Functional characterisation of a SNP in the ABCC11 allele—Effects on axillary skin metabolism, odour generation and associated behaviours

Journal of Dermatological Science 73 (2014) 23–30

Contents lists available at ScienceDirect

Journal of Dermatological Science journal homepage: www.jdsjournal.com

Functional characterisation of a SNP in the ABCC11 allele—Effects on axillary skin metabolism, odour generation and associated behaviours Mark Harker a,*, Ann-Marie Carvell a, Vernon P.J. Marti a, Svetlana Riazanskaia a, Hailey Kelso a, David Taylor a, Sally Grimshaw a, David S. Arnold a, Ruediger Zillmer a, Jane Shaw a, Jayne M. Kirk b, Zee M. Alcasid c, Sheila Gonzales-Tanon c, Gertrude P. Chan d, Egge A.E. Rosing e, Adrian M. Smith f a

Unilever Research & Development, Quarry Road East, Port Sunlight, United Kingdom Waters Corporation, MS Technologies Centre, Manchester, United Kingdom c Unilever Philippines Inc., 1351 United Nations Avenue, Paco, 1007 Manila, Philippines d Clinical Trial Management & Testing Associates, Inc., Unit 1207, 2301 Civic Place, Civic Drive, Filinvest Corporate City, Alabang, Muntinlupa City, Philippines e Unilever Research Laboratory, Oliver van Noortlaan 120, 3133 AT Vlaardingen, The Netherlands f Unilever Discover, Colworth Science Park, Sharnbrook, United Kingdom b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 18 March 2013 Received in revised form 1 August 2013 Accepted 30 August 2013

Background: A single nucleotide polymorphism (SNP), 538G!A, leading to a G180R substitution in the ABCC11 gene results in reduced concentrations of apocrine derived axillary odour precursors. Objective: Determine the axillary odour levels in the SNP ABCC11 genotype variants and to investigate if other parameters associated with odour production are affected. Methods: Axillary odour was assessed by subjective quantification and gas chromatography headspace analysis. Metabolite profiles, microbiome diversity and personal hygiene habits were also assessed. Results: Axillary odour in the A/A homozygotes was significantly lower compared to the G/A and G/G genotypes. However, the perception-based measures still detected appreciable levels of axillary odour in the A/A subjects. Metabolomic analysis highlighted significant differences in axillary skin metabolites between A/A subjects compared to those carrying the G allele. These differences resulted in A/A subjects lacking specific volatile odourants in the axillary headspace, but all genotypes produced odoriferous short chain fatty acids. Microbiomic analysis revealed differences in the relative abundance of key bacterial genera associated with odour generation between the different genotypes. Deodorant usage indicated a high level of self awareness of axillary odour levels with A/A individuals less likely to adopt personal hygiene habits designed to eradicate/mask its presence. Conclusions: The SNP in the ABCC11 gene results in lower levels of axillary odour in the A/A homozygotes compared to those carrying the G allele, but A/A subjects still produce noticeable amounts of axillary odour. Differences in axillary skin metabolites, bacterial genera and personal hygiene behaviours also appear to be influenced by this SNP. ß 2013 Published by Elsevier Ireland Ltd on behalf of Japanese Society for Investigative Dermatology.

Keywords: Glutamine conjugate Malodour Odoriferous Skin microbiome Skin metabolomics Volatiles

1. Introduction The gene ABCC11, encodes an ATP-driven efflux pump for amphipathic anions [1,2], that displays a single nucleotide polymorphism (SNP), 538G!A, leading to a G180R substitution in the corresponding protein. The SNP (R180) variant lacks Nlinked glycosylation resulting in rapid proteasomal degradation of the variant protein [3]. The SNP was first described in individuals

* Corresponding author at: Unilever R&D, Quarry Road East, Bebington CH63 3JW, United Kingdom. Tel.: +44 151 641 3992; fax: +44 151 641 1854. E-mail address: [email protected] (M. Harker).

displaying dry earwax as a consequence of the A/A genotype in contrast to G/A and G/G, which gives rise to a wet earwax phenotype [4]. This dimorphic trait is prevalent among East Asians (80–95%) and rare among European and African populations (0– 3%). ABCC11 was shown to be expressed in the secretory cells of the cerumen apocrine glands and proposed to play a key role in their secretory activity [4]. Subsequently, ABCC11 was also observed in axillary apocrine glands with the SNP (R180) variant leading to a loss of the secretion of metabolites associated with the formation of odour [5]. However, this study and others have presumed that A/ A individuals are nonodourous [6] without specifically evaluating the levels of axillary malodour by olfactive perception in this phenotype. The extent to which the SNP (R180) variant of ABCC11

0923-1811/$36.00 ß 2013 Published by Elsevier Ireland Ltd on behalf of Japanese Society for Investigative Dermatology. http://dx.doi.org/10.1016/j.jdermsci.2013.08.016

