antibody (mouse anti-human IgE, 10 μg/mL; antibodies-online, ABIN135676) was used in combination with peroxidase-conjugated streptavidin (3 μg/mL; Dianova, 016-030-084) to detect antibody binding. The plates were washed, and the reaction was developed by adding 50 μl of the activated substrate solution (TMB substrate, BD Biosciences, 555214) and incubated in the dark for 10 min. The reaction was stopped by adding 20 μl of 2 N H₂SO₄. Absorbance was measured at 450 nm in an ELISA microtiter reader (Infinite M200 Pro plate reader, Tecan Austria GmbH). Single OD measurements were performed, and the blank value in the absence of the serum was subtracted from the measured sample values.

Statistical analysis was performed with GraphPad Prism 7.0 software (GraphPad Software, USA). The unpaired nonparametric Mann-Whitney test was used for the comparison of 2 groups, showing non-Gaussian distribution.

To study the immune memory of S. epidermidis in healthy adults and AD patients, we first compared serum IgG1 and IgG4 binding to the bacterial extracellular proteins (ECPs) using an ELISA. IgG4 served as a surrogate of IgE because both immunoglobulin classes are generated under similar microenvironmental conditions (34, 46). The serum concentrations of S. epidermidis-specific IgG1 and IgG4 were much higher in AD patients than in the controls ( Figure 1A ). This indicates extensive exposure of the immune system to S. epidermidis and memory formation in AD. Using serum samples from AD patients, automated 1D immunoblots revealed a peak of IgG4 binding at the molecular mass of around 35 kDa ( Figure 1B ). We observed this signal in most individuals.

To identify the responsible bacterial protein, we analyzed IgG4 binding to the ECPs of S. epidermidis using 2D immunoblotting followed by mass spectrometry. This revealed Esp as the dominant IgG4-binding protein ( Figure 1C ).

To measure specific serum antibodies and to confirm that Esp was correctly identified as an IgG4-binding protein, we expressed the enzyme using recombinant technology, and used it as antigen in an ELISA. Recombinant triacylglycerol lipase (GehD) of S. epidermidis was included as a control antigen.

IgG1 binding to Esp and GehD was detectable in most individuals ( Figure 2A ). However, IgG4 binding differed substantially between Esp and GehD: There was strong binding to Esp – with pronounced interindividual differences – but binding to GehD was weak or absent. This was true for both AD patients and healthy controls ( Supplementary Figure S1A ). This difference is highlighted by the IgG4/IgG1 ratios of antibody binding, which were much higher in the case of Esp than GehD ( Supplementary Figure S1B ).

We then asked whether AD patients would develop an allergic response to Esp. Since the hallmark of allergy is the production of specific IgE antibodies, we developed an ELISA to measure specific serum IgE against Esp and GehD. IgE binding to Esp was significantly higher in AD patients than in healthy individuals. In contrast, the level of IgE binding to GehD was low and did not differ between the controls and the AD patients ( Figure 2B ). To compare the IgE binding to Esp and GehD, we calculated the IgE/IgG1 ratios. These corroborated the IgE bias and, hence, the type 2 immune polarizing potential of the bacterial protease Esp ( Figure 2C ).

To characterize the T cell response to Esp in the 8 AD patients ( Supplementary Table S2 ) and the 16 healthy individuals, we measured the secreted cytokines 9 days after stimulation with the recombinant proteins. The cytokine profiles of the T cell response to Esp differed substantially between healthy adults and AD patients. In healthy adults, the Esp-specific response was dominated by Th1/Th17 cytokines (IFN-γ, IL-17A, IL-17F, and IL-22) as well as by IL-6 and IL-10 ( Figure 3A ). Strikingly, the Esp-stimulated T cells from AD patients released almost no IL-17A or IL-17F and only low levels of IL-22 and IL-10. IFN-γ production was also significantly lower in AD than in healthy subjects ( Figure 3A ). In contrast, T cells from AD patients produced more Th2 cytokines than those from healthy adults (significant in the case of IL-13 and by a trend for IL-5). No differences were seen in IL-4 and IL-9 levels ( Supplementary Figure S2 ). The ratios of the Th2 cytokines and IFN-γ clearly show the Th2 bias of the Esp-specific T cells in AD as compared to healthy subjects ( Figure 3B ). In summary, the pronounced Th3 polarity of the Esp-specific T cell memory in healthy persons is distorted toward a Th2 profile in AD.

In search of molecular mechanisms underlying Esp’s type 2 polarizing potential, we reasoned that – similar to other allergenic proteases – Esp might process and activate the alarmin IL-33. We generated enzymatically active Esp by cleaving the recombinant proenzyme with the bacterial protease thermolysin ( Supplementary Figure S3 ) and confirmed its proteolytic activity in a casein cleavage assay ( Figure 4A ). The mature enzyme also showed proteolytic activity against recombinant full-length human IL-33. The apparent molecular weight of the generated IL-33 fragments ranged from 10 to 25 kDa ( Figure 4B ). Using mass spectrometry, we identified one cleavage site in the central domain of the cytokine (resulting fragment: IL-33_93-270) and three cleavage sites in the IL-1-domain (resulting fragments: IL-33_122-270, IL-33_173-270, and IL-33_205-270) ( Figure 4C , Supplementary Table S4 ). We then tested the activity of the Esp-digested IL-33 using human monocytes and monocyte-derived dendritic cells (moDCs) in cell culture. Measuring IL-6 as the readout, we noticed that monocytes and moDCs produced more IL-6 when stimulated with Esp-processed IL-33 than with the full-length IL-33 ( Figure 5 ).