Malassezia sympodialis Mala s 1 allergen is a potential KELCH protein that cross reacts with human skin

Abstract Malassezia are the dominant commensal yeast species of the human skin microbiota and are associated with inflammatory skin diseases, such as atopic eczema (AE). The Mala s 1 allergen of Malassezia sympodialis is a β-propeller protein, inducing both IgE and T-cell reactivity in AE patients. We demonstrate by immuno-electron microscopy that Mala s 1 is mainly located in the M. sympodialis yeast cell wall. An anti-Mala s 1 antibody did not inhibit M. sympodialis growth suggesting Mala s 1 may not be an antifungal target. In silico analysis of the predicted Mala s 1 protein sequence identified a motif indicative of a KELCH protein, a subgroup of β-propeller proteins. To test the hypothesis that antibodies against Mala s 1 cross-react with human skin (KELCH) proteins we examined the binding of the anti-Mala s 1 antibody to human skin explants and visualized binding in the epidermal skin layer. Putative human targets recognized by the anti-Mala s 1 antibody were identified by immunoblotting and proteomics. We propose that Mala s 1 is a KELCH-like β-propeller protein with similarity to human skin proteins. Mala s 1 recognition may trigger cross-reactive responses that contribute to skin diseases associated with M. sympodialis.


Introduction
Malassezia is an abundant genus of 18 yeast species (Vijaya Chandra et al. 2021 ) that includes the dominant commensal fungi of the human skin microbiota (Findley et al. 2013 ), with most pr e v alent distribution in skin areas of high sebum enrichment (Byrd et al. 2018 ). Besides being commensal skin colonizing yeasts, Malassezia is also linked to different skin diseases such as seborrheic dermatitis and atopic eczema (AE) (Saunders et al. 2012 ) and bloodstream infections in immunosuppressed individuals (Vijaya Chandra et al. 2021 ).
AE is a c hr onic inflammatory skin disease affecting up to 20% of c hildr en and 3% of adults (Nutten 2015, Brunner et al. 2017 ). Genetic and environmental factors leading to impaired skin barrier function have been associated with the pathogenesis of AE (Nutten 2015 ). Se v er al human pr oteins hav e r oles in maintaining skin barrier function, for example filaggrin and filaggrin-2 are important proteins in barrier formation and skin moisturization (Wu et al. 2009 ), while epiplakin and desmoplakin are essential in epidermal integrity and wound healing (Vasioukhin et al. 2001, Ishikawa et al. 2010. Skin barrier impairment results in increased contact with differ ent micr obes, suc h as the commensal yeast Malassezia, per petuating skin dama ge, as well as inducing local and systemic inflammatory reactions (Nutten 2015 ). Ten IgE-binding proteins (Mala s 1 and Mala s 5-13) have been characterized in M. sympodialis (Gioti et al. 2013 ), one of the most frequent species colonizing the skin of both healthy individuals and AE patients in Europe (Falk et al. 2005, Jagielski et al. 2014 ). These proteins act as aller-gens, inducing both IgE and T-cell reactivity in more than 50% of patients with AE (Scheynius et al. 2002, Vilhelmsson et al. 2007a, Balaji et al. 2011. One hypothesis linking M. sympodialis to the pathogenesis of AE is cr oss-r eactivity gener ated by Malassezia aller gens, whic h ar e highl y homologous to human pr oteins. Cr oss-r eactivity between M. sympodialis allergens and human proteins has been documented for Mala s 11 and Mala s 13 (Sc hmid- Gr endelmeier et al. 2005, Vilhelmsson et al. 2007a, Balaji et al. 2011. Mala s 11 is predicted to be a manganese superoxide dismutase (MnSOD) (Vilhelmsson et al. 2007a, Gioti et al. 2013. Up to 36% of AE individuals hav e high le v els of specific IgE a gainst human MnSOD and positiv e skin tests against this allergen (Schmid-Grendelmeier et al. 2005 ). Human and fungal MnSOD induce pr olifer ation of human peripheral blood mononuclear cells (PBMCs) from sensitised AE patients (Sc hmid-Gr endelmeier et al. 2005 ). In addition, CD4 + T-cells from AE individuals, specifically sensitised to the M. sympodialis Mala s 13 aller gen, wer e r eactiv e to human thior edoxin (Balaji et al. 2011 ). The high similarity with the corresponding mammalian homologs suggests that autor eactiv e T cells could contribute to tissue inflammation in AE.
Mala s 1 is a 37 kDa protein predicted to be secreted and, to date, has not been shown to have sequence homology to any specific human protein (Zargari et al. 1997, Gioti et al. 2013. The crystal structure of Mala s 1 r e v ealed it to be a β-propeller protein that binds lipids such as phosphatidylinositols (Vilhelmsson et al. 2007b ). Mala s 1 is similar to Tri14, a m ycoto xin synthesis pr otein involv ed in virulence and plant inv asion in Gibberella zeae (anamorph Fusarium graminearum) (Vilhelmsson et al. 2007b ). Mala s 1 is mainly localized in the M. sympodialis cell wall and in the budding area, as visualized by confocal laser scanning microscopy (Zargari et al. 1997 ). T herefore , Mala s 1 may be important for the replication of Malassezia and, thus, a potential therapeutic target.
Malassezia sympodialis cells can communicate with host cells, such as PBMCs and keratinocytes, through the release of extracellular nanosized vesicles, designated MalaEx (Gehrmann et al. 2011, Johansson et al. 2018, Vallhov et al. 2020. MalaEx are enriched in Mala s 1 compared with the allergens produced by whole M. sympodialis cells (Johansson et al. 2018 ). MalaEx induce a different inflammatory cytokine response in PBMCs from patients with AE sensitized to M. sympodialis compared to healthy controls (Gehrmann et al. 2011 ), supporting the link between AE and M. sympodialis . Recently, the presence of small RNAs in MalaEx was identified, suggesting that they have the potential to deliver functional mRNAs and microRNA-like RNAs to recipient host cells, thereby interfering with the host RNAi machinery to silence host immune genes and cause infection (Rayner et al. 2017 ).
Another potential mechanism of communication with host cells is related to the β-propeller protein structure of Mala s 1 (Vilhelmsson 2007b ). β-pr opeller pr oteins contain differ ent motifs allowing their classification into groups or families (Kopec and Lupas 2013 ). One of these β-propeller families contains a KELCH motif (Kopec and Lupas 2013 ). The KELCH motif is commonly found in bacterial and eukaryotic proteins with diverse enzymatic functions, and is composed of 50 amino acids, which fold into four stranded beta sheets (Interpro: IPR015915). The interest in KELCH proteins has increased in recent years as these β-propeller pr oteins ar e involv ed in m ultiple cellular functions and pr oteinpr otein inter actions with a wide cellular distribution in intracellular compartments, cell surface, and extracellular milieu (Adams et al. 2000 ).
In this study, the aims were to characterize Mala s 1 to (a) examine its precise cellular localization in M. sympodialis using high pr essur e fr eezing (HPF) tr ansmission electr on micr oscopy (TEM) , (b) investigate its potential as a drug target by examining if an anti-Mala s 1 antibody interferes with M. sympodialis growth, and (c) e v aluate potential cr oss-r eactivity between Mala s 1 and human skin proteins, including KELCH proteins, using an anti-Mala s 1 antibody in a human skin explant model.

