Microbiological and biochemical origins of human axillary odour


Correspondence: A. Gordon James, Unilever Discover, Colworth Science Park, Sharnbrook, Bedford MK44 1LQ, UK. Tel.: +44 1234 222706; fax: +44 1234 222552; e-mail: gordon.james@unilever.com


The generation of malodour on various sites of the human body is caused by the microbial biotransformation of odourless natural secretions into volatile odorous molecules. On the skin surface, distinctive odours emanate, in particular, from the underarm (axilla), where a large and permanent population of microorganisms thrives on secretions from the eccrine, apocrine and sebaceous glands. Traditional culture-based microbiological studies inform us that this resident microbiota consists mainly of Gram-positive bacteria of the genera Staphylococcus, Micrococcus, Corynebacterium and Propionibacterium. Among the molecular classes that have been implicated in axillary malodour are short- and medium-chain volatile fatty acids, 16-androstene steroids and, most recently, thioalcohols. Most of the available evidence suggests that members of the Corynebacterium genus are the primary causal agents of axillary odour, with the key malodour substrates believed to originate from the apocrine gland. In this article, we examine, in detail, the microbiology and biochemistry of malodour formation on axillary skin, focussing on precursor–product relationships, odour-forming enzymes and metabolic pathways and causal organisms. As well as reviewing the literature, some relevant new data are presented and considered alongside that already available in the public domain to reach an informed view on the current state-of-the-art, as well as future perspectives.


In the underarm (axilla), a large and permanent population of microorganisms thrives on secretions from the eccrine, apocrine and sebaceous glands. Traditional culture-based microbiological studies inform us that this resident microbiota consists mainly of Gram-positive bacteria of the genera Staphylococcus, Micrococcus, Corynebacterium and Propionibacterium (Leyden et al., 1981). It has been evident since the 1950s that malodour generation on axillary skin, as with other bodily sites, is due to the biotransformation of odourless natural secretions into volatile odorous molecules, by members of the resident microbial communities (Shelley et al., 1953). A study by Taylor et al. (2003), examining the relationship between microbial numbers and malodour intensities, supported the popular view that corynebacteria are the primary causal agents of underarm odour. As discussed in the following sections, this study further probed the Corynebacterium population of axillary skin using the genetic technique of 16S rRNA gene sequencing, although bacteria subjected to this were initially isolated by culturing on a selective medium. Until relatively recently, our understanding of the microbial communities on human skin sites was based wholly on traditional culture methods, which rely on the cultivability of genera in the selected media and cannot normally identify individual species without further biochemical or genetic analysis. Increasingly, however, culture-independent metagenomic approaches are being adopted to probe the skin microbiota, including in the axilla (Costello et al., 2009; Grice et al., 2009; Egert et al., 2011), and with the advent of high-throughput DNA sequencing technologies, in particular pyrosequencing (Margulies et al., 2005), the breadth and depth of data generated is increasing massively.

As this review will highlight, a consensus has emerged that volatile fatty acids (VFAs) and thioalcohols are the primary causal molecules of axillary malodour. In the recent past, 16-androstene steroids, which undoubtedly are present in axillary apocrine sweat, have been heavily implicated (Gower et al., 1994) and indeed are still highlighted as such in some current literature (Kanlayavattanakul & Lourith, 2011). However, we will argue later, on the basis of a study by Austin & Ellis (2003), as well as other evidence, that the contribution of steroids to axillary malodour is probably much less then previously believed. The involvement of short-chain (C2-C5) VFAs in axillary malodour has long been acknowledged, and various metabolic routes to these, and some structurally related medium-chain acids, from substrates readily available on human skin, have been elucidated (James et al., 2004a, b), again as discussed later. Since the early 1990s, structurally unusual medium-chain (C6–C10) VFAs, in particular the trans (E) isomer of 3-methyl-2-hexenoic acid (3M2H), have been shown to contribute to underarm odour (Zeng et al., 1991). Spielman et al. (1995) argued that this branched, unsaturated VFA was carried to the skin surface noncovalently bound to two proteins, apocrine secretion odour-binding proteins 1 and 2 (ASOB1 and ASOB2), with ASOB2 subsequently identified as apolipoprotein D (apoD), a member of the lipocalin family of carrier proteins (Zeng et al., 1996). Subsequently, however, Natsch et al. (2003) showed that 3M2H, and the structurally related 3-hydroxy-3-methylhexanoic acid, are covalently bound to l-glutamine residues in apocrine secretion and released by the action of a corynebacterial enzyme, Nα-acylglutamine aminoacylase, the encoding gene for which was cloned and expressed during the study. The same group later reported that a wide range of medium-chain VFAs are present in apocrine sweat as l-glutamine conjugates, released on skin by corynebacterial Nα-acylglutamine aminoacylase, with 3-hydroxy-3-methylhexanoic acid and 3M2H being the dominant species in terms of relative abundance (Natsch et al., 2006). Natsch and colleagues went on to investigate, with some success, whether a common pattern of these VFAs might contribute to an inherited individual-specific body odour type (Kuhn & Natsch, 2009; Natsch et al., 2010). However, the most significant recent breakthrough in establishing a genetic link to human body odour, as discussed later, was the discovery that a functional ABCC11 allele is essential for the secretion of the l-glutamine conjugates of 3M2H and 3-hydroxy-3-methylhexanoic acid, as well as the l-cysteinylglycine conjugate of the thioalcohol 3-methyl-3-mercaptohexan-1-ol, from the axillary apocrine gland (Martin et al., 2010).

The prominent role of thioalcohols in axillary malodour emerged in the public domain in three simultaneous publications in the mid-2000s (Hasegawa et al., 2004; Natsch et al., 2004; Troccaz et al., 2004). In particular, these studies implicated 3-methyl-3-mercaptohexan-1-ol as a major contributor to underarm odour, although in one report (Natsch et al., 2004) three additional thioalcohols were also identified, including 3-mercaptohexan-1-ol and 2-methyl-3-mercaptobutan-1-ol, validating work carried out by ourselves, as discussed later. Natsch et al. (2004) also proposed that the precursors of these malodorants are S-hydroxyalkyl-l-cysteine conjugates, enzymatically cleaved on skin by bacterial carbon–sulphur (C–S) β-lyases. In a repeat of their pioneering work on Nα-acylglutamine aminoacylase (Natsch et al., 2003), this group cloned, sequenced and heterologously expressed the gene encoding a corynebacterial C–S β-lyase. Again, our internal studies support the Natsch et al. (2004) findings, although we have also recorded C–S β-lyase activity in axillary Staphylococcus isolates, validating work published by Troccaz et al. (2008), as reviewed in the following sections. An alternative hypothesis was proposed by Starkenmann et al. (2005), who postulated that the direct thioalcohol precursors are S-hydroxyalkyl-l-cysteinylglycine conjugates, which are cleaved, in particular, by Staphylococcus haemolyticus. However, while accepting that l-cysteinylglycine conjugates are the true thioalcohol substrates, Emter & Natsch (2008) demonstrated that, at least in the case of the investigated axillary strain, the sequential action of a corynebacterial dipeptidase and C–S β-lyase is required for the release of 3-methyl-3-mercaptohexan-1-ol from its secreted dipeptide precursor.

