Fig. 4. Characterization and modelling of GAuDi.

a Design. GAuDi cells constitutively express EL222. Blue light triggers expression of Cre and recombination, and expression of ATAF1 transcription factor. ATAF1 drives the overexpression of a FAR1 mutant (that arrests cells in G1) and mScarlet-I, a “marker” for differentiation. An additional copy of ATAF1 driving itself was necessary for effective arrest. b GAuDi ODE model fit. An ODE model for GAuDi was developed phenomenologically. The model was fitted to dynamical data with repeated light stimuli of different durations. Solid circles represent data and solid lines signify model predictions. We note that before t = 0, a persistent non-zero differentiation fraction was present possibly due to leakiness in ATAF1 expression in a subset of the population (Supplementary Note 4). c GAuDi model validation. The fitted model was validated by predicting dynamic responses to light signals that were not used for the fit. Data are represented as in (b). Globally, model predictions were in good agreement with the data (Supplementary Fig. 15). d Setup for single reactor control. Cells harbouring the GAuDi system were continuously cultured in the exponential phase. The culture was set to a target level of differentiation and continuously monitored via automated flow cytometry measurements that were analysed online. The system state was estimated from analysed data and sent to the model predictive control (MPC) module. The MPC module provided an optimized light sequence to maintain the culture at the desired set point (Supplementary Note 5). Due to presence of dead cells, data were filtered and only live cells were used for subsequent analysis. e Single reactor, single strain control of a microbial consortium. Cultures were targeted to 25, 40, and 70% differentiation. Control started at t = 0 h. Circles signify differentiated fractions. Each colour corresponds to a different control experiment and the dashed line reflects the target set point.