Malassezia sympodialis and Candida albicans culture conditions
Malassezia sympodialis (ATCC 42132) was used for all experiments. Malassezia sympodialis was gr own fr esh fr om −70 • C gl ycer ol stoc ks for each experiment by plating on Dixon agar containing 3.6% w/v malt extract, 2% w/v desiccated ox-bile, 0.6% w/v bacto tryptone, 0.2% v/v oleic acid, 1% v/v Tween 40, 0.2% v/v gl ycer ol, 2% w/v bacto agar (mDixon) at 35 • C for 4-5 da ys . For culture in mDixon br oth (bacto a gar omitted), one colon y was selected fr om an a gar plate, inoculated into 10 ml of the mDixon culture medium and incubated at 37 • C for 3 days in a shaking incubator at 200 rpm. The culture was centrifuged at 200 x g , supernatant discarded, cell pellet washed twice with phosphate buffered saline solution (PBS) and resuspended in 1 ml PBS.

Sample processing and HPF TEM imaging
Malassezia sympodialis yeasts were cultured as described above on mDixon a gar plates, scr a ped off using a loop, and suspended in a minimal amount of distilled water to form a paste. Appr oximatel y 2 μl of each sample was processed by HPF using a Leica EM PACT 2 (Leica, Milton Keynes, UK). After HPF, samples wer e tr ansferr ed to a Leica AFS 2 embedding system for freeze substitution in a solution with 2% OsO 4 in 100% acetone for 40 min. Samples were then placed in 10% Spurr resin:acetone (TAAB Laboratories , Berks , UK) for 72 h follo w ed b y 30% Spurr r esin ov ernight, 8 h of 50% Spurr resin, 12 h of 70% Spurr resin, 90% Spurr resin for 8 h, and finally embedded in 100% Spurr resin at 60 • C for at least 24 h. Sections (90 μm) were cut with a diamond knife onto nickel grids using a Leica UC6 ultr amicr otome. Sections wer e contr ast stained with Ur an yLess EM stain (TABB Gr oup, Ne w York, US) and lead citrate in a Leica AC20 automatic contrasting instrument. Samples were imaged using a JEM 1400 plus transmission electron microscope with AMT ultr aVUE camer a (JEOL, Wel wyn Garden City, UK). These experiments were performed on three biological replicates.

