2.1 Animals and reagents

C57BL/6 mice of 8 weeks years old (n = 120) were purchased from Shanghai jiesijie experimental animal Co., Ltd. All animal experiments were in accordance with the requirements of the Animal Experiment Ethics Committee of Fudan University and were approved. Animals were maintained on commercial normal diet (proteins 24.02%, fat 12.95%, carbohydrates 63.03%, 3.44 kcal/g) from Beijing Keao Xieli Feed Co., Ltd. Complete diet composition is listed in Table S1. The environment was kept at 23 ± 2°C, and 12 h light/dark cycle with light on at 06:00 a.m.

Reagents: Sodium dodecyl sulfate (SDS), 1-hydroxybenzotriazole monohydrate (HOBt), sodium borodeuteride (NaBD₄), formic acid (FA), trifluoroacetic acid (TFA), sodium hydroxide (NaOH), super-2,5-dihydroxybenzoic acid (super-DHB), 10 × phosphate buffered solution (PBS), dimethyl sulfoxide (DMSO) and Sepharose CL-4B were purchased from Sigma Aldrich. Acetonitrile (ACN) and ethanol (EtOH) were provided by Merck Millipore. NP-40 and peptide-N-glycosidase F (PNGase F) were from New England Biolabs (Inc.). The FiltrEX™ 96-well Clear Filter Plates with 0.2 μm PVDF Membrane were from Corning. Milli-Q water (MQ) was provided by Milli-Q system.

2.2 Diets and samples collection

Prior to the diet intervention, the food was weighed at the same time every day for 2 weeks, and the average of the reduced food intake was used as the normal intake. At the start of the diet, 12-week-old animals were randomly divided into two groups, AL (ad libitum) group (30 male and 30 female) and CR group (30 male and 30 female). AL group were maintained in groups of five animals; each mouse of CR group was kept in a separate cage to prevent food competition. CR mice were fed once per day 2 h after the lights turned off, and food was placed onto the floor of cage. AL group were fed in the hopper. CR mice get 10 percent less per week until they reached 30 percent calorie restriction, while AL group receiving normal diet ad libitum. All mice were provided with ad libitum access to water. All of the mice were weighed every 2 weeks.

Blood samples were collected from the eye socket of mice by using capillary tubes without sacrificing the mice. Serum samples were collected on 15w, 19w, 23w, 27w, 31w, and 35w between 14:30 and 16:00 p.m. Samples were centrifuged at 3000 rpm for 10 min after 3 h quiescence. The serum was separated and stored at −80 degrees. Serum samples had not been frozen and thawed more than three times.

After 18 weeks of CR diet, 30-week-old male mice (n[AL] = 3 and n[CR] = 4) were euthanized by cervical dislocation between 14:30 and 16:00 p.m. Liver tissues were rapidly harvested and then flash-frozen in liquid nitrogen and stored at −80°C. At 60 weeks old, all of the remaining mice were euthanized in the same way. Blood was obtained before sacrificed by enucleation of the eyeball, and serum was separated by the method described above. White adipose tissues (WAT) were collected and fixed in formalin.

2.3 Serum protein quantification and adipose tissue staining

According to the manufacturer's method, the serum of three time points (15w, 27w, 31w) were selected randomly for protein content quantification by Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific).

Formalin-fixed WAT were embedded in paraffin (three mice per group). All the samples were cut into 5–6 μm paraffin sections and stained with hematoxylin and eosin (H&E). Finally, a Nikon light microscope (Nikon) was used to scanned for the entire field of view of paraffin sections, and images at 20× and 40× resolution were obtained.

2.4 N-glycans release and Bionic Glycome preparation

As for N-glycans release, 10 μL 2% SDS was added to 5 μL of serum and incubated at 60°C for 10 min. After denaturation, adding 10 μL enzymatic hydrolysis buffer (4% NP-40 and 5 × PBS) and 1 μL PNGase F and the reaction was carried out overnight at 37°C.

