Present addresses: Markus Egert, Faculty of Science, Coburg University of Applied Sciences, Coburg, Germany. Inke Schmidt, Department of Veterinary Disease Biology, University of Copenhagen, Frederiksberg, Denmark. Tim Lachnit, Zoological Institute, Christian-Albrechts-University, 24098 Kiel, Germany.
Correspondence: Roland Breves, Corporate Analytics, Microbiology and Product Safety, Henkel AG & Co. KGaA, 40191 Düsseldorf, Germany. Tel.: +49 211 797 9630;fax: +49 211 798 2245;e-mail: email@example.com
The activity of the human armpit microbiota triggers the formation of body odor. We used differential 16S rRNA gene (rDNA)- and rRNA-based terminal-restriction fragment length polymorphism fingerprinting in combination with cloning and sequencing to identify active members of the human armpit microbiota. DNA and RNA were isolated from skin scrub samples taken from both armpits of 10 preconditioned, healthy males. The fingerprint profiles indicated pronounced similarities between the armpit microbiota in the right and the left axillae of an individual test person, but larger differences between the axilla microbiota of different individuals. Using 16S rDNA and rRNA sequence data, the majority of peaks in the armpit profiles were assigned to bacteria affiliated with well-known genera of skin bacteria. The relative abundances of all groups were similar among the rDNA and rRNA samples, suggesting that all groups of armpit bacteria were active. Surprisingly, the relative abundance of sequences affiliated with Peptoniphilus sp. was by far and with statistical significance the highest in the rRNA samples of the right armpits. Thus, bacteria affiliated with Peptoniphilus sp. might have been particularly active in the right axillae of the test persons, possibly owing to the handedness of the test persons, which might cause different environmental conditions in the right axillae.
The human commensal skin microbiota clearly fulfills several beneficial functions for their host, for example protection against pathogenic microorganisms by stimulation of the skin immune systems or by contribution to the skin acid protection mantle (Cogen et al., 2008; Krutmann, 2009). On the other hand, skin microorganisms also cause negatively perceived skin phenomena, such as impure skin (Brüggemann et al., 2004) or the formation of body odor. In the case of the latter, the microbial community of the human axilla clearly plays a key role (Leyden et al., 1981). Body odor is caused by a complex mixture of compounds, excreted from the skin as water-soluble, nonvolatile, and nonmalodorous precursors, from which the volatile parts are subsequently released by microbial enzyme activities (Gautschi et al., 2007). Hence, a better understanding of the formation of human body odor is closely linked to a better understanding of the activity of the human armpit microbiota.
Owing to its importance for human health and well-being, the human skin is one of the major targets of the Human Microbiome Project (Turnbaugh et al., 2007). Just recently, studies of unprecedented comprehensiveness have unraveled microbial diversity in several niches of the human skin, such as the palm (Fierer et al., 2008), forearm (Gao et al., 2007), and inner elbow (Grice et al., 2008). Links between changes in skin microbial community composition and certain skin diseases are currently under investigation (Dekio et al., 2007; Gao et al., 2008). Also, the microbial community of the human axilla was recently investigated in considerable detail, during the course of a study that created a topographical map of the human skin microbiome (Grice et al., 2009). However, all of the above-mentioned studies were DNA based, i.e. focused on the structure of the skin microbiota rather than on its activity.
It is widely accepted that rRNA-based analyses rather target the active members of microbial communities as the cellular content of rRNA is linked to cellular activity (Molin & Givskov, 1999). Studies directly comparing rRNA gene (rDNA)- and rRNA-based data on microbial community composition have been conducted in other ecosystems such as water (Stoeck et al., 2007) or soil (Lueders & Friedrich, 2002) and were successful in unraveling active groups of microorganisms and helped to better understand their role in the ecosystem functioning.
Hence, the aim of this study was to identify particularly active members of the human armpit microbiota by comparing rDNA and rRNA-based community fingerprints of the armpit microbiota of healthy male. To the best of our knowledge, our study represents the first rRNA-, i.e. activity-based microbial community analyses of the human armpit microbiota.
