Cornell researchers have used artificial intelligence to uncover a previously unknown way cells control how proteins move inside them, a process essential for growth, communication and movement, and one that is often disrupted in cancer.
The study, published Sept.9 in the Journal of Cell Biology, shows that a little-understood protein called Avl9 acts as an “off switch” for another protein, Arf1, which helps direct where materials go inside cells. Turning Arf1 off at the right time is critical as it regulates cellular transport, otherwise cells become disorganized which leads to negative effects. The researchers also found that Avl9 is part of a broader group of proteins that may perform the same function, pointing to a previously unrecognized system cells use to keep this process in balance.
Led by Chris Fromme, professor of molecular biology and genetics at the College of Agriculture and Life Sciences and faculty in the Weill Institute for Cell and Molecular Biology, the team used AlphaFold, an artificial intelligence software that predicts the structures of proteins and how they might interact, to search for previously unknown partners of Arf1.
Instead of traditional lab screening, which can take months, the AI software quickly generated a shortlist of likely protein interactions for experimental testing.
“We discovered an unexpected function of a poorly understood protein, which helps us understand how cells control intraceullar trafficking,” Fromme said.
Arf1 acts as a molecular switch that directs the movement of proteins and other cargo to the right places inside the cell. Because it is involved in many essential processes, scientists expected additional proteins must exist to control its activity.
“Arf1 controls the movement of proteins that need to move from the Golgi to several other destinations in the cell,” Fromme said. “Arf1 is essential for many cellular behaviors, including secretion, migration, and basic internal organization.”
“Arf1 is so important and involved in so many different cellular pathways, we felt there had to be more to how cells control its activity than what was already known,” he said.
The AI analysis pointed to Avl9, which had been linked to secretion and cancer cell movement but whose function was unknown. Laboratory experiments confirmed that Avl9 shuts down Arf1 after it has carried out its role.
“The function of Avl9 was the exact opposite of what you would guess based on what was known about its amino acid composition,” said Ryan Vignogna, postdoctoral associate at the Weill Institute, and first author for the study.
To confirm the finding, the team tested normal and mutated versions of Avl9 in cells. A single amino acid change eliminated its ability to regulate Arf1. In human lung cancer cells, it reduced their ability to move, which linked the protein’s molecular role to a behavior important in cancer.
“These cell biology experiments are important because they show that the cell actually cares about the thing you think is interesting,” Fromme said.
The researchers then examined similar proteins. These proteins, part of a group known as DENN domain proteins, were previously thought to perform as “on switches.” The new findings suggest that some instead act as “off switches” like Avl9, revealing a new role for this family regulating how materials are transported inside cells.
“It means that the previous assumptions were wrong, and since this belongs to a family of related proteins, it prompted us to look for other examples,” Fromme said.
The team confirmed that at least one related human protein behaves in the same way, suggesting this is part of a broader system cells use to regulate internal transport.
The findings may help researchers better understand how disruptions in this system contribute to disease. In cancer, changes in how cells move and transport materials can influence how tumors grow and spread.
“The discovery of this new class of proteins helps other researchers design of experiments to characterize related proteins in cells and disease,” Fromme said.
Looking ahead, Fromme said the study shows how artificial intelligence can speed up discovery by helping researchers quickly identify promising biological questions.
“I think it allows us to ask more exploratory questions with lower effort, and so we can cast a wide net looking for interesting things, and then choose which ones we think are important enough to spend time on with in-depth experiments,” he said.
The study was funded by the National Institutes of Health and supported by the Imaging Facility at Cornell’s Institute for Biotechnology.
Stephen D’Angelo is the communications manager for biological systems at Cornell Research and Innovation.
