Detective mice assess relatedness in baboons using olfactory cues

Since Hamilton's work on inclusive fitness theory (Hamilton, 1964), significant advances have been made in understanding the role of genetic relatedness and kin recognition in both nepotism (the preferential treatment of kin) and inbreeding avoidance in animals (Bateson, 1983; Hepper, 1991; Sherman et al., 1997). In many group-living mammals, including some primates, individuals adjust their behaviour according to the degree of relatedness they share with those individuals with whom they interact (Dugatkin, 1997; Smith et al., 2003; Chapais, 2006). The mechanisms by which animals discriminate conspecifics may be multiple, including indirect mechanisms through contextual cues (e.g. spatial or temporal proximity) and direct mechanisms through phenotype recognition. Among direct mechanisms, two main hypotheses for kin recognition have been formulated: (i) familiarity (or association) among individuals that have been closely associated during early development (Waldman, 1988), and (ii) ‘phenotype matching’, i.e. the assessment of phenotypic similarities (including allele recognition, a particular case of phenotype matching) (Hepper, 1991; Mateo and Johnston, 2003). This latter capacity is expected to evolve in situations where familiarity does not necessarily correlate with genetic relatedness (Widdig, 2007). This mechanism may, for instance, have been favoured in some primates living in multimale multifemale groups where females are highly promiscuous and consequently paternity uncertainty is high. In particular, male baboons apparently recognise their genetically related offspring among familiar juveniles, suggesting the use of phenotype matching (Buchan, 2003).

Kin-recognition systems based on phenotypic similarities involve three components: the production of kin labels (also called expression), the perception of these labels by another individual, and the action (behavioural responses) of this individual according to the perceived similarities (e.g. nepotism or inbreeding avoidance) (Holmes and Sherman, 1982; Beecher, 1988; Reeve, 1989; Gamboa et al., 1991; Sherman et al., 1997; Hauber et al., 2001). The nature of kin labels supporting these mechanisms is probably variable between and within species but potentially includes visual appearance (Dasser, 1988; Vokey et al., 2004; Alvergne et al., 2009), acoustic cues through vocalisations (Rendall et al., 1996), personality traits (Gosling, 2001; Weiss et al., 2006) and body odours (Hübener and Laska, 1998; Widdig, 2007; Alvergne et al., 2009).

Olfaction is a sensory modality that is important in social communication across a wide array of species from insects to humans. Social odours, which are here defined as chemical signals involved in regulating social interactions (Snowdon, 2006), have a correspondingly broad range of functions. Odours can provide directional cues for orientation, serve as signals of alarm, mark territory boundaries, unify groups, direct foraging behaviour, attract mates, indicate reproductive and social status, and provide information about species, subspecies, group, kin and individual identity (Wyatt, 2003). Kin recognition and discrimination through olfactory cues occurs in several species across various taxonomic groups, ranging from invertebrates (Greenberg, 1979; Gamboa et al., 1996) to vertebrates, including fish (Reynolds and Sheldon, 2003; Neff, 2003) and birds (Bonadonna and Nevitt, 2004; Mardon and Bonadonna, 2009). More specifically, in mammals, the importance of olfactory signals in the context of kin recognition is gaining appreciation, as olfactory signals have been shown to advertise individual identity in rodents (Schwagmeyer, 1988; Todrank et al., 1998; Mateo, 2002; Busquet and Baudouin, 2005; Hurst et al., 2001) and carnivores (Zhang, 2005) and Old World monkeys (Setchel et al., 2010), as well as relatedness in prosimians [ring-tailed lemurs (Charpentier, 2008; Boulet et al., 2009)] and New World monkeys [common marmosets (Smith, 1997)].

