Sex-speciﬁc social aging in wild African lions

,


RESULTS
Understanding how individuals' social behavior changes with age (referred to as ''social aging'') can help illuminate the role of aging individuals in structuring societies [1][2][3][4][5] and in processes such as disease transmission and information spread. 2,16,17Multiple social aging hypotheses have been proposed, including within-individual changes (e.g., competition avoidance and alteration of spatial behavior) and population-level changes (e.g., selective disappearance and demographic shifts). 3Throughout adulthood, inter-individual variation in social behavior is common and often emerges between the sexes. 18As such, important between-sex differences in social aging are likely but remain understudied.Understanding sex-based variation could provide insights into drivers of social aging in wild populations and highlight the underlying evolutionary and ecological mechanisms.
Additionally, inter-individual variation in social behavior across the lifespan has been found to impact longevity variably across species. 6,8,19Such phenomena are increasingly well documented, particularly in primates, [4][5][6][7] but are often restricted to a single sex. 3 Owing to the often-divergent social strategies of males and females, [20][21][22][23] lifetime sociality is likely to differentially impact fitness and longevity between the sexes. 24As such, drawing broad inferences concerning the impact of sociality on longevity remains difficult, with findings often limited to particular taxa and single-sex examinations. 25Particularly lacking are considerations of such processes in social carnivores.Owing to the unique ecological and evolutionary dynamics of such species, they may provide insights into the drivers and mechanisms of social aging that remain poorly understood.
The African lion (Panthera leo) is unique amongst felids: lions are highly social, with individuals living in egalitarian fissionfusion social groups. 11,14Dispersal is highly sex skewed; most females remain with their natal pride (though some disperse to form new prides), 13 but all males disperse aged 2-4 and spend time in nomadic coalitions before taking residence in their first pride. 14,26Coalitions can be resident within multiple prides at once, with larger coalitions more successful in retaining residencies and enjoying greater reproductive success. 27Aging males exhibit reproductive senescence, likely due to their increasing vulnerability to intruding infanticidal males. 14All female pride-mates breed at similar rates until reproductive senescence begins around $10 years old, 15  shared. 14,28,29While individuals of both sexes spend significant time alone, essential behaviors (e.g., territorial defense, cub rearing, and hunting) are regularly executed in groups, 10,29 meaning singletons struggle to successfully reproduce and survive. 14,27espite the vast body of lion behavior research, little is known about intra-pride social associations, how sociality changes across a lifetime, and the role of sex-specific aging in mediating patterns of sociality.Given the variation in (and timing of) behaviors such as dispersal and reproduction between the sexes, we predict sex-based divergence in investment in the number of associates and strength of associations across the lifetime.
Using 30 years of data from 665 adult lions of the wild Serengeti population, we employed spatial and social network analyses to investigate sex-specific social aging and the association between fine-scale sociality and longevity.We compared individuals' (1) average group size, (2) number of associates (degree), (3) total sum amount of weighted social associations to others (strength), (4) social bond strength to associates (mean strength), and ( 5) the connectivity of their associates to one another (clustering) across their lifetimes within the population.Using an integrated nested Laplace approximation (INLA) model, we directly considered spatial autocorrelation in social network positions, accounting for each individual's average location at each timepoint.In addition, to decipher the effects of fine-scale, intra-pride social associations, we controlled for average annual group size as well as observation frequency, natal vs. foreign birth status, individual GPS collar status, year, current pride, natal pride, and individual identity.We analyzed male and female data separately to allow the comparison of sex-specific social aging and the impacts of sociality on longevity between sexes.
All models were substantially improved by considering spatial autocorrelation, indicating that social connectedness was spatially autocorrelated, so we report those models throughout.All response variables except male clustering showed at least a slight model improvement (DeltaDIC > 2), and the majority showed marked improvements (DeltaDIC > 100; Figure S1).However, accounting for these had little impact on other effect estimates, showing that spatial behavioral changes were unlikely to be driving the observed social aging trends.

