Paternally inherited HLA alleles are associated with
women’s choice of male odor.

Nature Genetics v.30 Feb02

Suma Jacob1, Martha K. McClintock1*, Bethanne Zelano1 & Carole Ober2*
*These authors contributed equally to this work. Published online: 22 January 2002, DOI: 10.1038/ng830

The major histocompatibility complex (MHC) is a source of unique individual odors that influence individual recognition, mating preferences, nesting behavior and selective block of pregnancy in animals1–10. Such phenomena have been difficult to study in humans, because the human leukocyte antigen (HLA, human MHC) loci are the most polymorphic loci in the human genome11, with the potential to generate millions of unique combinations of genotypes. In addition, high variability in background odors, encoded by the rest of the genome and influenced by cultural practices, contribute to a low signal-to-noise ratio that could mask HLA-based olfactory cues. Here we show that women can detect differences of one HLA allele among male odor donors with different MHC genotypes. Notably, the mechanism for a woman’s ability to discriminate and choose odors is based on HLA alleles inherited from her father but not her mother. The parents’ HLA alleles that she does not inherit show no relationship with odor choice, despite exposure to these HLA-encoded odors throughout her life. Our data indicate that paternally inherited HLA-associated odors influence odor preference and may serve as social cues.

1Institute for Mind and Biology and Department of Psychology, 5730 South Woodlawn Avenue, The University of Chicago, Chicago, Illinois 60637, USA.

2Department of Human Genetics, University of Chicago, Chicago, Illinois, USA. Correspondence should be addressed to M.K.M. (e-mail: mkm1@midway.uchicago.edu).

Introduction

It is often assumed incorrectly that humans have a poorly developed sense of olfactory perception12,13. In fact, there are some molecules to which humans are extremely sensitive (the threshold for detecting 2-bromophenol is 10–4.6 ppm)14. People can also detect the odors encoded by genetic information, discriminating between nearly identical strains of mice that differ only at one or a few MHC loci15. In addition, individuals have described body odors to be pleasant when they are from people who have few HLA alleles that match their own16,17.

We sought to determine the resolution of the human ability to discriminate among HLA-associated odors by investigating women’s odor choice based on the number of matches to their own HLA. Because odor choice could be based solely on exposure to MHC-associated odors from one’s family during development18 –20 or could require information from one’s inherited HLA, we carried out a human odor choice study in an isolated community for which the HLA types of two generations are known21. For the first time, we were able to determine whether a woman’s choice of HLA-associated odors were based on alleles inherited from her father, mother or both, or on exposure to nontransmitted parental HLA alleles. In addition, in contrast to the virtually unlimited number of HLA haplotypes present in outbred populations, there are only 67 HLA haplotypes in this community22. Forty-nine unmarried women who had never been pregnant participated in the olfactory choice sessions.

We selected as odor donors men of diverse ethnicity and a different ethnicity compared with that of the isolated community, but who nonetheless carried HLA alleles found in the community as well as completely foreign alleles (Fig. 1). At each of the five HLA loci studied, there was a possibility of up to two allele matches, for a maximum of ten matches. Our sample contained a median of two allele matches (range 0–7) between the smellers and the odor donors. To collect body odors, each donor wore the same T-shirt for two consecutive nights. The women were not told the source of the odors. They rated each T-shirt odor for four attributes: familiarity, intensity, pleasantness and spiciness. Unexpectedly, the women in our study rated the human odors in absolute terms as slightly pleasant and more pleasant than common household odors (Table 1).

Results

Odor choices and HLA matches

To determine whether women prefer odors on the basis of an individual’s HLA type, we asked the women which odor they would choose and which they would not choose, if they had to smell it all the time. There were no differences among donors in their likelihood of being chosen either as the most or the least preferred by the women as a group (χ2=2.3, df=5, P=0.81). There were significant differences only relative to each womanthat is, the combination of a particular woman and odor donor. The donor of a woman’s most preferred odor had significantly more HLA allele matches with her own alleles than did the donor of her least preferred odor (Fig. 1; Wilcoxon paired sign test, P=0.0016). Specifically, a woman’s most preferred odors were from donors who had 2.3 ± 0.2 matches with her own HLA alleles, significantly more matches than with donors of her least preferred odors (1.5 ± 0.2; paired t-test: t=3.65, P=0.0007). Thus, women can discriminate between odors from donors with different HLA types whose number of HLA matches differed on average by only one allele. Notably, this discrimination took place even though women were tested without regard to menstrual cycle phase, which modulates olfactory perception23.

