Lesional skin of seborrheic dermatitis patients is characterized by skin barrier dysfunction and correlating alterations in the stratum corneum ceramide composition

Seborrheic dermatitis (SD) is a chronic inflammatory skin disease characterized by erythematous papulosquamous lesions in sebum rich areas such as the face and scalp. Its pathogenesis appears multifactorial with a disbalanced immune system, Malassezia driven microbial involvement and skin barrier perturbations. Microbial involvement has been well described in SD, but skin barrier involvement remains to be properly elucidated. To determine whether barrier impairment is a critical factor of inflammation in SD alongside microbial dysbiosis, a cross‐sectional study was performed in 37 patients with mild‐to‐moderate facial SD. Their lesional and non‐lesional skin was comprehensively and non‐invasively assessed with standardized 2D‐photography, optical coherence tomography (OCT), microbial profiling including Malassezia species identification, functional skin barrier assessments and ceramide profiling. The presence of inflammation was established through significant increases in erythema, epidermal thickness, vascularization and superficial roughness in lesional skin compared to non‐lesional skin. Lesional skin showed a perturbed skin barrier with an underlying skewed ceramide subclass composition, impaired chain elongation and increased chain unsaturation. Changes in ceramide composition correlated with barrier impairment indicating interdependency of the functional barrier and ceramide composition. Lesional skin showed significantly increased Staphylococcus and decreased Cutibacterium abundances but similar Malassezia abundances and mycobial composition compared to non‐lesional skin. Principal component analysis highlighted barrier properties as main discriminating features. To conclude, SD is associated with skin barrier dysfunction and changes in the ceramide composition. No significant differences in the abundance of Malassezia were observed. Restoring the cutaneous barrier might be a valid therapeutic approach in the treatment of facial SD.


| INTRODUC TI ON
Seborrheic dermatitis (SD) is an inflammatory, eczematous skin disease of the face and scalp with a multifactorial underlying pathophysiology.
SD is characterized by the development of erythematous, scaly and itchy skin on seborrheic areas with high sebaceous gland activity such as the nasolabial folds, eyebrows and upper chest (Figure S1). 1 The exact pathophysiology of SD remains unclear due to its multifactorial and complex aetiology.Three major interdependent driving factors of the aberrant immunological responses behind SD are (I) individual susceptibility due to an imbalanced immune system leading to inflammation, (II) cutaneous microbial dysbiosis with pronounced colonization by Malassezia species and (III) a perturbed epidermal barrier. 1,2ile these three hallmarks all contribute towards the development of SD, much emphasis has been on the microbiome and especially the involvement of Malassezia.Malassezia is a commensal yeast which is regarded as a key pathogen due to its concurrence with lesional skin and the clinical response of SD to antifungals. 3It is hypothesised that the predilection of Malassezia for sebum-rich skin sites is due to its inability to perform de novo fatty acid synthesis, necessitating the processing of exogenous lipids which disturbs the epidermal barrier integrity and enables inflammation. 4,5While this provides rationale for why inflammation is limited to these areas, it has been shown that neither the amount of Malassezia 5,6 nor an increased level of sebum production 1,2,6 are strictly tied to the development of SD.
Due to the implication that external triggers such as Malassezia and its metabolites can penetrate the skin, 7 the cutaneous barrier function itself has been proposed to be involved in SD pathogenesis. 6The epidermal barrier function is located in the stratum corneum which consists of layers of cornified cells embedded in a lipid matrix mainly composed of cholesterol, fatty acids and ceramides. 8anges in the composition and consequently the lipid organization of this matrix directly impacts skin permeability. 9,10Barrier perturbations and simultaneous alterations of the lipid matrix composition, such as reduced chain length and changes in ceramide subclass composition, have been observed in other inflammatory skin diseases such as atopic dermatitis and psoriasis. 11,12This raised the question whether barrier repair can be exploited as a treatment option. 13e apparent contribution of host immunity, the microbiome and cutaneous barrier to the development of SD warrants a multimodal assessment for phenotyping SD.In this study, we established cutaneous inflammation by clinical scoring complemented with imaging.We elucidated the bacterial and fungal composition as both are implicated in SD, 14 with additional species-level profiling of Malassezia.Lastly, the cutaneous barrier function was characterized in-depth by trans-epidermal water loss (TEWL) measurements complemented with ceramide profiling using lipidomic analysis.This might yield new insights into how these modalities are implicated in disease.

| ME THODS
An extended version of the methods can be found in the supporting information provided online.

