Behavioral, physiological, and genetic drivers of coping in a non-human primate

Summary Animals experience stressful situations, from predation to social conflicts, but mostly deal with them successfully. This adaptive mechanism, coping, reduces the adverse effects of stressors, and its failure may result in reduced fitness. Substantial inter-individual variation in coping is observed, yet little is known about how behavioral, physiological and genetic drivers regulate coping holistically and contribute to such variations. We assessed behavioral coping styles (n=30), emotional arousal (n=12), and personalities (n=32) of long-tailed macaques (Macaca fascicularis) and also investigated the association of coping with a valine/methionine polymorphism encoded by a critical human stress regulatory gene, catechol-O-methyltransferase (COMT) (n=26). Personality and the human equivalent COMT Val/Met polymorphism were associated with “nonaggression-based” and “aggression-based” coping styles. Compared to nonaggression-based, aggression-based copers maintained higher average facial temperatures, indicating potentially lower emotional arousal, as measured using infrared thermography. These findings demonstrate a complex interplay of various proximate mechanisms governing coping in a non-human primate.

traits such as aggression, activity, boldness, and so forth) and stress reactivity (i.e., a quantitative dimension: through Hypothalamic-pituitaryadrenal/HPA axis activity). 157][18] A proactive behavioral coping style indicates an active response to stressors with a higher propensity to take risks and form rigid routines, while a reactive behavioral coping style indicates a passive response through social support-seeking or distancing, and higher behavioral flexibility. 19Typically, proactive copers exhibit low inhibition control, high novelty seeking and frequent aggressive and impulsive behaviors 20,21 ; on the contrary, reactive copers do not directly engage with the stressors or show aggressive and impulsive behaviors.At the physiological level, reactive coping is associated with higher HPA axis reactivity than proactive coping. 19,22However, according to the two-tier model, reactive copers can show a low HPA axis-or ''emotional'' reactivity (cf. 15,23) due to the independence of behavioral coping style and HPA axis reactivity.Notably, emotionality in coping has primarily been investigated by measuring stress hormone concentrations (see 24,25 ).Nevertheless, from a neurophysiological perspective, emotional states are associated with the activities of sympathetic and parasympathetic nervous systems, which may increase and decrease blood pressure, resulting in blood flow alterations underneath the skin. 26,27This change in temperature, especially in the face, can thus be used as a reliable alternative for estimating emotional arousal during coping.However, limited information is available on whether such a non-invasive methodological tool can capture or provide information for emotionality in coping.
A key proximate driver of variation in coping styles may be personality.9][30] While both personality and coping styles can be attributed to individual-level differences, the question remains whether coping styles are mere reflections of personalities in stressful contexts.2][33][34] Coping, unlike personalities, was found to be less stable over time with adjustments 32,35 and had very low heritability. 33Furthermore, little to no evidence suggested an overlapping genetic basis of coping styles and personalities. 33In contrast to the human literature, non-human animal research often considers coping styles and personalities synonymous.For instance, the 'boldness-shyness' personality dimension 36 (typically used in birds, fishes, and so forth) is used interchangeably with proactive-reactive coping styles. 19Consequently, the concepts of coping (in human and non-human animals) and its (in)dependence on personality need more clarity as variable trait characteristics are subject to selection pressure and are of immense importance.
Conventionally, research on coping has taken an integrated approach of behavior and neurophysiology in non-human animals.As a result, candidate genes of the HPA axis are yet to be significantly explored as potential causal mechanisms.8][39] For example, zebrafish (Danio rerio) lines selected for proactive and reactive coping styles show distinct basal neurotranscriptomic states, 40 and the Swiss sublines of Roman highand low-avoidance rats selected for good versus poor performance in exploration tasks exhibit divergent stress responses and coping styles. 41These findings emphasize the need for studies on key candidate genes regulating stress coupled with behavioral and physiological mechanisms.Such a comprehensive approach is essential for gaining a holistic understanding of coping in non-human animals.
A key candidate gene underlying varying coping styles in humans and potentially other animals is the so-called COMT (Catechol-O-methyltransferase) gene.COMT is a key enzyme encoded by the COMT gene in catecholamine catabolism. 42This enzyme is responsible for the inactivation of dopamine, adrenaline and noradrenaline neurotransmitters 43 and is accountable for more than 60% of dopamine clearance in the brain. 44,45In humans, a single nucleotide polymorphism (SNP) located in the coding region of the COMT gene (codon 158, exon 4) has been shown to have functional consequences.At this locus, a transition from guanine (G) to adenosine (A) results in changing the amino acid product from valine (Val) to methionine (Met).The resulting gene products differ significantly in their thermostability and, subsequently, in COMT enzymatic activity.The Val variant is considered more stable and active than the thermolabile Met encoding product. 46This functional COMT Val 158 Met polymorphism (dbSNP: rs4680) has been studied extensively in humans.Individuals with the G/G (Val) variant are associated with phenotypes, such as higher aggression 47 and impulsive behavior, 48 whereas emotional problems 49 and increased reactivity to stress 50 are found in individuals with the A/A (Met) variant.In non-human primates (NHP), COMT polymorphism is linked with aggression, dominance, and stress. 51,52To our knowledge, however, in the NHP genus Macaca, the human equivalent COMT Val/Met polymorphism was only identified in Assamese macaques, where a moderate relation between rank and aggression was found. 52Although crucial from a mechanistic point of view, no studies to date have investigated the association of COMT polymorphism and coping in non-human animals.
The current study uses a multidisciplinary framework to investigate the coping behavior of long-tailed macaques (Macaca fascicularis), a group-living NHP species.We conducted ecologically relevant predator exposure experiments repeatedly (see 53,54 ) to assess coping styles.Encounters with predators can be unpredictable and uncontrollable, following the definition of a stressor, 1 which forces individuals to cope with the situation.An infrared thermal imaging method was employed to detect facial temperature changes during predator exposure, a proxy for emotional arousal. 55A drop in nose temperature is typically associated with high emotional arousal in NHP 27,56 and thus can provide an estimate of emotionality in coping.We furthermore examined personalities in the long-tailed macaques using a comprehensive multimethod ''bottom-up'' approach of repetitive behavioral observations and experiments (cf. 53).Since relatively steep hierarchical structures are found in long-tailed macaque societies, 57 we also determined the dominance-rank relationships as potential predictors of coping and its variations.Finally, we extracted DNA from blood samples of a subset of subjects to investigate the potential existence of COMT polymorphism in the human equivalent COMT Val/Met site (exon 4) of long-tailed macaques.
We hypothesized that personality and underlying COMT genotype would moderate behavioral coping styles and emotional arousal, explaining individual variations.Due to the nature of the ''bottom-up'' approach, personality dimensions were not predetermined and thus, informed predictions could not be made.Yet broadly, we predicted that bold-and aggression-related traits would be positively associated with aggression-based coping styles, and social support-seeking behavior would be positively associated with nonaggression-based coping styles (cf. 12).As high-ranked individuals are often associated with boldness in despotic societies, 58,59 we expected them to have aggressionbased coping styles compared to their subordinate counterparts.Furthermore, we expected behavioral coping styles and emotional arousal to correlate.Thus, compared to a baseline measure, individuals with aggression-based coping styles are expected to maintain higher average nose temperatures (i.e., potentially low emotional arousal) or recover faster from an initial drop in nose temperature than individuals with nonaggression-based coping styles.Finally, we expected the underlying COMT genotypes, if the human-like SNP is present in long-tailed macaques, to be associated with different coping styles in our test subjects.Given the higher emotional resilience attributed to the human G/G genotype (Val carriers), [60][61][62] we would expect long-tailed macaques having the COMT G/G genotype (Val) to be less emotionally aroused and to exhibit more aggression-based coping styles compared to macaques carrying the A/A (Met) variant.
Considering the behavioral variables we expected to relate to coping (see Table S1), two non-correlating latent factors were extracted using an exploratory factor analysis (EFA).These two factors explained 37.6% and 22.4% of the total observed variance, i.e., 60% cumulatively.The first factor included positively loaded variables of close ground, conspecific affiliation, and a negatively loaded variable of far (Figure 1).In contrast, the second factor had only two positively loaded variablespredator aggression and conspecific aggression (Figure 1).Clearly, the second factor had variables relevant to aggression, both toward conspecifics and the stressor; thus, we labeled it as ''aggression-based coping'' (cf. 12).The first factor, apart from a close distance, did not indicate any direct engagement with or aggression toward the stressor.Instead, a potential reliance on conspecifics through affiliation was found.We labeled the factor as ''nonaggression-based coping'' (cf. 12).In summary, the obtained coping styles indicated direct interactions with conspecifics in the form of aggression or affiliation and direct (with aggression) or indirect (without aggression) engagement with the stressor.

