The evolution of brain neuron numbers in amniotes

Significance The evolution of brain processing capacity has traditionally been inferred from data on brain size. However, similarly sized brains of distantly related species can differ in the number and distribution of neurons, their basic computational units. Therefore, a finer-grained approach is needed to reveal the evolutionary paths to increased cognitive capacity. Using a new, comprehensive dataset, we analyzed brain cellular composition across amniotes. Compared to reptiles, mammals and birds have dramatically increased neuron numbers in the telencephalon and cerebellum, which are brain parts associated with higher cognition. Astoundingly, a phylogenetic analysis suggests that as few as four major changes in neuron–brain scaling in over 300 million years of evolution pave the way to intelligence in endothermic land vertebrates.

Figures S1 to S10 Tables S1 to S8 Other supplementary materials for this manuscript include the following: Datasets S1 to S2 Supplementary Figure 1. Neuron densities go down with increasing brain structure mass across amniotes. Log-log plot of neuron densities against whole brain or brain part mass. The lines represent PGLS regression for the different groups.
Supplementary Figure 3. Scaling of brain neurons with brain mass as estimated by bayou analysis for the 251 amniote species.
Supplementary Figure 4. Scaling of telencephalic neurons with brain mass as estimated by bayou analysis for the 251 amniote species.
Supplementary Figure 5. Scaling of cerebellar neurons with brain mass as estimated by bayou analysis for the 251 amniote species.
Supplementary Figure 6. Scaling of rest of brain neurons with brain mass as estimated by bayou analysis for the 251 amniote species. Figure 7. Shifts in neurons-body scaling in amniotes. (A) Tree colors correspond to neuron numbers relative to body mass, with blue colors indicating low neuron numbers and red colors high neuron numbers. The arrows indicate the branches with shifts in allometric relationship between body mass and neuron number (resulting in either an increase in neuronsarrow up, or a decrease in neurons -arrow down) for the whole brain, telencephalon, cerebellum and rest of brain, identified by reversible-jump Markov chain Monte Carlo analysis with posterior probability > 0.7 for clades including more than 3 species. (B-E) Log-log plots of neuron number for body mass with regression lines for the distinct regimes identified by PGLS analysis.

Supplementary
Supplementary Figure 8. Phenograms showing the evolution of telencephalic and pallial neuron numbers relative to body mass over time. Numbers of telencephalic and pallial neurons relative to body mass in mammals and birds are plotted side by side for comparison. The x-axis is flipped in birds, so that 0 (the present) is in the middle and the axis extends symmetrically left and right Supplementary Figure 9. Large relative brain size is positively associated with high relative neuron density within primates. Relative brain size and relative neuron density refer to residuals from regression of neuron density on brain size and brain mass on body mass, respectively. The line is a PGLS regression line for primates. Figure 10. Example of how the fold change in the number of neurons for body mass was calculated. The scaling coefficients are from PGLS regression of log10-transformed values. The mean difference refers to the difference in trait mean relative to body mass compared to reptiles. NA, "not applicable". ΔAIC and p-value refer to the comparison between the best-fit model and the null model (no allometric shifts). ΔAIC and p-value refer to the comparison between the best-fit model and the null model (no allometric shifts). Table S4. Neuron-brain structure scaling rules for the allometric grades identified by PGLS analysis in the dataset with imputed data for olfactory bulbs and striatum. ΔAIC and p-value refer to the comparison between the best-fit model and the null model (no allometric shifts). ΔAIC and p-value refer to the comparison between the best-fit model and the null model (no allometric shifts). Values represent ratios between rates of the group in the row and the groups in the columns. NA means "not applicable". Statistically significant differences between groups: ***, p < 0.001; **, p 0.01 -0.001; *, p 0.01 -0.05; no symbol, p > 0.05 Values represent ratios between rates of the group in the row and the groups in the columns. NA means "not applicable". Statistically significant differences between groups: ***, p < 0.001; **, p 0.01 -0.001; *, p 0.01 -0.05; no symbol, p > 0.05. Rest of brain 0.0023 0.75 0.83 NA σ 2 indicates the rate of evolution of allometric residuals. Values represent ratios between rates for the structure in the row and the structures in the columns. NA means "not applicable". Statistically significant differences between groups: ***, p < 0.001; **, p 0.01 -0.001; *, p 0.01 -0.05; no symbol, p > 0.05.