Protein lifetimes in aged brains reveal a proteostatic adaptation linking physiological aging to neurodegeneration

Abstract

Aging is a prominent risk factor for neurodegenerative disorders (NDDs); however, the molecular mechanisms rendering the aged brain particularly susceptible to neurodegeneration remain unclear. Here, we aim to determine the link between physiological aging and NDDs by exploring protein turnover using metabolic labeling and quantitative pulse-SILAC proteomics. By comparing protein lifetimes between physiologically aged and young adult mice, we found that in aged brains protein lifetimes are increased by ~20% and that aging affects distinct pathways linked to NDDs. Specifically, a set of neuroprotective proteins are longer-lived in aged brains, while some mitochondrial proteins linked to neurodegeneration are shorter-lived. Strikingly, we observed a previously unknown alteration in proteostasis that correlates to parsimonious turnover of proteins with high biosynthetic costs, revealing an overall metabolic adaptation that preludes neurodegeneration. Our findings suggest that future therapeutic paradigms, aimed at addressing these metabolic adaptations, might be able to delay NDD onset.

Fig. 1.. Precisely measured protein lifetimes in the cortex and cerebellum of aged mice.

(A) Experimental workflow. Mice aged 21 months were metabolically labeled for 14 or 21 days, as previously described (19), and protein lifetimes were calculated for the brain cortex, cerebellum, and the respective synaptic fractions (18). (B) Venn diagram showing the LC-MS/MS protein identifications and their overlap in the four fractions. A total of 90,305 peptides (5354 protein groups) were identified in the homogenates, and 63,619 peptides (3947 protein groups) were identified in the synaptic fractions (see also fig. S1 and table S1). Numbers in parentheses show the number of proteins in each fraction. (C) Correlation matrix (Pearson’s r²) of calculated protein lifetimes from cortex and cerebellum, as well as respective synaptic fractions. (D) Lifetimes of 1571 proteins measured in cortex homogenate, subdivided in 36 categories, according to their organelle and/or functional affiliation [see (19) and table S1]. Each point corresponds to a single protein lifetime. Thicker lines indicate mean ± SEM for each category. The segmented line represents the average of all categorized proteins. (E to G) STRING networks (77) and graphical representation of protein lifetimes (expressed in days) in the brain of aged mice for a selection of proteins implicated in AD (E), PD (F), ALS, and HD (G). The four circles in each box represent in which sample type each lifetime was measured. The legends in the lower part of each panel formally clarify their association with the respective pathways. Significances are indicated as false discovery rates (FDRs) calculated for the specified pathways and reflect the relevance of the represented proteins for each neurodegenerative disease. Note a wide distribution of lifetimes within each pathology. NDDs are not generally coordinated by lifetimes.

Fig. 2.. Specific lifetime changes in the aging brain.

(A and B) Venn diagrams of lifetimes measured here [“aged mice”; 21 months (21m)] or previously published [“young adult mice”; 5 months (5m); (19); see also table S3]. (C and D) Comparison of lifetimes in cortex (C) or cerebellum (D) of 5- and 21-month-old mice (nonparametric Wilcoxon matched-pairs signed-rank test, ****P ≤ 0.0001; boxplots represent median, 25th to 75th percentile, with whiskers showing 5th to 95th percentile). Turnover is significantly lower in both cases. (E and F) Summary of 50 proteins whose lifetime after rescaling is either relatively longer-lived (rLL) (E) or relatively shorter-lived (rSL) (F) in at least three of four turnover datasets analyzed here (brain cortex, cerebellum, synaptic cortical, and synaptic cerebellar fraction; see also table S3). Heatmaps show lifetimes color-coded as z scores. Gray boxes, not measured. Log₂ fold change (log₂FC) summarizes ratios of protein lifetimes of 21- versus 5-month-old mice (±SEM). Red boxes in (E) indicate proteins implicated in NDDs (see table S4). (G and H) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of rLL in aged mice in all four datasets (G) (N = 72) or always rSL in aged mice (H) (N = 128), showing lifetime changes of at least >10% (for details, see table S3 and fig. S6). Several mitochondrial proteins linked to NDDs appear relatively shorter-lived in aged mice as summarized in (H) and in fig. S6B. (I) Proteins either rLL or rSL in aged mice according to functional affiliation [see (19) and table S5]. Mean ± SEM of the log₂FC (21 months versus 5 months). Proteins per category are indicated in parentheses. Significance against the consistently changed lifetimes was calculated using Brown-Forsythe and Welch ANOVA followed by Dunnett multiple comparison correction (*P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001). Only significant categories are reported (for the remaining list, see fig. S7).

Fig. 3.. Cell type, regional, and synaptic specificities of protein lifetime changes in the aged versus the young adult brain.

(A) Heatmap summarizing lifetime changes for 1143 proteins in the aged brain with respect to the young adult brain [representing the fold change expressed as log₂ (log₂FC) between 21- and 5-month-old mice as a z score], showing modifications in lifetimes that in aging affect more specifically either brain cortex or cerebellum. Because of space limitations, only the 15 rSL proteins (top) or the 15 rLL proteins in the cortex versus the cerebellum (bottom) are shown (for a detailed list, see table S6). (B) Gene set enrichment analysis of (A), summarizing the molecular function, biological process, and cellular component of the gene ontologies (GOs) significantly enriched with an FDR of <0.05 (for the whole results, see table S6). (C and D) Comparison [represented as in (A) and (B)] between lifetime changes in the total homogenate and the respective enriched synaptic fractions that were obtained as previously described (19).

Fig. 4.. Correlations between protein features and protein lifetime changes reveal a generalized shift of metabolic resources in the aged brain.

(A to D) Analysis of biochemical properties linked to lifetime changes in aged versus young brain. We obtained biochemical properties of proteins (see Materials and Methods and table S7), and we then measured (i) their correlations to change of lifetime in aged mice versus young mice (B) and (ii) difference for each parameter in three protein subgroups defined as the rSL quarter of the aged proteome (<25th percentile), the middle quarter (37.5th to 67.5th percentile), and the rLL quarter (>75th percentile) (D). Pearson’s coefficients (B) show correlations, where bars represent mean ± SEM and points are changes in the individual datasets (Pearson’s P values: *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001). (C) Pearson coefficients summarizing the correlations of lifetime log₂FCs to amino acid composition of proteins. Bars represent mean ± SEM, and points are changes in the individual datasets (Pearson’s P values). Amino acids in orange (expensive) have a positive correlation with lifetime change, while cyan ones (affordable) show an opposite trend. The percentile analysis (D) reinforces these findings (significance was calculated with either one-way ANOVA and Tukey multiple comparison correction or Brown-Forsythe and Welch ANOVA test with Games-Howell multiple comparison correction if SDs were significantly different between groups). (E) We considered proteins, on average, either short-lived, middle-lived, or long-lived. (F) Within these groups, we calculated average lifetime change in aged brain versus young adult brain (log₂FC, 21 months versus 5 months). Short-lived proteins tend to live relatively longer in the aged brain than in the young adult brain, while longer-lived proteins live shorter in the aged brain, pointing to a compression of proteome lifetimes with age (see also fig. S7). Boxplots: median, 25th to 75th percentile; whiskers, 5th to 95th percentile.