Molecular Signatures of Neurodegenerative Diseases Identified by Proteomic and Phosphoproteomic Analyses in Aging Mouse Brain

Abstract

A central hallmark of neurodegenerative diseases is the irreversible accumulation of misfolded proteins in the brain by aberrant phosphorylation. Understanding the mechanisms underlying protein phosphorylation and its role in pathological protein aggregation within the context of aging is crucial for developing therapeutic strategies aimed at preventing or reversing such diseases. Here, we applied multi-protease digestion and quantitative mass spectrometry to compare and characterize dysregulated proteins and phosphosites in the mouse brain proteome using three different age groups: young-adult (3–4 months), middle-age (10 months), and old mice (19–21 months). Proteins associated with senescence, neurodegeneration, inflammation, cell cycle regulation, the p53 hallmark pathway, and cytokine signaling showed significant age-dependent changes in abundances and level of phosphorylation. Several proteins implicated in Alzheimer’s disease (AD) and Parkinson’s disease (PD) including tau (Mapt), Nefh, and Dpysl2 (also known as Crmp2) were hyperphosphorylated in old mice brain suggesting their susceptibility to the diseases. Cdk5 and Gsk3b, which are known to phosphorylate Dpysl2 at multiple specific sites, had also increased phosphorylation levels in old mice suggesting a potential crosstalk between them to contribute to AD. Hapln2, which promotes α-synuclein aggregation in patients with PD, was one of the proteins with highest abundance in old mice. CD9, which regulates senescence through the PI3K-AKT-mTOR-p53 signaling was upregulated in old mice and its regulation was correlated with the activation of phosphorylated AKT1. Overall, the findings identify a significant association between aging and the dysregulation of proteins involved in various pathways linked to neurodegenerative diseases with potential therapeutic implications.

Graphical Abstract

Highlights

• •Multi-protease approach reveals age-related phosphoproteome changes in the brain.
• •Phosphoproteins are involved in neurodegeneration and glutamatergic signaling.
• •Many kinases showed increased abundance and phosphorylation in the brains of old mice.
• •Results show evidence of the age-dependent upregulation of the PI3K-AKT-mTOR pathway.

In Brief

Proteome and phopshoproteome analyses of the brains of mice across three different age groups (young-adult, middle-aged, and old mice) were performed using a multi-protease digestion strategy. Proteins associated with senescence, neurodegeneration, inflammation, cell cycle regulation, and cytokine signaling show significant age-dependent changes in phosphorylation levels. Specific proteins implicated in Alzheimer’s and Parkinson’s diseases, such as tau (Mapt), Nefh, and Dpysl2, were hyperphosphorylated in old mice. Results highlight the link between aging and dysregulated proteins in neurodegenerative pathways.

Aging is a significant risk factor for various diseases, including cancer and neurodegenerative diseases (1, 2). Omics studies on aging have traditionally focused on transcriptional regulation of proteins and their cellular functions (3). Such studies have provided valuable insights into age-dependent changes in gene expression patterns, revealing differences between normal aging and diseased aging, and molecular differences among various age-related pathologies (4). However, gene expression analysis alone cannot fully ascertain changes in proteins. It has been previously shown that only 27% of the variation in protein abundance can be explained by mRNA abundance, while processes related to translation and protein degradation can explain about 40% of protein variations (5), suggesting that proteomic studies can confirm findings from gene expression, and also identify molecular signatures of aging that occur independent of gene expression changes (6). The heterogeneity of age-related changes between transcripts and proteins could be a result of multiple mechanisms such as reduced proteasome activity (7), ribosome occupancy (8) and post-translational regulations (9) such as phosphorylation. Previous proteomic studies have investigated aging-related changes in protein abundance in kidney, heart, and liver but similar studies using brain tissues are limited.

Phosphorylation plays a significant role in the aging process and is associated with many age-related neuropathological changes (10). Information about protein phosphorylation and other types of post-translational modifications (PTMs) cannot be obtained by gene expression analysis (6). Changes in protein phosphorylation influence key regulatory pathways involved in cell growth, proliferation, differentiation, and apoptosis, which are all critical factors in aging and age-related diseases (11). Dysregulation of protein phosphorylation is a hallmark of cancer as many oncogenes and tumor suppressors are regulated by phosphorylation, and dysregulation of these proteins contributes to cancer development (12). In addition, aging is one of the most common risk factors associated with various neurodegenerative diseases including Alzheimer’s disease (AD) and Parkinson’s disease (PD). A common feature of importance in these neurodegenerative diseases is the pathological irreversible accumulation of disease-specific proteins into insoluble aggregates (13, 14). Aberrant protein phosphorylation, which is thought to be due to aging, genetic, and environmental factors, leads to the formation of toxic protein aggregates, such as hyperphosphorylation of tau and α-synuclein, and such molecular changes are associated with neurodegenerative diseases (15, 16, 17). A central hallmark of AD pathology is the irreversible aggregation of amyloid β (Aβ) in extracellular senile plaque and hyperphosphorylated tau protein in intracellular neurofibrillary tangles. Similarly, the hallmark pathology in PD is the irreversible aggregation of α-synuclein, which is associated with serine 129 phosphorylation (18, 19). These pathological hallmarks are associated with synapse loss, neural cell death, and cognitive impairment (20, 21). However, the role of specific PTMs in disease pathogenesis and their relationship to cellular toxicity and aging are not fully understood. As a result, comprehensive proteomic and phosphoproteomic analyses of mouse brain tissues at different ages can reveal molecular signatures of aging, specifically related to neurodegeneration, potentially leading to the discovery of new drug targets.

Proteomic and phosphoproteomic studies are commonly performed using peptides derived from tryptic digestion (22, 23) because trypsin is highly specific, effective, and generates a higher number of peptides compared to other proteases (24, 25). Therefore, the majority of published proteomic and phosphoproteomic studies have predominantly used trypsin as a protease. However, many potentially interesting sequences and phosphosites might be missed by this workflow, as many peptides generated by tryptic digestion may not always possess appropriate chemical properties that make them suitable for detection by LC-MS/MS (26). While there are reports of increased proteome and phosphoproteome coverage using multiple proteases for digestion (24, 26, 27), this has not been systematically applied in brain proteome and phosphoproteome analysis. To increase the coverage of the brain proteome and phosphoproteome of different age mice, we used four different proteases for protein digestion (trypsin, chymotrypsin, endoproteinase Asp-N, and endoproteinase Glu-C) in this study. Low abundant phosphopeptides in the digested samples were subsequently affinity purified by PolyMAc phosphopeptides enrichment kit (Tymopra Analytical Operations, LLC), which is based on a metal-ion functionalized soluble nanopolymer to chelate phosphopeptides (28).

