Hindawi Oxidative Medicine and Cellular Longevity Volume 2021, Article ID 8849328, 17 pages https://doi.org/10.1155/2021/8849328

Research Article Noninvasive Analysis Using Data-Independent Acquisition Mass Spectrometry: New Epidermal That Reveal Sex Differences in the Aging Process

Shirui Chen ,1,2,3,4,5,6 Hui Zhang,1,2,3,4,5,6 Mengting Liu,1,2,3,4,5 Yaochi Wang,1,2,3,4,5 Cong Xin,1,2,3,4,5 Jing Ma,1,2,3,4,5 Xiaodong Zheng,1,2,3,4,5 Xuejun Zhang ,1,2,3,4,5 Liangdan Sun ,1,2,3,4,5 and Sen Yang 1,2,3,4,5,6

1Department of Dermatology, No. 1 Hospital, Anhui Medical University, Hefei 230032, China 2Institute of Dermatology, Anhui Medical University, Hefei 230032, China 3Key Laboratory of Dermatology, Anhui Medical University, Ministry of Education, Hefei 230032, China 4Anhui Provincial Institute of Translational Medicine, Hefei 230032, China 5Inflammation and Immune Mediated Diseases Laboratory of Anhui Province, Hefei 230032, China 6Anhui Ferry Dermatological Institute, Hefei 230032, China

Correspondence should be addressed to Liangdan Sun; [email protected] and Sen Yang; [email protected]

Shirui Chen and Hui Zhang contributed equally to this work.

Received 13 September 2020; Revised 31 January 2021; Accepted 24 February 2021; Published 9 April 2021

Academic Editor: Alexandros Georgakilas

Copyright © 2021 Shirui Chen et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

The development of mass spectrometry has provided a method with extremely high sensitivity and selectivity that can be used to identify biomarkers. Epidermal proteins, lipids, and cornified envelopes are involved in the formation of the skin epidermal barrier. The epidermal protein composition changes with age. Therefore, quantitative proteomic changes may be indicative of skin aging. We sought to utilize data-independent acquisition mass spectrometry for noninvasive analysis of epidermal proteins in healthy Chinese individuals of different age groups and sexes. In our study, we completed high-throughput protein detection, analyzed protein differences with MaxQuant software, and performed statistical analyses of the proteome. We obtained interesting findings regarding ceruloplasmin (CP), which exhibited significant differences and is involved in ferroptosis, a signaling pathway significantly associated with aging. There were also several proteins that differed between sexes in the younger group, but the sex differences disappeared with aging. These proteins, which were associated with both aging processes and sex differences, are involved in signaling pathways such as apoptosis, oxidative stress, and genomic stability and can serve as candidate biomarkers for sex differences during aging. Our approach for noninvasive detection of epidermal proteins and its application to accurately quantify protein expression can provide ideas for future epidermal proteomics studies.

