Introduction

With the development of DNA-sequencing technologies, such as next-generation sequencing, human microbiome studies have revealed the characteristic features of not only internal organs, such as the gut and mouth, but also external organs, such as the skin1,2,3,4. The human skin is the largest organ in the body, and its surface is colonized by a wide variety of bacteria5,6,7,8,9. Some of these bacteria have a symbiotic relationship with their host and can provide protection against invasion by pathogenic microorganisms10,11,12. However, an imbalance between commensal and pathogenic bacteria, known as dysbiosis, can cause skin disorders13. Human skin samples for bacterial community structure analysis are commonly collected using a swabbing method4,5,14,15. As an alternative sample collection method, tape stripping16, scraping17, and biopsy18,19,20 have also been used in some bacterial community structure studies of human skin.

Relationships have been reported between scalp bacterial community structure and human health. Several studies on the scalp microbiome have revealed an association between dandruff and bacterial community structure21,22,23,24. Additionally, the overall scalp microbiome composition notably differs between normal and diseased groups in other scalp diseases, such as seborrheic dermatitis25,26 and alopecia areata27,28,29. Scalp sample collection for microbiome analysis is mostly performed using the swabbing method21,22,23,24,25,27,29, although tape stripping26,30 and punch biopsy28 are also used.

Scalp hair, which is part of the skin, is also composed of site-specific bacterial community structures5,31,32. We have previously reported a specific feature of the bacterial community structure on scalp hair and the scalp33,34,35,36,37. We analyzed this structure on scalp hair shafts and hair roots and found that the major bacteria on scalp hair shafts were indigenous and derived from hair roots, with a similar structure from the tip to the base of the long scalp hair shaft34. As for bacterial localization on scalp hair, we observed a certain number of bacterial cells on the surface of scalp hair using SEM, which corresponded well with the bacterial copy numbers quantified by qPCR34. Furthermore, the bacterial community structure on scalp hair was more similar within than between individuals and was specific to each individual33. We reported that this structure may have been affected by host intrinsic factors (gender), as well as by extrinsic factors (hair styling and chemical treatments)36. Furthermore, regarding the relationship between the bacterial community structure on human scalp hair and scalps, it was found that major bacterial species were commonly present at both sites, but the bacterial community structure on hair was specific and distinguishable from that on the scalp35. Although the bacterial community structure on scalp hair may also be associated with scalp health and disease due to the relationship between the scalp and scalp hair microbiome, previous studies—except for the study using the swab method5—have used plucked or cut hair for microbiome analysis. Non-invasive swabbing methods have commonly been used for skin microbiome analysis, although no reports have compared the outcomes of these methods for bacterial community structure analysis of scalp hair.

The aim of this study was to compare bacterial community structure differences between scalp hair and scalp samples collected using non-invasive swabbing and cutting/removal of scalp hair through 16S rRNA gene sequencing.

Results

Alpha diversity of the bacterial community structures

Figure 1 shows the two alpha diversity indices: observed OTUs and the Shannon index. The observed OTU numbers and Shannon index on hair shaft, hair swab, and scalp swab samples were 25.4 ± 5.3 and 2.3 ± 0.3, 29.4 ± 10.0 and 2.5 ± 0.4, and 17.3 ± 4.1 and 1.9 ± 0.3, respectively. Both indices were significantly higher (p < 0.0001) in hair shaft and hair swab samples than in scalp swab samples. Notably, both indices were the highest in hair swab samples, although there were no significant differences between hair shafts and hair swabs.

Fig. 1
figure 1

Comparison of alpha diversity of bacterial community structures on the hair shaft, hair swab, and scalp swab, based on (a) observed OTUs and (b) Shannon’s index. The values are obtained from a clustering of 1,000 reads per sample. The different letters mean that there are significant differences (Kruskal–Wallis test, p < 0.01).

