Project design
Safety is reflected not only in the experimental process but also in the profound consideration of project outcomes during the project design phase. With regard to the treatment of T2DM using live biotherapeutics, our project regards safeguarding human health as a prerequisite, minimize potential harm to the biological environment, and further explore its therapeutic efficacy while investigating potential economic benefits. Therefore, we incorporated biosafety design as a significant module of our project from the very beginning of GlucoXpert's design. Through continuous iterative optimization, from chassis selection to genetic circuit design, we are committed to developing intelligently and precisely controllable live biotherapeutics.
Chassis Selection of Escherichia coli
Though several instructors noted that E. coli Nissle 1917 is unsuitable for long-term intestinal colonization in humans and present at low abundance in the gut microbiota, we decided to primarily focus our follow-up wet-lab experiments on E. coli for developing our microbial agents [1], given our limited timeframe, and considering that EcN has been widely applied in gastrointestinal disease treatment studies, with its safety profile fully validated in such research [2].
Also, E.coli is also a common laboratory strain frequently used by iGEM teams over the years. Its extensive research history and clear genetic background make it very suitable for genetic engineering and carrying heterologous genes.
Our experiments involve diverse E. coli strains, including TOP10, Trelief® α, BL21(DE3), and EcN. For experiments related to genetic circuits and biosafety where frequent plasmid construction and circuit function verification are required, with no high demands for protein expression levels, we primarily used commercial competent cells including TOP10, Trelief® α as chassis strains for characterization.
For experiments focused on the protein module, which require secretory expression and purification of proteins, while it is also essential to preliminarily confirm whether EcN can synthesize and secrete GLP-1, we therefore first employed BL21(DE3) as the plasmid expression host. After verifying the plasmid's function, we introduced the plasmid into EcN via electroporation, and subsequently characterized EcN's capacity to synthesize and secrete modified GLP-1.
In summary, we will characterize different E. coli strains and preliminarily verify GLP-1 expression in EcN. If follow-up experiments proceed smoothly, we will continue optimizing the experimental design for EcN, and move to further advance wet-lab work using other low-risk strains with better intestinal colonization, while simultaneously follow up on safety verification.
Design of the OR-gate Kill-Switch
While EcN is not inherently hazardous, the genetic elements we introduce into it and the engineered microorganisms themselves could lead to horizontal gene transfer (HGT). This poses potential ecological risks, and in severe cases, could threaten public health and the surrounding environment. By summarizing recent synthetic biology research and biosafety designs from iGEM teams, we categorize biosafety control strategies as follows [3].
Table 1. Various biosafety control strategies
| Strategy | Example |
|---|---|
| Auxotroph |
• dapA knockout auxotrophy • thyA knockout auxotrophy |
| Suicide regulatory gene |
• ATc-regulated Deadman suicide gene switch • LacI-GalR fusion transcription factor-based Passcode suicide gene switch |
| Toxin-antitoxin systems |
• Temperature-responsive CcdB-CcdA toxin-antitoxin pair • pH-responsive Doc-Phd toxin-antitoxin pair |
| Genetic element disassembly |
• Geneguard plasmid host-dependent system • AHL-regulated quorum sensing system |
| DNA degradation |
• Cas9 system targeting exogenous genes • Temperature-responsive intein-DNA endonuclease system |
| Mutation reduction |
• Genomic insertion sequence (IS) element knockout • Increased target gene copy number • JVCI-syn3.0 |
| Physical isolation |
• GelMA-encapsulated engineered bacteria • Alginate and HA-EGCG encapsulated engineered bacteria • Engineered bacteria-integrated electronic diagnostic devices |
To achieve dual active-passive regulation of engineered bacterial survival and further enhancing biosafety, we designed a dual-input "OR-gate" kill-switch responsive to arabinose and temperature signals. Following multiple rounds of iterative optimization of the circuit, we simultaneously further explored the feasibility of integrating novel biosafety regulatory mechanisms, aiming to propose a more forward-looking genetic circuit design.
1st-Generation Kill Switch Design
First, we introduced TlpA*, a passively regulated temperature-sensing protein derived from Salmonella. After sequence optimization, the half-maximal fluorescence value (EC50) of this protein is regulated to 35.6°C. When the temperature is below 35°C, TlpA* can undergo dimerization, which in turn blocks the transcription of its cognate promoter PtlpA[4].
To construct the "OR-gate" logic circuit, we introduced the arabinose operon as an active regulatory element. Specifically, by placing the repressor protein AraC [5] under the PtlpA-tlpA* self-repressive regulatory system [4], controlling the expression of the downstream toxin gene ccdB via the PBAD promoter, and regulating the expression level of the antitoxin gene ccdA to ensure that the engineered bacteria function normally in the dynamic in vivo host environment [6], the first-generation biosafety genetic circuit can be constructed, ultimately achieving the programmed suicide of EcN.