| Study design and population
The study was conducted at the Centre for Human Drug Research

| Clinical characteristics
Disease severity was scored using the Seborrheic Dermatitis and Severity Index (SDASI) adapted from Baysal, et al., 15 5-point IGA and percentage affected body surface area (%BSA).Patient reported outcomes included the 0-100 Numeric Rating Scale (NRS) itch, 5-Domain Itch Questionnaire 16 and Dermatology Life Quality Index (DLQI). 17,18See Supplementary Material for further details.

| Standardized photography
Standardized two dimensional cross-polarized images of the face were taken using a VISIA-CR (Canfield Scientific).Erythema Index calculations were performed based on a method by Yamamoto, et al. 19 See Supplementary Material for further details.

| Optical coherence tomography
Lesional and non-lesional skin was imaged by Vivosight Dx optical coherence tomography (OCT) (Michelson Diagnostics, Kent) and epidermal thickness, superficial roughness and average epidermal vascularization were determined.See Supplementary Material for further details.

| Microbiome composition
Skin swabs were collected by rubbing the skin for 10 s.Swabs were extracted and 16 s rRNA and internal transcribed spacer (ITS) sequencing was performed to determine the bacterial and fungal composition, respectively.After genus level classification, microbes contributing <1% of the total were excluded and relative abundances determined.See Supplementary Material for further details.

| Malassezia culturing
Agar plates with modified Dixon medium were pressed against the skin for 20 s and cultured for Malassezia.Malassezia species determination by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) was performed on mycological isolates as described by Kolecka et al. 20 See Supplementary Material for further details.

| Skin barrier integrity by trans-epidermal water loss
Subjects were allowed to acclimatize to controlled environmental conditions (humidity <60%, temperature 22 ± 2°C) in rested state for at least 15 min prior to measurements.TEWL was measured using an AquaFlux AF200 (Biox Systems Ltd.).See Supplementary Material for further details.

| Skin barrier lipidomics
Stratum corneum was harvested with four polyphenylene sulfide tape (Nichiban) after applying pressure using a D500 D-squame Pressure Instrument (CuDerm Corporation).Tapes were extracted and the ceramide fraction analysed through a validated liquid chromatography-mass spectrometry (LC-MS) setup as described by Boiten et al. 21Figure S2 lists the 12 most prevalent ceramide classes included in the analysis using the nomenclature by Motta et al. 22 Sebum levels were determined using a Sebumeter SM815 (Courage+Khazaka, Köln, Germany).See Supplementary Material for further details.

| RE SULTS
In total, 37 patients were enrolled into the study.The patient population exhibited mild-to-moderate SD as shown by a SDASI score of 7.0 ± 4.3 and IGA score of ≤3 for 97% of all patients (Table 1).Patient reported disease burden was rated mild-to-moderate as evident from the DLQI (7.2 ± 5.5/30; 'moderate effect on patient's life'), av-

| Inflammation
Apart from clinical scoring, hallmarks of inflammation were assessed using standardized 2D-photography and OCT.Lesional skin had a significantly higher erythema index compared to non-lesional skin

| Microbiome
After establishing the presence of inflammation, the facial microbial composition was investigated using 16 s rRNA and ITS sequenc- skin sites with M. sympodialis being the most prevalent at both sites (21.6% and 18.9% on lesional and non-lesional skin, respectively), followed by M. slooffiae (13.5%) on lesional and M. globosa (13.5%) on non-lesional skin (Figure 2E).

| Skin barrier
Finally, the skin barrier was studied as it represents the interface between external pathogens and the established epidermal inflammation.TEWL was used as an endpoint for skin barrier integrity and was significantly higher in lesional skin compared to non-lesional skin  3F).
Visualization of the roughly 300 individual ceramide responses using dimension reduction analysis by PCA showed two distinguishable populations when stratifying for skin site (Figure 3G).This indicates the ceramide profile of lesional skin is alike between subjects, but differs from non-lesional skin.Plotting the TEWL values against ceramide parameters revealed a positive correlation for ceramide

| Integration of results
Integration of quantitative clinical characteristics (Figure 1A,C,D), microbial properties (Figure 2A-D) and barrier parameters (Figure 3A-F) was performed using the entire dataset.The resulting PCA shows two distinguishable sets of data points with minimal overlap when stratifying for site (Figure 4A).The abundance of Cer[NS], the carbon chain length and the Cer[NS]:Cer[NP] ratio contribute most to the overall differential analysis (Table S4).Min-max normalized visualization of the most significant findings in a radar chart underline that the biggest differences are observed in the barrier compartment followed by inflammation (Figure 4B).Small differences between lesional and non-lesional skin were observed in the microbial parameters, with only the abundance of Staphylococcus being markedly different.

