Shift of the insoluble content of the proteome in the aging mouse brain

Abstract

During aging, the cellular response to unfolded proteins is believed to decline, resulting in diminished proteostasis. In model organisms, such as Caenorhabditis elegans, proteostatic decline with age has been linked to proteome solubility shifts and the onset of protein aggregation. However, this correlation has not been extensively characterized in aging mammals. To uncover age-dependent changes in the insoluble portion of a mammalian proteome, we analyzed the detergent-insoluble fraction of mouse brain tissue by mass spectrometry. We identified a group of 171 proteins, including the small heat shock protein α-crystallin, that become enriched in the detergent-insoluble fraction obtained from old mice. To enhance our ability to detect features associated with proteins in that fraction, we complemented our data with a meta-analysis of studies reporting the detergent-insoluble proteins in various mouse models of aging and neurodegeneration. Strikingly, insoluble proteins from young and old mice are distinct in several features in our study and across the collected literature data. In younger mice, proteins are more likely to be disordered, part of membraneless organelles, and involved in RNA binding. These traits become less prominent with age, as an increased number of structured proteins enter the pellet fraction. This analysis suggests that age-related changes to proteome organization lead a group of proteins with specific features to become detergent-insoluble. Importantly, these features are not consistent with those associated with proteins driving membraneless organelle formation. We see no evidence in our system of a general increase of condensate proteins in the detergent-insoluble fraction with age.

Keywords: aggregation; aging; neurodegeneration; protein homeostasis; proteomics.

Fig. 1.

Detergent-insoluble proteins accumulate in the brain tissue of 100-wk-old mice. (A) Outline of the experimental workflow. The MA icon will be used to indicate data generated from the meta-analysis. (B) Comparison of the composition of detergent insoluble fractions obtained from the cortex of 15- and 100-wk-old (n = 4, n = 4) mice using a two-sample paired t test. (C) Euclidean distance clustering of z-scored intensities from the significantly altered proteins. (D) Top GO terms for proteins enriched in the pellet fraction of old mice (red) and young mice (purple). Numbers of protein groups and P values (Benjamini–Hochberg corrected) in each category are indicated.

Fig. 2.

Validation of the aggregation of candidate proteins. FTA of (A) ASPA and (B) CRYAB using the cortex of 15- and 100-wk-old mice (n = 3, n = 3). The FTA signals are normalized to the mean of the corresponding young data points. Western blot signal from (C) ASPA and (D) CRYAB is normalized the α-tubulin loading control. Intensity ratios are then normalized to the mean of the young data points. The same six lysates were analyzed in A–D. SD and P values of two-sample paired t tests are indicated.

Fig. 3.

Detergent-insoluble protein feature analysis reveals features that separate datasets by age. (A) PCA analysis of the 12 different datasets assessed in the meta-analysis. Datapoints are colored by age and labeled by mouse phenotype. The binned old and young datasets are indicated by red or purple highlighted regions. The asterisk represents data collected in this study. (B–H) Violin plots comparing the distribution of the datasets binned into young (<15 wk; purple) and old (>80 wk; red) and of the mouse proteome (PME; gray) for the indicated features. P values (Hochberg-adjusted Wilcoxon test), number of proteins assessed for a given analysis (n), and effect sizes calculated as Z-scores (z) are shown. Secondary structure calculations were done using SCRATCH and intrinsic disorder using DISOPRED3. The MA icon indicates that the data are from the meta-analysis and arrows are proportionate to z.

Fig. 4.

Young datasets are enriched for features and proteins associated with MLOs. Violin plots comparing the distributions of (A) P-Scores, (B) RBPPred scores that predict RNA binding proteins, and (C) membraneless organelle and Granule Z-Scores for each protein (only MaGS >1.16 are considered to be likely part of a condensate). P values for all violin plots were calculated using a Benjamini–Hochberg-adjusted Wilcoxon test. The MA icon indicates that the data is from the meta-analysis, PME designates the proteome control group and arrows are proportionate to z-score as in Fig. 3. (D) Representation of Fisher test results indicating enrichment of MLO proteins as annotated by the drLLPS. Dot size represents the odds ratio and green intensity represents the Bonferroni-adjusted P values. (E) Violin plots comparing the average intensity of proteins identified in the pellet and supernatant fractions (P/S) in old and young mice in our study (indicated by the mouse with the hourglass). P values were calculated using a Benjamini–Hochberg-adjusted Wilcoxon test. Number of proteins assessed for a given analysis are shown. P/S ratios of all quantified proteins (gray), of MLO proteins in the young mice (purple), and of MLO proteins in the old mice (red) are shown. (F) Schematic representation of the accumulation of the detergent-insoluble proteins in young or old tissues that are associated with different features. A fraction of proteins that share features with condensate proteins (purple) such as an enrichment for intrinsically disordered regions (IDR) remains insoluble, whereas another subset of proteins that consist of more structured and hydrophobic proteins (red) accumulate upon aging, possibly due to increased misfolding.