We obtained swabs from two body sites (left lateral side of the nose and/or scalp) of healthy volunteers seen at the Skin Research Institute of Singapore, Singapore. 48 subjects were sampled on the side of the nose/aral crease and 44 subjects were sampled on the left temporal scalp. Sampling for both cohorts was performed separately (Supplementary Figure 1). For detailed yeast species characterization, gender and race comparisons from nasal skin swabs, only data from the 37 individuals from four major ethnic groups (10 Chinese, 7 Malay, 10 Indian, and 10 Caucasian) of equal gender distribution were analyzed. Subjects of mixed ethnicity were excluded in ethnically segregated species comparisons. Sampling was performed from December 2017 to March 2018.

Only one body site (left lateral side of the nose) was sampled using identical methods as above for 20 healthy volunteers (10 males, 10 females) seen at the Department of Dermatology at the University Hospital of Zürich, Switzerland as part of an epidemiological survey of healthy skin. For detailed yeast species characterization, only data from the 11 culture positive Swiss individuals were analyzed. Sampling for this cohort was performed in June–July of 2017.

All sampling was performed according the Declaration of Helsinki and approved by the respective ethics committees in Singapore (National University of Singapore, Institutional Review Board, NUS-IRB-237) or Zurich, Switzerland [Cantonal Ethics Commission (KEK), Zurich, KEK-ZH-NR: 2010-0117/0 and the Swiss Ethics Committees on research involving humans (BASEC), BASEC-Nr. 2016-00301]. Participants were between 22 and 65 years of age, having been living in Singapore/canton of Zürich, Switzerland for at least 2 weeks with no overseas travel and judged to be healthy based on the absence of overt skin diseases.

Briefly, sterile diagnostic swabs (COPAN Italia, Brescia, Italy or Fortuna Scientific, Singapore) were moisturized with saline and the whole left lateral side of the nose was swabbed for 10 s. Swabs were then directly inoculated onto modified Leeming Notman agar (mLNA) (Kaneko et al., 2007) plates containing 50 μg/ml of chloramphenicol and 100 μg/ml streptomycin (Sigma-Aldrich, Singapore) by streaking and rolling across the entire surface of the plate up for up to three times to maximize colony yield. Plates were incubated at 32°C for 2 weeks. No significant differences in sampling with different swab brands was observed (data not shown) and choice of swab was made based on availability in the respective geographical regions.

Modified Leeming Notman agar plates were monitored over a course of 1–2 weeks for colony growth. Characteristics such as morphology, texture, color, speed of growth, and phenotype changes were recorded. After 2 weeks, up to 10 colonies of the same phenotype were picked for further culture expansion. Briefly, cultures were lysed in universal DNA extraction buffer (Goldenberger et al., 1995) containing 20 μg/ml of Proteinase K (Sigma-Aldrich, Singapore). The 18S ITS (ITS1-5.8S-ITS2) ribosomal region was amplified and sanger sequenced (Ciardo et al., 2006). Species identification was performed with the recorded species/strain having the highest BLAST score with a minimal match of 97%.

To compare the reproducibility of culture-dependent methods, the right lateral side of the nose (parallel to the side from which cultures were derived) was tape stripped 50 times using a D-Squame^® standard sampling disks (Cuderm^® Cooporation, Dallas, TX, United States). This was performed for four male Singapore subjects, one from each ethnicity. DNA extraction from each tape was performed using the MasterPure^TM Yeast DNA purification kit (Epicentre, United States) as per manufacturer’s instructions combined with bead beating with Lysis Matrix E (MP Biomedicals, Singapore) during the lysis step. Species-specific PCR amplification was performed as described (Zhang et al., 2013).

Amplified products of the ITS regions for each species were cloned into a pCR4.0 vector using TOPO TA Cloning Kit (Life Technologies, Singapore) as per manufacturer’s instructions. Real-time qPCR was carried out with the respective species-specific primers against each respective ITS plasmid standard across seven dilutions over the dynamic range of 3,000,000–3 ITS copies as quantified via Qubit dsDNA HS Assay Kit (Thermofisher Scientific, Singapore). Briefly, a 10 μl reaction mix was prepared containing 1 μl template, 5 μl of Luna Universal qPCR Mastermix (NEB, Singapore), 0.25 μl of each forward and reverse primer (10 μM), and 3.5 μl UltraPure water. Limit of detection (LoD) and limit of quantitation (LoQ) were calculated as described (Armbruster and Pry, 2008).

