Cybergenetic circuits to test composability of gene networks – CyberCircuits

To create hybrid synthetic circuits in which a part of the circuit is replaced by a computer simulation, a number of tools from different scientific fields are required.

Biology:
On the experimental side, circuit parts are constructed using cloning techniques and need to be interfaced with optogenetic systems to allow for light control of cells.

Platforms:
To signal to biological circuit parts we use microscopy platforms equipped with digital micromirror devices that allow one to target light signal at individual cells. To connect cells and computers in real time we make use of U-net based image analysis to segment and track cells on microfluidic chips.

Mathematical and computational methods:
To use virtual circuit parts that exist only in the form of a computer algorithm, we require detailed stochastic models (continuous-time Markov chains) of the circuit parts that the computer is supposed to virtually replace. To calculate efficiently with such models we require fast approaches for approximating the solution of the chemical master equation. To build such models, we require approaches for learning parameters from single cell microscopy data. Bayesian inference methods can be used but require sufficiently efficient techniques to approximate likelihoods for continuous-time Markov chain models.

We have constructed a synthetic circuit library of repressilator versions in which any of the 3 promoters has been replaced by an optogenetic promoter and any of the 3 repressor proteins by a fluoresecent reporter. This provided us with a plethora of biological circuit parts that can be used in cybergenetic circuits.

We have developed a microscopy platform at Institut Pasteur that uses single cell optogenetics to connect yeast cells and computers in real time and we developed appropriate software for operating the platform. MicroMator, is a software environment that extends the microscopy control software MicroManager and allows the user to perform reactive microscopy experiment in which the experimental plan does not need to be pre-specified but can be updated in response to incoming data.

We have developed a mathematical modeling framework and a Python tool for calculating with stochastic models of biochemical processes. The Flips solver approximates the chemical master equation, coupled to population level processes such as growth or selection, with a Fokker-Planck equation and incorporates a numerical scheme for solving the resulting partial differential equations. The software is available at gitlab.inria.fr/dlunz/flips.

We have made significant progress on platforms and mathematical methods and have now all tools available to realize hybrid bio-digital circuits both in bacteria and in yeast. Future efforts will therefore be focused on improving biological circuit parts and bringing experimental and methodological work together to realize the project objectives.

J Ruess, M Pleška, CC Guet, G Tkacik, Molecular noise of innate immunity shapes bacteria-phage ecologies, PLoS Computational Biology 15 (7), e1007168, 2019. doi.org/10.1371/journal.pcbi.1007168

D Lunz, G Batt, J Ruess, To isolate or not to isolate: a theoretical framework for disease control via contact tracing, medRxiv, 2020. doi.org/10.1101/2020.05.26.20113340

D Lunz, G Batt, J Ruess, JF Bonnans, Beyond the chemical master equation: stochastic chemical kinetics coupled with auxiliary processes, HAL-Inria, 2020. hal.inria.fr/hal-02991103

D Lunz, On continuum approximations of continuous-time discrete-state stochastic processes of large system size, HAL-Inria, 2020. hal.inria.fr/hal-02560743

ZR Fox, S Fletcher, A Fraisse, C Aditya, S Sosa-Carrillo, S Gilles, F Bertaux, J Ruess, G Batt, MicroMator: Open and Flexible Software for Reactive Microscopy, bioRxiv, 2021.
doi.org/10.1101/2021.03.12.435206

A key property of engineered systems is their composability: the behavior of a complex network is computable from the wiring diagram and the known characteristics of its constituent parts. Despite of all the advances in systems and synthetic biology, it is unknown if genetic regulatory networks are at all composable yet the answer is fundamental for our ability to understand the behavior of evolved biological networks, as well as to engineer complex new ones.

We have recently developed a unique experimental platform that permits image-based tracking of hundreds of cells in a precisely manipulated microfluidic environment, on-line analysis of the observed data, as well as optogenetic signalling to individual cells. This gives us unprecedented opportunities to perturb, observe and control gene networks in single cells, to study stochastic gene regulation dynamics, and most importantly, to implement real-time communication between a computer and single cells. We can therefore construct hybrid bio-digital circuits in which a part of the genetic network is effectively implemented as a biological network whereas another part exists only virtually in the form of a model in the computer.

In this project, we will make use of these novel technological capacities to resolve or to circumvent what is arguably the most profound problem in quantitative biology: our failure to understand, or even just to predict, the dynamics of non-trivial biochemical processes in living cells and our resulting inability to rationally design complex synthetic circuits that reliably fulfill quantitative specifications. There are thus two major goals that we want to reach: to construct models that explain and predict better how composed circuits function in vivo and to construct synthetic circuits that function in vivo as previously specified.