24

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30

reduces the overall level of perceived axillary odour remains to be established. Axillary apocrine sweat is a lipid rich, viscous liquid, which in its native form is odourless. It is the action of skin commensal bacteria, which generates malodourous volatiles from this secretion [7]. A number of studies have demonstrated that aerobic lipophilic corynebacteria are the key organisms involved in the generation of axillary malodour [8,9]. Although Staphylococcus spp. have been implicated in the generation of a more general ‘‘sweaty’’ odour attributable to isovaleric acid [8]. Typical compounds associated with axillary odour include 3-methyl-2-hexenoic acid (3M2H) and 3-hydroxy-3-methyl-hexanoic acid (HMHA) [10], which are released from N-acyl glutamine conjugates secreted from axillary apocrine glands [11]. Several odoriferous sulfanylalkanols also contribute to axillary malodour [12,13], these are generated by the sequential action of a bacterial dipeptidase, tpdA, and a cysteine b-lyase on cysteine-(S) or cysteine-glycine-(S) conjugates [14]. Volatile steroids were once considered important malodour constituents of axillary malodour, but subsequent findings have questioned their relevance [10]. Body sites that do not host apocrine glands, but do support high microbial loads can also generate appreciable levels of odour, i.e. feet, due to the microbial generation of short chain fatty acids primarily from aliphatic amino acids derived from sweat and skin [15]. In our study, we aimed to complement previous work demonstrating the role of ABCC11 as an important component of the secretory mechanism of axillary apocrine glands. We assessed the axillary malodour levels of the ABCC11 genotype variants in a native population where the A allele was present at an intermediate frequency, through the subjective quantification by expert odour assessors. We also undertook detailed targeted analysis of specific odour precursors and metabolomic profiling of axillary skin washes obtained from these subjects. Specific odourous volatile compounds residing in the axillary headspace were also examined. As axillary malodour is also governed by the microbial population residing at the skin surface, microbiomic analysis was performed to determine if different ABCC11 genotype variants exhibited altered axillary microbial profiles. Finally, information regarding the personal care habits of the study population was also gathered to investigate the extent of selfawareness between the different genotypes of their level of axillary odour, reflected in their behaviours towards masking its presence. 2. Materials and methods 2.1. Subjects This study was conducted in accordance with the ethical principles of Good Clinical Practice and the Declaration of Helsinki. Ethics committee approval was sought before commencement of the study and all subjects gave written informed consent. The study was performed at the Unilever Consumer Studies Centre, Manila, Philippines. In total 200 female subjects (aged 18–40 years old) who were natural born Filipino citizens with established Filipino lineage were screened for participation in the study. The initial screening phase consisted of a phenotypic analysis to assign either wet or dry earwax type by a qualified dermatologist. During the screening phase, once 60 subjects with dry earwax had been identified, further subjects found to have dry earwax were excluded from any further participation in the study.

DNA sample in the form of a buccal swab, Isohelix buccal swabs (Cell Projects, Harrietsham, UK). The DNA was extracted using the Isohelix buccal DNA isolation kit (Cell Projects, Harrietsham, UK) according to the manufacturer’s instructions. To sequence the SNP (538G!A) in the ABCC11 gene, PCR amplification was carried out using primers forward MH1F (50 -AAGTCTCAAGGCGAGGGATT-30 ) and reverse MH1R (50 -CTAAGTGCCAGGGACATGGT-30 ). The resulting 300 bp product was purified using MultiScreen HTS PCR filter plates (Millipore, Billerica, USA). Sequencing was performed using an ABI Prism 3100 Genetic Analyser using the ABCC11 specific forward (MH1F) and reverse (MH1R) primers. 2.3. Malodour evaluation The odour intensity of each axilla (left and right) of each subject was evaluated by a group of 6 trained odour assessors after 5 and 24 h subsequent to a controlled axillary wash. Assessors for odour intensity assigned a score on a scale of 0 to 5, corresponding to the strength of underarm malodour encountered. A detailed record of how these assessments were made can be found in the Supplementary Materials and Methods. 2.4. Subject questionnaire Subjects (n = 164) also completed a questionnaire on variables which may be affected by ABCC11 activity such as, life-style, underlying non-pathological skin conditions and antiperspirant/ deodorant usage hygiene habits. 2.5. Axillary wash sampling Subjects were asked to lie down on a sampling bed in a supine position. A sterile teflon scrub cup was placed in one axilla, into which 1.5 ml of sterile distilled water was pipetted. The surface of the axillary skin, within the cup, was gently scrubbed with a teflon stick for 1 min and the fluid aspirated to a sample tube. This process was repeated at the same site in the same axilla and the two extracts pooled and stored at 80 8C. Subsequently, samples were defrosted at 4 8C and spun at 14,000 rpm for 10 min at 4 8C to remove any unwanted cellular material. Next the samples were concentrated to dryness overnight in a vacuum centrifuge. Following this concentration step the samples were re-suspended in 200 ml distilled water placed in a HPLC vial and stored at 80 8C until analysis. 2.6. Analytical-ultra performance liquid chromatography (UPLC) and high definition mass spectrometry (HDMS) UPLC was performed using an ACQUITY UPLC1 System (Waters, Milford, MA, USA) and SYNAPTTM G2 HDMSTM system (Waters, Manchester, UK). The remainder of the analytical details can be found in the Supplementary Materials and Methods. 2.7. Axillary skin volatile organic compound sampling and analysis Volatile odourant compounds present on the axillary skin surface were sampled using polydimethylsilicone (PDMS) membrane patches, as described previously [16]. Analytical procedures employed to analyse the patches can be found in the Supplementary Materials and Methods. 2.8. Axillary skin microbiome analysis

