Selective vulnerability and resilience to Alzheimer's disease tauopathy as a function of genes and the connectome

Brain regions in Alzheimer's disease exhibit distinct vulnerability to its hallmark pathology with the entorhinal cortex and hippocampus succumbing early to tau tangles while others like the primary sensory cortices remain resilient. The quest to understand how local/regional genetic factors, pathogenesis and network-mediated pathology spread, together govern this selective vulnerability (SV) or resilience (SR) is ongoing. Although many Alzheimer's risk genes are known from gene association and transgenic studies, it is still unclear whether and how their baseline expression confers SV/SR to pathology. Prior analyses have yielded conflicting results, pointing to a disconnect between the location of genetic risk factors and downstream tau pathology. The spatial distribution of vulnerability doesn't always align with genetic factors, suggesting a role for non-cell-autonomous mechanisms like transneuronal tau transmission. We hypothesize that a full accounting of the role of genes in mediating SV/SR would require modelling of network-based vulnerability, whereby tau misfolds, aggregates and propagates along fibre projections. We employed an extended network diffusion model (eNDM) and fitted it on tau PET data from 196 patients from the Alzheimer's Disease Neuroimaging Initiative. The fitted eNDM then becomes a reference from which to assess the role of innate genetic factors. Using the residual (observed - model-predicted) tau as a novel target outcome, we obtained its association with 100 Alzheimer's risk genes, whose baseline spatial transcriptional profiles were obtained from the Allen Human Brain Atlas. Our eNDM was successful in capturing tau pathology distribution in patients. After regressing out the model, we found that while many risk genes have spatial expression patterns that correlate with regional tau, many others showed a stronger association with residual tau. This suggests that direct vulnerability aligned with the network, as well as network-independent vulnerability, are conferred by risk genes. We report four classes of risk genes: network-aligned SV (SV-NA), network-independent SV (SV-NI), network-aligned SR (SR-NA) and network-independent SR (SR-NI), each with a distinct spatial signature and associated vulnerability to tau. Remarkably, using gene ontology analysis, we found that the identified gene classes have distinct and sometimes surprising functional enrichment patterns. Network-aligned genes broadly participate in cell death, stress response and metabolic processing; network-independent genes in amyloid-β processing and immune response. These previously unreported segregated roles point to multiple distinct pathways by which risk genes confer vulnerability or resilience in Alzheimer's disease. Our findings offer new insights into vulnerability signatures in Alzheimer's disease and may prove helpful in identifying potential intervention targets.