Integrative structural modeling of nuclear pore complexes

Abstract

Historically, structural biologists have applied mainly two experimental methods for the structure determination of proteins: X-ray crystallography and nuclear magnetic resonance (NMR) spectroscopy. Recently, cryo-electron microscopy (cryo-EM) was also added to the regular toolkit of structural biologists due to significant experimental and computational-based advances. These developments allowed the elucidation of complex structural architectures that were previously impossible to study due to the target’s complexity or sub-cellular localization. Despite all the recent cryo-EM successes, though, most of the structures of large protein systems that were resolved by such experiments exhibited high-resolution features mostly towards the protein’s core. At the same time, more flexible peripheral domains were usually poorly imaged. Such cases were very prominent in cryo-electron tomography (cryo-ET) applications where the solved structures typically range from mid-to-low resolution, thus hindering the process of understanding the function of these complexes at the atomic level, especially in their native cellular context. A state-of-the-art method that has been increasingly applied to obtain structural models of large protein assemblies is computational integrative/hybrid structural modeling. A typical integrative modeling workflow combines existing information from complementary techniques, such as Xray crystallography, cryo-EM, NMR, cross-linking mass spectrometry, or homology modeling to produce ensembles of models with higher accuracy and precision compared to models produced by individual computational or experimental approaches. However, the currently available integrative modeling software and relevant protocols are either limited to complexes of simple architectures or require extensive custom modifications, and thus major expertise from the users, when applied to more complex protein systems. In this Thesis, I describe the functionalities and algorithmic novelties of Assembline, a versatile modeling package for efficient and accessible integrative modeling of protein complexes with very complex architectures. Assembline is implemented using the Integrative Modeling Platform library, on top of which it includes custom algorithms for efficient sampling of the conformational space, versatile protein system configuration language, and graphical interface for input preparation, modeling, and analysis, as well as additional custom restraints. Flexible and symmetrical modeling are also supported, hence establishing Assembline as one of the most straightforward and intuitive software for integrative modeling currently available. I successfully applied Assembline functionalities to build structural models of the nuclear pore complex (NPC) for two yeast species. In the first case, the S. cerevisiae NPC models I built were primarily based on in-cell cryo-ET datasets and revealed striking structural differences compared with the published integrative NPC model built earlier based on in vitro data. The new in-cellbased NPC models appeared significantly dilated compared to their in vitro-based equivalents. At the same time, major structural elements comprising them, such as the mRNA export platform, seemed reoriented entirely in the in-cell NPC models. For the second case, I modeled a significantly less studied NPC from S. pombe. These integrative NPC models were also based on in-cell cryo-ET maps acquired under native and energy depletion cellular conditions to capture structural snapshots of possible conformational changes. The S. pombe NPC models revealed the unusual cytoplasmic side of this NPC that does not form a continuous ring, which challenges the long-standing dogma of the conserved three-ringed NPC architecture. Additionally, the NPC models from the energy-depleted cells revealed a significantly constricted architecture compared to the native NPCs and highlighted the structural plasticity of these complexes in response to physicochemical cues. Both modeling applications on the NPCs highlight the need to study the structures of such important protein complexes in their native cellular context. They are striking examples of the combination of state-of-the-art experimental methods with the newly developed integrative modeling methods.