The Bionic Glycome Method was used to make the internal standards (Qin et al., 2019). We selected mouse serum randomly to make a serum mix. The mix was precipitated with twice the volume of ice EtOH. The supernatant was transferred to a new centrifuge tube; 1% volume of FA was added and incubated at 37°C for 2 h. 2 M of freshly prepared NaBD₄ solution was used for internal standard reduction and reacted at 60°C for 2 h. Moreover, we optimized the purification part from before. We used Sepharose beads instead of the HILIC SPE column of cotton to simplify the procedure without compromising the purification effect. This facilitates the preparation of internal standards in large quantities. The activated Sepharose CL-4B beads were added to the 96-well PVDF membrane, and the samples were added to the wells and incubated for 15 min. After centrifugation, the beads were washed with 200 μL of 95%ACN containing 1%TFA to fully remove impurities. Finally, the glycans were eluted with MQ and concentrated by vacuum centrifugation, dried and stored at −80°C.

2.5 Sialic acid derivatization and purification of N-glycans

Two μL of released N-glycans or internal standard (10 μL of MQ redissolved) was added with 20 μL of derivatization reagent (250 mM EDC and 250 mM HOBt dissolved in EtOH) and reacted for 1 h at 37°C. After cool down to room temperature, 22 μL of -ACN was added and the reaction was carried out at −20°C for 15 min. The precipitated proteins were removed, and the glycans were purified using HILIC-SPE tips made of cotton as reported in previous studies (Reiding et al., 2014). Purified glycans finally eluted in 10 μL of MQ.

2.6 MALDI-TOF-MS analysis

Bruker ultrafleXtreme laser equipped with Smartbeam-II was used to detect the purified glycans in a reflected positive ion mode combined with the support of proprietary software Flexcontrol 3.4 (Bruker Daltonics). Calibration was performed using the Bruker Peptide Calibration Standard II. Each purified sample was mixed with an equal volume of the internal standard and dot 1 μL on a MTP 384 target plate polished steel BC (Bruker Daltonics), with three duplicate dots per sample. After air-dried, 1 μL of substrate (newly prepared 5 mg/mL super-DHB, dissolved in 1 mM NaOH in 50% ACN) was added and allowed to dry at room temperature. The range of m/z ratio was set from 700 to 3500. Laser emission was set at 10,000 laser shots and strafed samples with 100 shots per grating spot by complete random walk at a frequency of 1000 Hz; the laser voltage is set at 68 V. Glycans structures were confirmed by tandem mass spectrometry (MALDI-TOF-MS/MS), and GlycoWorkbench (version 2.1) (https://code.google.com/p/glycoworkbench/) was used for fragmentation spectra analysis.

2.7 Data processing and statistical analysis

The mass spectrograms were analyzed using flexAnalysis (Bruker Daltonics). Through the known m/z ratio (1647.5865 (H4N4F1), 1836.65066 (H4N4Ge1), 2479.88342 (H5N4F1Ge2), 1998.70346 (H5N4Ge1), 2287.77848 (H5N4Ge1Gl1), 2333.82552 (H5N4Ge2), 2622.90054 (H5N4Ge2Gl1)), the instrument was calibrated to reduce mass drift between sample batches. A total of 58 pairs of glycans were identified (S/N > 3). The data processed by flexAnalysis software were further imported into the commercial BioPharma Compass software (Bruker Daltonics) to extract peak signal intensity of each N-glycan spectrum. The relative peak signal intensity (light/heavy peak intensity) of each pair of target peaks was used for quantification, and the average signal intensity of three repeats was included in the quantitative analysis. A total of 58 glycans were identified, of which 45 glycans with coefficient of variation (CV) less than 25% in the reproducibility test were selected for quantification, and 18 derived glycosylation traits were further calculated (Tables S2 and S3). The formula used is shown in Table S4.