| DISCUSS ION
SD is a multifactorial disease in which the interplay between the cutaneous microbiome, especially the presence of Malassezia, impaired skin barrier function and abnormal immunological responses seem integral to its pathogenesis. 1In this study we comprehensively and non-invasively characterized the clinical representation of mild-to-moderate SD.We demonstrated a profound involvement of cutaneous barrier dysfunction with only small alterations in the microbiome, including the abundance of Malassezia, based on differences between lesional and non-lesional skin.This trial was performed in a sizeable number of 37 patients with similar disease burden after appropriate wash-outs and screening.

| Inflammation objectively quantified by OCT and standardized imaging
Visual assessment of SD, which includes the evaluation of erythema, is frequently used in daily clinical practice for disease monitoring.
However, visual examination of the skin can be hampered by limited sensitivity, observer bias and overall intra-and inter-rater variability. 23,24Therefore, we selected an objective approach to quantify cutaneous inflammation.Digitalized erythema assessments have been reported but not applied to SD. [25][26][27][28] Here, standardized crosspolarized light photography is used to capture consistent images and enhanced erythema, 29 resulting in a clear differentiation between lesional and non-lesional skin.Optical biopsies by OCT enabled the determination of additional (sub)cutaneous parameters.
Inflammatory characteristics such as increased perfusion 30 and epidermal thickness, corresponding to acanthosis, 31,32 were observed in lesional skin.Furthermore, a rougher lesional skin surface corresponds to the scaly phenotype of the disease. 1 Increased blood flow and epidermal thickness have been observed by OCT in the involved skin of psoriasis and atopic dermatitis patients compared to uninvolved skin and healthy controls. 33,34Stand-alone, but also combined, standardized photography and OCT qualify as valuable non-invasive tools which enable sensitive and selective endpoints for disease monitoring and detection of treatment responses in clinical trials. 35

| Malassezia and Staphylococcus dominate the lesional microbiome
Bacterial analysis of the skin surface showed an increased abundance of Staphylococcus and decreased abundance of Cutibacterium on lesional skin which concurs with previous SD profiling studies. 14,36Staphylococcus, and especially S. aureus, is considered to be a pro-inflammatory mediator in atopic dermatitis. 37While limited phylogenetic resolution prevents the identification of S. aureus in this study, the observed increase of Staphylococcus combined with reports that S. aureus is more abundant in SD patients compared to healthy controls might indicate bacterial involvement in SD pathogenesis. 38Despite reports of bacterial involvement in SD, microbial involvement remains primarily focussed on Malassezia as a key pathogen.However, no differences were observed between the abundance of Malassezia on lesional and non-lesional skin.This is in accordance with findings showing the presence of Malassezia is neither limited to lesional skin nor SD patients. 3,6,14Additionally, facial skin of healthy volunteers also showed a seemingly small and Malassezia dominated mycobiome. 39,40Although it is hypothesised that specific Malassezia species might be responsible for instigating inflammation as virulence factors differ between Malassezia species, [41][42][43] no significant differences on species level were observed in this study.Culturing led to successful isolation of Malassezia species in approximately half of the subjects, illustrating the known challenges of isolating Malassezia from clinical samples. 44Except for M. slooffiae, all species are relatively frequently isolated from healthy and SD skin. 45Remarkably, M. slooffiae has been reported to have little virulence when directly compared to M. globosa and M. sympodialis. 42However, intra-species variation in virulence has been reported indicating that microbial activity rather than abundance is an important factor for the association of specific species to lesional skin. 46,47sed on these results, it seems too straightforward to attribute SD pathogenesis to the presence of Malassezia alone.as barrier component. 48,49In this study, we show a substantially impacted barrier in lesional skin on functional grounds by TEWL and demonstrate concomitant changes in the ceramide profile.

