It has been known for centuries or even longer that dogs are able to identify individual people and track them over long distances.

Recent evidence also indicates that mothers can recognize the odor of their baby and, perhaps, babies can identify the scent of their mother.

How is this done? There must be an olfactory signature much like a fingerprint. We have called this an odorprint or an odortype. Our research over the past years has been designed to discover the nature of this odortype-what genes are involved, what roles it plays in normal animal and human behavior, and ultimately what chemicals are involved in the olfactory code.

The story of our research program began in 1974, when my mentor, collaborator and a former chairman of Monell, the late Dr Lewis Thomas, building on theories on histocompatibility genes and olfaction, suggested that histocompatibility complex genes, might impart to each individual a characteristic scent.

Dogs, he surmised, might be able to distinguish different human histocompatibility types by the sense of smell. During the past three decades, we have investigated this hypothesis; our studies and more recently, a number of other laboratories have succeeded in verifying the fundamental basis for this prediction.We have shown that mice that differ in as little as a single DNA base pair mutation in one of these genes exhibit different odors. Indeed, theoretically every mouse and every person (except identical twins) have a unique odor determined by these genes. Our studies have encompassed the behavior, physiology and chemistry of this phenomenon.


Our research program has made use of animal model, "the mouse", because we can easily produce mice with known genetic differences.

To study MHC-determined odortypes, several group of mice were produced. All individuals in each group were genetically identical.


Genes in the Major Histocompatibility Complex (MHC) have been identified by immunologists as critical for immune response. Animals and humans that are similar at this set of about 50 genes can mutually accept transplants for example.

The MHC of human is called HLA and of mice is called H-2. A unique feature of this set of genes is their extreme variability. Since the MHC is the only group of genes exhibiting this extreme variation, it was hypothesized- correctly as it turned out- that they would be involved in controlling odortypes.


An odortype (we defined as) is the genetically-determined body odor that distinguishes one individual of a species from another. The MHC-determined odortype is that portion of the animal's total odortype that is under the control of genes in the Major Histocompatibility Complex.

Mating Preference

A large number of experiments have confirmed H-2 associated mating preference are generally characterized by preference for non-self H-2 type.

Such preference could act to promote valuable heterozygosity at H-2.

This was evidently the first example of reproduction behavior and selection that has been traced to variation of a particular gene or gene complex. We next demonstrated this MHC-associated communication system involved chemical sensation by using a Y-maze.


Air is conducted through two odor chambers, containing either animals or urine samples exposed in Petri dishes, to the two arms of the maze.  During the training session, water-deprived sensor mice were rewarded with a drop of water for each correct response. Successful discrimination is indicated by 80% concordance over the course of the individual session. Generalization trials involve blind testing of coded samples, which is possible because no reward is provided in association with the blind odor sources. These coded samples are from novel sources, thus ensuring that the trained animals are responding to the MHC difference and not learning about specific individuals during training.

We demonstrated that mice are able to distinguish individuals on the basis of genetic differences at the major histocompatibility complex of genes, and that urine is a prime odor source. And also we have shown that mice have the ability to distinguish the odor of a number of different mutant mice from the odors of non-mutant mice.

Other studies have shown that odortypes of fetuses are expressed in the body odors of the pregnant mouse or woman and that these fetal odors persist in the female mouse for at least 15 days after the pups have been born and removed from the mother. Thus the "immunization memory" of the fetus is somehow expressed in the mother's body odor.

Drs. Schaefer and Restrepo, University of Colorado, had utilized this model system to test whether odortypes elicit unique maps of neuronal activity in the main olfactory bulb (MOB). Such distributed patterns of odor-induced neuronal activity likely contribute to the encoding of odortype information.

Recent advancements in understanding how chemosignals are used to detect MHC haplotype necessitates the assessment of the relative role of volatile and non-volatile chemosignals for detection of MHC haplotype.

Human Odortypes

For the first time we investigate for human odortypes, whether genetically determined HLA-based odortypes can be used to identify individual people.

We use a mouse bioassay system (Y-maze), to detect an HLA-determined odor signature. Mice trained to distinguish urine samples of two individuals from different supertype groups generalize this response to some but not all pairs with "similar" supertype differences. These data provide the strongest indications yet that MHC type in humans, as in mice, is involved in provisioning an individual with a distinct, genetically-based body odor.

Family study

Training on these 2 family members takes much longer than training on 2 unrelated individuals. Previous mouse training studies indicated that difficulty of training is likely directly related to degree of odor difference. Hence, as would be expected, family members likely have similar body odors. This could be due to genetic similarities (most likely) and/or to similarities in environment (e.g. diet).

In addition to these works, my laboratory is also involved in research on model studies of how odors may signal preclinical disease and how body odor changes with age.

Disease Detection

It has long been recognized that various genetic and metabolic disorders are characterized by changes in body odor; this is not surprising since many of those conditions alter body chemistry, resulting in an increase or change of odorous substances. Little systematic study of this area has been conducted because of the vast individual differences in human body odor due to genetic and environmental influences. An animal model system for which genetic and environmental factors are held constant and only the presence or absence of the disease vector is allowed to vary provides a rigorous model system to study body odor and disease. Mouse mammary tumor virus (MMTV) affords a favorable model with which to test for changes in odor profiles that arise from infection with an oncogenic retrovirus or from its premalignant (mammary nodules) and malignant sequelae. We also use a mouse bioassay system (Y-maze). Mice infected with MMTV can be discriminated by scent from uninfected, genetically identical control mice long before the development of tumors. Recently, we investigated the chemical nature of the MMTV-related odor signal. The most striking result was the finding that the amount of 3,4-dehydro-exo-brevicomin (DHB -reported to be a male mouse pheromone) was dramatically increased in urine of infected female mice. This result suggests that DHB plays an important role in the recognition of MMTV-infected mice and it raises interesting questions about virus interactions with mouse chemosensory communication systems. Mating preference experiments to be conducted as part of this work will help answer this question.

Lung cancer is the leading cause of cancer-related death throughout most of the world. A potential strategy for early diagnosis is to identify metabolic signatures of the disease (so called "biomarkers") using non-invasively obtained samples such as urine. To this end, we used two mouse models of human lung cancer to investigate whether changes in the patterns of urinary volatiles would be indicative of tumor growth. Sensor mice could be trained to discriminate between odors of mice with and without experimental tumors. Urinary volatile organic compounds were analyzed with solid-phase-micro extraction (SPME), followed by gas chromatography coupled with mass spectrometry. Thus, lung cancer causes characteristic changes in the concentrations of volatile compounds in the urine, generating a detectable disease-specific volatile signature.

Animal model studies are powerful in determining what is theoretically possible for human work and for getting suggestions on common mechanisms. The successful results of this mouse work encourage us to conduct similar studies using similar techniques on human patients. Of course the ultimate goal is to develop reliable, non-invasive, inexpensive tools to help in the diagnosis of human lung (and other) cancers.