Microbial adaptation and feedback-driven learning

Feedback-driven adaptation in S. cerevisiae

We study a feedback-driven mode of cellular adaptation that operates between rapid physiological responses and slow genetic evolution. In this regime, cells can explore new gene-regulatory configurations and stabilise them when changes in gene expression are coupled to improvement of their environment.

To probe this experimentally, we developed a microfluidic platform in which Saccharomyces cerevisiae populations control their ambient temperature through a closed-loop feedback system. Real-time microscopy is used to measure expression of a fluorescently tagged gene, and a computer converts this signal into temperature changes according to a predefined reward function. In this setup, stochastic fluctuations in gene expression can be reinforced by stress relief.

Using this system, we showed that yeast populations can stabilise increased expression of a gene and use it to regulate temperature, even though the gene has no native role in thermal stress. Under feedback, populations converge to a new expression state associated with a favourable environment, while identical control populations do not. This constitutes a form of reward-driven regulatory rewiring, in which a gene is repurposed into a novel adaptive role.

We are now using this platform to map how different feedback rules shape adaptation: which reward landscapes lead to reliable convergence, how fast learning occurs, how stable the resulting states are, and whether new regulatory relationships can be induced between multiple genes. Together, these experiments aim to identify general principles by which feedback can canalise stochastic exploration into new, persistent regulatory states.

Adaptation under changing environments

Environmental stress can present at different rates. In ongoing work, we study how this rate shapes evolutionary dynamics in large microbial populations. Using our in-house developed barcoding method to follow yeast populations evolving in turbidostats, we have shown that gradual and abrupt temperature increases, despite reaching the same final value, lead to distinct population trajectories, differences in lineage diversity, and markedly different patterns of clonal turnover.

Abrupt stress promotes large, transient growth advantages that allow low-frequency lineages to rise rapidly and restructure the population, whereas gradual stress preserves diversity and favours more reproducible outcomes dominated by lineages that were already relatively fit. These findings suggest that evolutionary dynamics depend not only on the magnitude of environmental stress, but also on its temporal structure.

By combining lineage-resolved measurements with controlled environmental dynamics, we aim to identify general principles linking environmental history, population structure, and evolutionary outcomes.