Precisely measured protein lifetimes in the mouse brain reveal differences across tissues and subcellular fractions

Abstract

The turnover of brain proteins is critical for organism survival, and its perturbations are linked to pathology. Nevertheless, protein lifetimes have been difficult to obtain in vivo. They are readily measured in vitro by feeding cells with isotopically labeled amino acids, followed by mass spectrometry analyses. In vivo proteins are generated from at least two sources: labeled amino acids from the diet, and non-labeled amino acids from the degradation of pre-existing proteins. This renders measurements difficult. Here we solved this problem rigorously with a workflow that combines mouse in vivo isotopic labeling, mass spectrometry, and mathematical modeling. We also established several independent approaches to test and validate the results. This enabled us to measure the accurate lifetimes of ~3500 brain proteins. The high precision of our data provided a large set of biologically significant observations, including pathway-, organelle-, organ-, or cell-specific effects, along with a comprehensive catalog of extremely long-lived proteins (ELLPs).

Fig. 1

A highly calibrated method for determining protein lifetimes in vivo. a The interpretation of protein turnover in vivo is complicated by the initially non-labeled pool of amino acids that can be reused following the degradation of pre-existing proteins. b To solve this problem, a thorough protocol has been optimized and validated with several independent methods (as discussed in detail in the text and in Supplementary Figs. 1–9). Note that all steps highlighted in green have been performed for the first time in vivo

Fig. 2

The lifetimes of brain proteins. a Upper panel: distribution of 2381 lifetimes calculated in the brain cortex homogenate. Lower panel: lifetimes of 1266 proteins organized in 36 groups, accordingly to their organelle and/or functional affiliation, performed by integrating previously published categorizations^– (see also Supplementary Data 1). Each data point corresponds to a single protein lifetime. The black lines indicate the mean and the standard error of the mean (SEM) for each group. The analysis of variance (ANOVA) on the right summarizes P-values for the indicated comparisons, following Bonferroni post hoc test (*≤0.05, **≤0.01, ***≤0.001). b Lifetimes are positively correlated to protein abundances (determined with iBAQ), isoelectric point, and grand average of hydropathy (GRAVY). Lifetimes are negatively correlated to protein length and intrinsic disorder, while the correlation to mRNA abundances is not significant. The lines with error bars indicate averaged bins of 2 days, with SEM. Brain cortex homogenate data have been used in this figure and the protein and mRNA abundances were measured in this study, in the same preparation (brain cortex of the same mice)

Fig. 3

Detailed lifetimes of cytoskeletal, mitochondrial and synaptic proteins. a A subdivision of lifetimes for cytoskeletal, mitochondrial, and synaptic proteins subgroups^–. The analysis of variance (ANOVA) on the right summarizes P-values for the indicated comparisons, following Bonferroni post hoc test (*≤0.05, **≤0.01, ***≤0.001). b The lifetimes of 90 proteins of particular interest, subdivided in 8 groups, are detailed. For each protein, the 95% confidence interval is reported as a measure of the error. c A section through a 3D model of the synaptic bouton, indicating 60 proteins shown in the copy numbers (as in ref. ) and pseudo-colored as determined by their lifetime values. The two insets on the right side represent scatter plots of the lifetimes vs. either the hydropathy (upper panel) or the percentage of beta sheet stretches in the secondary protein structure (lower panel). In both cases there is a positive correlation between lifetimes and these values, indicating that hydrophobic and more structured proteins tend to live longer in the presynaptic bouton (ANOVA P-values ≤ 0.05). d A 3D view of a modeled synaptic vesicle, showing the lifetimes of 20 proteins (or protein complexes), as determined from the synaptic vesicle fraction of the brain cortex, see also Fig. 6. The proteins are color-coded from short-lived (green) to long-lived (red). The numbers indicate the lifetimes, in days and the error corresponds to the 95% confidence interval. The inset in the upper right corner is a synaptic vesicle, with proteins color coded for their lifetimes and shown in the appropriate copy numbers. For the synaptic vesicle the data from the cortex subcellular fractionation has been used (see also Fig. 6), while all other data is from brain cortex homogenate

Fig. 4

Lifetimes of proteins from sorted neuronal and glia cell nuclei. a Schematic representation of the pulsing strategy, followed by fluorescence-activated cell sorting (FACS), mass spectrometry analysis, and lifetime determination. b Representative scatter plot of sorting events, with the positive (green) and negative (blue) sorted populations highlighted vs. the forward scatter height (FSC-H). c Representative images of sorted neuronal (NeuN⁺) and glial (NeuN⁻) nuclei. d Scatter plot of protein lifetimes for neuronal and in glial nuclei. e Detailed lifetimes for components of the large and of the small ribosomal subunit. All differences are significant, with a Bonferroni adjusted P-value < 0.001. In all cases the ribosomes that are enriched in the nuclear envelope/rough endoplasmic reticulum are shorter-lived in glial cells than in neurons. See also Supplementary Data 1 for a detailed list of protein lifetimes in the nuclei of neurons and glial cells. See also Supplementary Fig. 18 for the string analysis of proteins either significantly longer-lived in neurons or in glial cells. Ribosomal constituents, focal adhesions and nuclear parts are significantly longer-lived in neurons vs. glial cells. The lower false discovery rates (FDR; an adjusted form of P-value to account for false-positive hits) observed in glia cells indicate that overall there are fewer long-lived groups of proteins in these cells