Figure 1. Schematic diagram of the 1st-generation kill switch
2nd-Generation Kill Switch Design
During the experiment, we found that this genetic circuit failed to achieve clear temperature-dependent discrimination between 30 °C and 37 °C under arabinose-free conditions. Specifically, the fluorescence intensity of the downstream reporter protein GFP remained low at 30 °C, and the overall signal intensity was relatively weak. We hypothesized and initially verified (see our Engineering Page) that this was due to the excessively strong regulatory effect of the PtlpA-tlpA* self-repressive system under low temperatures, resulting in almost no expression of the repressor protein AraC. Upon further literature review, we discovered that AraC exerts a dual role on the PBAD promoter: Activation of PBAD expression requires the simultaneous presence of AraC and arabinose [5].
To address this, we expressed AraC independently via a constitutive promoter, and constructed a complex circuit by integrating the Tet (tetracycline) operon, insulator element RiboJ, and dual-promoter "OR-gate" logic, ultimately developing the second-generation biosafety genetic circuit.
Figure 2. Schematic diagram of the 2nd-generation kill switch
3rd-Generation Kill Switch Design
Previously, we found that the conformational regulation mechanism of the arabinose operon is relatively complex, so we planned to select other operons with simpler regulation principles. Through literature and patent searches, we identified a type of fucose operon [7] that not only possesses the safe and edible property but also seems to be simpler. Based on this, we replaced the original arabinose operon, constructed the third-generation biosafety genetic circuit using the fucose operon, and conducted characterization experiments.
Figure 3. Schematic diagram of the 3rd-generation kill switch
4th-Generation Kill Switch Design
During the experiment, we found that in the second-generation circuit, even when the expression of the TetR repressor was regulated by temperature under arabinose-free conditions, the dual-promoter system could not be activated. We hypothesized that the repressor protein AraC would still block the transcriptional regulation of the Ptet promoter, resulting in the failure of downstream gene expression.
For the third-generation kill switch, due to repeated failures in plasmid construction, the fucose repressor FucR could hardly be transformed and expressed in E. coli TOP10. It was assumed that this repressor might be toxic to TOP10, so characterization of this circuit was not continued.
Based on these issues, we constructed the fourth-generation circuit. We attempted to relieve the excessive regulation of the dual-promoter system by AraC in the absence of arabinose via adjusting the expression intensity of the AraC repressor. The constitutive promoter J23105 in the original circuit was replaced with J23117 and J23115 to improve temperature response sensitivity. However, this also reduced the detection limit of arabinose to some extent, which could easily cause regulatory leakage of arabinose in daily diet and lead to erroneous suicide of the engineered bacteria.
Figure 4. Schematic diagram of the 4th-generation kill switch
5th-Generation Kill Switch Design
After integrating the aforementioned genetic circuits, by regulating the expression of the repressor protein AraC, we have essentially achieved temperature-dependent suicide regulation in the presence of arabinose. For the arabinose-free state, we simultaneously optimized the dual-promoter tandem system by replacing the upstream Ptet promoter with the constitutive promoter J23101, thereby enabling the system to normally complete transcriptional read-through in the absence of the repressor protein AraC. When the temperature reaches 37℃ and AraC is expressed, the expression of downstream suicide genes will be repressed.
Figure 5. Schematic diagram of the 5th-generation kill switch
Future Design
1. Bacterial Autolysis System
In the GLP-1 protein engineering module, we achieved GLP-1 secretion by fusing a signal peptide and cell-penetrating peptide to the protein. However, some of our advisors pointed out that this method could not effectively release GLP-1 extracellularly. Additionally, we recognized that due to the inherent defects of the temperature-sensitive promoter, the aforementioned biosafety designs cannot completely prevent the escape of a small number of engineered bacteria. Combining these two considerations, we came up with a bacterial autolysis technology to synchronize engineered bacteria suicide and GLP-1 release.
Therefore, we designed a bacterial autolysis circuit based on the bacterial quorum sensing (QS) effect [8][9], with the LuxI/LuxR system from Vibrio fischeri. The LuxI protein catalyzes the synthesis of the signaling molecule N-acylhomoserine lactone (AHL). Since AHL can diffuse freely between adjacent cells, the intracellular AHL concentration in engineered bacteria increases synchronously with bacterial density. When the concentration reaches a certain threshold, it binds to and activates the LuxR transcriptional regulator, which in turn induces the expression of the phage lysis gene φX174E under the control of the PluxI promoter.
