It’s the little things that count – so let’s count the little things

To design new drugs and gain a deeper understanding of how and why biomolecules behave the way they do, researchers need to know what they look like. A 3D protein structure gives scientists the image they need and can provide insights into the mechanisms behind how these molecules function.

3D structures are created by averaging images from thousands of copies of that protein, usually collected via X-ray crystallography or cryo-electron microscopy. However, these structures predominantly represent snapshots of a protein frozen in time. It might showcase how the protein appears 95% of the time, but what about the other 5% that isn’t being observed?

Researchers at St. Jude designed a new tool to look at alternative conformations (shapes) of amino acids, the individual components that make up protein structures and identify key differences in seemingly identical HSP90 isoforms (functionally or structurally comparable proteins that have a similar but not identical amino acid sequence).

With this tool, the researchers, led by Marcus Fischer, PhD, St. Jude Department of Chemical Biology and Therapeutics, can view the invisible of protein structure and highlight not just the less populated states of proteins but differentiate between seemingly identical structures based on minute differences in how they interact with their environment.

“Our work goes beyond frozen structural snapshots towards exploring regions of the protein conformational landscape that have remained unchartered,” said Fischer. “Targeting those rare states can provide opportunities to achieve selectivity and, hence, reduce side effects, even in structures that appear nearly identical.”

“FLEXR” algorithm captures missing conformations

To address the issue of missing alternative amino acid conformations, scientists from the Fischer lab designed a new structural biology tool called FLEXR. The tool, published in Acta Crystallographica Section D, carefully skims through the electron density map, the blueprint generated from X-ray crystallography experiments, which allows scientists to build the 3D structural model. It highlights features of structures that researchers may have missed in manual model building. Users of structural models rarely look back at those maps for missing features but trust the models. Using the electron density maps, FLEXR is agnostic to what was included in the model. It will call attention to a missing amino acid conformation which might also point in a different direction and interact with different neighboring amino acids or ligand atoms than what was assumed.

This is a hugely valuable tool for drug design, opening new protein conformations for researchers to target. But it is also a tremendous help for researchers working on 3D protein structure elucidation.

“This tool bridges the gap for crystallographers,” said first author Tim Stachowski, PhD, St. Jude Department of Chemical Biology and Therapeutics. “They might have these subtle features in their electron density map, and they either overlook or miss them when they are modeling in the protein.”

Unlocking HSP90 for drug design

The practical uses for FLEXR are significant when you consider the nearly 200,000 protein structures submitted to the Protein Data Bank (the premier resource for scientists to download and examine the structures of proteins), with many copies of the same protein in different states or bound to different drugs.

Heat shock protein 90 (HSP90) is one of those superstars of the structure world, being the 12^th most characterized protein in the databank and a highly sought-after target for cancer treatment. This high demand is due to the wide net it casts with respect to function. As a molecular chaperone (a protein that aids in folding other proteins), any dysregulation of it has downstream effects throughout the cell, such as in cell cycle control and many signaling pathways. Despite HSP90 being so well studied, no approved therapeutic targets it.

“The appeal to studying HSP90 was the availability of vast amounts of structural data, from which one may conclude there is nothing more to learn about this biomedical target,” explained Fischer. “This is contrasted by the lack of a drug approved for use by the US Food and Drug Administration (FDA) despite decades of intense research efforts in industry and academia.”

HSP90 has four isoforms of the protein, creating a structural and chemical research problem. The unique functions of each isoform confound an uncanny structural resemblance. How can scientists target one HSP90 isoform in one pathway without targeting them all and causing chaos?

To tackle this, the Fischer lab solved the structures of all four human isoforms bound to the same ligand and looked very closely at those small population states that researchers often miss. Published in Protein Science, they found that the changes in how each isoform bound a ligand can come down to the most minute details, even the position of water molecules.

“Slight changes in the positions of waters or changes in the conformation of amino acid side chains can give you very different results for ligand binding,” Stachowski explained, “so, for example, you can design ligands that hydrogen bond with a water molecule in just one isoform.”

Improving the way scientists use protein structures for drug discovery

This research pushes forward the idea of carefully revisiting “old” drug targets with a new mindset. It also encourages researchers to look at new protein structures more rigorously and explore the untapped potential of the less populated states of protein structures.

As Fischer put it, “Combining a careful, apples-to-apples approach with new tools to extract hidden signatures of binding sites can inform our ability to target old targets in new ways.”