| Substantial functional and compositional barrier alterations
These compositional changes correlated with the degree of barrier impairment as judged by TEWL.In line with our study in SD, changes in the Cer[NS]:Cer[NP] ratio, 50 degree of unsaturation, 51 ceramide chain length 12,52 and presence of extremely short chain Cer[NSc34] 12,51,53,54 in lesional skin have been observed in atopic dermatitis where barrier involvement is firmly established.These changes to the lipid profile appear to be induced by inflammation as lipid alterations can be evoked by atopic dermatitis and psoriasis pro-inflammatory cytokines in vitro [55][56][57] and normalize in response to anti-inflammatory treatment in atopic dermatitis patients. 52Whether these alterations are therefore primarily linked to inflammation and only coincide with barrier dysfunction has been investigated in mechanistic studies using lipid model systems.
Lacking an inflammatory component, these models have shown that increased Cer[NS], 58,59 increased unsaturation 60  skin and acne, but without much effect on the relative ceramide subclass composition as reported on in the current study. 62,64,65ditionally, the impact of these factors might be limited as patients serve as their own control.The predilection of SD lesions with areas known for increased water loss such as the mouth, eyelids and nasolabial folds might be confounding for the increased TEWL. 49,66Indeed, healthy volunteers have shown TEWL values at the nasolabial fold that approach the lesional values observed in this study with comparable TEWL values at the cheek or forehead, sites where non-lesional measurements were often conducted. 67,68erefore, increased TEWL values in lesional skin may not reflect barrier impairment but rather indicate differences in normal physiological functioning between skin sites.However, the concurrent correlations between the TEWL values and ceramide composition reaffirm the interdependence of SD functional barrier impairment and ceramide-compositional alterations.

| Integrative data analysis emphasizes importance of barrier dysfunction
An integrative approach was taken to visualize the data after investigating the three hallmarks of SD separately.Using PCA, we elucidated which parameters of our dataset predominantly contribute to the SD phenotype.The abundance of Cer[NS] and ceramide chain length showed to be the most important discriminating features.
Indeed, the radar plot directs emphasis towards barrier function with little differences in the micro-and mycobiome.While the contribution of Malassezia seems to be negligible when only considering relative abundances, it should be re-emphasized that SD appears to be neither caused solely by barrier dysfunction nor microbial involvement but rather by the interplay between factors.This finding correlates with the shifting belief in literature that Malassezia might not be solely responsible for causing SD. 5,6 This highlights the added value of a multimodal and integrative approach to disease profiling which enables in-depth characterization with the possibility to unravel part of pathogenesis. 35

| CON CLUS ION
In conclusion, this study demonstrates the importance of the barrierinflammation axis in mild-to-moderate SD which seems to be more prominently involved compared to the microbiome.While not incorporating an internal healthy control group, our results are compared thoroughly with prior research in healthy volunteers through literature.Moreover, the results agree with and support existing literature regarding inflammation and microbial involvement in SD while complementing the current understanding of barrier dysfunction in SD.
Barrier impairment parallels that of atopic dermatitis where emollients are used effectively. 69Treating SD by improving the skin barrier function has been proposed as a potential adjuvant therapy, 70 but the management of SD remains focussed on anti-inflammatory and anti-fungal treatments. 71Taken together, the incorporation of emollients, humectants or other barrier repair agents should not be neglected in the management of SD and might support current treatment modalities.
final manuscript.The other authors have no conflicts of interest to declare.

(
Leiden, the Netherlands) from November 2018 to January 2020 following the Declaration of Helsinki principles after approval by the medical ethics committee Stichting Beoordeling Ethiek Biomedisch Onderzoek (Assen, the Netherlands).Patients gave written informed consent prior to participation in the study.The use of SD medication was prohibited prior to enrolment for a period of 2 weeks for topical treatments, including dandruff shampoos, 3 weeks for phototherapy and 4 weeks for systemic treatments.In total, 37 patient exhibiting mild-to-moderate SD defined by an Investigator's Global Assessment (IGA) score of 2-3 after verification by a dermatologist were included and assessments performed during a single visit.Due to the heterogeneous presentation of SD and the invasiveness of assessments, assessments are performed on different sites of the face as listed in

K
E Y W O R D S barrier, ceramides, Malassezia, seborrheic dermatitis, staphylococcus Data visualization and statistical testing were performed using Prism 9 (Graphpad Software).Two-way ANOVA, or a mixed effects model in the case of missing data points, was performed using Bonferroni's multiple comparison test in the case of multiple variables and paired t-test in the case of two variables.p-values are denoted as ns: p > 0.05, *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001.Integrative data graphing by principal component analysis (PCA) and min-max radar plotting has been performed through Python version 3.8.0(Python Software Foundation).See Supplementary Material for further details.
erage itch rating scale (23.6 ± 22.5/100) and 5-Domain itch scale rating (11.6 ± 3.2/25).TA B L E 1 Baseline demographics including clinical scoring and patient reported outcomes from the study population.The minimal and maximal values are indicated among their respective scores.