1. Introduction neously participate. Both inherent aging and photoaging affect changes in epidermal proteins such as collagen and Since 1980, many studies have demonstrated that the epider- other extracellular matrix proteins. mis is structurally and biochemically diverse in terms of its In past studies, genomics and transcriptomics have been metabolic state while playing key roles in epidermal barrier used to investigate several biological pathways of aging and function and skin senescence. Epidermal lipid and protein the expression of human aging-related [4–7]. Proteo- compositions and the number of stratum corneum (SC) mics research on skin aging is still scarce. Epidermal proteins layers affect these processes [1–3]. Skin aging is a complex are involved in the skin barrier [8], and proteomics studies physiological process in which several mechanisms simulta- provide more direct information about biological pathways 2 Oxidative Medicine and Cellular Longevity than other types of studies. Epidermal proteins continue to Brilliant blue G-250, 40% ethanol, 10% acetic acid, ethylene- change with age [9]. Therefore, research on epidermal proteins diaminetetraacetic acid (EDTA), and acrylonitrile (ACN) can provide insight into biological markers of aging [10]. were obtained from Wallis Technology, Beijing, China. Many proteins, such as members of the sirtuin family [11], Ultrafiltration membranes (10K MWCO, 1.5 ml, plate) were are involved in aging-related signaling pathways. Among these obtained from Pall Corp. (NY, USA). Pierce C18 pipette tips pathways, the most well known are the mTOR, AMPK, WNT, (10 μl bed), Empore™ C18 47 mm extraction discs (model MAPK, and p53 signaling pathways, which ultimately affect 2215), a ThermoMixer (MS-100), and a CentriVap vacuum the cell cycle, genomic instability, apoptosis, and ultimately concentrator and accessories were obtained from Thermo the outcome of aging [12, 13]. There are also differences in Fisher Scientific (Shanghai, China). The samples were ana- the rates of aging between the sexes that may be related to oxi- lyzed using a Q Exactive high-frequency mass spectrometer dative stress, hormone levels, immunity, and underlying and an UltiMate 3000 high-performance liquid chromatog- mechanisms related to the sex [14, 15]. Men raphy system (Thermo Fisher Scientific, San Jose, CA, USA). have thicker skin than women, but women have thicker sub- cutaneous tissue than men, and men’sskinismoresusceptible 2.1.2. Sample Preparation. The palm side of the forearm was to environmental stress and UV radiation than women’sskin gently wiped with a sterile cotton ball to cleanse it of epider- [16, 17]. Women also generally live longer than men [18], and mal contaminants. A 3:0 ∗ 3:0cmpiece of a 3M medical tape studies have shown that proteins that change with age also dif- was prepared to fully and precisely cover the back of the slide. fer between sexes. Notably, aging does not proceed at a linear The slide was then moved back and forth quickly with even speed; there are obvious inflection points at different ages, and pressure for 2 minutes. Even pressure was applied so that there are differences between men and women [19]. Many the slide was covered. Once the sample was collected, another studies thus far have suggested that the speed of aging differs slide was placed over it to protect the skin sample. To reduce between men and women, so research on sex- and age- sampling variation, the same technician collected samples related intersections will be very valuable. Therefore, we eval- from all volunteers during the study. uated the proteomes of different sexes at different ages and observed whether there were correlations between them. 2.1.3. Protein Extraction and Enzymolysis In this study, we analyzed the differences in the epidermal protein expression profiles of 20 healthy men and women in (1) Protein Extraction. Small pieces (0:5cm∗ 0:5cm)of China with a data-independent acquisition (DIA) method tape/skin samples were cut out of the slides with a sterile [20, 21]. We selected the age categories with consideration blade and then placed into corresponding 1.5 ml centrifuge of the significant changes that occur in the skin during men- tubes according to the sample numbers. An appropriate opause [22]. Noninvasive techniques were used to obtain amount of L3 lysis buffer without SDS was added, and cock- epidermal proteins and to explore proteins with differences tail containing EDTA was added to a final concentration of and associations at the levels of both age and sex. To better ×1. The samples were placed