Fig. 1 Number of allele matches between each smeller/donor pair. (figure 1 tables not included)
The genotypes at the five HLA loci of each of the six donors are shown in the top six rows. The HLA alleles on the paternally inherited and maternally inherited haplotypes are shown for each of the 49 smellers in the first six columns. The final row gives the mean, standard error of the mean and range of allele matches with all 49 women for each odor donor. The superscript M indicates the most preferred donor odor, and the superscript L indicates the least preferred, for each woman. The following donor alleles were not present in the study population: B58 (donor 3), DR12 (donor 4), A25 (donor 5), and A29 (donor 6).

Table 1 • Average ratings for attributes of each type of odor (paired t-test comparisons within subjects of men’s odor versus bleach, clove and control)

                                                     Odor
Attribute     Scale     Men           Bleach         Clove          Control
Familiarity   0 to 10   5.5 ± 0.29    9.6 ± 0.14a    7.3 ± 0.44b    6.2 ± 0.35c
Intensity     0 to 10   5.1 ± 0.22    9.1 ± 0.21a    9.1 ± 0.20a    4.4 ± 0.28d
Pleasantness  –5 to 5   0.2 ± 0.27   –0.08 ± 0.50   –0.90 ± 0.48c   1.1 ± 0.25d
Spiciness     –5 to 5  –1.2 ± 0.28    1.6 ± 0.53a    4.5 ± 0.20a   –2.6 ± 0.29a

Difference from men’s odor: aP≤0.0001; bP≤0.001; cP≤0.05; dP≤0.01.

Odor characteristics and identification

Before making their odor choices, women had rated their most preferred odors as more pleasant (mean 1.2 ± 0.41 versus –0.7 ± 0.42; paired t-test: t(48df )=3.77, P=0.0004) and less spicy (mean –1.3 ± 0.47 versus 0.1 ± 0.50; paired t-test: t(48df )=–2.62, P=0.01) than their least preferred odors, but did not rate them as more familiar (mean 5.4 ± 44 versus 5.2 ± 0.45; paired t-test: t(48df)=0.27, P=0.79) or intense (mean 5.8 ± 0.40 versus 6.1 ± 0.35; paired t-test: t(48df)=–0.87, P=0.39). Although the most preferred odors were from donors with more HLA matches than the least preferred, they were not perceived as being more familiar. Moreover, the women were unaware of the source of the odors, and only 19% of the women described the donor odors as possibly being related to anything associated with humans (such as a closet). Most of the women, therefore, made these judgments without being conscious of them as natural human odors.

Paternally versus maternally inherited HLA

We next asked whether these HLA-associated odor choices were related to the parental origin of the matched alleles. We examined separately the number of matches between each donor and the HLA alleles that the woman inherited from her mother versus those that she inherited from her father. The results indicate that a woman’s choices were based on matches to the alleles inherited from her father (1.39 ± 0.15 matches with most preferred donor versus 0.55 ± 0.10 matches with least preferred donor, paired t-test, t(48df)=4.57, P<0.0001), but not on matches to the alleles inherited from her mother (0.90 ± 0.11 matches with most preferred donor versus 0.94 ± 0.13 matches with least preferred donor, paired t-test, t(48df)=–0.31, P=0.76). To determine whether there was a linear dose-dependent relationship between increasing number of allele matches and strength of preference, we examined the number of paternally inherited matches for each woman and the preference rank she assigned to each odor donor. We found that the more matches to the paternally inherited alleles there were, the higher the donor’s rank, in contrast to the maternally inherited alleles, for which there was no such relationship (Fig. 2). In addition, the regression between the number of matching inherited alleles and preference rank paternal and maternal alleles (heavy was significantly different for the heavy red lines in Fig. 2).

Developmental mechanisms

Finally, we sought to determine whether the observed HLA odor choices required the inheritance of paternal alleles or could be explained by exposure to both the inherited and noninherited paternal HLA alleles during development. Because the HLA type of each parent was known, we were able to ask whether a woman had to inherit her father’s HLA alleles that matched those of the donor to manifest a preference, or whether postnatal exposure to her father’s HLA was sufficient. In the latter case, there would also be a correlation between the HLA alleles of preferred donors and those present in her father that she did not inherit. A woman’s preference was associated only with matches to inherited paternal alleles and not to the paternal alleles that she did not inherit (three-factor repeated measures ANOVA, Fig. 3). Moreover, she did not base odor choices on matches with her mother’s HLA alleles that she did not inherit (Fig. 3). Thus, women’s odor choice is associated with inherited HLA alleles and not just on postnatal exposure to the parents’ HLA, although environmental exposures could affect patterning choices.

