Ubiquitin-Proteasome System Dysregulation in Alzheimer's Disease Impacts Protein Abundance

Abstract

Alzheimer's disease (AD) is a relentlessly progressive, fatal neurodegenerative disorder associated with widespread aberrant proteomic changes. The full extent of protein dysfunctions in AD and their impact on cellular physiology remains unknown. Here, we used plexDIA, an approach that parallelizes the acquisition of samples and peptides, to characterize proteomic changes in AD. Using human dorsolateral prefrontal cortex tissue, we identified 281 differentially abundant proteins in AD. By systematically analyzing cellular compartment-specific shifts in protein abundance, we identified an AD-specific decrease in levels of the 20S proteasome, the catalytic core of the cell's primary protein degradation pathway. This alteration was accompanied by widespread decreases in proteasome subunit stoichiometries. Many proteasome substrate proteins were negatively correlated with 20S levels and increased in AD, suggesting that reduced 20S levels leads to abnormal protein accumulation. By analyzing proteins increased in AD, we identify key properties of such proteins. Namely, they have highly specific subcellular localizations and fast degradation rates, they contain signal sequences that allow them to be targeted for proteasomal degradation, and they are targeted by quality control pathways that recognize mislocalized proteins. Furthermore, we identify coherent sets of ubiquitin system enzymes, proteins that target substrates for proteasomal degradation, whose levels robustly discriminate AD from non-AD samples. One subset exhibited consistent increases in AD, while another exhibited consistent decreases, revealing complex alterations to the ubiquitin system in AD. Taken together, our results suggest that decreased ubiquitin-proteasome system capacity and impaired clearance of short-lived and mislocalized proteins contribute substantially to proteopathic burden in AD.

Figure 1:

Study overview and AD proteomic changes. A. Schematic of the cohort and proteomic profiling approach. B. Volcano plot of proteins exhibiting differential abundance (q value less than 0.05 and an absolute fold change greater than 1.5) between AD and non-AD samples. C. Compartment-specific shifts were scaled and plotted to visualize the largest shifts between AD and non-AD, revealing that the shift in proteasome 20S subunit proteins was the largest negative shift between AD and non-AD samples. D. / E. Rank order plots showing relative fold change ranks of 20S (D.) and 19S (E.) proteasome subunits between AD and non-AD samples.

Figure 2:

Proteasome subunit stoichiometries in non-AD and AD samples. Correlations between the abundance of individual proteasome subunits were visualized as a heatmap for non-AD and AD samples. A. / B. non-AD samples (A.) showed significantly higher correlations among components of the 19S regulatory particle and 20S core particle as compared to AD samples (B.).

Figure 3:

Proteome-wide correlations to 20S levels. A. Proteome-wide correlations to 20S proteasome core particle subunit levels. B. Proteome-wide correlations to 19S proteasome regulatory particle levels. C. – E. Levels of all proteins were correlated to 20S subunit median abundance and protein set enrichment analysis was run on the resulting set of correlations. Plots display significantly enriched (q ≤ 0.05) Gene Ontology Biological Process (C.), Molecular Function (D.), and Cellular Compartment (E.) terms. Black lines show individual proteins with the correlation magnitude displayed on the y axis. Red lines show the median for all proteins mapping to the indicated term. The full set of 20S correlations is shown for reference at the bottom of each plot in C - E. F. Proteins from the “ATP dependent protein folding chaperone” GO term were stratified by chaperone type and each protein’s correlation to 20S levels was visualized. As shown, proteins of the chaperonin containing TCP-1 (CCT) complex were significantly more negatively correlated to 20S levels than HSP chaperones. G. Proteins from the GO “Nucleocytoplasmic carrier activity” term were stratified by their direction of transport. There were no significant differences in 20S correlations between transporter types.

Figure 4:

AD-associated changes in 20S substrates. A. Volcano plot of log₂ fold change versus q for 20S substrates. 20S substrates are more often increased in AD. B. 20S correlation and log₂ fold change are strongly negatively correlated, both among 20S substrates and all detected mitochondrial ribosome proteins. C. Two-step pathway of mitochondrial ribosome assembly. Based on previously-published results. D.-F. Correlation of rate constants measuring pre-assembly degradation rate (k_a, D.), complex assembly (k_ab, E.), and post-assembly degradation (k_b, F.). Mitochondrial ribosomal protein abundance is significantly correlated with both degradation rates, but not complex assembly. Note that the pre-assembly degradation rates are much higher than post-assembly (compare y-axis in D. / E.)

Figure 5:

Properties of differentially abundant proteins between AD and non-AD samples. A. The log₂ fold change versus degradation rate was visualized for all differentially abundant proteins stratified by the fold change direction. B. Binning proteins by fold change revealed a subset of proteins that had large increases in AD and high degradation rates (rightmost box). C. Proteins of increased versus decreased abundance do not differ in the fraction of N-terminal disordered residues. Proteins decreased in AD have a significantly higher fraction of disordered C-terminal residues. Termini were defined as the first (N-terminus) or last (C-terminus) 100 amino acids of a protein. D. Proteins of increased versus decreased abundance in AD have similar lengths. E. Proteins of increased abundance have significantly more degron-containing peptide sequences than those decreased in AD. F. The log₂ ratio of the number of degron motifs in proteins of increased vs. decreased in abundance was stratified by E3 ligase. The largest and most significant difference was the enrichment of Bag6 degrons in proteins increased in AD (top row). G. The structure of TMEM94, a protein significantly increased in AD samples, is shown with its cytosol-facing N-terminus and Bag6 degron recognition motifs highlighted. H. The degron strength of peptides from the TMEM94 N-terminus were Z-transformed such that more potent degrons have higher values. Data are from a prior study. I. Model depicting substrate recognition and processing the components of the Bag6 ubiquitin ligase complex. J. Correlations among the relative abundance of Bag6 complex subunits were plotted, revealing a significant decrease in subunit stoichiometry in AD samples.

Figure 6:

Ubiquitin system enzyme alterations in AD. A. Hierarchical clustering was used to identify highly correlated subsets of ubiquitin system enzymes. A total of 16 clusters, numbered from top left to bottom right were identified. B. To determine the statistical significance of individual clusters, we used bootstrap-based resampling to generate 1,000 random clusters. Clusters exceeding the 99th percentile of the resulting empirical null distribution were considered significant. C. The discriminative ability of each cluster was evaluated using two complementary methods, the scaled Euclidean distance between AD and non-AD samples (x axis) and the F statistic from an ANOVA of AD versus non-AD samples. By both metrics, clusters 3 and 13 were clear outliers. D. Normalized levels of all cluster 3 and 13 ubiquitin system enzymes are shown, with each cluster’s mean at right.