Correlated evolution of social organization and lifespan in mammals

Mammals vary widely in both social organization (solitary, pair-living, and diverse group-living forms) and maximum lifespan, which spans roughly 100-fold across species. Prior evidence linking sociality and longevity has largely come from single-species studies, showing that affiliative bonds can enhance survival and reduce physiological dysregulation in humans and some nonhuman primates, though exceptions exist (e.g., negative associations reported in yellow-bellied marmots). Cross-species analyses have primarily focused on eusocial or cooperatively breeding species, leaving uncertainty about whether other social organizations are broadly associated with longevity across mammals. Mechanistically, hypotheses include stress buffering, parasite exposure and immune defenses, and the pace-of-life framework, whereby species with slower life histories invest in long-term social bonds that may yield survival benefits. This study aims to test, at a broad phylogenetic scale, whether social organization and longevity have co-evolved in mammals, and to investigate molecular correlates in brain transcriptomes that might underpin this association.

Evidence from humans and several mammalian species indicates that strong, stable social relationships can reduce mortality and extend lifespan, as seen in chacma baboons and rhesus macaques, while some species show the opposite pattern (e.g., yellow-bellied marmots). Previous multi-species work emphasized eusocial and cooperative breeders, with mixed findings and potential confounding by ecological factors. Proposed mechanisms include stress-buffering effects of social support, social modulation of pathogen transmission and immune function, and links between sociality and slow life histories. Conceptual frameworks for social complexity exist but consensus, quantitative metrics applicable across mammals are still developing, complicating large-scale comparative tests.

Data compilation: The authors assembled a dataset for 974 mammalian species with social organization (solitary, pair-living, group-living, allowing polymorphism), adult body mass, maximum lifespan, activity, lifestyle, and fossoriality from literature and databases (PanTHERIA, PHYLACINE, AnAge, etc.; last search Aug 5, 2022). Maximum lifespan (absolute longevity) and body-mass-adjusted residuals (relative longevity) were used, with long-lived thresholds set at >26 years (3rd quartile) and residuals >1.38; alternative cutoffs (17/35 years; 0.93/1.83) tested robustness. Missing life-history data for some species were imputed using PhyloPars after benchmarking multiple imputation approaches. Phylogeny and signals: The TimeTree mammalian phylogeny was used (with an alternative tree from Upham et al. for sensitivity). Phylogenetic signal (Pagel’s λ) was estimated for social organization (fitDiscrete) and longevity (phylosig). Evolutionary models of social organization and longevity: Using BayesTraits V3 in a Bayesian framework, four models for social transitions among the three states were compared: ER (equal rates), IC (stepwise transitions disallowing solitary↔group direct transitions), ARD (all rates different), and an RJ-MCMC-derived model. Chains ran 100 million iterations (50 million burn-in), hyper-priors on exponential means U[0,2], branch lengths scaled to mean 0.01; stepping-stone sampling (1000 stones, 10,000 iterations per stone) estimated marginal likelihoods; ten independent runs assessed stability. Longevity evolution (long-lived vs short-lived, both absolute and relative) was modeled analogously with ER/ARD/RJ-MCMC comparisons. Comparative analyses and correlated evolution: Phylogenetic ANOVAs (phytools) compared longevity across social states; MCMCglmm regressions modeled longevity (Gaussian) against social state, adult body mass, activity, lifestyle, and fossoriality with phylogeny as a random effect. Correlated evolution was tested using BayesTraits Discrete (binary recodings: solitary vs non-solitary; pair-living vs non; group-living vs non; and absolute/relative longevity states), comparing dependent vs independent RJ-MCMC models via Log Bayes Factors across ten runs. Taxonomic subsampling (50–95% of species, multiple replicates) and alternative longevity cutoffs assessed robustness; analyses were also repeated on an alternative phylogeny. Comparative transcriptomics: Brain RNA-seq was assembled for 267 samples from 94 species (14 orders), combining new data (166 samples from 54 species) and public datasets. Orthologous genes were identified by reciprocal BLAST to human CDS, retaining genes present in ≥70% of species (13,402 orthologs). Reads were aligned with STAR to species-specific ortholog references; counts obtained with featureCounts; batch effects removed by combat_seq; expression normalized by TMM and RPKM, then log2-transformed. For each gene, MCMCglmm models associated expression with social organization (binary or 3-state codings), longevity, body mass, activity, diet, lifestyle (aerial vs non-aerial), with phylogeny as covariance. Significance used pMCMC<0.05 and posterior mean magnitude thresholds determined by fitted distributions. Selection analyses and pathway enrichment: Coding sequences were aligned (MAFFT via TranslatorX; Gblocks filtered). RELAX (HyPhy) tested for intensified (K>1) or relaxed (K<1) selection on test branches defined by social states and long-lived vs short-lived. Pathway-level enrichment used polysel, summarizing gene scores (posterior means for expression analyses or RELAX K values for selection analyses) into SUMSTAT per pathway (Reactome/GO sets), with pruning for overlapping genes and rescaling for gene length/species-number correlations. Significance threshold P<0.05 or |log10 P|>1.3. Cross-talk between expression and selection was evaluated for overlapping genes and pathways.

