Integrated Transcriptome, Proteome, Acetylome, and Metabolome Proling of Mouse Liver During Normal Aging

Aging is a complex biological process accompanied by a time-dependent functional decline that affects most living organisms. We aimed to obtain an integrated aging-associated prole of the mouse liver using a multi-omics approach. We performed a combined transcriptome, proteome, acetylome, and metabolome analysis of liver tissues from young and old mice under physiological conditions. Old mice were frequently obese with a fatty liver, and the observed prole changes in different omics were generally moderate. Specically, transcriptome, proteome, and acetylome analyses revealed different patterns in old and young mice, but metabolome analysis did not. Functional enrichment analysis showed that metabolic pathways were broadly altered during normal aging. Notably, the genes, proteins, and metabolites involved in pyrimidine and glutathione metabolisms were signicantly affected in all these four omics. Moreover, we observed increased arachidonic acid metabolism and decreased complement and coagulation cascades in old mice, suggesting an alteration in the immune function during normal aging. We conducted a multi-omics investigation of normal liver aging in mice and generated comprehensive datasets for aging research. Further analysis revealed that impairment of pyrimidine and glutathione metabolisms and immune function may be critical for hepatic aging and may provide targets for aging interventions.


Background
Aging is a complex biological process accompanied by a time-dependent functional decline that affects most living organisms [1]. Aging and aging-associated diseases have brought great suffering and economic burden to individuals and society [2]. Aging-associated alterations include genomic instability, epigenetic alterations, loss of proteostasis, and metabolic manipulation [1,3]. Several interventions, including rapamycin, senolytics, NAD precursors, sirtuin-activating compounds, metformin, exercise, and calorie restriction, can potentially increase the health span and/or lifespan [4]. However, pivotal methods for interventions in aging are still de cient as our global view of normal aging is rather incomplete.
Omics analyses offer the advantage of obtaining an overall pro le of biological processes. Transcriptome [5], proteome [6], metabolome [7], single-cell transcriptome [8], and many other omics analyses have been utilized alone or in combination to study aging and present an increasingly detailed landscape of the aging process in different species. The liver is the biggest metabolic organ in mammals.
Liver aging deserves adequate research, as metabolism and epigenetics are intricately linked and work together to in uence aging [3]. Although the physiological aging of the liver shows relatively modest changes [9,10], omics studies have discovered some signi cant molecular alterations. During aging, DNA methylation in the liver is largely remodeled, can be accelerated by obesity, and may affect downstream gene expression [11,12]. In the aged mouse liver, in ammation is common [13,14], and disruption of metabolic homeostasis and circadian metabolism are observed [15,16]. As aging is a highly complex process spanning gene expression to metabolism, we expect that a comprehensive multi-omics analysis would improve our understanding of liver aging.
In this study, we performed a multi-omics analysis of the transcriptome, proteome, acetylome, and metabolome of livers from 2-month-old and 18-month-old mice under normal physiological conditions. We found that old mice frequently had obesity and a fatty liver. Transcriptome, proteome, and acetylome pro les distinguished young and old livers, but metabolome pro les did not. Further analysis revealed that dysregulation of pyrimidine and glutathione metabolisms and immune function might be critical for hepatic aging, which may provide targets for aging interventions. In addition, our results provide comprehensive multi-omics datasets for future aging research.