Materials and methods
Armpit sampling and extraction of nucleic acids
Armpit microbiota samples were taken using a skin scrub sampling technique (similar to the scrub methods described by Williamson & Kligman, 1965; Keswick & Frank, 1987) from the right and left axillae of 10 healthy, male volunteers (#1–#10), who had been instructed not to have a shower or to use deodorants for 48 h before sampling and not to wash themselves for 24 h before sampling. Samples were taken by standardized scrubbing of the middle of the axilla region with 2 mL of phosphate-buffered saline (pH 8.0, 1% Tween, 0.1% tryptone) using a sterile glass bar (rotated 10 times clockwise on the skin) and a sterile plastic cylinder held firmly onto the axilla. Subsequently, the buffer was taken up with a sterile pipette and mixed with 4 mL of RNAprotect Bacteria Reagent (Qiagen, Hilden, Germany) to preserve the nucleic acids. After 5 min of incubation at room temperature, the buffer samples were centrifuged to form a pellet, which was subsequently taken up in 600 μL of RNAprotect Bacteria Reagent, divided into equal shares, and frozen at −20 °C until further analysis.
For DNA extraction, the cells contained in 300 μL of the armpit sample were harvested by centrifugation (6000 g, 10 min) and subsequently mechanically extracted in 180 μL of 20 mM Tris-HCl lysis buffer (pH 8.0; 2 mM EDTA, 1.2% Triton X-100, and 20 mg mL−1 lysozyme) using a vibration mill (MM301, Retsch, Haan, Germany). After 1 h of incubation at room temperature, 25 μL of proteinase K and 200 μL of AL buffer (both Qiagen) were added. After 1.5 h of incubation at 56 °C and a short centrifugation step (2 min, 6000 g), the supernatant was extracted using the DNeasy Blood and Tissue Kit (Qiagen) following the manufacturer's protocol for Gram-positive bacteria. After a final elution with 30 μL of AE buffer, the DNA concentration in the extracts was determined using a Nanodrop spectrophotometer (Peqlab, Erlangen, Germany). Finally, the DNA extracts were stored at −20 °C until further analysis.
For RNA extraction, the cells contained in 300 μL of the armpit sample were harvested by centrifugation (6000 g, 10 min) and taken up in 700 μL of lysis buffer (200 mM NaCl, 3 mM EDTA). Subsequently, 400 μL of water-saturated phenol/chloroform/isoamyl alcohol (Roth, Karlsruhe, Germany) was added and the cells were mechanically extracted in FastProtein Blue tubes (QbioGene, Heidelberg, Germany) using a FastPrep cell disruptor (FP120, QbioGene). After centrifugation (20 000 g, 5 min, 4 °C), the RNA contained in the aqueous supernatant was extracted using KingFisher mL technology (Thermo Electron, Waltham) and the MagNA Pure LC RNA High Performance Isolation Kit (Roche, Mannheim, Germany) following an optimized protocol (HGW4; Jürgen et al., 2005) based on the manufacturer's instructions. The RNA eluted by the KingFisher mL was subsequently treated with DNase and purified using the RNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. Finally, the RNA was eluted with 30 μL of water, quality checked using the Nanodrop spectrophotometer (Peqlab), and stored at −80 °C until further analysis. The absence of contaminating rDNA was verified by means of a standard PCR.
Amplification of bacterial 16S rDNA and rRNA
In order to analyze the composition of the bacterial community in the armpit samples, bacterial 16S rDNA fragments were amplified using a pair of primers (27f, 907r) targeting all bacteria (Egert et al., 2003). The PCR reaction mixture contained 1 ×Taq PCR master mix (Qiagen), 0.5 μM of each primer, and 1–20 ng of template DNA in a total volume of 50 μL. The thermal profile has been described previously (Egert et al., 2003). To analyze the bacterial armpit community composition on an rRNA basis, the 16S rRNA was reverse transcribed into cDNA using the Sensiscript Reverse Transcription Kit (Qiagen) using ∼4 ng of RNA as a template and 0.6 μM of primer 907r. After reverse transcription (RT) and thermal inactivation of the enzyme, the reactions were cleaned (QIAquick PCR Purification Kit, Qiagen) and quantified using the Nandrop spectrophotometer. Subsequently, the cDNA was used as a template in an rDNA-targeted PCR as specified above.