Gold imm unola belling
Gold immunolabelling was carried out before the contrast staining step described abo ve . Samples on nickel grids were incubated in blocking buffer (1% w/v bovine serum albumin (BSA) and 0.5% w/v Tween 80 in PBS) for 20 min. Samples were then incubated three times for 5 min in an incubation buffer (0.1% w/v BSA in PBS), then incubated in 5 μg/ml primary antibody in incubation buffer (anti-Mala s 1 mouse monoclonal IgG 1 antibody (9G9) fr om Kar olinska Institute (Za gari et al. 1994, Sc hmidt et al. 1997 or mouse monoclonal IgG 1 antibody (Agilent Dako X093101-2; Santa Clara, USA) as an isotype control for 90 min. Samples were washed in incubation buffer six times for five min. The secondary donk e yanti-mouse IgG (H&L), Aurion 10 nm gold antibody (810.322; Aurion CliniSciences , Nanterre , F rance) w as added to samples for 60 min (1:40 dilution). Samples were washed six times for 5 min in incubation buffer. Samples were then washed three times in PBS for 5 min. Finally, samples were washed in double deionized, filter ed water: thr ee times for 5 min. After these steps, contr ast staining was performed as described abo ve . Yeasts wer e gr own on a gar and in broth as described abo ve . Inoculum was pr epar ed to obtain an absorbance between 0.05 and 0.1 at 600 nm. Working inocula w ere obtained b y diluting the inoculum 1:10 in mDixon broth and 100 μl of inoculum was added to all wells except for negative con-trols. Dilutions of each treatment (AmBD, voriconazole, anti-Mala s 1, and IgG 1 control) in mDixon broth was added to corresponding wells to give a final volume of 200 μl and concentration of 0.03 to 16 μg/ml. Samples were assayed in triplicate, with a positiv e contr ol (no drug) and a negative control (medium only) added. Gro wth w as monitor ed e v ery 6 h ov er 36 h by measuring optical densities at 530 nm using a microplate reader (VersaMax™ with Softmax Pro 7 software. Molecular Devices LLC, UK) incubating at 37 • C. MICs were also determined by optical density measurements using the same microplate reader at 36 and 96 h of incubation and defined as inhibition of 50% growth for voriconazole or anti-Mala s 1 antibody and 100% growth inhibition for AmBD. These experiments were performed on three biological replicates.

Skin samples and culture conditions
Human skin tissue without an adipose layer from abdominal or br east sur geries w as pur c hased fr om Tissue Solutions ® (Glasgow, UK). Ethical a ppr ov al is the r esponsibility of the compan y who quote ' Human tissue provided by this company is obtained according to the legal and ethical r equir ements of the country of collection, with ethical a ppr ov al and anon ymous consent fr om the donor or near est r elativ e. Tissue Solutions ® also compl y with the United Kingdom Human Tissue Authority (HTA) on the importation of tissues.' Skin explants from six different healthy donors, obtained under surgical aseptic conditions, were used to set up a skin model as pr e viousl y described (Corzo- Leon et al. 2021 ). In brief, the 1 cm 2 explants were inoculated by a ppl ying 1 × 10 6 yeast cells in 10 μl dir ectl y onto the epidermis. Skin samples with medium only were included as controls in all experiments. Skin samples were recovered after 6 days (1 sample), 7 days (3 samples), 8 days (1 sample), and 14 days (1 sample). Tissue samples for histology were placed into moulds, embedded in OCT Embedding Matrix (Cellpath Ltd, Ne wtown, UK) and flash-fr ozen with dry ice and isopentane . T hese wer e stor ed at −70 • C until use. Tissue samples for RNA-pr otein extr action wer e cut into small pieces ( < 5 mm) and placed in a microcentrifuge tube containing RNAlater ® solution (Merck Ltd, Dorset, UK) and stored at −70 • C until further use.

Immunofluorescent staining of skin tissue samples
For histological examination of human explant skin samples (see above), 6 μm skin tissue sections, cut from frozen OCT blocks using a cry ostat, w ere placed on poly-L-lysine-coated slides and fixed with 4% methanol-free paraformaldehyde in PBS for 20 min before further staining. After fixation, skin sections slides were washed in ice cold PBS in a Coplin jar. Samples were then permeabilised using 0.2% Triton X-100 for 10 min and washed three times with PBS at r oom temper atur e. Slides wer e incubated in blocking buffer (1% BSA, 0.1% glycine, 0.1% Tween 20 in PBS) for 1 h at room temperature.
Primary antibodies were prepared as described for gold immunolabelling in 1% BSA in PBST (0.1% Tween 20 in PBS). Slides containing 4 sections per slide and 8 sections per sample were incubated with primary antibodies at 4 • C overnight and then washed three times with PBS before incubating them with goat anti-mouse IgG (H&L) secondary antibody (Alexa Fluor ® 488 in 1% BSA PBST, 1:300; Abcam #ab150113) for 1 h at room temperature in the dark. Samples were then washed three times in PBS. Slides were incubated in 20 μg/ml calcofluor white (Sigma, UK) and 1 μg/ml propidium iodide (HPLC grade; Sigma, UK) for 30 min in a Coplin jar and washed three times with PBS. Slides were left in the dark to air dry. Mounting medium (Vectashield ® without DAPI, Vector laboratories Ltd, UK) was added, then a coverslip was added and sealed. Slides wer e ima ged using a Spinning Disk confocal microscope (Zeiss, UK).

