The upper buttock is a standard location for invasive procedures to study skin injury in human subjects 11. As this body site was not included in previous microbiome studies 18, we first analyzed its microbiome composition for comparison with other known locations. We obtained samples from five healthy volunteers in order to analyze the microbial composition of the upper buttock skin, the forehead (sebaceous environment) and from two moist environments of the body (armpit and inner elbow). In total, 116,291 bacterial 16S rRNA sequences were analyzed (Additional file 1). Hierarchical clustering using weighted UniFrac as distance measure (details in the Materials and methods) separates samples from different body environments into three groups (indicated by the numbers 1 to 3 in Figure 1a), with all armpit samples in group 1, 4 out of 5 forehead samples in group 2, and 4 out 5 samples from the upper buttock in group 3 (Figure 1a). Notably, the samples of the inner elbow did not cluster together and appeared divided over the three groups. The phylogenetic composition is distinct for upper buttock skin compared to the other three body sites with the exception of individual HV2 (Figure 1a; Additional file 2). We found that the upper buttock had the largest bacterial diversity of the analyzed sites: rarefaction curves show that the phylogenetic diversity observed in different body sites is highest in upper buttock skin for this limited number of samples (Figure 1b). In the moist region of the armpit we identified a high proportion of reads assigned to the Firmicutes phylum (72.4%) and a lower fraction of reads assigned to the Actinobacteria phylum (27.2%), whereas the sebaceous environment of the forehead was dominated by the Actinobacteria phylum (90.7%). At other locations of the skin the relative abundances of the phyla Firmicutes and Actinobacteria were differently distributed (respectively 30.7% versus 65.7% on the inner elbow, and 41.3% versus 46.4% on the upper buttock). Furthermore, the higher microbial diversity on upper buttock skin was reflected by the detection of other, relatively low abundant phyla, such as Proteobacteria (10.3%), Bacteroidetes (1.3%), Cyanobacteria (0.1%) and Acidobacteria (< 0.1%).

To investigate the composition of microbial communities in different stratum corneum layers we sampled the skin of 12 healthy volunteers at different depths following repeated tape stripping as a method of mechanical removal of stratum corneum layers (for detailed description see Materials and methods and Additional file 3). Samples for microbiome analysis were subsequently taken by swabbing normal skin and the deeper layers of the stratum corneum as exposed by tape stripping. In total, 495,709 bacterial 16S rRNA sequences were analyzed, resulting in an average of 5,901 (range 2,430 to 12,035) reads per sample (Additional file 4). The effect of volunteer origin, gender and stripping depth on microbiota composition were analyzed by redundant analysis, revealing volunteer-specific signatures (Figure 2, red arrows pointing to distinct directions) and strong effect of gender (Figure 2, green arrows) on microbiota composition. Redundant analysis also indicated important differences in microbial composition between the superficial and the deeper layer of the stratum corneum (blue arrows).

Hierarchical clustering of all samples showed that the individual volunteer is the strongest denominator of microbiota composition (Figure 3). UniFrac analysis showed no distinct clustering caused by either tape stripping or gender variables. However, even though the signal is small compared to the differences between volunteers, a small, but potential biologically relevant, signal could easily be obscured by the noise introduced by the individual nature of the samples. In-depth analysis of gender differences revealed bacterial taxa that were significantly different in relative abundance, and broadly supported gender differences in microbiota composition of the upper buttock skin (Figure 4). Males have relatively high proportions of Streptococcus, Eremococcus, Finegoldia, Anaerococcus, Veillonella, Sporacetigenium, Corynebacterium, Dermabacter, Brachybacterium, Microbacterium, Dermacoccus and Capnocytophaga when compared to females. Furthermore, females have relatively high proportions of Lactobacillus, Propionibacterium, Gardnerella and Enhydrobacter. The significance of these differences was determined using a Mann-Whitney U rank test (Figure 4). There is also a tendency for higher ratios of Staphylococcus, Janibacter and Brevibacterium in females. Most interestingly, we identified enrichment of Lactobacillus and Gardnerella to be specific for females and Corynebacterium in males. In addition, one of the Lactobacillus operational taxonomic units (OTUs) associated with females (P-value 0.034) could be assigned to the vaginal inhabitant Lactobacillus iners, as a microbial BLAST search showed the sequence to be 100% identical over the full 337 nucleotides to this species, validating the biological relevance of our statistical approach. Overall, males showed a higher bacterial diversity of the microbiome compared to females (Table 1) for all four diversity metrics tested (Shannon, Chao1, whole tree phylogenetic diversity, and observed OTUs), reflected by the presence of more small bacterial groups in males (Figure 4). Because we have observed significant differences in microbial composition between the superficial and deeper layers of the stratum corneum (Figure 2), we also used taxonomy-based classification provided by the Ribosomal Database Project combined with statistical mining to identify significant differences in microbiota composition (Figure 5). In the deeper layers of the stratum corneum (STR5 and STR10) we found a relative increase in the proportion of the following genera: Streptococcus, Staphylococcus, Rothia, Kocuria, Dermacoccus, Brevibacterium, Bifidobacterium and Pseudomonas. Furthermore, a relative decrease of Granulicatella, Propionibacterium and Sporacetigenium was found in these deeper layers. Translated to the phylum level, an increase in the relative abundance of Firmicutes (represented by Staphylococcus) in the deeper layers of the stratum corneum and a decrease of Actinobacteria (Propionibacterium) was observed (Figure 5).