Nevertheless, progress in understanding the importance of olfactory signals is limited by methodological difficulties, because it is difficult to study a sensory modality that we, as observers, cannot perceive directly. In contrast with the optical and acoustical properties of an object, there is as yet no easy way to ‘record’ and quantify the olfactory properties of objects and chemical signals, especially under field conditions (Epple, 1986). Although significant advances in semiochemistry (Zhang et al., 2005; Scordato et al., 2007) have taken us a step forward in revealing the information potentially available within a chemosignal (Belcher et al., 1986; Belcher et al., 1990; Smith et al., 2001), the analysis of the composition of chemical signals through gas chromatography–mass spectrometry still remains relatively expensive and often technically challenging, especially for non-volatile components such as proteins. However, these compounds have been shown to play a crucial role in communication in many species (Belcher et al., 1990; Hurst et al., 1998; Nevison et al., 2003). Another possible approach to investigating olfactory cues is to perform behavioural assays, which can reveal the information perceived by the signal receiver through behavioural, physiological or neural responses under controlled conditions (Drea et al., 2002; Mateo, 2003; Mateo, 2006a; Mateo, 2006b; Nevitt and Bonadonna, 2005). However, although behavioural assays are informative about both the production and the perception of a chemical signal, the experimental settings they require are difficult to implement, particularly in natural populations, because they require extensive manipulations as well as a long phase of conditioned learning for the animals involved.

In addition to these methodological constraints, the production and perception of olfactory signals have been poorly investigated in some taxa, particularly anthropoid primates (Heymann, 2006). This is largely due to the view that olfactory cues may not represent promising candidates for kin discrimination in a taxon believed to have shown a decreasing reliance on smell during its evolution due to the increasing importance of trichromatic vision (Dominy and Lucas, 2001; Gilad et al., 2004). However, a growing body of evidence indicates that the sense of olfaction in humans is more important than previously thought (Weisfeld et al., 2003; Shepherd, 2004), and it seems possible that a similar pattern may be found in our close relatives. Moreover, it has been proposed that the evolution of kin discrimination might have preceded the reduction in olfactory sensitivity of anthropoid primates (Widdig, 2007), suggesting that the use of olfactory cues by anthropoid primates in the context of kin discrimination is plausible.

The goals of this study are twofold. Our first aim is to develop and validate a cost-effective tool to assess similarity between animal odours, through an indirect behavioural method. Our approach uses mice as ‘noses’ in an improved habituation–discrimination paradigm conducted in a hole-board apparatus. We call this tool a ‘biological olfactometer’ because it allows qualitative measurement (‘meter’) of odours (‘olfacto’) using the olfactory abilities of mice as a ‘biological’ analyser. Mice were chosen to act as ‘noses’ because a number of experiments have shown that rodents have a highly efficient sense of smell (Breer, 2001; Novotny, 2003; Hurst and Beynon, 2004). Additionally, mice can both discriminate between odours of congenic mice differing only at one major histocompatibility complex (MHC) locus, and smell slight environmental differences among individuals (Penn and Potts, 1998). Furthermore, a recent study using a basic classical habituation–discrimination paradigm demonstrated that rats were able to discriminate human odours varying in their degree of relatedness (Ables et al., 2007). Overall, these data indicate a prodigious olfactory acuity in rodents in general and mice in particular, coupled with a good capacity to work in the context of the habituation–discrimination paradigm. The behavioural apparatus and procedures described in our study constitute significant methodological innovations to the classical habituation–discrimination paradigm. The second aim of the study is to use our validated ‘biological olfactometer’ to investigate the similarity in odour production in related individuals from a natural population of chacma baboons (Papio ursinus), a species where chemical communication has rarely been investigated but might be important (Clarke et al., 2009). If individual odours can serve as objective kin labels, we predict that individual odour patterns, as perceived by mice, will show higher similarity between kin than between non-kin.

Baboon odours were collected from a total of 77 wild chacma baboons (Papio ursinus Kerr 1792), including adults and juveniles, males and females, living in Tsaobis Leopard Park, on the edge of the Namib Desert in Namibia, Southern Africa [for details of the site and population, see Cowlishaw (Cowlishaw, 1999)]. The baboons, belonging to two troops (containing 32 and 57 individuals, respectively), were captured in October 2006 in order to gather biological samples (including tissue samples for microsatellite genotyping) using individual cages baited with corn cobs, and which were set-up at dusk. The baboons were captured at dawn, anaesthetised using tiletamine–zolazepam, and they were all processed within a day in order to be released together the following morning when fully awake. For each baboon, a swab made of viscose microfibres (obtained from the French forensic police and used for criminal investigation) was applied against the axillaries (armpits) for two minutes (one minute on each side), and then against the inguinal region (groin) for two minutes (one minute on each side), using new vinyl gloves for each individual. Once impregnated with baboon odours, the swabs were individually stored in opaque glass bottles, and refrigerated at 4°C. For this study, only those odours from adult females were used, to avoid sex and age effects. Tests were then performed on the odours collected from a subsample of 14 adult females (out of the 77 odour samples available).