RNA and protein extraction
Skin samples in RNAlater ® solution were recovered from stora ge (see abov e). RN Alater ® solution w as discar ded and the tissue was washed with sterile water. RNA and protein extractions wer e performed fr om the same samples in a single two-day process based on pr e vious publications (Corzo- Leon et al. 2019 ). For M. sympodialis yeasts, mDixon broth cultures were harvested by centrifugation, the pellet r ecov er ed and r esuspended in Trizol for RNA and protein extraction, similar to skin samples. Protein pellets wer e stor ed at −20 • C in 100 μl re-solubilisation buffer (1% w/v DTT, 2 M thiourea, 7 M urea, 4% w/v CHAPS detergent, 2% v/v carrier ampholytes, 10 mM Pefabloc ® SC serine proteinase inhibitor (T hermo-Fisher Scientific , UK)), with pr otein concentr ation determined by Coomassie G-250 Bradford protein assay kit following manufacturer's instructions (Sigma, UK).

SDS-PAGE gel electr ophor esis
After protein quantification, 10 μg protein was used for gel electr ophor esis and Western Blot. Non-reducing denaturing conditions were used for SDS-PAGE gel electr ophor esis. NuPAGE LDS Sample Buffer 4X (Thermo-Fisher Scientific, UK) was added at a 1:3 ratio to protein sample up to a final volume of 10 μl and incubated at 70 • C for 10 min. Samples were loaded onto a NuP a ge 4%-12% Bis-Tris gel along with a combination of SeeBlue:Magic Marker (T hermo-Fisher Scientific , UK) 4:1 ratio (final volume 5 μl) and the combination loaded into the first gel lane . T he gel was cov er ed in 1X running buffer (Novex NuPAGE MES SDS buffer, T hermo-Fisher Scientific , UK) within an electr ophor esis c hamber, 500 μl of antioxidant (Invitrogen NuPAGE™ Antioxidant; Thermo-Fisher Scientific, UK) was added to the running buffer covering the surface of the gel. Electr ophor esis was carried out for 40-55 min at 150 V.

Protein membrane transfer
Pr oteins wer e tr ansferr ed to a PVDF membr ane (Thermo-Fisher Scientific, UK) using ice cold 1X transfer buffer (NuPAGE transfer buffer, 10% methanol in deionized water). The PVDF membrane was activated prior to transfer in 100% methanol for 1-2 min, distilled water and tr ansfer buffer. Tr ansfer was carried out at 25 V for 2 h 15 min. The chamber was k e pt on ice during the transfer process.

Western blots
PVDF membr ane wer e r etrie v ed after tr ansfer and washed once with distilled water. The membrane was blocked at room temper atur e for 1 h in blocking buffer (5% skim milk in 0.1% PBS-Tween 20). The membrane was incubated overnight with the primary antibody (1:1000 in 5% BSA in 0.1% PBS-Tween 20) at 4 • C in 50 ml plastic tubes on a tube roller. The membrane was washed five times for 5 min with 0.1% PBS-Tween 20 before being incubated for 1 h at room temperature with goat ant-imouse IgG H&L secondary antibody (1:2000 in 5% skim milk in 0.1% PBS-Tween 20). Finall y, membr anes wer e washed five times for 5 min with 0.1% PBS-Tween 20 before being treated with SuperSignal TM West F emto (T hermo-Fisher Scientific , UK). T he chemiluminescent signal was ca ptur ed with the Peqlab camera (VWR, UK) and analysed using Fusion molecular imaging software v 15.5 (Fisher Scientific, UK).

Proteomics and bioinformatics analysis
Western blots (as described above) of M. sympodialis proteins and pr oteins fr om human skin wer e performed to determine cr ossreactivity of the Mala s 1 antibody. The gel was stained with Coomassie Brilliant Blue (CBB) G250 (Thermo-Fisher Scientific, UK) and bands cut out for automated in-gel digestion and identification by liquid c hr omatogr a phy with tandem mass spectrometry (LC-MS/MS) using a Q Exactive Plus (Thermo Fisher Scientific, UK) at the University of Aberdeen proteomics facility.
Gel-band proteins identified by mass-spectrometry were analysed by a ppl ying se v er al filters . T he first filter applied selected proteins by the expected protein size, based on the predicted kDa of the band detected by the anti-Mala s 1 antibody on the western blot. Then, only proteins with two or more identified peptides and two or more peptide-spectrum matches (PSM) were selected. Finall y, pr oteins wer e selected by cellular localization using the Uniprot database ( https:// www.uniprot.org/ ).