Transcriptome pro les distinguish old and young mice livers
We carried out high-throughput RNA-Seq on old and young mice livers to assess transcriptional changes during aging (brief quality information in Table S1). Over 50% FPKM > 1 in at least one group enabled the identi cation of 13,275 transcripts. Replicate correlation calculation revealed that the transcriptomic expression mode between samples was highly similar (Pearson correlation coe cient above 0.92, Fig.  S1A). Hierarchical clustering analysis (unsupervised Euclidean distance) clearly separated livers of the old and young mice ( Fig. 2A), and the same classi cation was supported by principal component analysis (PCA ; Fig. 2B). The aged liver transcriptomes were more individually variable than the young ones, as the Euclidean distance values between old mice were higher ( Fig. 2A). FC and adjusted p-value were calculated for differential expression analysis using the edgeR R package. A total of 1,439 transcripts were assigned as differentially expressed (810 up-regulated and 629 down-regulated in old mice), as shown in the volcano plot (Fig. 2C). We considered transcripts with over 50% FPKM > 1 in young mice, but less than 50% FPKM > 1 in old mice as young-unique, and transcripts with the contrary feature were considered old-unique. Up-regulated transcripts in old mice contained 169 old-unique ones, and down-regulated transcripts contained 130 young-unique ones (Fig. 2D). To predict localization, total, up-, and down-regulated transcripts were submitted to Ingenuity Pathway Analysis (IPA). Compared to the total transcriptome, up-regulated transcripts contained a reduced proportion of nuclear genes and increased plasma membrane genes, and down-regulated transcripts contained a reduced proportion of cytoplasm genes (Fig. S1B). For functional enrichment analysis, up-and down-regulated transcripts in old mice were submitted to KOBAS, a web server, to perform KEGG analysis. Notably, differentially expressed transcripts during aging were enriched in various metabolic pathways, some of which both contained upand down-regulated transcripts (Fig. 2E, 2F). Cytochrome P450, glutathione S-transferase, and UDP glucuronosyltransferase 2 family genes were enriched in both up-and down-regulated pathways (Table  S2). Several lysosome-associated and oncogenic genes, such as Prkca (PKCα) and Wnt5a, were upregulated, whereas multiple histones and innate immune system-associated proteins decreased during aging (Table S2).
Proteomic pro ling of young and old mouse livers Proteome data were obtained using LC-MS/MS. A total of 541,511 spectra were submitted to MaxQuant, and 77,726 (14.35%) were matched to 70,092 peptides (of which 40,796 were unique). This study identi ed 5,831 proteins and quanti ed 4,712 (Fig. S2A). For quality control, we assessed peptide length (Fig. S2B), spectra count per peptide (Fig. S2C), Andromeda score distribution (Fig. S2D), and Pearson correlation coe cient between samples (Fig. S2E). After ID conversion (Ensemble IDs of transcripts and SwissProt IDs of proteins were both converted to Symbol IDs), 4,316 proteins were matched to mRNAs, the expression of which revealed a low positive expression correlation (Spearman correlation coe cient = 0.366829). Both hierarchical clustering analysis (Fig. 3A) and PCA (Fig. 3B) of the quanti ed proteins showed that the hepatic proteome of young mice clearly differed from that of old mice. We performed Signi cance A analysis using Perseus (both sides, Benjamini-Hochberg FDR < 0.05) to de ne differentially expressed proteins and found 114 increased and 81 decreased proteins in old livers. Similar to the transcriptome analysis, we assessed cellular localization of the total, up-, and down-regulated proteins.
Generally, the proportion of cytoplasm genes in the proteome was higher than that in the transcriptome and differentially expressed proteins contained a greater proportion of extracellular proteins than the total proteome (Fig. 3C). KEGG enrichment analysis showed that differentially expressed proteins were especially enriched in metabolic pathways (Fig. 3D, 3E). At the protein level, cytochrome P450 and glutathione S-transferase family proteins were also found in both up-and down-regulated pathways. The levels of cytochrome c oxidase subunits, prostamide/prostaglandin F synthase, and leukotriene-B(4) omega-hydroxylase 2 tended to increase, and those of glutaminase tended to decrease in the mouse liver during aging (Table S3).