The φX174E gene from λ phage encodes protein E, an inhibitor of translocase MraY. MraY is an essential membrane-embedded enzyme that catalyzes the formation of the peptidoglycan precursor Lipid I, by combining UDP-N-acetylmuramic acid-pentapeptide with the lipid carrier undecaprenol phosphate. It plays a crucial role in the synthesis of peptidoglycan, a core component of the cell wall [10]. Protein E encoded by φX174E specifically inhibits MraY, blocking cell wall synthesis and leading to bacterial lysis.
When the bacterial density decreases to a certain level, the expression of the lysis gene ceases due to the reduced concentration of AHL, and bacteria resume growth, thus achieving continuous GLP-1 expression and release.
Figure 6. Schematic diagram of the bacterial autolysis system
2. Lysogenic Phage-mediated Host Lysis System
Phages exhibit strong strain-targeting specificity, making them suitable tools for specific chassis engineering. Phages are classified into virulent phages and temperate phages, also known as lysogenic phages [11]. Temperate phages can integrate into the host bacterial genome as prophages, stably inheriting alongside host proliferation. When environmental conditions change, prophages can undergo prophage induction, converting into virulent phages and lysing the host bacteria [12]. The lysogenic conversion mechanism of λ phage relies on interactions between two regulatory proteins (CI and Cro), three promoters (PR, PL, PRM), and six adjacent homologous operator elements. Additionally, various factors can indirectly affect the lysogenic/lytic switch by acting on the regulatory network, with the regulatory proteins CI and Cro playing decisive roles. In summary, CI promotes the conversion from the lytic to the lysogenic state, while Cro promotes the transition from the lysogenic state to the lytic state [13].
Approximately 20% of the genes in the λ phage genome can be replaced by exogenous genes without affecting its infection and integration capabilities. We will seek to utilize specific phages as delivery vehicles for biosafety genetic circuits, replacing the lytic gene φX174E in the aforementioned circuit with the lysis-associated Cro gene, and integrating this circuit into the phage genome. This enables intestinal Escherichia coli to sense cell density via QS and regulate their own lysis. After pre-infecting some engineered bacteria with the modified phages to achieve genomic integration, followed by thorough biological validation, we can formulate them into oral bacterial agents. By combining biosafety requirements with the need for GLP-1 release, we aim to develop more efficient and dynamically stable probiotic therapeutic strategies.
Figure 7. Schematic diagram of the lysogenic phage-mediated host lysis system
Design of Gut Delivery Carriers
We use sodium alginate-calcium chloride hydrogels as delivery carriers for engineered bacteria. These hydrogels possess advantages including high biocompatibility, capacity for loading multiple active substances, good tunability, site-specific delivery and controlled release, making them ideal carriers for oral probiotic delivery. Encapsulating probiotics within hydrogels can effectively isolate engineered bacteria from the external environment, ensuring their release only after reaching the intestine [14]. This facilitates the long-term colonization of EcN in the intestinal tract while avoiding damage to the human digestive tract, significantly reducing medication safety risks.
Figure 8. Microstructure of hydrogels observed under optical microscopy at 400× magnification.
Risk Assessment of Artificial Intelligence in Project Design
In the design and execution of our project, artificial intelligence (AI) technology was employed as an auxiliary tool. Regarding the application of AI technology in our project, we proactively conducted a biosafety risk assessment.
Current applications of AI in synthetic biology include predicting off-target effects of gene editing, de novo gene design, gene sequence modification, pre-synthesis gene screening, protein design, protein structure prediction, vaccine development, genetic circuit design, data analysis, library screening, drug screening, automation of laboratory experiments and risk prediction of biological research activities. Potential AI applications in our project encompass protein design and structure prediction, library screening, and diagrams and data analysis, among others.
Overall, several key challenges persist in AI technology application:
(1) AI may pose certain threats to data privacy and security;
(2) The accuracy of AI model predictions may be limited or biased due to constraints in the quality of training data;
(3) The transparency and interpretability of AI decision-making remain challenging, as the logical pathways behind its predictions and decisions are often not apparent;
(4) The reliability of AI model predictions is uncertain and requires rigorous validation;
(5) AI technology may raise concerns regarding data and intellectual property theft.
AI-driven prediction of how protein sequence and structure changes impact function greatly facilitates the design of novel proteins and modification of existing ones. However, existing AI technologies for aiding protein design still present several risks and vulnerabilities:
(1) Dual-use risk: AI-assisted protein design could be misused for malicious purposes, posing severe societal harm;
(2) Error-related risks: Inadequate guarantees for accuracy and efficiency of current AI in protein design;
(3) Unintended consequence risks: AI may fail to predict implications of newly designed or modified proteins;
(4) Potential biosafety risks: such as unintended biological effects;
(5) Ethical challenges: such as accountability for design outcomes;
(6) Others: Overreliance on AI predictions from excessive usage and oversight gaps in the protein design process.
