Growing Glycans in Rosetta: Accurate de novo glycan modeling, density fitting, and rational sequon design

Abstract

Carbohydrates and glycoproteins modulate key biological functions. However, experimental structure determination of sugar polymers is notoriously difficult. Computational approaches can aid in carbohydrate structure prediction, structure determination, and design. In this work, we developed a glycan-modeling algorithm, GlycanTreeModeler, that computationally builds glycans layer-by-layer, using adaptive kernel density estimates (KDE) of common glycan conformations derived from data in the Protein Data Bank (PDB) and from quantum mechanics (QM) calculations. GlycanTreeModeler was benchmarked on a test set of glycan structures of varying lengths, or "trees". Structures predicted by GlycanTreeModeler agreed with native structures at high accuracy for both de novo modeling and experimental density-guided building. We employed these tools to design de novo glycan trees into a protein nanoparticle vaccine to shield regions of the scaffold from antibody recognition, and experimentally verified shielding. This work will inform glycoprotein model prediction, glycan masking, and further aid computational methods in experimental structure determination and refinement.

Fig 1. Near native structures from de novo modeling.

(Top Scoring models for each glycan in the benchmark set) Yellow = Native, Cyan = Model.

Fig 2. De novo predictions, farthest from native.

(Top Scoring models for each glycan in the benchmark set) Yellow = Native, Cyan = Model

Fig 3

Best and worst results from density-guided modeling: a. Structural comparison of 3gml 165A glycan; 0.08Å RMSD; cyan = model | yellow = native b (and e). RMSD vs. Score (funnel) plot, top 80% by energy. c (and f). Funnel plot of top 10% models with pNear metrics. d. Structural comparison of 1gai 171A glycan; 0.88Å RMSD; cyan = model | yellow = native

Fig 4

Density-guided modeling quality: a. Boxplot of the RMSD to native of the best-scoring decoy for each of the benchmarked input glycans. b. Boxplot of the funnel quality for each of the benchmark glycans as measured by the pNear metric. A value closer to 1.0 indicates a high-quality funnel.

Fig 5. Characterization and reduced immunogenicity through glycosylation of the trimeric component of the I53-50 two-component protein nanoparticle scaffold that is used in clinical-stage subunit protein nanoparticle immunogens.

a. Schematic of protein design models. On the left, twenty I53-50A trimers (gray) and twelve I53-50B pentamers (orange) self-assemble into I53-50 protein nanoparticles [51]. Rosetta sugarcoating design protocols were used to glycosylate the outer surface of I53-50A trimers with 4 N-linked glycans (green) per protomer to form I53-50 particles with 240 N-linked glycans (middle). The inset on the right is a close-up view of glycosylated I53-50A trimers with 12 total glycans on the outward-facing surface. b. Characterization of bare versus glycosylated I53-50 particles using negative stain transmission electron microscopy (nsTEM; scale bar, 100 nm), SDS-PAGE, dynamic light scattering (DLS), and size exclusion chromatography (SEC) on a Superose 6 Increase 10/300 GL column (GE Healthcare). In the SEC chromatogram, both I53-50 and I53-50(gly) particles reach peak elution at 12.5 mL; unassembled I53-50A and I53-50B components elute at ~18 mL. c. ELISA curves (left two plots) and corresponding EC₅₀ titers (right bar plot) showing reduction in anti-I53-50A antibody responses when mice were immunized with I53-50(gly) versus I53-50. BALB/c mice were immunized intramuscularly at 0, 3, and 6 weeks with 5.57 μg of I53-50 or I53-50(gly) and serum antibody binding to I53-50A trimer (left) or I53-50A(gly) trimer (right) was quantified via ELISA using 8-week sera (N = 5 mice/group). For statistical analysis, Mann-Whitney tests were used to compare among the experimental groups.

Fig 6. Glycan modeling diagram.

a. Glycan trees building layer by layer. Numbers indicate distance to root of the glycan tree, which is the first residue. b. After a layer is built, Glycan Sampling is performed on the new layer, and then all layers, before building the next layer. c. Diagram showing major components of the GlycanSampler. The GS is a weighted random sampler, indicating that each DOF is sampled with a specific probability (S1 Text).

Fig 7. Schematic of benchmarking protocol.