Autonomous and Assisted Control for Synthetic Microbiology

Abstract

The control of microbes and microbial consortia to achieve specific functions requires synthetic circuits that can reliably cope with internal and external perturbations. Circuits that naturally evolved to regulate biological functions are frequently robust to alterations in their parameters. As the complexity of synthetic circuits increases, synthetic biologists need to implement such robust control "by design". This is especially true for intercellular signaling circuits for synthetic consortia, where robustness is highly desirable, but its mechanisms remain unclear. Cybergenetics, the interface between synthetic biology and control theory, offers two approaches to this challenge: external (computer-aided) and internal (autonomous) control. Here, we review natural and synthetic microbial systems with robustness, and outline experimental approaches to implement such robust control in microbial consortia through population-level cybergenetics. We propose that harnessing natural intercellular circuit topologies with robust evolved functions can help to achieve similar robust control in synthetic intercellular circuits. A "hybrid biology" approach, where robust synthetic microbes interact with natural consortia and-additionally-with external computers, could become a useful tool for health and environmental applications.

Keywords: control; cybergenetics; microbial consortia; relative sensing; robustness; synthetic biology.

Figure 1

Relative sensing of population parameters. (A) Theoretical perfect ratio sensing. The mean gene expression output of a reporter population (Response) is insensitive to changes in the total density of the co-culture, but sensitive to the relative abundance of the individual populations. (B–D) Natural microbial intercellular signaling networks composed of distinct cell populations that perform ratio sensing. A stimulatory signal (red circle) accumulates in the media in proportion to the density of the signal emitter and stimulates green fluorescent protein (GFP) production (green). The concentration of the signal also depends on antagonistic activity, which balances out activation. For this, S. cerevisiae (B) uses an extracellular protease (yellow), which directly degrades the signal produced by partner cells. B. subtilis (C) depletes the signal by internalizing it through active pumps (purple) and degrading it (∅ symbol) internally. In Enterococcus faecalis (D), cells carrying a conjugative plasmid (dotted line) produce an inhibitory signal (blue circle) from a plasmidial gene (blue arrow), which antagonizes the interaction between the signal produced by plasmid-free cells (red) and its cognate transcription factor (not shown). The thick arrows correspond to the genes encoding the corresponding products. (E) Ratiometric sensing of distinct extracellular signals in mammalian cells. One of the two signals (blue) forms receptor–signal complexes with low activity, while the other (red) forms high-activity complexes, such that one signal (blue) competitively inhibits activation by the other stronger ligand (red). Receptors are represented by the purple and orange symbols. Thick and thin arrows pointing to GFP (green) represent high and low activity of the receptor complexes, respectively. (F) Synthetic intercellular toggle switch. Signals (blue and red) produced by each cell from their respective genes (thick blue and red arrows, respectively) inhibit the production of signals by the other cell in a co-repressive circuit, via signal-specific expression of a transcriptional repressor (its coding gene is shown in grey). In this case, the per-cell output level is directly proportional to the ratio of the blue signal-producing strain to the red signal-producing strain (“majority wins”). The opposite pattern (“minority wins”) can be obtained by changing the circuit such that GFP is directly inducible by the red signal instead of repressed by the induced transcription factor (grey).

Figure 2

Internal and external control strategies from cybergenetics. (A) Computer-aided control. A network of interacting molecular components (circles) with a fluorescent protein output (green) interacts with a computer, which measures and acts on the output by delivering network inputs. (B) Dynamic compensation. Native (blue) and negative-feedback knockout (black) outputs for a signaling pathway, compared to computer-controlled negative feedback expression (orange). In this case, regular input pulses (bottom) restored wild-type behavior. (C) Application of external control to maintain a bistable system—in this case, a synthetic toggle switch circuit—close to its unstable equilibrium point. The circuit is switched on or off (blue) using two specific chemicals (arrows). By maintaining one input at roughly constant levels (not shown) and adding the other periodically (bottom), the system is maintained at its unstable point and remains undecided (orange). (D) Incoherent feed-forward loop (IFFL) based robust constitutive expression. Constitutive expression machinery activates the target gene (green, GFP) and its repressor gene (pink, transcription-activator-like effector (TALE) protein); the repressor gene is encoded upstream in the cassette and binds non-cooperatively to the target gene. Promoters and terminators are represented by right-angled and T-shaped lines, respectively. Sources of perturbations in the capacity for constitutive expression are shown in the upper three boxes. (E) Adaptation of output levels to an induced step increase (bottom) in the copy number of the incoherent feed-forward loop cassette (IFFL, orange) and a regular non-feedback system (blue). (F) A biomolecular (embedded) antithetic integral feedback controller based on inactivation (∅) by molecular titration (darker blue square, see text) controls the network, which reports a fluorescent output (as in A). The setpoint is determined by the ratio of the titrated elements (brown and beige circles). (G) Robust perfect adaptation of the antithetic integral feedback controller (AIF). Normally, a perturbation (a step function that activates degradation of circuit component; bottom) brings the system to a new steady state (blue). Using the control loop, the system is brought back to its set point (orange).

Figure 3

Ecosystem intervention. (A) Various ecosystems such as a natural water source, the human gut, or a bioreactor are susceptible to interventions using engineered cell populations to enable, e.g., remediation, therapy, or optimization processes, respectively. The engineered cells (green) coexist and interact with the natural microbiota (purple and yellow) via secreted molecules (corresponding colored halos around cells). (B) External control. Synthetic cells (center) can detect and report specific properties of the natural population (right). Data collection and computer-aided analysis (left) can be used to modify the detector itself (e.g., replenish the detector strain to avoid its extinction) or the natural population (e.g., add a specific dose of a species-specific toxin). (C) Internal control. The synthetic cell interacts bidirectionally with the natural microbiota. Using the information collected, the synthetic cell can both control itself (e.g., maintain its relative abundance) and the natural system (e.g., secrete killing agents).