After PCR, amplicons were checked for correct size by standard agarose gel electrophoresis and ethidium bromide staining, purified (QIAquick or QiaMinElute PCR Purification Kit, Qiagen), quantified using the Nanodrop photometer, and stored at −20 °C for further analysis.
Terminal-restriction fragment length polymorphism (T-RFLP) fingerprinting
For T-RFLP profiling (Schütte et al., 2008) of the compositions of the microbial communities, PCRs were conducted using a 6′-carboxyfluorescein-labeled 27f-primer. Seventy nanograms of purified and FAM-labeled PCR product was restricted for 3 h at 37 °C with 2.5 U of MspI (Invitrogen, Karlsruhe, Germany) in a total volume of 10 μL. After inactivation of the restriction enzyme (20 min at 60 °C), the reaction mixtures were frozen. Size-separation of fluorescently labeled restriction fragments and subsequent size-calling analyses were performed at the Fraunhofer Institute for Molecular Biology and Applied Ecology (Aachen, Germany) using an ABI 3730 DNA analyzer (Applied Bioystems, Weiterstadt, Germany), LIZ1200 as the internal size standard and the genemapper V3.7 software (Applied Biosystems).
Before analysis of the T-RFLP profiles, all restriction fragments smaller than 50 bp in length were excluded from the profiles to exclude possible primer dimers. Profiles were then normalized by calculating the relative peak heights, i.e. referring the height of each terminal-restriction fragment (T-RF) peak to the cumulative height of all the peaks in the profile. Minor peaks, arbitrarily defined as those with <2% relative peak height, were excluded from the analysis. To ensure that the total peak height of all major peaks equaled 100%, they were normalized to the total peak height of all the major T-RFs in the profile (Egert et al., 2004). To facilitate comparisons, T-RFs similar in size to within ± 1–2 bp were grouped into operational taxonomic units. Finally, the pairwise similarity of T-RFLP profiles was calculated using the Morisita index of community similarity as described previously (Egert et al., 2004). Morisita indices range from 0 to 1, with 1 indicating the complete (100%) identity of two communities.
Cloning and phylogenetic analyses of 16S rDNA and 16S rRNA fragments
Forty-eight randomly chosen bacterial 16S rDNA and 48 16S rRNA fragment PCR products were obtained from the left axilla samples of three volunteers (#1, #2, #4). They were cloned and sequenced by AGOWA (Berlin, Germany) using standard protocols. The sequences obtained were analyzed using the arb software package (Ludwig et al., 2004), i.e. they were added to the arb database and aligned using the integrated aligner. The alignments were corrected manually where necessary. The sequences were also compared with public databases using blast (McGinnis & Madden, 2004): closely related sequences were retrieved and added to the alignment. Trees were calculated using the neighbor-joining algorithm and a base frequency filter for Bacteria provided with the arb package. For constructing phylogenetic trees, 649 sequence positions (Escherichia coli numbering 73–851) were used. Chimeric sequences were identified by fractional treeing (Egert et al., 2003).
Sequences representing all clusters of phylotypes displayed in Fig. 1 were deposited in public databases (Table S1). Clones designated B1, B3, and B5 were obtained from DNA extracts of the left armpits of volunteers #1, #2, and #4, respectively (rDNA clones; accession numbers FN658918–FN658984). Clones designated B2, B4, and B6 were obtained from RNA extracts of the left armpits of volunteers #1, #2, and #4, respectively (FN658841–FN658917, rRNA clones).
The relative abundances, based on rRNA and rDNA analyses, of all the major groups of bacteria from both armpits (left and right) were analyzed using one-factor anova, followed by Tukey's honest significant differences test using statistica (Microsoft, Tulsa). One factor was ‘bacterial genus’; the response variable was the relative abundance of this genus. Shapiro–Wilk's W statistic was used to test for normal distribution; Cochran's C test (for normally distributed data) was used to test for homogenous variances. After arcsinus transformation, all data were normally distributed and variances were homogenous.
Results and discussion
The aim of this study was to identify specifically active members of the human armpit microbiota by comparing rDNA- and rRNA-based fingerprinting profiles. To the best of our knowledge, such a comparative approach – although common in microbial ecology research (Lueders & Friedrich, 2002; Stoeck et al., 2007) – has never been applied to a human skin ecosystem before. Here, the microbiota of the human armpit was investigated. Its activity is a key factor in triggering the formation of body odor (Leyden et al., 1981; Rennie et al., 1991; Guillet et al., 2000).