Extensive acetylation of histones and metabolic pathway proteins Acetylome data were also obtained using LC-MS/MS. A total of 146,519 spectra were submitted to MaxQuant, and 18,601 (12.7%) were matched to 13,808 peptides (1,690 proteins). A total of 13,606 peptides (6,626 of which unique) in 1,669 proteins were recorded as acetylated, and 5,640 unique acetylated sites were identi ed (Fig. S3A). For quality control, we assessed peptide length (Fig. S3B), spectral count per acetyl peptide (Fig. S3C), mass error (Fig. S3D), and Andromeda score distribution (Fig.  S3E). Each acetylated peptide could contain at most four acetylation sites, but most acetylated peptides contained only one (Fig. S3F). The localization probabilities of acetylation sites ranged from 0 to 1, and every quantile was grouped into one class. Class I sites (localization probability > 0.75) occupied 99.4% of all the acetylated ones (Fig. S3G), and 4,818 of them (1,367 proteins) were quanti able. Pearson correlation coe cient showed a high similarity between samples (Fig. S3H). Hierarchical clustering analysis (Fig. 4A) and PCA (Fig. 4B) of the quanti able Class I acetylation sites revealed that acetylome pro les distinguished livers of old and young mice. Motif analysis showed that glutamate or aspartate was frequently adjacent to an acetylated lysine (Fig. 4C). Acetylated proteins were mostly located in the cytoplasm. Compared to all acetylated proteins, the downregulated ones contained no plasma membrane proteins, and up-regulated ones contained a smaller proportion of nuclear proteins (Fig. 4D). The proteome and acetylome shared 1,262 proteins (Fig. 4E), which were enriched in multiple metabolic pathways and ribosomal proteins (Fig. 4F). Except for metabolic pathways, non-acetylated proteins were mainly enriched in spliceosome, endocytosis, protein processing in the endoplasmic reticulum (ER), and lysosome, which are associated with intracellular macromolecular homeostasis (Fig. 4G). Histones were the major proteins with altered pro les identi ed only in the acetylome (Fig. 4H). However, histone pro les in the transcriptome, proteome, and acetylome analysis showed no consistent alteration (Table S4). The Spearman correlation coe cient was only 0.38168 between proteome and acetylome pro les. We thus performed Signi cance A analysis for acetylome alone to de ne differentially expressed acetylated sites and proteins, nding 60 acetylation sites in 39 proteins increased, and 53 sites in 38 proteins decreased during aging. KEGG analysis showed that proteins containing differentially expressed acetylation sites were predominantly enriched in metabolic pathways and protein processing (Fig. S4A, S4B). Proteins in the enriched metabolic pathways are associated with fatty acid, amino acid, and nucleic acid metabolism, but contained only one cytochrome (cytochrome P450 4A14) whose family members broadly changed in the transcriptome and proteome pro les (Table S5).
Transcriptome, proteome, acetylome, and metabolome pro les show dysregulated pyrimidine and glutathione metabolisms during hepatic aging To identify the key alteration in the mouse liver during aging, we obtained the intersection of enriched pathways (corrected p-value < 0.05) differentially regulated in the transcriptome, proteome, and acetylome. Up-regulated transcripts and up-regulated proteins were both enriched in 16 KEGG pathways, 5 of which contained differentially expressed acetylation sites (Fig. 5A). Down-regulated transcripts and proteins shared 17 KEGG pathways, 7 of which contained differentially regulated acetylation sites ( Fig. 5B). Considering that metabolites re ect the results of complex biological processes, we performed metabolome analysis of the liver samples using LC-MS/MS to further narrow the functional alterations during hepatic aging.
In this study, 242 and 399 metabolites were identi ed in negative-and positive-ion modes, respectively. Pearson correlation coe cients between the four samples for quality control (QC samples) were nearly 1.00, and those between testing samples were all greater or equal than 0.80 (Fig. S5A, S5D), suggesting high-quality metabolome data. PCA showed that QC samples were condensed, but neither negative-or positive-ion mode could distinguish young and old livers (Fig. 6A, 6B). Although OPLS-DA analysis barely separated the two groups ( Fig. S5B, S5E), permutation test results revealed that the metabolite pro les of young and old mouse livers were similar (Fig. S5C, S5F). More metabolites with reduced than increased levels were identi ed by negative-ion mode and metabolites identi ed in positive-ion mode were distributed more symmetrically (Fig. 6C, 6D). We selected metabolites based on MS2 score and FC, 43 of which were up-regulated and 63 down-regulated (Table S6). MetaboAnalyst pathway analysis suggested that up-regulated metabolites were signi cantly (p < 0.05) associated with ribo avin, starch, sucrose, fructose, and mannose metabolisms (Fig. 6E), whereas down-regulated metabolites were signi cantly associated with pyrimidine, glycerophospholipid, and glutathione metabolisms (Fig. 6F). Taking all the four omics results into consideration, dysregulated pyrimidine and glutathione metabolisms are especially notable during aging.

