The transcriptional landscape of Alzheimer’s and Parkinson’s diseases

Alzheimer’s disease (AD) and Parkinson’s disease (PD) are the two most common neurodegenerative disorders worldwide. Although the aetiology, affected brain region and clinical features are particular to each of these diseases, they nevertheless share common mechanisms such as mitochondria dysfunction, neuronal loss and tau protein accumulation. The major risk factor for those disorders is ageing, the age of onset of both AD or PD being around 65 years old. Together, they account for 50 million cases worldwide, a number expected to increase due to the fact that the world population is living longer than ever. Most of AD and PD cases are sporadic and, despite all the research during the last centuries to better understand their molecular nature, current treatments are still symptomatic. Therefore, the development of effective therapies requires a better comprehension of the diseases’ aetiology and underlying mechanisms as well as finding disease-specific targets for drug discovery. A common strategy to identify biological pathways and cellular processes altered in neurodegenerative disorders is to compare gene expression profiles between age-matched diseased and non-diseased post-mortem brain tissues. However, the expression profiles derived from whole brain tissue mRNA highly reflect alterations in cellular composition, namely the well-known AD- or PD-associated loss of neurons, but not necessarily the disease-related molecular changes in brain cells. The advent of single-cell transcriptomes has made it possible to tackle this limitation, enabling the determination of reference gene expression profiles for each major brain cell type (namely neurons, astrocytes, microglia and oligodendrocytes) that can then be used to computationally estimate the cell type-specific content of bulk brain sample’s in healthy and diseased conditions, decoupling the neurodegeneration effect (i.e. the relative loss of neurons) from the intrinsic systemic or cell type-specific disease effects. This approach has already been applied in determining the effects of age and psychiatric disorders on the cellular composition of human brain, or the contribution of each cell type in shaping the pathological autism transcriptome. The same principle was applied in AD by modelling the expression of its risk genes as a function of estimated cellular composition of brain samples. For instance, APP, PSEN1, APOE and TREM2 had their expression levels associated with the relative abundance of respectively neurons, oligodendrocytes, astrocytes and microglia. Additionally, two recent studies profiled single nuclei of major brain cell types in AD and non-AD post-mortem brain samples, unveiling cell type-specific transcriptional changes. All these studies highlight the importance of charactering disease-associated cell type-specific phenotypes that can not only unveil the cellular and molecular bases of pathological mechanisms but also be therapeutically targeted. However, some of these studies still lack independent validation and have not fully dissected the nature of transcriptomic alterations in AD brains. Moreover, to our knowledge, similar approaches have not yet been applied to PD, despite increasing evidence regarding the importance of modelling cellular composition in neurodegenerative disorders. We therefore used scRNA-seq data to derive gene expression signatures for the major human brain cell types and estimate the cellular composition of idiopathic AD and PD post-mortem brain samples from their bulk transcriptomes, investigating whether neuronal loss could be confounding or masking the intrinsic disease effects on gene expression, and validating the results in independent datasets. Additionally, since AD and PD might share the same mechanisms of disease progression, we also investigated the similarities between the transcriptomic alterations induced by AD and PD in human brain tissues. This approach allowed the novel identification of genes and pathways whose activity in the brain is intrinsically altered by AD and PD in systemic and cell type-specific ways. Additionally, we pinpoint the genes that are commonly altered by these major neurodegenerative disorders as well as those specifically perturbed in each illness. Moreover, using chemical perturbagen data, we computationally identified candidate small molecules for specifically targeting the profiled AD/PD-associated molecular alterations. Thus, we unveil a set of novel candidates that can potentially be targeted in AD and PD therapeutics. Moreover, we herein demonstrate the potential of modelling cellular composition in transcriptomics analyses in the discovery of therapeutic targets for other neurodegenerative diseases