Phylogenetic signal: Social organization showed strong phylogenetic signal (Pagel’s λ≈0.94 for both multi-state and uni-state datasets), and longevity also showed strong signal (λ≈0.97; n=974, P<0.001). Social organization evolution: The ARD model was best supported (vs RJ-MCMC: Log BF=9.24; vs ER: 33.36; vs IC: 71.70), indicating heterogeneous transition rates. The transition from pair-living to solitary was ~14 times higher than solitary to pair-living (e.g., qpair→solitary ≈ 4.00 ± 1.55 × 10⁻3 vs qsolitary→pair ≈ 0.29 ± 1.50 × 10⁻1), suggesting instability of pair-living. Longevity evolution: RJ-MCMC and ARD outperformed ER. Transitions from long-lived to short-lived occurred ~4 times more frequently than the reverse (absolute longevity: qlong→short ≈ 2.09 ± 3.85 × 10⁻1 vs qshort→long ≈ 0.54 ± 1.42 × 10⁻1; similar for relative longevity). Association between sociality and longevity: Phylogenetic ANOVA showed group-living species live longer than solitary species (absolute longevity: t=12.40, Padjust=0.04; relative longevity: t=12.01, Padjust≈4.8×10⁻2). MCMCglmm (controlling for body mass, activity, lifestyle, fossoriality, and phylogeny) found longer lifespans in pair-living and/or group-living relative to solitary (multi-state dataset n=947: pair vs solitary post mean=0.10, PMCMC=1.11×10⁻3; group vs solitary post mean=0.06, PMCMC<6.0×10⁻4; uni-state dataset n=897: pair vs solitary post mean=0.10, PMCMC<6.0×10⁻3; group vs solitary post mean=0.06, PMCMC<1.11×10⁻3). Longevity correlated strongly with mass (Spearman r=0.71, P<2.2×10⁻16). Correlated evolution: BayesTraits Discrete favored the dependent model for solitary vs non-solitary with absolute long-lived (Log BF=3.18) and for group-living vs non-group-living with absolute long-lived (Log BF=9.58); no support for pair-living. Using relative longevity, dependent models were also supported (solitary Log BF=17.68; group-living Log BF=8.28). Taxonomic subsampling and alternative longevity cutoffs and trees generally upheld these patterns. Directionality: Transitions from short- to long-lived states were higher in non-solitary vs solitary (qnon-solitary ≈ 12.44 ± 1.84 × 10⁻3 vs qsolitary ≈ 2.77 ± 3.79 × 10⁻3) and in group-living vs non-group-living (qgroup ≈ 11.86 ± 1.55 × 10⁻3 vs qnon-group ≈ 2.86 ± 9.89 × 10⁻4), indicating that social (especially group-living) lineages are more likely to become long-lived. Conversely, transitions to group-living were similar in long- and short-lived species, suggesting longer lifespan does not drive the evolution of group-living. Transcriptomics: Of 13,402 orthologs across 94 species, hundreds of genes were associated with social states (solitary: 366 up/254 down; pair-living: 393 up/66 down; group-living: 162 up/321 down). There were 262 genes significantly correlated with longevity across four MCMCglmm models. Thirty-one genes overlapped between social-organization-associated and longevity-associated sets. These clustered into (1) immune-related functions (e.g., UBL7, TNNT3, XRCC6, ATP2A2, NPHS1, KALRN, CIQC, MCL1, ZFP36), and (2) hormonal/neural/signal transduction (e.g., MTM1, SLC29A2, ATP2A2, KALRN, RHOBTB2, SLC6A19, MCL1). Notable examples: XRCC6 was downregulated in solitary, upregulated in group-living, and positively correlated with lifespan; ZFP36 related to immune regulation and neuroprotection. Pathways: Enrichment pointed to hormone signaling and immunity. For example, the eicosanoid synthesis pathway showed opposing correlations with solitary vs group-living and positive association with longevity; immunoregulatory interactions between lymphoid and non-lymphoid cells were downregulated in solitary, upregulated in group-living, and negatively associated with longevity. Selection signatures (RELAX): In solitary branches, more genes showed intensified selection (5448) than relaxed (3200). Pair-living showed 3747 intensified and 4589 relaxed. Group-living showed more relaxed (5570) than intensified (3170). Long-lived lineages had more intensified (5364) than relaxed (3564) genes. Overlap patterns indicated that long-lived states in group-living mammals tend to exhibit relaxed selection relative to solitary (chi-squared tests P<0.001). Twenty pathways were under selection for both social organization and longevity. Expression–selection cross-talk: Eight genes (SHKBP1, MTM1, XRCC6, UBL7, VWASA, PUS3, MCL1, COX7A1) showed both differential expression and selection related to social organization and longevity. XRCC6 experienced intensified selection in solitary (K=2.58, P=4.34×10⁻13), relaxed in group-living (K=0.31, P=1.00×10⁻17), and relaxed in long-lived (K=0.70, P=1.06×10⁻5). The sulfur relay system pathway showed concordant associations in both expression and selection for social organization and longevity.

