Epigenetic predictors of species maximum lifespan and other life history traits in mammals

Abstract

Maximum lifespan is an intrinsic characteristic of a biological species and is defined as the longest time an individual of a species has been reported to survive. By analyzing 15K samples derived from 348 mammalian species representing 25 taxonomic orders we previously identified CpG methylation sites associated with maximum lifespan. Here we present accurate DNA methylation-based (DNAm) predictors of maximum lifespan (r=0.89), average gestation time (r=0.96), and age at sexual maturity (r=0.85). Our DNAm maximum lifespan predictor indicates a potential innate longevity advantage for females over males in 17 mammalian species such as humans, red deer, and cattle. The DNAm maximum lifespan predictions do not vary significantly by caloric restriction and partial reprogramming. Genetic disruptions in the somatotropic axis, which includes growth hormone, IGF-1, and their related receptors, have an impact on DNAm maximum lifespan only in select mouse tissues. Cancer mortality rates in major mammalian orders show no correlation with our epigenetic estimates of life history traits. The DNAm maximum lifespan predictor does not detect variation in lifespan between individuals of the same species, such as between the breeds of dogs. We also present the first prototypes of accurate pan mammalian DNAm predictors of sex and tissue type.Collectively, our findings indicate that maximum lifespan is determined, at least in part, by an epigenetic signature that is an intrinsic property of each species and is distinct from the signatures that relate to individual lifespan, which is unaffected by interventions influencing the mortality risk of individuals.

Cross-species and tissue imputation of species-level DNA methylation samples across mammalian species

“…We next applied the imputed species-tissue combination mean samples to analyze the relationship between species-level methylation values and species’ maximum lifespan 10,16 . For this we followed a similar approach to Li et al 16 and performed linear regression to predict the logarithm of a species’ maximum lifespan based on methylation data.…”

“…We next applied the imputed species-tissue combination mean samples to analyze the relationship between species-level methylation values and species’ maximum lifespan 10,16 . For this we followed a similar approach to Li et al 16 and performed linear regression to predict the logarithm of a species’ maximum lifespan based on methylation data. Specifically, we first performed a linear regression analysis in a tissue-agnostic setting based on the average of combination mean samples within a species to see if similar predictive performance could be achieved with CMImpute’s imputed data for tissue types without observed data as could be with observed data for tissue types in which observed data was available (Methods).…”

“…For comparison, we also conducted a similar set of clustering and visualizations for the data imputed from the baseline methods (Supplementary Fig. [15][16][17]. This showed that the logistic regression and tissue baselines did not show clear clustering of samples corresponding to species (Supplementary Fig.…”

“…There is thus extensive interest in profiling DNA methylation in humans 5,6 as well as other species 4,[7][8][9] . In addition to studying DNA methylation profiles in individual species, insights have been gained from comparative epigenomic analyses across species as epigenetic information from one species will likely be informative to another species [10][11][12][13][14][15][16] . Along with varying at the species level, DNA methylation levels typically vary significantly across different tissue types and thus associate with cell and tissue identity.…”

“…CMImpute specifically imputes samples representing a species’ mean methylation within a specific tissue type, henceforth referred to as a species-tissue combination mean sample or for short combination mean sample . Given the association of DNA methylation with species-level characteristics and cell and tissue identity, these types of combination mean samples have proven useful in cross-species epigenetic studies 10,11,16 . CMImpute takes as input exclusively methylation data with corresponding species and tissue labels to output combination mean samples.…”