Discussion
In this study, we performed the rst combined transcriptome, proteome, acetylome, and metabolome analyses of young and old mouse livers under physiological conditions. Transcriptome, proteome, and acetylome pro les revealed different expression patterns in old and young mice, in contrast to metabolomic pro les. Although many metabolic alterations were observed in all four omics, pyrimidine and glutathione metabolisms were clearly dysregulated during hepatic aging.
Aging is commonly accompanied by a progressive decline of cellular functions, but the aging liver appears to preserve its function relatively well [9,18,19]. In this study, the aging-related alterations in the four omics were generally mild, although fat accumulation was clearly observed in the liver. Transcriptome analysis identi ed changed levels of cytochrome P450 family members, whereas proteome analysis found alterations of both cytochrome P450 and cytochrome c family member levels. However, cytochromes are involved in many metabolic processes of endogenous or exogenous compounds [20,21]. Although aging is a risk factor for NAFLD [9,22], lipid metabolism may not be the critical pathway during hepatic aging. We, therefore, sought additional lines of evidence to further explore hepatic aging using acetylome and metabolome analyses.
Acetylation is a post-translational modi cation that integrates key physiological processes with gene regulation [23]. Acetylation is sensitive to intracellular metabolic alterations because acetyl-CoA derived from nutrient metabolism, especially lipid-derived acetyl-CoA, is its major carbon source [24]. The metabolome represents the collection of small molecules involved in metabolism, and improvements in the relevant analytical technologies provide signi cant information for biomarker and mechanism analyses [25,26]. In this study, transcriptome, proteome, acetylome, and metabolome analyses each presented characteristic changes during aging. However, alterations in pyrimidine and glutathione metabolisms were especially notable because they were observed in all four omics results in the aging liver.
Down-regulation of nucleic acid metabolism occurs in Caenorhabditis elegans and mouse heart during aging [7,27]. Intermediates of pyrimidine or purine metabolism, such as uridine, cytidine, and hypoxanthine, extend the lifespan of C. elegans [7,28]. Our metabolome data identi ed increased levels of deoxycytidine, D-ribose 5-phosphate, AMP, and adenine, and decreased levels of dihydrofolate, dihydrouracil, deoxyuridine, uracil, cytidine, thymidine, xanthine, AICAR, IMP, and GMP in the aging mouse liver. Disruption of nucleic acid metabolism is associated with increased mutagenesis, genomic instability, and tumorigenesis. Alterations of intracellular deoxyribonucleoside triphosphate (dNTP) pools may impair DNA synthesis and DNA replication, causing cell cycle dysregulation and double-stranded DNA breaks [29][30][31]. During hepatic aging, we observed down-regulation of Cdk1, Cdkn2c, Ccnd2 (cyclin D2), Ccne2 (cyclin E2), Ccnl1 (cyclin L1), Ccnl2 (cyclin L2), and Ccnt2 (cyclin T2), and up-regulation of Inca1 (inhibitor of CDK, cyclin A1 interacting protein 1) in the transcriptome. In the proteome of the aging liver, we observed decreased levels of cyclin-dependent kinase inhibitor 1B (Cdkn1b). These changes in the levels of cell cycle-associated transcripts and proteins indicate that cell cycle dysregulation is likely to happen in the mouse liver during aging [32,33], and may partially contribute to pyrimidine metabolism dysregulation. Moreover, down-regulation of histones, together with dysregulated pyrimidine metabolism, may exacerbate aging-associated genomic instability. Considering that the liver is the major organ for nucleic acid metabolism in mammals, a dysregulated pyrimidine metabolism in the liver is likely to change the levels of nucleic acids in the whole body and accelerate systemic aging.
The decrease in glutathione levels during aging was found decades ago [34,35]. Glutathione de ciency increases the cellular risk for oxidative damage, and glutathione imbalance is observed in a wide range of pathological conditions [36]. In this study, however, metabolome analysis only identi ed decreased levels of glutathione disul de, 5-L-glutamyl-L-alanine, gamma-L-glutamyl-L-valine, and gamma-L-glutamyl-Lglutamic acid in the aging mouse liver. We expect that improvements in metabolome technology may help to identify more metabolites and detect variations in glutathione levels directly in future studies.
This study also revealed that immunological function is altered during hepatic aging. Previous studies reported broadly up-regulated interferon signaling with aging across tissues and species [13,14]. In this multi-omics aging study, transcriptome analysis revealed up-regulated arachidonic acid metabolism, including prostamide/prostaglandin F synthase and leukotriene-B(4) omega-hydroxylase 2, during aging, and metabolome analysis con rmed increased arachidonic acid levels in the aging liver. In addition, the complement and coagulation cascades were decreased in both the transcriptome and proteome. Decreased complement may contribute to decreased hepatic protein synthesis ability and/or in ammation-associated complement consumption. As Xia et al. have summarized, aging-associated adaptive immunity decline is called immunosenescence, and an increase in the body's proin ammatory status with advancing age is called in amm-aging [37,38]. Thus, in amm-aging and immunosenescence may simultaneously occur during hepatic aging.
Mammalian aging is a highly complex process spanning gene expression to metabolism, and thus multiomics analysis can strongly support aging research. However, although each omics analysis provided abundant information, integrated analysis among them is quite a challenge. In this study, we pooled the liver samples and performed proteome and acetylome analyses using an MS2-based TMT strategy and the Signi cance A algorithm. MS2-based TMT can identify peptides precisely but introduce ratio compression [39]. We probably obtained a shortlist of differentially expressed proteins and excluded the interference of individual differences to some extent. We calculated Spearman correlation coe cients and found that the correlation between transcriptome and proteome and between proteome and acetylome were both low. Differences between each omics brought obstacles to reconstruct complete biological processes and signaling pathways but also prompted us to view aging from new perspectives. We expect technical and analytical improvements to increase identi cation accuracy and help future multi-omics analyses. We also hope that more omics methods can be applied and integrated for aging research. It is important to investigate other organs and both sexes in future studies to avoid biases [40] and obtain a comprehensive pro le of liver aging and systematic aging.