The findings demonstrate a broad, phylogenetically pervasive association between social organization and lifespan: group-living species generally live longer than solitary species, and lineages that are group-living are more likely to transition into long-lived states. This addresses the long-standing hypothesis that sociality and longevity co-evolve, beyond eusocial or cooperative breeders. The lack of evidence that longer lifespan promotes transitions to group-living suggests causality (at macroevolutionary scales) runs primarily from social organization to longevity rather than the reverse. Mechanistically, the transcriptomic and pathway analyses indicate hormonal regulation and immune processes as shared foundations linking social organization and lifespan, consistent with stress-buffering, social modulation of immunity, and neuroendocrine pathways. Selection analyses reveal that while longevity often shows intensified selection, group-living lineages exhibiting long lifespans display signatures of relaxed selection on many genes, implying that longevity in social species can arise through relaxation in some constraints coupled with coordinated expression changes in key pathways. Together these results integrate macroevolutionary transitions, comparative lifespan patterns, and molecular correlates, offering a cohesive picture of how social systems shape mammalian longevity.

This study provides comparative and molecular evidence that social organization and longevity have co-evolved across mammals, with group-living associated with extended lifespan and elevated probabilities of transitions to long-lived states. Brain transcriptomics implicate immune and hormonal/neural signaling genes and pathways in this association, and selection analyses identify overlapping pathways under selection for both traits. Future work should (1) develop and apply standardized, quantitative measures of sociality and social complexity across taxa; (2) integrate long-term field data on kinship, social networks, cooperation, and conflict; (3) expand multi-tissue, multi-omic analyses to resolve causal mechanisms; and (4) perform experimental validation in model systems to dissect the roles of candidate genes and pathways.

Social organization was categorized into three broad states assuming a gradient from solitary to group-living, but mammalian societies vary widely in composition, dynamics, and social complexity; more precise, consensus measures of sociality are needed. Maximum lifespan (and residuals) were used as longevity measures, and some values were imputed, which may introduce uncertainty. Although analyses controlled for body mass, activity, lifestyle, fossoriality, and phylogeny, unmeasured ecological or life-history confounders may persist. The correlated evolution tests rely on binary thresholds for longevity which, despite robustness checks with alternative cutoffs and trees, can influence inferences. Transcriptomic analyses used brain tissue from diverse species with some heterogeneity in sampling (sex, age, regions), and are correlative rather than causal. Selection inferences from coding sequences (RELAX) may miss regulatory evolution, and pathway-level enrichments depend on current annotations.