Predicting glycan structure from tandem mass spectrometry via deep learning.

Glycans constitute the most complicated post-translational modification, modulating protein activity in health and disease. However, structural annotation from tandem mass spectrometry (MS/MS) data is a bottleneck in glycomics, preventing high-throughput endeavors and relegating glycomics to a few experts. Trained on a newly curated set of 500,000 annotated MS/MS spectra, here we present CandyCrunch, a dilated residual neural network predicting glycan structure from raw liquid chromatography-MS/MS data in seconds (top-1 accuracy: 90.3%). We developed an open-access Python-based workflow of raw data conversion and prediction, followed by automated curation and fragment annotation, with predictions recapitulating and extending expert annotation. We demonstrate that this can be used for de novo annotation, diagnostic fragment identification and high-throughput glycomics. For maximum impact, this entire pipeline is tightly interlaced with our glycowork platform and can be easily tested at https://colab.research.google.com/github/BojarLab/CandyCrunch/blob/main/CandyCrunch.ipynb . We envision CandyCrunch to democratize structural glycomics and the elucidation of biological roles of glycans.

UniCarb-DB: An MS/MS Experimental Glycomic Fragmentation Database.

Glycosylation is a unique posttranslational modification that dynamically shapes the surface of cells. Glycans attached to proteins or lipids in a cell or tissue are studied as a whole and collectively designated as a glycome. UniCarb-DB is a glycomic spectral library of tandem mass spectrometry (MS/MS) fragment data. The current version of the database consists of over 1500 entries and over 1000 unique structures. Each entry contains parent ion information with associated MS/MS spectra, metadata about the original publication, experimental conditions, and biological origin. Each structure is also associated with the GlyTouCan glycan structure repository allowing easy access to other glycomic resources. The database can be directly utilized by mass spectrometry (MS) experimentalists through the conversion of data generated by MS into structural information. Flexible online search tools along with a downloadable version of the database are easily incorporated in either commercial or open-access MS software. This chapter highlights UniCarb-DB online search tool to browse differences of isomeric structures between spectra, a peak matching search between user-generated MS/MS spectra and spectra stored in UniCarb-DB and more advanced MS tools for combined quantitative and qualitative glycomics.

A monoallelic UXS1 variant associated with short-limbed short stature.

BACKGROUND: Serine residues in the protein backbone of heavily glycosylated proteoglycans are bound to glycosaminoglycans through a tetrasaccharide linker. UXS1 encodes UDP-glucuronate decarboxylase 1, which catalyzes synthesis of UDP-xylose, the donor of the first building block in the linker. Defects in other enzymes involved in formation of the tetrasaccharide linker cause so-called linkeropathies, characterized by short stature, radio-ulnar synostosis, decreased bone density, congenital contractures, dislocations, and more. METHODS: Whole exome sequencing was performed in a father and son who presented with a mild skeletal dysplasia, as well as the father's unaffected parents. Wild-type and mutant UXS1 were recombinantly expressed in Escherichia coli and purified. Enzyme activity was evaluated by LC-MS/MS. In vivo effects were studied using HeparinRed assay and metabolomics. RESULTS: The son had short long bones, normal epiphysis, and subtle metaphyseal changes especially in his legs. The likely pathogenic heterozygous variant NM_001253875.1(UXS1):c.557T>A p.(Ile186Asn) detected in the son was de novo in the father. Purified Ile186Asn-UXS1, in contrast to the wild-type, was not able to convert UDP-glucuronic acid to UDP-xylose. Plasma glycosaminoglycan levels were decreased in both son and father. CONCLUSION: This is the first report linking UXS1 to short-limbed short stature in humans.

The evolutionarily ancient FOXA transcription factors shape the murine gut microbiome via control of epithelial glycosylation.