Figure 9. AI integration with our project: Applications and challenges
According to the assessment framework outlined in Biosecurity Risk Assessment for AI Use in Synthetic Biology, the risk level of AI-assisted protein design is moderate [15]. In our project, all AI use is strictly limited to educational and research purposes and AI-generated results serve only as references and guidance. Comprehensive risk assessments and hazard screenings will be performed before any practical experiments.
AI application in literature retrieval and biochemical entity screening enhances search efficiency and accuracy. Since the AI used here is not directly linked to genetic manipulation, its associated risk level is assessed as low. When utilizing AI for data screening, we strictly comply with relevant laws and regulations, and conduct manual secondary screening of results to further mitigate potential risks.
References
[1] Sonnenborn, U. (2009). The non-pathogenic Escherichia coli strain Nissle 1917 – features of a versatile probiotic. Microbial Ecology in Health & Disease, 21(3-4),158-158.
[2] Lynch, J. P., Goers, L., & Lesser, C. F. (2022). Emerging strategies for engineering Escherichia coli Nissle 1917-based therapeutics. Trends in pharmacological sciences, 43(9), 772–786.
[3] Gan, M., Zuo, J., & Cao, Y. (2025). Biocontainment strategies of engineered bacteria. Synthetic Biology Journal. https://doi.org/10.12211/2096-8280.2025-010
[4] Rottinghaus, A. G., Ferreiro, A., Fishbein, S. R. S., Dantas, G., & Moon, T. S. (2022). Genetically stable CRISPR-based kill switches for engineered microbes. Nature communications, 13(1), 672.
[5] Schleif R. (2000). Regulation of the L-arabinose operon of Escherichia coli. Trends in genetics : TIG, 16(12), 559–565.
[6] Nanda, B., Bhowmick, J., Varadarajan, R., & Sarma, S. P. (2024). Backbone assignment of CcdB_G100T toxin from E.coli in complex with the toxin binding C-terminal domain of its cognate antitoxin CcdA. Biomolecular NMR assignments, 18(2), 285–292.
[7] Jiangnan University. (2024). An operon for fucose-induced gene expression, its construction method, and application [China Patent No. CN202311579074.0].
[8] Din, M. O., Danino, T., Prindle, A., Skalak, M., Selimkhanov, J., Allen, K., Julio, E., Atolia, E., Tsimring, L. S., Bhatia, S. N., & Hasty, J. (2016). Synchronized cycles of bacterial lysis for in vivo delivery. Nature, 536(7614), 81–85.
[9] You, L., Cox, R. S., 3rd, Weiss, R., & Arnold, F. H. (2004). Programmed population control by cell-cell communication and regulated killing. Nature, 428(6985), 868–871.
[10] Zheng, Y., Struck, D. K., & Young, R. (2009). Purification and functional characterization of phiX174 lysis protein E. Biochemistry, 48(22), 4999–5006.
[11] Feiner, R., Argov, T., Rabinovich, L., Sigal, N., Borovok, I., & Herskovits, A. A. (2015). A new perspective on lysogeny: prophages as active regulatory switches of bacteria. Nature reviews. Microbiology, 13(10), 641–650.
[12] Brady, A., Felipe-Ruiz, A., Gallego Del Sol, F., Marina, A., Quiles-Puchalt, N., & Penadés, J. R. (2021). Molecular Basis of Lysis-Lysogeny Decisions in Gram-Positive Phages. Annual review of microbiology, 75, 563–581.
[13] Lee, S., Lewis, D. E. A., & Adhya, S. (2018). The Developmental Switch in Bacteriophage λ: A Critical Role of the Cro Protein. Journal of molecular biology, 430(1), 58–68.
[14] Zhang, H., Liu, Z., Fang, H., Chang, S., Ren, G., Cheng, X., Pan, Y., Wu, R., Liu, H., & Wu, J. (2023). Construction of Probiotic Double-Layered Multinucleated Microcapsules Based on Sulfhydryl-Modified Carboxymethyl Cellulose Sodium for Increased Intestinal Adhesion of Probiotics and Therapy for Intestinal Inflammation Induced by Escherichia coli O157:H7. ACS applied materials & interfaces, 15(15), 18569–18589.
[15] De Haro L. P. (2024). Biosecurity Risk Assessment for the Use of Artificial Intelligence in Synthetic Biology. Applied biosafety : journal of the American Biological Safety Association, 29(2), 96–107.