Structure search suggests the function of thousands of viral proteins

Over the millennia, viruses have evolved to become excellent manipulators of biology. But there is much that scientists don't know about how viruses work, in part because they have relatively little insight into how viral proteins work. In a new study, scientists in Jennifer Doudna's group at the Innovative Genomics Institute present a database of predicted structures of viral proteins and use them to find the functions of previously mysterious proteins (Nature 2024, DOI: 10.1038/s41586-024-07809-y).

Researchers often learn about the function of an unknown protein from what they know about related proteins with similar sequences. But the sequences of tens of thousands of viral proteins bear little or no resemblance to known proteins. Scientists have found that different sequences can sometimes fold into unexpectedly similar 3D structures. When it comes to studying distant relationships, “structures have the advantage over amino acids in that they are conserved much, much longer,” says Martin Steinegger, a computational biologist at Seoul National University who developed a structural homology search tool.

Postdoctoral fellow Jason Nomburg of the Doudna lab began a study of the unannotated viral proteome by computationally predicting the structures of about 70,000 proteins from viruses that infect eukaryotic cells. He and his colleagues then sorted the proteins into related clusters based on structural similarity. Some of these clusters contained many proteins from distantly related viruses; others contained only single, enigmatic proteins. Some seemed to mimic host proteins. By developing further structural insights, the researchers were able to deduce the function of many more viral proteins than they could have guessed from sequences alone.

One cluster the team looked at in more detail appeared to have the same shape as phosphodiesterases – enzymes that break bonds between nucleic acids – from viruses that infect bacteria. When host cells recognize a molecule that suggests the presence of a virus, they produce nucleotides with unusual linkages, such as cGAMP. These messages trigger STING (stimulator of interferon genes) signals, a pathway that can trigger an immune response and that has attracted the interest of drug developers. Viruses that can disrupt this pathway may be better able to evade a cell's attention.

Nomburg and his colleagues demonstrated biochemically that the cluster of viral phosphodiesterase-like proteins can degrade the nucleotide message molecules. The structure “seems to have been adapted in different ways to attack different substrates,” Nomburg says: Viruses with a DNA genome use it to degrade cGAMP, a cyclic nucleotide from the DNA-sensing STING pathway, while RNA viruses use it to cleave a linear nucleotide from an RNA-sensing pathway. Although virologists had previously observed viral interference with STING signaling, they had never seen so many families of viruses using a similar protein.

Steinegger says the database will be a useful resource for virologists. According to computational biologist Pedro Beltrao of the Swiss Federal Institute of Technology (ETH) in Zurich, this utility could benefit both virologists interested in specific viruses and biologists interested in learning more about the dynamics of evolution. However, he adds that some caution is needed until follow-up studies can confirm a protein's role. Predicted structures run the risk of being inaccurate, and it's possible that proteins take on a similar shape without sharing a common function or evolutionary history.