This study was approved by the Institutional Review Board of Yonsei University Severance Hospital (IRB no. 4-2018-0334).

For clinical evaluation of HND patients, a database of approximately 5,007 patients with AD who visited Yonsei University Severance Hospital’s Department of Dermatology in 2011 was reviewed retrospectively (10). A query search of patients’ electronic medical records with keywords “red face syndrome,” “red,” “redface,” and “Head and Neck Dermatitis” in both English and Korean was also performed.

A total of 312 patients were diagnosed with AD at the Department of Dermatology, Severance Hospital, Yonsei University College of Medicine. The diagnosis was made according to the Hanifin and Rajka diagnostic criteria (11). Patients’ age and sex were also noted. The severity of AD was assessed using the eczema area and severity index (EASI). Patients with diffuse erythematous patches on the facial skin were categorized into the HND group, whereas the remaining subjects were categorized to the non-HND group. The patients were categorized according to different age groups in which childhood was defined as age less than 12 years, adolescents were aged between 12 and 18 years, and adults were categorized as those older than 18 years. Sensitization to M. furfur was assessed using CAP immunoassay. Sensitized status to M. furfur was defined by elevation in the levels of IgE specific to M. furfur above 0.70 kU/L (Class 2).

For histological evaluation of HND, a histological database at the Yonsei University Severance Hospital was utilized. A query search of AD patients from 2011 who underwent facial skin biopsy was performed, and five patients were randomly selected from 9 candidates.

Histological analysis of non-HND face specimens was performed through a query search of AD patients who underwent skin biopsy on the face from 2013 for suspected concomitant vitiligo (usually the biopsy is conducted with non-lesional normal skin and lesional skin with vitiligo to compare the melanocyte population). Crude age filtering was performed to age-match AD patients. Among the 10 candidates, five patients were randomly selected for image analysis.

At 200x magnification, the longest distance from the subcorneal level to the basal layer was chosen arbitrarily for epidermal thickness after calibrating the scale bar to pixels. The number of vessels/mm² was counted in the dermis of each slide section within a 100 µm distance from the epidermal–dermal junction.

Immunohistochemical staining was performed using paraffin-embedded sections with antibodies against factor VIII-related antigen (1:100, ab236284, Abcam), stromal cell-derived factor-1-alpha (SDF1-α) (1:100, ab25117, Abcam, Cambridge, United Kingdom), Interleukin-1-beta (IL-1-β) (1:100, ab2105, Abcam), tumor necrosis factor-alpha (TNF-α) (1:50, ab1793, Abcam), transforming growth factor-beta (TGF-β) (1:100, ab66043, Abcam), and vascular endothelial growth factor (VEGF) (1:200, ab1316, Abcam). Staining intensity was determined at 400x magnification at a randomly chosen area of the upper dermis. Images were quantified using ImageJ analysis tools (National Institutes of Health, Bethesda, MA).

To calculate the stained area of the antibody, we converted the original image to an 8-bit grayscale image (ImageJ>Image>8-bit), applied a binary threshold, and calculated the percentage positive for the stained part in the standard image. Quantification was performed relative to the entire selected region. The threshold for each staining was set as the average threshold of multiple immunostaining analyses performed by three independent experimenters.

Primary normal human epidermal keratinocytes were cultured at 37°C in 5% CO₂ in Epilife medium supplemented with human keratinocyte growth supplement (Gibco, USA). Human microvascular endothelial cells (HMVECs) were cultured at 37°C in 5% CO₂ in EBM-2 basal medium supplemented with EBM-2 growth medium (Lonza, USA).

M. furfur (ATCC 12078) was cultured at 30°C on Difco YM agar supplemented with 1% olive oil. S. epidermidis (Staphylococcus epidermidis, ATCC 12228) was cultured at 37°C on Difco tryptic soy agar. C. acnes (Cutibacterium acnes, ATCC 6919) was cultured at 37°C on forced clostridial medium (CM0149; Oxoid) with 2% agar. To induce hypoxia, a BD GasPakTM EZ Pouch was used. All the media were sterilized by autoclaving at 121°C for 15 min.

Organisms were harvested by centrifugation, and the pellet was suspended in the corresponding media. The organisms were heat-killed by incubation at 80°C for 3 min, and then co-cultured with normal human epidermal keratinocytes or human microvascular endothelial cells for 24 h at a density of 1 × 10⁵ cells/mL. To induce allergic environments, recombinant thymic stromal lymphopoietin (TSLP) (50 ng/mL) or IL-4 (50 ng/mL) was used.