Full size image

Bacterial community structures at the phylum level

Three major phyla—Actinomycetota, Pseudomonadota, and Bacillota—were commonly found in all samples, and their total abundances were 99.0, 98.8, and 99.5% in hair shaft, hair swab, and scalp swab samples, respectively (Fig. 2). Actinomycetota was the predominant phylum in all sample groups, with significant differences between hair and scalp swab samples (p < 0.01). Pseudomonadota abundance was significantly higher (p < 0.01) on hair shaft (45.8 ± 15.6%) and hair swab (42.0 ± 14.2%) samples than on scalp swab samples (14.9 ± 12.8%). Conversely, Bacillota abundance was significantly lower (p < 0.01) on hair shaft (6.6 ± 4.9%) and hair swab (12.8 ± 11.8%) samples than on scalp swab samples (29.3 ± 20.8%). There were no significant differences in the relative abundance of the three major phyla between hair shafts and hair swabs.

Fig. 2
figure 2

Average relative abundances of the major phyla in bacterial community structures the on hair shaft, hair swab, and scalp swab. The different letters mean that there are significant differences of each bacterial phyla between the sample groups (Kruskal–Wallis test, p < 0.01).

Full size image

Bacterial community structures at the OTU level

Based on the results obtained from all the individuals, we assigned the top 15 most abundant OTU sequences to the most closely related bacterial species (Fig. 3). Of these, the five major OTUs (Phylum, pairwise similarity) showing > 5% abundance in any of the sample groups were OTU2 related to Cutibacterium acnes (Actinomycetota, 99.7%), OTU3 related to Pseudomonas allii (Pseudomonadota, 99.7%), OTU4 related to Staphylococcus epidermidis (Bacillota, 99.7%), OTU5 related to Lawsonella clevelandensis (Actinomycetota, 99.7%) and OTU7 related to Curvibacter gracilis (Pseudomonadota, 99.7%).

Fig. 3
figure 3

Phylogenetic tree and abundance heatmap of the 15 major OTUs in 12 individuals. The percentages color mapped represent the relative abundance of the OTU from each site sampled.

Full size image

The relative abundance of the five major OTUs present in each sample group are shown in Fig. 4. OTU2 related C. acnes was the most predominant species in all sample groups. Relative abundance was not significantly different among the sample groups (Fig. 4a). The relative abundance of OTU3 related to P. allii and OTU7 related to C. gracilis were significantly higher (p < 0.01) in hair-related samples than in scalp swab samples (Fig. 5b and e). In contrast, the relative abundance of OTU4 related to S. epidermidis and OTU5 related to L. clevelandensis were significantly lower (p < 0.01) in hair-related samples than in scalp swab samples (Fig. 4c and d). Similar to the results observed at the phylum level, there were no significant differences in the relative abundances of the five major OTUs between hair shaft and hair swab samples.

Fig. 4
figure 4

Comparison of the relative abundance of the five major OTUs between the scalp swab, hair swab, and hair shaft. Relative abundance of (a) OTU2 related to Cutibacterium acnes, (b) OTU3 related to Pseudomonas allii, (c) OTU4 related to Staphylococcus epidermidis, (d) OTU5 related to Lawsonella clevelandensis, and (e) OTU7 related to Curvibacter gracillis. The different letters mean that there are significant differences of each bacterial phyla between the sample groups (Kruskal–Wallis test, p < 0.01).

Full size image
Fig. 5
figure 5

Beta diversity of bacterial community structures on the hair shaft, hair swab, and scalp swab in 108 samples from 12 volunteers and their biplot analysis at phylum level. The PCoA is based on the weighted UniFrac distance. The PC1 axis is associated with abundance of Actinomycetota, while the PC2 axis is associated with abundance of Pseudomonadota.