To exclude the effect of outliers, we removed outliers larger than Q3 + (1.5) × IQR or smaller than Q1 − (1.5) × IQR, where IQR = Q3 − Q1 (Q1 is 25% quantile and Q3 is 75% quantile). Statistical analysis was performed by IBM SPSS Statistics 20 (IBM Corporation), GraphPad Prism7.0 software (GraphPad Software Inc.), and R 4.1.0.(R Foundation for Statistical Computing) were used to create the diagrams. The difference between glycans was compared using an unpaired Student's t-test. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.

2.8 Validation by isotope-labeled permethylation method

To validate our results, we designed an isotope-labeled permethylation derivatization method for the quantification of glycans. The samples were subjected to methylation derivatization as shown previously (Blomme et al., 2011). Briefly, NaOH pellets and DMSO were grind thoroughly, and 0.2 mL was added to the dried sample purified by Sepharose. Add 0.2 mL of iodomethane, shake thoroughly for 1 h and then add water to quench the reaction. Add chloroform to extract the glycans and remove the aqueous phase. Multiple additions of water were used for extraction to achieve purification. In addition, the mixed serum was purified and derivatized with deuterated iodomethane, and the rest of the processing was the same. After spin-drying and re-dissolving with 50% methanol, the sample and the internal standard were mixed for MALDI-TOF-MS analysis.

2.9 Liver transcriptome analysis

Liver tissues were used for transcriptome analysis. Genome-wide gene expression profiling with 150 bp paired-end reads length was performed on Illumina platform. Briefly, mRNA was purified from total RNA using poly-T oligo-attached magnetic beads. cDNA library constructed by reverse transcription synthesis and was detected by Agilent 2100 bioanalyzer to conform the integrity and total amount. According to the manufacturer's instructions, the clustering of the index-coded samples was done, and the library preparations were sequenced on an Illumina Novaseq platform and 150 bp paired-end reads were produced.

Differential expression analysis was performed by the DESeq2R package (1.20.0). To control the false discovery rate, the obtained p-values were adjusted by Benjamini and Hochberg's method. Gene Ontology (GO) enrichment analysis of differently expressed genes and statistical enrichment in Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were implemented by the cluster Profiler R package. Gene Set Enrichment Analysis (GSEA) is a computational method used to determine whether a given set of genes can show significant consistency differences between two biological states. We use the local version of the GSEA analysis tool (http://www.broadinstitute.org/gsea/index.jsp), and GO and KEGG were used for GSEA independently. The RNA-seq data have been submitted to the Gene Expression Omnibus (GEO) database, and the accession number is GSE219147.

3.1 Calorie restriction caused a generalized downward in serum N-glycome without protein decline

The experimental design is illustrated in Figure 1. A total of 120 C57BL/6 mice of 12 weeks old were randomly divided into two groups and exposed to ad libitum (AL) and calorie restriction (CR) diet. Mice in CR group (n = 60) were kept in separate cages and reduced by 10% per week until they stabilized at 30% calorie restriction after 3 weeks. During this process, food was placed at the bottom of the cage for easy access, and no food was left. The AL group (n = 60) was free to eat every day. The daily food intake of CR mice was 2.6–3.0 g, and energy was 8.944–10.32 kcal.

In order to comprehensively analyze the dynamic changes in the serum N-glycome of mice under long-term CR, 562 serum samples were collected from the eye sockets. We selected several serum samples randomly to mix and prepared Bionic internal standards (IS), so as to better cover the total N-glycome of serum proteins and avoid the lack of IS peaks. After derivatization and purification, internal standards and samples were mixed for MALDI-TOF-MS detection (Figure 1). Representative spectrogram was shown in Figure S1 (some glycans are not indicated). After caloric restriction, the body weight of both male and female mice decreased significantly, and gradually stabilized after reaching 30%CR (Figure 2a,b). Through the hematoxylin and eosin (H&E) stain of WAT, we found that the adipocyte size of WAT tissue of mice was significantly reduced after calorie restriction (Figure 2c,d).