Female social aging and longevity
While average group size (average number of other individuals the focal lion was observed with per sighting) provides a simple measure of grouping behavior within a fission-fusion society, degree (number of unique individuals the focal lion was observed with over that year) provides an indication of overall gregariousness.Female lion networks were characterized by a substantial increase in group size with age (estimate 0.081; 95% CI 0.034, 0.127; p < 0.001; Figures 1 and 2).Once an individual's average group size was controlled for, we found no additional relationship between age and the unique number of associates females had (Figures 1 and 2).
The females' total sum of weighted social associations to other females (strength-a measure of an individual's overall social connectivity) decreased significantly with age (estimate À0.072; 95% CI À0.097, À0.048; p < 0.001; Figures 1 and 2).Aging studies commonly test for quadratic effects due to the possibility of non-linear relationships between response variables.
The observed social aging trends contrasted strongly with the relationship between sociality and female longevity.Despite decreasing with age, female intra-sex strength was associated with greater longevity (estimate 0.039; 95% CI 0.012, 0.066; p = 0.004; Figures 1 and 3), confirming that the longevity effects cannot be attributed to selective disappearance.This was similarly true for the relationship between female-male strength and longevity (estimate 0.035, 95% CI 0.003, 0.068; p = 0.030; Figures 1 and 3).These results suggest that females benefit from investing in their social associations across their lifetime.
There was a small negative association between femalefemale clustering and longevity (estimate À0.063, CI À0.126, À0.001; p = 0.047).As such, there is potentially a survival cost to existing within tightly connected social cliques with female pride-mates, compared with groups with more fluid membership.Notably, there was no significant relationship between the number of associates females had (degree) and longevity (Figure 1).

Male social aging and longevity
Similarly to females, there was a positive relationship between age and male average group size (estimate 0.172; 95% CI 0.064, 0.279; p = 0.002; Figures 1 and 2).Even when controlling for individuals' average group size, males also showed a strong and significant increase in their unique female associate numbers (degree) with age (estimate 0.172; 95% CI 0.087, 0.258; p < 0.001; Figures 1 and 2).There was a further negative quadratic effect of age, meaning the number of unique female associates peaked in mid-life (estimate À0.137, 95% CI À0.199, À0.075; p < 0.001; Figures 1 and 2).There was no significant change in the number of male-male associates with age (Figures 1 and 2).

DISCUSSION
Our findings demonstrate strong, sex-dependent, and oftennonlinear changes in social behavior with age in a wild carnivore, the African lion.These relationships existed even after accounting for individuals' group size changes, and the metrics of finescale social behavior were also related to longevity.It is plausible that the strongly divergent patterns of sex-based social aging relate to the profound differences in reproductive biology and social behavior between males and females.Similar to the orthodoxy from studies of other mammals (e.g., primates and ungulates 2,3 ), aging females appeared to broadly lose social connectivity while retaining a similar absolute number of associates.In contrast, males' connectivity was highly nonlinear and peaked in early and late life, showing that there are strongly divergent drivers governing the expression of social behavior.This illustrates social aging patterns should be considered more broadly in social animals, potentially uncovering a range of further processes governing the underlying causes and consequences of ecological and behavioral processes. 1,2As theorized in primates specifically, 5 animals with highly divergent sex roles in their society may exhibit similar divergence in social aging patterns due to shifts in motivation and power; further comparative work may help to elucidate the wider drivers and consequences.
Group size and number of associates Due to the fission-fusion structure of lion society, individuals' group size and the total number of unique associates they are observed with (degree) provide two informative yet relatively simple social measures.There was a significant increase in female sub-group size with age but a lack of change in associate numbers (when controlling for group size changes).This could partly be explained by reproductive senescence, 15 as mothers  S1.

Figure 2. Social aging patterns vary between the sexes
Age related changes in (from top to bottom) group size, degree, strength, mean strength, and clustering for females (left column) and males (right column).Red points and lines indicate associations to females, blue points and lines indicate associations with males, and purple points and lines indicate associations to both sexes combined.Where significant, quadratic age effects are included in the plot, generating n-or u-shaped curves.The black lines represent the mean of the posterior distribution for the age effect estimate; the colored lines are 100 random draws from the posterior to represent uncertainty.The points represent individual-by-year replicates, with transparency to allow for visualization of overplotting.For females, group size, strength, and mean strength plots show significant relationships.For males, group size, degree (to females only), strength, and mean strength (to males only) plots show significant relationships.See also Figure S1 and Table S1.