In this article, we examine, in detail, the microbiology and biochemistry of odour formation on axillary skin, based on published work by ourselves and others. As well as reviewing this literature, some relevant new data are presented, mainly concerning thioalcohol-based malodour, and considered alongside that already available in the public domain to reach an informed view on the current state-of-the-art, as well as future perspectives. While the area has been reviewed recently by Kanlayavattanakul & Lourith (2011) and Barzantny et al. (2012a), neither publication does full justice to the significant recent advances made in our understanding of the microbiological and biochemical origins of axillary malodour. The former has a strong focus on the treatment of body odours, encompassing both current and potential future active ingredients for combating axillary and foot malodour. The review by Barzantny et al. (2012a) is scientifically rigorous, while relying heavily on insights derived from bioinformatic analyses of the whole-genome sequences of Corynebacterium species. However, much of it is focussed on fatty acid β-oxidation, a metabolic pathway previously heavily implicated in axillary malodour (James et al., 2004a), but whose contribution we are now questioning, as outlined later. The focus of the current article is on defining precursor–product relationships, and identifying the principal odour-forming enzymes, pathways and causal organisms.

The axillary microbiota

Pioneering studies undertaken in the early 1950s demonstrated that the generation of malodour on underarm skin can be attributed mainly to the action of the resident microbiota on odourless natural secretions from the apocrine gland, an exocrine organ unique to a few bodily locations including the areolae, genitalia and external auditory meatus (ear canal), as well as the axilla (Shelley et al., 1953). Extensive studies on the axillary microbiota and its relationship with underarm odour have been reported, in particular, by Leyden et al. (1981) and Taylor et al. (2003). While the earlier publication now appears somewhat dated, with its reference to defunct taxonomic terms such as ‘diphtheroid’, and historical misclassifications such as the assignment of the Staphylococcus genus within the Micrococcaceae family, it warrants re-examination in the light of our current understanding of the problem, over 30 years later. Leyden et al. (1981) used traditional culture methods to quantitatively determine the axillary microbial communities of over 200 subjects and correlated the results with subjectively determined odour quality. They established that the microbiota consists mainly of ‘Micrococcaceae’ (predominantly Staphylococcus species), ‘lipohilic and large-colony aerobic diphtheroids’ (predominantly Corynebacterium species) and Propionibacterium species. Significantly higher numbers of both total bacteria, and in particular corynebacteria, were associated with a ‘pungent, apocrine’ odour quality, while high numbers of staphylococci correlated with a ‘faint acid, nonapocrine’ odour quality. This was backed up by in vivo experiments where cultured axillary bacteria were incubated with apocrine sweat (also sampled from the axilla) on the volar forearm. Only corynebacteria generated ‘apocrine odour’, while staphylococci produced an ‘acid odour’ attributed to the VFA isovaleric acid. Overlaying our current level of knowledge on the origins of underam odour, as discussed in the sections below, it is tempting to translate the results of Leyden et al. (1981) as indicative of causal links between (1) corynebacteria and thioalcohol and/or medium-chain (C6–C10) VFA malodorants, and (2) staphylococci and the contribution of short-chain (C2–C5) VFAs.

The more recent report by Taylor et al. (2003) on the microbial diversity of axillary skin, determined using a range of selective and nonselective culture media, is broadly supportive of the Leyden et al. (1981) findings. In total, 61 axillae were sampled from 36 volunteers, and a large variation in microbial colonisation was observed, as summarised in Table 1. In general, however, axillae were either dominated by Staphylococcus species or aerobic coryneforms (predominantly Corynebacterium species), although in a minority of individuals, the dominant colonisers were Propionibacterium species. When microbial counts were compared with subjectively determined malodour intensities for individual axillae at the time of sampling, highly significant associations (P < 0.0001) were found in particular for total aerobes and aerobic coryneforms (essentially Corynebacterium species). A significant association was also found between Micrococcus species and malodour intensity, but the low prevalence and density of this genus probably excludes it from a contributory role (Table 1). This result may be explained by the ability of micrococci to fully metabolise some VFA-based malodorants (James et al., 2004a, b) such that these bacteria thrive in the presence of increased VFA levels, but not at sufficient densities to impact on malodour intensity. No association with malodour was found for staphylococci, propionibacteria or Malassezia fungi (Taylor et al., 2003).

Table 1. Microbial ecology of axillary skin (n = 61) and its relationship to malodour intensity (adapted from Taylor et al., 2003)
Genus/typePrevalence (%)Density range (log10 CFU cm−2)Mean density (log10 CFU cm−2)Association with malodour
Staphylococcus 98.41.72–7.005.76No
Aerobic coryneforms93.40.72–7.516.07Yes (P < 0.0001)
Micrococcus 45.00.72–4.723.19Yes (P < 0.05)
Propionibacterium 87.50.72–7.075.36No
Malassezia 21.70.72–2.671.20No

With the culture-based microbial ecology data providing further evidence of the pivotal role of corynebacteria in malodour generation, a study was undertaken to further characterise these organisms, using the genetic technique of 16S rRNA gene sequence analysis (Taylor et al., 2003). Initially, an example was taken of the predominant colony type on aerobic coryneform-selective agar plates from 13 individuals. The 16S rRNA gene sequence was obtained for each of these isolates, and directly compared with sequences lodged on the ribosomal database project (RDP) website (http://rdp.cme.msu.edu/). The highest similarity match for the majority (9) of these was Corynebacterium sp. G-2 CDC G5840, while the remainder closely matched Corynebacterium mucifaciens DMMZ 2278. These results were confirmed in a subsequent large-scale study on a further 61 isolates from colonies sampled from aerobic coryneform-selective plates. The majority of these (67%) were again a close match for Corynebacterium sp. G-2 CDC G5840, while the next most dominant species match was C. mucifaciens DMMZ 2278, and examples were also found of Corynebacterium afermentans, Corynebacterium amycolatum, Corynebacterium genitalium, Corynebacterium riegelii and Corynebacterium striatum (Taylor et al., 2003). We have recently undertaken a re-analysis of the Corynebacterium sp. G-2 CDC G5840 16S rRNA gene sequence on RDP and found that it is an extremely close match with a group of organisms designated as Corynebacterium tuberculostearicum (96.7–99.2% similarity), with the closest match to the type strain CIP 107291.

