Whole brain generative model identifies neurotransmitter alterations underlying Alzheimer's disease progression

Background Alzheimer’s disease (AD) involves heterogeneous aberrations in multiple biological processes, although their causal molecular mechanisms are not fully understood. As the primary neuronal signalling molecules, neurotransmitters regulate a variety of biological processes pathological in AD, such as the coupling between neural activity and vascular response. However, due to the infeasibility of mapping most neurotransmitters in‐vivo, alterations during disease progression are not well characterized. We propose multi‐modal, subject‐specific models of neuroimaging alterations as functions of local neurotransmitter concentrations, which we combine with clinical data to identify specific neurotransmitters prominently altered during AD progression. Method (1) We used the methods described in [Iturria‐Medina et al., Neuroimage, 152:60–77, 2017] to pre‐process longitudinal neuroimaging data (amyloid β and tau protein distributions, cerebral blood flow, r‐fMRI, glucose metabolism, and grey matter density) for 141 healthy and 302 diseased subjects from ADNI. (2) Using the densities of 15 neurotransmitter receptors from multiple cortical areas [Palomero‐Gallagher & Zilles, Neuroimage, 197:716–741, 2019] and 4 serotonin‐type PET templates [Lanzenberger et al., Biological psychiatry, 61(9):1081–1089, 2007], we built subject‐specific, generative models of changes in neuroimaging in terms of (i) other neuroimaging modalities, (ii) network propagation, and (iii) first‐order receptor‐moderated interactions. (3) By evaluating the monotonicity of receptor‐associated coefficients with aging‐related cognitive measures (memory performance, executive function, MMSE, CDR, and ADAS), we identified receptors with a significant causal role in AD progression. Result Notably, the model explains 50%‐77% of observed variance in all imaging bio‐markers across clinical groups (HC, EMCI, LMCI, AD). Significant receptor‐imaging interactions with AD progression were identified for (i) NMDA and cholinergic receptors for amyloid distribution, (ii) nicotinic acetylcholine and serotonergic receptors for tau distribution, (iii) glutamatergic NMDA and D1 dopaminergic receptors for blood flow, (iv) GABAergic, cholinergic, dopaminergic and serotonergic receptors for neural activity, (v) GABA receptors for glucose metabolism, and (vi) GABAergic and serotonergic receptors for gray matter density. Conclusion A generative, multi‐modal neuroimaging model including 19 receptors allowed, for the f