van der Mers (vdMs), a unit of protein structure, facilitate de novo design of proteins that bind small molecules.
Protein design strategies have seen tremendous improvements. It is possible to design proteins that fold into desired tertiary structures, and attempts are being made to optimize the binding interfaces. However, design strategies often include rounds of experimental library screening or directed evolution, and the de novo design of proteins from scratch that not only fold successfully but also specifically bind a small molecule of interest is challenging. The design of ligand-binding proteins requires amino acid side chains to interact favorably with the target small molecule, but traditional approaches have had limited success, especially with the design of polar cavities that can bind hydrophilic molecules.
Nick Polizzi and Bill DeGrado from the University of California, San Francisco wanted to be able to make proteins that bind to virtually any small molecule or molecular surface. To this effect, they developed the concept of a van der Mer (van der Waals contact + rotamer), a unit of protein structure, that defines the placement of the backbone atoms of a contacting residue relative to key chemical groups in the ligand. They found each amino acid in the nonredundant Protein Data Bank that interacts with a particular chemical group (for example, the carboxamide of a glutamine or asparagine residue) and then extracted and clustered the backbone and chemical-group coordinates to obtain discrete vdMs. “vdMs build on the concept of rotamers. Rotamers are used to place side chains in energetically favorable orientations relative to the backbone during protein design. vdMs instead allow one to locate the positions of chemical groups that preferentially interact with a given residue type and position the interacting group relative to the backbone,” explain the researchers via e-mail. This allows more rapid and accurate design of binding sites.
The design strategy and search algorithm, called Convergent Motifs for Binding Sites (COMBS), works by dividing the target molecule into a collection of chemical groups that must non-covalently interact with the protein designed to bind to it. Then candidate backbones that might simultaneously bind the chemical groups as they occur in the full target molecule are generated using equations to create helical bundles of the right size and shape to bind the target molecule. Each backbone in the generated set is individually considered to see whether it can collectively engage each of the chemical groups of the ligand; vdMs assist in this process. Once a backbone is found that can position vdMs to engage all the targeted chemical groups, the remainder of the protein sequence is computed using flexible backbone design implemented in Rosetta. Using the approach, the researchers designed six de novo proteins to bind the drug apixaban. Two bound with low-micromolar to sub-micromolar affinity, and a high-resolution X-ray crystal structure of one showed that the drug–protein complex closely resembled the designed model.
“Previous methods did not work well enough to create binders with equivalent affinity, and instead relied on evaluating much larger numbers of designs and then iteratively improving them by yeast display and/or directed evolution,” explains DeGrado. He expects vdMs to be as widely used and influential for protein design as backbone fragments and rotamers have been. “I think our work highlights that by relating the binding function directly to the backbone coordinates, as observed in the PDB, we don’t need to worry up front about the molecular details of the side chain interactions. If a backbone can’t support the given binding function (as assessed by vdMs), then we move on to a different backbone without the time-consuming enumeration of rotamers,” adds Polizzi. They plan on releasing a user-friendly, open-source version of the python-based code and hope that more people will use and improve upon the method.
Polizzi, N. F. & DeGrado, W. F. Science 369, 1227–1233 (2020).
About this article
Cite this article
Singh, A. De novo design of small-molecule-binding proteins. Nat Methods 17, 1073 (2020). https://doi.org/10.1038/s41592-020-00995-3