Conclusions
In summary, we provide the rst integrated transcriptome, proteome, acetylome, and metabolome pro ling of mouse liver during normal aging, offering a comprehensive data resource for future aging research. The transcriptome, proteome, and acetylome pro les were clearly different in young and old livers, but metabolome pro les were not. Metabolic alterations in the mouse liver during aging were extremely complex, but dysregulated pyrimidine and glutathione metabolisms seemed notable. In addition, we found increased arachidonic acid metabolism and decreased complement and coagulation cascades in the aging mouse liver, suggesting that in ammatory and immune responses change during aging. Hepatic aging may contribute to systematic aging, may be a target for aging interventions, and deserves further exploration.

Animals
Wild-type C57BL/6 male mice were allowed to take food and water ad libitum. Colony rooms were maintained at a constant temperature and humidity with a 12:12 light/dark cycle. All animal protocols were approved by the Animal Care and Use Committee of the Institute of Basic Medical Sciences, Chinese Academy of Medical Sciences, and Peking Union Medical College.

Sample preparation
Two-month-old mice were de ned as young and eighteen-month-old mice as old. Mice were sacri ced by massive bloodletting from the orbital vessels after anesthesia with tribromoethanol. Next, the whole liver was detached from each mouse, immediately dissected, and stored separately. Samples acquired for metabolome, transcriptome, proteome, and acetylome analyses were immediately frozen in liquid nitrogen and transferred to -80 ℃ until use. Samples for H & E and Masson analyses were quickly placed in 4% paraformaldehyde (PFA). Samples for oil red O analysis were appropriately embedded into optimal cutting temperature compound (OCT) and stored at -80 ℃ until use.

Morphological analysis H & E and Masson staining
Liver samples were xed in 4% PFA overnight. Fixed tissues were dehydrated by placing in 75% ethanol for 4 h, followed by 85% ethanol for 2 h, 90% ethanol for 2 h, 95% ethanol for 1 h, absolute ethanol for 30 min twice, ethanol-dimethylbenzene for 5 min, and dimethylbenzene twice for 10 min. Dehydrated tissues were embedded in para n and cut into 4 µm-thick sections. The para n-embedded sections were successively placed in dimethylbenzene twice for 20 min each, absolute ethanol for 10 min twice, 95% ethanol for 5 min, 90% ethanol for 5 min, 80% ethanol for 5 min, 70% ethanol for 5 min, and washed with distilled water.
For H & E staining, hydrated sections were placed in hematoxylin solution for 3-8 min, 1% hydrochloric acid/ethanol differentiation solution for seconds, 0.6% ammonia for seconds, and eosin solution for 1-3 min.
For Masson staining, hydrated sections were processed according to the manufacturer's protocol of the Masson staining kit (Wuhan Goodbio Technology Co., Ltd, G1006).
Stained sections were subsequently transferred into 95% ethanol for 5 min twice, absolute ethanol for 5 min twice, and dimethylbenzene for 5 min twice. Next, the sections were dried and sealed with neutral gum. Pictures were taken with a Nikon Eclipse CI imaging system.