Discussion

This is the first study in humans, or in any species, to demonstrate that HLA-associated odor choices are determined by paternally inherited HLA alleles and that exposure to the HLA-associated odors of one’s family is not sufficient for determining preference. Our results indicate that women have an exquisitely sensitive olfactory system that allows them to make choices based on small differences in HLA alleles: preferred odor donors had an average of 1.4 allele matches, and the least preferred donors had an average of 0.6 allele matches to a woman’s paternally inherited haplotype (Fig. 3).

Fig. 2 Number of matching HLA alleles and odor choice. Individual regressions between the number of donor alleles matching a woman’s alleles and her preference rank of the six donors (a rank of 6 was the most preferred). a, This analysis includes only the 22 women who had at least three different paternal matching scores (individual ranges between 0–2 and 0–5 matching alleles) among the six donors. Average slope coefficient=0.62 ± 0.15, t=4.0, P=0.001; average intercept=2.7 ± 0.2, t=–4.8, P=0.001. b, Individual regressions for the 18 women whose matches between maternal and donor alleles met the same range criteria as in a. Average slope coefficient=–0.27 ± 0.19, t=–1.34, P=0.19; average intercept=3.8 ± 0.2, t=1.3, P=0.20. Heavy red lines in a and b represent the ten women who met the above criteria for both their paternally inherited and maternally inherited alleles and thus contributed to both panels in the figure. Their paternal and maternal regressions were significantly different (slope coefficients: 0.70 ± 0.08 versus –0.38 ± 0.23, paired t=4.2, P=0.002; intercepts: 2.6 ± 0.01 versus 3.8 ± 0.2, paired t=4.9, P=0.001).

Fig. 3 Effects of inheritance and parental origin of HLA alleles on odor choice. Number of allele matches between the paternally and maternally inherited alleles of each woman (n=49) and the alleles of her most preferred odor donor, compared with the number of matches with her least preferred donor. The results of a three-factor repeated-measures analysis of variance were: interaction of odor preference (most- versus least-preferred donor) × parental origin (paternal versus maternal alleles) × inheritance (inherited versus not inherited alleles) F(1,48)=11.16, P=0.002; interaction of odor preference × inheritance F(1,48)=11.63, P=0.001; main effect of odor preference F(1,48)=1.69, P=0.20; main effect of parental origin F(1,48)=0.01, P=0.93; main effect of inheritance F(1,48)=0.98, P=0.33.

No other studies have revealed a preference for a small number of MHC allele matches over fewer matches (that is, one or none). This is not inconsistent, however, with the results of previous studies in which the choices offered were between individuals with a small number of allele matches and those with identical or nearly identical MHC. Studies of inbred mouse strains4,7,9 and humans21,24 have revealed avoidance of mates with a very high number of matching alleles (with one or two haplotypes identical to one’s own). When given a choice of self or different MHC, there is a consistent preference for differences. Our study investigated choices within the lower range of 0–7 allele matches; the results show that women avoid odors from donors with 0 or 1 HLA allele matches to their own HLA alleles and prefer odors from donors with more HLA matches. Our study is thus consistent with previous literature, showing that a small, intermediate number of MHC matches is preferred over either zero matches or identical MHC.

This finding is different from the results of two studies by Wedekind and colleagues16,17, who reported that female students tended to describe body odors as pleasant when from someone with few HLA allele matches to their own HLA (r=–0.15, one-tailed P=0.07)17. Although these studies examined HLA matching in the same range as in our study, they did not use an odor preference choice test (‘wanting’)25, which has a different neural substrate than judgments of pleasantness (‘liking’)25. Moreover, they did not evaluate the individual effects of the maternally inherited and paternally inherited alleles on odor perception. Additional differences in study design, such as selecting donors and recipients from the same population, make it difficult to directly compare Wedekind and colleagues’ study16,17with the results of this study or the existing literature.

Why would an intermediate number of MHC matches be optimal? It has been shown that female mice prefer males that are just slightly unfamiliar over those that are either very familiar or very unfamiliar26. This preference might result in an optimal balance between inbreeding and outbreeding costs27,28. More recently, it was hypothesized29 that a preference for mates with intermediate levels of MHC matching may be an optimal evolutionary strategy to preserve immunocompetence of offspring. Our data are consistent with these proposals and suggest that when people are presented with choices among individuals with levels of MHC disparity that would typically be encountered in outbred, natural populations, they have a preference for more matches.