Coping scores
Individual scores extracted from the factors (i.e., aggression-and nonaggression-based coping) were plotted against each other (see Figure S2).An initial visual inspection was done based on the position of the individuals in the quadrant, where quadrants one, two, three, and four represented mixed, nonaggression-based, low reactivity, and aggression-based coping styles, respectively (see Figure S2).
The aggression-based and nonaggression-based factors had median values of 0.15 and À0.11, respectively.Out of the 30 individuals tested: (i) six belonged to quadrant 1 (mixed coping style); these individuals scored higher than the median scores of both nonaggression-(mean G standard deviation: (0.30 G 0.25) and aggression-based (0.80 G 0.28) coping factors; (ii) Seven individuals belonged to quadrant 2 (nonaggression-based coping style); they scored higher than the median value of the nonaggression-based factor (1.14 G 0.82) but lower than the aggression-based factor (À0.54 G 0.58); nonaggression-based factor scores of these individuals were significantly higher than the aggression-based factor scores (Wilcoxon Signed-Rank test: z = 2.41, r = 0.91, p = 0.01); (iii) Seven individuals, who belonged to quadrant 3 (low reactivity category), scored lower than the median values of both the coping factors (aggression-based: À0.88 G 0.65, nonaggressionbased: À0.76 G 0.76); aggression-and nonaggression-based factor scores were also comparable (Wilcoxon Signed-Rank test: z = 0.72, r = 0.27, p = 0.46); (iv) ten individuals belonged to quadrant 4 (aggression-based coping style), of which nine scored higher than the median score of the aggression-based factor (0.56 G 0.23); however, all ten of them scored lower than the median value of the nonaggression-based  S9.
factor (À0.45 G 0.32).We decided to categorize the particular individual who had an ambiguous value to low reactivity due to the lack of clear evidence of having an aggression-based coping style.Nevertheless, for validation and to check if our downstream analyses were affected by this, we conducted them with and without this individual included as an aggression-based coper.The nine individuals scored higher on the aggression-than nonaggression-based factor (Wilcoxon Signed-Rank test: z = 2.88, r = 0.96, p = 0.003).Interestingly, the six individuals categorized as mixed copers had significantly higher absolute values for the aggression-based factor in comparison to their scores on the nonaggression-based factor (Wilcoxon Signed-Rank test: z = À2.15,r = 0.88, p = 0.03).We combined aggression-based and mixed copers, which further helped us deal with our relatively low sample size.To summarize, except for low reactivity, where specific coping scores could not be assigned (n = 8), individuals belonging to aggression-(n = 15) and nonaggression-based (n = 7) coping styles had specific coping scores from their corresponding factors, i.e., the factor on which they scored (loaded) significantly higher.At the group level, we found variations in percentages of aggression-and nonaggression-based copers.In Gr.1, $36% of the individuals had aggression-based coping styles, $53% had nonaggression-based coping styles, and $0.9% had low reactivity.Gr. 2 had $53% aggression-based copers and, interestingly, no nonaggression-based copers, but $46% of the animals had low reactivity.We found 75% and 25% of the monkeys in Gr.3 to have aggression-and nonaggression-based coping styles, respectively.