Evolutionary adaptation of multicellular organisms to a closed gut created an internal microbiome differing from that of the environment. Although the composition of the gut microbiome is impacted by diet and disease state, we hypothesized that vertebrates promote colonization by commensal bacteria through shaping of the apical surface of the intestinal epithelium. Here, we determine that the evolutionarily ancient FOXA transcription factors control the composition of the gut microbiome by establishing favorable glycosylation on the colonic epithelial surface. FOXA proteins bind to regulatory elements of a network of glycosylation enzymes, which become deregulated when Foxa1 and Foxa2 are deleted from the intestinal epithelium. As a direct consequence, microbial composition shifts dramatically, and spontaneous inflammatory bowel disease ensues. Microbiome dysbiosis was quickly reversed upon fecal transplant into wild-type mice, establishing a dominant role for the host epithelium, in part mediated by FOXA factors, in controlling symbiosis in the vertebrate holobiont.

Ion mobility-tandem mass spectrometry of mucin-type O-glycans.

The dense O-glycosylation of mucins plays an important role in the defensive properties of the mucus hydrogel. Aberrant glycosylation is often correlated with inflammation and pathology such as COPD, cancer, and Crohn's disease. The inherent complexity of glycans and the diversity in the O-core structure constitute fundamental challenges for the analysis of mucin-type O-glycans. Due to coexistence of multiple isomers, multidimensional workflows such as LC-MS are required. To separate the highly polar carbohydrates, porous graphitized carbon is often used as a stationary phase. However, LC-MS workflows are time-consuming and lack reproducibility. Here we present a rapid alternative for separating and identifying O-glycans released from mucins based on trapped ion mobility mass spectrometry. Compared to established LC-MS, the acquisition time is reduced from an hour to two minutes. To test the validity, the developed workflow was applied to sputum samples from cystic fibrosis patients to map O-glycosylation features associated with disease.

Characterization of intestinal O-glycome in reactive oxygen species deficiency.

Inflammatory bowel disease (IBD) is characterized by chronic intestinal inflammation resulting from an inappropriate inflammatory response to intestinal microbes in a genetically susceptible host. Reactive oxygen species (ROS) generated by NADPH oxidases (NOX) provide antimicrobial defense, redox signaling and gut barrier maintenance. NADPH oxidase mutations have been identified in IBD patients, and mucus layer disruption, a critical aspect in IBD pathogenesis, was connected to NOX inactivation. To gain insight into ROS-dependent modification of epithelial glycosylation the colonic and ileal mucin O-glycome of mice with genetic NOX inactivation (Cyba mutant) was analyzed. O-glycans were released from purified murine mucins and analyzed by hydrophilic interaction ultra-performance liquid chromatography in combination with exoglycosidase digestion and mass spectrometry. We identified five novel glycans in ileum and found minor changes in O-glycans in the colon and ileum of Cyba mutant mice. Changes included an increase in glycans with terminal HexNAc and in core 2 glycans with Fuc-Gal- on C3 branch, and a decrease in core 3 glycans in the colon, while the ileum showed increased sialylation and a decrease in sulfated glycans. Our data suggest that NADPH oxidase activity alters the intestinal mucin O-glycans that may contribute to intestinal dysbiosis and chronic inflammation.

Glycan Complexity and Heterogeneity of Glycoproteins in Somatic Extracts and Secretome of the Infective Stage of the Helminth Fasciola hepatica.

Fasciola hepatica is a global helminth parasite of humans and their livestock. The invasive stage of the parasite, the newly excysted juvenile (NEJs), relies on glycosylated excreted-secreted (ES) products and surface/somatic molecules to interact with host cells and tissues and to evade the host's immune responses, such as disarming complement and shedding bound antibody. While -omics technologies have generated extensive databases of NEJs' proteins and their expression, detailed knowledge of the glycosylation of proteins is still lacking. Here, we employed glycan, glycopeptide, and proteomic analyses to determine the glycan profile of proteins within the NEJs' somatic (Som) and ES extracts. These analyses characterized 123 NEJ glycoproteins, 71 of which are secreted proteins, and allowed us to map 356 glycopeptides and their associated 1690 N-glycan and 37 O-glycan forms to their respective proteins. We discovered abundant micro-heterogeneity in the glycosylation of individual glycosites and between different sites of multi-glycosylated proteins. The global heterogeneity across NEJs' glycoproteome was refined to 53 N-glycan and 16 O-glycan structures, ranging from highly truncated paucimannosidic structures to complex glycans carrying multiple phosphorylcholine (PC) residues, and included various unassigned structures due to unique linkages, particularly in pentosylated O-glycans. Such exclusive glycans decorate some well-known secreted molecules involved in host invasion, including cathepsin B and L peptidases, and a variety of membrane-bound glycoproteins, suggesting that they participate in host interactions. Our findings show that F. hepatica NEJs generate exceptional protein variability via glycosylation, suggesting that their molecular portfolio that communicates with the host is far more complex than previously anticipated by transcriptomic and proteomic analyses. This study opens many avenues to understand the glycan biology of F. hepatica throughout its life-stages, as well as other helminth parasites, and allows us to probe the glycosylation of individual NEJs proteins in the search for innovative diagnostics and vaccines against fascioliasis.

