Making invisible excited-state structures of pro-interleukin-18 visible by combining NMR and machine learning

Area of Science:

• Biophysics
• Structural Biology
• Computational Biology

Background:

• Nuclear Magnetic Resonance (NMR) spectroscopy is vital for studying protein dynamics across vast timescales.
• NMR can detail transient, high-energy protein conformations but often lacks sufficient data for complete structural ensembles.
• The precursor form of interleukin-18 (pro-IL-18) exhibits sparsely populated, transient excited-state conformations that are difficult to structurally resolve using NMR alone.

Purpose of the Study:

• To develop a novel protocol integrating NMR data with generative machine learning to characterize rare protein conformers.
• To identify and structurally define the transient excited-state conformations of pro-IL-18.
• To combine experimental and computational approaches for a more comprehensive understanding of protein energy landscapes.

Main Methods:

• Utilized NMR spectroscopy to probe protein dynamics and identify regions of conformational exchange.
• Employed the generative machine learning model AlphaFlow to predict potential protein structural ensembles.
• Developed a protocol to select candidate conformers from AlphaFlow predictions based on NMR data, followed by experimental validation.

Main Results:

• Identified two distinct, sparsely populated (<0.5%) excited-state conformations of pro-IL-18 with millisecond (ms) lifetimes.
• Localized the conformational exchange to a pair of short β-strands within pro-IL-18.
• Successfully validated selected conformers against experimental NMR data, confirming their correspondence to the observed excited states.

Conclusions:

• The integrated approach of NMR and machine learning (AlphaFlow) effectively characterizes transient protein conformations.
• This combined strategy provides a more complete picture of the pro-IL-18 energy landscape than either method alone.
• The developed protocol offers a powerful new tool for studying challenging protein dynamics and rare conformational states.