The cells were harvested using trypsin-EDTA (0.25%) and centrifuged. Total RNA was extracted using the RNeasy Plus Mini Kit (Qiagen, Germany), following the manufacturer’s instructions. Next, cDNA was generated using a Veriti thermal cycler (Applied Biosystems). Real-time quantitative PCR was performed with cDNA supplemented with the appropriate TaqMan primers using a StepOnePlus PCR system (Applied Biosystems). mRNA expression level was calculated using the 2-ΔΔCT method. The primers used are as follows; VEGF (vascular endothelial growth factor) (Hs00900055_m1), VEGFR (VEGF receptor) (FLT1; Hs01052961_m1), IL-31 (Hs01098710_m1), IL-33 (Hs00369211_m1), TGF-β (Hs00998133_m1), TNF-α (Hs00174128_m1), IL-1β (Hs01555410_m1), and GAPDH (Hs02786624_g1).

The human stratum corneum was collected using D-squame tape (22 mm in diameter; CuDerm, Dallas, Tex) from the facial skin of five subjects from both HND and non-HND groups who provided informed consent. An additional number of five people who did not have any underlying disease history including dermatological conditions were recruited as the control group. Six consecutive D-squame tape strips were collected from each group. The first tape disc was discarded and the remaining tape discs were placed in separate tubes. The tapes were vortexed in 5 mL of methanol for 30 s. The tape debris was removed immediately, and the methanol solution was dried under a nitrogen stream at 30°C.

Total lipid extracts were separated into three fractions using silicic acid column chromatography: neutral lipids (TAG), free fatty acids, and ceramides. Half of the dried extracts were reconstituted in 200 μL of chloroform and loaded into silicic acid columns that were preconditioned with chloroform. After sample loading, each column was washed with 20 mL of chloroform to elute the neutral lipids, 20 mL of chloroform containing 0.2% acetic acid to elute free fatty acids, and 30 mL of methanol to elute the ceramides.

The fatty acid composition of SC lipid fractions was analyzed after derivatization (saponification and methylation) using a gas chromatograph equipped with an SPB-5 capillary column (5% phenol, 30 m, 0.25 mm ID, film thickness: 0.25 μm). The GC operating conditions were as follows: injector temperature, 305°C; detector temperature, 310°C FID; oven temperature, 175°C; 5°C/min; 300°C (20 min); and carrier gas, He.

The TAG and ceramide fractions of SC lipids were further analyzed using LC-MS/MS for profiling (12). Skin lipids were analyzed using a Nexera2 LC system connected to a triple quadruple mass spectrometer (LCMS 8060; Shimadzu, Kyoto, Japan) with a reversed-phase Kinetex C18 column (100 × 2.1 mm, Phenomenex, Torrance, CA, USA). Mobile phases were water/methanol mixture (1:9, v/v) with 10 mM ammonium acetate (A) and isopropanol/methanol mixture (5:5, v/v) with 10 mM ammonium acetate (B). The gradient elution was performed as follows: 0 min (30% of B), 0–15 min (95% of B), 15–20 min (95% of B), and 20–25 min (30% of B). Quantitation was conducted by selected reaction monitoring (SRM) of the [M + H]⁺ or [M + NH₄ ⁺] ion and related product ion for each lipid species. The concentration of each target lipid species was calculated as the ratio of the target analyte to the internal standard (IS) multiplied by the concentration of the IS. Single-point calibrations of each target lipid species were conducted using a selected IS for each lipid class (NS(d18:1/12:0), NdS(d18:0/12:0), NP(t18:0/8:0), AS(d18:1/18:1), AdS(d18:0/12:0), AP(t18:0/6:0), E(18:2)O(16)S(18), E(18:2)O(16)P(18), A(18:1)NS(d18:1/17:0), A(18:1)NS(d18:1/17:0), sphingosine d₉, OP(t18:0/16:0), OP(t18:0/16:0), OP(t18:0/16:0), and TG 45:0(15:0/15:0/15:0) for NS, NdS, NP, AS, AdS, AP, EOS, EOP, 1-O-Acyl-NS, 1-O-Acyl-AS, long chain base (LCB), OS, OP, OH, and TG, respectively).