Emotional arousal
According to the best-fitted model (see Table S2) on our thermography data, we found a significant effect of coping style on the nose temperature change (LME: t = À2.490,h 2 = 0.45, p = 0.03).In comparison to an initial 2-10 min time window (baseline), individuals with aggressionbased coping styles regained and maintained higher average mid-nose temperatures (average change in temperature = 2.54 G 3.24 C) than those with nonaggression-based coping styles (average change in temperature = 0.03 G 1.44 C, Figure 2).The temperatures remained consistent for either coping style during the two different time windows (2-10 min vs. 10-20 min -LME: t = 1.815, p = 0.08; 2-10 min vs. 20-30 min -LME: t = 1.889, p = 0.07).Note that the baseline nose temperatures remained comparable between aggression-and nonaggression-based copers (Mann Whitney U Test: z = À1.314,p = 0.17, see Figure S3).We did not find any effect of sex on nose temperature change (LME: t = 1.408, p = 0.19).
A validation test found no significant difference between aggression-and nonaggression-based copers regarding general activity (aggression-based copers: 654.28  suggesting that the observed nose temperature changes were not due to general movement and foraging, rather an effect of the activities of the sympathetic autonomic nervous system, confirming the emotional arousal we were aiming to measure.

Personality
Based on the behavioral variables extracted from focal observations and novelty experiments, a principal component analysis (PCA) provided us with three PCs (with eigenvalues >1), cumulatively explaining 78% of the total observed variance (Table 1).We labeled the PCs ''Activitysociability,'' ''Affiliation,'' and ''Exploration.''We did not find any effect of age and sex on affiliation; however, independent effects of age and sex were found on activity-sociability and exploration.We found a significant negative effect of age on activity-sociability (GLMM: p = 0.006, see Table S3) and exploration (GLMM: p < 0.001, see Table S4).Males were, furthermore, found to score higher than females in both activitysociability (female: À0.42 G 0.66; male: 0.70 G 1.10; GLMM: p = 0.002, see Table S3) and exploration (female: À0.37 G 0.73; male: 0.62 G 1.10; GLMM: p = 0.02, see Table S4) personality traits.Individual scores were obtained for the three traits (see Table S5).

Personality, dominance hierarchy, and behavioral coping styles
We did not find any effect of the three personality traits, age and sex, on aggression-based coping (see Table S6).Upon the inclusion of the individual with unclear coping style (see section -coping scores; data point highlighted by the solid red dot in Figure S2) as aggression-based coper, our results did not change (null vs. full model comparison: X 2 = 8.075, p = 0.152).Therefore, we reported and retained the first model results as the main outcome.

Catechol-O-methyltransferase valine/methionine polymorphism and behavioral coping styles
Sequencing results revealed two alleles equivalent to the human COMT Val 158 Met encoding polymorphism in long-tailed macaques.Out of the 26 individuals genotyped, 16 were homozygous A/A (Met encoding allele), six were heterozygous (A/G), and four were homozygous G/G (Val encoding allele).Accordingly, the targeted polymorphism was detected at 61.53%, 23% and 15.38% frequencies, respectively (allele frequencies: A = 0.73; G = 0.27).
We analyzed a subset of individuals for whom COMT polymorphism information was available, and a specific coping style could be assigned (n = 18).Due to a small sample size, we could only compare homozygous A/A (n = 6) and G/G (n = 3) individuals from the aggression-based coping category and found a difference (Wilcoxon Rank-Sum test: z = À2.20,h 2 = 0.6, p = 0.02, Figure 4).Homozygous G/G individuals (1.07 G 0.19) had significantly higher coping scores than their homozygous A/A counterparts (0.60 G 0.15).This was in contrast to the personality traits, which did not correlate with the genetic polymorphism (see Figure S4).Finally, if we visually inspect the thermography data, it is interesting to see that the two individuals with the G/G genotype of which we also had thermal data were the ones that recovered most (marked by high nose temperatures) from the stressor (see Figure 2).