This study aimed to investigate age-dependent changes in the proteome and phosphoproteome of whole brain tissues collected from wild-type male mice across three different age-groups: adult, middle-aged, and old. The data revealed significant age-dependent alterations in the brain’s proteome and phosphoproteome landscape, particularly proteins associated with several crucial biological pathways and processes such as cellular senescence, cell cycle regulation, the p53 hallmark pathway, neurodegeneration, cytokine signaling, and inflammatory responses. The results provide a comprehensive view of age-dependent alterations in the brain proteome and can potentially serve as markers or indicators of age-related vulnerabilities in the brain associated with specific proteins and phosphoproteins.

Experimental Procedures

Animals

Animal protocols were approved by the Purdue Animal Care and Use Committee and were in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals (#2008002069). Mice were housed individually in ventilated cages in 12 h light-dark cycles at temperature (70–75°F) and humidity (40–60%) controlled vivarium and maintained under ad libitum food and water. Brain tissue samples were collected from 3, 4, 10-, 19-, 20-, and 21-month-old wild-type (WT) male C57BL/6J littermates from the A53T SynGFP mouse line (29) which is currently maintained at Purdue University. Every effort was made to minimize the number of mice used and their suffering. Mice were anesthetized with 2% isoflurane in 100% oxygen to a surgical board.

Experimental Design and Statistical Rationale

Brains from five young-adult male mice (3–4 months), hereafter referred to as “adult mice,” five old adult male mice (19–21-months) referred to as “old mice,” and three mid-age mice (10 months), referred to as “mid-age mice,” were used for this analysis. Statistical analysis was performed using a one-way ANOVA for both global and phosphoproteomics data. False Discovery Rate (FDR) corrected p-values, hereafter referred to as q-values, were considered statistically significant if q ≤ 0.05. Multiple comparisons were corrected by Tukey's HSD (honestly significant difference) tests.

Tissue Collection and Homogenization for Protein Extraction

Whole brains were collected from each mouse and washed with phosphate-buffered saline (1×). After harvest, the brains were flash-frozen in liquid nitrogen and stored at −80 °C until further processing. The whole brain tissues were cut and shredded with the Shredder SG3 Kit (Pressure Biosciences, Inc) in 50 mM TEAB (triethylammonium bicarbonate) in deionized water supplemented with protease and phosphatase inhibitors. The phenylmethyl sulfonyl fluoride (PMSF) was used as a serine protease inhibitor to a final concentration of 1 mM. A mixture of sodium fluoride (NaF), Na-molybdate (Na₂MoO₄), Na-orthovanadate (Na₃VO₄), and Na-beta glycerophosphate (C₃H₁₉Na₂O₁₁P) to a final concentration of 10 mM was used as phosphatase inhibitors. The ground tissues were transferred to a Precellys soft tissue homogenizing CK14 tubes (Bertin Corp) and homogenized six times at 3200 rpm for three cycles, 20 s in each, followed by probe sonication (3 s pulses for 6 times, with 5 s pause in between). The protein concentration was measured by bicinchoninic acid (BCA) assay using BSA as a standard. Volumes equivalent to 1 mg of protein were taken of each sample lysate for subsequent digestion and phosphopeptides enrichment.

Multiprotease Protein Digestion and Sample Preparation for LC-MS/MS Analysis

Brain lysate volumes were adjusted to 50 uL with TEAB, and further diluted in 50 μl urea (final Conc. of 6 M urea in 50 mM TEAB) and incubated at room temperature for 30 min for complete solubilization. Disulfide bonds were reduced with 10 mM dithiothreitol (DTT) for 45 min at 37 °C. Samples were then cooled to room temperature and cysteines were alkylated with 20 mM iodoacetamide (IAA) for 45 min in the dark, at room temperature. After alkylation, DTT was added to the samples (5 mM) and incubated at 30 °C for 20 min for scavenging residual IAA. For digestion with trypsin, samples were diluted with 460 μl of 50 TEAB containing 20 μg trypsin (ThermoFisher) for digestion at 1:50 trypsin:substrate ratio. To the digestion mix, 30 μl acetonitrile (ACN) and 0.65 μl of 1M CaCl₂ were added before samples were incubated at 37 °C for 18 h in a thermomixer. The digestion was halted by the addition of 0.65 μl trifluoroacetic acid (TFA). The same protocol was used for chymotrypisn, AspN and GluC; with ZnSO₄ used in place of CaCl₂ for AspN. Peptides were then desalted with the Pierce peptide desalting spin columns (ThermoFisher). A total of 20 μg of peptides from each sample were separated at this step for global analysis. Phosphopeptide enrichment was then carried out using the remainder of the samples with the PolyMac spin tips (Tymora Analytical), following manufacturer’s recommendation (28).

LC-MS/MS Data Acquisition

Dried samples were reconstituted in 3% ACN, 0.1% formic acid (FA) and separated using a 25 cm nanoflow Aurora Series UHPLC packed emitter columns, by reverse-phase chromatography in a Dionex UltiMate 3000 RSLC system (ThermoFisher) (30). Global analysis was performed with a 130-min gradient method, with a flowrate of 400 nl/min. Briefly, samples were injected with 2% mobile phase solution B (80% ACN with 0.1% FA in water). Mobile phase B was then increased to 8% at 5 min, and linearly increased, reaching 27% B at 80 min. B was then increased to 45% until 100 min, and to 100% at 105 min. The gradient was then held constant until 112 min, at which point it was reverted to 0% B until the end of the run. MS analyses were performed in the Orbitrap Fusion Lumos Tribrid Mass Spectrometer, with the orbitrap detector, with a MS1 resolution of 60,000 and MS2 resolution at 15,000. Quadrupole isolation was set to “True.” The scan range used was between 350 to 1600 m/z. Radio Frequency (RF) lenses was set to 30%, Automatic Gain Control (AGC) target was set to “Standard”, with a maximum injection time of 50s and 1 microscan. A dynamic exclusion duration of 60s was used, with exclusion of isotopes. Data-dependent mode “Cycle Time,” with 3s between scans were used. HCD (Higher Energy Collisional Dissociation) was used for fragmentation, with HCD collision energy set to 30%. We also used an MS/MS maximum injection time of 50s and 1 microscan.