Protein N-glycosylation in the bronchoalveolar space differs between never-smokers and long-term smokers with and without COPD.

Chronic obstructive pulmonary disease (COPD) kills millions of people annually and patients suffering from exacerbations of this disorder display high morbidity and mortality. The clinical course of COPD is associated with dysbiosis and infections, but the underlying mechanisms are poorly understood. Glycosylation of proteins play roles in regulating interactions between microbes and immune cells, and knowledge on airway glycans therefore contribute to the understanding of infections. Furthermore, glycans have biomarker potential for identifying smokers with enhanced risk for developing COPD as well as COPD subgroups. Here, we characterized the N-glycosylation in the lower airways of healthy never-smokers (HNS, n = 5) and long-term smokers (LTS) with (LTS+, n = 4) and without COPD (LTS-, n = 8). Using mass spectrometry, we identified 57 highly confident N-glycan structures whereof 38 oligomannose, complex, and paucimannose type glycans were common to BAL samples from HNS, LTS- and LTS+ groups. Hybrid type N-glycans were identified only in the LTS+ group. Qualitatively and quantitatively, HNS had lower inter-individual variation between samples compared to LTS- or LTS+. Cluster analysis of BAL N-glycosylation distinguished LTS from HNS. Correlation analysis with clinical parameters revealed that complex N-glycans were associated with health and absence of smoking whereas oligomannose N-glycans were associated with smoking and disease. The N-glycan profile from monocyte-derived macrophages differed from the BAL N-glycan profiles. In conclusion, long-term smokers display substantial alterations of N-glycosylation in the bronchoalveolar space, and the hybrid N-glycans identified only in long-term smokers with COPD deserve to be further studied as potential biomarkers.

Quantification of chondroitin sulfate, hyaluronic acid and N-glycans in synovial fluid - A technical performance study.

Objective: To validate a quantitative high performance liquid chromatography (HPLC) assay for chondroitin sulfate (CS) and hyaluronic acid (HA) in synovial fluid, and to analyze glycan-patterns in patient samples. Design: Synovial fluid from osteoarthritis (OA, n â€‹= â€‹25) and knee-injury (n â€‹= â€‹13) patients, a synovial fluid pool (SF-control) and purified aggrecan, were chondroitinase digested and together with CS- and HA-standards fluorophore labelled prior to quantitative HPLC analysis. N-glycan profiles of synovial fluid and aggrecan were assessed by mass spectrometry. Results: Unsaturated uronic acid and sulfated-N-acetylgalactosamine (ΔUA-GalNAc4S and ΔUA-GalNAc6S) contributed to 95% of the total CS-signal in the SF-control sample. For HA and the CS variants in SF-control the intra- and inter-experiment coefficient of variation was between 3-12% and 11-19%, respectively; tenfold dilution gave recoveries between 74 and 122%, and biofluid stability test (room temperature storage and freeze-thaw cycles) showed recoveries between 81 and 140%. Synovial fluid concentrations of the CS variants ΔUA-GalNAc6S and ΔUA2S-GalNAc6S were three times higher in the recent injury group compared to the OA group, while HA was four times lower. Sixty-one different N-glycans were detected in the synovial fluid samples, but there were no differences in levels of N-glycan classes between patient groups. The CS-profile (levels of ΔUA-GalNAc4S and ΔUA-GalNAc6S) in synovial fluid resembled that of purified aggrecan from corresponding samples; the contribution to the N-glycan profile in synovial fluid from aggrecan was low. Conclusions: The HPLC-assay is suitable for analyzing CS variants and HA in synovial fluid samples, and the GAG-pattern differs between OA and recently knee injured subjects.

