Light-controlled evolution produces proteins that switch, oscillate, and compute

Directed evolution, the technique that won the 2018 Nobel Prize in Chemistry, has a blind spot. It excels at producing proteins that are always strongly active, because traditional selection methods apply constant pressure favoring maximum performance. But biological signaling does not work that way. Cells need proteins that turn on briefly, then off, then on again. They need molecular switches that change states on cue and logic gates that integrate multiple inputs before making a decision.

A team at EPFL, led by Sahand Jamal Rahi, has now developed a directed evolution method that can select for exactly these dynamic, multi-state protein behaviors. They call it optovolution, and it works by linking protein function to cell survival through the cell cycle, using light to control the timing. The study was published March 6, 2026, in Cell.

Rewiring the cell cycle as a fitness test

The system is built in budding yeast (Saccharomyces cerevisiae). The researchers rewired the yeast cell cycle so that cell division depended on the protein being evolved. The protein's output signal was connected to a cell-cycle regulator that is essential at one stage of division but toxic at another.

This creates an elegant pass-fail test. If the protein stays on too long, the cell dies from the toxic effect. If it stays off too long, the cell stalls because it cannot progress through division. Only cells in which the protein oscillates correctly, switching on and off at the right moments, survive to divide. Each roughly 90-minute cell cycle acts as one round of selection, and the process runs continuously without manual screening.

The researchers controlled the protein's behavior using optogenetics, delivering precisely timed light pulses to flip the protein between states. This external control sets the selection criteria: the light pattern defines what switching behavior the protein must exhibit to keep its host cell alive.

Nineteen new variants, including green-light responders

The team first applied optovolution to a widely used light-controlled transcription factor. The method produced 19 new variants with improved properties. Some were more sensitive to light. Others had lower background activity in the dark, reducing unwanted gene activation. Most notably, several variants responded to green light rather than only blue, a shift that researchers in the optogenetics field had considered extremely difficult to engineer based on how these proteins absorb light.

Evolution, apparently, found structural solutions that rational design had not anticipated.

A red-light system that lost its chemical crutch

The researchers also evolved a red-light optogenetic system to eliminate its dependence on an externally supplied chemical cofactor. In the original system, yeast cells needed to be fed a specific light-sensitive molecule for the red-light response to work. The evolved variant carried a mutation that disabled a normal yeast transport protein, which unexpectedly allowed the system to use light-sensitive molecules already present inside the cell.

This is a practical improvement. Requiring an external cofactor adds complexity and cost to experiments. Removing that requirement makes the red-light system significantly easier to deploy.

A single-protein computer

The most ambitious demonstration was evolving a transcription factor that behaves as a logic gate. The resulting protein activated genes only when two different inputs were present simultaneously: one light signal and one chemical signal. Either input alone was insufficient. This AND-gate behavior emerged from evolution rather than from the kind of modular protein engineering that synthetic biologists typically use to build such circuits.

Evolving computational behaviors in single proteins is significant because it suggests that complex information processing can be achieved with simpler molecular components than previously assumed.

Beyond light-sensing proteins

While the demonstrations focused on optogenetic tools, the method is not limited to proteins that sense light. Any protein whose activity can be toggled by an external signal, whether light, chemical, or temperature, could in principle be subjected to optovolution. The key requirement is that the protein's switching behavior can be linked to the yeast cell cycle in a way that creates survival pressure for correct dynamics.

This opens possibilities for evolving signaling proteins, molecular sensors, and biological circuit components that need to respond to changing conditions rather than simply performing one function at maximum output.

Yeast proof of concept

The system currently operates in yeast, which limits its immediate applicability to proteins that function in this organism. Extending optovolution to mammalian cells or bacteria would require adapting the cell-cycle coupling mechanism to those different biological contexts. The method also relies on the protein of interest being expressible and functional in yeast, which is not always the case for proteins from other organisms.

Still, the conceptual advance is clear. By making dynamic protein behaviors continuously selectable inside living cells, optovolution brings directed evolution closer to the way biology actually works.