Tissue-specific landscape of protein aggregation and quality control in an aging vertebrate

Abstract

Protein aggregation is a hallmark of age-related neurodegeneration. Yet, aggregation during normal aging and in tissues other than the brain is poorly understood. Here, we leverage the African turquoise killifish to systematically profile protein aggregates in seven tissues of an aging vertebrate. Age-dependent aggregation is strikingly tissue specific and not simply driven by protein expression differences. Experimental interrogation in killifish and yeast, combined with machine learning, indicates that this specificity is linked to protein-autonomous biophysical features and tissue-selective alterations in protein quality control. Co-aggregation of protein quality control machinery during aging may further reduce proteostasis capacity, exacerbating aggregate burden. A segmental progeria model with accelerated aging in specific tissues exhibits selectively increased aggregation in these same tissues. Intriguingly, many age-related protein aggregates arise in wild-type proteins that, when mutated, drive human diseases. Our data chart a comprehensive landscape of protein aggregation during vertebrate aging and identify strong, tissue-specific associations with dysfunction and disease.

Keywords: aggregates; aging; multi-tissue; protein homeostasis; proteomics; systems biology.

Figure 1.. Quantitative proteomics of tissue lysate and aggregate fractions in young, old, and old telomerase mutant killifish.

(A-B) Experimental design. Seven tissues from 3 young, 3 old, and 3 old TERT^Δ8/Δ8 male killifish were fractionated into tissue lysate (TL) and a high molecular weight subfraction enriched in protein aggregates (AGG) before analysis with quantitative TMT mass spectrometry. (C) Silver stained SDS-PAGE of unboiled samples from young, old, and old TERT^Δ8/Δ8 killifish brain and liver. The brain image is also reported in (Harel, et al., 2023). (D) Reproducibility of aggregate abundance between individuals by tissue (log2 transformed normalized peptide spectra counts for all observed proteins). r, Pearson’s correlation coefficient; a.u., arbitrary units. (E) Principal component analysis (PCA) of the different samples, with each symbol representing an individual fish. The brain PCA was also reported in the accompanying paper (Harel, et al. 2023).

Figure 2.. Tissue-specific changes in tissue lysate, aggregates, and aggregation propensity with age.

(A) Heatmap of proteins with increased abundance in TL, AGG, and aggregation propensity (PROP) in old vs. young animals (log2-transformed fold change (log2FC) > 0, two-sided Student’s t-test p-value < 0.05). The log2FC represents the average among 3 fish. (B) Percentage of significant shared vs. tissue-specific increases across the entire dataset. Statistics as in A. P-values, Chi-squared test of independence. (C) Example proteins with tissue-specific aggregation increases in old fish. Here and in all subsequent box plots, the box shows the quartiles of the data while the whiskers show the range of the data, here protein abundance in young (Y) or old (O) samples. P-values, Student’s t-test from three animals per group; n.s., not significant; a.u., arbitrary units. (D) Gene Set Enrichment Analysis (GSEA) of functional and pathway enrichments in age-related aggregates. Strongest normalized enrichment scores (NES) are shown; full list in Table S6. The heatmap is colored by rank statistics (the product of -log10(p-value) and NES). (E) Subcellular localization of proteins with significant aggregate accumulation in old tissues. Two examples are highlighted, right. Arc lengths are proportional to the fraction of age-related aggregates within that compartment that are derived from each tissue. The average value from all tissues is in the center of each doughnut. Localization data from UniProt (orthologous human proteins); proteins in multiple compartments were double counted; see also Table S7.

Figure 3.. Features of aggregation-prone proteins across aged killifish tissues.

(A) Schematic representation of Monte Carlo simulations to identify sequence feature enrichments and assign statistical significance. (B) Biophysical features most strongly associated with protein aggregate burden and aggregation propensity in young and old animals. Thresholds and statistics as in previous figures. Full results in Table S8A-D. Tissue-specific features surrounded with dotted boxes expanded in C. (C) Distinct biophysical properties in proteins that aggregate with age. Prion-like domains (predicted by PLAAC), local charge (net charge per residue), and hydrophobicity (Kyte-Doolittle scale), each calculated with localCIDER (sliding window of 5 residues).

Figure 4.. Validation of protein aggregation and identification of contributors to protein aggregation in vivo.