Oil Red O staining
Liver samples embedded in OCT were moved to a freezing microtome and cut into 8 µm-thick sections at -20 ℃. The sections were dried at room temperature for 10 min, xed with 4% paraformaldehyde (PFA) for 15 min, and washed with phosphate-buffered saline (PBS) for 5 min three times. Sections were transferred into oil red O solution (G1016, Goodbio Technology Co., Ltd) for 10 min, followed by 75% ethanol for 2 s, and then washed with water for 1 min. Next, the sections were transferred to hematoxylin solution for 1 min, 1% hydrochloric acid/ethanol differentiation solution for 3 s, and 0.6% ammonia for 3 s, after which they were washed with water. Excess water was removed, and glycerin gelatin was used to seal the sections. Pictures were taken using a Nikon Eclipse CI imaging system (Japan).

RNA sequencing and analysis
RNA isolation, library preparation, and sequencing were performed by Novogene Bioinformatics Technology Co., Ltd (Tianjin, China). Brie y, a total of 3 µg RNA per sample was used as input material for RNA sample preparation. First, ribosomal RNA was removed with the Epicentre Ribo-zero™ rRNA Removal Kit (RZH1046, Epicentre, USA), and rRNA free residue was cleaned up by ethanol precipitation. Subsequently, sequencing libraries were generated using the rRNA-depleted RNA by NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina® (NEBE7770, NEB, USA) following the manufacturer's recommendations. Fragmentation was carried out using divalent cations under elevated temperature in NEBNext First-Strand Synthesis Reaction Buffer (5×). First-strand cDNA was synthesized using random hexamer primer and M-MuLV Reverse Transcriptase (RNase H). Second strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. In the reaction buffer, dTTP was replaced by dUTP. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities.
After adenylation of 3′ ends of DNA fragments, NEBNext Adapter with hairpin loop structure was ligated to prepare for hybridization. To select cDNA fragments of preferentially 150 ~ 200 bp in length, library fragments were puri ed with AMPure XP system (Beckman Coulter, Beverly, USA). Next, 3 µL USER Enzyme (NEB, USA) was used with size-selected, adaptor-ligated cDNA at 37 ℃ for 15 min, followed by 5 min at 95 ℃ before PCR. PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index (X) Primer. At last, products were puri ed (AMPure XP system), and library quality was assessed on an Agilent Bioanalyzer 2100 system. Clustering of the index-coded samples was performed on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer's instructions. After cluster generation, the libraries were sequenced on an Illumina HiSeq 4000 instrument, and 150 base pair and paired-end reads were generated. For quality control, raw data in fastq format were rst processed using Novogene Perl scripts. Clean data were obtained by removing reads containing adapters, reads containing poly-N, and low-quality reads from the raw data. In addition, the Q20, Q30, and GC contents of the clean data were calculated. All downstream analyses were based on the clean data with high quality. RNA sequencing data were deposited in the Sequence Read Archive under the BioProject ID PRJNA609589.
Reference genome and gene model annotation les were downloaded from the Ensembl website (genome: ftp://ftp.ensembl.org/pub/release-97/fasta/mus_musculus/dna/Mus_musculus.GRCm38.dna.primary_assembly.fa.gz; gtf: ftp://ftp.ensembl.org/pub/release-97/gtf/mus_musculus/Mus_musculus.GRCm38.97.gtf.gz). HISAT2 (v2.0.5) was used to build the reference genome index and align paired-end clean reads to the reference genome. Then, StringTie (v1.3.3) was used to assemble the mapped reads of each sample and calculate FPKMs of coding genes. FPKM means fragments per kilo-base of exon per million fragments mapped, calculated based on the length of the fragments, and reads count mapped to this fragment. Transcripts with FPKM values > 1 in over 50% of the samples in either group were considered validated. The edgeR R package (v3.243) provided statistical routines for determining differential expression in digital transcript or gene expression data using a model based on a negative binomial distribution. Transcripts with adjusted p-values < 0.05 were considered to be differentially expressed. Up-and down-regulated transcripts were determined based on the log2 fold-change (FC) (generated by edgeR, old mice/young mice) > 0 or < 0, respectively.

Proteome and acetylome analyses
Protein extraction Samples stored in -80 ℃ were separately ground into a powder after submersion in liquid nitrogen and transferred to individual 5-mL centrifuge tubes. Four volumes of lysis buffer (8 M urea, 2 mM ethylenediaminetetraacetic acid (EDTA), 3 µM trichostatin A, 50 mM nicotinamide, 10 mM dithiothreitol, and 1% Protease Inhibitor Cocktail) were added to the cell powder, followed by sonication for three times on ice using a high-intensity ultrasonic processor. Cell debris was removed by centrifugation at 12,000 × g at 4 ℃ for 10 min. Each supernatant was collected, and protein concentrations were determined with a BCA Kit (P0011, Beyotime, Shanghai, China) according to the manufacturer's instructions. Total protein samples were split into four groups and then pooled into Young 1 (1-7 of young mice), Young 2 (8-14 of young mice), Old 1 (1-5 of old mice), and Old 2 (6-10 of old mice) groups.

