Spider silks represent a subclass of intrinsically disordered proteins (IDPs) that are challenging to study because they are highly repetitive, extremely large, and contain a central region that lacks a single native structure. These difficulties diminish the usefulness of traditional techniques such as nuclear magnetic resonance (NMR) which suffers from extreme spectral overlap in this case. By applying multiple NMR techniques and several approaches to simulating fragments of the MaSp1 spidroin silk, we were able to build a statistical model showing how recurrent geometric motifs throughout the IDP ensemble explain the NMR data.

We used a library of multidimensional ¹H-detected and X-detected NMR experiments, which better resolve the overlapping regions of the spectrum, to generate distance, phi/psi angular and dynamical time constraints for short regions of the large central repeat units of the spidroin proteins. These regions can be modelled using geometry optimisation (OPTIM, GMIN and PATHSAMPLE) which offers a powerful method for efficiently generating a large database of minima and transition states. In-house software (using Sqlalchemy) was written to store NMR constraints and simulations and concurrently perform geometric analysis to explain the data.

Starting structures were generated using Alphafold (<48hrs computational time) and an atomistic conversion of a coarse grain Gromacs Martini model (4 weeks). Finding a pathway between these states created a large set of minima from the space of disordered configurations of the full protein. Many simulations of smaller fragments of the full protein were also included in the database. Preliminary analysis of these data revealed a basin hopping move to explore the energy landscape more broadly using GMIN.

NMR experiments were performed on labelled natural and recombinant silk and silk fragments corresponding to the simulated fragments. Chemical shift assignments were made for fragments of the protein sequence and these were fed into TALOS-N for phi/psi angle predictions. These predictions combined with homonuclear nuclear Overhauser effect (NOE) measurements were analysed to generate initial NMR structures. Backbone relaxation data provided information on the overall chain dynamics of the different fragments.

We are able to explain the observed characteristics of the NMR data in terms of recurring geometric features that persist across the landscape of these intrinsically disordered proteins.

Chris Forman

Appearances