In life-science and synthetic-biology research, safety is always the core principle. Throughout the design and execution of our project we strictly follow biosafety guidelines, ethical requirements, and applicable laws and regulations, while integrating risk assessment, standardized procedures, and safety controls to protect personnel, non-target organisms , and the environment to the greatest extent, and to guard against AI-related hazards. Through refined experimental design, safety switches in genetically modified microorganisms, and forward-looking safety planning for future applications, we ensure that both research and deployment are conducted in a manner that is safe, controllable, traceable, and socially as well as ecologically responsible.

In the life sciences, we stand at an unprecedented crossroads: gene-editing tools can rewrite the code of life, synthetic biology enables the creation of entirely new living systems, and artificial intelligence lets us decode the mysteries of biology at a speed once unimaginable. Yet greater power brings greater responsibility.

• Every technological leap carries inherent risks:
• An inadequately assessed gene-drive system could upset ecological balance;
• An experiment conducted without proper ethical review could inflict irreversible harm on participants;
• A single data breach could hand the genetic information of millions to malicious actors.

Being responsible is therefore not merely a moral choice—it is an imperative for the sustainable advance of science itself.

To curb the brown citrus aphid (Toxoptera citricida) in citrus groves, we designed APHiGO, a green, end-to-end system that warns, prevents and rapidly intervenes. The platform has three core modules:

1. Early-Warning Module
An engineered Bacillus subtilis detects aphid-secreted sucrose and converts the invisible infestation density into a real-time volatile signal by producing methyl salicylate (MeSA).

2. Long-Lasting Prevention Module
A trunk-injected RNAi formulation gives the tree inside-out protection, overcoming the rapid degradation and uneven coverage of conventional foliar sprays.

3. Targeted Intervention Module
An engineered strain of Metarhizium anisopliae acts as a precision bio-insecticide that quickly knocks down aphid outbreaks.

An IoT layer continuously records micro-climate and MeSA levels; AI algorithms push field-specific advice to farmers through a WeChat mini-program, closing a smart protection loop that supports sustainable citrus production.

Fig 1. CitrusShield Smart Ecosystem

REGULATORY COMPLIANCE

Our project uses GMOs and synthetic-biology products: Bacillus subtilis, Metarhizium anisopliae, E. coli BL21, E. coli DH5α and Agrobacterium. Check-In Forms for every chassis have been approved by the Safety & Security Committee; all work is confined to BSL-1 laboratories. VLPs and RNAi molecules produced in E. coli never leave the lab. Designs for a "safety switch" in Metarhizium and Bacillus were completed only in silico; no wet-lab work was performed.

We explicitly state that "trunk injection" and "field testing" are future deployment concepts, not activities carried out for this competition.

Off-target Risk Management

To guarantee the environmental safety and species specificity of our RNAi-based biopesticide, a rigorous off-target risk assessment is an indispensable step. Because RNAi acts through complementary base-pairing, a short hairpin RNA (shRNA) that shares high homology with essential genes of non-target organisms—especially humans, domestic animals, natural enemies and pollinators—could cause inadvertent gene silencing and potential ecological damage. We therefore implemented a three-tier bioinformatic pipeline:

1. Stringent human-safety verification
Both the human whole-genome and transcriptome databases were queried in parallel to eliminate any possibility of off-target effects on human genes, ensuring absolute safety.

2. Precision evaluation of key ecological species
Transcriptome databases of natural enemies, pollinators, citrus trees and mice were systematically screened. Since RNAi acts directly on mRNA, this approach accurately quantifies potential impacts on non-target gene expression.

3. Proactive survey of the putative target spectrum
All RNAi sequences were broadly aligned against the non-redundant Core nucleotide database to identify other species that might be affected. The results confirm that the molecule is specific to the target pest and its close relatives, displaying high target fidelity.

In Figure 2 we assign risk levels based on the above comparisons. The RNAi constructs show virtually no off-target potential for human genes or citrus crops and are classified as absolutely safe. For key ecological species such as pollinators, the off-target risk is very low or negligible, whereas the constructs exhibit high specificity toward citrus pests. The proactive survey further reveals that any species falling outside the safety threshold are either citrus pests or their near relatives.