Sialidases and fucosidases of Akkermansia muciniphila are crucial for growth on mucin and nutrient sharing with mucus-associated gut bacteria.

The mucolytic human gut microbiota specialist Akkermansia muciniphila is proposed to boost mucin-secretion by the host, thereby being a key player in mucus turnover. Mucin glycan utilization requires the removal of protective caps, notably fucose and sialic acid, but the enzymatic details of this process remain largely unknown. Here, we describe the specificities of ten A. muciniphila glycoside hydrolases, which collectively remove all known sialyl and fucosyl mucin caps including those on double-sulfated epitopes. Structural analyses revealed an unprecedented fucosidase modular arrangement and explained the sialyl T-antigen specificity of a sialidase of a previously unknown family. Cell-attached sialidases and fucosidases displayed mucin-binding and their inhibition abolished growth of A. muciniphila on mucin. Remarkably, neither the sialic acid nor fucose contributed to A. muciniphila growth, but instead promoted butyrate production by co-cultured Clostridia. This study brings unprecedented mechanistic insight into the initiation of mucin O-glycan degradation by A. muciniphila and nutrient sharing between mucus-associated bacteria.

Reproducing extracellular matrix adverse remodelling of non-ST myocardial infarction in a large animal model.

The rising incidence of non-ST-segment elevation myocardial infarction (NSTEMI) and associated long-term high mortality constitutes an urgent clinical issue. Unfortunately, the study of possible interventions to treat this pathology lacks a reproducible pre-clinical model. Indeed, currently adopted small and large animal models of MI mimic only full-thickness, ST-segment-elevation (STEMI) infarcts, and hence cater only for an investigation into therapeutics and interventions directed at this subset of MI. Thus, we develop an ovine model of NSTEMI by ligating the myocardial muscle at precise intervals parallel to the left anterior descending coronary artery. Upon histological and functional investigation to validate the proposed model and comparison with STEMI full ligation model, RNA-seq and proteomics show the distinctive features of post-NSTEMI tissue remodelling. Transcriptome and proteome-derived pathway analyses at acute (7 days) and late (28 days) post-NSTEMI pinpoint specific alterations in cardiac post-ischaemic extracellular matrix. Together with the rise of well-known markers of inflammation and fibrosis, NSTEMI ischaemic regions show distinctive patterns of complex galactosylated and sialylated N-glycans in cellular membranes and extracellular matrix. Identifying such changes in molecular moieties accessible to infusible and intra-myocardial injectable drugs sheds light on developing targeted pharmacological solutions to contrast adverse fibrotic remodelling.

Amplification of Inflammation by Lubricin Deficiency Implicated in Incident, Erosive Gout Independent of Hyperuricemia.