Figure 3. Sociality correlates with longevity differentially between the sexes
Longevity effects of (from top to bottom) group size, degree, mean strength, and clustering for females (left column) and males (right column).Red points indicate associations to females only, while blue points indicate associations with males only, and purple points and lines indicate associations to both sexes combined.The black lines represent the mean of the posterior distribution for the age effect estimate; the colored lines are 100 random draws from the posterior to represent uncertainty.The points represent individual-by-year replicates, with transparency to allow for visualization of overplotting.For females, strength and clustering plots show significant relationships.For males, the degree plot shows a significant relationship.See also Figure S1 and Table S1.
are highly dependent on small creche groups while cub rearing. 28,29Furthermore, females frequently hunt alone or in substantially smaller groups than the maximum pride size. 14,29In other social apex predators, declines in hunting success with age have been observed, 30 and aging female lions do show signs of deteriorating physicality that could impair their hunting ability.It is plausible that aging females become increasingly dependent on the pride for food, rather than hunting themselves, and so are observed more frequently in bigger group sizes.Understanding female hunting behavior across the lifespan could elucidate this further.
Similarly, male sub-group size increased with age, which could reflect the increasing likelihood of a male's pride being taken over by intruding males. 14Further research into whether aging males spend an increasing proportion of their time with the bulk of the pride, in recognition of their physical vulnerability, would be useful.While female associate numbers did not change with age, males exhibited an increase in female associate numbers to a mid-life peak even when controlling for group size changes.This is likely because, in the Serengeti, many coalitions are resident within multiple prides simultaneously.These prides grow in size to a tipping point; however, pride residencies can also be surrendered by coalitions as they age, particularly if their non-dispersing daughters have reached reproductive maturity. 14cross the lifetime, neither yearly average group size nor yearly number of associates held statistically significant relationships with female longevity, suggesting that the extent of grouping behavior individuals display does not directly relate to survival.However, male longevity increased when their associate numbers were higher, and this was true when only considering their associations to females, or to other males.This could be partly indicative of selective disappearance 3 in that larger coalitions are more successful in maintaining pride residence, and therefore persisting in the study area, when compared to smaller coalitions. 14Further, males resident within larger prides (or multiple prides) may simultaneously have access to better quality territory, 10 which would positively contribute to their survival.

Connectivity and bond strength
Lions demonstrated significant reductions in their overall connectivity to same-sex individuals (intra-sex strength) and average bond strength to their associates (mean strength) with age.Sex-specific differences were again found: female lions exhibited a peak in both metrics in mid-life to males only.Males exhibited early-and late-life peaks in strength and mean strength when considering their male associates only.This contrasted with the evident mid-life peak in strength to females.
These findings may reflect the differing influence of reproductive behavior on social aging between the sexes.Females are reliant on the creche for cub rearing and protection from infanticidal males during their prime adult years. 31As reproductive senescence progresses, investment in creche-mate associations may decrease.Additionally, the early-life peak in mean strength mirrors the time of highest vulnerability for females that do disperse, 14 so this could be when they are most dependent on their cohort for survival.For resident males, the biggest threat to their reproductive output is intruding male competitors.Counteracting this threat requires early-life investment in coalition bonds during the natal pride and nomadic life stages.These within-coalition bonds may once again become important as males age and become more vulnerable to intruding males. 13onversely, aging males' drop off in overall connectivity to females (male inter-sex strength) could reflect reproductive senescence and perhaps even post-pride residence social behavior when intra-sex associations will dominate. 14cross their lifetime, lionesses that invested more in their social associations (with greater intra-and inter-sex strength) appeared to live longer.The direct causal mechanisms underpinning this remain unclear; however, where similar patterns exist in primates, it has been hypothesized that females experience fitness benefits from their associates through increased buffering against stressful events 6 and improved cooperation in shared behaviors. 7Both could tangibly impact female lion longevity given the threat frequently posed by infanticidal males, and the extensive cooperation required during territorial defense, hunting and cub rearing.Another possibility is that females are more likely to emigrate from the study area if they are less strongly connected to others, generating the perception that increased connectivity improves longevity.

Cliquishness
For females, there was a slight but significant negative relationship between longevity and their embeddedness in highly clustered intra-sex groups (as opposed to groups with relatively frequently shifting membership).This may reflect the higher rate of fission-fusion in larger prides, which hold better quality territories. 10Similarly, this suggests that a tendency for individuals to be overly cliquish (and less integrated in the wider population) holds relative costs, compared with being more socially transient and having connections outside of their immediate clique.

Ecological consequences
These findings will have a range of ecological and evolutionary implications for age-structured lion societies.For example, lionesses' decrease in social connectedness could drive reduced exposure to pathogens, while males' tendency to gain extra unique associates might disproportionately predispose them to acquiring novel parasites. 16,32Such changes in exposure rates could interact with their reproductive investment and resulting changes in immune resistance, 33 as well as with immunosenescence. 16Given that this population has previously experienced high-mortality disease outbreaks, 34,35 understanding these phenomena could be important for predicting the resilience of the Serengeti lion population in the future.Similarly, there could be important consequences for social learning, as knowledge about the environment is likely to improve with age. 36As such older lionesses are likely to be key resources for prides, but their tendency to become less connected could reduce the ability for information to spread.Yet more research into how behaviors spread socially in wild social networks with diverse structures is now needed. 37