DISCUSSION
We investigated coping mechanisms in an NHP species using a multidisciplinary research design.We found two non-correlating aggressionand nonaggression-based coping factors, characterized by direct aggressive engagement with stressors and conspecifics, and exhibition of affiliation toward conspecifics, respectively, during predator exposure experiments.Further inspection revealed aggression-based, nonaggression-based, mixed coping styles, and low coping reactivity in these animals.Aggression-based copers exhibited lower emotional arousal than nonaggression-based ones, as examined using a non-invasive infrared thermal imaging method.Personality traits partially predicted coping, particularly the nonaggression-based coping style, where higher affiliation was associated with lower nonaggression-based coping scores.We furthermore provided the first evidence of the human equivalent COMT Val/Met polymorphism in long-tailed macaques.
Although identified in a few individuals only, we could show that the G variant in a homozygous state (G/G, two alleles encoding for Val) was associated with aggression-based coping.This suggests a genetic basis for a complex behavior, such as coping.Our findings underscore the importance of multidisciplinary approaches in understanding coping mechanisms and their evolution.
Our results align with previous studies on coping styles and associated variations among individuals. 15,19,32,33,37,63However, unlike a proactive-reactive continuum, 15 which is predominantly described in the non-human animal literature, we found non-correlating aggression-and nonaggression-based coping styles, suggesting independent coping strategies in long-tailed macaques (see 64 ).The aggression-based coping style was associated with direct engagement with the stressor through aggressive behavior and also the exhibition of aggression toward conspecifics.Out of all the individuals to whom a specific coping style could be assigned (n = 22), we found $68% of them to have aggression-based coping styles.This is in line with the idea that a more despotic species, such as long-tailed macaques, would exhibit lower inhibition control than more socially tolerant species. 65On the other hand, the intensity of aggression toward conspecifics can be attributed to ''frustration'' and ''redirection,'' critical signs of societies with low social tolerance. 57,66The aggression-based copers retained higher average facial temperatures after an initial 10-min of predator exposure than their nonaggression-based counterparts, thus, possibly lower emotional arousal; it might indicate their predisposed neural underpinnings, which made them ''proactive,'' 67 i.e., better equipped to engage and cope with stressful situations.In other words, the aggression-based copers were more efficient in controlling (albeit not voluntarily) emotional arousal than nonaggression-based copers with regard to the activities of the autonomic sympathetic nervous system.On the other hand, the nonaggression-based coping dimension had positively loaded variables of close (distance to stressor) and conspecific affiliation and negatively loaded variable of far (distance to stressor).Even though the variables close and far were not mutually exclusive (see Table S1), they seem to have covaried.Yet interestingly, it suggests that individuals with nonaggression-based coping styles were not entirely avoiding the stressor but, at the same time, not directly engaging with it like their aggression-based counterparts.These results suggest the complex and multifaceted structure of coping and the less distinctive extremes of its strategies. 14,349][70] Thus, a nonaggression-based coping strategy might also resemble the emotional coping style and seem beneficial in long-tailed macaques, where close social bonds are prerequisites for maintaining group cohesion, cooperative interactions and sociality. 71,72 relatively new and non-invasive measure of infrared thermography provided information for emotional arousal in coping in our study.As opposed to the aggression-based, persistently lower facial temperatures in nonaggression-based copers were found, potentially indicating higher emotional arousal.Thus, the behavioral coping styles correlated with HPA axis reactivity.These findings corroborate that nonaggression-based copers require prolonged social support from conspecifics to reduce arousal and cope with stressors.While the apparent nonindependence of behavioral coping styles with HPA axis reactivity may seem contrasting to the prediction of the two-tier model, it should be noted that thermography data from the low reactants could not be obtained.Although they neither significantly engaged in predator aggression nor relied on conspecific support, the missing information on emotional arousal has made it challenging to conclude whether behavioral coping styles and reactivity are (in)dependent components of coping.Nevertheless, they seem to have covaried.For example, while iScience Article personalities did not affect coping styles in wild baboons, neuroticism was predicted by stress reactivity. 73Finally, despite controlling for obvious confounding factors (e.g., indoor/outdoor temperature differences, distance and angle of the subjects from the thermal camera, and so forth), other unprecedented factors (see 74 ) can cause a reduction in the precision of the thermal measures.Therefore, a methodological comparison between infrared thermography and endocrine measures would be recommended to better understand the HPA axis reactivity in coping.
Using the comprehensive multi-method approach of behavioral observations and experiments, we found three personality traits in longtailed macaques -activity-sociability, affiliation, and exploration.Traits similar to activity-sociability and affiliation had previously been identified in long-tailed macaques. 75While activity-sociability and affiliation in our study represented variables from an observational perspective, repetitive ''rare'' behaviors were captured using experiments in the exploration dimension.This validates the concept of objective assessment of personalities in non-human animals. 53,76Age and sex were found to influence personality traits.Age had a negative effect on activity-sociability and exploration.This effect may be attributed to the active and playful underpinnings of some loaded variables within those traits, e.g., social play, handling container, and object manipulation.Play behavior is essential for the locomotive, cognitive, and social development of young individuals. 77Therefore, younger individuals are expected to score higher on these traits than older individuals, in which activity budgets and priorities may have shifted more toward goals such as reproduction and competition. 78Alongside age, sex was also found to significantly predict personality scores on activity-sociability and exploration, in which males scored higher than females.This may be attributed to female philopatry and male dispersal in long-tailed macaques, where males tend to leave their natal group around the age of four years and join either a bachelor or a new social group. 79Although personalities, following a strict definition, should not be sex-dependent, there is growing evidence for sex-specific personality dimensions.Differential selection pressures and varying life history trajectories are considered the underlying mechanisms for the evolution of sex-dependent personality traits. 80e found partial support for our hypothesis of personalities predicting coping styles.The observed personality traits did not affect aggression-based coping styles.This indicates that the aggression-based coping style in itself may be an independent personality trait, equivalent to boldness, and the associated variables of our aggression-based coping may have captured boldness-related behaviors.However, it is also important to emphasize that we did not find any evidence of ''boldness-explorative'' behavioral syndromes, as reported in previous studies. 81,82Nonaggression-based coping, on the other hand, was inversely associated with personality trait affiliation.This again supports the social buffering hypothesis that (seeking) affiliative interactions can potentially alleviate the impact of stress.Macaque societies are characterized by strong positive relationships among individuals, such as friendships, 71 thus presenting opportunities to seek support from group members in distress.This strengthens the effectiveness of close social relationships in reducing stress in group-living species.Interestingly, we noted that females scored higher than males in the nonaggression-based coping dimension.While this observed sex difference may be explained by a general tendency of females to be more anxious than males, 83 and/or possibly female philopatry (see 79 ), our results here should be interpreted with caution due to a low sample size of males.Contrary to our prediction, although linear and steep hierarchies in all three groups were noticed, we did not find any relationship between dominance-rank relationships and coping styles.A recent study on wild baboons (Papio anubis) also found no relationship between social hierarchies and coping styles. 84These results are unsurprising because dominance and coping styles are group-and individual-level properties, respectively, and might not always covary.
In our final analysis, we were interested in whether potentially different COMT genotypes modulate differences in the coping styles of longtailed macaques.In humans, a functional polymorphism located in exon 4 of the COMT gene (Val 158 Met) has garnered significant scientific attention due to its counterbalancing effect on emotional resilience, stress, anxiety and cognition. 60Even though the catecholaminergic system appears highly conserved among vertebrates, the COMT gene has received little attention in NHP studies.To our knowledge, the Val/ Met polymorphism has only been reported in Assamese macaques 38 among Macaca and thus has not been studied extensively regarding emotional arousal or coping styles in NHP species.We found an association between the COMT Val/Met polymorphism and aggressionbased coping style; G/G individuals (Val/Val) scored higher on coping than individuals carrying the A/A (Met/Met) variant, as hypothesized. 62his finding suggests a potential genetic basis for coping styles in long-tailed macaques. 85,86Our results align with the ''warrior-worrier'' model used to explain the existence of two alleles that induce distinct behavioral phenotypes in humans.The model claims that the Val-allele confers an advantage in confronting aversive stimuli (greater stress resilience and lower anxiety levels, i.e., warrior), whereas the Met-allele is beneficial in cognitive tasks (e.g., memory and attention tasks; worrier).Due to a low sample size, driven by just one individual carrying the G/G variant, comparisons could not be made among nonaggression-based copers, which would be an important avenue for future studies.In addition, as highlighted, the two least emotionally aroused individuals (as measured by infrared thermography) in our study were aggression-based copers with the G/G variant, potentially indicating an effect of COMT Val/Met polymorphism on emotional arousal.To what extent the COMT Val/Met polymorphism can influence animal emotion would thus be interesting to explore in future.It is, however, crucial to highlight that the same polymorphism had no relationship with the observed personality traits in this study, potentially indicating non-overlapping genetic underpinnings of coping styles (and potentially emotionality) and personalities.
The candidate genes of the serotonergic and dopaminergic systems are known to mediate coping in humans 87 and thus may explain the underlying causal mechanisms.However, as pointed out by earlier studies, there is an urgent need to consider the candidate genes while understanding the causal mechanisms of coping.Similar results in humans and long-tailed macaques regarding the COMT Val/Met polymorphism and coping further highlight an evolutionary basis of this mechanism and stress regulation in general.Although the effect of Val/Met on human behavior has been intensively studied, we know that findings on the contribution of single candidate genes to a complex behavioral trait must be interpreted with caution. 88,89We call out for the replication of our study, which is the first to reveal an impact of COMT on coping behavior in an NHP species.Genome-wide association studies can help decipher the involvement of other gene variants, thereby providing a more detailed picture of the interaction of underlying genotype and coping behavior.In addition, comparative studies on other species may reveal the evolutionary history of these traits and their genetic bases.
Using a comprehensive research design, we provided evidence of two non-correlating coping factors in a group-living NHP species.We identified the complex interplay of proximate mechanisms pertaining to behavior, physiology (emotional arousal) and genetics.The complexity and dimensions of coping can be wide and should be investigated carefully.While varying concepts and methodologies are currently being applied, multidisciplinary research designs with novel methodologies are crucial to understanding the overarching process of coping.Although species-typical responses can be at play, identical research designs should be applied to a wide range of taxa to get a grip on the evolution of coping.

