Catalog
With the emergence and development of synthetic biology, genetically modified microorganisms are increasingly used in biomedical, industrial, and environmental applications. To prevent engineered microorganisms from leaking into the natural environment and to stop engineered genetic material from spreading to unintended host cells, it is essential to develop safe and reliable methods to inhibit their proliferation. Methods to prevent leakage of engineered microorganisms include utilizing engineered nutrient-deficient strains for containment, inducing cell death through suicide switches for active containment, and barriers to gene flow, among others [1]. The most commonly used method is the toxin gene suicide switch. Therefore, we initially designed and incorporated a suicide switch into our engineered strains to ensure biosafety.
We drafted a gene construct containing a standard suicide switch, where the toxin gene will be expressed in non-laboratory environments to induce cell death.
Based on our design, we predicted that our suicide switch would function normally in BZ-11. Before conducting actual laboratory experiments, our Human Practice (HP) team communicated with food regulatory agencies, feed manufacturers, and consumers to assess the acceptability of this system in animal feed production.
During HP interviews with food regulatory agencies, feed manufacturers, and consumers, we were informed that the use of toxin genes and antibiotic resistance markers in food-related products is unacceptable. Regulatory agencies emphasized that strains containing toxin proteins or resistance genes do not meet food safety approval standards. Moreover, consumers expressed significant concerns about the presence of such elements in food.
In light of this feedback, we contacted our instructor to seek possible solutions and ultimately decided to abandon the use of the suicide switch before conducting any practical experiments. We shifted to a "Resistance Removal" method, which ensures that the antibiotic resistance marker disappears in the final engineered strains through the Cre-loxP system, thereby avoiding the leakage of antibiotic resistance genes into the environment. This approach maintains the necessary control and safety requirements while addressing regulatory and consumer concerns, allowing us to proceed without using toxin genes.
Metabolic networks are interconnected biochemical pathways within living cells that assemble the components or compounds required for cellular functions or generate energy and materials through the breakdown of biomolecules. Enzymes act as catalysts in the vast majority of reactions, thus playing a central role in metabolic networks, and the interactions of their components determine the dynamic behavior of the entire network [2]. The presence of numerous uncertainties in biological systems complicates the study of biological metabolic networks, making it a complex and vast undertaking. To better understand our chassis organism, Aureobasidium melanogenum BZ-11, modeling in systems biology aids our comprehension of metabolic mechanisms and provides essential insights for our genetic modifications.
Genome-scale metabolic models (GEMs) are formed using gene-protein-reaction (GPR) associations based on genomic annotation data and experimental information. These models encompass system information such as metabolites, reactions, genes, and compartments, describing the metabolic network of an entire biological system and predicting metabolic fluxes of system metabolites through optimization techniques [3].
Since Aureobasidium melanogenum BZ-11 currently lacks a genome-scale metabolic network, we compared it with the closest species available in databases based on the phylogenetic tree of Ascomycota species [4], identifying Aspergillus niger’s GEMs network.
 |
|
Figure 1: The phylogenetic tree of Ascomycota species, adapted from previous work [4]. |
We analyzed the GEMs network of Aspergillus niger and attempted to fit the known pathways of Aureobasidium melanogenum BZ-11 one by one.
We found that many pathways in the genome-scale metabolic network of Aspergillus niger do not match those known for Aureobasidium melanogenum. For example, in melanin synthesis, Aureobasidium melanogenum employs the dihydroxynaphthalene (DHN) pathway, while Aspergillus niger utilizes the L-3,4-dihydroxyphenylalanine (L-DOPA) pathway [5]. This significantly impacted our analysis. Additionally, we realized that this method of searching pathways one by one is inefficient and undesirable. Therefore, we decided to construct the metabolic network ourselves. After gathering information, we discovered that building a GEM-level metabolic network requires whole-genome sequencing of the strain. However, this posed a challenge for us. Given the current situation and the problems encountered, we consulted our instructor, who agreed with our solution and generously provided us with the whole-genome sequencing information for the Aureobasidium melanogenum P16 strain, which greatly assisted us.