LC-MS Data Analysis

Raw MS/MS data were searched against the UniProt (31) mouse database (Downloaded on February 14th, 2023; 63,617 entries, including protein isoforms) using the MaxQuant (32) platform (Ver 2.0.3.1). Appropriate enzymes were selected for specific digestion, with up to two missed cleavages (or 4 for phosphoproteomics). Variable modifications were set for “methionine oxidation” and for phosphoproteomics “STY phosphorylation”; and “Carbamidomethyl” was set as a fixed modification. A false discovery rate (FDR) of 1% was used for both peptides and proteins identification. Additionally, 10 ppm was selected as the main search peptide tolerance value, and 20 ppm was set for the MS/MS match tolerance. Peptide quantitation was performed using “unique plus razor peptides”. The MaxQuant output files were processed and analyzed using the Perseus (33) biostatistics platform for statistical analysis. “Contaminants”, “reverse”, and “only identified by site” proteins were filtered out, and LFQ (label free quantitation) intensity values were Log₂ transformed. All downstream analyses were performed after further data filtering to retain only proteins identified by a at least 2 MS/MS counts and detected in minimum valid values in at least one group. Missing values were imputed based on the normal distribution of LFQ values. Intensity values and localization probability scores were used for phosphoprotein analysis. The raw phospho STY data file was filtered for proteins with a localization probability ≥0.75, and 70% valid intensity values in at least one group. Missing values were again imputed with values drawn from the normal distribution. Statistical significance was inferred based on ANOVA tests. Proteins with a q-value ≤0.05 for global and phospho datasets were considered significantly regulated. Gene ontology (GO) was performed using Metascape (34) online software.

Development and Analytical Validation Using Targeted MS Analysis

The MRM analysis was carried out as described previously (30). The Skyline software (v22.2) (skyline.ms) (MacCoss Lab Software) was used for developing MRM methods. The SRM Atlas (Complete Human SRM Atlas database) was used to select tryptic peptides and transitions for each protein surveyed, using y-ions as the product ions. At least four proteotypic tryptic peptides (no missed cleavages) per protein were then imported into Skyline, and the transition list with collision energy parameters were generated (Supplementary Table S8). A TSQ Endura Triple Quadrupole (QqQ) Mass Spectrometer (ThermoFisher Scientific) with the Flex ESI-interface in SRM mode in positive polarity was used to carry out all MRM analyses. The MS analysis was conducted with the spray voltage set at 2400 V, and the transfer capillary temperature was set at 275 °C. The MRM transitions were acquired with a cycle time of 0.8 s, Q1 resolution of 0.7 (full width at half maximum), and Q3 resolution of 1.2. This assay is classified under a Tier three level analysis. Cysteine carbamidomethylation was set as a fixed modification. Peaks containing at least three measured transitions, with consistent retention time across all samples were considered for further analysis. The log₂ transformed peak areas for each peptide then compared using two-sample t-Tests. The liquid chromatography (LC) setup used was the same as described in the previous section, run in a 60-min gradient as follows: samples were injected in 2% B and increased in a linear fashion until 27% B was reached at 40 min, 45% B at 45 min, and 100% B at 50 min. The concentration of B was then held at 100% for min before returning to 2% B and maintained at 2% B until the end of the run.

Western Blotting

Western blots were performed as described previously (30). Briefly, 40 uL of lysate was cleared by centrifugation at 16,000g for 5 min, supernatants were transferred to new tubes and precipitated with four volumes of cold acetone overnight at −20 °C. Samples were then centrifuged at 20,000g for 15 min, supernatant was discarded, and pellets were washed with 80% cold acetone, incubated at −20 °C for 20 min, centrifuged and 20,000g for 10 min before discarding the supernatant. The wash procedure was repeated a total of 3 times. The samples were then resolubilized in 40 uL of 1% SDS for 1 h at room temperature with constant shaking. The total protein concentrations were then measured by BCA assay, and 20 ug were used for the analysis. Volumes from all samples were adjusted, then mixed with NuPAGE LDS Sample Buffer (ThermoFisher Scientific) at a 3:1 ratio and heated to 70 °C for 10 min. Samples were then loaded into a NuPAGE 4 to 12% Bis-Tris gel (ThermoFisher Scientific), and separated at 200 V. Proteins were then transferred into a nitrocellulose membrane using a Mini Blot Module (ThermoFisher Scientific) for 1 h at 15 V, following the manufacturer’s recommendation. After the transfer, the membrane was blocked with 3% BSA in 1xTBS-T for 40 min at room temperature, before the addition of the primary antibodies. All antibodies were incubated with the membrane in 3% BSA in 1xTBS-T at 4 °C overnight, before washing and blotting with the appropriate secondary antibody. The following primary antibodies were using for blotting: Actin mAb (mAbGEa) (# MA1-744; ThermoFisher Scientific; 1:3000 dilution), DJ-1 (D29E5) mAb (# 5933; Cell Signaling Technologies; 1:1000 dilution), Phospho-CRMP2 (Ser522) Ab (# PA5-143675; ThermoFisher Scientific, 1:1000 dilution).

Results

Aging Leads to an Extensive Modulation of the Brain Proteome and Phosphoproteome

Physiological, morphological, and molecular changes that occur in the brain because of organismal aging have been a topic of great interest. Several studies have reported changes that occur at the RNA and protein levels using high-throughput techniques (26, 35, 36). However, there are still very few studies that explore the differential regulation of PTMs during biological aging. To fill in this gap, we employed integrated proteomic and phosphoproteomic analyses using a multi-protease digestion strategy, with four different enzymes (Trypsin, Chymotrypsin, AspN, and GluC) to obtain a comprehensive and in-depth view of the aging brain proteome and phosphoproteome (Fig. 1A). This multi-protease digestion approach has been previously demonstrated to greatly improve the sequence coverage of identified proteins, and, consequently, significantly increase the number of phosphosites that can be quantified in a mass spectrometry analysis (26). Using this approach, we identified 4641 protein groups (Supplemental Table S1 and Supplemental Fig. S1A), as well as 18,492 phosphorylated peptides and 10,474 class I phosphosites (phosphosites with probability >0.75) in this study (Supplemental Table S2 and Supplemental Fig. S2A).