Limitations of the study
Although we employed a relatively novel infrared thermography approach to measure emotional arousal, the methodology has some limitations.Relatively large time windows were chosen to investigate temperature differences.This step enabled us to increase the number of individuals for analyses, but at the cost of precise temperature detection in much shorter windows.Also, for inclusion in the thermography analyses, we only relied on individuals who approached close to the fences, indicating a potential sampling bias.However, this is true for most behavioral studies conducted in group settings, where dominant, bold and explorative individuals are potentially chosen as sampling subjects.Nevertheless, employing two thermal cameras during the experiments to maximize sampling effort, and thermal data on individuals with different coping styles (albeit posthoc) suggests that our experiments and the results were not greatly affected by this bias.Furthermore, as mentioned above, a methodological comparison should be carried out between infrared thermography and endocrine measures in quantifying the emotionality of coping.The sample size of the aggression-and nonaggression-based copers with the different COMT Val/Met genotypes was very small, but it depended on the allele frequencies.Besides, while we first report the presence of a human equivalent COMT Val/Met polymorphism in long-tailed macaques, complex behavioral mechanisms, such as coping, may have complex gene interactions.Therefore, the replication of our study with a larger sample of animals should be conducted, and other potential candidate genes must be included in the research design to check for any interactive effects of those genes.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: inside a 180 cm x 70 cm x 70 cm wooden box with only the front exposed to the to be tested group (Gr.3).All groups were habituated to these modifications (without the predators) on the days before testing for at least 24 hours.Besides, on the days of testing, the immediate next enclosure compartments belonging to the other groups were emptied.This is standard practice during enclosure cleaning and maintenance at the BPRC; therefore, we assume that individuals from the focal groups will react similarly on testing and non-testing days.The predator exposure experiments lasted half an hour and were recorded using two video cameras (Canon Legria HF G25 and Sony FDR AX100E 4K) mounted on tripods from different angles.
The facial (nose bridge, tip and mid-nose) temperature data of the subjects were collected using infrared thermography (cf. 93) throughout the predator exposure experiments.Facial temperatures were collected using two FLIR E96 (640 x 480 thermal resolution) thermal cameras by two experimenters who either sat on the ground (for Gr.1 and Gr.2) or stood (Gr.3) at least 1 metre away from the enclosures.The emissivity was set at 0.97; typically, a value of 0.97 or 0.98 is used for non-human primates. 94Reflective-and ambient temperatures and humidity measures were set at the thermal cameras before using them.We used two (for reliability) ThermoPro TP50 thermo-hygrometer devices to record ambient temperature and humidity.Any ambient temperature and humidity changes were noted and subsequently adjusted in the thermal cameras during the experiments.All subjects were habituated to the presence of the experimenters for over a month and at least a week to the thermal cameras.Three key criteria were followed to collect reliable facial temperature data to assess emotional arousal: (i) the distance between the experimenter and an individual subject is within 2 meters, (ii) an individual is facing the thermal camera at an angle not more than 50 , (iii) an individual is not moving abruptly and staying relatively still (cf. 27,56,93,94).positions when released.Apart from multiple identical items, the puzzles were placed in a way (e.g., in different compartments) that potentially minimised monopolisation by higher-ranking individuals.The experiments started with monkeys regaining access to the enclosures and ended either after one hour or after all puzzles had been solved (i.e., all rewards were obtained), whichever was earlier.
The novel food items used during the first phase of Gr.1 and Gr.3 were rambutans (see Figure S6C) and dragonfruits (see Figure S6D).However, due to veterinary advice and reconsideration, the rambutans were replaced with starfruits during the second phase of experiments.The monkeys had no experience with either of these fruits.In Gr.2, starfruit and dragonfruits were used in both phases.During each experiment, multiple pieces of either intact (rambutans) or halved (dragonfruit and starfruit) fruits were placed in the enclosure ground.The number of novel food items was adjusted according to the group sizes to avoid monopolisation.Like food puzzles, the experiments started with monkeys regaining access to their enclosures and ended after one hour or when all items were eaten.
The two novel objects used were blue and green coloured plastic egg containers (14 cm x 13 cm x 4 cm; see Figure S6E) and wooden massage rollers (24 cm x 10 cm x 4 cm; see Figure S6F).For Gr.1 and Gr.2, six identical objects were used during each experiment and spread about 100 cm apart from each other in the indoor enclosure compartments.In comparison, we used two objects of each type for Gr.3.Since the individuals could carry the objects freely, we recorded the monkeys' activities in indoor and outdoor enclosures.Each experiment lasted for an hour.