Fig. 5

The lifetimes from specific pathways are different between brain cortex and cerebellum. a Scatter plot of protein lifetimes in the cortex homogenate vs. the cerebellum homogenate. b–d Proteins significantly different in the two tissues for several pathways identified by the classification of lifetime changes (Bonferroni adjusted P-value < 0.001; see also Supplementary Fig. 20). Several exo-endocytosis cofactors are shorter-lived in the cerebellum when compared to the cortex (b), as well as specific adhesion molecules of the brain (c, left side). With the exception of septin 8, most septins (3, 5, 6, 7, and 11) are also longer-lived in the cortex (c, right side). On the contrary, histones are more stabilized in the cerebellum than in the cortex (d)

Fig. 6

Protein lifetime changes in the synaptic fractions. a Scatter plots of protein lifetimes between the homogenate and either the synaptosomal or the synaptic vesicle (SV) fractions. The blue plots represent data obtained from the cortex, while the green plots correspond to the cerebellum. In all cases, synaptic-enriched fractions are longer lived, as indicated by the upward trend vs. the diagonal identity line (represented as a segmented trait). All lifetimes are reported with their confidence intervals in Supplementary Data 1. b Venn diagrams showing the overlap of proteins significantly changed in the four datasets (25% change with Bonferroni adjusted P-values < 0.001). While some proteins change specifically in only one dataset, there are many proteins that changed in the same manner across different datasets. c Average lifetime changes in the synaptic fractions (only shown for proteins significantly different in at least two of the four fractions, with a change >25% and P-value < 0.001). A positive change indicates longer lifetimes in the synaptic fractions. In accordance to the scatter plots, the majority of lifetimes are increased in the synaptic fractions. Synaptic molecules already discussed in Fig. 2 are detailed and color-coded as in Fig. 3b. Several exo-endocytosis cofactors (turquoise) are stabilized, together with AP-1/3 adaptor proteins (light blue). Several tubulin subunits are stabilized (orange), as well as GluR-3. A number of other synaptic components are indicated in black. In general there is a clear differential regulation of adhesion molecules, with some either stabilized or destabilized at the synapse, suggesting that their localization might be a predominant determinant of their stability. See also Supplementary Fig. 21c-f for the detailed gene ontology analysis of the lifetime changes among different fractions

Fig. 7

Protein lifetime differences upon chronic environmental enrichment. a, b Scatter plots of protein lifetimes upon prolonged environmental enrichment in the cortex homogenate (a) or in the cortex synaptosomes (b) vs. the control mice. c Among the proteins whose lifetime have been precisely determined, there are 794 common hits. d Some of the proteins are significantly different (Bonferroni adjusted P-value < 0.001), and of these only 20 are common among the two cellular fractions. e A precise analysis of these proteins identifies synaptic components that are turned over at a higher speed following environmental enrichment, such as the presynaptic adhesion molecule Neurexin-4, the scaffold molecule CASK, the phosphoprotein synapsin-1 and the neuronal RasGAP SynGAP1 (see Supplementary Table 2 for details). Some mitochondrial components implicated in the metabolism of glutamate and acetyl-CoA are also turned over at a higher speed following environmental enrichment (Supplementary Table 2). On the contrary, two myelin components (PLP1 and Claudin 11) are stabilized upon environmental enrichment (Supplementary Table 2). f String analysis of the 20 common proteins. This identifies three functional protein clusters changed upon environmental enrichment (Supplementary Table 2). The detailed analysis indicates that the most important differences between these two cohorts of mice are at the level of myelin, the mitochondrial inner membrane and the synapse

Fig. 8

Protein lifetimes across different cells or tissues. a, b Scatter plots of the lifetimes in the cortex homogenate versus the lifetimes of primary rat neurons (published elsewhere; a). Non-mitochondrial proteins are represented in blue while mitochondrial proteins are represented in green. The lifetimes of proteins in vitro are shorter than in vivo, both for mitochondrial and non-mitochondrial proteins. Even accounting for this overall difference, presynaptic proteins are shorter living in cultured cells than in vivo (see Supplementary Table 3 for details). This difference might be due to the fact that culture neurons are still growing and developing axons and synapses at the time of the measurements. b Scatter plots of the lifetimes in the cortex homogenate versus the lifetimes in heart samples. In the heart the turnover of proteins is overall faster than in the brain, but the mitochondria proteins tend to be longer-lived than the rest of the proteome. The gene ontology analysis suggests that the main lifetime differences observed between the two tissues are at the level of cell cycle and metabolic processes, in line with their different requirements of the two tissues (Supplementary Table 3). c Scatter plots of the lifetimes in the cortex homogenate versus the lifetimes of skeletal muscle (gastrocnemius). Protein turnover in muscle is also faster than in the brain. If compared with the heart, the mitochondrial turnover in the muscle is relatively faster, probably reflecting a difference in the metabolic requirements of the two tissues (please compare b and c). In the muscle the turnover of proteins involved in DNA biosynthesis is relatively faster than in the brain, probably because of the limited cell renewal and DNA synthesis observed in the nervous system (Supplementary Table 3)