Fig. 1.

Global proteomicsof theaging mice brain. A, experimental workflow. Brains from adult, mid-age and old mice were harvested and homogenized as described in methods. Proteins were then digested using four different enzymes (AspN, Chymotrypsin, GluC and Trypsin) in parallel, and phosphopeptides were enriched prior to LC-MS/MS. The raw LC–MS/MS data was then processed using the MaxQuant (www.maxquant.org) and Perseus (www.maxquant.org/perseus) platforms. B, Violin plots depicting the overall distribution of Log₂(LFQ) values in the global dataset. Yellow hue corresponds to adult mice, cyan mid-age and red corresponds to old mice. C, PCA plot of all brain samples analyzed in the global proteomics. Age groups of the mice are indicated by the same hues as described in (B). D, Volcano plot analysis of all quantified proteins. Significantly regulated (q < 0.05 & logFC >|0.5|) proteins are color coded, and top regulated proteins are labeled. E, Heatmap representation of all significantly changing (q < 0.05) proteins in our analysis. Hue represents the Z-score normalized Log2(LFQ) values. Red hue indicates upregulated proteins and blue represents downregulated proteins. Proteins were clustered into 5 clusters by k-means. F, Z-scored Log₂(LFQ) values from proteins from each k-mean cluster, across replicate mice from adult, mid-age, and old groups. G, top four Reactome Gene Set terms enriched for the significantly changing proteins found each k-means cluster represented in Figure 1, E and F.

Our global analysis showed a consistent protein abundance distribution as displayed by the log₂ transformed LFQ (label-free quantitation) values for all quantified protein groups across all samples (Fig. 1B). Pearson correlation analysis of all replicates across age groups (Supplemental Fig. S1C) also showed a remarkable consistency and biological reproducibility among different runs and confirmed a high-quality proteomic dataset. A principal component analysis (PCA) revealed distinct clusters that can be defined by the age groups of each replicate mouse; indicating that the variance in our data arises from physiological aging (Fig 1C and Supplemental Fig. S1B).

To determine how the old mice differed from the adult mice in terms of protein levels, we displayed all quantified proteins as a volcano plot, highlighting the proteins with the highest log transformed fold-change and lowest q-values. Namely, Hapln2, Tspan2, and Tcp11I1 were among the topmost abundant proteins, and Chtop, Pdcd10, and Atp1a4 were among the lowest abundant proteins between the old and adult mice (Fig. 1D). Statistical analysis using one-way ANOVA identified 445 protein groups that were significantly different among the three age groups (q < 0.05) (Fig. 1E and Supplementary Table S3). K-mean clustering grouped them into five distinct clusters. To visualize the abundance patterns of the significant proteins in each cluster, we plotted them using Z-score values across all samples as line charts (Fig. 1F). Cluster 1 includes proteins that had their levels increased only in old mice; cluster two contains proteins that decreased in their levels only in old mice; cluster three depicts proteins decreased in their levels in middle-aged mice and old mice, and cluster four represents proteins more abundant in middle-aged and old mice. Cluster five contains proteins that increase in abundance as a function of age.

To contextualize the proteins contained in each cluster (Fig. 1F), we performed an enrichment analysis using the Reactome Gene Set database (RGS), via the Metascape platform (34). Figure 1G shows the top four terms enriched for each cluster. “Neurotransmitter release cycle”, “HSP90 chaperone cycle for steroid hormone receptors (SHR) in the presence of ligand” and “Transmission across Chemical Synapses” were among the top enriched terms for clusters in which protein levels were significantly elevated in old mice. These pathways are linked to neurodegeneration, with studies suggesting that dysregulation in neurotransmission and impairments in synaptic function underlies cognitive decline (37). Proteins found under the gene sets enriched in this study are listed under Supplemental Table S4.

On the other hand, “L13a-mediated translational silencing of Ceruloplasmin expression”, “Signaling by Rho GTPases”, “Mitotic Prometaphase” and “Post-translational protein phosphorylation” were the top-ranked terms among the Reactome gene sets with lowest q-values for clusters comprising downregulated protein groups in old mice. “L13a-mediated translational silencing of Ceruloplasmin expression”, the term with the lowest q-value enriched in cluster 2, primarily comprises ribosomal and elongation factor proteins. The Rho GTPases, on the other hand, play an important role in maintaining synaptic plasticity, which is crucial for brain functions such as learning and memory (38). Thus, a decrease in proteins involved in Rho GTPases signaling may indicate age-associated cognitive impairment. The downregulation of post-translational protein phosphorylation in old mice prompted us to investigate the dynamics of protein phosphorylation in the brain during aging.

Aging Brain Proteome is Associated With Protein Hyperphosphorylation

Similarly to the global data, the phosphoproteome dataset displayed a comparable distribution of LFQ values (Supplemental Fig. S2B). Our Principal Component Analysis (PCA) plot showed that samples from the same age group clustered together (Fig. 2A), suggesting that the observed variance in the data likely originated from physiological aging. Our Venn diagram shows minimal overlap among the class I phosphosites mapped by each individual enzyme showcasing the advantages of a multi-enzyme digestion approach for comprehensive phosphoproteomic analysis (Fig. 2B). This is also illustrated by our ability to map hundreds of phosphosites to various kinases spanning most of the mouse kinome (Fig. 2C).

Fig. 2.

Phospho proteomics oftheaging mice brain. A, PCA plot of all brain samples analyzed in the phospho proteomics analysis. Age groups of the mice are indicated by the different hue, with yellow corresponding to adult mice, cyan to mid-age and red corresponding to old mice. B, Venn Diagram representing the overlap between phosphosites identified after proteolysis with Trypsin, Chymotrypsin, GluC and ApsN. C, kinase tree depicting all phosphorylated kinases that could be mapped using our multi-enzyme digestion strategy. Illustration reproduced courtesy of Cell Signaling Technology, Inc. D, Heatmap representation of all significantly changing (q < 0.05) phosphosite proteins in our analysis. Hue represents the Z-score normalized Log2(LFQ) values. The yellow hue indicates upregulated phosphosites and the back represents downregulated phosphosites. Proteins were clustered into 5 clusters by k-means.