We investigated the temporal changes in the microbiome composition of the skin surface following mechanical removal of the stratum corneum by tape stripping. Again, the host seems to be the most important denominator of microbiota composition as most of the samples were grouped by individual using hierarchical UniFrac clustering (Additional file 5). Initially, we observed that the microbiota profile of injured skin is considerably disturbed up to day 14 when compared to the microbiome of the superficial skin layer before stripping (Figure 6). It has to be noted that tape-stripped skin at day 14 in general has fully recovered from the initial injury as assessed by clinical (not shown) and microscopic criteria (Additional file 6). Even if the samples from different volunteers differ widely in their constituent taxa, the behavior and dynamics in time can be similar. The consensus tree shows that after tape stripping, the microbiota composition (visualized by the pie charts in Figure 6) on day 1 reflects the superficial layer, which then diverges within 14 days in the direction of the STR10 state. No difference between partial or complete removal of the stratum corneum was found with respect to microbiome changes in the process of recolonization. This is illustrated by the pie charts, which respectively correspond to 15 times tape-stripped skin (F1 to F3 and M1 to M3) and skin with a completely removed stratum corneum (F4 to F6 and M4 to M6). Gender differences also appeared in the recolonization process during the first days after tape stripping, as we observed a relative increase of Propionibacterium, which was a more dominant bacterial group in women, at the cost of Staphylococcus, Corynebacterium and Micrococcus. The microbiome shifts in response to skin injury were less dramatic in males than in females (Additional file 7).

As reported above and described in earlier studies, the microbiota composition varies considerably between individuals 151718. This applies both to diversity and the dominance of certain taxonomic groups, for example, as shown for females F1 (Propionibacterium dominance) and F6 (Staphylococcus dominance) in Figure 6. The factors that determine the skin microbiome composition are not known, but host factors involved in innate immunity of the skin are likely to play a role in shaping the microbiome under steady state conditions and following disturbance of homeostasis (for example, by skin injury). Previous work from our lab has shown that expression of antimicrobial proteins is strongly induced following skin barrier disruption 114041. In this part of the study we compared the responses of five healthy controls with respect to induction of eight different antimicrobial proteins. Note that these individuals were distinct from the volunteers in the tape stripping study from whom no biopsy material was available. Gene expression profiles at baseline and 24 hours after tape stripping of the upper buttock skin were analyzed by quantitative PCR (qPCR). In normal healthy skin antimicrobial proteins are expressed at moderate levels (SLPI, MRP8, psoriasin), low levels (hBD-3, elafin, lysozyme) or undetectable levels (hBD-2, LL37) (Figure 7a). An indication for the basal antimicrobial protein expression levels are the mean Ct values for SLPI (Ct = 26), MRP8 (S100A8, Ct = 27), psoriasin (S100A7, Ct = 28), hBD-3 (DEFB103, Ct = 30), elafin (PI3, Ct = 28), lysozyme (LYZ, Ct = 33), hBD-2 (DEFB4, Ct = 37) and LL37(CAMP, Ct = 37). Upon tape stripping, all genes showed upregulation, with the highest fold-increases for elafin, MRP8 and psoriasin. Although the eight genes showed induction in all individuals studied, we observed a hitherto unrecognized large variation between individuals. For example, in individual 1 there was abundant induction of all antimicrobial protein genes except for elafin, whereas individual 3 showed low responses for hBD-3, SLPI, MRP8, psoriasin, LL37, and lysozyme compared to the other individuals while the expression levels of hBD-2 and elafin are the highest in this person. This phenomenon was also confirmed at the protein level (Figure 7b), where hBD-2 and elafin protein expression levels at 48 hours after tape stripping varied between individuals.

