Mouse tissue glycome atlas 2022 highlights inter-organ variation in major N-glycan profiles

Abstract

This study presents "mouse tissue glycome atlas" representing the profiles of major N-glycans of mouse glycoproteins that may define their essential functions in the surface glycocalyx of mouse organs/tissues and serum-derived extracellular vesicles (exosomes). Cell surface glycocalyx composed of a variety of N-glycans attached covalently to the membrane proteins, notably characteristic "N-glycosylation patterns" of the glycocalyx, plays a critical role for the regulation of cell differentiation, cell adhesion, homeostatic immune response, and biodistribution of secreted exosomes. Given that the integrity of cell surface glycocalyx correlates significantly with maintenance of the cellular morphology and homeostatic immune functions, dynamic alterations of N-glycosylation patterns in the normal glycocalyx caused by cellular abnormalities may serve as highly sensitive and promising biomarkers. Although it is believed that inter-organs variations in N-glycosylation patterns exist, information of the glycan diversity in mouse organs/tissues remains to be elusive. Here we communicate for the first-time N-glycosylation patterns of 16 mouse organs/tissues, serum, and serum-derived exosomes of Slc:ddY mice using an established solid-phase glycoblotting platform for the rapid, easy, and high throughput MALDI-TOFMS-based quantitative glycomics. The present results elicited occurrence of the organ/tissue-characteristic N-glycosylation patterns that can be discriminated to each other. Basic machine learning analysis using this N-glycome dataset enabled classification between 16 mouse organs/tissues with the highest F1 score (69.7-100%) when neural network algorithm was used. A preliminary examination demonstrated that machine learning analysis of mouse lung N-glycome dataset by random forest algorithm allows for the discrimination of lungs among the different mouse strains such as the outbred mouse Slc:ddY, inbred mouse DBA/2Crslc, and systemic lupus erythematosus model mouse MRL-lpr/lpr with the highest F1 score (74.5-83.8%). Our results strongly implicate importance of "human organ/tissue glycome atlas" for understanding the crucial and diversified roles of glycocalyx determined by the organ/tissue-characteristic N-glycosylation patterns and the discovery research for N-glycome-based disease-specific biomarkers and therapeutic targets.

Figure 1

A strategy for the construction of the mouse glycome atlas. (a) An image representation for the glycoblotting-based solid-phase glycan enrichment analysis for the construction of mouse tissue glycome atlas. (b) In the present study, sixteen frozen organs/tissues in addition to the serum and serum-derived exosomes were employed for the general protocol enabling one-pot rapid and efficient glycan enrichment and subsequent chemical manipulations on the polymer-solid beads. Pie charts represent ratio of the terminal sugar residues of each organ revealed by glycotyping analysis^,,, of the present results. (c) General biosynthetic pathway of major N-glycans showing maturation from high mannose-type to hybrid-type and complex-type glycoforms in the posttranslational modification of mammalian proteins. This schematic representation does not involve the biosynthesis of minor glycoforms having GalNAc terminals, Lewis type antigenic structures, lactosaminoglycans (LacNAc repeats), sulfated sugar moieties, and so on. (d) The sugar compositions of N-glycans assigned based on the reported structures may often predict multiple candidates of the isomeric glycoforms found in the Expasy GlycoMod Tool (https://web.expasy.org/glycomod/ and https://glyconnect.expasy.org/browser/compositions/453).

Figure 2

Heatmap representation of the absolute expression levels (pmole/100 mg total protein) of N-glycan structures (103 glycoforms listed in Fig. S3) identified in the mouse organs/tissues, serum, and serum-derived exosomes. The MALDI-TOFMS spectrum is a result for one of the five samples of the brain samples tested. The heatmap shows an absolute glycan level estimated from the peak areas compared with that of the internal standard spiked. The grey rectangles show that the peaks are not detected. The numbers of 1–5 in a right column show the sample number for each organ/tissue, serum, and exosomes described in the Table S2. The numbers of 1–104 (compound 59 is an internal standard spiked before glycoblotting) in a bottom row represent the glycoforms listed in Table S1 and Fig. S2, respectively.

Figure 3

Posttranslational protein glycosylation is dependent strongly on the individual organs. (a) Differences in the expression levels of gross N-glycans represented as pmole/100 mg protein. (b) Total number of N-glycans (glycoforms) identified in the organs/tissues, serum, and exosomes, respectively. Column charts indicate the medians of the results for five independent experiments.

Figure 4

Heatmap representation of the scaled expression levels of N-glycan structures shown in Fig. 2 (103 glycoforms listed in Fig. S2) identified in the mouse organs/tissues, serum, and serum-derived exosomes. The color key in the heatmap shows the column z-score of level for each glycoform. The grey rectangles show that the peaks are not detected. The numbers of 1–5 in a right column show the sample number for each organ/tissue, serum, and exosomes described in the Table S2. The numbers of 1–104 (compound 59 as an internal standard is not shown) in a bottom row represent the glycoforms listed in Table S1 and Fig. S2, respectively.

Figure 5

Glycotyping analysis of 103 mouse organ glycoforms based on the general N-glycan taxonomy (a), the most matured terminal sugar residue defined by the hierarchy in N-glycan biosynthesis (b), the numbers of HexNAc except core structure or the numbers of antennae of the N-glycan structure (c), number of sialic acid residue (Neu5Ac and Neu5Gc) (d), and number of Fuc (e), respectively. Colored columns show the percentages (average of five samples) to total N-glycans estimated in the individual organs except the internal standard compound spiked.

Figure 6

The hierarchical clustering analysis of the mouse N-glycome datasets. The N-glycosylation pattern dissimilarities in all samples were aggregated into 18 groups by using hierarchical clustering method (unsupervised learning). The Canberra distances were used as the dissimilarities among the N-glycosylation patterns in the samples. The dendrogram represents the aggregation by Ward’s method.

Figure 7

Machine learning analysis of the organ N-glycosylation patterns expands applicability of the mouse tissue glycome atlas. (a) The F1 scores in predicting organs by discriminant analysis using four machine learning algorithms with organ/tissue N-glycan profiles. The test data were randomly selected one of five replicates by stratified sampling, while other four of five samples were used as learning data. (b) The F1 scores in predicting mouse strains by discriminant analysis using four machine learning algorisms with lung N-glycan profiles (Tables S2 and S4). Red: decision tree, orange: neural network, sky-blue: random forest, blue: SVM. ddY: outbred mouse Slc:ddY, DBA: inflammatory inbred mouse DBA/2Crslc, MRL: the model mouse of systemic lupus erythematosus (SLE) syndrome MRL-lpr/lpr.