The Accelerating Medicines Partnership – Alzheimer’s Disease (AMP-AD) Consortium was formed to accelerate the development of diagnostics and therapeutics for this devastating and complex disease. One roadblock to development is the lack of good pre-clinical models, as efficacy in clinical trials often does not match the promising results seen during development. Recently, the AMP-AD Consortium published a manuscript describing the use of systems biology to investigate transcriptional changes in the human disease1. This work can enable the development of improved translational models by providing a transcriptional map of disease.
The basis for this bioinformatics work is 3 published, sizeable, and independent RNA-seq datasets of human post-mortem samples from 7 brain regions, a subset of which came from Alzheimer’s patients. All datasets were re-analyzed for standardization, and 5 separate gene co-expression module methods were employed separately on the data from each region. 30 modules were significantly enriched for differentially expressed genes in the Alzheimer’s samples and were considered further. These modules were also enriched for previously published AD-related genes and pathways, highlighting the relevance of these genes. The 30 modules covering all 7 brain regions were compared to each other and found to overlap significantly in gene content, indicating that similar genes are co- and differentially expressed across brain regions in AD. Using this overlap, the 30 modules can be summarized into 5 “consensus clusters”, termed clusters A-E, that are consistent across the various studies and brain regions.
All but one of the consensus clusters were highly associated with specific cell type functions, even if they were not primarily composed of cell type specific genes. The final cluster (cluster E) was most associated with proteostasis, a clearly dysregulated aspect of AD pathology. When assessing the sex-specific differences in these modules, the 4 cell type-associated modules were more differentially expressed in females than males. By contrast, cluster E was more strongly impacted in males than females. In comparing the modules generated by this meta-analysis to previously published single study modules, most had similarity to the meta-analysis. One outlier, module 109, contained genes that were distributed across meta-analysis clusters and was enriched for differentially expressed genes in both male and female AD samples. Finally, only 3 of the 5 clusters were enriched for current therapeutic targets in AD, suggesting that additional targets or pathways may be present and bear further investigation.
This gene expression analysis has important implications for therapeutic development and research into Alzheimer’s disease. Most importantly, the meta-analysis confirms and extends previous work to describe the disorder on a transcriptional level. The study was sufficiently powered to confirm strong sex-based differences in gene expression, which may indicate that different therapeutics may need to be employed, and different risk alleles may exist. These modules represent conserved changes across multiple patients, studies, and brain regions that define the transcriptional landscape of AD. One outcome is that new mouse models can now be developed to best recapitulate these transcriptional changes.
The process of creating of better mouse models for human AD is costly and time-consuming, particularly due to the need to age mice to develop Alzheimer’s-like pathology. Streamlining this process would benefit from tools to screen mice, such that the transcriptional changes can be detected quickly and reproducibly in large numbers of animals and ideally also in younger mice. The MODEL-AD consortium, funded by a grant from the NIH including Indiana University, The Jackson Laboratory, and Sage Bionetworks, and NanoString have developed the nCounter® Mouse AD gene expression panel for evaluation of mouse models of AD with analysis of 770 genes selected based on the 30 clinically relevant gene co-expression modules discovered in the AMP-AD consortium study of human brain tissue described1. Data generated on a pilot version of the panel has already been shown to correlate well with human RNA-seq results, demonstrating the panel’s utility in comparing the known modules across various cohorts of AD mice.
- Logsdon BA, Perumal T, et al. Heterogeneity across human AD coexpression modules identified by meta-analysis of the human brain transcriptome. bioRxiv, 2019. 10.1101/510420.
FOR RESEARCH USE ONLY. Not for use in diagnostic procedures.