Animals vary consistently in traits such as boldness and stress reactivity, which influence fitness, movement, and social dynamics. Identifying genetic variants linked to these behaviours can clarify proximate mechanisms and evolutionary trade-offs. Dopaminergic (DRD4) and serotonergic (TPH2) pathways are known to modulate exploration and emotionality, while LDHA contributes to neuroenergetics and endurance, yet their combined role in shaping behavioural diversity remains unclear. Here we show, in 137 homing pigeons, that a SNP at position C382T in DRD4 and T185A in TPH2 are significantly associated with boldness. Birds with DRD4 T/T and TPH2 T/A genotypes emerged more quickly from the shelter. We also found that DRD4 C/C and TPH2 T/A genotypes were associated with slower recovery from acute social isolation stress, quantified by eye infrared thermography. In addition, a microsatellite polymorphism in LDHA intron 6 interacted with the presence of a mirror companion, suggesting a genotype-dependent effect of social buffering. These findings provide evidence that common genetic variants in neuromodulatory and metabolic genes contribute jointly to behavioural and physiological syndromes in birds, and social context can interact with genotype to influence stress resilience. We demonstrate how candidate markers are useful in exploring the ecological and evolutionary consequences of personality variation.

2.1. Subjects

We tested 137 homing pigeons (Columba livia), bred and housed in two adjacent pigeon lofts at The John Krebs Field Station, University of Oxford, Wytham, UK. The ages of the birds ranged between one and 20 years. The sex of the birds was identified by genotyping the chromo-helicase DNA-binding gene (CHD) as detailed in the Genotyping section. There were 75 males, 60 females, and 2 undetermined due to low-quality DNA sampling. Although normally free-ranging, due to the implementation of an Avian Influenza Prevention Zone just prior to our study, the birds had no access to the outside during the study. Food and water were provided ad libitum during daily provisioning, and birds were routinely weighed and inspected for health. Metadata for all birds is available in the Supplemental Data.

2.2 Genotyping

Feather samples were collected from all 137 individuals. DNA was extracted from feather roots using the QIAGEN DNeasy Tissue Kit (QIAGEN, Valencia, CA, USA).

We chose the widely polymorphic regions known to affect biochemical synthesis, which are the most informative for association testing in our study. For genotyping of DRD4 exon 3 and TPH2 (part of exon 10 + intron 10 + exon 11 + part of 3 UTR), primer pairs were designed using Primer 3 plus Version 2.3.6 software [56], based on the registered pigeon DRD4 (NC_088606.1and) and TPH2 (NC_088602.1) as shown in Table 1. PCR was performed in a 15 μl reaction mix containing 20 ng of genomic DNA, 2x PCR buffer, dNTPs at 400 μM, each primer at 0.3 μM, and 0.5 U of LA-Taq DNA polymerase (TaKaRa, Shiga, Japan). The PCR cycle conditions are shown in Table 1. The amplified products were purified using a PCR purification kit (Roche, Mannheim, Germany), and the resultant products were sequenced using the same primers with Big Dye Terminator ver. 3.1 Cycle sequencing kit (Applied Biosystems, Foster City, CA, USA) according to the standard protocol, and electrophoresed on an ABI PRISM 3130xl sequencer (Applied Biosystems, Foster City, CA, USA). BLAST software [57] was used for sequence identification and confirmation. FinchTV 1.4.0 (Geospiza, Inc., Seattle, WA, USA; http://www.geospiza.com), MEGA 11 [58], and Bioedit 7.0.5.3 [59] software were used for sequence alignment and polymorphism detection.

For genotyping of LDHA (part of exon 5 + intron 5 + part of exon 6) (Supplementary Figure 1) and CHD (Supplementary Figure 2), their microsatellites were amplified using fluorescently labelled (6-FAM) forward primers (Table 1) [34]. PCR reaction mixture was 10 μL, including 20 ng of genomic DNA, 2x GC buffer I, dNTPs at 400 μM each, forward/reverse primers at 0.3 μM, and 0.5U of LA-Taq DNA polymerase (TaKaRa, Shiga, Japan). Thermocycling conditions are shown in Table 1. PCR products were electrophoresed on an ABI 3130xl DNA Sequencer (Applied Biosystems, Foster City, CA, USA). Fragment sizes were estimated based on the fluorescently labelled forward primer in GENE MAPPER (Applied Biosystems, Foster City, CA, USA). Estimations of genotypic and allelic frequencies, heterozygosity (H_E), number of effective alleles (N_E), and Hardy-Weinberg equilibrium were carried out using GENALEX version 6.0 [60].

2.3. Behavioural tests

2.3.1. Boldness

The widely used “emergence test” of boldness was used to determine each bird’s behaviour on the bold-shy axis: birds were individually placed in a shelter, and we measured the latency before they voluntarily exited this shelter and entered a novel, open space. The shelter was made of a cardboard box (34 × 27 × 27 cm), and the open space was a 27-cm-wide and 140-cm-long passage with wood-board sides [13]. A small pile of multigrain-mixed food was situated at the far end of the passage. The birds’ daily feeding routine was shifted during the experiment, such that they were only fed in the lofts after the experiments were finished every day. At the start of each test, a bird was placed in the shelter, covered in front by an opaque shutter. The bird was then allowed to habituate for 3 minutes, followed by the removal of the shutter. The bird was then left undisturbed, and the experimenter remained silent and out of the bird’s view. The time taken by the subject to emerge from the shelter was recorded, and once emerged, the bird was immediately removed from the apparatus and returned to the loft. If a bird did not emerge within 10 minutes, the trial was concluded, and the bird was returned to the loft. No bird was allowed access to the food at the end of the passage, to avoid reinforcing the emergence. Each test was filmed from above by a tripod-mounted video camera (HC-V520, Panasonic, Japan). Each bird was tested twice with an interval of 36 days between the two tests to quantify repeatability in the time taken for a given subject to emerge from the box. All 137 pigeons participated in our boldness assays.