Of the organisms subjected to rRNA gene sequence analysis, only the 13 in the initial study were also analysed biochemically for their ability to metabolise fatty acids (Taylor et al., 2003), this being the determinant of the A/B phenotyping of corynebacteria that we previously linked to their malodour-generating ability (James et al., 2004a). This established that none were capable of catabolising long-chain fatty acids, and all were thus members of the nonodiferous Corynebacterium (B) subgroup, although as highlighted later, we are now questioning the importance of this A/B classification of corynebacteria, and the production of VFAs via fatty acid metabolism by the (A) subgroup, in axillary malodour. Nonetheless, in a separate study of 19 fatty acid-catabolising Corynebacterium (A) strains, isolated from the axillae of two individuals, the majority of the rRNA gene sequences obtained most closely matched C. amycolatum NCFB 2768, although examples were also found with a high degree of similarity to Corynebacterium suicordis and Corynebacterium sp. 59614. In a further analysis of their odour-forming capability, each of these isolates was also shown to possess appreciable levels of C–S β-lyase activity, although none matched the level of activity of Corynebacterium jeikeium NCIMB 40928 (subgroup A) or three separate strains of C. tuberculostearicum (subgroup B), including the type strain CIP 107291 and two axillary isolates. It is tempting to infer from these results that C. tuberculostearicum represents the dominant Corynebacterium (B) species on axillary skin, and C. amycolatum the principal representative of subgroup (A). However, more data would be required to validate or refute such a statement, and anyway, as discussed later, we now doubt the importance of this A/B phenotyping of axillary corynebacteria in the context of malodour. Nevertheless, there is a need for metagenomic approaches to be adopted to characterise the axillary microbiota to a higher resolution. While the studies described earlier and by Taylor et al. (2003) employed a genetic approach, 16S rRNA gene sequencing, to probe the Corynebacterium population of axillary skin, organisms subjected to this had originally been isolated by culturing on a selective medium. However, completely culture-independent approaches are now being applied to study the human skin microbiota, including in the axilla (Costello et al., 2009; Grice et al., 2009), and with the advent of high-throughput DNA-sequencing technologies, the breadth and depth of data generated is increasing massively. Of particular relevance to the microbial world has been pyrosequencing (Margulies et al., 2005) which enables up to a million reads per run (each of up to 1000 bp) to be obtained in < 24 h at relatively low cost (454 Life Sciences, http://my454.com/), although this is now being rivalled by a new wave of high-throughput platforms such as sequencing-by-synthesis (Bennett, 2004) (Illumina, http://www.illumina.com/systems/sequencing.ilmn) and nonoptical semiconductor sequencing (Rothberg et al., 2011) (Ion Torrent, http://www.iontorrent.com/) that provide similar read lengths at equivalent or lower cost. Applying these technologies, it is possible to obtain a high-resolution picture of microbial communities by a culture-free metagenomic approach, either directly by shotgun sequencing or with polymerase chain reaction (PCR) amplification of a phylogenetically relevant target gene (for bacteria, usually 16S rRNA gene), normally termed microbiomics (Nelson & White, 2010). Similarly for microbial whole-genome sequencing projects, costs have reduced and throughput increased to levels unimaginable only a few years ago. Major initiatives such as the US National Institutes of Health-sponsored Human Microbiome Project (HMP, http://www.hmpdacc.org/) and the European Commission-funded MetaHit project (http://www.metahit.eu/) are increasing our understanding of the diversity of the gut and skin microbiota and identifying many new candidate organisms for whole-genome sequencing. While the latter initiative is wholly focussed on the human intestinal microbiota, HMP includes skin as a target site and has thus been of great value to the skin microbiology community in the postgenomic era.

Culture-independent studies on the bacterial microbiota of human skin appear to confirm the traditional view that the dominant genera on most surface sites are Propionibacterium, Staphylococcus and Corynebacterium (Costello et al., 2009; Grice et al., 2009). However, other taxa are undoubtedly present and, at the species level in particular, even within these most prevalent genera, the diversity is much greater than culture-based methods had revealed (Grice et al., 2009), although species-level discrimination remains a significant challenge. Specifically in relation to the axillary microbiota, early microbiomics studies suggest it is dominated by members of the Staphylococcus genus (Firmicutes phylum) and Propionibacterineae (probably Propionibacterium species) and Corynebacterineae (probably Corynebacterium species) suborders (Actinobacteria phylum), as well as members of the Micrococcineae suborder (probably Micrococcus species), although the presence of taxa not previously indicated by culture-based methods is also evident, notably additional members of the Firmicutes and Actinobacteria phyla (Costello et al., 2009; Egert et al., 2011). Of particular interest among the Firmicutes are representatives of a heterogeneous group of organisms traditionally classified as Gram-positive anaerobic cocci (GPAC), notably members of the Anaerococcus and Peptoniphilus genera, which are present on multiple skin sites (Costello et al., 2009) and were recently confirmed as significant components of the resident axillary skin microbiota (Egert et al., 2011). While these obligate anaerobes are commonly isolated from clinical specimens (Song et al., 2007), they had not previously been reported as commensal inhabitants of normal human skin. Future challenges stemming from these early microbiomics studies include defining axillary skin diversity at the species-level, for example, to determine whether the results of the combined culture-based and genetic studies described earlier and by Taylor et al. (2003) on dominant Corynebacterium species can be validated by an exclusively culture-free approach. A further challenge will be to establish what contribution, if any, GPAC such as Anaerococcus and Peptoniphilus species make to axillary malodour.

Steroid-based malodour

Odorous steroids, in particular the 16-androstenes 5α-androstenol and 5α-androstenone, are known to be present in apocrine sweat, were previously heavily implicated in axillary malodour (Gower et al., 1994) and indeed still form the basis of some current literature (Kanlayavattanakul & Lourith, 2011). An extensive study on the biochemical origins of 16-androstenes on axillary skin, including a re-assessment of their contribution to malodour was reported by Austin & Ellis (2003). This showed that axillary bacteria can only produce 16-androstenes from precursors already containing the C16 double bond. In incubations with one such substrate, 5,16-androstadien-3-ol, some mixed populations of axillary corynebacteria, particularly the (A) subgroup as defined later, generated many different steroid metabolites identifiable by gas chromatography (GC) with mass spectrometry (MS), including 5α-androstenol and 5α-androstenone, which were assembled into a proposed metabolic map of axillary 16-androstene biotransformations (Austin & Ellis, 2003). When individual Corynebacterium strains were isolated and individually incubated with 5,16-androstadien-3-ol, a few were capable of efficient, rapid reactions, but no single isolate could carry out the full complement of biotransformations observed with the mixed cultures. A key observation was that very few organisms capable of 16-androstene transformations were recovered from axillary skin. For example, from a panel of 21 individuals, only 4 of 18 mixed corynebacteria populations, and 4 of 45 Corynebacterium isolates could metabolise 5,16-androstadien-3-ol. Of course, it is possible that both we and others have thus far failed to isolate the most active bacteria or identify the true substrate. However, even the staunchest proponents of 16-androstenes have failed to produce any direct evidence connecting them to axillary malodour at the low levels (pmol cm−2) present physiologically (Gower et al., 1994). The link between these steroids and underarm odour originates from studies in the 1970s highlighting that the porcine pheromone 5α-androstenone is detectable in the human axilla (Claus & Alsing, 1976), the fact that it possesses a urine-like smell characteristic of axillary odour and the belief that the levels present would exceed its low olfactory threshold (Gower et al., 1994). However, there are no reports in the literature where GC or GC-MS studies incorporating organoleptic assessments have been used to implicate 16-androstenes in underarm odour, and in fact Zeng et al. (1991), who used such an approach, stated explicitly that this was not the case. In the light of this evidence, and the high levels of specific anosmia reported for 16-androstenes (Amoore, 1977), it is postulated that the contribution of these steroids to axillary malodour is probably much less than previously believed.