on ice for 5 min, and DTT identify aging-sex correlates, we performed statistical analy- was added to a final concentration of 10 mM. The samples ses and linked information about these changes to aging- were soaked overnight and then centrifuged at 25,000 × g related signaling pathways. The highlighted proteins may and 4°C for 15 min. The supernatant was obtained, and be sex-specific markers of aging, and their identification DTT was added to a final concentration of 10 mM. The sam- may help elucidate the reasons for the difference in the rate ples were incubated in a 56°C water bath for 1 hour, after of aging between men and women. which iodoacetamide (IAM) was added to a final concentra- tion of 55 mM. The samples were placed in a dark room for ° 2. Materials and Methods 45 min and then centrifuged at 25,000 × g at 4 C for 15 min to obtain the supernatant, which contained the proteins. 2.1. Study Participants. Skin samples were collected from 9 healthy Chinese males and 11 healthy Chinese females, none (2) Protein Extraction and Quality Control. The protein con- of whom had experienced excessive light exposure. The par- centrations were measured using the Bradford method [23] ticipants were divided into two groups according to age: the with acceptable enzymatic efficiency, and it was determined 20–30 y group (3 males and 7 females) and the 55–70 y group that the extracted protein content was sufficient. For the (6 males and 4 females). Individuals were excluded from the Bradford quantitative assay, protein standards (0.2 μg/μl study if they had an allergy to tape, were long-term outdoor BSA; 0, 2, 4, 6, 8, 10, 12, 14, 16, and 18 μl) were added to wells workers, or had a skin disease or other systemic condition A1 through A10 of 96-well MICROLON ELISA plates, and involving skin problems. The participants did not use topical then, pure water (20, 18, 16, 14, 12, 10, 8, 6, 4, and 2 μl) was emollients or other cosmetic products for 24 hours prior to added. Then, 180 μl of Coomassie Brilliant blue G-250 quan- the experiment. This study was conducted in accordance titative work solution was added to each well. A linear stan- with the recommendations of the Medical Ethics Committee dard curve was prepared based on the optical density at of Anhui Medical University, and informed consent was 595 nm (OD595) and the protein concentration. The protein obtained from all enrolled subjects. sample solutions were diluted several times, 180 μlofthe quantitative solution was added to 20 μl of protein solution, 2.1.1. Materials. Sodium dodecyl sulfate (SDS), L3 lysate and the OD595 was read. The protein concentration of each without SDS, trypsin, dithiothreitol (DTT), Coomassie sample was calculated according to the standard curve and Oxidative Medicine and Cellular Longevity 3 the sample OD595. The purity of the extracted proteins was After liquid chromatography, the peptides were separated verified by SDS-PAGE and Coomassie Brilliant blue staining. by nanoelectrospray ionization (ESI) and detected in a DDA mode with a Q Exactive HF (Thermo Fisher Scientific, (3) SDS-PAGE. For each sample, 30 μg of protein was added San Jose, CA) tandem mass spectrometer. The main param- to an appropriate amount of loading buffer. After thorough eter settings were as follows: the ion source voltage was mixing, the proteins in buffer were heated at 95°C for 1.6 kV, the primary mass spectrometer range was 5 min, centrifuged at 25,000 × g for 5 min, and placed into 350~1,500 m/z, the resolution was 60,000, the MS2 starting the wells of a 12% SDS polyacrylamide gel. After electropho- m/z was fixed at 100, and the resolution of the second-stage resis, Coomassie Brilliant blue staining was conducted for 2 mass spectrometer was 15,000. The screening criteria for hours. An appropriate amount of decolorization solution the precursor ions in the MS2 scans were a charge of 2+ to (40% methanol and 10% acetic acid) was added to each sam- 7+ with a peak intensity of over 10,000 and a ranking in the ple, and the samples with decolorization solution were placed top 20. The main parameter settings for DIA mode detection in a shaker 3 to 5 times for 30 min each. were as follows: the ion source voltage was 1.6 kV, the MS1 scanning range was 350–1,500 m/z with a resolution of (4) Proteolysis. Trypsin (2.5 μg) was added to 100 μg of pro- 120,000, and the 350–1,500 Da range was divided into 40 tein for each sample at a protein : enzyme ratio of 40 : 1, and windows for fragmentation and signal acquisition. Higher- the protein was hydrolyzed at 37°C for 4 hours. Trypsin energy collisional dissociation (HCD) was used for the was added once more according to the above ratio, and DDA and DIA modes of mass spectrometric detection of enzymolysis was continued for 8 hours at 37°C. The enzy- ion fragmentation patterns. The fragmented ions were matically digested peptides were desalted with a Strata-X detected in an Orbitrap. The dynamic exclusion time was column and extracted under vacuum. With this method, set to 30 s. The automatic gain control (AGC) settings were we extracted sufficient amounts of proteins with qualified as follows: level 1: 3E6 and level 2: 1E5. enzymatic efficiency. 