2.3.2. Stress response to social isolation

Birds were caught from the loft and brought indoors in a transit box, followed by a minimum of 20-minute habituation time outside the test room, where they maintained social contact with their conspecifics inside the transit box. After the habituation time had elapsed, birds were brought individually into the test room with constant ambient lighting. The bird was immediately introduced into a novel cage, which comprised two equal-sized compartments separated by a wire-mesh divider. The focal bird was placed in the near compartment with no perch, food, or water provided and left undisturbed for 10 minutes (i.e., they experienced social isolation in a novel environment). The interior space was deliberately constrained to the extent that the bird could stand and turn only a limited amount. In this no-mirror condition, the far compartment was empty. For the duration of the 10-minute test period, the bird was videotaped by a thermal imaging camera (FLIR T620bx, FLIR Systems, United States) from a set distance of 1 m. The experimenter was hidden behind a plain white cloth and controlled the photography equipment remotely using the FLIR Tools software to minimise any effect of observer presence.

Video footage was subsequently analysed with a focus on changes in maximum eye temperature, a reliable indicator of stress in birds when using infrared thermography [30], as temperature varied in different parts of the eye. Maximum eye temperatures of the birds were obtained from the thermal videos using the software FLIR Tools, at one-minute intervals. Since birds were not fully restrained and hence their eyes were not always within the camera's view, a time window of 10 s either side of the one-minute mark was allowed for camera adjustment to capture the eye image.

Following completion of the above, the test was repeated for each subject, with the only modification being that a mirror (297 mm × 210 mm) was placed at a distance of 20 cm in front of each pigeon, in the far compartment. We used a flat mirror as a standardised visual social cue to approximate social presence without physical contact, which [61]has been shown to attenuate isolation-related stress in other bird species such as European starlings [62] and domesticated chickens [63]. As such, we treat the mirror condition as a pragmatic social-cue manipulation. 134 of our total of 137 birds participated in the two social isolation tests.

2.3.3. Route efficiency

We quantified the route efficiency of a subset of our subjects using data from a previous study [13]. GPS-mounted birds were released individually from a novel site (Church Hanborough: 51°48’44.3” N, 1°22’38.3” W; distance to home: 5.2 km, direction to home: 128°) and their navigational performance was quantified from GPS data in terms of the ratio of straight-line distance to actual flight distance (termed route efficiency). Higher route efficiency implied better navigational performance by a given bird. 20 of our total of 137 birds participated in these navigation trials.

2.4. Data analyses

All statistical analyses were performed in R Studio, using established and publicly available statistical packages, as detailed below.

2.4.1. Boldness

The distribution of times taken by subjects to emerge from the shelter (latency) was non-Gaussian and also right-censored, given the 600-second limit on each trial. To test for repeatability in the measure of boldness, we thus fitted two different models using the rpt() function (rptR package). The first model was binary and predicted whether or not the individual emerged from the box, and the second tested the latencies of individuals that emerged both times using a Gaussian distribution. For both models, trial number, age, weight, and sex of the subject were added as control variables. To test for the relationship between the different genes and the boldness response, we fitted a Cox proportional hazards regression model on the latency and emergence status with the fixed effect of genotypes and the random term of individual ID. To rule out the confounding effects of age, weight, sex, or trial number, we added each term to the regression model and used a likelihood ratio test to compare the models. 

2.4.2. Stress response to social isolation

Maximum eye temperature was measured in all birds for 10 minutes following transfer to the experimental cage. The mean change in maximum eye temperature over the test period was computed for each genotype to determine the effect of genetic polymorphism on stress-induced eye temperature change. For some subjects, there were a few time windows when no eye image could be captured (1.21% of the dataset), usually because the bird turned its head away for a short period, and these incomplete time series were completed by the R package imputeTS by interpolation.

We first tested if different genotypes had different baseline temperatures at the beginning of the social isolation test. To do so, we ran a linear mixed model predicting the temperature at timestep zero by the genotype of the individual (testing the different genes in three different models). We added the age, weight, sex, and a random effect of the individual ID as control variables.

Second, we tested whether the evolution of the eye temperature over time in response to the social isolation was different between genotypes. We analysed the evolution of the temperature by subtracting the baseline temperature at timestep zero from each timestep’s temperature. We then checked the presence of a difference in the evolution of the temperature by testing the effect of the interaction between timestep and genotype in a set of linear mixed models (lmer() function from the lme4 package). In all models, the evolution of the temperature over the timesteps was approximated using a sigmoid curve (SSlogis() function from the stats package). The first model included a triple interaction timestepgenotypecondition (presence or absence of a mirror) to also account for a potential differential effect of the genotype on the evolution of the temperature depending on the presence of the mirror. As this interaction was not found to be significant, we used a second model testing the interactions (timestepgenotype, timestepmirror, and mirror*genotype) separately. We tested all three genes in different models, and added the age, weight, sex, as well as a random effect of the individual ID as control variables in all models.

We verified all model assumptions by checking the distribution of the residuals in diagnostic plots (histogram of the residuals, qq-plot, and plot of the residuals against fitted values).

2.4.3. Route efficiency

The means of individual route efficiencies were computed for each genotype and fitted to a Generalized Linear Model for one-way ANOVA to determine the effect of genetic polymorphism on individual flight characteristics during solo homing.