The evolutionary significance of MHC-based odor preferences has been narrowly interpreted as a potential mechanism for mate choice; however, MHC-based odor choices may influence broader aspects of social behavior, such as communal nesting10 and other forms of kin-biased behavior. For example, adult female baboons (Papio cynocephalus) groom and aid their paternal half-sisters more than non-kin, biasing their social behavior as much or even more than they do toward maternal half-sisters30. This ability to recognize paternal kin is notable because in this promiscuous species, paternity is unknown and adult males do not participate in childrearing. Olfactory recognition of paternally inherited MHC could help a female to recognize paternal kin in this and other species where paternity is not certain.

The main olfactory system has the potential to decode MHC information31. The multiple receptors and neural pathways of the olfactory system distinguish small differences in a large repertoire of molecular structures32 and thus could mediate the discrimination between gene products of one HLA allele, as seen in this study. In humans and mice, and perhaps other animal species as well, the detection of MHC-mediated body odor may result from the close linkage between the MHC loci and olfactory receptor genes33,34. MHC-specific odors may be soluble MHC proteins or their vagile components, odor molecules bound selectively to MHC proteins, or by-products of MHC-specific bacteria colonization in skin or axillae35–37. Moreover, an inherited parent-of-origin effect on odor preferences has been reported in mice38, albeit for general body odor and not specifically for MHC-derived odors, showing that olfactory systems can distinguish between paternally and maternally inherited genetic information.

This is the first double-blind study of MHC-derived odor choices in humans, and the first to reveal a paternally inherited component associated with these choices. We have shown that a woman’s odor choice is associated with inherited HLA alleles rather than exposure to HLA-associated odors from her family during development. Consistent with earlier studies, these data indicate that there is not one most preferred human male odor for everyone, but that odor preference is relative, based in this case on the degree of HLA differences between a man and a woman.

Methods

Sample composition. Members of the community are of German-Austrian descent. Unmarried women (n=49; average age 25 ± 1.6 y; range=13 y to 56 y) participated in olfactory sessions. They were all raised by their biological parents and all but one were members of large families (median sibship size=7; range 4–11). All were at least one year past menarche and all but one (age 56) had menstruated during the month preceding the study. None of the women in the study were using hormonal contraception or had ever been pregnant. The women were blind to the hypotheses being tested and the identity of the contents and odors of the boxes. The six male odor donors, derived from Ashkenazi Jewish, Dutch, English, German, Polish, Scottish, Sikh and Spanish ancestry, ranged in age from 23 y to 47 y (average 31.3 ± 3.8 y).

Odor collection.

Male donors each wore three T-shirts; each shirt was worn for two consecutive nights. The T-shirts and bedding were washed with fragrance-free detergent. During the study period donors avoided a list of 21 foods, including garlic and asparagus, and filled out a diet log. We encouraged them to follow a bland diet similar to that of the isolated community. Men showered with Ivory soap before putting on T-shirts and refrained from using scented products such as deodorant, lotion, cologne and aftershave. They also were instructed to avoid scents such as cigarette smoke and pets as well as sexual activity or contact with other individuals during odor collection. Men stored shirts of consecutive wear in open plastic bags during the day. After the second night, we cut each Tshirt in half with sterilized scissors, placed them in separate freezer containers and stored them at –80 ºC to prevent degradation of the chemosignals. We kept shirt halves frozen until the morning of the olfactory sessions. These procedures were approved by the Institutional Review Board of the University of Chicago.

Olfactory sessions.

Before each session, we put portions of the donor Tshirts in unmarked cardboard boxes lined with foil, with a triangular hole through which subjects smelled the contents. Subjects placed their noses in the hole but could not see the contents of the box. Control T-shirts were from the same stock, washed in the same fragrance-free detergent and presented in the same box type as worn T-shirts. The controls thus differed only in their lack of additional human or household odors. We lined all boxes with foil wrap to decrease box cardboard odor, to create a replaceable lining around the triangular hole for hygiene and to prevent odor contamination by the smellers. Each woman was asked to rate ten odor boxes (six T-shirts from different male donors, one shirt with chlorine bleach, one shirt with clove oil and two unworn shirt controls). Two male testers from outside the community and blind to our specific hypotheses followed a protocol with scripted questions. They presented boxes and recorded women’s responses to the odors. Women held the boxes so that the nose and chin were in the triangular opening of the box and, taking their time, smelled the contents with active nasal inspirations. They verbally stated a rating for each box, on the basis of a rating scale. The scale for both familiarity and intensity ranged from zero to ten, with zero representing unfamiliar or undetectable odors and ten very familiar or very intense odors. The scale for pleasantness and spiciness ranged from –5 to 5, with –5 representing very unpleasant or bland odors, and 5 representing very pleasant or spicy odors. Ratings were standardized by z-score relative to ratings of the carrier odor (freshly laundered T-shirts in a cardboard box).