2.2. Genotyping of the ABCC11 allele In total 164 out of 200 subjects screened progressed to the genotyping phase of the study. Each subject provided a genomic

A detailed description of sampling conditions and experimental protocols to determine the bacterial profiles of all the subjects can be found in the Supplementary Materials and Methods.

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30 Table 1a Five hour mean malodour scores of the different genotype groups.

25

Table 1c Twenty-four hour mean malodour scores of the different genotype groups.

Group

MMS 5 h

Group

MMS 24 h

AA GA GG

2.59 3.26 3.21

AA GA GG

2.60 3.40 3.50

3. Results 3.1. Genotyping In total 164 subjects were screened to determine their ABCC11 genotype. Two sequencing reactions were performed for each genomic DNA sample, using the forward (MH1F) and reverse (MH1R) primers to facilitate the accurate genotype designation of each subject. The amplified products of each reaction provided amplicons of the expected size (300 bp). Sequencing data confirmed that the correct region of the ABCC11 locus had been amplified. From the 164 subjects analysed there were 31 G/G, 87 G/ A and 46 A/A, 25 individuals from each of these subgroups were randomly chosen to participate in the full study. 3.2. Malodour intensity The mean malodour score (MMS) of the right and left axillae of each subject was assessed (blind) 5 and 24 h after a controlled wash on four consecutive days. After 5 h the MMS of the A/A genotype group was significantly lower than that of the G/A and G/G groups (Tables 1a and 1b). A comparison of the MMS of the G/A vs. G/G groups demonstrated no significant difference (Table 1b). After 24 h the MMS of the A/A genotype group was again significantly lower than that of the G/A and G/G groups (Tables 1c and 1d), with the differences being slightly larger than they were at 5 h. Again, the MMS for the G/A vs. G/G genotype groups demonstrated no significant difference (Table 1d). However, an MMS score of >2 affords a level of malodour that is noticeable to others, indicating that A/A subjects do generate perceivable levels of axillary odour. 3.3. Personal hygiene behaviours All of the 164 subjects genotyped in the screening phase of the study completed a questionnaire regarding their personal hygiene habits, skin type and lifestyle. Analysis of the data failed to yield any significant differences between genotype vs. skin type or lifestyle (data not shown). However, a very clear link was established between ABCC11 genotype and the daily hygiene habits of subjects, when considering their regular use of either an underarm antiperspirant or deodorant product (Tables 2a and 2b). Significantly fewer A/A subjects vs. G/A and G/G subjects used an underarm antiperspirant or deodorant on a daily basis (Tables 2a and 2b). There were no significant differences in the use of antiperspirants or deodorants between the G/A vs. G/G genotype subjects. 3.4. Odour precursor concentrations in axillary washings The most abundant malodour conjugate precursor in the G/G and G/A genotypes was Na-3-methyl-3-hydroxy-hexanoylglutamine

(HMHA-Gln) who produced significantly more HMHA-Gln than the A/A group, but there were no significant differences between the G/G and G/A subjects (Fig. 1a). The second most abundant precursor was Na-3-methyl-2-hexenoyl-glutamine (3M2H-Gln), where again the G/G and G/A groups produced significantly more 3M2H-Gln than the A/A group, but again there were no significant differences between the G/G and G/A subjects (Fig. 1b). The 3-methyl-3sulfanyl-hexanol-cysteine-glycine (Cys-Gly-3M3SH) conjugate precursor was not detected in any of the A/A subjects, consequently the G/G and G/A groups produced significantly more than the A/A group and again there were no significant differences between the G/A and G/G subjects. 3.5. Metabolite profiles of axillary washings Principal component analysis (PCA) was performed for the acquired metabolomic data to identify if the metabolite profiles of the different ABCC11 groupings could be used to differentiate and possibly assign genotype. The scores and loadings plot for the PCA model of the three different genotypes is shown in Fig. 2a. Two groups were separated on the basis of the PCA analysis, which closely reflected two different genotype classes A/A vs. G/A and G/ G. Internal cross validation of the model gave a goodness of fit, R2, of 0.90 and the goodness of predication, Q2, of 0.69. The loadings plot indicates the exact mass retention time pairs (EMRT’s) that contributed towards the groupings in the scores plot, Fig. 2b. The EMRT’s highlighted contribute significantly in differentiating the groups. Analysis of these EMRT’s separating the two groups suggested that HMHA-Gln and 3M2H-Gln were major contributors responsible for this differentiation. Data from the PCA analysis was exported to SIMCA-P11.5 for partial least squares – discriminant analysis (PLS-DA) and orthogonal partial least squares – discriminant analysis (OPLS-DA). The scores plot of the PLS-DA data further separated the A/A from the G/A and G/G genotypes (Fig. 2c). The three genotypes were split into two classes (GA/GG and AA) for OPLS-DA modelling. The data were visualised using an S-plot (Fig. 2d). The axis represents the reliability and magnitude of each EMRT. The EMRT’s with the greatest magnitude and reliability for class GA/GG were identified as HMHA-Gln (m/z 275, ESI+) and 3M2H-Gln (m/z 257, ESI+). 3.6. Axillary volatile organic compound profiles Two-dimensional gas chromatography coupled with time-offlight mass spectrometry analysis of polydimethylsilicone (PDMS) membrane patches, used to sample volatile organic compounds from the axillary skin surface, revealed that sulfanylalkanols were detected in the majority of the G/A and G/G subjects (Fig. 3a). No sulfanylalkanols were detected in the patches used to sample A/A subjects. A similar result was also observed for the detection of