Conclusion
We found evidence of sex-differentiated patterns of social aging, and links between lifetime sociality and longevity, in African lions.The individual-based and long-term nature of our data, as well as our incorporation of both age and longevity effects, confirm that seasonal calendar, running from the start of the dry season to the end of the wet season.Following the ''gambit of the group'' approach, 38 individuals observed in the same group according to the date and GPS location recorded (as described in the ''Data collection'' section above) were classed as associating at that time-point.This process generated a ''group-by-individual'' matrix, and the Simple Ratio Index (SRI) was used for deriving dyadic association matrices (i.e., social networks) between individuals, where the proportion of overall sightings in which two individuals were observed together was calculated rather than the absolute count (see Farine and Whitehead 39 for more details).All SRI values range between 0 (individuals never seen together) and 1 (individuals always seen together).
Using the annual networks, we derived 5 commonly used metrics to characterize lion social behavior.
(1) Group size: the average number of other individuals the focal lion was observed with per sighting (2) Degree: the number of unique individuals the focal lion was observed with over the course of the year (i.e., number of the node's network edges) (3) Strength: the total weighted sum of the focal lion's associations within the network over the course of the year (i.e., sum of the node's weighted edges, thereby quantifying overall connectivity) (4) Mean strength: the average value of the focal lion's weighted associations to its non-zero associates over the course of the year (i.e., mean of the node's weighted edges, thereby quantifying bond strength) (5) Clustering: the propensity for a focal individual's associates to also be associated with one another (with higher values indicating greater ''cliquishness'')

Statistical analysis
We fitted linear models using the Integrated Nested Laplace Approximation (INLA), which is a deterministic Bayesian modelling framework.We fitted three broad classes of models: those examining the population as a whole, only females, or only males.We only included individuals aged 3 and above.This analytical setup is based on an established method for differentiating withinand between-individual age effects 40 which we have used in social aging models before 3 : by fitting age, longevity, and individual identity, the model controls for selective disappearance of certain individuals (i.e., of the more or less social) through mortality.This allows us to identify whether age-related changes in social network position are resulting from within-individual behavioural aging, or whether such patterns might emerge through population-level structural changes.
As such, each model included the following fixed explanatory variables: age (continuous, in years); Age 2 to detect quadratic shapes (continuous, centred around the mean value for age); Longevity (continuous, calculated as age at disappearance from study site in years; Observations (continuous, calculated as the number of times an individual was seen that year); Group size (continuous, calculated as an annual mean from the observation data); Foreign (binary variable, representing whether or not the individual was born in the study area); and Collared (binary variable, representing whether or not they had a GPS collar).We also included a suite of random effect factors: Natal pride (the pride in which an individual was born); Pride (the pride in which an individual was most often seen each year); Year; and Individual identity.We did not fit a Maximum Pride Size factor in addition to the Group Size and Pride based on the comparable results of prior analyses (see Figure S2 for SPDE model including Max Pride Size; Figure S3 for correlation between Max Pride Size and Group Size).In the overall models, we also fitted some effects to examine differences between sexes: first a main sex effect, and then interactions between sex and our continuous variables (Age, Age 2 , Longevity, Observations, and Group size).
INLA models allow the fitting of a stochastic partial differentiation equation (SPDE) effect to account for spatiotemporal autocorrelation.Controlling for spatial autocorrelation is potentially important because in this and other wildlife populations, older individuals could exhibit changes in spatial behaviour that drive changes in social behaviour. 3,16Further, spatial patterns of environmental structure or demography could drive social changes across the study area in a way that confounds with age structuring. 1 As such, isolating the spatial and social components of behavioural aging patterns is therefore likely to be important for ecological understanding.
To investigate spatial dependence in this way, we used each individual's average X and Y coordinate in each year (i.e., their annual spatial centroid); the effect modelled the expectation that individuals will have more similar response values (i.e., social network positions) when their locations were closer in space.To assess whether the models were significantly spatially autocorrelated, we first fitted the base model, and then added the SPDE effect and compared the model fit using deviance information criterion (DIC).We took 5 DIC to differentiate between competitive models, such that if the SPDE effect reduced DIC by more than 5 it was taken to be significantly autocorrelated.The base model + SPDE effect estimates for each of the single sex models can be found in Table S1, and the spatial DIC changes can be seen in Figure S1.
with cubs raised in creche groups where caring and nursing responsibilities are Current Biology 34, 4039-4046, September 9, 2024 ª 2024 The Author(s).Published by Elsevier Inc. 4039 This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Figure 1 .
Figure 1.Associations between age, longevity, and social metrics Estimates taken from the SPDE single-sex model for females (left) and males (right) showing the age and longevity effects for each social metric tested.Points represent the model estimate with error bars showing the 95% credibility intervals (estimated according to the 95% upper and lower quantiles of the posterior density distribution for each effect).Bold points indicate a significant relationship; faded points indicate a non-significant relationship.See also Figure S1 and TableS1.