All volunteers in this study were selected according to the inclusion/exclusion criteria as approved by a protocol from the National Institutes of Health (NIH) Human Microbiome Project (protocol number HMP-07-001, 30 March 2009; protocol available at the NIH Human Microbiome Project website) 58. The exact inclusion/exclusion criteria and study procedures that we presented to the volunteers can be found in Additional files 10 and 11. Medical ethical committee (Commissie Mensgebonden Onderzoek Arnhem-Nijmegen) approval and individual written informed consent were obtained in advance of sample collection. The study was conducted according to the Declaration of Helsinki principles.

Five healthy volunteers (one male and four females, Caucasian, aged 26 to 31 year) were included in the topographical study. Subjects were instructed not to wash or use body lotion for 24 hours prior to sampling and to avoid swimming in a chlorinated pool, using a hot tub, sauna/steam bath or tanning bed for 48 hours prior to sampling (Additional file 11). All samples were taken in April 2010. Samples were collected from body sites with different microenvironments: two moist regions (armpit, inner elbow), a sebaceous region (forehead) and a presumed dry region (upper buttock). Samples were obtained by swabbing 4 cm² skin areas using Sterile Catch-All™ Sample Collection Swabs (Epicentre Biotechnologies, Madison, WI) soaked in sterile SCF-1 solution (50 mM Tris buffer (pH8), 1 mM EDTA, and 0.5% Tween-20). The swabbing technique was executed as follows: stretch the skin with one hand while the other hand holds the swab so that the shaft is parallel to the skin surface. Subsequently, the swab is rubbed back and forth approximately 50 times applying firm pressure. Immediately after swabbing, each swab is swirled in a 2 ml collection tube containing 300 μl MicroBead Solution (MO BIO Laboratories, Carlsbad, CA). The swab sponge should be pressed against the tube wall multiple times for 20 seconds to ensure transfer of bacteria from swab to solution. Samples were stored at -20°C until further processing. To minimize sample cross-contamination, a fresh pair of sterile gloves was worn by the person sampling each individual.

To study microbiome dynamics of human epidermis following skin barrier disruption we used the well-described 'tape stripping' method 10. This method involves the repeated application of adhesive tape to the skin surface, thereby removing stratum corneum layers. This method creates a superficial wound showing slight skin irritation (erythema) and loss of barrier function (transepidermal water loss). An example of the morphology of tape-stripped skin and subsequent regeneration of human epidermis is depicted in hematoxylin and eosin stained sections in Additional file 6. For this study (all samples were obtained in February 2011), we recruited 12 healthy Caucasian volunteers, 6 male and 6 female, aged 21 to 56 years, according to the Human Microbiome Project criteria as described above. Four of these volunteers participated also in the topographical study (Additional file 12). For tape stripping experiments we selected the upper buttock skin (just under the waistband) for two reasons. First, this area of the body contains a relatively high richness of observed OTUs (Results section; Figure 1; Additional files 1 and 2). In addition, the skin at this body site is a convenient area for tape stripping experiments or biopsy taking, which may sometimes result in abnormal wound healing with poor cosmetic results.

To study the human skin microbiome in different epidermal layers, two areas on the right upper buttock measuring 2 cm² (1 × 2 cm) each were tape-stripped 5 (STR5) and 10 (STR10) times, respectively, by application and removal of adhesive tape (n = 12). Using the swabbing technique described above, the two barrier-disrupted skin areas were sampled, as well as 2 cm² healthy non-barrier disrupted (STR0) right upper buttock skin (see Additional file 3 for experimental set-up).