Since the Aureobasidium melanogenum BZ-11 strain lacks a genome-scale metabolic network and it's difficult to use the metabolic networks of closely related species, we chose to use the whole-genome sequencing information of the Aureobasidium melanogenum P16 strain to construct a GEM for facilitating the prediction of the relationship between β-glucan yield and its growth.
After reviewing literature and gathering information, we decided to use the CarveMe reconstruction tool. CarveMe is an automated reconstruction tool for genome-scale metabolic models that uses the Diamond sequence aligner to compare sequences in databases. It can infer an organism's uptake/secretion capabilities solely from genetic evidence, thereby constructing metabolic networks, making it particularly suitable for organisms that cannot be cultured under well-defined media [6].
We constructed a genome-scale metabolic network using the CarveMe reconstruction tool by running it in Linux Ubuntu.
We analyzed the constructed GEMs network using flux balance analysis (FBA) methods, simulating it with COBRApy, and performed two optimizations with β-glucan and growth generation as objective functions.
 |
|
Figure 2: Relationship Between β-glucan and Growth Flux |
From the analysis results, we found that the β-glucan flux does not affect growth until it reaches 175 mmol/gDW/h. After reaching a maximum yield of 667 mmol/gDW/h, growth decreases by about 87%. Therefore, to maximize β-glucan production, it is necessary to minimize the impact on cell growth and reduce system load. Additionally, we still could not determine which specific secondary metabolites' synthesis would affect β-glucan synthesis. To gain insights, we reviewed past iGEM teams' wikis and ultimately found further analysis methods in the 2023 iGEM Leide team's wiki [7].
The determination and study of in vivo metabolic flux are known as flux analysis, which plays a central role in metabolic engineering and is a fundamental determinant of cellular physiology, providing key parameters for various pathways in metabolism [8]. OptForce is a computational optimization framework based on flux variability analysis (FVA) that aims to identify all potential genetic modifications leading to the overproduction of desired metabolites in microorganisms through flux balance analysis (FBA). A primary challenge in metabolic engineering is to maximize the conversion of renewable resources into useful products. OptForce improves these methods by integrating experimental flux data from wild-type strains and focusing on which fluxes must change to achieve overproduction goals. It compares the range of metabolic fluxes between wild-type and overproducing strains to determine which reactions must be increased (MUSTU) or decreased (MUSTL) to meet target yields. Once all necessary flux changes are identified, OptForce employs a dual-layer optimization program to determine the minimal set of genetic interventions required for the expected overproduction (FORCE set). This method ensures efficient strain engineering by precisely identifying key modifications while minimizing the number of genetic changes needed for the target phenotype [9].
Inspired by the 2023 iGEM Leide team, we used the OptForce algorithm based on the MathWorks COBRA Toolbox for our construction [7].
Through reasonable constraints and calculations, we obtained the first-order Must and second-order Must sets, and simulated the OptForce algorithm, returning the top 10 single reaction force sets.
 |
|
Figure 3: OptForce result |
We ultimately focused on the pullulan_01 pathway for the synthesis of the secondary metabolite pullulan. We also assessed the preceding reaction, pullulan_00, considering it to significantly affect β-glucan synthesis as well. Meanwhile, our wet experiments revealed the presence of pullulan through the analysis of extracellular polysaccharide components, indicating that knocking out this pathway could enhance β-glucan yield while avoiding impurity issues during the purification of extracellular polysaccharides. To improve the quality of the final product, we decided to knock out the melanin synthesis pathway; however, upon performing FBA simulations and gene knockout simulations again, we found that controlling these sub-reactions was not sensitive, making it difficult to draw conclusions from the simulations.
After analyzing the GEMs metabolic network, we narrowed our focus to the metabolic pathways of β-glucan, melanin, and pullulan. Based on this, we decided to use traditional kinetics for modeling to conduct a more refined analysis and further identify the key reactions we want to knock out.
