Synthetic biology, driven by advancements in DNA synthesis, automation, and biological part characterization, is increasingly impactful in biotechnology and medicine. However, constructing synthetic circuits remains challenging due to unexpected cross-talk, loading effects, stochastic cellular environments, and lengthy design cycles. Circuits often perform differently in their final application context than during development. Changing culture conditions or host organisms can significantly impair circuit performance. Various strategies mitigate these environmental effects. Improving network architecture enhances circuit reliability, while feedback regulation, mimicking endogenous biological systems, enhances robustness. Feedback regulation has been used to improve bioprocess yields and circuit robustness but lacked perfect adaptation – maintaining constant activity despite external disturbances. Achieving this requires integral action. While theoretical biomolecular controllers (integral controllers using chemical reactions) exist, experimental demonstrations remain scarce, with only a few in vivo and in vitro examples. Translating circuit specifications to biomolecular realizations is complex, requiring consideration of component availability, characterization, and cross-talk. Mathematical modeling helps, but can be inaccurate if the controlled system isn't quantitatively defined. Hardware-in-the-loop (HIL) engineering is a solution for complex real-time embedded controllers, where the controller interacts with a system simulation. This shortens design cycles and minimizes real-system testing. This paper introduces Cyberloop, a hybrid in vivo/in silico framework for testing and optimizing synthetic circuits under realistic conditions. Cyberloop interfaces the in vivo system at a single-cell level with in silico biomolecular controllers, enabling fast, cost-effective prototyping. Using fluorescence measurement and optogenetic light stimulation, Cyberloop closes the loop with the true biological system, eliminating assumptions about the system’s structure or parameters. The framework's fluorescence measurement and optogenetic activation is versatile and applicable to various cell types, allowing parallel targeting of many cells. The paper uses a genetically engineered *Saccharomyces cerevisiae* strain to show how a biomolecular controller's behavior in a deterministic setting changes in a stochastic cellular environment. It examines the Antithetic Integral Control motif's robustness and performance under physiological dilution rates, and analyzes extensions of this motif to enhance closed-loop performance.

The introduction thoroughly reviews existing literature on synthetic biology challenges, feedback regulation in biological systems, previous attempts at implementing integral control in vivo and in vitro, and the use of hardware-in-the-loop (HIL) methods in engineering. It highlights the limitations of previous approaches, such as the lack of perfect adaptation in feedback regulation and the challenges of translating theoretical biomolecular controllers into experimental realizations. The review emphasizes the need for a framework that allows for rapid prototyping and testing of synthetic circuits in a realistic cellular environment, leading to the introduction of the Cyberloop framework.

The Cyberloop framework uses a combination of in vivo experimental techniques and in silico computational modeling. The in vivo component involves using a genetically modified strain of *Saccharomyces cerevisiae* equipped with optogenetic tools for gene expression control and fluorescent proteins for real-time visualization and quantification of nascent RNA counts. The cells are cultured in a monolayer on an agarose pad under a microscope. The in silico component involves a computer simulation of biomolecular controller reaction networks. The workflow between consecutive sampling involves the following steps: 1. Image acquisition: Brightfield and fluorescence images are captured using a Nikon Ti-Eclipse inverted microscope. 2. Image analysis: Cell segmentation and tracking are performed using CellX and custom software. Nascent RNA counts are quantified from the fluorescence images. 3. Controller simulation: Based on the quantified RNA counts, a stochastic simulation of the biomolecular controller network is run using Gillespie's Stochastic Simulation Algorithm (SSA) to determine the light intensity needed for the next stimulation cycle. 4. Light delivery: A DMD-based custom-built blue light projection system delivers the computed light intensity to individual cells. The Cyberloop experiments employed different biomolecular controller designs: Autocatalytic Integral Controller, Antithetic Integral Controller, and Antithetic Integral Rein Controller. The effects of various parameters, such as controller gain, basal production rates, annihilation reaction rates, and dilution rates, were investigated. Data analysis involved computing time-course averages, steady-state values, and frequency responses of single-cell output trajectories. The frequency analysis used Discrete Fourier Transform (DFT) to extract the frequency response from the single-cell output trajectories. The experiments included controls for comparison. Specifics on growth conditions, plasmid and yeast strain construction, culture media and initialization, agarose pad preparation, and imaging and light delivery system are all detailed in the methods section.

The study yielded several key findings: 1. **Autocatalytic Integral Controller:** This controller, while exhibiting perfect set-point tracking in deterministic settings, quickly reaches an absorbing state (V=0) in a stochastic environment, rendering it ineffective. Introducing a low basal production rate of the controller species eliminated this absorbing state, enabling improved performance. 2. **Antithetic Integral Controller:** This controller showed robust set-point tracking in the presence of significant stochastic fluctuations, even with low controller species copy numbers. Increasing the sequestration reaction rate improved transient dynamics. Dilution rates comparable to yeast growth rates had negligible impact on steady-state tracking performance, suggesting the complexity of schemes to mitigate dilution-induced errors might not be necessary in all systems. 3. **Antithetic Integral Rein Controller:** This modified controller showed improved transient dynamics, including faster settling time and lower variance, compared to the original Antithetic controller. However, this improvement was less significant when the output degradation rate was high. 4. **Single-Cell Dynamics:** Frequency analysis revealed that the Antithetic controller can induce oscillatory single-cell output responses under specific parameter regimes. Adding a proportional negative feedback loop to create an Antithetic Proportional-Integral controller mitigated these oscillations.

The Cyberloop framework successfully bridges the gap between purely simulation-based design and trial-and-error biological implementations of biomolecular controllers. The Cyberloop avoids assumptions about the controlled network's structure or parameters by embedding it in its biological context. The study provided guidelines for choosing optimal parameter regimes, demonstrated the impact of stochasticity and dilution on controller performance, and validated the effectiveness of controller modifications. The findings show that even in the presence of significant stochasticity and dilution, robust and accurate control can be achieved through careful selection of the controller design and its parameters. The discovery of oscillatory behavior in single cells highlights the importance of considering single-cell dynamics alongside population-level behavior. The ability to mitigate these oscillations by adding a proportional feedback mechanism offers a valuable strategy for practical implementation.

The Cyberloop framework offers a valuable tool for rapid prototyping and optimization of biomolecular controllers in cellular environments. The study's findings provide practical guidelines for designing robust controllers that perform effectively even under stochastic conditions. Future research could explore the applicability of Cyberloop to other types of synthetic circuits and investigate the effects of different cellular contexts and controller architectures. Further exploration of single-cell dynamics and the development of more sophisticated control strategies remain important areas of study.

The study focused on a specific genetically modified *Saccharomyces cerevisiae* strain and a particular optogenetic system. The findings might not be directly generalizable to other cell types or systems without further investigation. The Cyberloop framework relies on accurate quantification of cellular readouts and precise light delivery, which can be challenging to achieve in all experimental setups. The choice of controller architectures and parameters remains partly subjective, necessitating further exploration to define optimal design principles more comprehensively.