AI-facilitated high-resolution cryo-EM analyses of tubular mastigonemes reveal the structural roles of N- and O-glycans

Glycans are essential mediators of multiple biological processes, yet their chemical complexity and non-template-driven biosynthesis pathways pose significant challenges for the establishment of the sequence-structure relationship. Here, we report cryo-EM structures of the tubular mastigoneme from a golden alga species. A large number of N- and O-glycans are observed at 1.8-2.2 Å resolutions. We developed EModelG, an AI-guided pipeline for automated atomic modeling of glycans. In addition to a number of canonical high-mannose or complex type N-glycans, we have discovered and confirmed a non-canonical N-glycan linked to Asn in the Ala-Asn-Asp (AND) motif. The surface spikes on the mastigoneme are made of dense O-glycans coating the stem of linear PSXX tetrapeptide repeats. The trihydroxylated proline anchor and sugar compositions, linkages, as well as acetylation, methylation, and sulfation of sugar residues, are unambiguously resolved from the > 2.0-Å EM map and further confirmed with LC-MS/MS analyses. Well-resolved water molecules, which account for more than 10% of the resolved volume, mediate extensive hydrogen bonding. Our integrative approach, combining cryo-EM, glycoproteomics, and AI modeling, establishes a scalable framework for deciphering glycan folding.