( 67 .
62 AU vs. 49.19AU, p ≤ 0.001, Figure1A).Despite high epidermal disorganization hampering the localization of the dermal-epidermal junction in multiple measurements, the epidermis of lesional skin was significantly thicker compared to non-lesional skin (0.15 mm vs. 0.10 mm, p ≤ 0.001, Figure1B).Superficial roughness of the skin was significantly increased in lesional skin (0.013 AU vs. 0.009 AU, p ≤ 0.001, Figure1C).Higher superficial vascularization was observed in lesional skin compared to non-lesional skin at a shallow skin depth of 0.1-0.25 mm (p ≤ 0.05-0.001,Figure1D) with no significant differences at greater depths.This culminates to an increased average vascularization between 0.1 and 0.25 mm of 0.079 in lesional skin compared to 0.058 in non-lesional skin (p ≤ 0.001, FigureS3).
ing for the bacterial and fungal microbiome, respectively.Neither Shannon indexes showed a significant difference between lesional and non-lesional skin sites (1.55 vs. 1.71, p > 0.24 and 0.58 vs. 0.66, F I G U R E 1 Erythema index as determined by standardized photography (A) and the epidermal thickness (B), superficial roughness (C) and the degree of vascularization at different depths in the epidermis (D) as determined by optical coherence tomography of lesional and nonlesional skin.Epidermal thickness could only be determined in 23 of 37 lesional and 34 of 37 non-lesional measurements.AU, arbitrary units; NS, not significant.p > 0.23, for bacteria and fungi respectively, Figure 2A,C), indicating biodiversity on average did not differ between skin sites.However, Staphylococcus was significantly overrepresented and Cutibacterium significantly underrepresented in lesional skin compared to nonlesional skin (44.05% vs. 19.50% and 23.04% vs. 38.20%,respectively, p ≤ 0.001, Figure2B).The mycobiome proved small with only three genera present over the detection thresholds used (Figure2D), of which over 80% comprised of Malassezia hits in both lesional and non-lesional skin without any significant differences (82.20% vs. 83,52%, p > 0.05).Due to the limitations associated with reliable identification of Malassezia at the species-level using ITS-sequencing, axenic culture plates were taken and subsequent MALDI-TOF MS was performed as a more specific qualitative alternative.Using Malassezia specific protocols successful isolation and identification of 16 from 37 lesional and 18 from 37 non-lesional samples was possible.No clear differences were observed between Cer[NS]:Cer[NP] ratio (r = 0.6474), amount of Cer[NSc34] (r = 0.5170) and degree of unsaturation (r = 0.5920) and a negative correlation against the ceramide chain length (r = −0.6668)(Figure3H-K).

F I G U R E 2 F I G U R E 3
Bacterial Shannon diversity index (A) and composition of the bacterial microbiome (B) by 16 s rRNA sequencing along with the fungal Shannon diversity index (C) and composition of the fungal microbiome (D) by ITS sequencing.The presented genera are filtered for minimal prevalence of 1% over all samples and presented relative to the total amount of microbes detected per analysis.Presence of different Malassezia species per site after isolation with contact plates and subsequent MALDI-TOF analysis (E).None of the samples yielded two or more different isolates.Barrier parameters of lesional compared to non-lesional skin demonstrate impaired barrier function in lesional skin.Functional barrier integrity is measured by trans-epidermal water loss (A).The ceramide profile is depicted after grouping individual ceramides per subclass (B), with the ratio between the abundance of Cer[NS] and Cer[NP] highlighted (C).The abundance of short ceramide species Cer[NSc34] (D) and degree of unsaturation (E) within Cer[NS] is shown.The average carbon chain length (CCL) of the combined sphingosine base and fatty acid tail within the non-Cer[EO] moiety is depicted (F).PCA analysis using all individual detected saturated ceramides yields two distinct population (G).Axes list the percentage of variance explained by the first two principal components, with the proximity of datapoints indicating similarity between samples.Correlations between TEWL and Cer[NS]:Cer[NP] (H), percentage Cer[NSc34] (I), percentage of unsaturation (J) and CCL (K) are shown with a line representing the optimal fit from linear regression analysis and 95% confident interval and Pearson's correlation coefficient.CCL, carbon chain length; MUCer, monounsaturated ceramide; TEWL, Trans-Epidermal Water Loss.
Until now, the limited studies that have demonstrated functional barrier impairment in SD have neglected the lipid compartment F I G U R E 4 Principal component analysis (PCA) of lesional and non-lesional skin using all individual datapoints (A).Axes list the percentage of variance explained by the first two principal components.Integrative visualization of major contributors within the three different axes by radar chart (B).The distance of a datapoint from the center represent the average value per parameter per site compared to the average value per parameter of both sites.AU, arbitrary units; CCL, carbon chain length; MUCer, monounsaturated ceramide; TEWL, Trans-Epidermal Water Loss.

Table S1 -
S3. See Supplementary Material for further details.