A cerebellar substrate for cognition evolved multiple times independently in mammals

Abstract

Given that complex behavior evolved multiple times independently in different lineages, a crucial question is whether these independent evolutionary events coincided with modifications to common neural systems. To test this question in mammals, we investigate the lateral cerebellum, a neurobiological system that is novel to mammals, and is associated with higher cognitive functions. We map the evolutionary diversification of the mammalian cerebellum and find that relative volumetric changes of the lateral cerebellar hemispheres (independent of cerebellar size) are correlated with measures of domain-general cognition in primates, and are characterized by a combination of parallel and convergent shifts towards similar levels of expansion in distantly related mammalian lineages. Results suggest that multiple independent evolutionary occurrences of increased behavioral complexity in mammals may at least partly be explained by selection on a common neural system, the cerebellum, which may have been subject to multiple independent neurodevelopmental remodeling events during mammalian evolution.

Keywords: cerebellum; convergence; evolution; evolutionary biology; intelligence; mammals; neuroscience.

Figure 1.. Artist’s rendering of scans of brain sections from representative species in the sample.

For each species, the left pane depicts a coronal section of the anterior cerebellum (near the facial colliculum of the rhomboid fossa), and the right pane of the posterior cerebellum (first available section in which white matter is no longer present). The dark overlay indicates the medial cerebellum. Note the absence of lateral cerebellar hemispheres in the barn owl. All illustrations are to scale, except for those of the barn owl. Original illustrations of Macaca mulatta, Ursus maritimus, Zalophus californianus, Bos Taurus, and Tursiops truncatus come from www.brainmuseum.org, Tyto alba comes from brainmaps.org, and Hylobates lar comes from the collection at the Vogt Institute for Brain Research (Zilles et al., 2011).

Figure 2.. Phylogenetic generalized least-squares analysis of lateral cerebellar size to medial cerebellar size (left pane; colors as in Figure 3), and of lateral-medial cerebellar reorganization to a measure of domain-general cognition in primates (right pane).

The original measure of domain-general cognition (Deaner et al., 2006) is inversely related to cognitive ability (low scores indicate high cognitive ability). Here, for the purposes of visualization, we inversed this measure so that higher scores indicate a higher cognitive ability. Phylogenetic confidence intervals were computed following Smaers and Rohlf (Smaers and Rohlf, 2016).

Figure 3.. Best-fit adaptive regimes and ancestral phenogram for lateral-medial cerebellar reorganization.

Best-fit adaptive regimes were estimated using a Bayesian reversible-jump procedure for fitting OU models (Uyeda and Harmon, 2014) and confirmed as significant grade shifts using a phylogenetic ANCOVA (Smaers and Rohlf, 2016). Posterior probabilities (PP) of regime shifts were estimated using the Bayesian reversible-jump procedure. Nodal values for the ancestral phenogram were estimated using a multiple variance Brownian motion approach (Smaers et al., 2016). Green data points and branches comprise the convergent regimes of apes, toothed whales and pinnipeds, blue data points and branches comprises those of feliformes and artiodactyls.

Figure 3—figure supplement 1.. Results of the Bayesian reversible-jump Ornstein-Uhlenbeck approach for lateral-medial cerebellar reorganization, relative cerebellum size, medial cerebellum size, and lateral cerebellum size.

Figure 3—figure supplement 2.. Ancestral phenograms using a reversible-jump (‘rjBM’), multiple variance (‘mvBM’) and standard Brownian motion (or ‘constant variance’ Brownian motion; ‘cvBM’) approach.

Figure 3—figure supplement 3.. Lineage-specific rates of evolution relative to a gradual model using the mvBM and rjBM approaches.

Figure 4.. Rate of evolution (σ²) for relative cerebellum size and lateral-medial cerebellar reorganization (left panel), and medial and lateral cerebellar volume (right panel) as estimated by a standard Brownian motion model (Revell, 2012).

Rate ratios and P values are calculated using Q-mode rate analysis (Adams, 2014). Reversible-jump (Venditti et al., 2011) and multiple variance (Smaers et al., 2016) Brownian motion models yield equivalent results.