Protein digestion
For digestion, 3.7 mg protein was reduced with 5 mM dithiothreitol for 30 min at 56 ℃ and alkylated with 11 mM iodoacetamide for 15 min at room temperature in the dark. The protein sample was then diluted by adding 100 mM triethylammonium bicarbonate (TEAB) to less than 2 M urea. Next, trypsin (V5280, Promega, Madison, WI, USA) was added at a 1:50 trypsin-to-protein mass ratio for the rst digestion overnight and a 1:100 trypsin-to-protein mass ratio for a second digestion for 4 h.

Tandem mass tag (TMT) labeling
After trypsin digestion, peptides were desalted using a Strata X C18 SPE column (Phenomenex, Torrance, CA, USA) and vacuum-dried. Peptides were reconstituted in 0.5 M TEAB and processed according to the manufacturer's instructions for TMT labeling (90068, Thermo Fisher Scienti c, Rockford, IL, USA). Brie y, one unit of TMT reagent was thawed and reconstituted in ACN. Peptide mixtures were then incubated for 2 h at room temperature and pooled, desalted, and dried by vacuum centrifugation.
High pH reversed-phase pre-fractionation of peptides TMT labeled peptides (10% for proteome analysis, and the remaining 90% for acetylome analysis) were fractionated by high pH reversed-phase high-performance liquid chromatography. For proteome analysis, a 300 Extend C18 column (5 µm particles, 4.6 mm inside diameter [ID], and 250 mm length; Agilent) was used. Peptides were separated into 60 fractions by stepwise increases in ACN concentration (8-32% in 60 min, 1 mL/min) at pH 9.0. Total fractions were split into nine groups and then pooled and vacuumdried. For acetylome analysis, a Betasil C18 column (5 µm particles, 10 mm ID, and 250 mm length; Thermo Fisher Scienti c) was used. The gradient and mobile phase times were the same as those used for the proteome. The obtained 60 fractions were split into four groups, pooled, and vacuum-dried.
Liquid chromatography (LC) tandem mass spectrometry (MS) analysis of peptide mixtures A Q Exactive™ HF-X mass spectrometer interfaced with an EASY-nLC 1200 nano ow LC system (Thermo Fisher Scienti c) was used for LC-MS analysis. Samples for proteome and acetylome analysis were separately resuspended in mobile phase A (0.1% formic acid and 2% ACN in water) and loaded onto the EASY-nLC 1200 nano ow LC system at a constant ow rate of 400 nL/min. Mobile phase B contained 0.1% formic acid and 90% ACN in water. For proteome analysis, the following gradient was used: 8%-22% B for 0-38 min, 22%-32% B for 38-52 min, 32-80% B for 52-56 min, and 80% B for 56-60 min. For acetylome analysis, the following gradient was used: 9%-23% B for 0-24 min, 23%-35% B for 24-32 min, 35%-80% B for 32-36 min, and 80% B for 36-40 min. For proteome analysis, a data-dependent strategy was used by rst obtaining MS1 data in the Orbitrap at a resolution of 120,000 (at an m/z ratio of 200 and a maximum injection time of 50 ms for target values of 3e6 ions in the 350-1600 m/z mass range). For the MS2 scan, the top 30 precursor ions (charge state from + 2 to + 5) were selected for fragmentation by higher-energy collision dissociation with a normalized collision energy (NCE) of 28%. A total of 5e4 ions were accumulated over 40 ms as the maximum permitted lling time for each scan. Dynamic exclusion time was set to 30 s to reduce the repeated fragmentation of precursor ions. For acetylome analysis, the data-dependent strategy was also used. MS1 was measured in the Orbitrap at a resolution of 120,000 (at an m/z ratio of 200 and a maximum injection time of 50 ms for target values of 3e6 ions in the 350-1600 m/z mass range). For MS2 scan, the top 20 precursor ions (charge state from + 2 to + 5) were selected for fragmentation by higher-energy collision dissociation with an NCE of 28%. A total of 1e5 ions were accumulated over 100 ms as the maximum permitted lling time for each scan. Dynamic exclusion time was set to 10 s.

