Lipidome determinants of maximal lifespan in mammals

Abstract

Maximal lifespan of mammalian species, even if closely related, may differ more than 10-fold, however the nature of the mechanisms that determine this variability is unresolved. Here, we assess the relationship between maximal lifespan duration and concentrations of more than 20,000 lipid compounds, measured in 669 tissue samples from 6 tissues of 35 species representing three mammalian clades: primates, rodents and bats. We identify lipids associated with species' longevity across the three clades, uncoupled from other parameters, such as basal metabolic rate, body size, or body temperature. These lipids clustered in specific lipid classes and pathways, and enzymes linked to them display signatures of greater stabilizing selection in long-living species, and cluster in functional groups related to signaling and protein-modification processes. These findings point towards the existence of defined molecular mechanisms underlying variation in maximal lifespan among mammals.

Figure 1

Dataset. (a) Phylogenetic tree of the 35 species used in this study. Colors show the clade identity and the MLS of species, with darker shades representing longer MLS. (b) Species’ MLS distribution in years. Each dot represents a species. Colors are as in panel a. (c) Number of lipid compounds measured in each tissue. (d) Relationship between species’ MLS and body mass. The colors are as in panel a. Open circles indicate species deviating from the linear relationship between MLS and body mass. The dashed line shows linear model fit to the remaining species (F-test, p < 0.05). MLS was normalized to the maximal MLS value within each clade. NMR – naked mole-rat, HM – human, GS – grey squirrel, CBWB – common bent-winged bat, RBFB – rickett’s big-footed bat, CHB – chinese horseshoe bat. (e) Number of lipid compounds of each tissue after removal of the compounds related to the confounding factors.

Figure 2

Long lifespan prediction. (a) Accuracy of predictive models in each tissue. (b) Proportion of lipid compounds identified as MLS predictors in each tissue. (c) Proportion of MLS predictors shared between tissues. Black borders indicate significant overlap (p < 0.05). (d) AUC of predictive models, constructed on training sets excluding an increasing proportion of randomly sampled individuals of one clade, with the models’ accuracy, tested on the clade individuals excluded from the training set. Identity of the test clade is indicated by the curve color. The horizontal gray line and shaded region represent mean and standard deviation of the random predictor accuracy calculated by randomizing the output variable 100 times.

Figure 3

Long-lifespan related lipids. (a) Concentration level change and double bond distribution of long MLS predictors belonging to two lipid sub-classes in six tissues. Change in the concentration level is calculated as the difference between mean of the concentration levels in the long-living species and all other species in each of the clades. Left panel shows median concentration change of long MLS predictors belonging to simple Glc series and triacylglycerol lipid sub-classes in long-living species. The error bars show the 0.25 to 0.75 inter-quintile intervals. Significant concentration shift towards lower and higher values in long-living species is indicated by blue and red colors, respectively. Significant enrichment of MLS predictors among detected compounds in a given lipid sub-class is shown by thick lines and asterisks. The vertical dashed line shows the median concentration change for all long MLS predictors. Right panel shows the median double bond number and its 0.25 to 0.75 inter-quintile intervals for the same lipid sub-classes as in the left panel. Significantly higher or lower number of double bonds in a given lipid group is shown by red and blue color respectively, as well as thick lines and asterisks. The vertical dashed line shows the median number of double bonds in all long MLS predictors. (b) Enzyme DNA sequence variation. Distributions of the dN/dS values of the enzymes linked to the lipid predictors of long MLS (darker colors), and the enzymes linked to other lipids detected in our dataset (lighter colors), in brain and non-neural tissues (liver, muscle, and kidney) in the long-living species. Asterisks indicate p-value range *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 4

Functional modules. The network nodes represent GO terms significantly enriched in enzymes linked to the lipid predictors of long MLS, and having lower dN/dS values in the long-living species compared to the short-living species. The edges represent enzymes shared between terms with line width proportional to the number of shared enzymes: from 1 to 12. Colored squares within each GO term node indicate the tissue in which these enzymes are linked to the lipid predictors of long MLS: green – liver, dark blue – muscle, light blue – kidney, magenta – heart, orange – brain cortex, brown – cerebellum. Sub-classes of the lipid predictors of long MLS linked to the enzymes in each functional group are listed next to the corresponding network module. The colors indicate the tissues where they are used as predictors of long MLS, according to the same color code as the GO terms.