Combining previous studies (Reiding et al., 2016; Rombouts et al., 2016) and tandem mass spectrum data, a total of 58 serum glycans were identified, of which 45 glycans were quantified (CV < 25%). To the best of our knowledge, this is the most N-glycans identified in the mice serum of CR. The accuracy of our quantitation was further verified by the permethylation method. We compared the glycans identified by both methods, two derivatization methods yielded consistent results (Figure S2). However, the number of glycans identified was less due to the lack of recognition of sialic acid-linked isomers by the permethylation method.

The change of each quantified N-glycans during aging in AL and CR groups is shown in the heat map (Figure 3). As shown in the figure, compared with fucosylated glycans, afucosylated glycans were more abundant in the serum N-glycans of mice. In addition, sialic acid linked to α2,3 and α2,6 can be distinguished in this study, and the sialic acid with O-acetylation modification was also detected. The structures of 6 glycans were further identified by tandem mass spectrometry (Figure S3). Most of the identified glycans showed a significant and stable lower level at every time points in CR group. In order to elucidate whether this decrease was caused by changes in protein abundance, serum protein concentrations were assessed for three time points (Figure S4). We found that the amount of protein did not decrease and even increased slightly. Thus, calorie restriction caused a global decrease in serum N-glycome levels without reducing serum protein levels.

3.2 Derived glycosylation traits changed during calorie restriction

In addition to analyzed and compared the total N-glycome at the level of individual glycan structures, derived glycosylation traits were further assessed. We analyzed and compared the level of glycosylation traits derived based on structural similarity between AL and CR group (Figure 4 and Figure S5). It is found that the abundance of three types of glycans (high mannose, hybrid, and complex) was decreased obviously, and glycans with different antenna numbers also decreased in CR group. As for the derived glycosylation traits based on the component of monosaccharide, we found that the level of galactosylation and sialylation of per antenna remained low after CR, while fucosylation was not affected.

We further investigated whether gender difference has an effect on the alteration in derived glycan traits. Considering previous researches have reported that higher levels of core fucosylation were in females than in males in humans and mice (Ding et al., 2011; Han et al., 2020), we analyzed the effect of gender on the fucosylation. We found that the fucosylation level was significantly higher in female mice than that in male for both AL and CR groups (Figure S6a,b). Male had higher levels of afucosylated glycans. The result was consistent with previous findings. After CR treatment, the level of fucosylation trait showed opposite changing trend in male and female mice. Among the 9 fucosylated glycans we analyzed, the male of CR group increased mainly in the two biantennary sialoglycans (H5N4F1Ge1Gl1 and H5N4F1Ge2). Females in the CR group showed a significant decline in most glycans (Figure S6c–k). This suggested that fucosylation is significantly affected by sex, and its change with calorie restriction may be adjusted by sex. In contrast, when we analyzed the effect of gender on the other glycan traits, we found they tended to decrease in both sexes after CR, and interestingly, more in male (Figure S7). Therefore, we considered them as a group for further analysis.

3.3 N-glycans with α2,3/α2,6-linked sialic acid decreased while with O-acetylated sialic acid interestingly elevated

Since the approach we employed allows detection and quantitation of individual N-glycans with α2,3/α2,6-linked sialic acid as well as O-acetylated sialic acid, which are considered to play an important role in biological processes, we also evaluated the effect of CR on the properties of these sialic acid-derived traits, as shown in Figure 5 and Figure S8. Neu5Gc, as the main form of sialic acid in mice serum, decreased in both linkage types after CR. At the same time, the sialoglycans in the form of Neu5Ac were less in the mice serum and also decreased in the CR group. Branched sialic acid showed the same change. Interestingly, the O-acetylated sialic acid was significantly elevated. It is worth mentioning that acetylation of sialic acid has been reported to have the effect of masking sialic acid (Sjoberg et al., 1994; Visser et al., 2021).

