Exploiting glycan topography for computational design of Env glycoprotein antigenicity

Mounting evidence suggests that glycans, rather than merely serving as a “shield”, contribute critically to antigenicity of the HIV envelope (Env) glycoprotein, representing critical antigenic determinants for many broadly neutralizing antibodies (bNAbs). While many studies have focused on defining the role of individual glycans or groups of proximal glycans in bNAb binding, little is known about the effects of changes in the overall glycan landscape in modulating antibody access and Env antigenicity. Here we developed a systems glycobiology approach to reverse engineer the complexity of HIV glycan heterogeneity to guide antigenicity-based de novo glycoprotein design. bNAb binding was assessed against a panel of 94 recombinant gp120 monomers exhibiting defined glycan site occupancies. Using a Bayesian machine learning algorithm, bNAb-specific glycan footprints were identified and used to design antigens that selectively alter bNAb antigenicity as a proof-of concept. Our approach provides a new design strategy to predictively modulate antigenicity via the alteration of glycan topography, thereby focusing the humoral immune response on sites of viral vulnerability for HIV.Carbohydrates on the HIV Env glycoprotein, previously often considered as a “shield” permitting immune evasion, can themselves represent targets for broadly neutralizing antibody (bNAb) recognition. Efforts to define the impact of individual glycans on bNAb recognition have clearly illustrated the critical nature of individual or groups of glycans on bNAb binding. However, glycans represent half the mass of the HIV envelope glycoprotein, representing a lattice of interacting sugars that shape the topographical landscape that alters antibody accessiblity to the underlying protein. However, whether alterations in individual glycans alter the broader interactions among glycans, proximal and distal, has not been heretofore rigorously examined, nor how this lattice may be actively exploited to improve antigenicity. To address this challenge, we describe here a systems glycobiology approach to reverse engineer the complex relationship between bNAb binding and glycan landscape effects on Env proteins spanning across various clades and tiers. Glycan occupancy was interrogated across every potential N-glycan site in 94 recombinant gp120 recombinant antigens. Sequences, glycan occupancy, as well as bNAb binding profiles were integrated across each of the 94-atngeins to generate a machine learning computational model enabling the identification of the glycan site determinants involved in binding to any given bNAb. Moreover, this model was used to generate a panel of novel gp120 variants with augmented selective bNAb binding profiles, further validating the contributions of glycans in Env antigen design. Whether glycan-optimization will additionally influence immunogenicity, particularly on emerging stabilized trimers, is unknown, but this study provides a proof of concept for selectively and agnostically exploiting both proximal and distal viral protein glycosylation in a principled manner to improve target Ab binding profiles