DN A extr action, ITS PCR and pGEM-T vector cloning
Genomic DN A w as extracted (Hoffman & Winston, 1987 ) from control skin explants (Corzo- Leon et al. 2021 ) or M. sympodialisinoculated skin explants stored in OCT Embedding Matrix. As a positiv e contr ol, DN A w as extr acted fr om Candida albicans (ATCC 90028) pure culture grown in YPD broth (1% w/v yeast extract, 2% w/v peptone, 2% w/v glucose). Distilled w ater w as included as a negative control.
To identify any fungi in the skin samples, the ITS1 primer pair (ITS1F (5 CTTGGTCA TTT A GA GGAA GTAA 3 ), ITS1R (5 GCTGCGTTCTTCA TCGA TGC 3 (Irinyi et al. 2015 , )) were used. The ITS1 region was amplified using 2X KAPA HiFi Hotstart ReadyMix following manufactur er's r ecommendations (Roc he, UK), with 0.4 μM primers, and 50 ng template DNA. PCR settings w ere: one c ycle of initial denaturation at 95 • C for 3 min, 30 cycles of denaturation at 98 • C for 20 s, annealing at 52 • C for 15 s, extension at 72 • C for 30 s and one cycle of final extension 72 • C for 1 min. PCR products were analysed on a 1% agarose TAE gel.
PCR pr oducts wer e purified with PEG/NaCl purification buffer for 60 min at room temperature follo w ed b y centrifugation for 1 h at 2 200 x g at 4 • C. The supernatant was r emov ed and DNA pr ecipitated in 70% ethanol, and the DNA pellets were recovered with centrifugation at 2 200 x g for 10 min at 4 • C and resuspended in sterile water. A-tailing of purified PCR products was performed using GoTaq pol ymer ase (Pr omega, UK) according to the manufacturer's instructions . T he A-tailed PCR pr oducts wer e ligated into Pr omega pGEM-T v ector following the manufactur er's instructions. Plasmid extraction was carried out with Qiagen QIApep ® spin miniprep kit following the manufacturer's instructions (QI-AGEN Ltd, UK). Inserts were confirmed by EcoRI restriction digestion and gel electr ophor esis . T he reco vered plasmids sequenced b y DN A Sequencing & Services (MRC I PPU, School of Life Sciences, University of Dundee, Scotland). Sequences were analysed using SeqMan Pro (DNASTAR, Madison, W,I US), and species identified by BLAST ( https:// blast.ncbi.nlm.nih.gov/ ).

In silico sequence analysis of mala s 1
The Mala s 1 protein sequence (ID M5E589) w as do wnloaded from Unipr ot ( www.unipr ot.org/). Pr esence of a KELCH domain in the Mala s 1 protein sequence was investigated using a pr e viousl y published consensus KELCH domain sequence (Adams et al. 2000 , Pr a g andAdams 2003 ). The potential epitope bound by the anti-Mala s 1 antibody was manually identified in the Mala s 1 protein sequence based on pr e vious publications (Sc hmidt et al. 1997, Gioti et al. 2013 ).

Mala s 1 is present in the cell wall of M. Sympodialis
Malassezia sympodialis yeast cells were processed by HPF and freeze substitution to localize Mala s 1 by TEM and immunohistoc hemistry. Gold imm unolabelling of yeast cells incubated with the anti-Mala s 1 antibody sho w ed that Mala s 1 was found mainly in the cell wall and in the budding area (Fig. 1 ). Gold particles bound to the Mala s 1 allergen were grouped in pairs , triplets , and quadruplets and binding was not specific to a particular region of the cell wall. Some gold labelling was also detected inside the yeast cells but mainly close to the cell wall rather than in the centre of the cell. No gold particles were observed either intracellularly or bound to the cell walls when yeast cells were incubated with the gold-conjugated secondary antibody only or the control IgG 1 antibody ( Supplementary Fig. 1).

