A consensus-based and readable extension of Linear Code for Reaction Rules (LiCoRR).

Systems glycobiology aims to provide models and analysis tools that account for the biosynthesis, regulation, and interactions with glycoconjugates. To facilitate these methods, there is a need for a clear glycan representation accessible to both computers and humans. Linear Code, a linearized and readily parsable glycan structure representation, is such a language. For this reason, Linear Code was adapted to represent reaction rules, but the syntax has drifted from its original description to accommodate new and originally unforeseen challenges. Here, we delineate the consensuses and inconsistencies that have arisen through this adaptation. We recommend options for a consensus-based extension of Linear Code that can be used for reaction rule specification going forward. Through this extension and specification of Linear Code to reaction rules, we aim to minimize inconsistent symbology thereby making glycan database queries easier. With a clear guide for generating reaction rule descriptions, glycan synthesis models will be more interoperable and reproducible thereby moving glycoinformatics closer to compliance with FAIR standards. Here, we present Linear Code for Reaction Rules (LiCoRR), version 1.0, an unambiguous representation for describing glycosylation reactions in both literature and code.

Integration of the transcriptome and glycome for identification of glycan cell signatures.

Abnormalities in glycan biosynthesis have been conclusively linked to many diseases but the complexity of glycosylation has hindered the analysis of glycan data in order to identify glycoforms contributing to disease. To overcome this limitation, we developed a quantitative N-glycosylation model that interprets and integrates mass spectral and transcriptomic data by incorporating key glycosylation enzyme activities. Using the cancer progression model of androgen-dependent to androgen-independent Lymph Node Carcinoma of the Prostate (LNCaP) cells, the N-glycosylation model identified and quantified glycan structural details not typically derived from single-stage mass spectral or gene expression data. Differences between the cell types uncovered include increases in H(II) and Le(y) epitopes, corresponding to greater activity of α2-Fuc-transferase (FUT1) in the androgen-independent cells. The model further elucidated limitations in the two analytical platforms including a defect in the microarray for detecting the GnTV (MGAT5) enzyme. Our results demonstrate the potential of systems glycobiology tools for elucidating key glycan biomarkers and potential therapeutic targets. The integration of multiple data sets represents an important application of systems biology for understanding complex cellular processes.

A mathematical model to derive N-glycan structures and cellular enzyme activities from mass spectrometric data.

Effective representation and characterization of biosynthetic pathways of glycosylation can be facilitated by mathematical modeling. This paper describes the expansion of a previously developed detailed model for N-linked glycosylation with the further application of the model to analyze MALDI-TOF mass spectra of human N-glycans in terms of underlying cellular enzyme activities. The glycosylation reaction network is automatically generated by the model, based on the reaction specificities of the glycosylation enzymes. The use of a molecular mass cutoff and a network pruning method typically limits the model size to about 10,000 glycan structures. This allows prediction of the complete glycan profile and its abundances for any set of assumed enzyme concentrations and reaction rate parameters. A synthetic mass spectrum from model-calculated glycan profiles is obtained and enzyme concentrations are adjusted to bring the theoretically calculated mass spectrum into agreement with experiment. The result of this process is a complete characterization of a measured glycan mass spectrum containing hundreds of masses in terms of the activities of 19 enzymes. In addition, a complete annotation of the mass spectrum in terms of glycan structure is produced, including the proportions of isomers within each peak. The method was applied to mass spectrometric data of normal human monocytes and monocytic leukemia (THP1) cells to derive glycosyltransferase activity changes underlying the differences in glycan structure between the normal and diseased cells. Model predictions could lead to a better understanding of the changes associated with disease states, identification of disease-associated biomarkers, and bioengineered glycan modifications.

Model-based analysis of N-glycosylation in Chinese hamster ovary cells.

The Chinese hamster ovary (CHO) cell is the gold standard for manufacturing of glycosylated recombinant proteins for production of biotherapeutics. The similarity of its glycosylation patterns to the human versions enable the products of this cell line favorable pharmacokinetic properties and lower likelihood of causing immunogenic responses. Because glycan structures are the product of the concerted action of intracellular enzymes, it is difficult to predict a priori how the effects of genetic manipulations alter glycan structures of cells and therapeutic properties. For that reason, quantitative models able to predict glycosylation have emerged as promising tools to deal with the complexity of glycosylation processing. For example, an earlier version of the same model used in this study was used by others to successfully predict changes in enzyme activities that could produce a desired change in glycan structure. In this study we utilize an updated version of this model to provide a comprehensive analysis of N-glycosylation in ten Chinese hamster ovary (CHO) cell lines that include a wild type parent and nine mutants of CHO, through interpretation of previously published mass spectrometry data. The updated N-glycosylation mathematical model contains up to 50,605 glycan structures. Adjusting the enzyme activities in this model to match N-glycan mass spectra produces detailed predictions of the glycosylation process, enzyme activity profiles and complete glycosylation profiles of each of the cell lines. These profiles are consistent with biochemical and genetic data reported previously. The model-based results also predict glycosylation features of the cell lines not previously published, indicating more complex changes in glycosylation enzyme activities than just those resulting directly from gene mutations. The model predicts that the CHO cell lines possess regulatory mechanisms that allow them to adjust glycosylation enzyme activities to mitigate side effects of the primary loss or gain of glycosylation function known to exist in these mutant cell lines. Quantitative models of CHO cell glycosylation have the potential for predicting how glycoengineering manipulations might affect glycoform distributions to improve the therapeutic performance of glycoprotein products.

A mathematical model of N-linked glycosylation.

Metabolic engineering of N-linked oligosaccharide biosynthesis to produce novel glycoforms or glycoform distributions of a recombinant glycoprotein can potentially lead to an improved therapeutic performance of the glycoprotein product. A mathematical model for the initial stages of this process, up to the first galactosylation of an oligosaccharide, was previously developed by Umana and Bailey (1997) (UB1997). Building on this work, an extended model is developed to include further galactosylation, fucosylation, extension of antennae by N-acetyllactosamine repeats, and sialylation. This allows many more structural features to be predicted. A number of simplifying assumptions are also relaxed to incorporate more variables for the control of glycoforms. The full model generates 7565 oligosaccharide structures in a network of 22,871 reactions. Methods for solving the model for the complete product distribution and adjusting the parameters to match experimental data are also developed. A basal set of kinetic parameters for the enzyme-catalyzed reactions acting on free oligosaccharide substrates is obtained from the previous model and existing literature. Enzyme activities are adjusted to match experimental glycoform distributions for Chinese Hamster Ovary (CHO). The model is then used to predict the effect of increasing expression of a target glycoprotein on the product glycoform distribution and evaluate appropriate metabolic engineering strategies to return the glycoform profile to its original distribution pattern. This model may find significant utility in the future to predict glycosylation patterns and direct glycoengineering projects to optimize glycoform distributions.