OBJECTIVE: In gout, hyperuricemia promotes urate crystal deposition, which stimulates the NLRP3 inflammasome and interleukin-1ß (IL-1ß)-mediated arthritis. Incident gout without background hyperuricemia is rarely reported. To identify hyperuricemia-independent mechanisms driving gout incidence and progression, we characterized erosive urate crystalline inflammatory arthritis in a young female patient with normouricemia diagnosed as having sufficient and weighted classification criteria for gout according to the American College of Rheumatology (ACR)/EULAR gout classification criteria (the proband). METHODS: We conducted whole-genome sequencing, quantitative proteomics, whole-blood RNA-sequencing analysis using serum samples from the proband. We used a mouse model of IL-1ß-induced knee synovitis to characterize proband candidate genes, biomarkers, and pathogenic mechanisms of gout. RESULTS: Lubricin level was attenuated in human proband serum and associated with elevated acute-phase reactants and inflammatory whole-blood transcripts and transcriptional pathways. The proband had predicted damaging gene variants of NLRP3 and of inter-α trypsin inhibitor heavy chain 3, an inhibitor of lubricin-degrading cathepsin G. Changes in the proband's serum protein interactome network supported enhanced lubricin degradation, with cathepsin G activity increased relative to its inhibitors, SERPINB6 and thrombospondin 1. Activation of Toll-like receptor 2 (TLR-2) suppressed levels of lubricin mRNA and lubricin release in cultured human synovial fibroblasts (P < 0.01). Lubricin blunted urate crystal precipitation and IL-1ß induction of xanthine oxidase and urate in cultured macrophages (P < 0.001). In lubricin-deficient mice, injection of IL-1ß in knees increased xanthine oxidase-positive synovial resident M1 macrophages (P < 0.05). CONCLUSION: Our findings linked normouricemic erosive gout to attenuated lubricin, with impaired control of cathepsin G activity, compounded by deleterious NLRP3 variants. Lubricin suppressed monosodium urate crystallization and blunted IL-1ß-induced increases in xanthine oxidase and urate in macrophages. The collective activities of articular lubricin that could limit incident and erosive gouty arthritis independently of hyperuricemia are subject to disruption by inflammation, activated cathepsin G, and synovial fibroblast TLR-2 signaling.

Metabolic reprogramming and membrane glycan remodeling as potential drivers of zebrafish heart regeneration.

The ability of the zebrafish heart to regenerate following injury makes it a valuable model to deduce why this capability in mammals is limited to early neonatal stages. Although metabolic reprogramming and glycosylation remodeling have emerged as key aspects in many biological processes, how they may trigger a cardiac regenerative response in zebrafish is still a crucial question. Here, by using an up-to-date panel of transcriptomic, proteomic and glycomic approaches, we identify a metabolic switch from mitochondrial oxidative phosphorylation to glycolysis associated with membrane glycosylation remodeling during heart regeneration. Importantly, we establish the N- and O-linked glycan structural repertoire of the regenerating zebrafish heart, and link alterations in both sialylation and high mannose structures across the phases of regeneration. Our results show that metabolic reprogramming and glycan structural remodeling are potential drivers of tissue regeneration after cardiac injury, providing the biological rationale to develop novel therapeutics to elicit heart regeneration in mammals.

Truncated lubricin glycans in osteoarthritis stimulate the synoviocyte secretion of VEGFA, IL-8, and MIP-1α: Interplay between O-linked glycosylation and inflammatory cytokines.

The primary aim of the study was to identify inflammatory markers relevant for osteoarthritis (OA)-related systemic (plasma) and local (synovial fluid, SF) inflammation. From this, we looked for inflammatory markers that coincided with the increased amount of O-linked Tn antigen (GalNAcα1-Ser/Thr) glycan on SF lubricin. Inflammatory markers in plasma and SF in OA patients and controls were measured using a 44-multiplex immunoassay. We found consistently 29 markers detected in both plasma and SF. The difference in their concentration and the low correlation when comparing SF and plasma suggests an independent inflammatory environment in the two biofluids. Only plasma MCP-4 and TARC increased in our patient cohort compared to control plasma. To address the second task, we concluded that plasma markers were irrelevant for a direct connection with SF glycosylation. Hence, we correlated the SF-inflammatory marker concentrations with the level of altered glycosylation of SF-lubricin. We found that the level of SF-IL-8 and SF-MIP-1α and SF-VEGFA in OA patients displayed a positive correlation with the altered lubricin glycosylation. Furthermore, when exposing fibroblast-like synoviocytes from both controls and OA patients to glycovariants of recombinant lubricin, the secretion of IL-8 and MIP-1α and VEGFA were elevated using lubricin with Tn antigens, while lubricin with sialylated and nonsialylated T antigens had less or no measurable effect. These data suggest that truncated glycans of lubricin, as found in OA, promote synovial proinflammatory cytokine production and exacerbate local synovial inflammation.

A Complex Connection Between the Diversity of Human Gastric Mucin O-Glycans, Helicobacter pylori Binding, Helicobacter Infection and Fucosylation.