Full size image

Beta diversity of the bacterial community structures

Figure 5 shows the beta diversity of the bacterial community structure in hair shaft, hair swab, and scalp swab samples obtained by principal coordinate analysis (PCoA) based on weighted UniFrac distances. The plots were grouped into two clusters, hair-related samples and scalp swab samples, although some were positioned in similar areas. This clustering was also confirmed within individual samples, except for some individuals (Fig. S1). We further performed a biplot analysis at the phylum level, and the results corresponded well with the relative abundance of major phyla in the bacterial community structure in each sample group (Fig. 1). The Actinomycetota and Bacillota biplots were was located in the middle of the two clusters and in the scalp swab cluster, respectively (Fig. 5). In contrast, the Pseudomonadota biplot was located in a hair-related cluster.

The clustering tendency of the PCoA plot for each sample group was confirmed by comparing weighted UniFrac distances (Fig. 6). The distances within hair-related samples (hair swab vs hair swab, hair swab vs hair shaft, and hair shaft vs hair shaft) were significantly lower (p < 0.01) than those between hair-related samples and scalp swab samples (hair shaft vs scalp swab and hair swab vs scalp swab). This result corresponded well with the comparison of weighted UniFrac distances within individual samples, except for some individuals (Fig. S2).

Fig. 6
figure 6

Comparison of weighted UniFrac distances within hair swab samples and between scalp swab or hair shaft samples. The different letters means that there are significant differences of each bacterial phyla between the sample groups (Kruskal–Wallis test, p < 0.01).

Full size image

Finally, to analyze the contribution of the PCoA plots in each sample group, weighted UniFrac distances were calculated using Adonis (Table 1). The sampling site-based category was statistically significant and affected the contribution to the variance (p < 0.001). In contrast, the hair sampling method-based category was not statistically significant.

Table 1 Effect sizes (R2) calculated using the weighted UniFrac distance.
Full size table

Discussion

Our previous study suggested that bacterial community structures on the human scalp hair shaft are formed from both hair-specific and scalp skin-derived bacteria and are distinguishable from other human skin microbiomes35. We verified whether non-invasive swab sampling—a commonly used human skin microbiome analysis—could reflect the bacterial community structure on scalp hair in concordance with cutting/removal of scalp hair, using 16S rRNA gene sequencing. Scalp samples were also analyzed to verify differences between swabs. We identified three bacterial community structure characteristics using different sampling methods. Firstly, there were no significant differences in the alpha diversity indices between hair shaft and hair swab samples. Secondly, the relative abundances of the three major phyla and five major OTUs were not significantly different between hair shaft and hair swab samples. Third, the PCoA plots based on weighted UniFrac distances were grouped into two clusters: hair-related samples and scalp samples, and there were significant differences only between hair-related samples and scalp swab samples. We further discuss each result individually below.

First, we compared two alpha diversity indices in each sample group and found that hair-related samples—hair shafts and hair swabs—had significantly higher indices than did scalp swab samples, while there were no significant differences between those of hair shaft and hair swab samples (Fig. 1). Generally, variations in the cutaneous microbiota of healthy volunteers—including sebaceous, dry, and moist environments—are associated with skin ecological zones6. Skin physiological conditions affect the formation of skin bacterial communities; thus, the alpha diversity of the bacterial community structure on sebaceous skin is lower than that on dry skin sites5,14,23. The higher hydrophobicity of scalp hair causes higher diversity compared with a sebaceous scalp. Hair bacteria are present on the scalp hair surface, as quantified by qPCR and SEM observations34, and bacterial community structure distribution on scalp hair shafts within individuals is relatively steady, even when the scalp hair sampling sites were different37. Therefore, swab sampling and hair shaft sampling methods can collect similar bacterial community structures on scalp hair in terms of alpha diversity, since hair-attached bacteria are present on the hair surface and their diversity is constant at any hair site.