Collection and storage of blood samples for genetic analyses
EDTA whole blood samples were obtained from the monkeys during annual veterinary health check-ups to isolate genomic DNA (gDNA) using a standard salting-out purification procedure.The check-ups are distributed over the year and have the target of sampling each animal with an interval of approximately 12 months.The individuals in our study did not show any clinical signs of diseases based on daily care and observations.Besides, monkeys were only included in the annual veterinary health check-ups when they were R 9 months old.The collection of blood samples took place in the morning, and monkeys did not receive food after 1700 hours on the previous day.Progressive refinement procedures were followed to minimise the stress of capture before the sedation process.Monkeys were trained to voluntarily enter the squeeze cage as a part of the refinement procedure.After an intramuscular injection of the sedative, blood sampling was carried out by qualified animal caretakers within a window of 20 minutes.The collection site was sterilised using 70% alcohol.The monkeys gained consciousness within 2 hours of the procedure.Notably, no behavioral experiments or observations were conducted within a week of the health check-ups.
From each monkey in the studied cohort, 30 ml of gDNA with a concentration varying between 46-649 hg/ml (on average 124,5 hg/ml) was shipped to Affenberg Landskron Field Research Lab (field research station of the University of Vienna; Landskron, Carinthia, Austria) for further analysis.

Behavioral coping styles
Self-directed behaviors are considered proxies for quantifying stress-related responses in non-human primates (see, 96,97 but see 98,99 ).To investigate whether the predator models indeed played the role of stressors, we first compared self-directed behaviors (autogroom and scratch combined) between predator exposure experiments (all phases and predator models combined) and regular focal data (phases combined).Frequencies of self-directed behaviors were calculated per minute at the individual level.We calculated the frequencies from focal data when no aggressive interactions were observed for at least five minutes to eliminate any potential effect of social stress.
Videos were coded using Solomon Coder (version beta 17.03.22).The durational (s/hour) and event behaviors (frequencies/hour) were calculated per individual and were corrected for the time spent out of sight.We set up a minimum threshold of 6 minutes of within-sight observation (10% of the total combined observation time/phase) for the individuals to be considered for analyses.Despite attempts to record all individuals with multiple cameras, we could not obtain adequate data from two individuals (two females above three years of age) belonging to Gr.2.Therefore, we revised our sample size for the behavioral part of coping from 32 to 30 individuals.In addition, even if two predators could elicit responses from monkeys quite differently (see 54 ), we decided to combine their data for each phase (at the individual level) to tackle the low occurrences of some of the variables.Three experimenters coded the behavioral coping videos, and inter-rater reliability was calculated.It was found to be high (based on 33%, i.e., 4 out of 12 predator exposure videos; (ICC (3,k) = 0.97, p < 0.001).
The consistency of the behavioral coping-related variables was examined using a two-way mixed model intraclass correlation (ICC (3,1)) analysis.Only variables with sufficient temporal consistency (ICC values R 0.3 and p < 0.05) were included in further analyses.ICC analysis yielded 9 repeatable variables with ICC values ranging from 0.32 to 0.80 (see Table S9).These repeated variables (average values between phases one and two) were used in the EFA with a principal axis factoring (PAF) extraction method.PAF was chosen as it does not assume a multivariate normal distribution, and the data might exhibit significant deviation from a normal distribution.After an initial run of the EFA, four variables, foraging, locomotion, freeze, and yawn, were removed due to low Kaiser-Meyer-Olkin measure of sampling adequacy (MSA) scores.The final model included 5 variables with an overall acceptable MSA score of 0.62.Bartlett's score of sphericity was significant (x 2 = 41.56,p < 0.001); thus, the assumptions of EFA were met.The latent factors were then extracted based on eigenvalues >1.The root mean square of the residuals showed a value of 0.02, which indicated a good model fit.In addition, a Tucker-Lewis Index (TLI) value of 0.95 was obtained, further suggesting that the model fit was sufficient. 100We performed a promax rotation method for meaningful interpretation of the potentially correlated factors and factor loadings.The resulting factors were labelled based on which behavioral variables loaded significantly (R 0.5, positive and negative) onto them (i.e., coping styles 12,15,34,37 ).The scores of each individual per factor were extracted.Due to a relatively small sample size, instead of k-means clustering (see 101 ), we decided to perform the categorization of individuals into different coping styles using a visual inspection followed by a median split approach (cf. 64).The median split approach, in particular, allowed us to employ a conservative categorization.Factor scores were assigned to individuals only if they scored higher than the factor median.Moreover, betweenfactor comparisons were made using Wilcoxon signed-rank tests before assigning them to the individuals as specific coping scores.Effects of sex and age were checked on the obtained coping styles.