VFA-based malodour

It is generally accepted that short- (C2–C5) and medium-chain (C6–C10) VFAs are among the causal molecules of axillary malodour. The involvement of short-chain (C2–C5) VFAs in body odour has long been acknowledged, and various metabolic routes to these, and some medium-chain acids, from substrates readily available on human skin, were reported by James et al. (2004a, b). Propionibacteria and staphylococci were shown to ferment glycerol, from triacylglycerol hydrolysis, and lactic acid, naturally abundantly present on skin, to acetic and propionic acid. Furthermore, staphylococci are capable of converting branched aliphatic amino acids, such as l-leucine, to highly odorous short-chain (C4–C5) methyl-branched VFAs, such as isovaleric acid, which are traditionally associated with the acidic note of axillary malodour (Leyden et al., 1981). Amino acids are present in eccrine sweat, but may also be provided, on axillary skin, by the breakdown of proteins from the keratinising epidermis and apocrine secretion, by bacterial proteases (Holland, 1993). However, we proposed that the major metabolic route to short- and medium-chain VFAs in the axilla was the partial degradation (via β-oxidation) of structurally unusual, particularly methyl-branched long-chain fatty acids, known to be present in sebum (Nicolaides, 1974), by a subgroup (A) of the Corynebacterium genus, while the other subgroup, corynebacteria (B), are incapable of growth on fatty acids. This assertion was based on in vitro kinetic data showing higher rates of VFA generation by corynebacteria (A) than staphylococci (James et al., 2004b), as well as the significant in vivo association found between total corynebacterial (A + B) counts, but not staphylococcal numbers, and malodour intensity (Taylor et al., 2003). However, given what we now know about the probable role of axillary Corynebacterium species in the production of thioalcohols and structurally unusual medium-chain VFAs from conjugated amino acid precursors, as described later, molecules more likely to possess the ‘pungent, apocrine’ odour quality described by Leyden et al. (1981), it is timely for us to revise our position. Although corynebacteria (A) metabolise fatty acids at a higher rate than staphylococci do branched aliphatic amino acids, ecology studies inform us that, on average, < 10% of axillary Corynebacterium isolates are of the (A) subgroup (G. James, G. Mycock & D. Taylor, unpublished data), while the prevalence and density of the Staphylococcus genus is universally high across test populations (Leyden et al., 1981; Taylor et al., 2003). Additionally, sebum does not contain significant levels of any iso-methyl-branched, odd carbon number long-chain fatty acids to act as precursors for isovaleric acid, the most prominent of this class of VFA malodorants, via β-oxidation by corynebacteria (A) (Nicolaides, 1974), while l-leucine, the substrate for this VFA via staphylococcal amino acid metabolism, is abundant on the skin (Holland, 1993). While there is no direct association between staphylococcal numbers and malodour intensity in the axilla (Taylor et al., 2003), this does not preclude a role for these microorganisms in the production of some classes of odorant, much as described by Leyden et al. (1981) and discussed earlier. Our current position, therefore, is that the major metabolic route to short-chain VFAs in the axilla is the biotransformation of branched aliphatic amino acids by Staphylococcus species, with the l-leucine to isovaleric acid conversion being the most prominent.

While the metabolism of glycerol, lactic acid, amino acids and fatty acids can explain the origins of most short- and some medium-chain VFAs on axillary skin, these pathways cannot account for the highly structurally unusual medium-chain VFAs, such as 3M2H, described originally by Zeng et al. (1991), and latterly by Natsch et al. (2003, 2006). Initially, it was believed that these were transported to the skin surface noncovalently associated with two proteins in apocrine sweat, ASOB1 and ASOB2, and released on the skin surface, presumably due to the action of bacterial proteases (Spielman et al., 1995). ASOB2 was later shown to be synonymous with apoD, a member of the α-microglobulin superfamily of carrier proteins otherwise known as lipocalins (Zeng et al., 1996). This raised the intriguing possibility that human axillary odorants are transported in a similar fashion to some nonhuman mammalian odorants involved in chemical signalling, where lipocalins are heavily implicated, for example the major urinary proteins of rodents (Cavaggioni & Mucignat-Caretta, 2000). More recently, however, Natsch et al. (2003, 2006) demonstrated that 3M2H, the closely related and even more abundant 3-hydroxy-3-methylhexanoic acid, and relatively lower levels of a wide range of other structurally unusual medium-chain VFAs, are in fact present in apocrine sweat as covalently-bound conjugates of l-glutamine, and released on the skin surface by a corynebacterial enzyme, Nα-acylglutamine aminoacylase. The earlier of these studies also provided the first report of an enzyme involved in human body odour being purified and characterised, and its encoding gene cloned, sequenced and heterologously expressed. Troccaz et al. (2009) measured significant (mg L−1) levels of Nα-3-hydroxy-3-methylhexanoyl-l-glutamine in the axillary sweat of 49 volunteers over 3 years, providing further evidence in support of this amino acid conjugate as a key malodour precursor. There was a recent attempt to combine the two conflicting hypotheses on the origins of medium-chain VFAs in the axilla by postulating that 3-hydroxy-3-methylhexanoic acid was covalently bound to the N-terminal amino acid of apoD (Akiba et al., 2011). Indeed, this residue is predicted to be l-glutamine, after cleavage of the putative signal peptide at residues 1–20 of the apoD precursor (http://www.ncbi.nlm.nih.gov/protein/NP_001638.1). However, given the levels of Nα-3-hydroxy-3-methylhexanoyl-l-glutamine reported by Troccaz et al. (2009), a 1 : 1 ratio of apoD to 3-hydroxy-3-methylhexanoic acid would require nonphysiologically high (g L−1) levels of this protein in axillary sweat, a seemingly impossible scenario.