2.2. Analysis of Epidermal Proteomics Data. The DDA data 2.1.4. High-pH Reverse-Phase Separation. A pooled sample from the machine were identified using the Andromeda was created from 10 μg of each sample, and 200 μg of the search engine integrated with MaxQuant [24], and Spectro- pooled sample was mixed with 2 ml of mobile phase A (5% naut was then used to build a spectral library with the results. acetonitrile, ACN, pH 9.8) and diluted into a Shimadzu LC- The resulting data were reviewed using the UniProtKB/S- 20AB liquid chromatography system. The samples were sep- wiss-Prot Homo sapiens proteome database. For large-scale arated in a liquid phase on a 5 μm 4:6 × 250 mm Gemini C18 DIA data, after constructing the spectral library information, column. Elution was conducted at a flow rate of 1 ml/min convolutional extraction of the data was completed using with the following gradient: 5% mobile phase B (95% ACN, Spectronaut, and data analysis and quality control were com- pH 9.8) for 10 min, 5% to 35% mobile phase B for 40 min, pleted using the mProphet algorithm, resulting in reliable 35% to 95% mobile phase B for 1 min, mobile phase B which protein quantification results. Ontology (GO) and lasted for 3 min, and equilibration at 5% mobile phase B for Kyoto Encyclopedia of Genes and Genomes (KEGG) path- 10 min. The elution peaks were monitored at 214 nm, and way functional annotations were performed. Based on the one fraction was collected every minute. The samples were high-quality quantitative results, we searched for differen- combined with the chromatogram elution peaks to obtain tially expressed proteins (DEPs) between groups. 10 fractions, which were then extracted by freezing. MaxQuant was used to complete the identification of the DDA data in order to create the spectral library for subse- 2.1.5. High-Performance Liquid Chromatography. The quent DIA data analysis. During this procedure, the original extracted peptide samples were redissolved (centrifuged with offline data were used as the input data, the corresponding mobile phase A (2% ACN and 0.1% formic acid (FA)) at parameters and database were configured, and then, the 20,000 × g for 10 min), and the supernatant liquid was sam- identification and quantitative analysis were performed. pled. Separation was performed with a Thermo UltiMate The identification information that met the criterion of a false 3000 UHPLC. Each sample was first injected into a trap col- discovery rate ðFDRÞ ≤ 1% was used to build the final spectral umn for enrichment and desalting before being run through library. The offline DIA data were analyzed using Spectronaut a self-loading C18 column (150 μm I.D., 1.8 μm pore size, [25], and iRT peptides were used to correct the retention time. 25 cm column length) coupled to tandem mass spectrome- Spectronaut integrates the mProphet scoring algorithm, ters. The samples were separated at a flow rate of 500 nl/min which can accurately reflect the degree of matching of isolated through the following effective gradient: 0–5 min, 5% mobile pairs. Then, based on the target-decoy model applied by phase B (98% ACN, 0.1% FA); 5–160 min, linear increase in SWATH-MS, the false-positive rate was controlled at 1% mobile phase B from 5% to 35%; 160–170 min, 35% to 80% FDR, which yielded significant quantitative results. mobile phase B; 170–175 min, 80% mobile phase B; and 176– This procedure was used to preprocess the data accord- 180 min, 5% mobile phase B. The output of the liquid chro- ing to the set comparison group, and a significance test matograph was connected directly to the mass spectrometer. was then performed based on the model. After that, the DEPs were screened with the criteria of a fold change (FC) value 2.1.6. Data-Dependent Acquisition (DDA) and Data- ≥ 1:5 and a p value < 0.05 (the top 10 DEPs are listed in Independent Acquisition (DIA) Mass Spectrometry Analyses. Table 1). The R package MSstats [26] from the Bioconductor 4 Oxidative Medicine and Cellular Longevity