Lyophilized ceramide (esterified omega-hydroxyacyl-sphingosine; EOS) was purchased from Avanti Polar Lipids (USA). EOS was added to sterilized media at 45°C and placed on a magnetic stirrer. M. furfur and C. acnes were cultured on appropriate culture media discs, and their colonization patterns were observed. The concentrations of ceramide (EOS) were as follows: control (no additional treatment), 2, 5, and 10 μg/mL. Colony area analysis was performed using ImageJ software (13). The observed colony area values were divided by the colony area of the control (no EOS) and compared between different EOS treatments.

Statistical Analysis

The Student’s t-test was used for dependent samples, and the nonparametric Mann–Whitney U test was used for comparison of quantitative values between two groups. The Kruskal–Wallis test was used to compare more than two groups. Tukey’s multiple comparisons test was used after Kruskal-Wallis test to compare each experimental group against each control group. Quantitative data are described as median and range or as mean ± standard deviation. Statistical differences were considered significant if p< 0.05. GraphPad Prism version 9.4.1 for Windows (GraphPad Software, San Diego, CA, USA) and SPSS (version 19.0; SPSS Inc., Chicago, IL, USA) were used to calculate statistical significance.

Among the 5,007 patients with AD, 120 (2.4%) had clinical records indicating HND. The 120 HND and 4,887 non-HND patients were age- and sex-matched using the exact matching method. After eliminating all missing values in the database, 74 HND and 74 non-HND patients were matched for comparison. There were no significant differences in the distribution of age, sex, and disease onset between the HND and non-HND groups ( Table 1 ). Regarding AD phenotypes, there was a higher percentage of extrinsic AD (total serum IgE > 200 IU/mL) in the HND group ( Table 1 ). Interestingly, AD patients with HND showed a higher disease severity, as measured by the EASI score ( Table 1 . p<0.0001) and elevated total IgE ( Table 1 . p-value<0.0001) during the initial clinic visit.

Next, we concentrated on the effect of “fungiome” composition to the HND in AD, especially on M. furfur, which has been associated with dandruff (14, 15), HND in AD (4, 16), and AD itself (17). To elucidate the differences in M. furfur sensitization status between the HND and non-HND groups, 312 patients were enrolled in the analysis. Of the 312 patients with AD, 75 were in their childhood, 61 were in adolescence, and 176 were adults. The average age of the patients was 24.45 ± 14.56, consisting of 169 male and 143 female AD patients.

The average level of IgE specific to M. furfur was 8.57 ± 18.89 kU/L. In the childhood group, the specific IgE level was 1.689 ± 3.46 kU/L, and the levels for the adolescence and adulthood groups were 11.53 ± 22.21 kU/L and 10.48 ± 20.8 kU/L, respectively. The difference between adolescents and adults was not significant, whereas the data were significant for both childhood and adolescence and between childhood and adulthood ( Figure 1A ; both p<0.001) The sex-specific level of specific IgE to M. furfur was also assessed, and it was found to be 8.53 ± 19.62 kU/L in males and 8.63 ± 18.05 kU/L in females ( Figure 1B ). The difference between the two groups was not statistically significant.

The association between HND and sensitization to M. furfur was evaluated. Interestingly, the levels of IgE specific to M. furfur was significantly higher in the patients with HND ( Figure 1C ; 20.61 ± 26.04 vs. 4.07 ± 12.80 kU/L, p<0.0001). In addition to laboratory characteristics, the association between the presence of HND and clinical severity, which was measured using the EASI score, revealed that clinical severity was significantly higher in HND patients ( Figure 1D ; 24.12 ± 15.90 vs. 14.28 ± 11.50, p<0.0001).

The patients were categorized according to their sensitization to M. furfur. Of the 312 patients, 142 were sensitized and the remaining 170 were non-sensitized. EASI scores of the M. furfur sensitized AD patients were significantly higher than those of the non-sensitized group ( Figure 1E ; 21.91 ± 14.59 vs. 12.83 ± 11.07, p<0.0001). These data indicate that sensitization to M. furfur not only affects the risk of HND in AD but might also contribute to the overall severity of AD.

To investigate histological differences between the HND and non-HND groups, skin biopsy specimens were examined by hematoxylin and eosin staining. In HND pathology, a general trend of hyperkeratosis, acanthosis, and parakeratosis was observed, with dense inflammatory cells surrounding the increased vasculature ( Figure 2A ). The average epidermal thickness in HND group (412.5(292.8–532.1) μm) was significantly higher than that of the non-HND group (179.6(140.9–205.3) μm) ( Figure 2B ; p<0.05).