To study recolonization of the human skin microbiome after skin barrier disruption, four areas on the left upper buttock measuring 2 cm² each were tape-stripped 15 times (n = 6, three males and three females) or as many times as required to obtain a glistening surface, which indicates complete removal of the stratum corneum. The latter procedure (usually 30 to 70 times tape stripping) exposes the upper living epidermal cell layer and was performed in six individuals, three males and three females. These tape-stripped areas were sampled over time for up to two weeks (see Additional file 3 for experimental set-up). This end point was chosen for the reason that, in general, injured skin (by tape stripping) recovers completely within this time period based on clinical and microscopic criteria. In a previous pilot experiment we found that 15 times tape stripping of healthy upper buttock skin minimizes bacterial counts below detectable levels with qPCR (Additional file 9). Analysis of 15 times tape-stripped skin up to complete removal of the stratum corneum shows that this lower compartment is practically devoid of detectable bacterial DNA. The reason for including skin samples with totally removed stratum corneum (30 to 70 times tape stripping) is that this results in a moist wound surface, which may have other recolonization characteristics than the mildly tape-stripped (15 times) wounds.

Finally, we analyzed archival material of five healthy volunteers from which we have obtained biopsies from the left upper buttock (males and females, Caucasian race, aged 24 to 66 years) to study the host response towards superficial injury. These individuals were previously tape-stripped on two areas on the upper buttock measuring 4 cm² each until the surface became slightly shiney. After 24 and 48 hours, 3-mm biopsies were taken from the tape-stripped area and from healthy skin for both RNA isolation and histology, respectively, to examine antimicrobial protein gene expression profiles of the host.

DNA was extracted from the swabs using the MO BIO Ultraclean Microbial DNA Isolation Kit with modifications. As described above, each swab was swirled in a 2 ml collection tube containing 300 μl MicroBead Solution. Subsequently, 50 μl MD1 Solution was added to the bacterial cells and the samples were heated for 10 minutes at 70°C. This bacterial solution was transferred to a MicroBead tube and horizontally vortexed for 10 minutes at maximum speed using the MO BIO Vortex Adapter tube holder. The remaining steps were performed as described in the manual as provided by the manufacturer. DNA samples (30 μl in MD5 solution) were stored at -20°C until further processing.

The following universal primers were applied for amplification of the V3-V6 region of the 16S rRNA gene: forward primer, 5'-CCATCTCATCCCTGCGTGTCTCCGACTCAGNNNNNNACTCCTACGGGAGGCAGCAG-3' (the italicized sequence is 454 Life Sciences primer A, and the bold sequence is the broadly conserved bacterial primer 338F; NNNNNN designates the sample-specific six-base barcode used to tag each PCR product); reverse primer 5'-CCTATCCCCTGTGTGCCTTGGCAGTCTCAGCRRCACGAGCTGACGAC-3' (the italicized sequence is 454 Life Sciences primer B, and the bold sequence is the broadly conserved bacterial primer 1061R).

Data from a pilot experiment revealed that samples derived from 2 cm² of skin on the upper buttock contain only small numbers of microorganisms (just a slight band around 800 bp of length on agarose gel after 16S rRNA gene PCR amplification). Therefore, we introduced a pre-amplification step with the same primers as described above excluding the barcodes and flag sequences. In this pilot experiment we established that this additional PCR step did not affect the results compared to a single amplification step with bar-coded primers; in other words, we observed no distortion or skewing of the taxa distribution. As a negative control we used a mock swab, which was only exposed to ambient air, and extracted the genomic microbial DNA from this sample. However, even after two rounds of PCR no visible band on agarose gel was observed. The first amplification PCR consists of 5 μl microbial genomic DNA, 16 μl master mix (1 μl KOD Hot Start DNA Polymerase (1 U/μl; Novagen, Madison, WI, USA), 5 μl KOD-buffer (10 ×), 3 μl MgSO4 (25 mM), 5 μl dNTP mix (2 mM each), 1 μl (10 μM) of each forward and reverse primer), and 29 μl sterile water (total volume 50 μl). PCR conditions were: 94°C for 2 minutes followed by 30 cycles of 94°C for 20 s, 55°C for 10 s, and 70°C for 15 s, ending with a last step of 72°C for 10 minutes to ensure complete amplification of the target region. The approximately 750 bp PCR amplicon was subsequently purified using the MSB Spin PCRapace kit (Invitek, Westburg, The Netherlands) and the concentration was checked with a Nanodrop 2000 spectrophotometer (Thermo Scientific, Wilmington, DE). All PCR amplicons were diluted to a concentration of 10 ng/μl, and 3 μl of this purified amplicon (total 30 ng) was used as input for a re-amplification PCR using the bar-coded primers as described above. PCR conditions were: 94°C for 2 minutes followed by 35 cycles of 94°C for 20 s, 55°C for 10 s, and 70°C for 15 s, ending with a last step of 70°C for 10 minutes. Again, the approximately 750 bp PCR product was purified (MSB Spin PCRapace kit) and the concentration was measured. A composite sample for pyrosequencing was prepared by pooling 100 ng of these purified PCR products of each sample. The pooled sample was elecrophorized on a 1% agarose gel and the approximately 800 bp band was excised and extracted from the agarose gel with the MinElute Gel Extraction kit (Qiagen, Venlo, The Netherlands). The concentration of the gel-extracted amplicon was determined and 50 μl (concentration 14.5 ng/μl purified PCR product) was submitted for pyrosequencing of the V3-V4 region of the 16S rRNA gene on the 454 Life Sciences GS-FLX platform using Titanium sequencing chemistry at DNAvision, Charleroi, Belgium.