Second, we identified three major common phyla in all sample groups: Actinomycetota, Pseudomonadota, and Bacillota (Fig. 2). For a more detailed analysis of the bacterial community structure, we extracted the top 15 most abundant OTU sequences and compared the relative abundances of the five major OTUs (Fig. 3 and 4). Similar to the alpha diversity comparison, the four major OTUs—except for OTU2 related to C. acnes—were significantly different between hair-related samples and scalp swabs, whereas there were no significant differences between hair shafts and hair swabs. These major phyla on the hair shaft and scalp are similar to those described in our previous study35. In addition, the four major OTUs—except OTU7, related to C. gracilis, which was isolated from well water38—were consistent with previous reports34,35. Although our study showed that Pseudomonas was more abundant, and Staphylococcus was less abundant, on scalp hair, some studies have shown that Corynebacteriaceae, Tissierellaceae, and Staphylococcus are the major bacterial families or genera of bacterial community structures on scalp hair31,32. Kerk et al.39 reported that P. aeruginosa colonizes and adheres to the scalp hair shaft when cultured on these shafts, whereas the growth of S. epidermidis was inhibited. In addition, hair-derived antimicrobial proteins or peptides have been identified in scalp hair shafts that may inhibit the growth of S. epidermidis40,41. Although Cutibacterium is known to carry genes for biotin biosynthesis, an essential nutrient for scalp hair growth24,42,43, Pseudomonas—the second major bacterial genus on hair—is not known to play a key role, and further detailed studies are required to clarify the bacterial ecology of scalp hair.

Third, we analyzed beta diversity and biplots at the phylum level to compare bacterial community structures among each sample group (Fig. 5). It was confirmed that the plots were grouped into two clusters: hair-related samples and scalp swab samples. In addition, the phylum Pseudomonadota contributed to the formation of a bacterial community structure specific to hair-related samples, while the phylum Bacillota contributed to that on the scalp. UniFrac analysis revealed significant differences between hair-related samples and scalp swab samples, but not between hair shafts and hair swabs (Fig. 6). In addition, we confirmed the contribution to the variance of weighted UniFrac distance using Adonis, and the sampling site category was statistically significant, but not the hair sampling method category (Table 1). Grice et al.44 reported that the bacterial community structure of the inner elbow of healthy human subjects obtained by swabs, scrapes, and biopsies had bacterial compositions comparable to those of all the employed sampling methods. Moreover, Ogai et al.16 reported that the bacterial community structure on the back skin of healthy young participants collected using tape stripping was comparable to that collected using the swabbing method. Even in lesional skin, such as diabetic foot ulcers, there were no significant differences in the overall bacterial communities recovered from swabs compared to tissue scrapings17. Prast-Nielsen et al.18 recently reported that the bacterial community structure obtained via skin biopsies and swabs revealed stark differences compared to that of normal skin. In lesional skin, such as psoriasis20 and diabetic foot ulcers19, different sampling techniques, such as swabs and biopsies, resulted in different bacterial community profiles. It is important to note that swabbing and tape stripping collected surface microbiota, whereas skin biopsies collected the full skin layer microbiota. Lange-Asschenfeldt et al.45 reported that 85% of skin surface bacteria were found within the corneocyte layers, and ~ 25% of the cutaneous bacterial population was localized within the hair follicles. Nakatsuji et al.46 also reported that different microbiota may be present in the dermis, epidermis, and superficial adipose tissues. Swab and hair shaft sampling methods can be used to collect similar bacterial community structures on scalp hair, as almost all hair bacteria are present on the hair surface.

In conclusion, the results here suggest that scalp hair bacteria collected using non-invasive swab sampling were comparable to those collected cutting/removal of scalp hair. While the focus of this study is the bacteriome, it is important to note that the fungal mycobiome plays an integral role in shaping the total scalp microbiome. Fungal data was not collected within the scope of this study. The data shared here does not take into consideration the metabolic activity of the bacteria identified, rather just communicates the relative abundance among the total bacterial load47. Also, despite the many advantages of non-invasive swabbing methodologies, one is unable to determine the role of the follicular microbiome in re-colonizing the scalp and hair surface48. Nevertheless, there are some limitations to this study, the establishment of a non-invasive hair sampling method for the analysis of bacterial community structure is expected to advance research on the relationship between hair bacteria and scalp diseases, such as dandruff and androgenetic alopecia. Recently, we evaluated the effect of bacteria on human keratinocyte cellular activity and reported that some predominant bacteria in scalp hair enhanced the expression of several longevity genes49. Further studies on isolating hair-specific bacteria to investigate the interactions between scalp hair and bacteria are currently underway.