Table 1: Top 10 differentially expressed proteins.

Upregulated Downregulated Protein ID log2FC p value Protein ID log2FC p value LG3BP 0.700 0.001 ABRX2 −0.750 0.001 HBA 3.606 0.001 PLD3 −0.698 0.003 RENR 1.014 0.001 VTI1B −0.587 0.004 APOA2 3.069 0.001 VAT1 −0.783 0.004 CLUS 1.667 0.001 KRT35 −1.645 0.004 Old vs young SG1D2 2.661 0.002 BAF −2.492 0.005 SEM7A 1.414 0.002 ECM1 −0.631 0.009 B2MG 1.590 0.002 K1C15 −0.600 0.010 S10A4 0.734 0.002 MYH9 −0.907 0.011 HBB 3.462 0.003 PSB7 −0.678 0.012 resource library was used to complete differential analysis of different age groups. Only when a p <0:05 between age the DEPs. groups was obtained were independent sample t-tests performed between the male and female subgroups within 2.3. Screening for DEPs. After the raw data were median- the age groups. Both the p value and the average value of normalized by the software, the resulting data were selected protein expression were recorded for the different sexes. fi for follow-up analysis. DEPs were identi ed through sym- GraphPad Prism 8.3.0 was used to create violin plots of the metrical scatter plot screening. The dots represent the relative proteins with significant differences in the t-tests in order 2 mean expression values (normalized and log transformed) to better visualize the distribution of the data and the differ- of the proteins in the two groups, and the log2ðFCÞ values ences in expression between sexes. GraphPad was also used were calculated using the R package MSstats. The dashed line to generate receiver operating characteristic (ROC) curves in the graph represents the threshold line at ∣log2ðFCÞ ∣ =1:5 and heatmaps for selected DEPs. . Thus, the volcano plot of DEPs was produced with the aver- age protein expression values, log2ðFCÞ values, and p values. 3. Results The color of the dot for each protein indicates the signifi- cance of the p value, with the blue to red gradient indicating 3.1. Protein Quantification and Data Quality Control. In this p values ranging from nonsignificant to significant. The project, mass spectrometry data were collected for 20 samples graph shows information about the gradient change. We using a Q Exactive HF instrument in the DIA mode. Peptide performed protein-protein interaction network analysis on quantification and protein quantification were completed all the DEPs and screened the proteins with more than 10 using Spectronaut and MSstats software. The specific quanti- nodes of neighboring proteins in the protein relationship tative information for each sample is shown in Figures 1(d) network. The proteins with the most significant differences and 1(e). In total, we identified 1,318 proteins in the 20 sam- in log2ðFCÞ values and p values were also selected (Table 2). ples, which were further analyzed with MSstats software, A gene-level network analysis was conducted for the selected resulting in 1,270 proteins. The quality of the data was ffi most promising candidate proteins, and the biological path- assessed by analysis of intragroup coe cients of variation, ways in which they were enriched were investigated by using principal component analysis, and quantitative sample corre- ’ ffi NetworkAnalyst 3.0 inline tool [27, 28], the value of the lation analysis. We calculated Pearson s correlation coe - interaction relationship between genes retained the interac- cients for the expression levels of all protein expressions ffi tion relationship with experimental evidence, and the confi- between the two groups and displayed these coe cients as dence score cutoff is set to 900. In addition, to observe the a heat map, as shown in Figure 2(b). We also acquired correlations between our differential expression analysis data DDA mass spectrometry data from the samples and then fi and other aging-related proteomics results in external data- used MaxQuant to complete library identi cation in order bases, we compared the DEP data with data from the Human to obtain nonredundant, high-quality MS/MS spectral infor- fi Aging Gene Database (https://genomics.senescence.info/ mation for subsequent quanti cation in the DIA mode. The genes/index.html) (last updated on February 9, 2020) and spectral library contained fragment ion intensities and reten- from plasma proteomics studies related to aging changes in tion times, which characterized the peptide spectral peaks. healthy populations [29]. With regard to statistical information for the peptides and proteins in the spectral library, we obtained the peptide dis- 2.4. Independent Sample t-Tests for within-Age Group tribution, protein amount distribution, and protein coverage Differences. Information was compiled on the DEPs identi- distribution (Figures 1(a)–1(c)). fied in the age groups in the DIA experiment. To reduce error caused by NA values in the experimental data, SPSS 25.0 soft- 3.2. Identification of 95 DEPs in Aging via Proteomics ware was used to perform independent sample t-tests for the Analysis. In order to obtain quantitative proteomic maps of Oxidative Medicine and Cellular Longevity 5