Mala s 1 is not an antifungal target
We then assessed whether co-culture of M. sympodialis cells with the anti-Mala s 1 antibody could interfere with yeast growth using a modified EUCAST broth microdilution antifungal susceptibility assay (Arendrup et al. 2017). Amphotericin B deoxycholate (AmBD) and voriconazole were used as positive antifungal controls and a nonspecific IgG 1 antibody used as a negative contr ol, with concentr ations r anging fr om 0.03 to 16 μg/ml. During the first 36 h of incubation, anti-Mala s 1 antibody-treated yeasts gr e w slo w er than untreated y easts (Fig. 2 a), ho w e v er, this same pattern was observed for yeasts treated with the IgG 1 control, as w ell as v oriconazole and AmBD (Fig. 2 b-d). After 36 h, all concentrations of voriconazole and AmBD inhibited growth by at least 50% (Fig 2 e). After 96 h, only voriconazole had more than 80% inhibitory activity at all concentrations tested (Fig. 2 f). After 96 h, samples were cultured on mDixon agar without drug to identify the minimal effective, or minimum fungicidal concentration (See Supplementaryl Fig. 2). Only v oriconazole w as fungicidal for M. sympodialis at all concentrations tested (from 0.01 μg/ml up to 8 μg/ml) (Supplementary Fig. 2).

Anti-mala s 1 antibody cross-reacts with proteins in human non-inoculated skin
Next, we investigated if the anti-Malas s 1 antibody recognized human skin proteins present in skin explants as indirect evidence of potential cr oss-r eactivity between Mala s 1 and human skin proteins. Non-inoculated tissue from six different healthy skin donors was analysed by imm unofluor escence micr oscopy using the anti-Mala s 1 antibody. For all donors, samples stained with the anti-Mala s 1 antibody sho w ed detectable fluorescent signal when compared to controls (only secondary antibody or IgG 1 isotype control) (Fig. 3 ). The anti-Mala s 1 antibody stained the epidermis (as shown by co-localization with propidium iodide-stained epidermal cells) with stronger signals in the basal la yers , intercellular junctions, as well as keratinocyte cytoplasm and nucleus ( Supplementary Fig. 3). There was some variation in staining between different donors (Supplementary Fig. 3). Calcofluor white staining was used to indicate the presence of fungal structures  AmBD. Gr owth curv es r epr esent the mean of thr ee biological r eplicates, with eac h colour r epr esenting a differ ent drug concentr ation, r anging fr om 16 to 0.03 μg/ml or untreated y easts. Gro wth w as plotted after 36 h (E) and 96 h (F) as per centa ge of gr owth when compar ed to no drug contr ol. Plotted colour lines correspond to the mean of three biological replicates, error bars correspond to SD and different colours correspond to different drugs or no drug control. on the skin. No chitin was detected, indicating that no fungi were present on the skin explants (Fig. 3 ).
These non-inoculated skin samples were checked for the presence of Malassezia DNA to rule out prior colonization of the skin samples by Malassezia species or contamination of the skin during experiments . T he ITS1 region was amplified and sequenced from DNA extr acted fr om the six non-inoculated skin samples and compar ed with contr ol skin samples pr e viousl y inoculated with M. sympodialis and C. albicans . Four out of the six non-inoculated skin samples amplified ITS1 PCR products, but no M. sympodialis sequences were identified in any of the six non-inoculated samples ( Supplementary Fig. 4). One skin sample (4-17) contained a sequence of Malassezia globosa (Supplementary Fig. 4). Sequence analysis identified C. albicans and M. sympodialis from inoculated skin samples, as expected.
Pr oteins wer e then extr acted fr om non-inoculated human skin fr om thr ee r andoml y selected donors to detect potential anti-Mala s 1 antibody targets in skin. The donors selected for the proteomics anal ysis wer e those in lane 2, lane 5 ,and lane 6 (Supplementary Fig. 4b). Pr oteins wer e also extr acted fr om M. sympodialis yeast cells grown on mDixon agar and in mDixon broth as Mala s 1 protein controls. Expression of the Mala s 1 protein was e v aluated by imm unoblotting. The anti-Mala s 1 antibody corr ectl y identified a 37 kDa pr otein corr esponding to the size of the Mala s 1 protein in M. sympodialis yeast cells grown on mDixon agar (Fig. 4 a) and in mDixon broth (data not shown). The anti-Mala s 1 antibody also identified a protein ≥200 kDa in pr otein extr acts fr om non-inoculated skin samples (Fig. 4 b). These gel bands wer e r ecov er ed and pr ocessed by automated ingel trypsin digestion and the peptides separated and identified by LC-MS/MS. To identify the potential protein(s) recognized by the Mala s 1 antibody, proteins identified by proteomics ( n = 180) wer e filter ed, initiall y by selecting pr oteins of expected size ( ≥ 200 kDa = 17 proteins found), then proteins where the number of peptides was ≥ 2 and peptide spectrum matches (PSM) was ≥ 2 ( n = 12). Next, proteins were filtered according to their cellular localization and protein structure (n = 6). MalaEx can also contain Mala s 1 and can be taken up by human keratinocytes and monocytes where MalaEx are mainly found in close proximity to the nuclei (Johansson et al. 2018 ). Ther efor e, pr oteins localizing to the cell nucleus were also included. Considering the antibody binding pattern observed in skin samples (Fig. 3 ), an extra filter was applied for proteins expressed in all epidermal layers and in nuclei, c ytoskeleton, and c ytoplasm. After a ppl ying all filters, two pr oteins remained as potential targets: epiplakin and desmoplakin (Fig. 4 c).

