Cell-in-the-loop pattern formation with optogenetically emulated cell-to-cell signaling

ARTICLE https://doi.org/10.1038/s41467-020-15166-3 OPEN Cell-in-the-loop pattern formation with optogenetically emulated cell-to-cell signaling 1234567890():,; Melinda Liu Perkins 1 ✉, Dirk Benzinger2, Murat Arcak1 & Mustafa Khammash2 ✉ Designing and implementing synthetic biological pattern formation remains challenging due to underlying theoretical complexity as well as the difficulty of engineering multicellular networks biochemically. Here, we introduce a cell-in-the-loop approach where living cells interact through in silico signaling, establishing a new testbed to interrogate theoretical principles when internal cell dynamics are incorporated rather than modeled. We present an easy-to-use theoretical test to predict the emergence of contrasting patterns in gene expression among laterally inhibiting cells. Guided by the theory, we experimentally demonstrate spontaneous checkerboard patterning in an optogenetic setup, where cell-tocell signaling is emulated with light inputs calculated in silico from real-time gene expression measurements. The scheme successfully produces spontaneous, persistent checkerboard patterns for systems of sixteen patches, in quantitative agreement with theoretical predictions. Our research highlights how tools from dynamical systems theory may inform our understanding of patterning, and illustrates the potential of cell-in-the-loop for engineering synthetic multicellular systems. 1 Department of Electrical Engineering, University of California, Berkeley, CA, USA. 2 Department of Biosystems Science and Engineering, ETH Zürich, Basel, Switzerland. ✉email: ; NATURE COMMUNICATIONS | (2020)11:1355 | https://doi.org/10.1038/s41467-020-15166-3 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15166-3 S patial patterning is crucial for the proper functioning of diverse multicellular biological systems from slime molds1 to developing embryos. The ability to synthetically engineer multicellular patterning will facilitate advances in designing microbial communities2–4, creating synthetic biomaterials5,6, and programming tissue and organ growth7–10, among other applications11. While recent efforts to synthetically engineer multicellular patterning have met with success (see refs. 12–14 for reviews), relatively few of these efforts15,16 have been guided by quantitative mathematical theory beyond numerical simulation. In contrast, conventional engineering approaches rely on the predictive power of theory both to design complex systems and to build the intuition necessary to envision new capabilities. Future progress in synthetic multicellular patterning will benefit from a firm understanding of the underlying theoretical principles, as well as scalable, efficient methods for implementing—and validating—these principles in practice. Gene expression patterning has received much focus in the theoretical literature17–23, and is also of particular interest in regenerative medicine, since it is central to the early stages of embryonic development and eventual cell fate determination7,24. There are a number of challenges associated with engineering spontaneous gene expression patterning into biochemical systems, including how to facilitate interaction among cells25 and achieve spatial precision in the resulting patterns26–28. Even when successful, these implementations are still constrained by time, expense, and the availability of biological parts satisfying parameter requirements29,30. Moreover, it may be difficult to measure or monitor particular system components in real time, which can hinder debugging and slow down the design-build-test cycle31. While numerical simulation is an important method for efficient prototyping, simulations are only as valid as the models underlying them, and simplifications or faulty assumptions can limit the experimental applicability of simulation results. We propose that future efforts in synthetic patterning would benefit from an intermediate step between pure simulation and full biochemical implementation, which could be used to validate theories or incrementally test synthetic designs before they are fully incorporated into the organism. Inspired by human-in-theloop approaches for engineering systems that must interact with complex, living individuals32, we propose a cell-in-the-loop approach in which physical signaling among cells is substituted with computer-controlled inputs calculated in silico from real- time measurements of gene expression. Cell-in-the-loop, by incorporating live cells into the simulation, eliminates the need to make assumptions about individual cell behavior during dynamic evolution, while retaining flexibility in testing parameters that remain under computational control. These benefits are particularly essential for patterning systems, in which the large number of interacting cells can make detailed simulations prohibitive or impossible. We implement cell-in-the-loop using optogenetics, which have been shown to afford excellent spatiotemporal precision in applications including feedback control33–36, and which were previously used to emulate cell-to-cell signaling for oscillatory synchronization37. We engineer Saccharomyces cerevisiae to respond to blue light38 by increasing gene expression as measured by a fast-acting fluorescent reporter39. We use an optogenetic platform capable of targeting individual cells independently of each other36, such that the light input to any given cell can be calculated based on the gene expression levels of other cells that are interacting with the target cell. Both the network architecture (which cells interact with which) as well as the exact form of interaction are programmed into the computer, allowing us to precisely modulate system parameters related to cell-to-cell signaling. We adapt a general theory for pattern emergence in large-scale lateral inhibition systems40,41 to inform our designs and predict steady-state outcomes. Lateral inhibition regulated by the NotchDelta signaling pathway is responsible for patterning in a range of developmental contexts, including proneural stripe formation42 and subsequent neural precursor selection43 in fruit flies, as well as patterning in the central nervous system44, inner ear45,46, and intestine47 of vetebrates48. Inspired by these systems, we program a computational signaling relation to emulate mutual inhibition among groups of cells and vary the strength of the inhibition by tuning a single digital bifurcation parameter. Once the network architecture and signaling relation are defined, inputs to cells are calculated solely based on measurements of those cells without any further external control, creating a self-contained dynamical system. Using this setup, we visualize gene expression levels of real cells by the brightness of square patches on a virtual grid (Fig. 1). We show spontaneous emergence of contrasting checkerboard patterns in which neighbori (...truncated)