Deep Proteome Analysis Identifies Age-Related Processes in C. elegans

Abstract

Effective network analysis of protein data requires high-quality proteomic datasets. Here, we report a near doubling in coverage of the C. elegans adult proteome, identifying >11,000 proteins in total with ∼9,400 proteins reproducibly detected in three biological replicates. Using quantitative mass spectrometry, we identify proteins whose abundances vary with age, revealing a concerted downregulation of proteins involved in specific metabolic pathways and upregulation of cellular stress responses with advancing age. Among these are ∼30 peroxisomal proteins, including the PRX-5/PEX5 import protein. Functional experiments confirm that protein import into the peroxisome is compromised in vivo in old animals. We also studied the behavior of the set of age-variant proteins in chronologically age-matched, long-lived daf-2 insulin/IGF-1-pathway mutants. Unexpectedly, the levels of many of these age-variant proteins did not scale with extended lifespan. This indicates that, despite their youthful appearance and extended lifespans, not all aspects of aging are reset in these long-lived mutants.

Graphical abstract

Figure 1

SILAC-Based Deep Proteome Analysis Reproducibly Identifies 9,398 Proteins in C. elegans (A) Schematic overview shows the methodology used in this study. (B) Area-proportionate Venn diagram (top) shows reproducibility in protein identifications across three biological replicates, with the number of proteins in each region of the Venn diagram indicated on a schematic (bottom). (C) Principal-component analysis (PCA) plot shows reproducibility in protein quantification across the three biological replicates. (D) Histogram shows log₁₀-transformed protein abundance (MaxQuant intensity) for day 1 measurements. (E) A cumulative plot of protein abundance, as estimated using median protein intensity measurements based on three biological replicates. 9,398 proteins were identified in total with at least one peptide per protein. 90% of the bulk protein mass is made up of only 2,015 proteins (21% of measured proteins). The remaining protein identifications (7,383 or 79%) comprise less than 10% of the bulk protein mass. See also Figures S1–S4 and S8.

Figure 2

Of the Detected C. elegans Proteome, 8.5% Changes by a Large Magnitude during Aging (A) Volcano plot of log₂-transformed SILAC ratios (fold change) against the negative log₁₀ of the FDR-corrected p value calculated using ANOVA. Shown are the graphs for day 5 versus day 1 fold changes (left), day 10 versus day 5 (center), and day 10 versus day 1 (right). The shaded dark gray box represents the mean ± 1.96 SD for each plot. Horizontal gray lines denote p value cutoffs of 0.05 (^∗), 0.01 (^∗∗), and 0.001 (^∗∗∗). The number of proteins that increase (orange) or decrease (blue) significantly (p value < 0.05) in each graph also is indicated. (B) GO term enrichment (biological process) of the 627 age-variant proteins (blue and orange dots from the three volcano plots in A with p values < 0.05) performed using DAVID and plotted using REVIGO. The size of the bubbles is indicative of the number of proteins annotated with that GO term; bubbles are color coded according to significance. (C) KEGG pathway enrichment of the 627 age-variant proteins calculated using DAVID is shown. (D) Heatmap shows age-dependent changes in abundance of the above proteins generated using R. (E) Hierarchical clustering of the 627 proteins into six groups based on trend profiles. Results from GO term enrichment analysis (p < 0.05) of each cluster using DAVID also are depicted. Only GO terms in clusters 1, 4, 5, and 6 remained significant (p < 0.05) after multiple hypothesis correction (Benjamini). See also Figure S4.

Figure 3

GSEA Uncovers a Downregulation of Cellular Metabolic Pathways with Age (A) Density plots depicting the protein abundance of self-defined gene sets (left, insulin-IGF signaling; right, mTOR) relative to the entire dataset collected in this study. C. elegans proteins and their human orthologs are indicated in the figure. Negative regulators, in the context of extended longevity, are annotated with an asterisk (^∗). (B) GSEA (KEGG pathways) output of the day 10 versus day 5 changes in protein abundance depicted schematically using Enrichment Map (Cytoscape). The thickness of the lines connecting nodes (i.e., KEGG pathways) indicates the overlap of quantified proteins common to both nodes. See also Figure S5.

Figure 4

Levels of the Peroxisome Import Protein PRX-5 Decrease during Aging (A) Schematic depicting an adaptation of the KEGG pathway peroxisome in humans (adapted from www.genome.jp/kegg/). C. elegans homologs of the human proteins are shaded yellow. Also indicated are the proteins quantified in this dataset, with orange, blue, or black arrows to denote whether the proteins were found to increase or decrease in abundance or to remain relatively unchanged between days 5 and 10 of adulthood, as measured in our study. (B) Graph shows the decrease in levels of the peroxisomal import protein PRX-5 during aging, at the indicated time points. (C) Graph shows the decrease in mRNA levels of prx-5 during aging observed in previously published datasets in wild-type nematodes (left; Youngman et al., 2011) and temperature-sensitive sterile mutants (right; Budovskaya et al., 2008). See also Figure S5.