Table 2: Functional role of the 21 proteins.

Protein UniProt Gene symbol Full name Function ID Hemoglobin subunit HBB P68871 HBB This protein includes iron ion binding and oxygen binding beta Hemoglobin subunit Involved in oxygen transport from the lung to the various peripheral HBA P69905 HBA1, HBA2 alpha tissues May stabilize the HDL (high-density lipoprotein) structure by its APOA2 P02652 APOA2 Apolipoprotein A-II association with lipids and affect the HDL metabolism Synaptic vesicle Possesses ATPase activity (by similarity) and plays a part in calcium- VAT1 Q99536 VAT1 membrane protein regulated keratinocyte activation in epidermal repair mechanisms Galectin-3-binding Promotes integrin-mediated cell adhesion and may stimulate host LG3BP Q08380 LGALS3BP protein defense against viruses and tumor cells Keratin, type I KRT35 Q92764 KRT35 A member of the keratin gene family cuticular Ha5 BRISC complex Required for normal induction of p53/TP53 in response to DNA ABRX2 Q15018 ABRAXAS2/FAM175B subunit Abraxas 2 damage Barrier-to- Plays fundamental roles in nuclear assembly, chromatin organization, BAF O75531 BANF1 autointegration factor gene expression, and gonad development Glyceraldehyde-3- The product of this gene catalyzes an important energy-yielding step G3P P04406 GAPDH phosphate in carbohydrate metabolism dehydrogenase This protein is associated with the transport of cholesterol and the APOA1 P02647 APOA1 Apolipoprotein A-I major protein component of high-density lipoprotein (HDL) This protein produces haptoglobin, which encodes a protein with HPT P00738 HP Haptoglobin antimicrobial activity against bacteria This protein is involved in several basic biological events such as cell CLUS P10909 CLU Clusterin death, tumor progression, and neurodegenerative disorders Alpha-1-acid This protein is classified as an acute-phase reactant and may be A1AG1 P02763 ORM1 glycoprotein 1 involved in aspects of immunosuppression Alpha-2- This gene can inhibit inflammatory cytokines and thus disrupts A2MG P01023 A2M macroglobulin inflammatory cascades CYTC P01034 CST3 Cystatin-C This protein was shown to have an antimicrobial function Fibrinogen gamma The protein encoded by this gene is the gamma component of FIBG P02679 FGG chain fibrinogen CERU P00450 CP Ceruloplasmin A glycoprotein with ferroxidase activity Fibrinogen alpha This gene encodes the alpha subunit of the coagulation factor FIBA P02671 FGA chain fibrinogen ANT3 P01008 SERPINC1 Antithrombin-III A member of the serpin superfamily and a plasma protease inhibitor FIBB P02675 FGB Fibrinogen beta chain The protein encoded by this gene is the beta component of fibrinogen Vitamin D-binding Involved in vitamin D transport and storage and scavenging of VTDB P02774 GC protein extracellular G-actin the stratum corneum of keratinocytes derived from elderly proteins with more than 10 nodes of protein interactions and young populations, we used data extracted from Max- (Figure 3(a)). We compared our DEPs with 307 human Quant and Spectronaut convolutions and applied bioinfor- aging-related genes from the GenAge database and identified matics analysis using 1% FDR analysis to identify 95 unique 6 genes that intersected with the DEPs, namely, CDC42, proteins that were differentially expressed between the age CLU, A2M, LMNA, VCP, and AIFM1 (Figure 4(c)). In a groups. The supplementary tables show important informa- proteomics study on the senescence-associated secretory tion about the identified proteins, including their UniProtKB phenotype [30], serine protease inhibitors (SERPINs) were accession numbers, IDs, protein and gene names, adjusted p shown to be significantly associated with age changes in values, and associated functions. The proteins that were human cohorts and to be potential biomarkers. In the significantly differentially expressed according to age induced aging model in that study, the secretion levels of the included 38 downregulated proteins and 57 upregulated pro- proteins SERPINE1, SERPINF1, SERPINF2, SERPING1, and teins (Supplementary Table 1 and Supplementary Table 2). SERPINH1 were significantly increased. In contrast, In an interaction analysis of the 95 DEPs, we counted the SERPINA3, SERPINB1, SERPINB6, and SERPINC1 were 6 Oxidative Medicine and Cellular Longevity

Count Count Count 300 382 400 273 258 258 750 714 300 600 218 218 2 0 200 8 208 208 300 282 2 200 177 177 1 177 177 1 9 2 500 3 226 226 439 439 400

7 200 7 177 1 177 170 200 170 101 101 7 2

1 100 Protein count Protein 217 100 217 2 112 112 1 77 77 7 71 71 250

7 200 4 2 Protein count Protein 0 77 7 77 74 7 74 129 129 Protein count Protein 50 100 50 56 56 1 6 6 100 61 6 36 3 36 3 36 36 21 2 8 37 37 3 2 18 18 12 12 3 1 0 0 0 3 1 8 9 1 2 3 4 5 6 7 10 ≥ 11 (0,10) 0-10% (60,70) (70,80) (80,90) (10,20) (20,30) (30,40) (40,50) (50,60) 10-20% 20-30% 30-40% 40-50% 50-60% 60-70% 70-80% 80-90%

Unique peptide number (90,100) (100,lnf) 90-100% Protein mass (kDa) Protein coverage (kDa) (a) (b) (c) Proteins identifed per sample Peptides identifed per sample 6000

1200 6289 6162 1218 1212 5610

1000 1097 5000 5275 5223 5162 5186 1041 1037 1041 1028 1038 5146 998 5098 998 5030 5032 5058 990 1014 4975 975 1003 4980 976 963 962 4934 4863 4852 4818 926 914 912 4506 800 4000 4494 600 3000 400 2000 Protein number Protein 200 1000

0 peptide number Unique

1 2 3 4 5 6 7 8 9 0 10 11 12 13 14 15 16 17 18 19 20 Sample 1 2 3 4 5 6 7 8 9 Sample 10 11 12 13 14 15 16 17 18 19 20 Young Old Young Old (d) (e)