Sixteen tetra- and di-nucleotide human microsatellite markers were polymorphic in P. ursinus with reproducible results, and thus retained for relatedness analysis. Briefly, DNA was extracted using a DNeasy Tissue Kit (Qiagen, Crawley, UK), following the manufacturer's instructions. PCR amplification was performed using a Qiagen Multiplex PCR kit, following the manufacturer's instructions. PCR conditions were as follows: initial denaturation (15 min, 95°C) and then either (i) 10 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 2.5 min and elongation at 72°C for 45 s, followed by 26 cycles of 94°C for 30 s, 58°C for 3 min, 72°C for 1 min, or (ii) 36 cycles of denaturation at 94°C for 30s, annealing at 47°C for 2.5 min and elongation at 72°C for 45 s and then a final elongation at 60°C for 7 min. Multiple PCR products with different fluorescent labels were run together on either an ABI373 or ABI377 sequencer (Applied Biosystems, Foster City, CA, USA). The software Genotyper (Applied Biosystems) was used for automatic analysis of allele size, combined with visual analysis. Apparent homozygotes were genotyped at least three times from independent amplifications to minimise the risk of genotyping error. The number of alleles per microsatellite locus ranged from 3 to 11 (mean ± s.d.=5.25±1.82), and observed heterozygosity ranged from 0.55 to 0.77 (mean ± s.d.=0.68±0.08) for the microsatellite loci. Further details on microsatellite genotyping in this population can be found elsewhere (Huchard et al., 2010).

Pairwise coefficients of relatedness based on microsatellite typing similarity were calculated between pairs of individuals, using a triadic likelihood estimator of relatedness (TL) based on a likelihood method that uses the genotypes of a triad of individuals to estimate pairwise relatedness (Wang, 2007). Using TL relatedness coefficients, the mean (±s.d.) value of pairwise relatedness (r) between 34 mother–offspring pairs known from behavioural observations was found to be 0.48±0.08 (the TL coefficients range in value from zero to one). Pairwise relatedness coefficients ranged from 0 to 0.80 (median=0.02; mean ± s.d.=0.07±0.12, N=21.945 dyads across six baboon groups for 210 individuals).

Pairs of females for which the pairwise coefficient of relatedness was higher than the average pairwise relatedness of known mother–offspring pairs minus one standard deviation were matched as relatives in this sample (mean pairwise relatedness ± s.d.=0.48±0.08), whereas pairs for which the pairwise coefficient of relatedness was lower than the average pairwise relatedness in the population were matched as non-relatives (mean pairwise relatedness of this sample ± s.d.=0.03±0.03).

Odour tests were performed in a 4 hole-board apparatus (45.0 cm length × 45.0 cm width × 30.0 cm height) enclosed by grey Plexiglas. We used 0, 1 or 2 holes among the 4, according to the phase of the odour test. The hole-board apparatus was placed on the floor of the room (3.0 m length × 3.0 m width × 2.4 m height). Holes [3.0 cm diameter (Ø) × 2.5 cm depth] were located in each corner of the board and were 6.0 cm away from the sidewalls, each hole delimiting an open space in which the odour sample was inserted (see Table 1 for apparatus configuration). When not used, the holes were sealed. The start box consisted of a grey PVC tube (20.0 cm Ø × 20.0 cm height) placed in the centre of the board where the mice were enclosed for 15 s at the beginning of each phase of the test. The apparatus was placed in a room exposed to 2 lx lights. Before each mouse was tested, and between each phase of a test, the apparatus was cleaned with 50% ethanol and then with water. Photocells placed in each hole were used to automatically evaluate the number of sniff bouts (head dips) in each hole (Célérier et al., 2004). When only one hole was active (i.e. contained one odour sample), the parameter used to evaluate odour investigation was the number of sniff bouts. When two holes were active, the parameter used to evaluate the investigation of one odour relative to the other was the percentage of sniff bouts, considering that the percentage expected by chance is 50%. This parameter allowed us to correct for differences between individual mice in their general exploratory behaviour, which is probably more dependent on variations of locomotor activity and personality across individuals than on the sensorial perception of odour differences.