Database searches
Raw MS/MS data were analyzed using the MaxQuant search engine (v.1.5.2.8) [41] against the Swiss-Prot Mouse database (updated on June 24, 2019; 17,014 entries) concatenated with the reverse decoy database. Trypsin/P was speci ed as the cleavage enzyme. Two missing cleavages were allowed for proteome analysis, and four missing cleavages were allowed for acetylome analysis. The mass tolerance for precursor ions was set to 20 ppm in the rst search or 5 ppm in the main search, and the mass tolerance of fragment ions was set to 0.02 Da. A carbamidomethyl group on a Cys residue was speci ed as a xed modi cation. Oxidation of Met and protein N-terminal acetylation were set as variable modi cations for proteome analysis, and acetylation of Lys, oxidation of Met, and protein N-terminal acetylation were set as variable modi cations for acetylome analysis. The false discovery rate (FDR) was adjusted to < 1%, and the minimum score for modi ed peptides was set to > 40. The MS-proteomics data were deposited in the ProteomeXchange Consortium via the iProX partner repository [42] under the dataset identi er PXD018003 (subproject ID of proteome: IPX0002001001; subproject ID of acetylome: IPX0002001002).

Data management
Proteins/peptides in the reverse decoy database and potential contaminant database were excluded for both proteomics and acetylomics analyses. In addition, the localization probability of acetylation in the acetylome ranged from 0 to 1. Peptides with localization probabilities of > 0.75 were grouped into Class I and were selected for further analysis. Normalization of the proteome and acetylome data was performed with Perseus (v.1.6.5.0) [43] by dividing the intensity by the median of each group. Signi cance A analysis was performed using Perseus, and a protein/site with Benjamini-Hochberg FDR < 0.05 was considered differentially expressed [41,43]. Up-and down-regulated proteins/peptides were determined as FC (mean values of old mice/mean values of young mice) > 1 or < 1, respectively.

Functional enrichment analysis
Kyoto Encyclopedia of Genes and Genomes (KEGG) [44] enrichment analysis was performed using KOBAS [45]. We chose the hypergeometric test/Fisher's exact test as the statistical method and QVALUE as the FDR correction method. KEGG terms with a corrected p-value of < 0.05 were considered as signi cantly enriched.

Motif analysis
To obtain the sequence characteristics of acetylated peptides, Class I acetylated peptides were submitted to MoMo modi cation motifs (http://meme-suite.org/tools/momo) [46] for motif analysis using the motif-x algorithm.

Metabolome analysis
Sample preparation, LC-MS analysis, peak extraction, and compound identi cation were performed in Dr. Zheng-Jiang Zhu's laboratory, as previously described [47]. Brie y, 20 mg tissue per liver sample was Raw data in wiff format were analyzed using an in-house software program developed in Dr. Zheng-Jiang Zhu's laboratory to perform peak extraction and compound quanti cation. Peaks appearing in over 50% of all samples were considered to represent true feature hits. The mass error tolerance for MS1 matches was set to ± 25 ppm, and the mass error tolerance for MS2 matches was set to ± 35 ppm. MS2 spectral similarity scores were set to range from 0 to 1, and compounds with a score of > 0.6 were further analyzed.
Data management was initially implemented by Shanghai Biotree Biotech Co., Ltd. Brie y, missing values were inserted as half of the minimum value, and the data were normalized by the total ion current. The collated data were entered into the SIMCA14 software program (v14.1, Sartorius Stedim Data Analytics AB, Umea, Sweden) for supervised orthogonal projections to latent structures-discriminate analysis (OPLS-DA), and the rst principal component of variable importance in the projection (VIP) was obtained. FC (mean values of old mice/mean values of young mice) was calculated, and Student's t-test was used to determine p-values. Compounds with both a VIP > 1 and a p-value < 0.05 were preliminarily selected as differentially expressed metabolites and as candidates for pathway analysis.
Compounds identi ed in POS and NEG mode were combined to obtain the overall set. For compounds identi ed in both POS and NEG mode, those with higher MS2 spectrum similarity scores were retained if they displayed consistent changing trends, whereas those with contradictory FC were discarded. Up-and down-regulated compounds were determined as those with FC > 1 or < 1, respectively. An arranged compound dataset was submitted to the MetaboAnalyst website (https://www.metaboanalyst.ca/) [