Table 1b Group differences in mean malodour scores at 5 h. Group

Group

Difference in MMS

Standard error difference

Lower 95% confidence limits

Upper 95% confidence limits

p-Value

GA GG GG

AA AA GA

0.67 0.63 0.05

0.21 0.21 0.21

0.25 0.20 0.38

1.10 1.05 0.47

0.002 0.004 0.828

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30

26

Table 1d Group differences in mean malodour scores at 24 h. Group

Group

Difference in MMS

Standard error difference

Lower 95% confidence limits

Upper 95% confidence limits

p-Value

GA GG GG

AA AA GA

0.80 0.90 0.10

0.22 0.22 0.22

0.36 0.46 0.34

1.24 1.34 0.54

0.0005 0.0001 0.6461

Table 2a Subject response to the question ‘‘Do you normally use an underarm antiperspirant or deodorant?’’. Deo/Ap user

corresponding uniquely occurring OTU’s in the G/G and G/A samples on the right hand side. In the centre are the OTU’s that occur in both groups and all three genotype variants.

Genotype

Users Non-users

AA

GA

GG

50% 50%

86% 14%

97% 3%

4. Discussion The reported association between the homozygous SNP (R180) variant and the dry earwax phenotype with near complete loss of

Table 2b Testing the difference in response counts between genotypes. Test

Chi square

Probability > chi square (p)

AA vs. GA

Likelihood ratio Pearson

25.13 25.90

>0.0001* >0.0001*

AA vs. GG

Likelihood ratio Pearson

22.94 18.89

>0.0001* >0.0001*

GA vs. GG

Likelihood ratio Pearson

1.78 1.49

*

0.18 0.22

Significant difference with 99% confidence.

3M2H, where no acid was detected in the patches from A/A subjects but was frequently detected in patches from the G/A and G/G individuals (Fig. 3b). However, there were no differences in the frequency of detection of short chain fatty acids derived from non-conjugated precursors in patches from all three genotypes (Fig. 3c), with these acids being ubiquitous amongst the entire subject population. 3.7. Axillary microbiome analysis Analyses of the mean relative abundances of the major bacterial genera as a function of host genotype revealed a predominance of Staphylococcus, Corynebacterium and Anaerococcus, accounting for between 89 and 94% of the total axillary skin microbiome (Fig. 4a, Fig. S1). Other genera were also observed but with lower relative abundance and these accounted for <10% of the total population. A directional increase in the mean relative abundance of Staphylococcus and Anaerococcus was observed in the G/G and G/A genotypes, which was found to be statistically significant for Staphylococcus for all genotype comparisons, except G/A vs. G/G (Table 3). A corresponding statistical decrease in Corynebacterium mean relative abundance in the G/A and G/G genotypes was also observed for all genotype comparisons. Statistical differences in the mean relative abundance of a number of genera, i.e. Paracoccus, Streptococcus and Peptoniphilus between the different genotypes were also observed. Network-based analysis was used to visualise the microbial community structure of the different ABCC11 genotypes, to display the partitioning of microbial phyla [17]. Phylum-level operational taxonomic units (OTU’s) and human subjects were designated as nodes in a bipartite network, in which OTU’s are connected to the subjects in which their sequences are found. Visualisation of the datasets revealed a core microbiome that is shared by all three genotypes (Fig. 4b). The Proteobacteria, Firmicutes and Actinobacteria are the most prevalent phyla observed. In Fig. 4c, on the left hand side of the network the A/ A samples and OTU’s in A/A genotypes are observed, with the

Fig. 1. Concentrations of axillary malodour precursors in axillary wash samples of subjects with different ABCC11 genotypes. (a) Average HMHA-Gln levels across the different genotype groups. (b) Average 3M2H-Gln levels across the different genotype groups. (c) Average 3M3SH-Cys-Gly levels across the different genotype groups.