Mala s 1 is a potential KELCH protein
Mala s 1 is a β-pr opeller pr otein (Vilhelmsson et al. 2007b ) containing 4-8 symmetrical beta sheets. We show here that the anti-Mala s 1 antibody-stained human skin sections and pr e vious data indicated that MalaEx vesicles were taken up intracellularly in keratinoc ytes and monoc ytes (Johansson et al. 2018 ). This led us to test the hypothesis that Mala s 1 might be a β-propeller protein with a KELCH motif and hence act as an allergen due to molecular mimicry with human skin proteins . T he 350 aa Mala s 1 protein sequence (Uniport ID M5E589) was analysed manually for KELCH motifs using the pr e viousl y published motif consensus (Adams et al. 2000 , Pr a g andAdams 2003 ). A 64 aa sequence was identified as having the consensus for a KELCH motif (Fig. 5 ). The anti-Mala s 1 antibody is known to recognize the peptide epitope (SFNFADQSS) (Schmidt et al. 1997 ) which lies out with the predicted KELCH motif of Mala s 1, but no sequences resembling the antibody target peptide were detected in the two potential human target proteins, epiplakin and desmoplakin. The antibody target epitope sequence was only found in the Mala s 1 protein sequence (Fig. 5 ).

Discussion
Immunostaining with TEM confirmed that Mala s 1 is localized in the M. sympodialis cell wall. Binding of the anti-Mala s 1 specific antibody to its cell wall target did not inhibit growth of the fungus suggesting that Mala s 1 is not an antifungal ther a peutic tar get. The anti-Mala s 1 antibody recognized and bound to unknown targets in full thickness human skin explants . T his cross-reactivity  In silico analysis of Mala s 1 identified a KELCH motif (highlighted in bold). Mala s 1 has a β-propeller structure (Vilhelmsson et al. 2007b ), and three hydrophobic amino acids (brown) are found before a double GG (red font), and six aa residues are located between Y (red font) and W (red font). The blue font indicates the peptide sequence reported as the epitope of anti-Mala s 1 antibody (Schmidt et al. 1997 ).
could be due to anti-Mala s 1 antibody recognising human proteins with similar tertiary structures or epitopes to Mala s 1 despite Mala s 1 not having a clear human homologue. Proteomic analysis of anti-Mala s 1 antibod y-reacti ve proteins isolated from western blotting r e v ealed a number of candidate human targets, but none contained the specific peptide epitope recognized by the antibody on Mala s 1. Bioinformatics analysis predicted that Mala s 1 contains a KELCH domain, a common motif for pr otein-pr otein interactions . T his poses the question whether the Mala s 1 KELCH domain is responsible for cross-reactivity between anti-Malassezia antibodies and human skin proteins.
Taking adv anta ge of an anti-Mala s 1 specific mouse monoclonal antibody (Zagari et al. 1994, Schmidt et al. 1997, we demonstr ated by TEM imm uno-gold labelling that Mala s 1 is mainly localized to the cell wall of M. sympodialis yeast cells, in a gr eement with Zargari et al. ( 1997 ). The immuno-gold labelling was found in a disperse pattern around the cell periphery of the mother cell wall, but also in the budding area and inside yeast cells (Fig. 1 ). As immuno-gold labelled Mala s 1 particles appeared to be enriched at the plasma membrane/cell wall interface, we cannot rule out that we are also visualizing the MalaEx vesicles (Gehrmann et al. 2011 ), whic h ar e enric hed in Mala s 1 (Johansson et al. 2018 ).
Although the anti-Mala s 1 antibody recognized and bound the M. sympodialis yeast cell wall, drug susceptibility testing did not show an antifungal effect against M. sympodialis (Fig. 2 ). This suggested that Mala s 1 is not essential for replication nor as a cell wall component; hence, Mala s 1 is not a potential target for antifungal ther a p y. AmBD w as also ineffective in the drug susceptibility testing performed in this study, whereas voriconazole was fungicidal at low drug concentr ations. Pr e vious studies hav e documented higher susceptibility of Malassezia species to triazoles, specifically triazoles with anti-mould activity (Miranda et al. 2007, Carrillo-Muñoz et al. 2013, Rojas et al. 2014, Leong et al. 2017 ). Contrary to this study, AmBD has been reported to achieve 100% growth inhibition at high MICs (r anging fr om 0.5 to 4 μg/ml) for cutaneous clinical isolates (Rojas et al. 2014, Leong et al. 2017 and MICs ≥ 8 μg/ml to AmBD were documented for bloodr ecov er ed Malassezia furfur (Iatta et al. 2014 ). These differences could be explained by the type of medium used for antifungal susceptibility testing, mDixon vs supplemented RMPI 1640 or the fungal isolates used in this study. In the future the anti-Mala s 1 antibody used in combination with antifungal drugs could be tested for synergy of the antibody with other antifungal ther a pies. Potential beneficial immunomodulatory effects and effector functions of the anti-Mala s 1 antibody could also be tested in assays with immune cells or in other ex vivo human-pathogen interaction models.