Materials and methods

Ethics approval and consent to participate

The study was conducted in accordance with the Helsinki Declaration on this subject. On recruitment, participants signed a document in which they consented to the confidential treatment of their personal information and consented to the publication of anonymized data. The study protocol and the informed consent form were reviewed and approved by the Research Ethics Board of the Tokyo University of Agriculture (Authorized No. 2211).

Samples and collection

Scalp hair shaft samples were collected from 12 healthy Japanese and Taiwanese adults (six females and six males), aged 21 to 45 years, who consented to participate in this study. No volunteer was taking any medication during the experimental period. Hair shafts, hair swabs, and scalp swabs were collected using nitrile gloves, and hair shaft samples were cut using sterile scissors. All samples of hair and scalp were taken from the top region of the scalp. Hair shaft samples were cut approximately 5 mm away from the scalp, and one full length of hair was used in the extraction for analysis. After sampling, the hairs were chopped into lengths of ~ 5 mm with scissors and placed in plastic microtubes. Hair swab samples were directly taken from the hair using cotton swabs (Mentip for hospitals, Nihon Menbou Corporation, Saitama, Japan) and pre-moistened with 50 μL of sterile distilled water. Scalp swab samples were collected directly from the head crown using cotton swabs by rubbed them onto the scalp surface (between hair strands) to cover a total surface area of 2.5 cm2. The head of each swab was removed from the handle and placed in an Eppendorf tube.

Extraction of bacterial DNA

Bacterial DNA was extracted using the NucleoSpin® Tissue kit (MACHEREY–NAGEL, Düren, Germany) according to the manufacturer’s instructions, with a slight modification. First, hair shaft, hair swab, and scalp swab samples were immersed in 100 μL lysozyme solution (20 mg/ml lysozyme derived from egg white [Wako Pure Chemical Industries, Osaka, Japan] in 20 mM Tris–HCl and 0.2 mM EDTA, pH 8.0) for 30 min at 37 °C, as previously reported34, and the DNA extracts obtained (100 μL) were stored at − 20 °C until use.

Analysis of bacterial community structures using 16S rRNA gene sequencing

To analyze the bacterial community structures using the MiSeq™ platform (Illumina Inc., CA, USA), a three-step PCR method was performed using the extracted DNA samples. In the first-step PCR amplification, a universal primer set for the V4 region of the bacterial 16S rRNA gene (515F, 5ʹ-GTG CCA GCM GCC GCG GTA A-3ʹ and 806R, 5ʹ-GGA CTA CHV GGG TWT CTA AT-3ʹ)50 was used. The 25 μL reaction mixture consisted of 12.5 μL of Kapa HiFi HotStart Ready Mix (Kapa Biosystems Inc., Wilmington, MA, USA), 0.5 μL of each primer (10 pM), and 11.5 μL of extracted bacterial DNA. The amplification program included an initial denaturation step at 95 °C for 3 min, followed by 40 cycles of denaturation at 98 °C for 30 s, annealing at 56 °C for 30 s, and elongation at 72 °C for 30 s. After electrophoresis through a 1.5% (w/v) agarose gel, the targeted bands were excised from the gel with sterilized cutters, and the DNA extracted using the FastGene® Gel/PCR Extraction Kit (NIPPON Genetics Co., Ltd., Tokyo, Japan), according to the manufacturer’s instructions. The DNA concentration was measured using a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). To prepare for 16S rRNA amplicon sequencing using MiSeq sequencing, templates are given tail, adapter, and index sequences in a two-step PCR; therefore, long-tailed primers are required, making amplification difficult. We were unable to perform direct amplification using a two-step PCR, probably because the amount of bacterial DNA obtained from the 3 cm hair shafts was very low. Therefore, we first performed a PCR using a universal primer set without any additional sequences. Consequently, we successfully obtained sufficient template fragments with a minimum number of reaction cycles.