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30

27

Table 3 Mean relative abundance, and multinomial genotype comparisons for the predominant bacterial genera. p-Values in bold  0.05. Genera

Staphylococcus Corynebacterium Anaerococcus Acinetobacter Peptoniphilus Streptococcusa Paracoccus Micrococcus Pseudomonas a

Mean relative abundance

p-Value

AA

GG

GA

Omnibus

AA vs. GG

AA vs. GA

AA vs. GX

GG vs. GA

0.5355 0.3497 0.0357 0.0144 0.0034 0.0018 0.0041 0.0019 0.0052

0.6837 0.1766 0.0792 0.0199 0.0010 0.0006 0.0014 0.0037 0.0045

0.7386 0.0822 0.0659 0.0209 0.0058 0.0177 0.0018 0.0022 0.0030

0.0106 0.0000 0.2545 0.7988 0.0075 0.0000 0.0239 0.4155 0.5027

0.0159 0.0035 0.1013 0.5321 0.0561 0.1227 0.0073 0.1837 0.7871

0.0059 0.0000 0.2300 0.6349 0.3929 0.0112 0.0275 0.6756 0.2843

0.0024 0.0000 0.1121 0.5405 0.9652 0.0897 0.0069 0.3032 0.4762

0.3792 0.0400 0.6762 0.9470 0.0015 0.0000 0.5212 0.3091 0.3559

Significance driven by one high observation in the GA communities.

Fig. 2. (a) PCA scores plot of the metabolite profiles of the three different genotypes, G/G, G/A, & A/A and pooled sample. (b) Corresponding loadings plot from PCA analysis of the three genotypes. (c) PLS-DA scores plot of the metabolite profiles, G/G, G/A and & A/A. (d) Corresponding loadings plot from OPLS-DA analysis of the AA, GA and GG genotypes.

axillary malodour remains largely anecdotal and to date has not been robustly interrogated [6,18]. In our study, homozygous subjects for the non-functional A allele produced significantly lower levels of axillary malodour compared to those with a functional protein (G/A and G/G). However, the perception-based measures still detected appreciable amounts of axillary malodour in the A/A cohort, at levels that would be expected to be perceivable to other individuals with normal levels of olfactory acuity. This finding is at odds with the current view that A/A individuals are largely nonodourous [5,6], but is in agreement with earlier reports which found that Asian subjects do generate appreciable levels of axillary odour, although the quantitative and qualitative differences observed between subjects in these studies may have been influenced by their ABCC11 genotype [19,20]. This is the first study of its kind to evaluate the levels of axillary malodour in A/A individuals and directly compare them to the axillary malodour levels in G/A and G/G subjects. In the current study, the perceived differences in odour intensities were due to A/A subjects producing lower levels of the amino acid conjugate precursors than those carrying the G allele, indicating an active role for the ABCC11 efflux pump in the secretion of these

precursors from axillary apocrine glands. Although there were no significant differences in the precursor levels between G/G and G/ A subjects, there was a trend for G/G subjects to exhibit higher precursor concentrations. This may indicate a potential gene dosage effect, with the heterozygotes expressing lower levels of functional protein, resulting in lower amounts of precursors being secreted. These differences in precursor concentrations did not result in significantly lower levels of axillary malodour between G/A and G/G subjects. The A/A genotype did not result in the complete absence of conjugate precursors, indicating that the SNP (R180) variant does not result in a complete loss of function, but displays an extremely low level of ABCC11 activity. This complements the known molecular consequence of this SNP at the level of the protein, where some low level of activity would be expected [3]. A metabolomic-based approach was successfully applied to differentiate two genotype classes based on the metabolite profiles of axillary wash samples. The key biomarkers driving discrimination between the two identified groups (A/A vs. GA/GG) included the Na-acyl-glutamine conjugates HMHA-Gln and 3M2H-Gln, providing further evidence that ABCC11 plays a key role in the transport

28

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30

Fig. 3. Two-dimensional gas chromatography coupled with time-of-flight mass spectrometry analysis of axillary skin volatile organic compounds sampled on PDMS patches. (a) Frequency of detection of sulfanylalkanols from subjects with different ABCC11 genotypes, 3-mercapto-3-methylhexan-1-ol, & 2-methyl-3mercaptobutan-1-ol, 3-mercapto-2-methyl-pentan-1-ol and 3-mercaptohexan-1-ol. (b) Frequency of detection of 3M2H from subjects with different ABCC11 genotypes. (c) Frequency of detection of various short chain fatty acids derived from presumed non-conjugated sources, from subjects with different ABCC11 genotypes, butyric acid, isovaleric acid, valeric acid and caproic acid.

of these odour precursors from axillary apocrine glands to the skin surface. Although a role for ABCC11 in the secretion of conjugated odour precursors from axillary apocrine glands has been demonstrated [5], the experimental design and subsequent conclusion that a functional ABCC11 allele is essential for axillary malodour formation, presupposed that no other mechanism for axillary malodour production in the homozygous A/A subjects existed.