An odour stimulus consisted of a 48–52 g swab sample impregnated with baboon odours. Each odour sample was placed in a clean phial, inserted between two squares of gauze (3 cm length × 3 cm width) and held in position by a perforated cap (2 cm Ø), so that mice could come into direct contact and perceive both the volatile and non-volatile components of the odour (Cheetham, 2007). The phial was inserted at the bottom of the active hole in the apparatus and fixed so that it could not be moved by the animals. The odour samples were stored at 4°C in hermetically sealed bottles and placed at room temperature 10 min before testing. Each odour sample was prepared and manipulated using new vinyl gloves.

The subjects used as ‘noses’ in our ‘biological olfactometer’ were 24 naïve adult male Swiss mice (two months old; 28–32 g) obtained from DEPRE (France). Animals were housed in groups of 12 with access to food and water ad libitum. They were kept in a temperature (21°C) and humidity (50%) controlled facility on a 12 h:12 h light:dark cycle. All test procedures were conducted during the light phase of the cycle, between 09:00 h and 16:00 h in a sound-attenuated and air-regulated experimental room, to which the mice were habituated at least 12 h before behavioural testing.

Habituation–discrimination is a procedure used to assess the perception of differences in odour signals (Halpin, 1986). Using this procedure, we assessed (i) whether mice investigate unfamiliar baboon odours more than familiar ones (Experiment 1), and (ii) whether mice are able to discriminate baboon odours based on their relatedness (Experiment 2). Before each experiment, mice were allowed to freely explore the apparatus in the absence of odour stimuli for two minutes, in order to get used to the apparatus. Then the animals were submitted to a habituation phase, followed by a discrimination phase.

For both Experiment 1 and Experiment 2, this phase consisted of the display of one referent baboon odour to a mouse for two trials of 10 minutes each, separated by a two-minute interval (the referent odour was replaced by a fresh odour, i.e. a new piece of swab from the same individual, between trials 1 and 2 to ensure that habituation was not due to odour degradation). A subject was considered habituated when it displayed a reduced interest in the odour between trials 1 and 2. Only habituated mice were tested in the second discrimination phase.

In Experiment 1, the mouse was simultaneously presented with two odours for five minutes: the referent odour (a fresh sample, i.e. odour A) and the test odour (the odour of a baboon unrelated to the referent baboon, to which the animal was not familiar, i.e. odour B). In Experiment 2, the mouse was simultaneously presented with two odours for five minutes: odour A belongs to a baboon closely related to the referent baboon (0.56>r>0.40) whereas odour B belongs to a baboon unrelated to the referent baboon and unrelated to odour A (r<0.03). Note that the referent odour used in the habituation phase was not used in Experiment 2. For both Experiment 1 and Experiment 2, if the subject perceived odour B as more different in quality from the referent odour than odour A, it should investigate odour B more often than expected by chance (>50%).

Each odour triad formed by the association between a referent odour (habituation phase) with an odour A and an odour B (discrimination phase) was tested by three mice, and each mouse was used only once. The habituation and test odours were always placed in different holes, their locations were counterbalanced across tests, and the experimenter was blind to the identity of odours.

Data were analysed using an exact permutation test for paired samples. For habituation data analysis, the number of sniff bouts in trials 1 and 2 were compared: the null hypothesis H0 is that the number of sniff bouts is identical in trial 1 and trial 2; the alternative hypothesis H1 is that the number of sniff bouts in trial 2 is lower than the number in trial 1.