Simbiology is widely used for pharmacokinetic simulations and serves as an efficient tool for simulating small metabolic systems. It allows observation of the network's dynamic behavior and provides more refined control and optimization strategies [10]. Biochemical network simulation can dynamically observe the changes of various substances over a certain time scale, aiding our understanding of the network's dynamic behavior. Considering that there are not many synergistic effects in the network, we employed a lightweight local sensitivity analysis to identify the parts of the system with high sensitivity.
Using MathWorks-based Simbiology for kinetic modeling, we ignored enzyme inhibition and activation effects, focusing on which reactions enzymes mainly participate in to further simplify the model load.
We observed the dynamic changes of system substances at low concentrations, analyzing the yield of the target product under different parameters. We iterated the parameters of the first three reactions that showed high sensitivity in the synthesis of pullulan and melanin, observing the corresponding yields and the degree of fluctuation.
 |
|
Figure 4: Analysis of the Melanin Synthesis Pathway (LSA) |
 |
|
Figure 5: Analysis of the Pullulan Synthesis Pathway (LSA) |
The sensitivity values for Kf_26 involved in pullulan synthesis and Kf_25 involved in melanin synthesis were the highest. Our experiments will focus on the melain_01 and pullulan_01 reactions for gene knockout. Ultimately, we decided to target the Ags2 and PKS genes for knockout in our wet experiments. However, due to the compact nature of our experimental section, some parameters were not fitted, and we currently cannot achieve complete quantitative analysis. Subsequently, we compared the results of the LMN back to the GEMs, seeking to visualize and analyze the deficiencies in GMN analysis, discovering many key pathways that were overlooked. This indicates the need for further model optimization and the addition of constraints for in-depth exploration in the future.
Vitreoscilla Hemoglobin (VHb) is a type of hemoglobin derived from Vitreoscilla, which serves as a terminal electron transport chain component, enhances the survival of microorganisms under low oxygen conditions, and promotes total protein synthesis in cells. It has been applied in industrial fermentation [11]. To accurately localize and attach VHb to the outer side of the cell wall of A. melanogenum BZ-11 to enhance oxygen uptake, we designed to combine the cell wall anchoring protein with VHb to ensure its localization at the cell wall.
Cell wall protein is a mannan protein consisting of an N-terminal signal peptide and a C-terminal anchoring sequence, where the N-terminal is responsible for protein secretion, and the C-terminal anchors the protein to the cell wall. This protein can bind to the structural 1,6-β-glucan present in the cell wall [12]. We also identified a cell wall protein anchoring sequence in A. melanogenum BZ-11. Additionally, we found cell wall proteins from other species, namely Yarrowia lipolytica (CWP) and Ogataea angusta (Sed, Tip), and compared their sequences with those from our species to score and find the most suitable sequence copy for A. melanogenum BZ-11 as the final sequence.
We retrieved the GPI-CWP sequences of these species from NCBI and successfully predicted the N-terminal signal peptide and C-terminal anchoring region of these sequences using the SignalP and big-PI Fungal Predictor tools. We then fused the predicted N-terminal and C-terminal sequences with the target protein.
We used SignalP to simulate the cleavage site and determine if any changes occurred. By gradually adding amino acids and comparing with the recombined sequence of the target protein, we identified the sequence with the least polarity change.
Based on the sequence alignment results, we preliminarily determined to use the AM.CWP protein from A. melanogenum BZ-11 as the anchoring protein. However, the sequence alignment data alone lacked sufficient convincing power and required further validation. Next, we plan to connect the cell wall protein sequence with EGFP to construct a plasmid for introduction into A. melanogenum BZ-11 and check the fluorescence intensity on the cell wall. To ensure the stability and functionality of the protein for better validation of EGFP, we also predicted the EGFP sequence by gradually adding amino acids, resulting in the addition of three amino acids (AYI) compared to the original sequence.
 |
|
Figure 6: EGFP Signal Peptide Prediction |
In previous work, we confirmed that the PKS gene is a key gene for melanin synthesis in A. melanogenum BZ-11, while the AGS2-1 and AGS2-2 genes are essential for pullulan synthesis in the same organism.
