The N-Glycosylation Processing Potential of the Mammalian Golgi Apparatus

Abstract

Heterogeneity is an inherent feature of the glycosylation process. Mammalian cells often produce a variety of glycan structures on separate molecules of the same protein, known as glycoforms. This heterogeneity is not random but is controlled by the organization of the glycosylation machinery in the Golgi cisternae. In this work, we use a computational model of the N-glycosylation process to probe how the organization of the glycosylation machinery into different cisternae drives N-glycan biosynthesis toward differing degrees of heterogeneity. Using this model, we demonstrate the N-glycosylation potential and limits of the mammalian Golgi apparatus, for example how the number of cisternae limits the goal of achieving near homogeneity for N-glycans. The production of specific glycoforms guided by this computational study could pave the way for "glycoform engineering," which will find uses in the functional investigation of glycans, the modulation of glycan-mediated physiological functions, and in biotechnology.

Keywords: Golgi apparatus; cisternal number; computational modeling; glycan biosynthesis; glycan heterogeneity.

FIGURE 1

N-glycan processing pathways to generate a fully sialylated bi-antennary glycan. (A) Pictorial examples of structural features of complex N-glycans that are referred to in this work. (B) N-glycan biosynthesis pathway for Fuc₁GlcNAc₄Man₃Gal₂NeuAc₂ starting with the two oligomannose glycans that enter the Golgi apparatus. The biosynthesis of complex N-glycans is characterized by increasing numbers of reaction paths that ultimately converge on structural nodes often due to the addition of terminating sialic acids.

FIGURE 2

Effect of cisternal number and transit time on glycan heterogeneity. (A) Total number of glycans (circles) and proportion of oligomannose glycans (bars) produced in silico when starting with parameters fitted for the WT HEK293T cell glycan profile altering the number of cisternal elements (CS) in the system from the four (striped bar) used in the WT HEK293T model. The distribution of enzymes across the model Golgi apparatus, the total effective rates for each enzyme and the total time spent in the Golgi apparatus was kept the same for each simulation. (B) As in “A” but varying the total time spent in the Golgi with four cisternae. Striped bar indicates transit time for the WT HEK293T cell model; the times shown for all other simulations are relative to this value. (C) Relative proportions of the three most abundant complex glycans predicted by the model Golgi apparatus when the number of cisternal elements are altered. For all simulations error bars show standard deviation for n = 3.

FIGURE 3

Glycan by-products of maximizing target complex glycans. Relative abundance of the oligomannose pool and the three most abundant complex glycans that are predicted as by-products when maximising the abundance of the indicated target glycans with the indicated type (fixed or variable enzyme localization) of fitting strategy.

FIGURE 4

Optimized Golgi distribution of enzymes to maximize the relative abundance of target glycans. Distributions of the effective enzymatic rates for the indicated enzymes throughout the Golgi apparatus following fitting with variable enzyme localization to maximize the abundance of the indicated target glycans. The predicted distribution in WT HEK293T cells (Fisher et al., 2019) is shown for comparison.

FIGURE 5

Flux analysis of Mgat2 substrates in WT, bi-Sia₁, and bi-Sia₂ simulations. (A) Proportion of biosynthetic flux carried by Mgat2 from its three dominant substrates in WT HEK293T cells (red) and the simulated profiles maximizing bi-Sia₁ (blue) and bi-Sia₂ (purple) fitted with variable enzyme localization. (B) The proportion of substrates for Mgat2 that are fucosylated for the simulations described in “A.”

FIGURE 6

Total enzymatic rate changes to maximize relative abundance of target glycans. Percentage changes in total effective enzymatic rates following fitting to maximize the abundance of bi-Sia₁ (blue), bi-Sia₂ (red), and tri-Sia₁ (green) fitted with variable enzyme localization (solid) and fixed enzyme localization (striped) relative to the parameters predicted for the WT HEK293T cell profile. Total effective enzymatic rates obtained from the different fits were adjusted to ensure that the transit time in each case matched that of the WT HEK293T simulation.

FIGURE 7

Hypothetical engineering of the Golgi apparatus to increase the abundance of tetra-antennary glycans. (A) The relative abundance of all tetra- and tri-antennary glycans produced in silico using effective enzymatic rates for: WT HEK293T (red), WT with the total effective enzymatic rates of Mgat4/5 increased 10-fold (dark blue), WT with the total effective enzymatic rates of Mgat4/5 increased 100-fold (light blue), WT with the distribution of enzymes in the model Golgi apparatus set to separate Mgat4/5 from GalT (yellow), and WT with the total effective enzymatic rates of Mgat4/5 increased 10-fold and the localization of enzymes in the model Golgi apparatus set to separate Mgat4/5 from GalT (yellow and dark blue striped). (B) Flux diagrams for selected N-glycosylation reactions for WT HEK293T, WT with the total effective enzymatic rates of Mgat4/5 increased 10-fold and WT with the total effective enzymatic rates of Mgat4/5 increased 100-fold. (C) Schematic showing the levels and relative localizations of key enzymes for WT HEK293T cells and the computationally engineered scenario predicted to maximize the production of tetra-antennary glycans.