Cellular prediction during variation in carbon availability

When sugar runs out, cells of budding yeast that shift their metabolism from fermentation to respiration exhibit faster growth and improved adaptation. Those cells capable of predicting future scarcity by sensing both the concentration and variation of nutrients should therefore have less lag in growth.
To address whether cells can predict future environments, we developed new tools: a microfluidic device combining single-cell ALCATRAS traps with a serpentine micro-mixer for high-resolution imaging during dynamic but controlled sugar concentrations, and a convolutional neural network for efficient, accurate, high-throughput quantification of transcription factor relocation.
Using these tools together and by following translocating transcription factors, we show that yeast cells likely predict low glucose levels typically an hour before they happen.