Premium
The effect of microglial genes on network diffusion of pathology in mouse models of tauopathy
Author(s) -
Anand Chaitali,
Maia Pedro D.,
Torok Justin,
Mezias Christopher,
Raj Ashish
Publication year - 2021
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.052382
Subject(s) - microglia , neuropathology , tauopathy , trem2 , biology , neuroscience , connectome , chronic traumatic encephalopathy , neurodegeneration , pathology , microbiology and biotechnology , disease , inflammation , medicine , immunology , concussion , functional connectivity , poison control , environmental health , injury prevention
Background Misfolded tau and amyloid, the hallmark pathologies of Alzheimer’s disease, undergo “prion‐like” trans‐synaptic transmission along the brain’s anatomic connectivity network (Frost, et al ,2010). Our lab has previously shown that Alzheimer’s neuropathology progression is well‐approximated by a network diffusion model (NDM) (Raj, et al ,2012). However, other processes in the brain may affect the neuropathology progression by mediating intraregional clearance/amplification and interregional spread. For instance, microglial activation is a common pathological feature of Alzheimer’s disease, with TREM2‐expressing microglia being integral to the inflammatory response. We show here that the NDM can be extended to include effects of microglial mediation of pathology. Method We modified the NDM to an extended NDM (eNDM) with additional model parameters related to microglia‐mediated spread/clearance and applied it to the 426×426 region mouse connectome. The eNDM included datasets of longitudinal spatiotemporal tau pathology in PS19 mice, expression levels of five microglial homeostatic signature genes ( Cx3Cr1 , C1qa , P2ry12 , Fcrls , Olfml3 ) as well as TREM2 and CD33 , microglial surface‐markers and phagocytosis regulators (Li, et al ,2018). We compared the fits of the NDM to those of the eNDM using each gene individually as well as the first two principal components of spatial gene expression (Lein, et al ,2007) after optimizing model parameters using a constrained nonlinear optimization algorithm. Results Inclusion of microglial gene expression improved the predictive power of the model beyond that of the NDM (R 2 =0.46). Using the principal components of gene expression generally provided better model fits than individual genes. Notably, exclusion of TREM2 before performing PCA had a significant effect on model fit ( p <0.001, R 2 =0.5) (Table 1, Figure 1), increasing the values of the optimal diffusion and spread parameters. Conclusion Although network spread models capture well the spread of pathology between brain regions, there are other factors that mediate this process. We demonstrated that an eNDM with microglial genes had an improved model fit even after accounting for differences in model parsimony. We also showed that exclusion of TREM2 from the model resulted in higher diffusivity parameter values, thus supporting the biological observation of TREM2 ‐knockout resulting in increased spread of pathology, putatively by affecting microglial phagocytic potential (Griciuc, et al , 2019).