Studies undertaken in our laboratory offer a degree of additional support to the Nα-acylglutamine aminoacylase hypothesis. We have been able to isolate axillary bacteria, mainly Corynebacterium species, with the ability to hydrolyse both a model substrate, carbobenzyloxy-l-glutamine (Z-Gln), and a chemically synthesised equivalent to the putative physiological precursor of 3M2H, Nα-3-methyl-2-hexenyl-l-glutamine (Table 2), using essentially the same assay conditions as Natsch et al. (2003). However, the Nα-acylglutamine aminoacylase activity of these isolates is insufficient to explain the high levels of 3M2H we have detected in axillary samples by GC-MS. Various medium manipulations were carried out in an attempt to increase yields of benzyl alcohol from Z-Gln, ranging from simple buffer (3 g L−1 KH2PO4, 1.9 g L−1 K2HPO4) to complex broth [Tween 80-supplemented Tryptone soya broth (TSBT), as described by James et al. (2004b)], but none were successful (G. James & D. Cox, unpublished data). The only test organism with sufficient aminoacylase activity was Corynebacterium sp. Ax20, the strain described by Natsch et al. (2003, 2006) and also an axillary isolate, which efficiently hydrolyses both Z-Gln and Nα-3-methyl-2-hexenyl-l-glutamine (Table 2). It is possible that our internal studies have thus far failed either to isolate the most active bacteria, or to identify the optimum physiological conditions in vitro for expression of the enzyme, although in light of the highlighted medium manipulation study, the former would appear to be the more likely scenario. Corynebacterium sp. Ax20 was originally described as C. striatum (Natsch et al., 2003, 2004), but in a later publication, this speciation is omitted (Emter & Natsch, 2008). Genetic analysis of this microorganism in our laboratory failed to find an exact match for its 16S rRNA gene sequence, although it was more similar to Corynebacterium glaucum than C. striatum. Perhaps significantly, none of the genetically analysed axillary Corynebacterium isolates described by Taylor et al. (2003) matched C. glaucum, while only one matched C. striatum. While this alludes to the possibility that we have thus far failed to isolate the most aminoacylase-active bacteria by traditional culture methods, it again emphasises the importance of defining axillary skin diversity at the species-level by culture-independent means, as discussed previously.

Table 2. Nα-Acylglutamine aminoacylase activity of axillary bacteria, based on conversion of 2 mM carbobenzyloxy-l-glutamine (Z-Gln) to benzyl alcohol (BA) in a whole-cell biotransformation assay
GenusStrains testedAminoacylase-positive strainsActivity range (BA yield)
  1. a

    Corynebacterium sp. Ax20 (from A Natsch).

Corynebacterium 2781–26% (a99%)
Staphylococcus 3112–15%
Micrococcus 111–2%

Thioalcohol-based malodour

During studies undertaken in our laboratory, analysis of axillary samples by GC-MS with subjective ‘sniff port’ detection (GC-MS-Nose) led to the identification of a series of four putative thioalcohols involved in underarm odour. These represented two sets of isomers, of molecular weight 120 and 134 Da, respectively, with, in each case, one isomer possessing a meaty, onion-like smell, characteristic of axillary odour, and the other a less objectionable, occasionally fruity odour. The less offensive-smelling 134-Da isomer was identified as 3-mercaptohexan-1-ol, while the meaty, onion-like 120-Da species was confirmed as 2-methyl-3-mercaptobutan-1-ol. The presence of these thioalcohols on axillary skin, and their role in underarm odour, was initially reported by Natsch et al. (2004), who further identified 3-mercaptopentan-1-ol (probably equivalent to our less malodorous 120 Da isomer) and 3-methyl-3-mercaptohexan-1-ol (molecular weight, 148 Da). The latter was simultaneously identified and implicated in malodour by Troccaz et al. (2004) and Hasegawa et al. (2004), and there is a growing acceptance that thioalcohols, along with structurally unusual medium-chain VFAs such as 3M2H and 3-hydroxy-3-methylhexanoic acid, are the primary causal molecules of underarm odour.

Using the same rationale as Natsch et al. (2004), we deduced that the thioalcohol precursors would be S-hydroxyalkyl-l-cysteine conjugates, based on the existence of similar substrate–product relationships in other biological environments, notably some Vitis vinifera (grape vine) varieties and the aroma of their resultant wines (Tominaga et al., 1998). In support of this, we were able to identify the direct precursor to 2-methyl-3-mercaptobutan-1-ol, S-((2-hydroxy-1-isopropyl)ethyl)-l-cysteine, in axillary samples using GC-MS (R. Calvert, unpublished data). Similarly, we reasoned that such precursors, secreted in apocrine sweat, would be cleaved on the skin surface by bacterial C–S β-lyases. A whole-cell biotransformation assay system was developed to screen representative axillary bacteria for their ability to cleave a model C–S β-lyase substrate, S-benzyl-l-cysteine, as well as a chemically synthesised equivalent to the putative 2-methyl-3-mercaptobutan-1-ol precursor, S-((2-hydroxy-1-isopropyl)ethyl)-l-cysteine. The results demonstrate that both corynebacteria and staphylococci, but not micrococci, possess C–S β-lyase activity, though in the case of the Corynebacterium isolates, there were a greater proportion of positive strains, and the activities tended to be higher (Table 3). It had previously been reported that, within the axillary microbiota, only corynebacteria can cleave S-hydroxyalkylcysteines (Natsch et al., 2004), but clearly this is not the case, at least for a minority of Staphylococcus isolates. Indeed, a gene encoding the most likely candidate enzyme with S-hydroxyalkylcysteine β-lyase activity, namely cystathionine β-lyase (EC, was cloned from an axillary S. haemolyticus isolate and heterologously expressed by Troccaz et al. (2008). The recombinant enzyme was shown to cleave the l-cysteine and l-cysteinylglycine conjugates of 3-methyl-3-mercaptohexan-1-ol, but activity was very low, and no odour was generated on incubation with sterile axillary sweat. Recombinant cystathionine β-lyase had previously been cloned and heterologously expressed from Corynebacterium sp. Ax20 by Natsch et al. (2004) and shown to be active against the l-cysteine conjugates of various thioalcohols, as well as releasing these odorants on incubation with axillary sweat extracts.