Fig 2. Species or gene pool BLAST risk level dot matrix diagram. 1-5 respectively represent our assessment of humans, citrus crops, and pollinators (Ladybug), pollinating insects (bees), threat and risk assessment of pests. The right side caption represents the sequence identity assessment classified by Per.Ident, and we have classified it into 5 levels (0: represents No significant similarity found in BLAST somewhat analysis or Per.Ident < 70%, no risk; 0.25: represents 70% ≤ Per.Ident < 80%, low risk; 0.5 represents 80% ≤ Per.Ident < 90%, medium risk; 0.75 represents 90% ≤ Per.Ident < 95%, high risk; 1 represents 95% ≤ Per.Ident, extremely high risk)

Project verification

Safety Measures Details ▼

Throughout the experimental validation of this project, the safety of laboratory personnel was treated as a top priority. The following specific measures were implemented to guarantee that every step—especially those involving RNAi synthesis, virus-like particle (VLP) production and insect handling—remained safe and fully controllable

Therefore, we visited Xie Ning, an expert from the Pesticide Registration Review Committee, to gain in-depth understanding of the relevant information on biopesticide standard setting.

1. RNAi synthesis
All double-stranded/small-hairpin RNA molecules are chemically synthesized inside a Class-II biological safety cabinet to minimize aerosol release and cross-contamination. Operators wear disposable lab coats, nitrile gloves and surgical masks at all times to prevent direct skin contact or inhalation of nucleic-acid samples.

2. VLP production
The MS2 VLPs used here are genome-free; only the coat-protein coding sequence is expressed to drive self-assembly. Consequently the particles are replication-incompetent and non-infectious. All consumables and waste liquids that contact VLPs are autoclaved at 121 °C before disposal. A dedicated bench area is reserved for VLP work and is clearly marked to avoid overlap with other experiments. Operators wear full PPE (gloves, goggles, lab coat) while inside this zone.

3. Insect handling
Aphid colonies are maintained in climate-controlled incubators fitted with fine-mesh escape-proof cages. Every manipulation (transfer, injection, collection) is performed inside a BSL-1 laminar-flow hood. Containers and tools that contact insects are disinfected with 70 % ethanol or 1 % sodium hypochlorite; dead insects are heat-inactivated (95 °C, 30 min) and then disposed of through the university's biological-waste officer. Racks and incubators are inspected and cleaned weekly to ensure no escaped individuals can survive or congregate.

4. Use and destruction of engineered microbes
All genetic work employs non-pathogenic strains—Bacillus subtilis (EPA-tier 1), Escherichia coli (K-12 derivatives) or Metarhizium anisopliae—certified free of virulence factors. During manipulation workers wear gloves and FFP-2 masks to avoid inhalation or skin contact. At the end of each experiment every culture, spore suspension or contaminated item is autoclaved (121 °C, 30 min) before discarding; no viable GMO leaves the laboratory.

The realization of our project chiefly involves two genetically engineered strains—Metarhizium anisopliae and Bacillus subtilis—which in the future will be deployed in field environments via spraying. Safety control of GMOs is extremely strict; therefore, we have designed dedicated safety switch systems for each organism. This strategy is crucial because genetic material may spread into indigenous microbes through horizontal gene transfer (HGT), posing potential threats to ecosystems and public health.

Metarhizium anisopliae

Metarhizium will be formulated as an instant-capsule pesticide preparation for application in citrus orchards. Farmers will dissolve the capsule in water and spray it onto citrus trees. To ensure the safe use of engineered Metarhizium in citrus groves, we designed a safety switch that allows the engineered fungus to survive only within the target host, while automatically triggering self-elimination in non-target environments.

We selected the promoter PMcl1, which responds specifically within the insect hemocoel (literature), to control expression of the photosensitive lethal protein SuperNova. When the engineered fungus is applied to the environment as spores, this promoter remains silent, ensuring stability and dormancy under natural conditions. Only after the fungus successfully infects aphids and enters their hemocoel is PMcl1 activated, initiating SuperNova expression. The protein is then transported into the nucleus via a fused nuclear-localization signal.

Crucially, the interior of the aphid hemocoel is dark, so SuperNova's photosensitive toxicity cannot be activated, allowing the engineered fungus to complete its infection and killing lifecycle normally. After the task is finished, when fungal hyphae emerge from the aphid cadaver into natural light, SuperNova is immediately activated by illumination, generating massive reactive oxygen species that cause cell death, thereby effectively preventing further spread and persistence of the engineered fungus in the environment. This minimizes the potential risk of residual and disseminated GMOs in nature and provides an important safeguard for the safe application of microbial pesticides (details can be found on our "Design" page).