However, the conjugated precursors do not act as the sole sources of axillary malodour, since even in their absence appreciable levels of axillary malodour are observed. This finding suggests that other biochemical pathways are involved in malodour production and provide a significant contribution to the overall level of axillary malodour, i.e. the biotransformation of branched aliphatic amino acids and partial catabolism of methyl-branched long chain fatty acids, both pathways producing short chain volatile fatty acids [15,21]. Direct measures of the odourous volatiles present in the axillary headspace confirmed that the prevalence of branched and straight chain acids was the same across all three genotypes. The presence of these acids in all subjects represents a residual level of odour irrespective of genotype that probably accounts for the majority of axillary odour detected in A/A individuals [19]. Differences in the presence of key odourants responsible for axillary malodour across the three genotypes help to explain previous general observations that certain Asian populations tend to exhibit a slightly acidic/cheesy axillary odour, compared to a more intense meaty/oniony smell proposed to originate from individuals of a more Western origin [19]. The microbiota of the axilla is known to play a key role in the generation of malodour [8,9]. The axillary microbiome profiles in the current study were dominated by Staphylococcus, Corynebacterium and Anaerococcus, findings which are in agreement with previous data [22,23]. Interestingly, the bacterial diversity of the three genotypes demonstrated some significant differences in the mean relative abundance of Staphylococcus and Corynebacterium, with Staphylococcus statistically higher in abundance, and Corynebacterium statistically lower in abundance in individuals carrying at least one G allele compared to A/A subjects, a finding which has not been demonstrated previously. This result was slightly surprising given the fact that in the A/A genotype the conversion of aliphatic amino acids into short chain volatile fatty acids appears to be the major biochemical route to odour formation, with Staphylococcus being the key organism implicated in this reaction [15], but being least abundant in A/A individuals. The data also suggests that in the G/A and G/G subjects there were still sufficient numbers of Corynebacteria present to convert the amino acid conjugates to odourant acids and thiols. The role these differences in the microbial community structure have on odour intensity levels in the different genotypes is unclear, but would be expected to be secondary compared to the role altered levels of conjugated odour precursors play. Humans have distinct odour profiles that are in part coded by an individual’s inherited complement of immune-system genes, notably those for human leucocyte antigens (HLA’s) [24]. Differences in odours originating from sweat can be correlated with HLA type and it has been postulated that humans avoid mating with individuals with HLA patterns similar to their own [25]. However, the role of the ABCC11 gene in olfactory driven HLA selection appears minimal, since it is unlikely that A/A individuals selectively seek out G/A or G/G individuals as mates [26], even though this would signify some degree of genetic distance between the two partners. Indeed, considering the speed of the dissemination of the A allele throughout parts of Asia [4], the available evidence would suggest that the opposite is true. Olfactory perception varies considerably from one individual to another in odour sensitivity, hedonic and semantic processing, and high olfactory cognition [27]. Ratings of odour pleasantness and acceptability are modulated to a large degree by experience during childhood, adolescence and into adulthood, which in turn are strongly influenced by the cultural and environmental context of the individual [28]. Across numerous cultures human body odours evoke negative social attitudes [29,30], to the extent that a lack of body odour has become a sign of health and high social standing in many societies [26]. The proximal basis for this strong disliking of

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30

29

Fig. 4. (a) Mean relative abundance of the 9 most predominant bacterial genera observed across all three genotypes. (b) Phylum-level (OTU’s) and human subjects were designated as nodes in a bipartite network, in which OTU’s are connected to their subjects in which their sequences were found. Squares are samples coloured by genotype, A/ A (orange), G/G (purple) and G/A (green). Lines indicate that an OTU was found in a given sample. Diamonds indicate OTU’s present in a sample and are coloured by phylum (OTU’s are not sized in proportion to their counts). All genotypes cluster centrally and demonstrate little or no differentiation, indicating no major differences in their microbiota. (c) Left hand side, A/A samples and connected OTU’s that occurs uniquely in the A/A genotype. Right hand side, G/A, G/G samples and connected OTU’s that occur uniquely in G/A and G/G genotypes. Centre, OTU’s that occur in both groups/all three genotypes.

human body odour derives from social learning, since the odour itself is not harmful and does not signal a potential threat to health, as is the case with some odours, i.e. decomposing food or those of faecal origin [24]. The rapid spread of the A allele may therefore be a consequence of positive mate selection, due to reduced levels of body odour presenting olfactory cues signalling a healthy individual and potentially one of greater social standing. Indeed, attractiveness of body odour is known to correlate with genetic quality of prospective mates, including; body symmetry [31], facial attractiveness [32], attractiveness of nonverbal behaviour [33] and with psychometric dominance [34]. Homozygous A/A individuals are far less likely to adopt certain personal hygiene habits, i.e. daily use of a deodorant to combat axillary odour compared to those carrying at least one G allele, suggestive of a high level of self awareness in both groups regarding their absolute levels of axillary odour. This corroborates a recent finding where A/A individuals living in the United Kingdom were five times less likely to use a deodorant than those carrying the G allele [35]. The residual malodour levels caused by volatile fatty acids derived from sources other than conjugated precursors appear not to prompt behaviours that drive some individuals to control this particular odour. In our study, A/A individuals were only approximately two times less likely to use a deodorant/antiperspirant compared to the rest of the population, suggesting these subjects may use such products to derive alternative benefits, i.e. sweat control. Genes underpinning quantitative behavioural phenotypes are relatively rare [36]. However, in this instance ABCC11 appears to play an indirect role in determining whether or not an individual is likely to use an underarm deodorant on a daily basis. This will of course also be governed by environmental influences, particularly