To identify significantly regulated class I phospho (STY) sites in our dataset, we utilized similar parameters for the statistical analysis as used in our global analysis (q-values <0.05 following ANOVA testing). Of the 10,474 class I phosphosites, 6155 with distinct multiplicity were present in 70% of the replicates in at least one age group, from which 1242 phosphosites with distinct multiplicities met our significance criteria (Supplemental Table S5). These significant phosphosites were subdivided into five distinct clusters, based on their abundance patterns. Among them, Clusters 3 and 5 clearly stood out, representing most of the significant phosphosites which were defined by elevated phosphorylation levels in the old mice, indicating that increased protein phosphorylation accompanies aging. Interestingly, clusters 3 and 5 were characterized by proteins involved in synaptic processes (Fig. 3A). We also observed an increased phosphorylation of proteins involved in neurodegeneration and cellular senescence, (Fig. 3, B and C) further highlighting the role of protein hyperphosphorylation in these biochemical processes (Supplemental Table S6).

Fig. 3.

Age-related changes in pathways governed by phosphoproteins are significantly dysregulated in old mice.A, cluster-wise KEGG enrichment analysis of significantly regulated proteins. Colors and numbers indicate the k-means cluster (Fig. 3D) for which the Gene Sets were enriched. B, differentially regulated phosphosites in proteins involved in the “Neurodegeneration” and (C) “Cellular Senescence” pathways.

Analyses using the Human Phenotype Ontology database term “Neurodegeneration” (Fig. 3B), show a hyperphosphorylation of both the Mapt and the Nefh proteins. Mapt phosphorylation has been linked to self-aggregation into filaments, leading to Mapt protein aggregation and plaque formation (39). We could not find prior evidence of Mapt phosphorylation at sites S147 and S148. Therefore, we may have identified novel Mapt phosphorylation sites in mouse brain under our experimental conditions. It is worth highlighting that S147 and S148 exhibit contrasting phosphorylation patterns in old mice. While S147 shows a significant increase, the phosphorylation signal of S148 declines in old mice compared to adult and middle-aged mice. Given the proximity of these two phosphorylated sites, we manually evaluated the MS/MS spectra of the phosphopeptides containing these residues (Supplemental Figs. S6 and S7). This manual evaluation confirmed the confident mapping and precise localization of each individual site. Additionally, we were able to detect these phosphorylated sites consistently in each mouse replicate (Supplemental Figs. S6C and S7C), confirming their high-confidence phosphorylation. The phenotypical consequence of this observation at the phosphorylation level requires further investigation.

Aging Promotes Kinome Changes Accompanied by Neuroinflammation and Synaptic Signaling Disruption

Kinases are responsible for the phosphorylation of their substrates, but their own activation hinges on PTMs (11). Thus, it is crucial to obtain a comprehensive map of all phosphorylated kinases identified in the phosphoproteome dataset, to identify major players and better understand underlying molecular mechanisms governing the changes in protein phosphorylation during aging. We depicted all significantly regulated kinases on a kinase tree and labeled them according to the cluster(s) at which their phosphosites belong in the heatmap. Our data revealed that most kinases that showed changes in their phosphorylation status belong to the CAMK, AGC, and STE groups (Fig. 4A).

Fig. 4.

The brain kinome suffers extensive changes during aging.A, kinase tree depicting differentially phosphorylated kinases in old mice. Colors and numbers indicate the k-means cluster (Fig. 2D) for which a phosphosite from this kinase falls into. B, PhosR Kinase activation enrichment analysis. Red hue indicates higher kinase enrichment scores C and D, enriched kinases can be categorized into 5 modules, that are composed of distinct target phosphosites.

To correlate the changes in kinase phosphorylation to kinase activity, we performed kinase-substrate enrichment analysis using the PhosR tool (40), specifically focusing on significantly upregulated phosphosites in the brains of old mice. Thus, we aimed to pinpoint specific kinases whose activity patterns changed as the brain ages. We identified 13 such kinases which were then plotted in a heatmap, against the top five phosphosites for each kinase (Fig. 4B). The enriched kinases could be divided into five distinct clusters, or modules, based on the enrichment score of their target sites (Fig. 4C). Interestingly, kinases from the same module clustered with distinct kinase networks (Supplemental Fig. S4, A and B), with AGC and CAMK kinases preferentially clustering together in module 1, and CMGC kinases forming another cluster in module two. It is important to note that Csnk2a1 is the only kinase that clustered in module 3 (Fig. 4D). The consensus phosphorylation motif for Csnk2a1, SxxE is one of the three most common phosphorylation motifs present in our dataset (Supplemental Fig. S4C). Taken together, these results demonstrate how signature proteins involved in the immune response and synaptic signaling changed at both their phosphorylation and global levels during aging.

We identified dozens of proteins involved in the Reactome term “cytokine signaling and immune system”, many of which were significantly different between old and adult mice (Supplemental Fig. S3A). Three such proteins, Itgb1, Crkl, and Prkaca, had both a q value < 0.05 and a fold-change >1.4 (Supplemental Fig. S3A). Prkaca was also significantly increased in the Wikipathway “Neuroinflammation and Glutamatergic Signaling” (Supplemental Fig. S3B), and it was one of the kinases enriched in our PhosR analysis (Fig. 4B), suggesting it may be one of the key players in the onset of age-associated neuroinflammation. This hypothesis was supported by the fact that the number of phosphosites associated with neuroinflammation pathways that significantly differ between old and adult mice were much higher than the number of significant proteins at the global levels. The differences in the phosphorylation levels of the phosphosites associated with neuroinflammation pathways between old and adult mice were also much more drastic (Supplemental Fig. S3, A–D). Thus, the neuroinflammation pathway may be an intriguing avenue for further investigation into the molecular mechanisms and functional consequences responsible for age-related differences in protein phosphorylation.