According to standard habituation–discrimination protocols, in the discrimination phase, if the subject perceives a greater difference between odour B (unrelated) and the referent odour than between odour A (related) and the referent odour, it is expected to investigate odour B more often than expected by chance (50%). Thus, for the discrimination data analysis, the percentage of sniff bouts of the unrelated odour (odour B) was compared with the percentage expected by chance (50%): the null hypothesis H0 is that the percentage of sniff bouts of the unrelated odour is at chance level (50%); the alternative hypothesis H1 is that the percentage of sniff bouts of the unrelated odour is above chance level.

The mice actively investigated all of the odours presented and showed no sign of avoidance to any stimuli in the two experiments. Three mice out of 24 were excluded from the data analysis: one because it did not fulfil our habituation criterion, and two because they did not investigate the two holes at least once during the test phase.

The subjects showed a significant decline of interest for the referent odour between trial 1 (16.4±3.3 sniff bouts) and trial 2 (6.8±1.8 sniff bouts) during the habituation phase (P=0.0002; Fig. 1A). Subsequently, the mice investigated the unfamiliar odours more than expected by chance during the discrimination phase (60.2%; P=0.045; Fig. 1B). The mice are thus able to detect odour similarities in baboons.

The subjects showed a significant decline of interest for the referent odour between trial 1 (25.3±5.5 sniff bouts) and trial 2 (13.5±3.2 sniff bouts) during the habituation phase (P=0.002; Fig. 2A). Subsequently, the mice investigated the unrelated odours more than expected by chance during the discrimination phase (64.4%; P=0.03; Fig. 2B). Thus, according to the behaviour of the mice, the similarity of body odours between two baboons is higher for related pairs than for unrelated pairs.

In this study, we designed and validated a cost-effective tool to assess the degree of similarity between animal odours. Using mice olfactory abilities in a new protocol of habituation–discrimination, and a highly sensitive apparatus automatically recording exploration behaviours, our results indicate that: (i) baboon odours are not considered repulsive for mice because avoidance behaviours were not observed, (ii) mice can detect odour differences between individuals of the same sex and age class in another mammal species (Experiment 1), and (iii) mice perceive a higher odour similarity between related baboons than between unrelated baboons (Experiment 2).

The method of habituation–discrimination developed in this study is an improvement of the classical procedure (Halpin, 1986) used in a previous study assessing objective odour qualities (Ables et al., 2007). Our method, using a hole board, offers several advantages. First, data collection is entirely automatic, which generates strict and unbiased measurements of behaviour. Second, it allows the tests to be conducted in a soft light and silent atmosphere without human presence; thus, minimising stress for the mice (Célérier et al., 2004). Third, this method is particularly well suited for mice because it is based on the exploration of a novel environment, on locomotor activity and on head dipping in a hole-board, all of which are behavioural aptitudes spontaneously expressed by mice (Kliethermes and Crabbe, 2006). Finally, it does not require any reinforcement or training. The ability of mice to detect individual odour differences in baboons confirms their prodigious acuity in olfaction coupled with a good capacity to work in the context of the habituation–discrimination paradigm. Further investigations are now needed to determine if this method is similarly efficient when applied to other sex and age classes of individuals, for instance using odours from males and juveniles. Taken together, these results suggest that the behavioural procedure we developed, using mice as ‘noses’, may be used in a quantitative way, to assess the degree of similarity between animal odours. Indeed, our method, which was tested here on baboon odours, is potentially suitable for other species.