Emotional arousal
Following the three criteria of collecting thermal images, we extracted information from the individuals across all phases and predator models.The number of thermal images and individuals was adjusted based on the specific coping styles of the monkeys.One experimenter coded all the thermal recordings; another experimenter coded 5% of the images ($64 images, including whenever individuals were visible) obtained from the videos to check for reliability; it was found to be high ((ICC (3,k) = 0.93, p < 0.001).Not all individuals were equally sampled, as the procedure heavily depended on monkeys approaching the enclosure fences.The valid thermal images were imported to FLIR Tools (version 6.4.18039.1003)for analysis.Nose-bridge, tip, and mid-nose temperatures were extracted from each image.Each image was magnified (606%) consistently, and an area size of 2 x 2 pixels was applied to the designated regions to extract temperature measures (see below figure).Temperature measures from the three regions were then checked for correlation using Spearman Rank correlation tests.High correlations between the nose bridge, tip, and mid-nose regions for facial temperature readings (Spearman's Rank Correlation: nose tip and midnose, rho = 0.99, p < 0.001; nose-tip and nose bridge, rho = 0.99, p < 0.001, mid-nose and nose bridge, rho = 0.99, p < 0.001) was found.Since the mid-nose region is less susceptible to temperature changes due to breathing, mid-nose temperature data were used for the analysis (see 93 ).
Even though the animals were trained to move between in-/out-door enclosures following the commands of caretakers, we could not discard the possibility of emotional arousal associated with this process.Therefore, instead of using potentially unreliable thermal measures (as control) before the introduction of the predator, we decided to divide the 30-min predator exposure period into small time windows for comparison.We did not code data for the first two minutes of the predator exposure to avoid any potential influence of temperature differences between the enclosures and emotional arousal due to the associated lock and release.Additionally, if an individual travelled, during the experiment, between the indoor and outdoor enclosures, a 1-min no coding window for the individual was applied to control for potential temperature differences; besides, data were not coded for 1-min if a focal individual was found vocalising (cf. 27,56,93).We divided the overall duration of the experiments into the following time windows: 2-10 min, 10-20 min, and 20-30 min.Such a division ensured maximised sampling effort from the individuals.Individuals were only included in the analysis if at least one valid data point was collected from each time window.In case multiple data points were found pertaining to one individual within a time window, the average value was used.Since the aim was not to determine how quickly the facial temperature changed upon predator exposure but rather how coping strategies worked throughout the session, we converted all temperature values of the individuals to '0' from the 2-10 min window.This was used as the baseline measure.However, this is not to say that we did not expect a potential drop in temperature upon predator exposure.Temperature changes in the following periods (i.e., 10-20 min and 20-30 min) were compared with the baseline.Finally, we conducted a validation test using locomotion and foraging behaviors to control for the potential confound of the general activity of the monkeys.Following the inclusion criteria, we could extract reliable data on emotional arousal from 12 individuals (six each for aggression-and nonaggression-based coping styles).