Table 3. C–S β-Lyase activity of axillary bacteria, based on conversion of 0.5 mg mL−1 S-benzyl-l-cysteine to benzyl mercaptan (BM) in a whole-cell biotransformation assay
GenusStrains testedC–S β-lyase-positive strainsActivity range (BM yield)
Staphylococcus 17319–47%
Micrococcus 1100%
Corynebacterium 281915–83%

Starkenmann et al. (2005) proposed an alternative hypothesis for thioalcohol production on axillary skin, based on the glutathione detoxification pathway, providing evidence that the direct precursors are S-hydroxyalkyl-l-cysteinylglycines, derived from S-hydroxyalkylglutathione conjugates in the apocrine gland. This was supported in a further study by Troccaz et al. (2009) who measured significant (μg-mg L−1) levels of the putative 3-methyl-3-mercaptohexan-1-ol precursor, S-[1-(2-hydroxyethyl)-1-methylbutyl]-(l)-cysteinylglycine, in the axillary sweat of 49 volunteers over 3 years. Starkenmann et al. (2005) showed that a strain of S. haemolyticus isolated from axillary skin was more efficient at cleaving the l-cysteinylglycine conjugate of 3-methyl-3-mercaptohexan-1-ol than either Staphylococcus epidermidis or Corynebacterium xerosis. It is not clear whether cleavage is mediated by a single β-lyase enzyme (though presumably not cystathionine β-lyase, as discussed earlier (Troccaz et al., 2008)) or by a combination of a carboxypeptidase or dipeptidase and a C–S β-lyase, although the existence of a single enzymic step is the preferred postulate. However, while confirming that l-cysteinylglycine conjugates are the major secreted thioalcohol precursors, Emter & Natsch (2008) provided evidence that the sequential action of a corynebacterial dipeptidase and C–S β-lyase is required for release of 3-methyl-3-mercaptohexan-1-ol from its dipeptide precursor. Again, this group strengthened their case by cloning, sequencing and heterologously expressing a gene encoding, in this case, a novel metallo-dipeptidase from Corynebacterium sp. Ax20 (Emter & Natsch, 2008). Clearly, therefore, question marks remain over the exact origin of thioalcohols on axillary skin, particularly in relation to the biochemical route employed by staphylococci. It is not clear whether the secreted precursors are S-hydroxyalkyl-l-cysteinylglycines or a combination of these and l-cysteine conjugates, although there is a concensus that the former are more abundant. The causal bacteria and exact biochemical route also remain unresolved, with disagreement on whether an uncharacterised S. haemolyticus β-lyase or a combination of a corynebacterial dipeptidase and cystathionine β-lyase is the main contributor.

Initial work undertaken in our laboratory, using a whole-cell biotransformation assay, showed that, of the bacteria tested, C. jeikeium NCIMB 40928 possessed the highest C–S β-lyase activity, although this level of activity was later matched by three strains of C. tuberculostearicum, including the type strain CIP 107291 as well as two axillary isolates. C. jeikeium NCIMB 40928 was selected for a study aimed at cloning and expressing the gene responsible for this activity. In the absence of N-terminal sequence data, a bioinformatics exercise was undertaken to identify and group enzymes with potential C–S β-lyase activity, and facilitate the design of degenerate primers. This resulted in eight phylogenetically distinct protein groups, of which two, categorised as the MetC- and MalY-type β-lyases, were judged as being the most likely homologues of an axillary Corynebacterium C–S β-lyase. The MetC/MalY classification originated from Escherichia coli, which possesses two cystathionine β-lyases, one encoded by the traditional metC gene, and the other by malY, encoding a bifunctional protein (in this organism) also involved in regulating the uptake and metabolism of maltose (Zdych et al., 1995). Sequence similarity within each of these groups was relatively low (36% identity for the MetC group, 25% for MalY), but both mainly contained enzymes assigned as cystathionine β-lyase (EC, responsible for a key step in l-methionine biosynthesis. Degenerate primers were designed, based on areas of close homology within these groups, in an attempt to amplify regions of a corresponding C. jeikeium NCIMB 40928 gene. The MalY group included the putative cystathionine β-lyase of the whole-genome sequenced Corynebacterium species Corynebacterium glutamicum, and primers for this group were thus designed with a bias for the corresponding gene AAK69425 (http://www.uniprot.org/uniprot/Q93QC6). The putative C. jeikeium NCIMB 40928 cystathionine β-lyase gene was indeed cloned following amplification with one of these malY primers, while no such success was achieved with any of the metC primers. Subsequent sequence analysis of the Corynebacterium sp. Ax20 C–S β-lyase cloned and heterologously expressed by Natsch et al. (2004) confirmed that this too is a member of the MalY group, as are the putative cystathionine β-lyases of all whole-genome sequenced Corynebacterium strains, while the recombinant S. haemolyticus cystathionine β-lyase described by Troccaz et al. (2008) is of the MetC-type, along with the corresponding predicted enzymes of all whole-genome sequenced Staphylococcus strains. Recombinant polyhistidine-tagged C. jeikeium MalY C–S β-lyase was expressed in E. coli and purified to near-homogeneity by immobilized metal affinity chromatography. The recombinant protein migrated to the same point on a nondenaturing gel as the native NCIMB 40928 enzyme, and also exhibited C–S β-lyase activity by zymogram staining using the model substrate S-benzyl-l-cysteine (Fig. 1). Lysis of a chemically synthesised equivalent to the putative physiological 2-methyl-3-mercaptobutan-1-ol precursor, S-(3-hydroxy-1,2-dimethylpropyl)-l-cysteine, was also demonstrated in a liquid-phase enzyme assay based on the same double-dye system used in the zymogram stain (C. Austin, unpublished data). By an interesting coincidence, we later showed that it was 100% identical, at both the nucleotide and amino acid level, to the putative cystathionine β-lyase gene and protein of C. jeikeium K411 (http://www.uniprot.org/uniprot/Q4JWQ6), the first axillary-relevant Corynebacterium species to be whole-genome sequenced, by a group from University of Bielefeld, Germany (Tauch et al., 2005). In collaboration with ourselves, the Bielefeld group went on to show that the repressor protein McbR acts as a transcriptional regulator of malY and other genes involved in the biosynthesis of sulphur-containing amino acids (l-methionine and l-cysteine) in C. jeikeium K411 (Brune et al., 2010). Subsequently, this group adopted a bioinformatics-based approach to deduce the complete transcriptional regulatory network of C. jeikeium K411 and its relationship with metabolic pathways involved in malodour formation (Barzantny et al., 2012b). This followed on from a previous bioinformatics-based analysis of the molecular basis of axillary odour, based on insights derived from the functional annotation and metabolic reconstruction of multiple corynebacterial whole-genome sequences (Barzantny et al., 2012a). However, both articles have a heavy focus on fatty acid metabolism, a pathway previously heavily implicated in axillary malodour (James et al., 2004a), but whose contribution is now in doubt, as outlined previously.

Figure 1.

Characterisation of recombinant Corynebacterium jeikeium NCIMB 40928 MalY C–S β-lyase by native PAGE with activity staining, using S-benzyl-l-cysteine as the substrate. The recombinant enzyme preparation (and negative control) was diluted 1000-fold, and a C. jeikeium cell lysate 20-fold, each in sample buffer (Invitrogen), to approximately equilibrate the C–S β-lyase levels. The activity-stained native and recombinant enzymes can be visualised migrating to the same position on the gel.

The work carried out by ourselves on thioalcohol-based malodour, both in terms of the whole-cell assays and the malY cloning, tends to support the view that, regardless of whether the secreted precursors are S-hydroxyalkyl-l-cysteinylglycines or a combination of these and S-hydroxyalkyl-l-cysteines, the primary causal organisms are members of the Corynebacterium, rather than Staphylococcus genus. However, taken as a whole, the evidence remains equivocal, and while further biochemical studies would be of value, a species-level definition of axillary skin diversity, as discussed previously, would also be highly beneficial.