For the second-step PCR, a universal primer set for the V4 region of the bacterial 16S rRNA gene and tailed sequences for MiSeq sequencing were used (1-515F, 5ʹ- TCG TCG GCA GCG TCA GAT GTG TAT AAG AGA CAG GTG CCA GCM GCC GCG GTA A-3ʹ and 1-806R, 5ʹ-GTC TCG TGG GCT CGG AGA TGT GTA TAA GAG ACA GGG ACT ACH VGG GTW TCT AAT-3ʹ)51. Although it has been reported that this primer set poorly amplifies Propionibacterium in human skin52, the results here showed good amplification of the predominant species in hair, Cutibacterium acnes (previously named Propionibacterium acnes). The 25 μL reaction mixture consisted of 1.0 µL of each primer (5 µM), which was heat-shocked at 95 °C for 5 min, 12.5 µL of Kapa HiFi HotStart Ready Mix, 12.5 ng of DNA obtained from the first-step PCR amplicon, and sterilized ultrapure water. The amplification program included an initial denaturation step at 95 °C for 3 min, followed by 20 cycles of denaturation at 98 °C for 30 s, annealing at 55 °C for 30 s, and elongation at 72 °C for 30 s. The PCR products were purified using the FastGene® Gel/PCR extraction kit according to the manufacturer’s instructions.

For the third-step PCR, a primer set with flow cell adapter sequences, index sequences, and tailed sequences was used (forward primer, 5ʹ-AAT GAT ACG GCG ACC ACC GAG ATC TAC AC-Index sequence-TCG TCG GCA GCG TC-3ʹ and reverse primer, 5ʹ-CAA GCA GAA GAC GGC ATA CGA GAT-Index sequence-GTC TCG TGG GCT CGG-3ʹ). The third-step PCR mixture (25 µL) comprised 12.5 µL of Kapa HiFi HotStart Ready Mix, 0.5 µL of each primer (10 pM), and 11.5 µL of the second-step PCR amplicon. The amplification program included an initial denaturation step at 95 °C for 3 min, followed by 8 cycles of denaturation at 98 °C for 30 s, annealing at 55 °C for 30 s, and elongation at 72 °C for 30 s. After electrophoresis in a 1.5% (w/v) agarose gel, the target bands were excised with sterilized cutters, and the DNA extracted using the FastGene® Gel/PCR extraction kit, as described above. The DNA concentrations of the third-step PCR amplicons were quantified using the Qubit dsDNA HS assay kit (Thermo Fisher Scientific Inc., USA) according to the manufacturer’s instructions. Purified PCR products from each sample were mixed, denatured, and sequenced with a MiSeq System (Illumina Inc., USA) using the MiSeq reagent kit v3 (300 bp × 2 cycles with paired end reads; Illumina, Inc., USA) according to the manufacturer’s instructions. Good’s coverage values (> 95%) were obtained for all hair samples using the DNA extraction kit and PCR conditions described above. Good’s coverage values were estimated using QIIME™ 1.9.1 software53.

Bioinformatics and statistical analysis

The index and universal sequences of each read were checked and reads with complete index sequences were selected as valid. USEARCH V8.1.1861 software54 was used to merge paired-end reads and remove chimeric sequences. After the chimeric check, the reads were grouped into operational taxonomic units (OTUs) with > 97% similarity. Clostridium and Escherichia were filtered from the OTU tables for further analysis because they are frequently used in our laboratory and have rarely been detected in previous reports34,35,36. Alpha diversity (observed OTUs and Shannon’s index) was evaluated at a 1% OTU distance using the QIIME™ software package53. For taxonomy-based analyses, the representative sequences of each OTU were analyzed using the EzBioCloud platform55. Statistical analysis of the bacterial community structure was performed using the Kruskal–Wallis test in XLSTAT software ver. 2014 (http://www.xlstat.com/en/).