Figure 1: (a) Identification of DDA profiles for subsequent DIA quantification: unique peptide distribution with the horizontal axis being the only matching peptide number per protein and the vertical axis being the number of proteins; (b) protein mass distribution with the horizontal axis being the protein mass interval (Kilodalton) and the vertical axis being the corresponding protein number; (c) protein coverage distribution with the horizontal axis being the percentage protein coverage interval and the vertical axis being the number of proteins. (d, e) Unique proteins and peptides were counted in each sample collected by Q Exactive HF in the DIA mode. significantly upregulated with age in our study. In an age- relation to aging included ferroptosis and apoptosis, both of associated proteomics study in a healthy population [29], 217 which are cell growth- and death-related metabolic pathways. age-associated proteins were present, 9 of which intersect Among the six proteins intersecting with GenAge data, AIFM1 with the DEPs in our study: ANXA2, S100A4, B2M, FGA, and LMNA are involved in apoptosis. In the gene-level FGB, FGG, SERPINA3, CST3, and HP (Figure 4). In the network analysis of the 21 most promising proteins we previous study, SERPINA3 and SERPING1 were positively screened (Figures 3(b)–3(e)), the significantly enriched correlated with age, while SERPINF2 was negatively pathways were the complement and coagulation cascades, correlated with age, which is consistent with our results. platelet activation, ferroptosis, apoptosis, autophagy, and PPAR signaling pathways. Among them, the PPAR signaling 3.3. Functional Analysis of the DEPs. Through DIA mass pathway may be closely related to the aging process [31]. (For spectrometry, we identified a total of 1,270 unique proteins, more detailed gene node information, please refer to of which 95 were differentially expressed in the different Supplementary Tables 4, 5, 6, 7, and 8 and the following age groups. The 95 identified proteins were analyzed using website: https://www.networkanalyst.ca/NetworkAnalyst/ the GO and KEGG databases (the functional classifications faces/Share?ID=_9qljssi8x.) A protein of particular interest of the 95 DEPs are shown in Supplementary Table 3). The was ceruloplasmin (CP), which had 16 neighboring proteins in functional pathways into which the DEPs were classified the protein-protein interaction network. CP is closely are shown in Figure 5(a) including the cellular process, associated with aging [32, 33], and the biological metabolic environmental information processing, genetic information pathway in which it participates is the well-known ferroptosis processing, human disease, metabolism, and organismal pathway (Figure 6), a signaling pathway that is closely system pathways. We analyzed the DEPs with the KEGG and associated with aging. Our age-associated proteomics results found that the most enriched pathways included platelet are consistent with the results obtained in other age-associated activation, complement and coagulation cascades, and the proteomics studies with independent cohorts (Supplementary VEGF signaling pathway. Among the DEPs, BANF1 was Tables 1 and 2 at 10.1038/s41591-019-0673-2). found to be responsible for the chromatin structure and dynamics, replication, recombination, and repair. Energy 3.4. Potential Candidate Biomarkers of Sex Differences in the production and transfer are also entry points for the aging Aging Process. We compared the proteins that changed most process, and LDHB and HBA1 were enriched in these significantly under the influence of two factors, age and sex, pathways. The signaling pathways that we focused on in according to a dual statistical threshold defined by the p value Oxidative Medicine and Cellular Longevity 7

Sample

Protein extraction

Extraction quality control

Proteolysis

Peptide separation afer sample mixing Individual sample

LC-MS/MS (DDA model) LC-MS/MS (DIA model)

DDA data

DIA data MaxQuant Protein database

Spectronaut Unqualifed

Quality control Qualifed MSstats diference analysis

Signifcant analysis of diferentially expressed proteins KEGG annotations Linear change analysis of pathway annotations diferentially expressed proteins (a)

Pearson_value

Control_10 0.92 0.909 0.923 0.93 0.942 0.905 0.909 0.886 0.918 0.924 0.75 0.948 0.94 0.937 0.939 0.943 0.938 0.911 0.938 1 1.0

Control_9 0.897 0.928 0.927 0.925 0.944 0.887 0.9 0.893 0.903 0.905 0.775 0.938 0.953 0.949 0.932 0.938 0.957 0.937 1 0.938

Control_8 0.887 0.892 0.885 0.896 0.924 0.867 0.871 0.896 0.885 0.894 0.743 0.888 0.917 0.90