Bacillus subtilis

To address the biosafety risk that endospores of Bacillus subtilis, when used as a chassis, may escape and survive for extended periods, we designed a redundant OR-logic-gate kill switch. The core of this system is the highly efficient MazF/MazE toxin–antitoxin system, which strictly limits strain survival by means of two independent environmental signals—sucrose (indicating heavy aphid presence) and population density (quorum sensing).

The key design point is that only when both conditions (high sucrose and high quorum) are satisfied can the strain survive. Once either safety signal is lost, massive expression of the MazF toxin is triggered. MazF is a ribonuclease that rapidly degrades cellular mRNA and causes cell death (literature). To cope with the inevitable leakiness inherent to OR-gate promoters, we use a weak constitutive promoter, PliaG (BBa_K823000), to continuously express the MazE antitoxin, neutralizing basal-level toxin (literature). Only when the OR gate is fully activated and MazF expression exceeds the buffering capacity of MazE does the killing program efficiently initiate.

This dual-signal design ensures reliable biocontainment: it ties bacterial survival to citrus trees harboring aphid honey-dew. Even if an escaped strain accidentally encounters sucrose in soil or on weeds, the lack of sufficient quorum-sensing signal (due to dilution or dispersion) will still activate the kill switch, serving as a final insurance to clear escapees.

Fig 3. Kill switch mechanism diagram

Sucrose-regulation module: The PsacB promoter senses environmental sucrose. PsacB is a sucrose-inducible promoter (activated when sucrose is present) responsible for producing LacI repressor protein . LacI then binds to lacO sites on the PgraC promoter, completely shutting it down (literature). Thus, under safe conditions where sucrose is present, PsacB is activated to produce LacI, thereby repressing PgraC (literature). Once the strain escapes into a sucrose-free environment, PsacB turns off, LacI repressor stops being expressed and degrades, PgraC is derepressed, and suicide is activated.

Quorum-sensing module: The quorum-sensing module serves as the second safety line for population density and is fully orthogonal to the LacI module. This system utilizes the native ComQXPA pathway of Bacillus subtilis: as cell density increases, pre-ComX expressed by ComQXPA is processed by ComQ and secreted outside the cell; the accumulated signaling molecule ComX phosphorylates the membrane receptor ComP, which in turn activates phosphorylation of the intracellular transcription factor ComA (ComA-P) (literature). We exploit the natural target of ComA-P, the PsrfA promoter (a density-activated promoter), to express the XylR repressor protein downstream (literature). At high population density (safe condition), PsrfA is continuously activated to produce XylR protein. XylR then efficiently binds to the xylO operator on the Pxyl promoter, repressing and shutting it down. Conversely, when strains are dispersed and lack quorum-sensing signal, XylR repressor stops being produced and degrades, Pxyl is immediately derepressed, and suicide is activated.

Fig 4. OR gate status diagram

Laws & Regulations ▼

To ensure that our project outcomes can be safely and compliantly implemented, we proactively conducted regulatory research on RNA-based biopesticides. We found that China's current regulatory system has core issues such as a lack of classification standards and clear registration data requirements. At the same time, there are significant differences in regulations across countries, and these complexities once created bottlenecks in our policy research.

Therefore, we visited Xie Ning, an expert from the Pesticide Registration Review Committee, to gain in-depth understanding of the relevant information on biopesticide standard setting.

When we attended the CCiC meeting, we asked Dr. Bao Yuhan for advice. He offered two suggestions:First, for the current Chinese pesticide-standard formulation process, carry out in-depth analysis of the core considerations—this step can be accomplished by conducting expert interviews to obtain first-hand information.Second, re-examine and adjust the project-design mindset, clarifying what factors must be considered for developing RNA pesticides under the existing pesticide-standard system.

During this discussion, he urged us to analyze the factors that lie behind the laws and regulations. We therefore performed an analysis of China's RNA-pesticide statutes;

Learn more in Integrated Human Practices.

At the same time, in order to gain a deep understanding of the relevant laws and to review our product development, we wrote the China RNA Pesticide Industry Development Blue Book: Analysis of International RNA Pesticide Regulations and Reflections Based on China's National Conditions. It contains our technical analysis and risk assessment of RNA pesticide products.

Learn more in Integrated Human Practices.

Product safety profile ▼

We have anticipated the form of the product, its intrinsic properties, the end-user, and the receiving environment. Although laboratory work is still ahead, we already commit to addressing every point listed below before any real-world release, accepting full responsibility for users and the ecosystem. Field trials, when initiated, will strictly follow all relevant protocols and regulatory guidelines.