Aging is Accompanied by Differential Regulation of CD Molecules

Cluster of differentiation (CD) molecules are a denomination of proteins found on the surface of cells. The expression of specific CD proteins is often used to characterize cell phenotypes and can serve as a marker of aging and age-associated diseases (41). CD molecules can also play an important role in the immune system, and in inflammatory processes (42). Our results demonstrate that the CD9 protein, a tetraspanin membrane protein that is involved in cellular adhesion, cancer metastasis, and in the immune response (43), was significantly more abundant in old mice (Fig. 5C). CD9 has recently been found to be upregulated in senescent endothelial cells. Specifically, CD9 regulates cellular senescence via the activation of the phosphatidylinositide 3 kinase-AKT-mTOR-p53 signaling pathway (43). These results are in accordance with the upregulation of phospho AKT1, as well as with the PhosR enrichment of AKT1 (Figs. 4C and 5A). In corroboration with these observations, the consensus motif for AKT1 (RxxS) was also among the top three motifs present in our dataset (Supplemental Fig. S4C). Our results suggest, therefore, that CD9 may be a key protein responsible for the development of the senescent phenotype in aging mouse brains, via the upregulation of AKT1 and activation of pro-inflammatory pathways.

Fig. 5.

Aging is accompanied by an extensive shift in multiple layers of regulatory processes, from transcription to PTMs.A, changes in the protein and phosphorylation levels of proteins mapped to the cytokine signaling in the immune system. B, Neuroinflammation and glutamatergic signaling. C, CD molecules protein and (D) proteasome protein pathways. E, correlation profile of age-related changes in the mouse brain transcriptome and proteome and. F, changes in the mouse brain proteome and phosphoproteome.

Aging is Associated With Increased Phosphorylation and Accumulation of Neurodegenerative Proteins

Cellular senescence, and, to an extent, aging, are associated with an acute increase in protein accumulation (44). This can be attributed to an increase in protein misfolding (45) or a decrease in proteasome function (46, 47). The loss of proteasome function has been associated with a downregulation in the expression of proteasome proteins, modifications of proteasome subunits, and proteasome complex disassembly (46). Interestingly, we only observed three proteasome subunits that were significantly different between old and adult mice: Psma7, Psma5, and Psmd11. Although statistically significant, all three proteins had a fold-change of less than 1.4-fold, and therefore, the differences in protein levels may not be physiologically relevant. However, our results showed that Psma5 had a significant decrease in the phosphorylation levels of site S56, while Psmda4 had a drastic increase in T250 phosphorylation in old mice. Although the biochemical and biological consequences of Psmd4 T250 phosphorylation have yet to be determined, it is relevant to note that Psmd4 has been implicated in the activation of the PI3K/Akt/GSK-3β signaling pathway (48).

To further investigate which proteins may accumulate in the brain during physiological aging, we performed a correlation analysis between RNA (protein coding) and protein levels, using a previously published RNA seq dataset (49), available at GSE75192. For this comparison, we used the RNA data from 2-months and 24-month-old mice. Our results show that many genes that were identified at both RNA and protein levels were only significantly increased at the protein level, with only Apod and Fabp7 being regulated at both the transcription and protein levels (Fig. 5E). We also observed that most significantly regulated genes at the RNA level were not present in our proteomics dataset (here represented as Log (FC) = 0). It is important to note that there is a greater number of significantly upregulated proteins than downregulated ones, suggesting an accumulation of such proteins in aged mice. Hapln2 (Hyaluronan and Proteoglycan Link Protein 2) stands out as the protein with the highest fold-change increase at the protein level, which was not reflected in its transcription levels. The increases in Hapln2 tissue content have been shown to increase α-synuclein-containing protein aggregates (50) and have been linked with the onset of PD (51). The observation that many genes are significantly increased in their levels at the protein level but not at the transcription level suggests that post-translational regulation such as phosphorylation plays a significant role in the aging process. As phosphorylation information cannot be captured by RNA analysis, this highlights the necessity of integrating multiple omics approaches to gain a comprehensive view of the molecular mechanisms underlying aging and age-related neurodegenerative diseases.

Our phosphoproteomics analysis also sheds light on a new layer of regulatory patterns that take place in aging brains. Our results show that a drastic number of proteins are only differentially regulated at the phospho-level, with a minimal overlap in proteins that are significantly regulated at both their global and phospho levels (Fig. 5F). Dpysl2, also known as Crmp2, was the protein with the highest change in its phosphorylation status, with a 49-fold increase in old mice compared to adult mice. Phosphorylation of Dpysl2 S522 is mediated by Cdk5 (52), and acts as a priming event, triggering the subsequent phosphorylation of nearby sites, namely S518, T514, and T509 by Gsk3 (52). T509 phosphorylation was also significantly increased in old mice, with a 3-fold increase in its phosphorylation levels. Both the kinases that control Dpysl2, Cdk5, and Gsk3b, were among the significantly enriched kinases in our PhosR analysis (Fig. 4B) and were clustered together under modules 2 and 4 (Fig. 4, C and D), indicating potential crosstalk between Cdk5 and Gsk3b signaling pathways, a crosstalk which has been previously suggested to underline AD pathogenesis (52, 53). To investigate if the increase phosphorylation of Dpysl2 at S522 is associated with the development of AD, we used the available phosphoproteomics dataset from human patients suffering from either a low pathology of plaques and tangles (control), mild cognitive impairment accompanied by Aβ pathology, and late-stage AD, reported by Bai et al (54). By comparing the normalized phosphorylation intensities between mild cognitive impairment and late-stage AD relative to control samples, there is an observed increase in phosphorylated Dpysl2 S522 in AD patients (Supplemental Fig. S8C), corroborating our observation in this current study.

Validation of Differentially Abundant Proteins in the Brains of Mice With Different Age Groups

We performed targeted MRM-based proteomics and Western blot analysis of Mtco2, Park7, Dpysl2 to validate our global proteomics and phosphoproteomics results. Mtco2 is a mitochondrial Cytochrome C-Oxidase II (Cox2) and was found to be more abundant in old mice compared to adult mice, showing consistency between untargeted and targeted MRM assays (Supplemental Fig. S5, A–C).

Additionally, we observed a significant increase in Park7 protein levels in the brains of old mice compared to the adult mice. Park7, also known as DJ-1, is an important player in cellular protection against oxidative stress, and its loss of function has been implicated in the early onset of PD. To validate our untargeted MS results, we performed both Western blot and MRM analyses using brain samples from adult and old mice. Our Western blot result corroborated the untargeted MS results, showing a modest age-dependent increase of Park7 (DJ-1) (Fig. 6A). The higher abundance of Park7 was further confirmed by MRM analysis, where we monitored one of its tryptic peptides (GAEEMETVIPVDVMR), detecting a significant increase in Park7 levels in old mice (Fig. 6B). We were able to detect up to five transitions from this peptide across all samples, consistently showing increased signals in old mice compared to adult mice (Fig. 6C). The increased signal of the GAEEMETVIPVDVMR peptide in the MRM analysis was remarkably consistent with observations in our global dataset (Fig. 6D).