Positive culture rate was calculated by taking the percentage of the number of subjects whose swabs yielded yeast colonies on mLNA after 2 weeks in culture over the total number of subjects sampled. To compare the number of unique species identified on each subject, an unpaired, two tailed t test was performed. For quantitative comparisons of the diversity of species across different subjects, the Simpson’s Index, Shannon–Wiener Species Diversity Index and Evenness were calculated as follows: Shannon-Wiener Species Diversity Index, H = -SUM[(pi) × ln(pi)] E = H/H_max where, SUM = Summation pi = Number of individuals of species i/total number of samples S = Number of species or species richness H_max = Maximum diversity possible E = Evenness = H/H_max. Simpson’s Diversity Index = 1 - D E_1/D = (1/D)/S Where, D = Simpson’s Index of Diversity S = Sum of numbers data.

Molecular typing was performed for isolates of M. furfur, M. sympodialis, M. globosa, M. restricta, and Malassezia slooffiae using the following loci – (1) Internal transcribed spacer regions (ITS1-5.8S-ITS2), (2) Intergenic Spacer Regions (IGS), (3) D1/D2 26S rRNA regions, (4) ß-tubulin, and (5) Elongation Factor-1 regions (Lorch et al., 2018). The number of strains analyzed was based on the minimum number of strains of each species available to us for analysis if not already available in GenBank. Sequencing of PCR products was performed using the BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems, United States). These sequences were pooled with locus-specific sequences available on the GenBank database¹. Sequence alignments were performed using the Multiple Sequence Alignment tool (Corpet, 1988) and/or the software Clustal Omega². Strain information from non-local isolates was sourced from prior publications (Theelen et al., 2001; Gupta et al., 2004; Cabañes et al., 2005; Jagielski et al., 2014; Lian et al., 2014).

Duplicate loci sequences from multiple strains were identified by ElimDupes³ and removed from further analysis. Phylogenetic tree construction and bootstrap analysis (1000 replicates) was performed using the Bionumerics 7.6 software package (Applied Maths, United States). Neighbor-joining, UPGMA, maximum-parsimony and maximum-likelihood methods were used for phylogenetic tree construction (Saitou and Nei, 1987; Felsenstein, 1988, 1996). Multiple correspondence analysis (MCA) and hierarchical clustering analysis was performed using the “factoMineR” package R-version 3.4.3 (R Development Core Team, Vienna, Austria).

Skin swabs from healthy Singapore subjects yielded a positive yeast culture rate of 87.5% (42 of 48 individuals) from the side of the nose and 95.5% (42 of 44 individuals) from the scalp. All positive cultures from swabs taken from the side of nose contained Malassezia. M. globosa (56.8%, 21 of 37 subjects), M. furfur (48.6%, 18 of 37), and M. restricta (40.5%, 15 of 37) were the dominant species in this Singapore cohort (Figure 1A and Table 1). Other Malassezia species included M. sympodialis 21.6% (8 of 37), M. dermatis 21.6% (8 of 37), and M. slooffiae 13.5% (5 of 37) (Figure 1A). Candida (e.g., Candida parapsilosis, Candida orthopsilosis) and Rhodotorula were cultured from 35.1% (13 of 37) and 13.5% (5 of 37) of the sampled subjects, respectively (Table 1). Fewer species of Malassezia were detected on the scalp, with M. globosa (47.6%, 20 of 42) and M. restricta (23.8%, 10 of 42) being the dominant colonizing species respectively (Figure 1A and Table 1). Other fungi such as Aspergillus, Aureobasidium, Pallidocercospora, Cladosporium, Hortaea, and Trichosporon were also observed sporadically in culture although their presence was not quantified.

The rate of positive culture from skin swabs taken from the side of the nose of healthy Swiss individuals with the same sampling methods and culture conditions was 55.0% (11 of 20), which was lower than the observed positive culture rate in Singapore (87.5%). M. restricta and M. sympodialis were the dominant cultured yeasts (47 and 25% of Swiss Malassezia isolates (Figures 1A,B).

