Abstract:In the initial conception of our project "Halo World", we planned to transform the extremophile Halomonas TD so that it could better utilise carbon dioxide electrochemical derivatives (CDE). In addition, we also planned to develop a thermosensitive bio-switch suitable for Halomonas TD and fully exploit the production potential of the modified Halomonas TD. Until today, we have successfully obtained Halomonas TD80, which can make fuller use of CDE, built a T-switch suitable for it, and produced three major products covering PHAs, amino derivatives and proteins. What's more, we have also developed a software for genome neutral site screening (which has been applied in our project) and a fermentation model that can realise real-time feeding.

We have successfully integrated the formate assimilation pathway into TD80, resulting in a 66% increase in the growth of TD80. Additionally, we have demonstrated that the THF system can be successfully applied in electrochemical derivatives with acetate as the main carbon source, and the introduction of the THF system has increased the growth of TD80 by 15%. Currently, both acetate and formate in the electrochemical derivatives can be absorbed and utilized by Halomonas, with only minimal non-toxic and harmless bicarbonates remaining.

We have built a thermal bio-switch（what we call "T-switch") suitable for Halomonas TD, which can achieve a leakage with a fluorescence intensity(FI) of less than 10 and a dynamic range of more than 15 fold between 30℃ and 37℃.

·PHAs:PHB&P34HB

The growth curve indicated that cysNC is not an essential gene. Gas chromatography analysis revealed that the gene clusters on the 321 plasmid were lost after the removal of screening pressure, leading some TD80 cells to stop producing 4HB and resulting in a decrease in the 4HB molar ratio. However, the absence of antibiotics did lead to an increase in dry weight. Therefore, identifying a new essential gene may be a viable strategy to enhance the 4HB mol% of P34HB, which will be a focus of our future efforts.

·Amino acid derivative: Tyrian purple

We learned that two key enzymes (Stth and TnaA) in the biosynthetic pathway of Tyrian purple can use tryptophan, which leads to the production of more by-products indigo. Therefore, we planned to verify the application of our T-switch through the production of Tyrian Purple.

·Protein: SOD/PhaP

We successfully integrated the SOD and phaP protein expression pathways into TD80. Currently, in a nitrogen-rich fermentation system, TD80 was able to produce higher amounts of SOD and phaP proteins induced by IPTG at a concentration of 20 mg/L using sodium acetate as a carbon source.

We successfully developed a five-step genomic locus screening program. By incorporating an improved KMP algorithm and the VWZ-curve promoter prediction method, we reduced the single-round runtime of locus screening by over 80% and the multi-round runtime by over 95%. This program successfully identified high-expression loci in Halo6.0 (8 loci), Nissel (6 loci), BL21 (4 loci), LY01 (11 loci), and LY03 (9 loci). Most loci exhibited good expression strength and stability, while a few loci showed no significant differences.

We added GO-term-based enzyme partitioning constraints to the ec-GEM, first quantifying the total enzyme amount for each term, and then iteratively calculating within each term. This approach addressed the issues of high computational complexity and excessive constraints associated with traditional ec-GEM calculations. We also employed deep learning tools, ProtParam and DL-kcat, to predict enzyme parameters (kcat, molecular weight, half-life, and hydrophilicity), enabling more detailed enzyme constraints. Finally, we determined the optimal substrate concentration ranges for cell proliferation and PHA synthesis from the solution space, achieving real-time fermentation feed flow control by establishing differential equations. The model demonstrates strong predictive performance, with an R^2 value of 0.85. The Percentage Root Mean Square Error (PRMSE) between the experimental data and the model's predictions is 13.17%.