(A) S. cerevisiae aggregation assay for eGFP fusions to killifish proteins (1) and criteria used to score aggregation (2). (B) Percentage of yeast cells with GFP puncta upon overexpression of indicated killifish proteins in S. cerevisiae. Proteins with a significant age-associated aggregation propensity were selected from different tissues. A 10% cutoff was used to define diffuse versus punctate proteins. Two independent experiments were performed for each protein (~120 eGFP-positive cells quantified in each). Full results in Table S9. (C) Representative images of example proteins. (D) Mean net charge vs. hydropathy poorly distinguishes aggregation in yeast relative to random chance (54% vs. 50%). (E) A support vector machine classifier trained on aggregation status in old killifish and the charge patterning parameter delta better distinguishes killifish proteins that aggregate in yeast relative to random chance (81% vs 50%). Right detail, two proteins with different delta values and aggregation behaviors. (F) Distribution of delta values from all aggregates detected and a model that predicts aggregation behavior in vivo based on charge patterning ‘delta’ and aggregation propensity during aging.

Figure 5.. Tissue-specific changes in protein quality control machinery in tissue lysate and aggregate fractions during aging.

(A) Age-related changes in protein quality control pathways in liver and muscle TL. Color reflects the log2-transformed average age-dependent change. (B) Chaperone abundance changes in AGG between old and young animals. Color indicates ranked statistics; size scaled by the -log10-transformed Student’s t-test p-value. (C) Age related AGG abundance changes of TRiC and proteasome components as well as their clients in liver. Circle color indicates ranked statistics; light gray, not detected. (D) AGG and PROP changes of TRiC clients actin and tubulin in aged liver. P-values, Student’s t-test. (E) Age-related changes in levels of chaperone-mediated autophagy (CMA) factors and their targets in brain TL. Color indicates ranked statistics; light gray, not detected. (F) Model of a vicious cycle that may amplify protein aggregation during aging.

Figure 6.. Analysis of changes in tissue lysate and high molecular weight aggregate fraction of old TERT^Δ8/Δ8 mutant fish.

(A) Heatmap of proteins with increased abundance in TL, AGG, and PROP in old TERT^Δ8/Δ8 vs. old wild-type animals (log2FC > 0, two-sided Student’s t-test p-value < 0.05, average of three individuals). See also Table S13. (B) Significant shared vs. tissue-specific increases across the entire dataset. Statistics as in A. P-values, Chi-squared test of independence. See also Table S13D. (C) Cell cycle progression in adult male transgenic animals carrying a dual FUCCI reporter. Data shown are averages from FACS analysis of testis from three individuals. See also Table S16. (D) Percentage of proteins with PROP changes (both up-regulation and down-regulation) between age-matched old TERT^Δ8/Δ8 and wild-type animals. All significant z-scores were counted for each protein at the tissue level; sector radius reflects the percentage of the proteome they represent for each tissue. Z-score magnitude denoted by color intensity. (E) Tissue-specific correlations between cells in G1/G0 phase estimated based on the FUCCI reporter and extent of aggregate remodeling. The extent of aggregate remodeling is quantified as the fraction of proteins with differing aggregation properties between old TERT^Δ8/Δ8 vs. old wild-type (WT) individuals. See also Table S16. (F) Scatter plot of AGG abundance in old TERT^Δ8/Δ8 mutants and age-matched old wild-type. Red, aggregate proteins significantly up-regulated in old vs. young (p-value <0.05, log2FC>0) and old TERT^Δ8/Δ8 mutants compared to age-matched wild-type (p-value <0.05, log2FC>0). (G) Box plot of Lamin A protein levels in skin from age-matched TERT^Δ8/Δ8 and wild-type fish. Each dot represents the AGG abundance from an individual. P-values, Student’s t-test; n.s., not significant. (H) Illustration of the telomerase complex (left) and bar plot of normalized dyskerin (DKC1) level in the aggregate fraction in age-matched old TERT^Δ8/Δ8 and wild-type killifish (right). Bar plot represents average and standard deviation. See also Table S17.

Figure 7.. Disease-associated proteins with increased aggregate or aggregation propensity during physiological aging.

(A-B) Cartoon showing the functions of proteins with an age-associated increase in AGG and PROP identified in heart and striated muscle. Each dot represents values for an individual. (C) Proteins with an age-associated increase in AGG and PROP that are also linked to Mendelian diseases. Protein and disease associations are obtained from Online Mendelian Inheritance in Man (OMIM) database. See also Table S17.