An integrated multi-omics approach to improve our understanding of ageing and neurodegeneration

Abstract

The gut-brain axis is a complex system that begins its developmental journey in utero and becomes functionally activated by the first colonising microbiome at birth. In this research project, we have developed a computational framework to independently analyse single-nucleus transcriptomics and shotgun metagenomics data types along with their integration at the pathway level. From our transcriptomics analysis on 84 healthy individuals across the human life span, we identified that the astrocytes in prefrontal cortex assist in brain development and switch their function from being neuroprotective to neurotoxic in older age, in response to neuroinflammation. We also identified that major proteostasis branches, namely autophagy-lysosome pathway and ubiquitin proteasome systems are differentially activated across age. This activation also shows sex dimorphism, that we also identified by differential expression of 26 X-chromosome associated genes across age. Our co-expression analysis suggested that many gene modules are preserved across age, except during the young age (20-35 years of age) and we consider this age group as a window of maximum variability. Our sex-specific co-expression networks highlighted that there is a high degree of preservation of module functions in both sexes for each age group, and thus it is important to identify further if there are global ageing differences between the two sexes. In parallel, our microbiome analysis on 1969 individuals from China, Germany and United States of America (USA) suggest that there are unique species-specific signatures associated with ageing and Parkinson’s disease. We also identified that these signatures were shared at the family level and at the functional level with shared signalling pathways. We identified 23 and 15 species to be uniquely associated with ageing and PD represented by 3766 strain variants with a diversity of functions. Comparing the performance of machine learning algorithms trained on the bacterial and virome gut composition, we observed bacteria to perform better as predictors of health. The integration of these two data types identified 46 pathways which were classified as either host-microbiome specific pathways or potentially host-microbiome interacting pathways. From these analyses, we propose that Bifidobacterium dentium could be one of the species interacting with astrocytes through synthesis of a neurotransmitter (GABA), while ferroptosis might represent another mechanism of host-microbiome interaction contributing to ageing and PD associated decline in cognition.