Local Probes for Collective Motions: Combining 2D-IR Experiments and Molecular Simulation to Characterize Low-Frequency Protein Modes

• Applicant: Professor Dr. Jens Bredenbeck
• Subject Area: Physical Chemistry of Molecules, Liquids and Interfaces, Biophysical Chemistry
• Term: since 2026
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 579751893

Protein motions are central to function, yet not all are equally relevant. Low-frequency vibrational modes below ~100 cm⁻¹ (3 THz) have been proposed to play a role in important aspects of protein function such as energy transport, ligand binding, and conformational changes. In particular, underdamped low-frequency modes (ULFMs) have been proposed to couple to enzymatic reaction coordinates and influence catalysis. Directly characterizing such modes in proteins in aqueous environment is a long standing challenge due to the high density of vibrational states of proteins and water. This project confronts this challenge by combining 2D-IR spectroscopy, site-specific IR labels, molecular dynamics (MD) simulations, and computational spectroscopy using tailored and machine-learned energy functions. IR labels absorb in a spectral window free from protein/water congestion. Their frequency depends on the local environment, thus periodic protein motions, the ULFMs, manifest as periodic modulation of the label frequency. This dynamic information is captured in the frequency–frequency correlation function (FFCF), which we will measure by 2D-IR and compute from MD. Thus, the IR label provides a local spectrum of protein modes coupled to its high-frequency vibration, enabling a “zoom” into selected sites of the protein. Using the label p-CN-Phe, we will target human carbonic anhydrase II (hCAII), for which functional roles of low-frequency modes have been hypothesized. In related bCAII we previously identified low-frequency oscillations from a serendipitously bound molecule at the active site. Our labeling strategy will now generalize this approach, allowing flexible label positioning across proteins and extension to many proteins. For MD, a cluster-based approach describes the potential energy surface of the labels. This involves reference electronic structure calculations for labels with up to 20 surrounding waters and fitting the Lennard-Jones interactions. Electrostatic interactions are represented as a machine learning-based minimal distributed charge model and bonded interactions for the spectroscopic probe are described by reproducing kernel Hilbert space representations. All proteins are simulated fully hydrated, yielding data for computing FFCFs that are directly compared with experiment for validation. Because MD simulations contain the full information about the conformational space sampled, the motions underlying the spectroscopic response can be analyzed and characterized. This joint experimental/computational approach also enables rational design of mutants for optimal label placement as well as suggesting labels best suited for a given position and mode. In summary, the project develops and applies a general, coherent extensible methodology for detecting, characterizing, and interpreting low-frequency protein modes, thereby providing the foundation to rigorously assess their biological relevance.

• DFG Programme: Research Grants
• International Connection: Switzerland
• Cooperation Partner: Professor Dr. Markus Meuwly