Formulation and mode-of-application safety
We will develop a stable formulation for our product and analyze its physicochemical stability during transport, storage, and application to ensure no improper degradation or leakage risks. Trunk injection, as a precise application technique, will be governed by standard operating procedures to avoid tree damage or human contact, while also analyzing the degradation and residue of the agent within the trunk. Spraying, as a common method, will be evaluated for the product’s degradation, residue, and dispersion in the natural environment after application. Currently, it is known that RNA pesticides degrade easily in the environment, and microorganisms can decompose RNA in the soil within a few days without accumulation.

Safety of RNAi-VLP
Although we have maximized the targeting of RNAi molecules through bioinformatics analysis, to compensate for the uncertainty of computer prediction, in the future, we will further verify its specificity in a system that simulates real-world environments. Specifically, we will use non-target organisms such as bees, ladybugs, and aquatic organisms to evaluate whether RNAi molecules have gene silencing effects on them, and simultaneously verify whether MS2 VLP and its functional peptides are toxic to the above organisms. Environmental Safety: Impact on soil microbial communities: Analyze the potential impact of product components (such as engineered bacteria, VLP) on soil microbial diversity and key ecosystem functions (such as nitrogen cycling) after they enter the soil. Water body and groundwater pollution threat and risk assessment: Assess the risk of product components leaching or running off from the application point to surrounding water bodies under rainfall or irrigation conditions to ensure that water sources are not polluted. Ecosystem-level impact study: By establishing a small-scale field trial ecosystem, long-term observation of the comprehensive impact of the "CitrusShield" system on the entire citrus orchard ecological network (including pests, natural enemies, pollinators, soil organisms, etc.) is conducted to ensure that its introduction does not disrupt the ecological balance.

Guaranteeing the safety of personnel and carrying out every procedure in a regulated manner are integral parts of our project. Laboratory safety is the bottom line of scientific research; a single lapse can lead to irreversible consequences. Our team has therefore established clear protocols and strengthened individual accountability to ensure that everyone who enters the lab gives constant priority to safety in thought and rigorously follows every safety requirement in practice.

Laboratory safety equipment: Our laboratory is classified as BSL-1 and is equipped with a BSL-2 biological safety cabinet; it has been rated as "good" in the university's laboratory safety management program.Our laboratory is fitted with a comprehensive array of equipment that safeguards both the facility and the personnel who work in it.

The fire hydrant is one of the most basic fire-fighting facilities in the laboratory.A fire extinguisher is positioned beside the laboratory door.Our laboratory keeps a first-aid kit stocked with gauze, pressure bandages, iodine solution, instant cold packs, adhesive strips and other emergency supplies.Our laboratory is equipped with an eyewash station; if chemicals splash into the eyes or onto the face, it is used immediately to dilute and remove the harmful substance and prevent further injury.There is an emergency-supply cabinet near the laboratory that is stocked to cope with any sudden incident.Our laboratory is equipped with an emergency response procedure manual, enabling us to handle unexpected incidents calmly and according to protocol.The laboratory is fitted with a fume hood; whenever we must handle volatile or otherwise hazardous reagents, we perform the manipulations inside the hood.We are equipped with a biological safety cabinet (BSC) to ensure maximum protection: personnel are not infected, samples are not contaminated, and pathogens do not leak out.

Last-Person-Check System: To ensure daily lab safety, we implement a "Last-Person-Check" system: whoever leaves last performs a full end-of-day inspection—confirming that all instruments are shut down, reagents are properly stored, and benches are clean. The results are recorded every day on the checklist.

Last-Person-Check Form

Laboratory Zoning Experiment: We strictly segregate different experimental areas (e.g., RNA-prep zone, protein-processing zone, agarose-gel zone) so that cross-contamination among samples or reagents is prevented and data accuracy is guaranteed. For instance, RNA is extremely vulnerable to contamination from other zones, which can easily produce false-positive results; zoning markedly reduces this risk.

Clear zone demarcation and labeling also help personnel instantly recognize the reagents and samples required for each experimental step, minimizing the chance of operational errors.

Schematic Diagram of Laboratory Zoning (Partial)

Laboratory Safety Management Regulations: School and laboratory administrators have formulated laboratory safety management regulations for us in various forms. Before joining the laboratory, we all studied them rigorously, and we strictly abide by these regulations during the experimental process.

