Milinski et al.  report human individuals perceiving their body odour as self-like if it is enriched with around 1 nanomol per armpit of a synthetic peptide that is a known ligand of one of the subject's own human leukocyte antigen (HLA) alleles. At the same time, they report specific brain activation patterns in test subjects challenged with air which has contacted water containing 25 µM of the same peptides.
This paper challenges several paradigms in human olfaction, and these observations, if they can be further substantiated, may have deep implications on our understanding of how we perceive other humans. Before we fully accept this paradigm shift, we should interrogate the implications these observations have and ask the critical open questions.
The implications are twofold: on the one hand, they challenge our physico-chemical understanding on what constitutes an odorant, and on the other hand they would redefine what an odour is at the perceptual level.
On the physico-chemical side, nonapeptides are very far away from the chemical nature of all known odorants perceived by humans. These are relatively hydrophobic, volatile molecules with usually no more than three functional groups and a molecular weight which in general is below 300 Da. Figure 1 shows the heaviest known odorant perceived by humans according to Ohloff et al.  in comparison with the A24 peptide reported as an olfactory cue by Milinski et al. . Owing to the high molecular weight (in this case 1184.4 Da) and extensive possibility for hydrogen bonding, peptides do not evaporate, and I am not aware of a study reporting volatile nonapeptides. The only way we can imagine them to reach the human olfactory epithelium is in the form of aerosols (or dust), which raises the (testable) question of whether a human sniffing his/her armpit generates an airstream which is sufficient to desorb the highly diluted peptides from the skin surface in the form of aerosols.
From the sensory perspective, the findings are even more intriguing. In an olfactometer study, panelists reported no difference in intensity when sniffing water or the peptide solutions. So far, we have therefore no indication that panelists are able to consciously detect the presence of the peptide in the solution. This could easily be tested with a standard sensory test (e.g. a triangle test), and a dose–response and a detection threshold for solutions containing the peptide could be determined as it is done in classical chemosensory science. If panelists are not able to detect presence of the peptide blindly, we would have to conclude that processing is subconscious. In summary, (i) the proposed ligands are far outside the physico-chemical realm of known odorants, (ii) there is specific processing in the brain, but (iii) potentially no conscious perception. These observations might indicate that this ‘peptide-sense’ could be a new chemosensory modality.
Surprisingly, there is no discussion on how the presented results affect our current understanding of human olfaction. The physico-chemically very unusual nature of peptides as odorants is not mentioned by Milinski et al.—air having been in contact with the peptide solution is just called ‘odour’. It is only mentioned in the results section that ‘peptides were delivered in aerosolized form’. An olfactometer is a device built to deliver odours in the gas phase, and the device used had been validated with the extremely volatile chemicals limonene and terpinenol . If this same device is assumed to deliver non-volatile peptides in the form of aerosols, this should be tested and analytically quantified, which then would allow one to relate the effects to the dose applied and to perform dose–response studies.
The behavioural study does raise a number of questions. The results are split into three groups, and smokers and people with a self-reported cold are excluded in a post hoc analysis. Only with these criteria applied do the results reach significance (p = 0.0167). In a typical sensory study, these would be a priori exclusion criteria. Furthermore, in this study, different assessments were pooled: panelists having neither of the two HLA alleles selected for the study received solvent instead of peptide as ‘self’ stimulus and panelists having both selected HLA alleles received solvent as ‘non-self’ stimulus. With this approach, both the ‘self’ and the ‘non-self’ stimulus could be either solvent or one or two peptides, and these three groups were pooled for analysis. This is based on the a priori assumption that the specifically selected combinations in each case provide a more ‘self’ versus ‘non-self’ stimulus. One may argue that these different treatments actually address three slightly different questions—(i) Can panelists discriminate ‘self’-peptides from placebo? (ii) Can panelists discriminate ‘non-self’-peptides from placebo? and (iii) Can panelists discriminate ‘self’ from ‘non-self’ peptides? In future studies, the same panelists could be tested in a cross-over design with both water versus ‘self-peptide’ and water versus ‘non-self-peptide’; preference for the peptide condition should then be inversed based on HLA type, and there would be no bias owing to the fact that water has been used itself as ‘self’ or ‘non-self’ condition in different panelists.
Research on major histocompatibility complex (MHC)/HLA-associated individual odours has led to surprises before. Originally, mice were shown to discriminate urine odour from inbred strains based on difference in MHC region only . This signal could be extracted from urine by organic extraction and survived sampling in the gas phase and gas chromatography [4,5], confirming its hydrophobic and volatile nature. HLA-associated body odours were also reported in humans , but a more recent study did not replicate the same findings . Currently, it is unclear whether the differences reported in studies with humans are owing to the outbred nature of the human population with a high diversity at the HLA locus. In mice, MHC-binding peptides where postulated as alternative chemosensory cues for MHC recognition, somewhat conflicting with the earlier studies on the volatile nature of the signal , but it was also proposed that these two systems could act in parallel . These studies, along with studies on major urinary proteins , indicate that non-volatile signals can indeed be involved in individual recognition in mice upon direct physical contact with urine.
So far the studies on MHC/HLA-binding peptides as chemosensory cues are based on synthetic peptides which were tested for neuronal activation and/or behavioural effects. This research field has therefore started with a priori assumptions on the chemical cues tested and has taken an approach radically different from classical work in chemical ecology, where semiochemicals were always identified based on isolation from the donor. If donor-derived extracts/mixtures triggered a response in the recipients, activity-guided fractionation and chemical analysis then led to a proposed active principle, which was resynthesized to replicate the behavioural effect. For an unequivocal proof that HLA-binding peptides secreted by human donors give specific responses in human recipients, this classical approach, namely starting with peptide fractions from sweat of HLA-matched and unmatched human donors, will be needed. With the current approach of working only with synthetic peptides, which were never detected in the bodily fluids (i.e. sweat in this human study), questions will remain. Sturm et al.  recently tried this approach in mice. With LC–MS analysis, they tested for MHC-genotype-related differences in the abundance of MHC-binding peptides, but no such differences in abundance of native peptides could be found. They then used immunological methods to detect an ovalbumin-derived peptide in urine of ovalbumin-transgenic mice. The immunological signal was only detected in mice expressing β2-microglobulin, indicating MHC-dependent peptide secretion in urine. However, we cannot yet infer from these studies that the native peptide profile in mouse urine is affected by the MHC genotype, nor can we draw conclusions on the situation in human sweat.
Science should always be ready for paradigm shifts, but before we accept a new human sensory modality detecting peptides by which we perceive our potential partners without actually noting an odour, we may call for more evidence and a more critical discussion on the open questions, which were not addressed in this study.
The author has been studying human body odours over the last decade. This comment is his personal view and does not reflect the opinion of his employer.
Comment to: Milinski M, Croy I, Hummel T, Boehm T. 2013 Major histocompatibility complex peptide ligands as olfactory cues in human body odour assessment. Proc. R. Soc. B 280, 20122889.
The accompanying reply can be viewed at http://dx.doi.org/10.1098/rspb.2013.2816.
- Received June 28, 2013.
- Accepted August 30, 2013.
- © 2013 The Author(s) Published by the Royal Society. All rights reserved.