Our experimental work is built around a structured, four-stage engineering pipeline for enhancing cyclic lipopeptides (CLPs) production in Bacillus velezensis. This step-wise strategy progresses from random mutagenesis combined with high-throughput screening strategy, to targeted regulated gene editing, followed by metabolic refactoring of the biosynthetic pathways, and CLPs synthesis module-nonribosomal peptide synthetase (NRPS) engineering. This section details the experimental design, methodologies, and results that form the foundation of our iterative 'Design-Build-Test-Learn' cycle.

In this round, we employed a non-rational approach to enhance CLP production by combining ARTP mutagenesis with high-throughput screening. A diverse mutant library of Bacillus velezensis was rapidly generated using the Atmospheric and Room-Temperature Plasma (ARTP) system ^[1]. We then implemented a two-tiered screening strategy, utilizing the high-throughput Oxford cup method for initial selection followed by the minimum inhibition concentration (MIC) assay^[2] to identify mutants with significantly improved CLP yields. Subsequent whole-genome sequencing of these superior performers was conducted to pinpoint key genetic mutations, thereby uncovering potential regulatory genes and providing critical targets for the rational design phase of our project.

Experimental Procedure

• ARTP Mutagenesis:The wild-type Bacillus velezensis strain HBMY was provided by Hubei WEL-SAFE Biotechnology Co., Ltd. Cells in the logarithmic growth phase were subjected to ARTP mutagenesis under the following parameters: power of 100 W, distance of 2 mm, and treatment time of 120 s. The lethality rate was controlled between 70–80%. After treatment, the cells were serially diluted to appropriate concentrations and plated on LB agar to determine cell viability and facilitate mutant isolation.
• Assessment of antimicrobial activity:The antimicrobial activity of the wild-type and mutant strains was evaluated using a dual-method approach: the Oxford cup diffusion assay provided a qualitative assessment, while the minimum inhibitory concentration (MIC) method offered a quantitative measure of antibacterial potency.
• Extraction of CLPs:CLPs were extracted from both the wild-type HMBY and mutant strains via acid precipitation. Briefly, the fermented broth was centrifuged, and the supernatant was collected. The supernatant pH was adjusted to 2.0 using 6 M HCl and left to stand at 4 °C overnight. The sample was then centrifuged at 12,000 rpm for 20 minutes to collect the precipitate, which was subsequently resuspended and diluted in methanol for further analysis.
• Analysis of CLPs:The structural composition of CLPs from strain HMBY was characterized using high-performance liquid chromatography–mass spectrometry (HPLC-MS). Thirteen CLPs were revealed, comprising three iturin, five fengycin and five surfactin. CLP yields were further quantified by HPLC.

In this round, we transitioned to a rational design approach to validate and characterize key genes regulating CLP biosynthesis. Comparative genomic analysis of the high-yield mutant HMBY-106 against the wild-type strain identified candidate mutations in three genes (relA, cdaA, cheB). To confirm their functional roles, we employed a CRISPR-Cas9 system to construct precise mutants of each gene in the wild-type HMBY. These engineered strains were then subjected to a comprehensive phenotypic analysis, including growth profiling, CLP yield quantification, and expression analysis of key Non-ribosomal Peptide Synthetase (NRPS) genes. The resulting data allowed us to elucidate the specific regulatory function of mutant gene and integrate these findings into a proposed mechanistic model for CLP biosynthesis.

Experimental Procedure

• Target identification via comparative genomics:Comparative genomic analysis was performed between the high-yield mutant HMBY-106 and the wild-type Bacillus velezensis HMBY, leading to the identification of candidate regulatory genes (relA, cdaA, cheB) potentially involved in CLP overproduction.
• Gene editing:To validate the function of these candidate genes, a CRISPR-Cas9 system based on the single-plasmid shuttle vector pJOE8999 ^[3] was employed for targeted gene editing in the wild-type HMBY. This system features a temperature-sensitive origin (pE194ts), a kanamycin resistance marker, a mannose-inducible Cas9, and an sgRNA expression cassette. Spacer sequences targeting each gene were cloned via BsaI sites, and mutant gene templates were inserted using SfiI sites. The constructed plasmids were introduced into B. velezensis HMBY by electroporation (2.0 kV, 200 Ω)^[4], followed by selection and screening of mutant strains.CLP yields of the mutants were quantified by HPLC.
• Phenotypic characterization & model building:Phenotypic effects of each mutant were systematically assessed. Bacterial growth was monitored by measuring OD₆₀₀ at hourly intervals. CLP production was quantified using HPLC, and the expression levels of NRPS genes(ituA-D, fenA-E and srfA-D) were analyzed by quantitative reverse transcription PCR (qRT-PCR). These data collectively informed the construction of a regulatory model for CLP biosynthesis in B. velezensis.