We amplified the 5' arm and 3' arm sequences of the PKS, AGS2-1, and AGS2-2 genes, and constructed linearized gene knockout vectors pFL4A-NAT-loxp-PKS, pFL4A-NAT-loxp-△AGS2-1, and pFL4A-NAT-loxp-△AGS2-2 using the NAT gene as a marker. These vectors were then sequentially introduced into the organism.
The strains containing the knockout vectors were able to grow on media containing nourseothricin
We verified the knockout effects using colony PCR.
 |
|
Figure 7: Electrophoresis to detect knockout results |
From the colony PCR results, the final BZW△ags2-1/2 strains showed that the PKS, AGS2-1, and AGS2-2 genes were all knocked out. However, the knockout of these genes does not necessarily mean that the metabolic pathways have been completely blocked; further validation is needed using high-performance liquid chromatography and infrared spectroscopy analysis.
During our Human Practices activities, we identified a widespread issue through conversations with aquaculture farmers: feed sinking to the bottom leads to significant waste, as much of the feed is not efficiently utilized, potentially limiting the growth of aquatic animals. Moreover, feed sinking and overfeeding can cause severe water pollution, sometimes even resulting in the mass death of aquatic species. Thus, "precision feeding" plays a crucial and indispensable role in improving both production efficiency and economic outcomes in aquaculture.
To prevent feed from sinking to the bottom, we designed a feed-holding dispenser composed of three parts: a floater, a connecting rod, and a feed holder. Since most aquatic animals do not deliberately seek food at the bottom and prefer fresh feed, we can place the feed directly on the holder, ensuring that it remains available to the animals before sinking. The aquatic animals can easily consume the feed from the holder (Fig. 8).
 |
|
Figure 8: Schematic diagram of hardware usage |
Using SolidWorks, we created a hardware model where the floater is a sphere with a diameter of 3 cm and a thickness of 2 mm, the connecting rod is 8 cm long with a 3 mm diameter, and the feed holder is a curved disc with a diameter of 10 cm and a thickness of 3 mm. We then created physical models using 3D printing methods (Fig. 9)
 |
|
Figure 9: Feeder modeling and 3D-printed model |
We simulated a water environment and then tested the hardware to determine the effectiveness of the hardware in use (Vedio 1).
|
Video 1: Feed tossing demonstration by the feeder |
During our feed placement experiments, we encountered an issue: it was challenging to accurately determine where to feed in an open water surface, resulting in a significant portion of the feed missing the holder and sinking to the bottom. To address this, we sought to design a method that would restrict the feeding area, ensuring that all feed could be precisely placed onto the feed holder, rather than sinking to the bottom again.
As a solution, we added a ring around the floater with a diameter smaller than that of the feed holder (8cm), to delineate the feeding area (Fig. 10). This ensures that the feed remains within the designated range and does not sink, effectively mitigating the issue of feed waste (Vedio 2).
In this Engineering process, we realized the importance of taking a comprehensive approach to hardware design to ensure it can be effectively applied in real-world scenarios.
 |
|
Figure 10: Feeder modeling and 3D-printed model |
|
Video 2: Feed tossing demonstration by the feeder |
When feeding we ah find that the upper floater takes up a lot of space on the surface of the water and is a significant hindrance to feeding. Therefore, we improved the floater design by combining the range-defining ring and the floating sphere into a single floating ring (Fig. 11). This modification significantly reduces the surface area occupied by the floater while maintaining its functional effectiveness.
 |
|
Figure 11: Feeder modeling and 3D-printed model |
We then created physical models using 3D printing methods (Fig. 11).