Increased vascularity in HND patients was confirmed by immunohistochemical staining for factor VIII-related antigens ( Figure 2C ). Average vessel count for HND group was 15 (12–19) with a concurrent average vessel area of 221.4 (159.8–273.4) µm². The average vessel count in non-HND histology was 4.5 (3–6) with an average vessel area of 101.0 (78.9–177.3) µm². The differences in vessel counts and average vessel sizes were statistically significant ( Figures 2D, E ; both p<0.05).

Next, immunohistochemical staining of pro-inflammatory cytokines and chemokines was performed to confirm the presence of inflammatory mediators in HND lesions. The evaluated molecules included TNF-α, TGF-β, IL-1β, SDF1-α, and VEGF. As a result of IHC staining, all molecules showed significantly higher signal in HND group, confirming that the HND patients’ facial skin show intense inflammatory circumstance than non-HND group. ( Figures 3A–F ; SDF1-α, p<0.01; IL-1β, p<0.05; TGF-β, p<0.01, TNF-α, p<0.01; VEGF, p<0.001),

Malassezia spp. colonization/abundance is known to be linked with AD pathogenesis, as AD patients often exhibit hypersensitization to Malassezia spp. with higher Malassezia-specific IgE levels (18–21). These findings from previous studies indicate the possibility of a positive association between Malassezia spp. abundance and hypersensitivity.

Thus, the effects of M. furfur colonization on skin were evaluated in vitro ( Figure 4A ). First, keratinocytes (normal human epidermal keratinocytes) were cultured with either M. furfur or S. epidermidis, and the expression levels of VEGF, VEGFR, IL-31, and IL-33 were evaluated. To create an environment similar to that of AD, TSLP and IL-4 were used. As shown in Figure 4B , the expression levels of VEGF, VEGFR, IL-31, and IL-33 were significantly increased when the cells were co-cultured with M. furfur and treated with TSLP (VEGF, p<0.01; VEGFR, p<0.01, IL-31; p<0.05; IL-33, p<0.05). No significant increase in expression was observed in keratinocytes co-cultured with S. epidermidis. The expression levels did not increase when cells were not treated with TSLP. As shown in Figure 4C , similar results were observed when cells were treated with IL-4, except for VEGF (VEGF, p = 0.1244; VEGFR, p<0.05, IL-31; p<0.05, IL-33; p<0.05). These results suggest that M. furfur induces the production of cytokines related to AD, further exacerbating disease severity, and that the expression of VEGF, which induces angiogenesis, may contribute to the erythema observed in HND of AD.

Figure 4D depicts the effect of M. furfur on endothelial cells. Human microvascular endothelial cells were co-cultured with M. furfur and treated with IL-4 using the same protocol as that used for keratinocytes. The increase in the expression levels of VEGFR, TGF-β, TNF-α, and IL-1β was most significant when the endothelial cells were co-cultured with M. furfur and treated with IL-4 ( Figure 4D ; VEGFR, p<0.05; TGF-β, p<0.001; TNF-α, p<0.001, IL-1β; p<0.001). However, this tendency was not observed in endothelial cells co-cultured with S. epidermidis. These results imply that M. furfur also affects endothelial cells, inducing the production of pro-inflammatory cytokines, such as TGF-β, TNF-α, and IL-1β, as well as VEGFR, which further promote vascular proliferation.

Malassezia spp. are well known for their ability to produce lipases (22) that break down cutaneous lipid components into unsaturated fatty acids, oleic acid, and arachidonic acid (23, 24). Therefore, we investigated the lipid composition of the stratum corneum, which is essential for the maintenance of skin barrier function.