Helicobacter pylori colonizes the stomach of half of the human population. Most H. pylori are located in the mucus layer, which is mainly comprised by glycosylated mucins. Using mass spectrometry, we identified 631 glycans (whereof 145 were fully characterized and the remainder assigned as compositions) on mucins isolated from 14 Helicobacter spp.-infected and 14 Helicobacter spp.-noninfected stomachs. Only six identified glycans were common to all individuals, from a total of 60 to 189 glycans in each individual. An increased number of unique glycan structures together with an increased intraindividual diversity and larger interindividual variation were identified among O-glycans from Helicobacter spp.-infected stomachs compared with noninfected stomachs. H. pylori strain J99, which carries the blood group antigen-binding adhesin (BabA), the sialic acid-binding adhesin (SabA), and the LacdiNAc-binding adhesin, bound both to Lewis b (Leb)-positive and Leb-negative mucins. Among Leb-positive mucins, H. pylori J99 binding was higher to mucins from Helicobacter spp.-infected individuals than noninfected individuals. Statistical correlation analysis, binding experiments with J99 wt, and J99ΔbabAΔsabA and inhibition experiments using synthetic glycoconjugates demonstrated that the differences in H. pylori-binding ability among these four groups were governed by BabA-dependent binding to fucosylated structures. LacdiNAc levels were lower in mucins that bound to J99 lacking BabA and SabA than in mucins that did not, suggesting that LacdiNAc did not significantly contribute to the binding. We identified 24 O-glycans from Leb-negative mucins that correlated well with H. pylori binding whereof 23 contained α1,2-linked fucosylation. The large and diverse gastric glycan library identified, including structures that correlated with H. pylori binding, could be used to select glycodeterminants to experimentally investigate further for their importance in host-pathogen interactions and as candidates to develop glycan-based therapies.

Sulfated glycan recognition by carbohydrate sulfatases of the human gut microbiota.

Sulfated glycans are ubiquitous nutrient sources for microbial communities that have coevolved with eukaryotic hosts. Bacteria metabolize sulfated glycans by deploying carbohydrate sulfatases that remove sulfate esters. Despite the biological importance of sulfatases, the mechanisms underlying their ability to recognize their glycan substrate remain poorly understood. Here, we use structural biology to determine how sulfatases from the human gut microbiota recognize sulfated glycans. We reveal seven new carbohydrate sulfatase structures spanning four S1 sulfatase subfamilies. Structures of S1_16 and S1_46 represent novel structures of these subfamilies. Structures of S1_11 and S1_15 demonstrate how non-conserved regions of the protein drive specificity toward related but distinct glycan targets. Collectively, these data reveal that carbohydrate sulfatases are highly selective for the glycan component of their substrate. These data provide new approaches for probing sulfated glycan metabolism while revealing the roles carbohydrate sulfatases play in host glycan catabolism.

Computational Modeling of O-Linked Glycan Biosynthesis in CHO Cells.

Glycan biosynthesis simulation research has progressed remarkably since 1997, when the first mathematical model for N-glycan biosynthesis was proposed. An O-glycan model has also been developed to predict O-glycan biosynthesis pathways in both forward and reverse directions. In this work, we started with a set of O-glycan profiles of CHO cells transiently transfected with various combinations of glycosyltransferases. The aim was to develop a model that encapsulated all the enzymes in the CHO transfected cell lines. Due to computational power restrictions, we were forced to focus on a smaller set of glycan profiles, where we were able to propose an optimized set of kinetics parameters for each enzyme in the model. Using this optimized model we showed that the abundance of more processed glycans could be simulated compared to observed abundance, while predicting the abundance of glycans earlier in the pathway was less accurate. The data generated show that for the accurate prediction of O-linked glycosylation, additional factors need to be incorporated into the model to better reflect the experimental conditions.

An Interactive View of Glycosylation.

The present chapter focuses on the interactive and explorative aspects of bioinformatics resources that have been recently released in glycobiology. The comparative analysis of data in a field where knowledge is scattered, incomplete, and disconnected from main biology requires efficient visualization, integration, and interactive tools that are currently only partially implemented. This overview highlights converging efforts toward building a consistent picture of protein glycosylation.