Fig. 6.

Global Proteomics Validation.A, Western blot analysis of Park7, phospho Dpsyl2 (pS522) and Actin between adult and old mice. B, MRM analysis of Park7 shows a significant upregulation in the brain of old mice compared to adult ones. C, representative peak of Park7 (GAEEMETVIPVDVMR peptide) in adult vs old mice in the MRM analysis. D, peak area for the GAEEMETVIPVDVMR peptide for each replicate in the targeted analysis compared to the untargeted global proteomics.

Furthermore, a significant increase in the phosphorylation levels of Dpsyl2 at 522 was also confirmed by Western blot analysis using an antibody specific to Dpysl2 at S522 phosphorylation (Fig. 6A). This was further validated by manually evaluating the MS/MS spectra corresponding to the phosphopeptide containing phosphorylated S522 site (Supplemental Fig. S8, A and B).

Discussion

Aging is one of the most common risk factors associated with neurodegenerative diseases including AD and PD, which affect tens of millions of people worldwide (55, 56). A central hallmark of such diseases is the irreversible accumulation of misfolded proteins as insoluble aggregates in different parts of the brain. For example, aggregation of amyloid β (Aβ) in extracellular senile plaque and hyperphosphorylation of tau protein in intracellular neurofibrillary tangles are the primary reasons for the development of AD (57). PD is one of the most common neurodegenerative disorders that is characterized pathologically by selective loss of dopaminergic neurons in the substantia nigra and the accumulation, aggregation, and spread of α-synuclein (58). With the increase in the aging population worldwide, the incidence of different neurodegenerative diseases is also rising and is estimated to double by 2050 (59). Identifying molecular signatures and understanding the underlying mechanisms that lead to the accumulation of pathological aggregates with age could provide insights for early detection and the development of disease-modifying interventions. Here we combine multi-protease digestion followed by quantitative mass spectrometry to explore age-dependent changes in the brain proteome and phosphoproteome and interrogated such changes linked to the processes associated with neurodegeneration. Our multi-protease digestion approach was more effective in identifying a higher number of phosphosites compared to single protease digestion with trypsin (Fig. 2A).

Age has a profound influence on brain phosphorylation with a significant increase in the level of protein phosphorylation in general with age (Fig. 2D). We also found specific age-related changes in proteins associated with neurodegeneration, glutamatergic signaling, and synaptic vesicle cycling. We identified several proteins whose phosphorylation is known to result in protein aggregation and the development of age-related neurodegenerative disorders. We also have identified several new proteins, and their novel phosphorylation sites for the first time. Tau is a microtubule-associated protein that promotes microtubule stability (60). In AD, the formation of neurofibrillary tangles (NFTs) involves tau hyperphosphorylation, a hallmark of pathological features of AD (20, 21). We found tau hyperphosphorylation and changes in dynamics in the brains of old mice compared to the adult and middle-aged mice, suggesting age-related changes may increase the vulnerability of older mice to AD, consistent with the general understanding that age is a major risk factor for neurodegenerative diseases. While involvement of tau in AD has been studied over several decades, how hyperphosphorylated tau and/or NFTs damage neurons in AD is still not well understood. Additionally, identification of new phosphorylation sites on the tau protein, specifically S147 and S148, adds to the comprehensiveness of our data, showcasing the depth of the investigation. This discovery also indicates that there might still be undiscovered phosphosites on tau and other proteins involved in protein aggregation and development of AD or PD, emphasizing the need for continuous expansion of advanced phosphoproteomic studies. While Aβ1-42 aggregation and tau hyperphosphorylation is widely accepted as a hypothesis for the development of AD (61, 62, 63), neuroinflammation has also been proposed recently as a mechanism to explain the complex etiology of AD (64). Upregulation of proteins and phosphoproteins associated with neuroinflammation in the brains of old mice compared to adult and middle-aged mice suggests that the aging process might contribute to changes in the inflammatory response and ultimately influence the progression of AD.

Like AD, PD is also one of the most prevalent neurodegenerative diseases affecting tens of millions of people worldwide. PD is characterized by the formation of Lewy bodies and selective loss of dopamine neurons in the substantia nigra (SN) (50, 65). It results from the accumulation, aggregation, and spread of α-synuclein (66), and several other proteins including tau (67) and ubiquitin (68). Despite the identification of a variety of risk factors, the pathogenesis of PD is still unclear (69). Hapln2 is known to promote α-synuclein aggregation and contribute to neurodegeneration in patients with PD (50). Previously, Hapln2 protein levels were shown to increase with the highest fold change among all upregulated proteins in the SN region of PD patient (65). Hapln2 was also one of the proteins with highest increase in its abundance in the brains of old mice in our study (Fig. 1D), suggesting the vulnerability of old mice to PD. Park7 is a protein with multifunctional properties and its mutation is also linked to early onset of PD (70). Upregulation of Park7 (DJ-1) protein levels in the brains of old mice (Supplemental Fig. S5, A and B) in this study is another indication that old mice are susceptible to PD. Dpysl2 was another protein with the highest fold-change in its phosphorylation level at S522, and this phosphorylation is known to be mediated by Cdk5(71), which triggers subsequent phosphorylation of nearby residues (S518, T514, and T509) by Gsk3 (52). Increased phosphorylation of Dpysl2 at S22 and T509 in old mice (Fig. 5F), coupled with enrichment of Cdk5 and Gsk3b, kinases known to phosphorylate Dpysl2, suggests a potential interaction between Cdk5 and Gsk3b pathways in aging mice. Such interaction between Cdk5 and Gsk3b has been previously demonstrated to disrupt neuronal homeostasis and advance neurodegenerative diseases (71). Phosphorylation of Dpysl2 S522, mediated by Cdk5 (52), acts as a priming event, triggering the subsequent phosphorylation of nearby sites, namely S518, T514, and T509 by Gsk3 (52). Identification of simultaneous upregulation of various protein markers in the brains of old mice brain linked to different neurological disorders may suggest that there may be common pathways and mechanisms involved in age-related changes that contribute to various neuropathology, including AD and PD, depending on the cellular context. This concept aligns with the idea that aging is a major risk factor for many neurodegenerative conditions. Understanding these shared pathways as a function of aging or under different physiological growth conditions may provide critical information and targets for therapeutic and disease-modifying interventions.

