AI-based structure prediction empowers integrative structural analysis of human nuclear pores

Area of Science:

• Structural biology
• Molecular cell biology
• Biophysics

Background:

• The eukaryotic nucleus is protected by the nuclear envelope, perforated by nuclear pore complexes (NPCs) for nucleocytoplasmic transport.
• Human NPCs are large protein complexes (~120 MDa) composed of nucleoporins (NUPs), categorized into scaffold and intrinsically disordered types.
• NPC architecture is dynamic, responding to nuclear envelope tension with conformational changes, posing challenges for high-resolution structural determination.

Purpose of the Study:

• To elucidate the human NPC scaffold architecture at atomic resolution, improving understanding of its function and dynamics.
• To overcome challenges in previous structural models, including limited coverage of human NUPs, difficulty in characterizing disordered linkers, and the NPC's membrane association.

Main Methods:

• Artificial intelligence (AI)-based prediction to generate structural models of human NUPs and subcomplexes.
• Cryo-electron tomography (cryo-ET) to obtain high-resolution maps of NPC conformational states (constricted and dilated).
• Integrative modeling to fit AI-predicted NUP structures into cryo-ET maps, including linker NUPs and membrane-associated NUPs.
• Molecular dynamics simulations to analyze NPC scaffold behavior within a membrane environment.

Main Results:

• Generated an extensive repertoire of accurate structural models for human NUPs, covering previously uncharacterized domains and interfaces.
• Obtained well-resolved cryo-ET maps of human NPC in constricted and dilated states, enabling detailed fitting of NUP structures.
• Developed a comprehensive architectural model, doubling structural coverage of the NPC scaffold and revealing precise anchoring sites for disordered NUPs.
• Simulations showed the NPC scaffold prevents excessive constriction of the membrane fusion pore.

Conclusions:

• A 70-MDa, atomically resolved model covering >90% of the human NPC scaffold has been established.
• The model captures NPC conformational changes and precisely identifies anchoring sites for intrinsically disordered NUPs.
• AI-driven structure prediction significantly accelerates the elucidation of complex subcellular architectures at atomic resolution.