For the odor choice test, the women smelled repeatedly three identical boxes containing T-shirts; we then asked which odor they would choose to smell all the time and which odor they would avoid. Choices were determined in two counterbalanced rounds that included all six donors. Then, without knowing that they had previously chosen the odors, they were asked to choose between the two boxes chosen in the previous rounds and between the two boxes avoided in the previous rounds. Each woman thus chose from all the donors her most preferred odor and her least preferred odor, without knowing that they were male body odors.

HLA typing and scoring.

We typed odor donors and subjects for HLA-A, -B, -C and -DR antigens by serology, and –DQB1 alleles using molecular techniques (Fig. 1)22. We determined matching scores for each female smeller/male donor pair by counting the number of alleles present in the male donor that matched an allele in the female. Homozygous alleles in the smeller that matched a donor allele were counted as two matches. At the five loci, there was a potential for a maximum of ten matches.

Acknowledgments

We thank D. Hayreh and P. Klimczyk for assistance on field trips, K. Beaman for serological HLA typing the donors in this study, C. Wedekind for sharing unpublished protocols and J. Brown for helpful comments. This work was supported by a MERIT Award from the National Institute of Mental Health, a grant from the John T. and Catherine D. MacArthur Foundation (to M.K.M.), a grant from the National Institute of Child Health and Development (to C.O.) and  n MD/PhD Training Grant from the National Institutes of Health (to S.J.).

Received 7 June; accepted 21 December 2001.

References

1. Yamazaki, K. et al. Recognition among mice: evidence from the use of a Y-maze differentially scented by congenic mice of different major histocompatibility types. J. Exp. Med. 150, 755–760 (1979).

2. Yamaguchi, M. et al. Distinctive urinary odors governed by the major histocompatibility complex. Proc. Natl Acad. Sci. USA 78, 5817–5820 (1981).

3. Beauchamp, G.K. et al. Chemosensory recognition of mouse major histocompatibility types by another species. Proc. Natl Acad. Sci. USA 82, 4186–4188 (1985).

4. Yamazaki, K. et al. Control of mating preferences in mice by genes in the major histocompatibility complex. J. Exp. Med. 144, 1324–1335 (1976).

5. Yamazaki, K. et al. Recognition of H-2 types in relation to the blocking of pregnancy in mice. Science 221, 186–188 (1983).

6. Brown, R.E., Singh, P.B. & Roser, B. The major histocompatibility complex and the chemosensory recognition of individuality in rats. Physiol. Behav. 40, 65–73 (1986).

7. Egid, K. & Brown, J.L. The major histocompatibility complex and female mating preferences in mice. Anim. Behav. 38, 4186–4188 (1989).

8. Eklund, A., Egid, K. & Brown, J.L. The major histocompatibility complex and mating preferences of male mice. Anim. Behav. 42, 693–694 (1991).

9. Potts, W.K., Manning, C.J. & Wakeland, E.K. Mating patterns in seminatural populations of mice influenced by MHC genotype. Nature 352, 619–621 (1991).

10. Manning, C.J., Wakeland, E.K. & Potts, W.K. Communal nesting patterns in mice implicate MHC genes in kin recognition. Nature 360, 581–583 (1992).

11. Parham, P. & Ohta, T. Population biology of antigen presentation by MHC class I molecules. Science 272, 67–74 (1996).

12. Herrick, C.J. Neurological Foundations of Behavior (Holt, New York, 1924).

13. Schaal, B. & Porter, R.H. "Microsmatic humans" revisited: the generation and perception of chemical signals. Adv. Study Behav. 20, 135–199 (1991).

14. Devos, M., Patte, F., Rouault, J., Laffort, P. & Van Gemert, L.J. Standardized Human Olfactory Thresholds (IRL Press at Oxford University Press, Oxford, 1990).

