AI-Designed Protein Switches Control Drugs with Speed

“Proteins float around in the cell and constantly bump into each other. How can you design rates at which biology happens fast?” Adam Broerman, a graduate student at the University of Washington (UW), posed as he described his PhD research designing functional coupling in protein systems in an interview with GEN.

Broerman, who works in the lab of Nobel Laureate David Baker, PhD, has recently focused his attention on controlling how tightly proteins bind to partner molecules to drive biological function, including immune signaling, metabolism, and more. While high-affinity binding, such as a drug interacting with a target, can lead to potent biological response, this interaction can take a slow time scale of hours to dissociate, creating a dangerous problem for medicines that must be halted in the presence of harmful side effects.

In a new study published in Nature titled “Design of facilitated dissociation enables timing of cytokine signalling,” Broerman, along with colleagues from UW, have now designed a generalizable molecular on/off switch that facilitates binder dissociation rates as high as 6,000-fold. The work offers a powerful tool for safer medicines that quell harmful side effects and activate drugs on cue, and sensitive biosensors for applications, such as viral testing.

“Protein design has focused on stable states and binding interactions,” said Baker, who is also a Howard Hughes Medical Institute (HHMI) investigator and the director of the Institute for Protein Design at UW, in an interview GEN. “The big advance here is designing the kinetics of interactions, how fast they occur, not just the thermodynamics, or interaction stability.”

Baker was awarded the 2024 Nobel Prize in Chemistry for his work in pioneering AI for protein design. In recent years, his lab has developed RFdiffusion, a diffusion-based AI model that generates proteins from scratch (de novo) to enable broad applications, including drugging historically “undruggable” intrinsically disordered proteins and designing highly specific protein binders to enhance the immune system’s ability to detect and fight disease.

Power stroke

The study applied the molecular switch to interleukin-2 (IL-2), a central immune cytokine investigated in cancer therapy and notorious for toxic side effects, and showed that IL-2-activated human immune cells were silenced on demand upon addition of an effector molecule. The new form of control could make cancer therapy more tunable, allowing physicians to administer high-dose, short-duration treatments to improve cancer-killing results.

Baker told GEN that the molecular switches demonstrated in the study are a facilitated dissociation proof-of-concept and not yet positioned for application in patients. However, the work provides a path for developing therapeutic safety mechanisms after drug administration or localized drug activation in disease-relevant areas, such as the tumor microenvironment.

The switch also significantly improved a SARS-CoV-2 sensor that responded approximately 70 times faster than previous protein-based tests for coronavirus. The same approach could yield rapid sensors for disease markers, environmental pollutants, and other important chemicals.

To achieve dissociation, the designed molecular switches bind to the excited state of the target molecule. The addition of an effector allosterically controls the switch to force the binder apart. The switch applies an induced fit power stroke mechanism, in which the fit of the effector is stabilized along each step of the conformational change and generates force akin to how motor proteins, such as myosin and kinesin, function as they walk along tracks.

Prior approaches for interfering with binder interactions have included direct steric overlap, which restricts design efforts to stay near the interaction site, and competitive binding, where dissociation rates are limited by the affinity of the existing binder. In contrast, the study’s allosteric approach provides a modular framework that is generalizable across broad binder interactions and designs rates directly to achieve speed.

Future directions of the study will design switches that respond to therapeutically relevant effector proteins, such as tumor marker proteins, to support drug activation during disease-relevant conditions.