This stage adopted a metabolic network reconstruction strategy, focusing on enhancing precursor supply and blocking competing pathways to direct metabolic flux efficiently and specifically towards the target product, Fengycin. By constructing a dual-gRNA-mediated multi-gene cluster knockout system, the surfactin (srfA-D), iturin (ituA-D), and polyketide synthase gene clusters (ppsCD) were simultaneously knocked out. This overcame challenges associated with long-fragment genetic manipulation, achieving streamlining and optimization of the metabolic pathway.

Experimental Procedure

• Dual-gRNA plasmid system construction:Using pJOE8999 as the backbone, a dual-guide RNA expression cassette was assembled through multiple rounds of PCR amplification and subsequently cloned into the BsaI and XbaI restriction sites of the vector. Dual gRNAs targeting the surfactin (srfA-D) and iturin (ituA-D) gene clusters were inserted using BsaI and SalI sites, while corresponding homologous repair templates were integrated via SfiI sites. The resulting plasmid constructs were then introduced into B. velezensis HMBY-rel via electroporation (2.0 kV, 200  Ω), and transformants were selected and screened to obtain the desired mutant strains.
• Assessemnt of antimicrobial activity and fengycin yield:The antimicrobial activity of the mutant strains was evaluated using the Oxford cup diffusion assay and the minimum inhibitory concentration (MIC) method by using Xanthomonas spp. QKHT-5 as indicator. fengycin yields of the mutants were quantified by HPLC.

In this round, we implemented a computational strategy to redesign the communication (COM) domains within the NRPS machinery. Through multiple sequence alignment, domain architecture analysis, and phylogenetic inference, we identified key residues governing COM domain interactions and specificity. Guided by predictions from our custom-developed software, we engineered targeted mutations in the COM pairing regions of the high-yielding strain. The resulting mutant strains were evaluated for antimicrobial activity and fengycin yield to experimentally validate the computational predictions and assess the success of NRPS reprogramming.

Experimental Procedure

• Software development and computational design: A systematic bioinformatic workflow was implemented to guide the reprogramming of NRPS COM domains. First, multiple sequence alignment of collected fengycin synthetase sequences was performed using MAFFT^[4] , followed by the construction of a maximum-likelihood phylogenetic tree with IQ-TREE^[4] . This step enabled the evolutionary positioning of target COM domains and the identification of conserved and variable regions. Subsequently, the InterProScan tool^[5] was employed to annotate essential catalytic domains (C, A, T, and G) and generate a domain architecture map for precise boundary definition. Finally, COM domain interfaces were automatically extracted (spanning 30 amino acids on each side of the module junction), and integrated evolutionary and structural analyses were applied to predict key residue regions influencing domain interaction, thereby providing focused targets for site-directed mutagenesis.
• COM domain reprogramming and fengycin variants analysis: To reprogram the COM domain interactions within the fengycin NRPS machinery, we performed precise genome editing using our CRISPR-Cas9 system, guided by computational predictions from our custom software. The resulting mutant strains were analyzed by HPLC-MS to characterize the structural profiles of the engineered fengycin variants. Antimicrobial activity was quantitatively assessed using the Oxford cup diffusion assay and the minimum inhibitory concentration (MIC) method, with Xanthomonas spp. QKHT-5 as the indicator strain.

[1]Wang, X., Lu, M., Wang, S., Fang, Y., Wang, D., Ren, W., et al. (2014) The atmospheric and room-temperature plasma (ARTP) method on the dextranase activity and structure. Int J Biol Macromol, 70, 284-291.

[2]Robas, M., Probanza, A., González, D., Jiménez, P. A. J. I. J. o. E. R. and Health, P. (2021) Mercury and Antibiotic Resistance Co-Selection in Bacillus sp. Isolates from the Almadén Mining District. International Journal of Environmental Research and Public Health, 18, 8304.

[3]Altenbuchner, J. (2016) Editing of the Bacillus subtilis Genome by the CRISPR-Cas9 System. Appl Environ Microbiol, 82, 5421-5427

[4]Tian, H., Liu, B., Yang, J., Zhou, C., Xu, X., Zhang, Y., et al. (2021) Genetic transformation system for Bacillus velezensis NSZ-YBGJ001 and curing of the endogenous plasmid pBV01. Biotechnol Lett, 43, 1595-1605.

[5]Katoh, K., Rozewicki, J., Yamada, K.D. (2019) MAFFT online service: multiple sequence alignment, interactive sequence choice and visualization. Briefings in Bioinformatics, 20, 1160-1166.

[6]Minh, B.Q., Schmidt, H.A., Chernomor, O., Schrempf, D., Woodhams, M.D., von Haeseler, A., et al. (2020) IQ-TREE 2: new models and efficient methods for phylogenetic inference in the genomic era. Molecular Biology and Evolution, 37, 1530-1534.