as deodorant use is relatively widespread in Filipino society. However, the fact that a single SNP plays such a large role in governing this specific aspect of human behaviour is quite a novel finding. The perception of a person’s own body odour leads to a slightly stronger and faster brain response, compared to the perception of body odour derived from someone else [37]. Suggesting that self and non-self body odours are processed differently and that the information conveyed by one’s own odour may be more important than that of others. The degree to which these signals drive both conscious and unconscious behaviours remains to be elucidated. But the fact that ABCC11 is so influential on determining this particular human behaviour is relatively rare, but not completely unique [35,38].

Funding sources All authors (except JMK) were employees of Unilever R&D during the collection of data and preparation of this manuscript, JMK is employed by Waters plc. and GPC is employed by Clinical Trial Management & Testing Associates. The funders played no direct role in the collection of data or preparation of this manuscript. Acknowledgements The authors would like to thank Clive Harding for critical review of this manuscript. We also thank Peta Jones and Teresa Robinson (Unilever R&D) for technical assistance in performing various aspects of the clinical study. We also thank Prof. Neil Hall and Dr. Linda D’More (University of Liverpool) for help with

30

M. Harker et al. / Journal of Dermatological Science 73 (2014) 23–30

pyrosequencing of the bacterial samples and Source Bioscience plc. Nottingham for sequencing of PCR products for subject genotype designation. Finally, we thank all subjects for their cooperation and participation in this study.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jdermsci. 2013.08.016. References [1] Chen ZS, Guo Y, Belinsky MG, Kotova E, Kruh GD. Transport of bile acids, sulfated steroids, estradiol, 17-b-D-glucuronide, and leukotriene C4 by human multidrug resistance protein 8 (ABCC11). Mol Pharmacol 2005;67:545–57. [2] Kruh GD, Belinsky MG. The MRP family of drug efflux pumps. Oncogene 2003;22:7537–52. [3] Toyoda Y, Sakurai A, Mitani Y, Nakashima M, Yoshiura KI, Nakagawa H, et al. Earwax, osmidrosis, and breast cancer: why does one SNP (538G4A) in the human ABC transporter ABCC11 gene determine earwax type? FASEB J 2009;23:2001–13. [4] Yoshiura K, Kinoshita A, Ishida T, Ninokata A, Ishikawa T, Kaname T, et al. A SNP in the ABCC11 gene is the determinant of human earwax type. Nat Genet 2006;38:324–30. [5] Martin A, Saathoff M, Kuhn F, Max H, Terstegen L, Natsch A. A functional ABCC11 allele is essential in the biochemical formation of human axillary odor. J Invest Dermatol 2010;130:529–40. [6] Preti G, Leyden JJ. Genetic influences on human body odor: from genes to the axillae. J Invest Dermatol 2010;130:344–6. [7] Gautschi M, Natsch A, Schro¨der F. Biochemistry of human axilla malodor and chemistry of deodorant ingredients. Int J Chem 2007;61:27–32. [8] Leyden JJ, McGinley KJ, Holzle E, Labows JN, Kligman AM. The microbiology of the human axilla and its relationship to axillary odor. J Invest Dermatol 1981;77:413–6. [9] Taylor D, Daulby A, Grimshaw S, James G, Mercer J, Vaziri S. Characterization of the microflora of the human axilla. Int J Cosmet Sci 2003;25:137–45. [10] Zeng X, Leyden JJ, Lawley HJ, Sawano K, Nohara I, Preti G. Analysis of characteristic odors from human male axillae. J Chem Ecol 1991;17:1469–92. [11] Natsch A, Derrer S, Flachsmann F, Schmid J. A broad diversity of volatile carboxylic acids, released by a bacterial aminoacylase from axilla secretions, as candidate molecules for the determination of human-body odor type. Chem Biodivers 2006;3:1–20. [12] Natsch A, Schmid J, Flaschmann F. Identification of odoriferous sulfanylalkanols in human axilla secretions and their formation through cleavage of cysteine precursors by a C-S lyase isolated from axilla bacteria. Chem Biodivers 2004;1:1058–72. [13] Troccaz M, Starkenmann C, Niclass Y, van de Waal M, Clark AJ. 3-Methyl-3sulfanylhexan-1-ol as a major descriptor for the human axilla-sweat odour profile. Chem Biodivers 2004;1:1022–35. [14] Emter R, Natsch A. The sequential action of a dipeptidase and a beta lyase is required for the release of the human body odorant 3-methyl-3-sulfanylhexan-1-ol from a secreted Cys-Gly-(S) conjugate by Corynebacteria. J Biol Chem 2008;283:20645–52. [15] James AG, Cox D, Worrall K. Microbiological and biochemical origins of human foot malodour. Flavour Fragr J 2013. http://dx.doi.org/10.1002/ffj.3136.

