Single-cell transcriptomic and proteomic analysis of Parkinson's disease brains

Abstract

Parkinson's disease (PD) is a prevalent neurodegenerative disorder, and recent evidence suggests that pathogenesis may be in part mediated by inflammatory processes, the molecular and cellular architectures of which are largely unknown. To identify and characterize selectively vulnerable brain cell populations in PD, we performed single-nucleus transcriptomics and unbiased proteomics to profile the prefrontal cortex from postmortem human brains of six individuals with late-stage PD and six age-matched controls. Analysis of nearly 80,000 nuclei led to the identification of eight major brain cell types, including elevated brain-resident T cells in PD, each with distinct transcriptional changes in agreement with the known genetics of PD. By analyzing Lewy body pathology in the same postmortem brain tissues, we found that α-synuclein pathology was inversely correlated with chaperone expression in excitatory neurons. Examining cell-cell interactions, we found a selective abatement of neuron-astrocyte interactions and enhanced neuroinflammation. Proteomic analyses of the same brains identified synaptic proteins in the prefrontal cortex that were preferentially down-regulated in PD. By comparing this single-cell PD dataset with a published analysis of similar brain regions in Alzheimer's disease (AD), we found no common differentially expressed genes in neurons but identified many shared differentially expressed genes in glial cells, suggesting that the disease etiologies, especially in the context of neuronal vulnerability, in PD and AD are likely distinct.

Fig. 1.. Single nucleus transcriptomic analysis of human brain prefrontal cortex reveals cell-type-specific changes in PD.

(A to C), UMAP plotting of human brain nuclei (n = 77,384 nuclei), colored by (A) major human brain cell types: excitatory (ExN) and inhibitory (InN) neurons, oligodendrocytes (Oligo), astrocytes (Astro), microglia (MG), oligodendrocyte precursor cells (OPC), endothelial cells (Endo) and brain-resident T cells (Tcells); (B) disease diagnosis of either Parkinson’s disease (PD) or healthy controls (HC); or (C) individuals (from 12 individuals of 6 PD and 6 HC). (D) Heatmap of top marker genes in each brain cell type. (E) Brain cell type proportion grouped by disease diagnosis of PD or HC. (F) Differentially expressed genes (DEGs) counts for each brain cell type (log fold change > 0.75 for T cells; log fold change > 0.25, FDR corrected P-value adjusted < 0.05 for all the other cell types). (G and H) Gene ontology (GO) pathway analysis of differentially expressed genes between PD and HC for each brain cell type for either (G) up-regulated pathways or (H) down-regulated pathways (H), with normalized z score for each pathway.

Fig. 2.. Correlation analysis of Lewy body pathology to chaperone gene expression in brains of patients with PD.

(A) Immunohistochemistry of pS129 α-synuclein (EP1536Y) in formalin fixed prefrontal cortex cortical sections detected α-synuclein-positive Lewy bodies and Lewy neurites. All scale bars represent 50 μm, except in the right-most panel scale bar = 25 μm. (B) Bar graph of Lewy body plus Lewy neurite score quantification across conditions (n = 12 with 6 PD and 6 HC, unpaired t-test P = 0.03). (C) Dot plots showing the differential expression of genes related the unfolded protein misfolding pathway in excitatory neurons (ExN) and inhibitory neurons (InN) of PD versus healthy controls (HC). (D) Dot plot of differential gene expression patterns of ExN versus InN in PD patients (log_fc: log fold changes). (E) Correlation between gene expression and pathology scores for each gene of the protein misfolding pathway. Spearman’s correlation coefficient (R²) was calculated based on normalized gene expression versus Lewy pathology score.

Fig. 3.. Cell-type-specific regulation of PARK genes and GWAS candidate genes in brains of patients with PD.

(A) Dot plots showing the expression of known PARK genes and GBA in human PD brains compared to healthy controls (HC). (B) Violin plots showing cell-type-specific differential expression of SNCA, LRRK2, PRNK, PINK1, GBA and PARK7 in major neuronal and glial brain cell types in human brains PD versus HC. (C) Heatmap for cell-type-specific expression of UTMOST identified most significant genes in PD brains compared to healthy controls, with normalized average gene expression value for each gene across cells for a specific cell type. (D) Heatmap for cell-type-specific differential expression of UTMOST prefrontal cortex most significant genes with effect size greater than 0 in human PD brains compared to healthy controls, with normalized average gene expression value for each gene in a specific cell type. (E) Heatmap for cell-type-specific differential expression of UTMOST prefrontal cortex most significant genes with effect size less than 0 in human PD brains compared to healthy controls, with normalized average gene expression value for each gene in a specific cell type.

Fig. 4.. Cell-cell communications between brain cell types in brains of patients with PD.

(A) Matrix showing the difference in cell-cell interaction pair numbers between PD (n = 6 individuals) versus healthy control (HC, n = 6 individuals) brains. (B) Dot blot showing the cell-cell interaction between excitatory neurons (ExN) and astrocytes (Astro) in PD and HC brains. (C) Dot plot showing the strength of cell-cell interactions between brain-resident T cells (Tcells) and excitatory neurons (ExN) in PD and HC brains (different font colors indicate - red: unique interactions in PD; blue: unique interactions in HC, black: common in HC and PD).

Fig. 5.. Paired proteomics and single cell transcriptomics of brains of patients with PD.

(A) Volcano plot showing significant differentially expressed proteins (n = 2,523 proteins, two-sided t test, P < 0.05) of PD (n = 6 individuals) versus health controls (HC, n = 6 individuals). (B) Heat map showing the upregulated (top) and downregulated (bottom) differentially expressed proteins (there were no significant differentially expressed proteins after Bonferroni or FDR correction). (C) Upregulated (top) and downregulated (bottom) gene ontology (GO) terms (n = 13 upregulated and 9 downregulated genes as input (P < 0.05), hypergeometric test, FDR correction). (D) Distinct protein modules (M1-M14) clustered by WGCNA across PD and HC. Each module’s label is the most significantly enriched GO term of that module proteins. (E) Eigen-proteins that are highly expressed in PD brains compared to HC in the most upregulated modules of M1 and M2. (F) Cell-type enrichment evaluated by module proteins against lists of RNA markers for different brain cell types from the single nucleus transcriptomic data using the one-tailed Fisher’s exact test. Blue line: P < 0.05. Red line: Adjusted P < 0.05, Bonferroni correction.

Fig. 6.. Comparison between PD and AD single cell transcriptomics and proteomics.

(A) The number of overlapping differentially expressed genes (DEGs) for each cell type between PD brain single nucleus transcriptomic data and Mathys et al’s AD brain single nucleus transcriptomic data (10) (n = 24 AD and 24 healthy controls). Over-representation test, (**): P ≤ 0.01, (***): P ≤ 0.001, (****): P ≤ 0.0001 (B) The number of overlapping (over-representation test, (****): P ≤ 0.0001) differentially expressed (DE) Gene ontology terms (hypergeometric test, FDR correction) for each cell type between PD brain single nucleus transcriptomic data and Mathys et al’s AD brain single nucleus transcriptomic data (10). (C) Top five overlapping differentially expressed (DE) pathways for astrocytes (Astro) and microglia (MG) between PD and AD. (D) The number of overlapping (over-representation test, (****): P ≤ 0.0001) differentially expressed (DE) proteins (P < 0.05) and pathways (genes with P < 0.1) between PD proteomics and Ping et al’s AD proteomics data (58) (there are n = 632 upregulated and 1222 downregulated genes for Ping et al’s data (58)). (E) Top five overlapping upregulated and downregulated differentially expressed (DE) protein pathways between PD and AD.