In vivo validation of late-onset Alzheimer's disease genetic risk factors

Abstract

Introduction: Genome-wide association studies have identified over 70 genetic loci associated with late-onset Alzheimer's disease (LOAD), but few candidate polymorphisms have been functionally assessed for disease relevance and mechanism of action.

Methods: Candidate genetic risk variants were informatically prioritized and individually engineered into a LOAD-sensitized mouse model that carries the AD risk variants APOE ε4/ε4 and Trem2*R47H. The potential disease relevance of each model was assessed by comparing brain transcriptomes measured with the Nanostring Mouse AD Panel at 4 and 12 months of age with human study cohorts.

Results: We created new models for 11 coding and loss-of-function risk variants. Transcriptomic effects from multiple genetic variants recapitulated a variety of human gene expression patterns observed in LOAD study cohorts. Specific models matched to emerging molecular LOAD subtypes.

Discussion: These results provide an initial functionalization of 11 candidate risk variants and identify potential preclinical models for testing targeted therapeutics.

Highlights: A novel approach to validate genetic risk factors for late-onset AD (LOAD) is presented. LOAD risk variants were knocked in to conserved mouse loci. Variant effects were assayed by transcriptional analysis. Risk variants in Abca7, Mthfr, Plcg2, and Sorl1 loci modeled molecular signatures of clinical disease. This approach should generate more translationally relevant animal models.

Keywords: APOE4; Abca7; Alzheimer's disease; Plcg2. Mthfr; Sorl1; Trem2; animal models; preclinical; transcriptomic analysis.

FIGURE 1

Strategy to prioritize loci and LOAD risk variants. Summary of strategies for variant selection for (A) LOAD and (B) neurovascular risk factors. (C) Gene expression analysis comparing human and mouse gene expression data to identify human LOAD modules that are altered by genetically engineered variants in mice.

FIGURE 2

Correlation between LOAD‐associated risk variants and 30 human AMP‐AD brain co‐expression modules using the NanoString Mouse AD panel. (A) Correlation between the effect of each mouse perturbation relative to the LOAD1 background in 4‐month‐old mice and 30 human co‐expression modules, also including the early‐onset transgenic model 5XFAD and the LOAD1 background relative to C57BL/6J. The 30 human co‐expression modules were grouped into five consensus clusters with similar gene content across the multiple studies and brain regions. Framed circles correspond to significant (p < 0.05) positive (blue) and negative (red) Pearson's correlation coefficients, with size and color intensity proportional to the correlation. The effects of multiple LOAD risk variants in mice were positively correlated (p < 0.05) with cell cycle and myelination‐associated modules in Consensus Cluster D and cellular stress‐response‐associated modules in Consensus Cluster E. (B) Correlation between effect of each mouse perturbation at 12 months and the 30 human co‐expression modules. LOAD risk variants showed significant correlation with functionally distinct AMP‐AD co‐expression modules. The effects of Abca7*A1527G, Shc2*V433M, Ceacam1 KO, and Slc6a17*P61P in aged mice correlated with the immune modules in Consensus Cluster B, while the effects of Sorl1*A528T and Plcg2*M28L correlated with the neuronal modules in Consensus Cluster C.

FIGURE 3

Correlation between effect of genetic variants and Gene Set Enrichment Analysis (GSEA). (A) Correlation between regression coefficients calculated for each genetic variant at 4 months. Color intensity and size of circles are proportional to Pearson correlation coefficient, with insignificant correlations (p > 0.05) left blank. (B) Correlation between regression coefficients calculated for each genetic variant at 12 months. The effects of Snx1*D465N, Plcg2*M28L, and Mtmr4*V297G risk variants in mice showed significantly positively correlation (p < 0.05) at 12 months. (C) GSEA results of selected AD‐associated pathways from Reactome library in presence of each LOAD risk variant in mice. Enriched pathways are grouped by their overlap with functional annotations of human AMP‐AD Consensus Clusters. Immune‐related pathways had increased expression in the presence of multiple risk variants such as Abca7*A1527G, Mthfr*677C > T, and Snx1*D465N, while neuron‐associated pathways had reduced expression in the presence of risk variants such as Abca7*A1527G, Mthfr*677C > T, Sorl1*A528T, Plcg2*M28L, Ceacam1 KO, Shc2*V433M, and Slc6a17*P161P.

FIGURE 4

Identification of specific AD‐associated processes in LOAD risk variants exhibiting transcriptomic changes similar to human LOAD in age‐dependent manner. For four new mouse strains the following are displayed: the six top enriched GO terms identified by GSEA and GO enrichment analysis of genes with common directional changes with human AD modules (top left); gene module networks with common directional changes with human AMP‐AD modules, where node colors correspond to human AMP‐AD Consensus Clusters A (orange), B (green), C (blue), D (turquoise), or E (pink) (top right); and the effects of each variant at multiple ages correlated across LOAD effects in 30 AMP‐AD modules, following the legend of Figure 3. Results for (A) Abca7*A1527G model, (B) Mthfr*677C > T model, (C) Plcg2*M28L model, and (D) Sorl1*A528T model. All results are relative to LOAD1 genetic background for all strains.

FIGURE 5

Correlation between effect of each mouse perturbation and molecular subtypes of LOAD. Two molecular LOAD subtypes inferred in ROSMAP cohort, three subtypes in Mayo cohort, and two subtypes in Mount Sinai Brain Bank (MSBB) cohort. Framed circles correspond to significant (p < 0.05) positive (blue) and negative (red) Pearson's correlation coefficients across all genes on the NanoString panel, with color intensity and circle size proportional to the correlation. (B) At 4 months, the Abca7*A1527G and Sorl1*A528T variants represent inflammatory subtypes of LOAD (Subtypes A) in each cohort, while Shc2*V433M and Clasp2*L163P variants mimic the non‐inflammatory subtypes of LOAD (Subtypes B). (C) At 12 months, the Abca7*A1527G and Ceacam1 KO variants recapitulate inflammatory subtypes of LOAD (Subtypes A), while the Snx1*D465N, Mtmr4*V297G, and LOAD1 variants model non‐inflammatory subtypes of LOAD (Subtypes B).