Intra- and interpersonal similarity of armpit communities
In this study, rDNA and rRNA were successfully extracted from samples taken by means of the skin scrub technique from the left and right armpits of 10 preconditioned, healthy, male test persons. Partial 16S rDNA PCR fragments were amplified from all 20 DNA extracts, and after RT, from all the RNA extracts. T-RFLP fingerprint profiles could thus be established from all the samples (Fig. 1).
For each individual test person, the rDNA- and rRNA-based fingerprints were very similar. The average pairwise similarity of rDNA and rRNA fingerprints, calculated as Morisita indices of community similarity, were 88 ± 6% for the right axillae and 91 ± 4% for the left axillae. In addition, the fingerprints obtained from the right and left axilla of each individual were very similar: 89 ± 5% average similarity, when based on DNA, and 87 ± 7%, when based on rRNA. In contrast, the fingerprints showed marked interindividual differences: averaged over all the pairwise comparisons that are possible for 10 test persons, the right axillae showed similarities of just 52 ± 7% (rDNA based) or 50 ± 5% (rRNA based). Almost identical low similarities were calculated for the left axillae: 53 ± 5% (rDNA based) and 55 ± 5% (rRNA based). In conclusion, the armpit flora appears to differ considerably between different individuals, but is very similar in the armpits of a single individual. Similar findings have been described before for the armpit using culture-based methods (Leyden et al., 1981) and for forearm skin using molecular tools (Gao et al., 2007).
Relative abundances of major armpit genera
In order to assign the dominant T-RFs in the armpit profiles to bacterial genera, bacterial 16S rRNA genes were cloned after PCR- or RT-PCR amplification from the corresponding rDNA and rRNA samples of the left armpits of three test persons. Between 44 and 48 16S rRNA gene sequences were obtained per sample. After the separation of potentially chimeric sequences, phylogenetic analyses of the remaining 247 sequences (Fig. 2) revealed that these were affiliated with genera well known for their association with the human skin (Dekio et al., 2005; Gao et al., 2007) and armpit (Leyden et al., 1981; Taylor et al., 2003; Grice et al., 2009), i.e. Staphylococcus, Corynebacterium, and Propionibacterium. In addition, two genera related to Peptostreptococcaceae (genera Anaerococcus and Peptoniphilus) were detected in relatively high numbers. A similar finding has been reported before, particularly for male test persons (Trebesius et al., 2006).
In order to assign the cloned sequences to the T-RFs in the fingerprint profiles, selected clones were checked for their in vitro T-RF formation patterns (Fig. 2; Table S1). The clonal T-RF profiles were checked for plausibility by an in silico search for restriction sites within the cloned sequences. The shorter and longer restriction fragments observed in addition to the true T-RF might result from an overdigestion of the amplicons (shorter additional fragments) or represent so-called pseudo-T-RFs (longer additional fragments; Egert & Friedrich, 2003). Also, the differences observed between the in silico predicted and actually determined T-RF length of 4–7 bp have been described before (Kaplan & Kitts, 2003).
Using the information from the clonal T-RFLP profiles, the vast majority of the armpit T-RFLP profiles could be assigned to distinct bacterial genera (Fig. 1; Table 1; Table S1). Subsequently, the relative abundances of these dominant genera were calculated based on the relative frequencies of the assigned T-RFs (Table 1). Statistical analyses indicated that there were no significant differences between the relative abundances of the genera Staphylococcus, Corynebacterium, Propionibacterium, and Anaerococcus, irrespective of whether the abundances were calculated on the basis of rDNA or rRNA fingerprints.
Table 1. Average relative abundances of major groups of bacteria in samples taken from the right and left armpits of 10 male test persons, based on the relative abundances of assigned fingerprint peaks in 16S rDNA and 16S rRNA T-RFLP profiles (Fig. 1)
Each value represents the average ± SE of the mean (n=10, each).