Gsk3β and Cdk5 are the two major kinases also involved in abnormal phosphorylation of tau (72). Cdk5 is a candidate for early tau phosphorylation, whereas Gsk3β is associated with the later stage of Tau phosphorylation (72). GSK-3β is a key enzyme associated with tau metabolism and accumulates at the cytoplasm of neurons developing NFTs in AD or other taupathies (73, 74). Cdk5 is also a ubiquitous protein integrating multiple signaling pathways that are important for neuronal viability and function (72). GSK3 is a remarkable kinase that is associated with neuronal and glial hyperphosphorylation of tau resulting in its deposit in AD and contributes to corticobasal degeneration (74). The corticobasal degeneration is a rare, sporadic, and progressive neurodegenerative disorder that affects brain regions that specifically control activities like movement, speech, and memory (74, 75). Increased phosphorylation of GSK3β may suggest its higher activity in old mice leading to higher phosphorylation of tau and other proteins linked to neurovegetative diseases.

While changes in gene expression at the transcriptional level provide valuable insights into cellular processes, they often do not fully capture the complexity of protein regulation. This is particularly true in the context of aging as dysregulation of protein PTMs such as phosphorylation has been implicated in the pathogenesis of age-related diseases. Thus, integrating phosphorylation studies with transcriptomic and proteomic analyses allows for a more comprehensive understanding of the molecular mechanisms underlying biological or pathological aging.

In conclusion, this study sheds light on proteomic and phosphoproteomic responses in aging mouse brains. By using multi-protease digestion and quantitative mass spectrometry, this study generated a comprehensive resource elucidating the changes in proteins and phosphorylation levels in mouse brain associated with biological aging. These results contribute to our understanding of the molecular processes underlying aging and provide valuable information on how such age-dependent changes may contribute to neurodegenerative diseases.

Limitations of This Study

While our study provides valuable insights into proteomic and phosphoproteomic landscapes during aging, we acknowledge several limitations. Our study does not include many other important PTMs, such as glycosylation, methylation, acetylation, SUMOylation, and ubiquitination, which are known to be modulated by aging (76). Although we used multi-protease digestion methods to increase the proteome coverage, we didn’t employ any fractionation strategy to improve the resolution and sensitivity of analyses. Therefore, the depth of the proteomic and phosphoproteomic coverage could be further increased with protein- or peptide-level fractionations and organelle separations. Furthermore, we used a relatively low number of animals per age group. Increasing the sample size could improve the robustness and accuracy of the data, enabling more reliable conclusions to be drawn. The brain is a heterogeneous organ with various cell types with distinct functions and vulnerabilities. Our findings are based on the average response of all cell populations to aging. Understanding how different cell types respond to aging by single cell and spatial omics is important for unraveling age-related changes in brain function and pathology. Finally, our study used only male mice. Validating the results using female mice and investigating sex-dependent and -independent changes in the brain proteome and phosphoproteome during aging are warranted to generalize our conclusions.

Data Availability

All LC-MS/MS raw data files related to proteomics, phosphoproteomics, and MRM analysis are deposited in MassIVE data repository with submission ID: MSV000094441 and have been made public. Annotated spectra for the proteomics and phosphoproteomics results can be found at MS-Viewer. For the proteomics, the spectra can be found using the key: o0f1ay7212. For phosphoproteomics, the spectra for the combined digestions can be found using the key: t9oybhmufm. Phospho results for each independent digestion can be found using the following keys: bfyoxmdsqh (AspN), xnoz8hmyxo (Chymotrypsin), xt4g3syu80 (GluC), mv6ov0sw9p (Trypsin). Note: The raw files labeled as “Adult-6” and “Old-6” have been relabeled in this manuscript as “Adult-1” and “Old-1”, respectively.

Supplemental data

This article contains supplementary datasets.

Conflict of interest

The authors declare no conflicts of interest with the contents of this article.

Supplementary Data

Supplemental figures

Supplementary Fig 1. Global proteomics data quality assessment.A, heatmap representation of all proteins quantified across all digestions. B, scree plot accompanying the PCA plot in Fig 1C. C, correlation plots between replicate mice.

Supplementary Fig 2. Phosphoproteomics data quality assessment.A, heatmap representation of all phosphosites with distinct multiplicities quantified across all digestions.B, violin plots depicting the overall distribution of log₂(intensity) values in the phosphoproteomics dataset. Yellow hue corresponds to adult mice, cyan Mid-Age, and red corresponds to old mice.

Supplementary Fig 3. Differentially regulated phosphosites in proteins involved in the (A) “Cytokine signaling in immune system,” (B) “Neuroinflammation and glutamatergic signaling,” (C) “Interferon signaling,” and (D) “Inflammatory response” pathways.

Supplementary Fig 4. A and B, kinase network accompanying Fig. 4, B–D. C, enriched phosphorylation motif in the phosphoproteomics dataset.

Supplementary Fig 5. A, MRM analysis of Mtco2 shows a significant upregulation in the brain of old mice compared to adult ones. B, representative peak of Mtco2 (VVLPMELPIR peptide) in adult versus old mice in the MRM analysis. C, peak area for the VVLPMELPIR peptide for each replicate in the targeted analysis compared to the untargeted global proteomics.

Supplementary Fig 6. A and B, MSMS spectra for the phosphorylation of MAPT at position S147. C, MS1 intensity of MAPT pS147 phosphosite for each sample analyzed.

Supplementary Fig 7. A and B, MSMS spectra for the phosphorylation of MAPT at position S148. C, MS1 intensity of MAPT pS148 phosphosite for each sample analyzed.

Supplementary Fig 8. A and B, MSMS spectra for the phosphorylation of DPSYL2 at position S522. C, Dpysl2 phsophorylation at S522 at different multiplicities in brain samples suffering from mild cognitive impairment and high AD pathology, available from the study published by Bai et al (54). Bar graphs show the fold-change between mild cognitive impairment and high AD pathology relative to low pathology control samples. x-axes represent different multiplicities for the reported phosphorylated peptide (K.TVTPASSAKTSPAK.Q).