Our findings also show that related adult females from a wild population of chacma baboons (P. ursinus) produce more similar odour cues than unrelated individuals of the same sex and age class, at least from a mouse's point of view. The next step is to identify the composition and the chemical properties of odour labels produced by baboons and used by mice, thanks to complementary experiments with mice, possibly coupled with chemical analyses. Previous studies, essentially in rodents, indicate that the scents associated with the MHC can be used to recognise kin (Yamazaki et al., 1983; Roser and Singh, 1991; Potts et al., 1991; Manning et al., 1992; Brennan and Peele, 2003), although similar effects are not documented in baboons. The MHC is a highly polymorphic gene complex that has been shown to influence the volatile scent profiles of mice (Yamazaki et al., 1979; Carroll et al., 2002), rats (Singh et al., 1987) and humans (Wedekind and Furi, 1997; Jacob et al., 2002). It is thus possible that individual odour differences perceived by mice result from MHC variation among individuals. Although attention has classically and largely focused on volatile compounds (such as odour variation influenced by MHC loci or by genes linked to these loci), it has lately become increasingly apparent that non-volatile proteins and peptides are also important in scent signals (Hurst et al., 2001). On the one hand, non-volatile components are more stable and persistent than volatile components, which have a reduced longevity and can rapidly deteriorate. On the other hand, the detection of non-volatile components occurs through the vomeronasal system, which requires a longer scent contact than the detection of volatile components through the main olfactory system (Halpern and Martinez-Marcos, 2003; Hurst and Beynon, 2004). Both types of components (volatile and non-volatile) may differentially contribute to the communication of information on individual identity. In our experiments, mice were allowed to investigate odours at close contact; thus, both volatile and non-volatile scent components of the olfactory signal were likely to have been detected. One possibility to test whether the olfactory signal of baboons detected by our mice contains some non-volatile components would be to compare our results with a similar experiment where the mice are not allowed to make physical contact with the odour source.

Across taxonomic groups, the ability to assess the degree of relatedness among conspecifics appears essential in the evolution of behaviour such as nepotism and mate choice [in insects (Greenberg, 1979; Gamboa et al., 1996); in fishes (Neff and Sherman, 2003; Neff and Sherman, 2008); in mammals (O'Riain and Jarvis, 1998; Heth et al., 1998; Mateo, 2003)]. Odour is one of the possible cues used to detect kin. Such an ability relies on phenotypic comparisons and occurs in three steps: (i) similarity in odour production by related individuals, (ii) perception of such similarity, and (iii) behavioural reaction to this perception. Our data indicate that the first step of this process is effective in chacma baboons because variability in baboon odour production is influenced by their relatedness. Thus, a baboon could potentially use such cues to identify his own kin by comparing a known phenotype (either his own odour phenotype or one learned under unambiguous circumstances) with the odour phenotype of another individual, or to assess kinship between two individuals that are not directly related to him (but might be related to each other) by matching their respective body odours. In any case, the effective use of olfactory cues by baboons to identify their kin remains to be investigated. It remains possible that the olfactory abilities of baboons may limit their capacity to detect cues that are perceived by mice (Dominy and Lucas, 2001; Gilad et al., 2004) or that mice may not focus on the same olfactory components that baboons do.

Our study opens perspectives for both biochemistry and behavioural ecology. First, we hope that this research, by contributing a simple and cost-effective tool to the study of chemical communication in animals, will encourage its exploration in a variety of taxa and contexts. Second, our finding that odours may play a role in both the signalling of individual characteristics and relatedness among female baboons suggest that this cue is possibly used for kin recognition in this species and possibly other anthropoid primates, and we hope it will encourage future research to investigate both intra-specific perception of odour similarity and its associated behavioural response in non-human primates.

We would like to thank the 2006 Tsaobis field team for their assistance in the baboon trapping. We are also grateful to S. Funk, A. Randall, A. Dillestone, H. Marshall, C. Staples, J. Osborn and M. Weill for their help with lab work, J. Wang and A. King for their help with parentage analysis and relatedness calculations. We thank the Ministry of Lands and Resettlement for permission to work at Tsaobis Leopard Park, the Gobabeb Training and Research Centre for affiliation and the Ministry of Environment and Tourism for research permission in Namibia. We would like to thank D. Béracochéa for lending a hole-board apparatus, G. Ganem for her help with the habituation-discrimination paradigm, S. Behamou for his assistance in statistical analysis, A. Charmantier for reading the manuscript and A. Malapert for bibliography. Two anonymous reviewers also provided helpful comments on the manuscript. We confirm that we have adhered to the Guidelines for the Use of Animals in Behavioural Research and Teaching (Animal Behaviour 2003, 65:249-255) and the legal requirements of the country (Namibia) in which the work was carried out. This work was funded by a Scientific Council (Montpellier II University) Project Grant, a Natural Environment Research Council (UK) Project Grant and Advanced Fellowship awarded to G.C. This paper is a publication of the ZSL Institute of Zoology's Tsaobis Baboon Project. Contribution ISEM 2010-019.