5.7 ± 1.6%
4.5 ± 2.2%
5.5 ± 2.3%
7.0 ± 2.9%
12.4 ± 6.1%
13.6 ± 6.3%
15.7 ± 7.2%
15.0 ± 8.1%
0.9 ± 0.5%
12.1 ± 4.4%
0.9 ± 0.6%
2.9 ± 1.5%
13.5 ± 8.3%
13.8 ± 7.8%
13.5 ± 7.1.%
17.7 ± 9.0%
60.2 ± 12.1%
52.0 ± 12.9%
56.1 ± 10.5%
51.0 ± 10.8%
7.3 ± 4.5%
4.0 ± 2.4%
8.4 ± 3.9%
6.4 ± 3.8%
Interestingly, the relative abundance of bacterial sequences related to Peptoniphilus sp. was by far the highest (12.1%) in the rRNA samples of right armpits. This abundance was significantly (P<0.05) higher than the relative abundance of this group in the right and left DNA armpit samples (0.9%, each). With limited significance (P<0.10), it was also higher than that in the rRNA samples from the left armpits (2.9%) (Table 1). These data suggest that bacteria affiliated with Peptoniphilus were particularly active in the right armpit of the test persons. This might be due to the handedness of the test persons, which might lead to different environmental conditions, for example aeration, mechanical agitation and/or sweat secretion, in the right armpits compared with the left armpits. Unfortunately, the actual handedness of the test persons sampled in this study was not recorded. However, as roughly 90% of human individuals tend to use their right hands for complex tasks (Cashmore et al., 2008), it is likely that the vast majority of the test persons were indeed right-handed. However, to substantiate the findings presented here, a study comparing the activity of the microbiota in both armpits of left- and right-handed person would be needed. Ideally, such a study should also include PCR-independent techniques, as PCR-based techniques are prone to certain forms of bias (Wintzingerode et al., 1997; Kanagawa, 2003).
Implications for understanding the functionality of the armpit microbiota
To the best of our knowledge, community analyses of the human armpit microbiota have up to now been based either on cultivation, well known to be affected by pronounced differences in cultivability of the target bacteria, or on DNA-based molecular methods, hence neglecting the fact that rRNA better reflects microbial activity. Our data show that the relative importance of a group of bacteria (Peptoniphilus sp.) can easily be underestimated when relative abundances are solely based on 16S rRNA genes (∼1–3%) instead of 16S rRNA (12%). Clearly, rRNA-based approaches, based as they are on microbial activity, can help to detect novel aspects of the functionality of the human armpit microbiota.
The large similarities between the rDNA- and the rRNA-based fingerprints suggest that the dominant members of the human armpit microbiota (staphylococci, corynebacteria, propionibacteria) are also rather active. This can be explained by the environmental conditions in the human armpit (warm, wet, occluded, eutrophic; Bojar & Holland, 2002), which are very favorable for microbial growth and activity.
However, our data also suggest that rRNA-based analyses are suitable for revealing differences in the composition of the microbial community that would otherwise not have been detected using rDNA-based techniques. The asymmetric activity pattern observed for bacteria affiliated with Peptoniphilus sp. was only detected using rRNA, but not with rDNA. As only molecular data were recorded here, further studies will have to target physiological reasons for this asymmetric activity pattern. It is tempting to speculate and worth investigating in more detail whether the handedness of the test persons could indeed be responsible for this phenomenon, for example by causing differential sweat secretion in the axilla affiliated with the working hand (Ferdenzi et al., 2009). However, so far, it has remained unclear as to which (sweat) factor might have caused the asymmetric activity pattern of Peptoniphilus sp., but not of the other genera.
Further studies should also address whether the asymmetric activity pattern can be correlated with differences in body odor. Recently, it has been reported that perceptual differences between the right and the left axilla of humans can occur (Ferdenzi et al., 2009). However, up to now, a role in the formation of body odor has only been shown for other groups of skin bacteria, mainly aerobic corynebacteria and micrococci (Taylor et al., 2003), albeit only using culture-based studies. Nevertheless, as obligate anaerobic Peptoniphilus sp. produce mainly (smelling) butyric acid from peptone (Ezaki et al., 2001), a potential contribution of this genus to body odor formation appears to be worth investigating in more detail.
The authors wish to thank Nicholas Kennedy for English suggestions.