15. Gilbert, A.N., Yamazaki, K., Beauchamp, G.K. & Thomas, L. Olfactory discrimination of mouse strains (Mus musculus) and major histocompatibility types by humans (Homo sapiens). J. Comp. Psychol. 100, 262–265 (1986).

16. Wedekind, C., Seebeck, T., Bettens, F. & Paepke, A.J. MHC-dependent mate preferences in humans. Proc. R Soc. Lond. B 260, 245–249 (1995).

17. Wedekind, C. & Füri, S. Body odour preferences in men and women: do they aim for specific MHC combinations or simply heterozygosity? Proc. R Soc. Lond. B 264, 1471–1479 (1997).

18. Quadagno, D.M. & Banks, E.M. The effect of reciprocal cross-fostering on the behaviour of two species of rodents, Mus musculus and Baiomys taylori ater. Anim. Behav. 18, 379–390 (1970).

19. McCarthy, R. & Southwick, C.H. Cross-species fostering: effects on the olfactory preferences of onychoms torridus and Peromyscus leucopus. Behav. Biol. 19, 255–260 (1977).

20. Yamazaki, K., Beauchamp, G.K., Kupniewski, D., Bard, J. & Thomas, L. Familial imprinting determines H-2 selective mating preferences. Science 240, 1331–1332 (1988).

21. Ober, C. et al. HLA and mate choice in humans. Am. J. Hum. Genet. 61, 497–504  (1997).

22. Weitkamp, L.R. & Ober, C. Ancestral and recombinant 16-locus HLA haplotypes in the Hutterites. Immunogenetics 49, 491–497 (1999).

23. Pause, B.M., Sojka, B., Krauel, K., Fehm-Wolfsdorf, G. & Ferstl, R. Olfactory information processing during the course of the menstrual cycle. Biol. Psychol. 44, 31–54 (1996).

24. Ober, C., Weitkamp, L.R. & Cox, N. in Chemical Signals in Vertebrates 8 (eds Johnston, R., Müller-Schwarz, D. & Sorensen, P.) 189–199 (Plenum, New York, 1999).

25. Berridge, K.C. & Robinson, T.E. What is the role of dopamine in reward: hedonic impact, reward learning, or incentive salience? Brain Res. Rev. 28, 309–369 (1998).

26. Gilder, P.M. & Slater, P.J. Interest of mice in conspecific male odours is influenced by degree of kinship. Nature 274, 364–365 (1978).

27. Bateson, P. in Mate Choice (ed. Bateson, P.) 257–277 (Cambridge University Press, Cambridge, 1983).

28. Ochoa, G. & Jaffe, K. On sex, mate selection and the Red Queen. J. Theor. Biol. 199, 1–9 (1999).

29. Penn, J.P. & Potts, W.K. The evolution of mating preferences and major histocompatibility genes. Am. Nat. 153, 145–164 (1999).

30. Smith, K. in Department of Psychology 167 (The University of Chicago, Chicago, 2000).

31. Schaefer, M.L., Young, D.A. & Restrepo, D. Olfactory fingerprints for major histocompatibility complex-determined body odors. J. Neurosci. 21, 2481–2487 (2001).

32. Mori, K., Nagao, H. & Yoshihara, Y. The olfactory bulb: coding and processing of odor molecule information. Science 286, 711–715 (1999).

33. Fan, W., Weiwen, C., Parimoo, S., Lennon, G.G. & Weissman, S.M. Identification of seven new MHC class I region genes around the HLA-F locus. Immunogenetics 44, 97–103 (1996).

34. Amadou, C. et al. The mouse major histocompatibility complex: some assembly required. Immunol. Rev. 167, 211–221 (1999).

35. Yamazaki, K., Singer, A., Curran, M. & Beauchamp, G.K. in Chemical Signals in Vertebrates Vol. 8 (eds Johnston, R.E., Muller-Schwarze, D. & Sorenson, P.W.) 173–180 (Plenum, New York, 1999).

36. Vincent, C. & Revillard, J.P. Characterization of molecules bearing HLA determinants in serum and urine. Tranplant. Proc. 11, 1301–1302 (1979).

37. Pearse-Pratt, R., Schellinck, H., Brown, R., Singh, P.B. & Roser, B. Soluble MHC antigens and olfactory recognition of genetic individuality: the mechanism. Genetica 104, 223–230 (1999).

38. Isles, A.R., Baum, M.J., Ma, D., Keverne, E.B. & Allen, N.D. Urinary odour preferences in mice. Nature 409, 783–784 (2001).