Human genetics and axillary odour

For many years, speculation has surrounded the existence of a genetic contribution to body odour, particularly in the axilla with its high concentration of apocrine glands (Shelley et al., 1953). This was recently investigated by Kuhn & Natsch (2009), building on their previous observation that such a wide range of VFAs are present in apocrine sweat as l-glutamine conjugates (Natsch et al., 2006). It was postulated that a common pattern of these VFAs would contribute to an inherited individual-specific body odour type, and indeed, a study conducted on 12 pairs of monozygotic human twins supported this hypothesis. A later attempt was made to link this genetically determined pattern of Nα-acylglutamines to polymorphic genes residing in the human leucocyte antigen complex, but the evidence here suggested this was not the case (Natsch et al., 2010). A genetic influence on human axillary odour that has been validated is a single nucleotide polymorphism (SNP) in the ABCC11 gene (538G→A), which had previously been demonstrated to be the determinant of human earwax type, a landmark discovery representing the first example of DNA polymorphism determining a visible phenotypic trait (Yoshiura et al., 2006). Martin et al. (2010) showed that the ABCC11 gene product, an apical efflux pump, is also responsible for the secretion of the l-glutamine conjugates of 3M2H and 3-hydroxy-3-methylhexanoic acid, as well as the l-cysteinylglycine conjugate of 3-methyl-3-mercaptohexan-1-ol, from the axillary apocrine gland. While the GG and GA genotypes correspond to high levels of these odorants and their amino acid conjugates in axillary extracts, the AA genotype leads to a near-complete loss of these malodorants and precursors in the underarm. The dominant G-allele prevails in populations of European and African origin, while the A-allelle is most frequent among East Asians, notably in China and Korea (Yoshiura et al., 2006). With intermediate frequencies also evident elsewhere in Asia, the selective pressures underlying this geographical distribution are unknown. Also lacking is a robust evaluation of the effect of this SNP on axillary odour intensity and quality, beyond anecdotal evidence that Caucasians and Africans generate a more intense malodour than many Asians, who apparently produce only a faint, acidic underarm odour (Martin et al., 2010).


Microbial sampling and maintenance

The microbial ecology of axillary skin was determined by sampling normal volunteers using the scrub-cup technique (Williamson & Kligman, 1965), and plating out serially diluted aliquots onto selective and nonselective culture media, as described by Taylor et al. (2003). A library of axillary bacteria (Corynebacterium, Brevibacterium, Propionibacterium, Staphylococcus and Micrococcus species), characteristic of the normal microbiota of axillary skin was established, including strains from the UK National Collection of Type Cultures (NCTC), the American Type Culture Collection (ATCC) and the UK National Collections of Industrial Food and Marine Bacteria (NCIMB), as well as isolates from the ecology studies. Corynebacterium sp. Ax20 was a gift from A Natsch (Givaudan). Bacterial isolates were classified to genus-level by their ability to grow on selective agar plates (Taylor et al., 2003), and where possible, to species-level by the use of API Staph, API Coryne and Rapid ID 32 A identification systems (bioMérieux). In some cases, axillary Corynebacterium isolates were further characterised by 16S rRNA gene sequencing analysis, as described previously (Taylor et al., 2003). Cultures were routinely grown and maintained on Tween 80-supplemented Tryptone soya broth (TSBT) or agar (TSAT) plates under aerobic or, for the propionibacteria, anaerobic conditions (James et al., 2004b).

Chemical sampling and analysis

Chemical analysis of axillary samples was carried out using GC with simultaneous MS and subjective ‘sniff port’ detection (Nose). GC-MS-Nose was carried out using a Hewlett Packard 5890 Series 11 gas chromatograph, fitted with an ATAS Optic 2 injector and HP-INNOwax (polyethylene glycol fused silica) capillary column (30 m × 0.53 mm; film thickness, 1 μm). After sample injection and chromatography (oven: 60 °C for 3 min; 60–240 °C at 3 °C min−1), separated components eluting from the column were split between a sniff port, for assessment by an odour judge, and a Hewlett Packard 5972 MSD mass spectrometer for chemical identification. Odour judges (n = 2) assessed each 1 h GC-MS-Nose run in two 30 min sessions – each sample was run twice with the judges alternating between the first and second session. The assessors were instructed to stand with their nose above the sniff port and indicate to a scribe the exact moment that they detected an odour. The scribe noted the time and asked the assessor to describe the aroma and whether it was characteristic of axillary malodour. Samples, obtained by the scrub-cup technique (Williamson & Kligman, 1965), were prepared for GC-MS-Nose by acidification, extraction in 1 volume of diethyl ether, and evaporation to 75 μL under a stream of nitrogen.

Microbial metabolism assays

Semi-synthetic media were developed for use in the in vitro assay systems employed to investigate odour formation by axillary bacteria, including the fatty acid biotransformation assay described previously by James et al. (2004a). The whole-cell C–S β-lyase assay consisted of screw-topped 7 mL glass tubes, to which were added 1–2 mL semi-synthetic medium (3.95 g L−1 Na2HPO4, 1.35 g L−1 KH2PO4, 0.5 g L−1 Casamino acids, 0.5 g L−1 MgSO4·7H2O), supplemented with 0.5 mg mL−1 S-benzyl-l-cysteine or S-((2-hydroxy-1-isopropyl)ethyl)-l-cysteine. Following inoculation with fresh bacterial biomass, pregrown in TSBT, to give starting optical densities (A590) of ~10.0, assays were incubated aerobically at 35 °C with agitation and analysed after 24 h. At the end of each assay incubation, culture viability was determined, either by BacLight (Life Technologies Ltd) live/dead ratio, or by total viable count analysis on TSAT plates. Yields of benzyl mercaptan or 2-methyl-3-mercaptobutan-1-ol were determined by GC using methodology similar to that described previously (James et al., 2004a, b). An identical protocol was employed for the whole-cell Nα-acylglutamine aminoacylase assay, except that the semi-synthetic medium was as described by Natsch et al. (2003) and the substrate was 2 mM carbobenzyloxy-l-glutamine (Z-Gln) or Nα-3-methyl-2-hexenyl-l-glutamine, resulting in formation of benzyl alcohol or 3M2H, respectively, both measurable by GC.