To better reflect real-world conditions, we introduced a water flow simulator to replicate water movement in aquaculture environments and assess the effectiveness of our device. Following this, we conducted a feeding simulation.
|
Video 3: Feed tossing demonstration by the automatic feeder |
During the simulation, we observed that without being anchored, the feeder would drift with the water current, resulting in unstable feeding positions (Video 3). This instability could make it difficult for aquatic animals to locate the feeding area, or cause the feeder to drift away from the group, leading to inefficiencies in feeding.
To address this issue, we attached a stabilizing rod to the feeder, ensuring that it remains stationary in flowing water. Furthermore, we replaced the stabilizing rod with an adjustable rope, allowing for flexibility in various aquaculture environments while maintaining the feeder’s stability and precision (Fig. 13 and Video 4).
We recognized that when conducting hardware simulations, it is essential to closely mimic real-world conditions. This approach helps us more easily identify potential issues that may arise during application, thereby increasing the feasibility of its use.
 |
|
Figure 13: Feeder modeling and 3D-printed model |
|
Video 4: Feed tossing demonstration by the automatic feeder |
[1] Moe-Behrens, G. H. G., Davis, R., & Haynes, K. A. (2013). Preparing synthetic biology for the world. Frontiers in Microbiology, 4. https://doi.org/10.3389/fmicb.2013.00005
[2] Dräger, A., & Planatscher, H. (2013). Metabolic Networks. Encyclopedia of Systems Biology, 1249–1251. https://doi.org/10.1007/978-1-4419-9863-7_1277
[3] Gu, C., Kim, G. B., Kim, W. J., Kim, H. U., & Lee, S. Y. (2019). Current status and applications of genome-scale metabolic models. Genome Biology, 20(1), 121. https://doi.org/10.1186/s13059-019-1730-3
[4] Xiao, D., Driller, M., Dielentheis‐Frenken, M., Haala, F., Kohl, P., Stein, K., Blank, L. M., & Tiso, T. (2024). Advances in Aureobasidium research: Paving the path to industrial utilization. Microbial Biotechnology, 17(8). https://doi.org/10.1111/1751-7915.14535
[5] Jia, S.-L., Chi, Z., Chen, L., Liu, G.-L., Hu, Z., & Chi, Z.-M. (2021). Molecular evolution and regulation of DHN melanin-related gene clusters are closely related to adaptation of different melanin-producing fungi. Genomics, 113(4), 1962–1975. https://doi.org/10.1016/j.ygeno.2021.04.034
[6] Machado, D., Andrejev, S., Tramontano, M., & Patil, K. R. (2018). Fast automated reconstruction of genome-scale metabolic models for microbial species and communities. Nucleic Acids Research, 46(15), 7542–7553. https://doi.org/10.1093/nar/gky537
[7] Modeling | Leiden - iGEM 2023. (2023). Igem.wiki. https://2023.igem.wiki/leiden/model
[8] Stephanopoulos, G. (1999). Metabolic Fluxes and Metabolic Engineering. Metabolic Engineering, 1(1), 1–11. https://doi.org/10.1006/mben.1998.0101
[9] Ranganathan, S., Suthers, P. F., & Maranas, C. D. (2010). OptForce: An Optimization Procedure for Identifying All Genetic Manipulations Leading to Targeted Overproductions. PLoS Computational Biology, 6(4), e1000744. https://doi.org/10.1371/journal.pcbi.1000744
[10] SimBiology. (2024). Mathworks.cn. https://ww2.mathworks.cn/products/simbiology.html
[11] Zhang, L., Li, Y., Wang, Z., Xia, Y., Chen, W., & Tang, K. (2007). Recent developments and future prospects of Vitreoscilla hemoglobin application in metabolic engineering. Biotechnology Advances, 25(2), 123–136. https://doi.org/10.1016/j.biotechadv.2006.11.001
[12] Jaafar, L., & Zueco, J. (2004). Characterization of a glycosylphosphatidylinositol-bound cell-wall protein (GPI-CWP) in Yarrowia lipolytica. Microbiolog, 150(1), 53–60. https://doi.org/10.1099/mic.0.26430-0