Figure 5

Peroxisomal Protein Import Is Impaired during Aging (A) Schematic depicting the correct peroxisomal localization of GFP targeted to the peroxisome (GFP-SKL) in young animals (left panel). If, as predicted by our MS analysis, the peroxisome import machinery is compromised during aging, GFP-SKL import into the peroxisome will be impaired (right panel). (B–D) Scale bar, 10 μm. (B) GFP-SKL is correctly localized to the peroxisome in the intestine of control animals, but mislocalized to the cytosol under conditions where prx-5 is knocked down with RNAi (n = 7). (C) GFP-SKL begins to accumulate in the cytoplasm of aging animals (day 5 adults) when compared to day 1 adults. Shown are representative images from the distal intestinal cells. Cytoplasmic (diffuse) GFP intensity was quantified, summed, and plotted from 81 individual z-sections at 0.5-μm intervals for each worm to minimize masking of the cytoplasmic signal by intense GFP foci (peroxisomes) as well as to minimize bleed through. The cytoplasmic signal was found to increase significantly with age (upper left graph; p value calculated using an unpaired t test = 7.68 × 10⁻⁵; n = 11 individuals for day 1 adults and n = 9 for day 5 animals). As the total hsp-16.2-driven GFP-SKL signal also was found to increase with age, cytoplasmic GFP measurements were corrected for total intensity in each image and age-dependent changes remained significant (upper right graph; p value from unpaired t test = 0.009). Also indicated is a count of the number of dots resolved (peroxisomes) at days 1 and 5 from a single z-section. This was found to increase significantly by day 5 (lower graph; p value from unpaired t test = 7.86 × 10⁻⁵). (D) As above, but we used worms that express GFP-SKL under the intestine-specific ges-1 promoter (strain VS15). Cytoplasmic GFP-SKL intensity increased significantly with age (p value from unpaired t test = 1.56 × 10⁻¹¹ for raw cytosolic intensities [left graph] and 4.54 × 10⁻¹⁰ for cytosolic GFP intensity normalized to total GFP intensity [right graph]; n = 15 for day 1 adults and n = 19 for day 5 adults). As in (C), the number of dots resolved also increased significantly with age (lower graph; p value from unpaired t test = 4.46 × 10⁻⁸). See also Figure S6.

Figure 6

Most Age-Variant Proteins Do Not Scale with Biological Age in Long-Lived Insulin/IGF-1 Mutants (A) Snapshots of GFP fluorescence at days 1, 5, and 10 of adulthood for the indicated GFP::protein fusions. Shown are high exposures at day 1 (upper panels) and comparable exposures at days 1, 5, and 10 (lower panels) (n = 20 per experiment). (B) HMG-11::GFP-expressing worms were grown on the indicated RNAi-bacteria (initiated at L1 stage) and GFP fluorescence was quantified on days 1, 5, and 10 of adulthood. Graphs show mean ± SD (n = 20 worms). Significance was calculated using an unpaired t test (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001; n.s., not significant). The daf-2 and hsf-1 dsRNA constructs are in different vector backbones, hence the corresponding empty vector control for each was used. Snapshots of images taken at day 5 (hsf-1 RNAi) and day 10 (daf-2 RNAi) also are shown. (C) As in (B), except we used SQST-1::GFP-expressing hermaphrodites (n = 20 except for day 10 hsf-1 RNAi where n = 11). (D and E) 88 proteins from the 627 age-variant proteins identified in this study were detected in a previous study by Walther et al. (2015) in wild-type, daf-2(e1370), and hsf-1(sy441) animals, and they were found to show similar abundance trends in wild-type animals when compared to our dataset. Of these, those that increased with age in wild-type animals also were quantified in daf-2(e1370) and hsf-1(sy441) worms (D). The stacked bar plot shows how the levels of these proteins vary in the different strains relative to wild-type animals at day 6 (hsf-1) or days 6 and 12 (daf-2) of adulthood, using raw data from Walther et al. (2015). Also indicated are representative examples of proteins present at lower, comparable, or higher levels in day 12 daf-2 mutants compared to wild-type controls. (E) As in (D), except that proteins whose levels were found to decrease in wild-type animals were analyzed. As in (D), raw data from Walther et al. (2015) were used to generate the bar plots. See also Figure S7.