Since we found that the anti-Mala s 1 antibody did recognize human pr oteins pr esent in non-inoculated human skin explants, we used a combination of bioinformatics and proteomics to investigate potential human skin orthologous proteins that may contribute to the host's inflammatory responses to this allergen. Although se v er al candidate pr oteins wer e identified, suc h as epiplakin and desmoplakin (Fig. 4 ), we cannot definitiv el y pinpoint that the anti-Mala s 1 antibody cross reacts with a single hu-man skin protein. Bioinformatics analysis revealed for the first time that Mala s 1 allergen is a potential KELCH pr otein. Pr oteins with KELCH motifs hav e wide-r anging functions including cellular communication, with examples that contribute to microbial pathogenicity and skin function r ele v ant to this study. Two important examples of these KELCH pr oteins ar e the human intracellular NS1-binding pr otein (NS1-BP) tar geted by NS1 protein of Influenza A virus during infection (Wolff et al. 1998 ) and the KELCH K13-pr opeller pr otein in Plasmodium f alciparum wher e SNPs ar e associated with artemisinin r esistance, whic h limits the eradication of malaria in areas where these mutations are highly prevalent, for example, Southeast Asia and China (Ménard et al. 2016 ). Another example of a KELCH protein complex that functions specifically in the skin is the nuclear factor erythr oid-2-r elated factor 2/KELCH-like ECH-associated protein 1 (Nfr2/KEAP1) complex (Helou et al. 2019 ). In this complex, Nfr2 initiates the transcription of antioxidant enzymes, limits both the le v els of r eactiv e oxygen species (ROS) and inflammatory r esponses, suc h as inflammasomes, while KEAP1 suppresses these functions (Helou et al. 2019 ). Dysregulation of the Nfr2/KEAP1 system has been associated with skin damage and psoriatic plaque formation in a psoriasis-mouse model (Ogawa et al. 2020 ). Considering the multiple functions of KELCH proteins in the skin and their roles in the pathogenesis of human infections, the study of Mala s 1 allergen and determining its role during interaction with skin is highly relevant.
In the current study, two potential human target proteins that cr oss-r eact with Mala s 1 antibody were identified: epiplakin and desmoplakin (Fig. 4 ). The distribution of these two proteins in the epidermis was compared to the observed binding pattern of anti-Mala s 1 antibody to human skin. Desmoplakin (260 kDa) is closer to the size of the immunoblot band than epiplakin (460 kDa). Desmoplakin has a Src homology 3 (SH3) domain containing 6 βstrands in 2 anti-parallel β-sheets that is k e y in substrate recognition, membrane localization and regulation of kinase activity (Kur oc hkina and Guha 2013 ). Desmoplakin contains the 8 highly conserv ed r esidues, pr edictiv e of a KELCH motif consensus sequence between residues 480 and 540. Four of these residues are localized within the SH3 domain. SH3 domains can interact with KELCH proteins . For example , SH3 domain in Tyr osine-pr otein kinase Fyn has been described as interacting with KLHL2 (KELCH Like Family Member 2), which has an important role in ubiquitination and actin cytoskeleton formation (Dhanoa et al. 2013 ). Desmoplakin is essential in forming desmosomes, which are cellcell adhesions necessary for skin integrity (Vasioukhin et al. 2001 ). Recessiv e m utations in DSP (the desmoplakin gene) have been associated with palmoplantar ker atoderma, acanthol ytic epidermolysis bullosa, dermatitis, and extensive skin erosion (McAleer et al. 2015 ).
In conclusion, we propose that the IgE binding allergen Mala s 1 of M. sympodialis is a potential KELCH protein with a motif that is found in a diverse set of proteins, including over 100 human proteins, which can fold into a β -propeller. We suggest that the similarity of Mala s 1 to human proteins may lead to cr oss-r eactivity and thereby contribute to the skin inflammation amongst AE patients sensitized to M. sympodialis.
Author contributions statement C .M., A.S ., and D.M. conceived the idea and designed the study. D.E.C.L. contributed to the experimental design and performed the experiments. C.M. and D.M. contributed to project management, and supervision. All authors participated in data inter pr etation. D.E.C.L. drafted the first version of the manuscript. All au-thors contributed in writing, critically reviewed and edited the manuscript, and a ppr ov ed the final v ersion.