Personality
We collected data on all 32 individuals.A total of 10,229 minutes of focal data (Phase 1: mean G standard deviation = 159.76G 2.67 observation minutes/individual; Phase 2 = 159.89G 2.64 min/individual) was collected in addition to the 36 personality experiments from the two Infrared thermal imaging to measure emotional arousal The left panel shows the experimental set-up using the predator hawk in one of the long-tailed macaque groups.The right panel shows a zoomed-in version (606%) where the mid-nose temperature is extracted from a focal individual using a 2 x 2-pixel area.The pixel area provides an upper (up arrow), a lower temperature (down arrow), and an average estimate.We used the average value for analysis.
phases.Four coders coded observational and experimental videos, and inter-rater reliability was estimated using a two-way mixed ICC coefficient (ICC 3,k = 0.96, p<0.001).We used 5% of the observational and experimental data to calculate the inter-rater reliability.
For each phase of the focal observation, durational variables (sec/min) and event behaviors (frequencies/min) were calculated per individual, and the varying observation minutes were corrected.The variables for each category of the personality experiments were coded (also see 53 ) as follows: latency to approach (sec) as the time it took for an individual to move from an initial 5-meter distance to a 1-meter radius of the novel object, novel food, or food puzzle.If no approach was made, latency was scored as the total length of the experiment; proximity (s/hour) as the total time an individual spent in proximity (% 1 meter) of the novel object, novel food, or food puzzle; manipulating (s/hour), or handling (s/hour), as the time an individual spent manipulating or handling the food puzzle or novel object, respectively.The coding of proximity was seized as soon as an individual began manipulating or handling the experimental items.All variables were standardised for meaningful interpretation of the next steps of analyses.
All three groups were used together to view the species-level personality constructs comprehensively.The temporal consistency of the variables was tested using a two-way mixed model ICC, which compared each observational and experimental variable between phases one and two.A conservative cutoff was set in which only variables with ICC values R 0.5 and p < 0.05 were considered temporally repeatable.Variables in which over half (i.e., more than 16 individuals) of the individuals had zero occurrences were excluded.
We found 33 repeatable variables from the ICC analysis with values ranging from 0.50 to 0.95.The average values of these variables between phases one and two were included in a PCA, but the number was further decreased to 17 after removing those with low communality scores (cut-off value = 0.7) and to even 12 after removing those loading in more than one component with similar magnitude.The communality scores of the remaining 12 variables ranged between 72.28% and 94.14%, with an overall Kaiser-Meyer-Olkin measure of sampling adequacy (MSA) value of 0.81 (see Table 1).A scree plot was generated using an unrotated PCA, and the eigenvalue of each potential principal component was reviewed alongside the percentage of variance explained individually and cumulatively.The number of principal components was subsequently decided on by inspecting the plot, eigenvalue scores (>1), and the amount of variance explained cumulatively (>70%).Since personality traits can theoretically be correlated and form behavioral syndromes, we used a direct oblimin rotation technique.Factor loadings R 0.5 (for both positive and negative loading) for the variables were considered significant.Finally, variables were removed if loaded on multiple components with similar magnitude.The resulting principal components were labelled based on the significant behavioral variables loaded on them.The scores of each individual per component, or personality trait, were extracted.Effects of sex and age on the personality traits were investigated.

Dominance hierarchy
Dominance rank relationships were calculated at the group level using a Bayesian Elo-rating method. 102The method was based on calculating winning probabilities, and we particularly used submissive behaviors coded from focal observations (avoid, be displaced, silent-bared teeth, flee, and social presence, see 53 ) to assess the steepness of hierarchies.We checked for the independence of the above behaviors from personality assessment.After constructing the hierarchies, individuals were plotted according to their respective estimated Elo values (see Figure S7).

COMT genotyping
We genotyped our study animals (n=26) for the equivalent site of the human Val 158 Met encoded polymorphism located in exon 4 of the COMT gene.However, in rhesus macaques, the COMT gene encodes a 270 amino acid (aa) long product, which is one aa shorter as compared to the human COMT protein.As such, the human COMT Val 158 Met encoded polymorphism in macaques is located at codon 157. 52Amplification of the target site and sample preparation for sequencing was conducted at the Affenberg Landskron Field Research Lab (field research station of the University of Vienna; Landskron, Carinthia, Austria).Amplification products were outsourced for sequencing to LGC genomics (Berlin, Germany).
We cycled 15 min at 95 C, following 40x 10 sec at 95 C, 40 sec at 60 C and 20 sec at 72 C. To ensure that only a single product resulted from amplification, we included a melting curve analysis ranging from 72 C to 95 C with 0.3 C/s.The experimental samples showing a single peak at the expected melting temperature were purified using the Omega E.Z.N.A. Gel extraction kit (Omega Bio-tek, Inc. Norcross, USA).The spin column was loaded with 20ml PCR sample mixed with 20ml binding buffer.Duplicates of the same experimental sample were pooled.For purification, we followed the manufacturer's instructions with only slight modifications.We added another washing step and refrained from loading the samples on an agarose gel to secure DNA yield.We used the melting curve results to verify the amplification products, and 10 ml of each sample plus 4 ml of forward primer were sent to LGC genomics.The resulting Sanger sequencing electropherograms were evaluated using the CodonCode Aligner software (version 10.0.2;CodonCode Corporation, Dedham, USA).Three additional

Figure 1 .
Figure 1.Exploratory factor analysis (EFA) showing two latent factors Latent factors, namely aggression-and nonaggression-based coping, were identified (n = 30).The behavioral variable loadings are indicated by the positive and negative values, see also TableS9.

Figure 2 .
Figure 2. Mid-nose temperature change during predator exposure The boxplot shows the average mid-nose temperature change during predator exposure.Individual data points are represented using solid dots, squares and triangles (n = 12).The squares and triangles indicate individuals with COMT Val/Met G/G genotype.Boxes represent interquartile ranges, and whiskers represent the upper and lower limits of the data.The horizontal bars within the boxes represent the median values, see also Figure S3.

Figure 4 .
Figure 4. COMT Val/Met encoding polymorphism and behavioral coping stylesThe boxplot shows the two coping strategies and their association with the COMT Val/Met genotype.A significant difference between the coping scores of homozygous G/G and A/A individuals was found (Wilcoxon Rank-Sum test: z = À2.20,h 2 = 0.6, p = 0.02, n = 9).Individual data points are represented using solid dots.Boxes represent interquartile ranges, and whiskers represent the upper and lower limits of the data.The horizontal bars within the boxes represent the median values.

Table 1 .
Output of the principal component analysis Variables, factor loadings (>G0.5),attribute communalities, and variance explained by each principal component are provided in the table.

TABLE
d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS d METHOD DETAILS B Coping B Personality B Collection and storage of blood samples for genetic analyses d QUANTIFICATION AND STATISTICAL ANALYSIS B Behavioral coping styles B Emotional arousal B Personality B Dominance hierarchy B COMT genotyping B Statistical models and packages