Brain to body allometry and patterns of encephalization in carnivorans
We measured crania of 361 specimens belonging to 158 terrestrial carnivoran species (Table S1). We completed the dataset with published data for 16 more species [58], for a total of 174 carnivoran species. Because all information on the ecology or metabolism are not available for every species in the literature, two different datasets were analysed. The ‘dataset 1’ includes all 174 terrestrial carnivoran species, and the ‘dataset 2’ includes the 124 species for which all information on ecology, environment, social complexity, and physiological predictors are available from literature.
As expected, our PGLS analyses identified a significant positive brain to body mass correlation within Carnivora (Figs. 1A, 1C). This emphasizes the major influence of the body mass on the evolution of brain size within carnivorous species. Indeed, the body mass alone explains 87 % and 91% of the observed brain size variation in both datasets (dataset 1: R² = 0.87, p < 0.001, λ = 0.56, dataset 2: R² = 0.91, p < 0.001). In addition, our analyses revealed the strong influence of the phylogenetic history in carnivoran brain evolution. In particular, we found a significant phylogenetic signal of the relative brain size (datasets 1 & 2, λ = 0.99; p < 0.001), body mass (datasets 1 & 2: λ = 0.98; p < 0.001), and encephalization quotient (dataset 1: λ = 0.56; p < 0.001; dataset 2: λ = 0.66, p < 0.001).
Many studies have already highlighted the need to consider the allometric relationship between brain and body masses when studying the evolution of brain size [50,51]. Regarding carnivoran brain evolution, our standardised major axis method (SMA) revealed no significant differences for the allometric body-brain slope between the 13 families included in this study (Table 1A, Fig. 1A). We found the same result when considering the two suborders Caniformia and Feliformia separately (Table 1B, Fig. 1C). In a consistent way with previous studies, the slope characterizing the encephalization within all carnivorans was 0.67, meaning that body mass increases faster that brain mass (Table 1C).
The encephalization quotient (EQ) that we quantified exhibits considerable variability within carnivoran species, ranging from 0.73 (the African striped weasel, Poecilogale albinucha) to 0.85 (the spotted linsang, Prionodon pardicolor) (Table S2, Fig. 1E). We found significant differences in the encephalization quotient distribution between families (F = 9.689; p < 0.001) (Fig. 1B). In particular, Ailuridae, Canidae, Ursidae, Procyonidae, Hyaenidae, Mustelidae and Felidae families display larger EQs whereas Herpestidae, Eupleridae, Nandiniidae, Viverridae, Mephitidae and Prionodontidae families possess smaller EQs, as expected following the allometric model (i.e., the body mass being the only predictor of brain size). Analyses performed at the suborders level highlighted a significant difference of encephalization between caniformian and feliformian species (F = 30.55; p < 0.001), with caniformians having a higher EQ on average than feliformians (Figs. 1C, 1D).
Tempo of encephalization
Using a phylogenetic ridge regression approach, our results identified three significant shifts in the evolution rate of the encephalization compared to the average rate computed over the rest of the tree branches (Fig. 2). Two nodes showed a significant increase in the evolutionary rate of the EQ in terrestrial carnivorans: the node including all Canidae species and the node within the Mustelidae family including the Helictindinae (i.e. badgers), Guloninae (i.e. martens, fisher, tayra and wolverine), Ictonychinae (i.e. Grisons and polecats), Mustelinae (i.e. weasels, ferrets, and minks) as well as Lutrinae species (i.e. otters). By using random topologies computed from the original phylogeny via tree swapping, we found that evolutionary rate shifts for EQ were correctly identified for these two nodes for 76% and 81% of the computed random trees respectively (Table S2). In contrast, we identified a significant decrease in the evolutionary rate of EQ for the node including the Herpestidae, Hyaenidae, and Eupleridae families (i.e. Malagasy carnivorans). Nevertheless, it is worth noticing that this decrease of evolution rate was identified in only 23% of the computed random trees (Table S2).
We investigated the influence of 13 variables classified into four different categories: ecological, environmental, social, and physiological (Table S3). We first tested for collinearity between predictors using both correlation test and variance inflation factors (VIFs) [59]. Although our results showed significant correlations between some predictors (Table S4, Fig. 3), VIFs results did not exceed 3.4 in all cases, meaning that all variables were more related to the response than to other predictors. We then assessed the correlation between the estimated EQ and all predictors prior to PGLS analyses (Supplementary Fig. 1). PGLS analyses were performed on the dataset 2 including all 124 carnivoran species for which all information were available.
The model that best explains the evolution of encephalization in our sample of carnivoran species includes the geographic range and the home range combined for both the ‘Group size’ and Social complexity’ analyses (Table 2A). However, this model only explains 12% of the variation in the encephalization observed. While the home range is positively associated with encephalization, we found that the geographic range exhibits a significant negative correlation with the EQ.
The evolution of encephalization in the suborder Caniformia is best explain by a model including the geographic range, the home range, and the ability to hibernate (Table 2B, Caniformia). This model explains 28% of the encephalization variation observed within caniformian species. We found a different result for the feliformian species with only the temperature being a good predictor for encephalization in this suborder (Tables 2B, Feliformia) which is negatively associated with the EQ. Nevertheless, this model only explains 7% of the EQ variation for this taxonomic group.
Our results also revealed different scenarios for the evolution of encephalization when analyses were conducted at the family level. Similarly to the analyses computed for all carnivoran species, both the geographic range and the home range are the best predictors for EQ evolution within Canidae species and this model explains 23% of the EQ variation observed (Table 2C, Canidae). For the Felidae family, the geographic range is the best predictor of encephalization, with the EQ being negatively correlated with it and explaining 21% of the EQ variation observed (Table 2C, Felidae). Finally, we found two different models explaining the EQ evolution within Mustelidae depending on whether the group size or the social complexity was used as social variables (Table 2C, Mustelidae). When considering the group size, the evolution of the Mustelidae EQ appears to be best predicted by a model including the litter size, the temperature, and the group size, which explains 49% of the EQ variation. When analyses were conducted with the social complexity as social predictor, we found that the litter size and temperature best predicted the EQ in Mustelidae species, with 45% of the observed EQ variation being explained. The litter size appears to be negatively correlated with the EQ whereas the temperature is positively associated with the EQ variation in this family.