Figure 5b shows the changes of the four O-acetylated sialic acid glycans we identified under caloric restriction. We found that the abundance of H5N4Ge1Gl1Ac1 and H6N5Ge1Gl2Ac1 decreased after CR. However, the glycans containing only with α2,6-linked sialic acid but not α2,3-linked sialic acid modifications increased obviously during CR treatment. Previous studies have demonstrated 9-O-acetylation and the regulation of SIAE are related to the formation of α2,6-linked sialic acid by ST6Gal1 (Grabenstein et al., 2021), suggesting a possible regulatory relationship between α2,6-linked sialic acid and O-acetylation in CR.

3.4 Liver transcriptome analysis showed that N-glycan synthesis was significantly affected

Liver is the secretory organ of most serum proteins, participating in the body's energy metabolism, and also is responsible for maintaining metabolic homeostasis. To gain more insights into the regulation of N-glycosylation under caloric restriction, liver transcriptome analysis for both AL and CR group was performed in this study. Samples within the group had a strong Pearson correlation (Figure S9a). Principal component analysis (PCA) of transcriptomic data (Figure S9b) indicated the significant differences between different feeding patterns. Venn diagram was used to show the differential expression of AL and CR groups (Figure S9c). There were 9969 shared transcripts, 614 specifically expressed transcripts in AL group and 513 specifically in CR group. The volcano plot showed that compared with the AL, 672 up-regulated genes and 752 down-regulated genes were found in CR (Figure S9d). Pathways and gene sets with noticeable changes were identified in enrichment analysis (Figure S10), among which processes related to xenobiotics and drug metabolism were up-regulated, including xenobiotic metabolic, exogenous drug catabolic, drug catabolic, and cytochrome P450 pathway, and the uptrend of fatty acid metabolic activity was found, including long-chain fatty acid metabolism, arachidonic acid metabolism, and unsaturated fatty acid metabolism process. In addition, the decreased of negative regulation of proteolysis and peptidase inhibitor activity were observed, which may further promote protein digestion.

In order to understand the changes of genes and pathways related to N-glycan synthesis, 148 key genes in nucleotide sugar synthesis pathway, nucleotide sugar transporters, glycosidases, and glycosyltransferases were analyzed. As shown in Figure 6a, 15 genes increase slightly, while 25 genes significantly decrease obviously (|log₂Fc| > 0.5 and p < 0.05). Among them, N-acetylneuraminic acid synthase (Nans), cytidine monophospho-N-acetylneuraminic acid synthetase (Cmas), and cytidine monophospho-N-acetylneuraminic acid hydroxylase (Cmah), which are the key enzymes involved in the synthesis of CMP-Neu5Gc, were significantly decreased, indicating that donor synthesis of sialic acid was blocked. What's more, a number of mannosyltransferases (Alg2, Alg8 and Alg12), sialyltransferases (st6gal1, st3gal1, st3gal3, st3gal4), and galactosyltransferases (B4galt1) were decreased.

Based on the above findings, we showed several pathways in GO analysis with significant changes about N-glycans biosynthesis (Figure 6b,c). It is found that in Molecular Function (MF), the activities of hexose transferase were down-regulated, and the activities of UDP-glycosyltransferase and sialyltransferase were also affected (Figure 6c). Secondly, the cellular component (CC) was found to be down-regulated in a variety of cellular sites enriched with glycosylation modifications (mainly endoplasmic reticulum and Golgi). However, increased synthesis of acetyl-CoA and increased degradation of long-chain fatty acyl-CoA were observed. This change may be related to the rise in acetylated sialic acid (Figure 6b). According to GSEA enrichment analysis (Figure 6d), mannose transferase activity (GO:0000030) and N-glycan biosynthesis (MMU00510) were remarkably declined after CR treatment, and many related genes were significantly affected.