Modelling AvianGuard

Modeling was a guiding element of our engineering strategy — not a separate task, but an active design tool that shaped our experimental workflow. From the beginning, our models evolved alongside the biological constructs, forming a continuous Design–Build–Test–Learn (DBTL) loop that translated theory into working devices.

From Understanding to Prediction

Our modeling journey began with a focused question: Can we predict and rationally tune promoter activity? The answer came through the quantitative characterization of the pLux promoter, detailed in both the Contribution and Engineering Success section.

There, a dynamic gene expression model was developed and validated experimentally. By coupling model predictions with measured calibrated GFP fluorescence, we identified the kinetic parameters that define promoter activation and translation efficiency (See Contribution). This predictive framework allowed us to design a new device — one that behaved as anticipated — demonstrating how modeling can move beyond explanation to become an engineering instrument (See Engineering Success).

Expanding to System-Level Understanding

Building on that foundation, the modeling presented on this page extends from a single promoter to the complete AvianGuard system. We constructed a compartmental dynamic model that integrates transcriptional activation (LuxR/AHL quorum sensing), protein transport, and intein-mediated cyclization. This model links molecular regulation to macroscopic function — connecting gene expression, protein secretion, and environmental dynamics into one predictive framework.

The same parameter values identified during the promoter characterization were used as inputs, ensuring that our higher-level simulations were grounded in experimentally verified biology. This approach closes the modeling hierarchy: from data-informed promoter behavior to system-wide prediction of passive immunity performance.

A Coherent Engineering Workflow

Each modeling phase informed the next design choice and guided experimental validation. The promoter model enabled accurate parameter identification; the system model used those results to simulate real biological constraints, such as timing, signal accumulation, and protein yield.

By interconnecting both levels of abstraction, we established a unified modeling architecture that supports predictive design across scales — from a single transcriptional unit to the behavior of an engineered chassis.

The model design includes the compartmental model of nanobody production to ensure that we are producing sufficient defences in the medium and long term. This dynamic model maintains the non-cycled fusion-protein (FP) as an intermediate species in the periplasm of the bacteria, prior to cyclisation. This is a realistic approximation of the current level of non-cycled fusion-protein synthesized by our Passive immunity device. This is more biologically realistic, more modular and easier to fit with experimental data.

This approach enables us to decouple:

• Translocation from the cytoplasm to the periplasm.
• Cycling (splicing) of the fusion-protein in the periplasm to generate the cycled functional nanobodies: our Cyclo-bodies.

At the population level, quorum sensing is a cell-to-cell communication mechanism that triggers AvianGuard and its Passive and Active synthetic devices. When there are enough cells in the population, the automatic signal activates the Plux promoter and the expression of the downstream sequences.

where molecules of AHL_ext passively diffuse across the population of N cells following Fick's dynamics.

Inside the cell, AHL_ext rapidly becomes AHL_int. Dimer (LuxR.AHL)₂ activates the inducible promoter Plux as a result of quorum sensing. LuxR is also constitutively produced in our device.

Transcription of the fusion-protein was modelled as a Hill function of order 2. The experimental characterization of the Plux promoter and the dose-curve for different AHL inductions levels is in Engineering DBTL.

Where:

• A_int is the intracellular concentration of AHL.
• R is the concentration of LuxR.
• k_dlux is the dissociation constant from the promoter Plux.
• k_d2 is the dissociation constant LuxR-AHL.
• CN is the copy plasmid numbers.
• α is the basal expression.

The Fusion-protein (FP_cyt) is translated as one compartment and includes the Inteins N and C, as well as the signal peptide Ups45. We use the signal peptide Ups45 to transport FP_cyt from the cytoplasm to the periplasm.

where v_sec describes the rate of a Plux-catalyzed reaction based on Michaelis-Menten kinetics.

In this compartment, the fusion-protein FP_cyt moves across the membrane to its correct cellular location guided by signal Ups45. This process involves the protein-conducting channel and produces molecules of non-cycled fusion-protein FP at the rate v_sec.

For analyzing the optimal induction level, we selected 3 different scenarios:

• No induction (0 AHL)
• Medium induction level (0.1 μM)
• High induction (1000 μM)

There are three important temporal windows: (1) the automatic signal of production through quorum sensing, (2) the number of cells increases in the bioreactor, so AHL_int is diluted and (3) cyclo-body production begins to decrease after the level of AHL_int falls below the required level.

Finally, in the last compartment and after a splicing process (cycling mediated by Inteins N and C), we produce cycled functional nanobodies or our Cyclo-bodies (Cb).

In Figure 3, we have also demonstrated that the production level of cyclo-bodies (Cb) is high (more than 200 nM per cell), since almost all non-cycled fusion-proteins (FP) are correctly cyclized.

Nanobody production titers vary widely depending on expression systems and strategies. Research shows that bacterial expression typically yields 1-25 mg/L, while yeast systems achieve 5-500 mg/L. These differences reflect the complexity of achieving high-level functional protein expression across platforms.

Bacterial expression typically yields lower titers (1-25 mg/L) but offers simpler processing. Advanced strategies like dynamic control systems can push yields higher, with reports of 20 mg/L from E. coli periplasmic expression using optimized two-stage induction protocols. The Chi.Bio micro-bioreactor platform enables precise environmental control for expression optimization.

Several parameters values were estimated for this study, parameters for transcription and translation of proteins can be found in [2, 4].

Table 1: Parameters used to simulate the model with secretion and splicing in the periplasm. Based on literature and values from the reference antithetic circuit together with the values obtained from the Engineering Success.

[1] Yadira Boada, Fernando N Santos-Navarro, Jesús Picó, and Alejandro Vignoni. Modeling and optimization of a molecular biocontroller for the regulation of complex metabolic pathways. Frontiers in Molecular Biosciences, 9, 2022.

[2] Yadira Boada, Alejandro Vignoni, and Jesús Picó. Engineered control of genetic variability reveals interplay among quorum sensing, feedback regulation, and biochemical noise. ACS Synthetic Biology, 6(10):1903–1912, 2017. PMID: 28581725.

[3] Jennifer N. Hennigan, Romel Menacho-Melgar, Payel Sarkar, Maximillian Golovsky, and Michael D. Lynch. Scalable, robust, high-throughput expression purification of nanobodies enabled by 2-stage dynamic control. Metabolic Engineering, 85:116–130, 2024.

[4] Ron Milo and Rob Phillips. Cell biology by the numbers. Garland Science, 2015.

[5] Pardon E, Laeremans T, Triest S, Rasmussen SG, Wohlkönig A, Ruf A, Muyldermans S, Hol WG, Kobilka BK, Steyaert J. A general protocol for the generation of nanobodies for structural biology. Nat Protoc, 85:674–693, 2014.