The average Singapore subject carries a higher number of Malassezia species compared to the average Swiss subject (2.03 species vs. 1.55 species, ^∗p < 0.05), with greater species diversity and eveness (Table 2). Variations were observed in the species isolated from different ethnic groups (Figure 1C). M. sympodialis, M. slooffiae, and M. dermatis were selectively absent in some ethnic groups. The Singapore and Swiss Caucasian cohorts showed different species distribution profiles, with M. globosa being absent in the Swiss cohort and M. restricta being two times more commonly isolated on Swiss Caucasian skin compared to Singapore Caucasian skin (Figure 1B). Multiple correspondence analysis (MCA) was performed to identify individuals of similar profile and associations between the presence of different commensal yeast species. Comparison of the Singapore and Swiss cohorts revealed that the Swiss cohort formed a distinct cluster (Figures 2A,B), despite both cohorts having Caucasian individuals of European decent. The Singapore Chinese cohort also formed a distinct cluster. Singapore Malay and Indian cohorts were closely associated and the Singapore Caucasian cohort fell midway between the Swiss Caucasian cohort and the general Singapore cohort (Figure 2B).

No significant differences in detected species numbers were observed when comparing gender and age (p > 0.05, Figure 2B and Supplementary Figures 2, 3) and no Malassezia species was specifically correlated with any one ethnic group (Figure 1C and Supplementary Figures 2–4). Hierarchical clustering showed similar ethnic group and geographical clustering (Figure 2C), although some individuals were observed to fall outside of their ethnic group or geographical cluster (Figure 2C circled). Some categorical factors such as the presence of C. orthopsilosis, M. globosa, M. slooffiae, and M. restricta were observed to be associated with gender (Supplementary Figure 5). However, the inclusion of other commensal yeast species did not significantly change clustering (Supplementary Figure 5).

We compared PCR detection methods and culture-based results from tape strips obtained from opposite sides of the nose for four male volunteers representing each major ethnic group (Chinese, Malay, Indian, Caucasian). PCR methods were able to detect six of the seven tested species of commensal yeasts. M. globosa and M. restricta were the most robustly detected by PCR. The only exception was M. furfur, which was detected by culture only in three of the four subjects but not detected in any of the subjects by PCR (Figure 3). This unlikely to be a result of lower sensitivity as Limit of Detection (LoD) and Limit of Quantitation (LoQ) values for M. furfur primers were established to be the lowest of all five tested species-specific primers sets (Supplementary Table 1). These results suggest that there are limitations in the respective culture-dependent and independent species detection methods, which warrants further study.

Five different loci – internal transcribed spacer region (ITS), 26S ribosomal region (26S), intergenic spacer region (IGS), Elongation Factor alpha 1 (EF-1α) and ß-tubulin were sequenced for six or more strains of M. furfur, M. sympodialis, M restricta, M. globosa, and M. slooffiae (Lorch et al., 2018). This includes reference strains in the CBS database, strains derived from primary isolates in Singapore and Zürich, and sequences derived from the NCBI sequence database. The 26S loci was 99% identical across all species of Malassezia and was not suitable for strain typing (Figure 4A). IGS, EF-1α and ß-tubulin were observed to be potential candidates for strain typing of M. sympodialis, M. globosa, and M. restricta (<99% conserved nucleotides). However, these loci could not be analyzed in all cases, as the primer amplification did not work well for all species despite extensive PCR optimization.

For M. furfur and M. sympodialis, ITS profiles provided the best resolution relative to other loci in their ability to enable the visualization of clear groups or clusters of strains. Based on data describing isolate backgrounds, we observed segregation of disease derived M. furfur isolates (CBS 1878 and CBS 14141) from healthy derived M. furfur isolates (Figure 4B). Other M. furfur isolates from healthy individuals were generally well represented in the international community (Singapore and Zürich), with a small subgroup found mainly in Singapore. We were unable to infer any phylogeographical or healthy/disease isolate clustering on dendrograms derived from ITS strain typing with M. globosa and M. sympodialis (Supplementary Figures 6A,B) due to limited information on the background of database strains (unclear where they were isolated or if they were from healthy or diseased skin). Some Malassezia strains were observed to be inter-individual or community specific, with strains from the same species being found only on certain groups of individuals. A single healthy host was observed to carry up to two to three culturable Malassezia species, often with multiple strains of the same species (Supplementary Figure 6B, Subject 031).