Pyrosequencing data were analyzed with a workflow based on QIIME v1.2 59, using settings as recommended in the QIIME 1.2 tutorial, with the following exceptions: reads were filtered for chimeric sequences using Chimera Slayer 60; the number of QC-passed sequences per sample are presented in Additional file 4 and OTU clustering was performed with settings as recommended in the QIIME newsletter of 17 December 2010 61 using an identity threshold of 97%. Diversity metrics were calculated as implemented in QIIME 1.2. Hierarchical clustering of samples was performed using UPGMA with weighted UniFrac as a distance measure as implemented in QIIME 1.2. The Ribosomal Database Project classifier version 2.2 was performed for taxonomic classification 62. Visualization of differences in relative abundance of taxa between different sample groups (Figures 4 and 5) was done in Cytoscape 63. Taxa (that is, nodes) were included in the visualization if they met the following criteria: all samples together have at least 10 reads assigned to the taxon and the sample groups have a fold-difference of at least 0.1 for the taxon, or the taxon has a child (that is, more specific taxonomic classification) meeting the first criterion. The significance of the difference in relative abundance of specific taxa between sample groups was calculated using the Mann-Whitney U test as implemented in SciPy 64. We provide additional information on the diversity metrics and statistics in Additional file 13. Additional data handling was done using in-house developed Python and Perl scripts.

Isolation of epidermal sheets for mRNA extraction was performed as previously described 65. RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany). A DNase I treatment was performed according to the manufacturer's protocol (Invitrogen, Carlsbad, CA, USA). Reverse transcriptase reactions and qPCR were performed as described previously 39. The amount of mRNA for a given gene in each sample was normalized to the amount of mRNA of the human ribosomal phosphoprotein P0 (RPLP0) reference gene in the same sample. Primers for qPCR (Biolegio, Nijmegen, the Netherlands) were only accepted if their efficiency was 100 ± 10%. Corrections were made for primer efficiency. Primer sequences and efficiency are shown in Additional file 14. Relative mRNA expression levels were calculated with the delta-delta Cycle threshold (ΔΔCt) method 66. In case of very low to absent mRNA levels (Ct value > 37), the relative quantity was set to 0. All mRNA expression levels in purified epidermal sheets are related to the lysozyme expression of the normal healthy skin of volunteer 3, which was set to 1.

Skin biopsies were immediately fixed in a 10% formalin solution (Baker Mallinckrodt, Deventer, The Netherlands) for 4 hours and subsequently embedded in paraffin. Sections were stained with hematoxylin and eosin or analyzed by immunohistochemistry as follows: sections were blocked for 15 minutes with 20% normal rabbit serum (for hBD-2) or normal goat serum (elafin) in phosphate-buffered saline and subsequently incubated with anti-hBD-2 (1:100; Abcam, Cambridge, UK), and anti-elafin (1:500) for 1 hour at room temperature. Next, sections were incubated for 30 minutes with a secondary antibody (biotinylated rabbit anti-goat or biotinylated goat anti-rabbit in phosphate-buffered saline containing 1% bovine serum albumin, Vector laboratories, Burlingame, CA, USA). After 30 minutes incubation with Avidin-Biotin complex (Vector Laboratories, Burlingame, CA), sections were treated with 3-amino-9-ethyl carbazole (Calbiochem, San Diego, CA, USA) for 10 minutes.

Sequence data from this study have been submitted to MG-RAST 67 as project number 2329 68. The barcodes for linking reads in the pooled samples to individual samples are provided in Additional file 4.