Tape stripping was performed on the facial skin of patients with HND and those without HND, with five subjects in each group. Five additional subjects without any underlying disease were also tested as controls. Patients with HND had significantly higher erythema index and transepidermal water loss levels in the facial area, which supported the presence of HND ( Table 2 ). Among the various ceramides from the stratum corneum, phytoceramide NP (nonhydroxyacyl-phytosphingosine), AP (α-hydroxyacyl-phytosphingosine), and EO (esterified omegahydroxyacyl) type ceramides (EOS (esterified omega-hydroxyacyl-sphingosine), and EOP (esterified omega-hydroxyacyl-phytosphingosine)) were found to be significantly reduced in the HND group when compared to the non-HND and control groups ( Figure 5A , NP; p<0.01, AP; p<0.05, EOS; p<0.01, EOP; p<0.01). However, when the non-HND group was compared to the healthy controls, no significant differences were found. Next, individual lipid components belonging to each ceramide type that varied significantly were compared. In the case of NP- and AP-type ceramides, there was individual lipids with no significant difference between non-HND group and healthy controls. Some individual lipid components showed statistical differences between HND group and non-HND/healthy controls ( Figure 5B , t18:0/24:0; p<0.05, t18:0/25:0; p<0.05, t18:0/26:0; p<0.05, t18:0/27:0; p<0.01, t18:0/28:0; p<0.01, Figure 5C , t18:0/24:0; p<0.05, t18:0/25:0; p<0.01, t18:0/26:0; p<0.01). However, all individual lipid components belonging to the EO-type ceramides exhibited statistically significant differences between HND group and non-HND/healthy controls while non-HND group and healthy controls showed no significant difference ( Figure 5D , E(18:2)O(28)S(18); p<0.01, E(18:2)O(29)S(18); p<0.01, E(18:2)O(30)S(18); p<0.01, E(18:2)O(31)S(18); p<0.01, E(18:2)O(32)S(18); p<0.01, Figure 5E , E(18:2)O(29)P(18); p<0.05, E(18:2)O(30)P(18); p<0.01, E(18:2)O(31)P(18); p<0.01, E(18:2)O(32)P(18); p<0.001, E(18:2)O(33)P(18); p<0.01, E(18:2)O(34)P(18); p<0.01).

In the above analysis, the decrease ratio between Non-HND/Control to HND group was more prominent in EO type ceramides (EOS, EOP) than that of other types. Among the four EO type ceramides, EOS ceramide is regarded as the main ceramide component of the epidermis (25). In the epidermis, EOS is converted to ω-hydroxyacyl-sphingosine (OS) ceramide. The converted OS is attached to the cornified envelope to maintain the skin barrier function (26–29). Based on the above results and that of previous studies, OS ceramides were analyzed after the extraction of hydrolyzed unbound ceramides. As a result, OS ceramides, which are the most abundant in the stratum corneum, were found to be significantly decreased in the HND group when compared to that in the non-HND and control groups ( Figure 6A , OS; p<0.01, OP; p = 0.3833, OH; p = 0.2839). In addition, when the individual lipid components were compared among the three groups, statistically significant differences were found mostly in those belonging to OS-type ceramides ( Figure 6B , t18:0/31:0; p = 0.2516, t18:0/32:0; p<0.01, t18:0/34:0; p = 0.538, Figure 6C , t16:1/36:0; p = 0.2839, Figure 6D , d17:1/30:0; p<0.01, d18:1/30:0; p<0.05, d18:1/31:0; p<0.01, d18:1/32:0; p<0.001, d20:1/30:0; p<0.01, d20:1/31:0; p<0.01, d20:1/32:0; p<0.01). These data indicate that ceramide levels, especially EOS-type ceramides, are downregulated in HND. Considering the ability of Malassezia spp. to produce lipases, it can be inferred that the Malassezia spp. colonization might have been involved in the decreased stratum corneum ceramide levels of HND patients.

Finally, we investigated whether the restoration of OS-type ceramides could reduce the colonization of Malassezia spp. in vitro ( Figure 7A ). To assess the effect of ceramide on the growth of M. furfur, 0, 2, 5, and 10  μ g of EOS (esterified ω-hydroxyacyl-sphingosine) ceramide were added to the growth medium of M. furfur. C. acnes, the most common bacteria in the normal flora of facial skin, was used as a control, and the same experimental protocol was applied. For C. acnes, the total area of colonization did not vary with the concentrations of ceramide ( Figure 7B ). In contrast, in the case of M. furfur, as the concentration of ceramide increased, the total area of colonization of M. furfur decreased significantly ( Figure 7C , Tukey’s multiple comparisons test; control vs. 2μg/mL; p< 0.001, control vs. 5μg/mL; p< 0.001, control vs. 10 μg/mL; p< 0.001). These data indicate that ceramide, especially EOS ceramide, inhibits the growth of M. furfur.