Laboratory Safety Management Regulations (Excerpt)

Experimental reagents: When reagents with potential hazards are involved, we first review their safety data sheets and fully understand their properties. Throughout the experiment every step is performed strictly according to standard protocols, and all leftover or waste reagents are handled in full compliance with laboratory regulations.

The integration of artificial intelligence and biology (AI + bio), as a promising and disruptive technological frontier, is triggering profound changes in the global scientific and industrial landscape. Its potential in drug discovery, disease diagnosis, synthetic biology, and precision medicine has been widely recognized, significantly improving research efficiency and clinical translation. Yet this convergence also brings severe potential hazards and systemic risks, calling for forward-looking responses in both technology development and governance.

Global Catastrophic Biological Risks (GCBRs) Framework

Referring to the "Global Catastrophic Biological Risks (GCBRs)" framework proposed by Johns Hopkins University in 2017, the risks facing the bio-AI fusion field mainly appear in five areas:

1. Lowering Technical Barriers：AI tools can simplify complex biological workflows, automating and democratizing operations that once required extensive expertise, enabling misuse or abuse by non-specialists.
2. Assisting Problem-Solving：AI systems can help researchers overcome traditional bottlenecks, but may also be used to circumvent biosafety controls, such as designing novel pathogens or evading existing detection methods.
3. Enhancing Pathogen Upper Limits：Through machine-learning modelling and generative design, AI can accelerate the development of microbes with higher transmissibility, drug resistance or virulence, amplifying biological threats.
4. Circumventing Oversight Mechanisms：Distributed, automated experimental platforms may fall outside current regulatory coverage, making tracking and control of high-risk biological activities harder.
5. Accelerating R&D Cycles：AI dramatically shortens the path from proof-of-concept to real-world application, causing regulatory policies and ethical standards to lag behind technological iteration.

Our Multi-Stakeholder Approach

To address these risks, our team has engaged in deep exchanges with experts in AI technology, bioengineering, data ethics and science policy, integrating multi-stakeholder advice systematically throughout project development.

Technical Solutions

Technically, our future software/hardware platform will be deeply integrated with state-of-the-art AI large models to build an experimental environment capable of self-auditing and risk perception. For example, embedding AI capacity into a PostgreSQL-based data-analysis pipeline enables more efficient and accurate gene-sequence screening and function prediction, markedly improving the performance and reliability of key modules such as dsRNA off-target analysis. We also plan to introduce Explainable AI (XAI) to increase transparency of model decision-making and reduce scientific uncertainty and safety risks arising from algorithmic "black boxes".

Regulatory and Governance Measures

On the regulatory and governance side, the platform will incorporate a "session-tracking mechanism" that assigns a unique ID to every bio-design task, logging the full workflow from sequence design and in-silico validation to physical synthesis, ensuring full traceability and auditability. All core code and technical documentation are released under the Creative Commons Attribution 4.0 International licence, guaranteeing openness and community auditability and accepting continuous oversight by the global research and regulatory community.

Public Education and Awareness

We further believe that meeting the fundamental challenge of "lowered technical barriers" requires not only stronger technical controls and legal constraints but also a systematic uplift of public scientific literacy and risk awareness. To this end, together with the CJUH-JLU-China team we produced the popular-science handbook Debunking Synthetic Biology Myths, using accessible content and multi-channel dissemination to bridge public knowledge gaps and reduce potential misuse driven by misunderstanding, misinterpretation or malicious information at the source.

Open Source Commitment and Security

We recognize that while open-source models foster collaborative innovation and social supervision, they may also be exploited by malicious actors. We therefore commit to implementing strict peer review and security scanning for merge requests on platforms such as GitLab, establishing a permission-graded, behavior-monitored access-control system, and continuously tracking the latest policy developments in international biosecurity and AI ethics. We will iteratively update the platform's security design and governance strategies, striving to maintain a balance between technological development and social responsibility.

Safety is not a check-box item tacked on at the end of our project; it is the invisible thread that runs through every design, every experiment, and every product deployment. From the moment a DNA sequence appears on a computer screen until the biological product actually leaves the bench and heads into the outside world, the first question we always ask is: "How do we ensure that no harm escapes our control?" We write "safety" into the very sequence of the molecule itself, we build multiple layers of protection for the operator inside biosafety cabinets and fume hoods, and we use data—not slogans—to prove to regulators, users, and the environment that our product carries a "passport of harmlessness." In our project, if it is not safe, it is not science!