[16] Riazanskaia S, Blackburn G, Harker M, Thomas CLP. The analytical utility of thermally desorbed polydimethylsilicone membranes for in-vivo sampling of volatile organic compounds in and on human skin. Analyst 2008;133: 1020–7. [17] Ley RE, Hamady M, Lozupone CA, Turnbaugh PJ, Ramey RR, Bircher JS, et al. Evolution of mammals and their gut microbes. Science 2008;320:1647–51. [18] Inoue Y, Mori T, Toyoda Y, Ishikawa T, Mitani Y, Yoshimura Y, et al. Correlation of axillary osmidrosis to a SNP in the ABCC11 gene determined by the Smart Amplification Process (SmartAmp) method. J Plast Reconstr Aesthet Surg 2010;63:1369–74. [19] Hasegawa Y, Yabuki M, Matsukane M. Identification of new odoriferous compounds in human axillary sweat. Chem Biodivers 2004;1:2042–50. [20] Akutsu T, Sekiguchi K, Ohmori T, Sakurade K. Individual comparisons of the levels of (E)-3-methyl-2-hexenoic acid, an axillary odor-related compound. Chem Sens 2006;31:557–63. [21] Barzantny H, Brune I, Tauch A. Molecular basis of human body odour formation: insights deduced from corynebacterial genome sequences. Int J Cosmet Sci 2011;34:2–11. [22] Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, Knight R. Bacterial community variation in human body habitats across space and time. Science 2009;326:1694–7. [23] Grice EA, Kong HH, Conlan S, Deming CB, Davis J, Young AC, et al. Topographical and temporal diversity of the human skin microbiome. Science 2009;324:1190–2. [24] Stevenson RJ. An initial evaluation of the functions of human olfaction. Chem Sens 2010;35:3–20. [25] Zavazava N, Westphal E, Muller-Ruchholtz W. Characterization of soluble HLA molecules in sweat and quantitative HLS differences in serum of healthy individuals. Int J Immunogenet 1990;17:387–94. [26] Soo M, Stevenson RJ. The moralisation of body odor. Mankind Quart 2006;47:25–57. [27] Ponceleta J, Rincka F, Bourgeata F, Schaald B, Roubya C, Bensafia M, et al. The effect of early experience on odor perception in humans: psychological and physiological correlates. Behav Brain Res 2010;208:458–65. [28] Schaal B, Soussignan R, Marlier L, Kontar F, Karima IS, Tremblay RE. Variability and invariants in early odour preferences comparative data from children belonging to three cultures. Chem Sens 1997;22:212. [29] Chrea C, Valentin D, Sulmont-Ross C, Ly Mai H, Hoang Nguyen D, Abdi H. Culture and odor categorization: agreement between cultures depends upon the odors. Food Qual Prefer 2004;15:669–79. [30] Schleidt M, Hold B, Attili G. A cross-cultural study on the attitude towards personal odors. J Chem Ecol 1981;7:19–31. [31] Gangestad SW, Thornhill R. Menstrual cycle variation in women’s preferences for the scent of symmetrical men. Proc R Soc Lond B 1998;265:927–33. [32] Rikowski A, Grammer K. Human body odour, symmetry and attractiveness. Proc R Soc Lond B 1999;266:869–74. [33] Roberts SC, Kralevich A, Ferdenzi C, Saxton TK, Jones BC, DeBruine LM, et al. Body odor quality predicts behavioural attractiveness in humans. Arch Sex Behav 2011;40:1111–7. [34] Havlicek J, Roberts SC, Flegr J. Women’s preference for dominant male odour: effects of menstrual cycle and relationship status. Biol Lett 2005;1:256–9. [35] Rodriguez S, Steer CD, Farrow A, Golding J, Day INM. Dependence of deodorant usage on ABCC11 genotype: scope for personalized genetics in personal hygiene. J Invest Dermatol 2013. http://dx.doi.org/10.1038/jid.2012.480. [36] Ebstein RP, Israel S, Hong-Chew S, Zhong S, Knafo A. Genetics of human social behaviour. Neuron 2010;65:831–44. [37] Pause BM, Krauel K, Sojka B, Ferstl R. Body odor evoked potentials a new method to study the chemosensory perception of self and non-self in humans. Genetica 1999;104:285–94. [38] Sookian S, Burgueno A, Gianotti TF, Marillet G, Pirola CJ. Odor perception between heterosexual partners: its association with depression, anxiety, and genetic variation in odorant receptor OR7D4. Biol Psychol 2011;86:153–7.