Cloning and expression of axillary Corynebacterium C–S β-lyase

A bioinformatics-based approach was adopted to identify and group enzymes with the potential to cleave β-C–S bonds, and facilitate the design of degenerate primers. Names of candidate enzymes were used to search the Swiss-Prot and TrEMBL protein databases (http://web.expasy.org/docs/swiss-prot_guideline.html), and the sequences obtained were aligned by clustalw (http://www.ebi.ac.uk/Tools/msa/clustalw2/) and submitted for BLAST database searching (http://blast.ncbi.nlm.nih.gov/). Alignment was assessed by studying the resultant phylogenetic tree of protein groupings. Two protein groups identified from the sequence alignment, tentatively labelled the MetC- and MalY-type β-lyases, each encoding cystathionine β-lyase (EC, were judged as being the most likely homologues of the axillary Corynebacterium lyase responsible for the cleavage of S-hydroxyalkyl-l-cysteine conjugates, and degenerate primers were designed, based on areas of close homology within these groups. Genomic DNA was extracted from C. jeikeium NCIMB 40928 using the PrepMan Ultra method (Applied Biosystems) and used as the template for degenerate primer PCR experiments. Following successful amplification of a region of NCIMB 40928 genomic DNA with a malY primer, several steps of genome walking were undertaken to determine the complete sequence of the putative C. jeikeium C–S β-lyase gene. This was then amplified and cloned into a pET-22 expression vector (Novagen), with a C-terminus polyhistidine-tag, which was used to transform competent E. coli BRL(DE3) cells (Novagen). Transformed cells were grown to an optical density (A590) of 0.6–0.8, harvested and lysed on ice by a combination of enzyme treatment (lysozyme) and sonication. Putative polyhistidine-tagged recombinant C–S β-lyase was purified from the cell lysate by immobilized metal affinity chromatography using Ni-NTA (Qiagen) and characterised by native and denaturing polyacrylamide gel electrophoresis (PAGE). C–S β-lyase activity was detected on native PAGE gels by zymogram staining, with S-benzyl-l-cysteine as the substrate and a double-dye system of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) and phenazine methosulfate (PMS), which react to form a purple precipitate in the presence of benzyl mercaptan.

Conclusions and future perspectives

In this article, we have reviewed, in detail, the microbiology and biochemistry of malodour formation on axillary skin, based on published work by ourselves and others, as well as some relevant new data. It is evident from traditional culture-based studies that the microbiota of axillary skin is usually dominated by staphylococci or corynebacteria, or occasionally propionibacteria, of which only the corynebacteria are significantly associated with malodour (Leyden et al., 1981; Taylor et al., 2003). Based on 16S rRNA gene sequencing, the most abundant Corynebacterium species isolated from underarm skin were C. tuberculostearicum, C. mucifaciens and C. amycolatum, but there is a need for more culture-independent metagenomic studies to be undertaken on the the axillary microbiota, particularly to define diversity at the species-level. Axillary microbiomics studies published thus far have confirmed the prevalence of staphylococci, corynebacteria and propionibacteria, but also revealed the presence of taxa not previously indicated by culture-based methods, notably GPAC such as Anaerococcus and Peptoniphilus species (Costello et al., 2009). It remains unknown what contribution, if any, these organisms make to axillary malodour.

A consensus has emerged that short- (C2–C5) and medium-chain (C6–C10) VFAs, along with thioalcohols, are the causal molecules of axillary malodour, and although the involvement of 16-androstene steroids remains popular with some, the evidence presented by Zeng et al. (1991) and Austin & Ellis (2003) suggests their contribution is minor. Our current view is that the major route to short-chain VFAs is the metabolism of branched aliphatic amino acids by Staphylococcus species, rather than the partial degradation of long-chain methyl-branched fatty acids by corynebacteria (A), as previously reported (James et al., 2004a, b). Meanwhile, the medium-chain VFAs most heavily implicated in axillary malodour, such as 3M2H and 3-hydroxy-3-methylhexanoic acid, originate from Nα-acyl-l-glutamine precursors (Natsch et al., 2003, 2006). However, a challenge remains to identify bacteria with sufficient Nα-acylglutamine aminoacylase activity to explain the high levels of these VFAs present on axillary skin. Along with structurally unusual medium-chain VFAs, thioalcohols such as 2-methyl-3-mercaptobutan-1-ol and 3-methyl-3-mercaptohexan-1-ol have emerged as the primary causal molecules of underarm odour (Hasegawa et al., 2004; Natsch et al., 2004; Troccaz et al., 2004). There is evidence that these originate from S-hydroxyalkyl-l-cysteine precursors, by the action of mainly corynebacterial, but also some staphylococcal C–S β-lyases. Two malY genes encoding this enzyme have independently been cloned and expressed from axillary Corynebacterium species, along with a functionally analogous but phylogenetically distinct metC gene from S. haemolyticus (Natsch et al., 2004; Troccaz et al., 2008). However, an alternative scheme has been proposed, based on the lysis of S-hydroxyalkyl-l-cysteinylglycine precursors, either by a single uncharacterised staphylococcal β-lyase enzyme or by a combination of a corynebacterial dipeptidase and cystathionine β-lyase (Starkenmann et al., 2005; Emter & Natsch, 2008), and clearly a challenge for the future is to fully elucidate the mechanisms underlying thioalcohol formation in the axilla. The main representatives from each of the principal structural groups of axillary malodorants are displayed in Table 4, alongside their corresponding precursor molecules, biochemical routes of formation and causal bacteria.

Table 4. Key representatives of the major structural classes of axillary malodorants and their precursors, with metabolic routes and causal bacteria indicatedThumbnail image of

Although a few knowledge gaps remain, the last decade or so has seen a step-change in our understanding of the microbiological and biochemical origins of axillary odour, as schematically illustrated in Fig. 2. Apart from the outstanding areas detailed previously, a further challenge for the future will be to understand the influence of human genetics on malodour intensity and quality, particularly in the light of recent revelations on the significance of a SNP in ABCC11 (538G→A), the gene encoding an apical efflux pump in the axillary apocrine gland (Martin et al., 2010). Finally, the definitive test will be to exploit all this knowledge to develop a new generation of deodorant systems based on targeting specific bacteria, metabolic pathways or key enzymes, a significant shift from the current reliance on fragrances and broad-spectrum antimicrobial agents.

Figure 2.

Schematic of the microbiological and biochemical origins of axillary malodour, adapted from Austin & Ellis (2003), James et al. (2004a, b), Natsch et al. (2003, 2004, 2006), Starkenmann et al. (2005) and Emter & Natsch (2008). The major routes are proposed to be the metabolism of branched aliphatic amino acids to short-chain (C2–C5) VFAs by staphylococci, the hydrolysis of Nα-acylglutamines to medium-chain (C6–C10) VFAs by corynebacteria, and the formation of thioalcohols from l-cysteine and/or l-cysteinylglycine conjugates by corynebacteria and/or staphylococci.


The authors would like to thank Della Hyliands and Elizabeth-Ann Simons (Unilever Discover Colworth, UK), Sally Grimshaw (Unilever Discover Port Sunlight, UK) and Prof. Keith Holland (formerly of University of Leeds, UK) for their helpful assistance in undertaking various aspects of the reported studies.