Across three injection sites, the mouse connectome predicted the seed-to-seed difference in alpha-synuclein pathology, but regional vulnerability explained more of each pattern

Models of misfolded protein spread fit linear diffusion on the Allen mouse connectome to the pathology caused by one injection site. A fit of this kind cannot separate axonal routing from regional vulnerability, which is the same for every injection site. We used published group means for alpha-synuclein fibrils injected into caudoputamen, nucleus accumbens or hippocampal CA1 of mice (213 Allen regions per hemisphere, 1 to 24 months). Fitting one site at a time, a vulnerability predictor made only from the other two sites gave a mean Spearman rho of 0.51, above the anterograde (0.37) and retrograde (0.31) diffusion models and distance from the injection (0.11). Predictions made from the wrong injection site still fitted; the gain from the correct site was 0.21 rho. We then compared two injection sites, which cancels vulnerability. The retrograde contrast gave rho 0.33 (0.31 after the distance contrast) against a null of 0.10 for 200 rewired connectomes that kept out-strength, target number and the weight-distance relation (p < 0.005). On the regions above the zero floor at both sites, the same contrast gave rho 0.50. Weights fitted on two sites and applied to the third reduced held-out error by 8.4% beyond vulnerability and distance, for every held-out site, against a null range of -2.4 to 1.9%. The result held over a thousandfold range of the diffusion rate, for somatic and neuritic pathology, and without the isocortex. Within one region, the injected side had more pathology in 94% of regions, and the size of the left-right difference followed the connectome (partial rho 0.17). Independent tau data from a hippocampal injection repeated the left-right result (partial rho 0.25). A split-half ceiling from those per-mouse data was 0.80 to 0.98, so most seed-specific structure stays unexplained. The connectome carries real routing information, but a fit to one injection site measures mostly vulnerability. Group means, a wild-type regional connectome and an ordinal pathology scale limit the test.

alpha-synuclein · tau · mouse connectome · pathology spread · network diffusion · selective vulnerability · injection site · Allen Mouse Brain Connectivity Atlas · structure and function · null models