Despite its importance, little is known on the evolution of the mammalian brain. Previous work suggests that body size and behavioral function are intertwined in their influence on the evolution of the brain 1,2. Most previous studies focused on examining gray matter and sometimes also white matter volume. Few studies, however, investigated white-matter connectivity maps. Because the underlying white matter connectome facilitates the connections between gray matter areas, it must have simultaneously evolved to support gray matter evolution. Hence the evolution of white matter connectivity is at least as important as that of gray matter. In this work we used a wide comparative approach relying on diffusion MRI based fiber-tracking to reconstruct whole-brain structural connectomes and explore its evolution. Whole brain samples of 98 species covering almost all mammalian orders (excluding monotremes) were scanned (Mammalian MRI database = MAMI). The samples were of wild-life animals or animals that expired in regional zoos whose brain was freshly excised, fixated in formaldehyde and scanned subsequently (after rehydration in PBS). For some of the species, more than one brain specimen was obtained resulting in 152 scanned samples. Brain samples who were smaller then 72mm in the anterior-posterior and left-right directions were scanned on a 7T small rodent scanner (biospec 30/70) and larger brains were scanned on a clinical MRI scanner (Prisma 3T). The imaging protocol included an anatomical scan (T1- or T2-weighted MRI) and a HARDI scan at 64 gradient directions with b=1000 s/mm2 and additional 3 images with b=0. The matrix size was kept similar across all samples so that the resolution was adjusted to the brain size. The matrix size for the anatomical scans was 160x120 over ~74 slices (in the human brain that corresponded to a 1.2x1.2x1.2mm3 resolution), the HARDI acquisition was performed with matrix size of 128x96 and ~58 slices (in the human brain that corresponded to a 1.7x1.7x1.7 mm3). SNR was kept roughly similar across samples (leading to variable acquisition time). Fiber tracking was performed on the HARDI data set using spherical harmonics deconvolutions 3. Following fiber-tracking a connectivity matrix (from each voxel to all others) was computed and graph related indices (e.g. mean short path, density, clustering coefficient, efficiency etc) were computed and compared across all samples. HARDI and fiber-tracking analysis were performed in ExploreDTI 4 while graph theory analysis was computed using in house matlab scripts.

Fiber-tracking and network analysis of the MAMI database (example shown in Fig. 1) revealed that:

1. The ratio between gray matter and white matter was linear on a logarithmic scale with a slope 1.23 similar to literature values (Fig. 2).

2. The number of fibers (streamlines) was roughly constant across mammals as well as the connectome mean short path (Fig. 3).

3. A large variability in the formation of callosal fibers was observed (with several species from different orders lacking the corpus callosum) which was compensated in the association mass connectivity efficiency (Fig. 4).

4. Network analysis revealed that the connectome across mammals can be regarded as a small world network. The averaged short path between two points in all brains is 5-6 nodes.

5. Many aspects of the connectome divert from the phylogenic tree.

The MAMI database and fiber-tracking analysis allows constructing a fine-grained description of brain connectivity crossing many different species from distinct evolutionary stages and diverse ecological environments. Surprisingly it was found that the efficiently of connectivity is evolutionary preserved and independent of brain size (for comparison: the smallest brain volume that was scanned was 0.14ml (trident bat) and the largest was 1980ml (stripped dolphin)). While the different species that were sampled differ dramatically in many aspects of the lives, the brain representation of their cognitive system must be different. However, it appears that enhancement of in one cognitive system may lead to different local connectivity pattern while preserving the total connectivity mass and its efficiency. The presented analysis is only the first analysis that was performed on the database and obviously much more could be revealed on the evolution of the connectome and specific brain regions. Characterizing the structure of the observed brain networks will allow tracking both the impact of the environment on brain connectivity, as well as its evolutionary development.