A combination of genome reduction and promoter engineering can enhance surfactin production by Bacillus amyloliquefaciens LL3

Background: Genome reduction and metabolic engineering have emerged as intensive research hotspots for constructing the optimal functional chassis and various microbial cell factories. Surfactin, a lipopeptide-type biosurfactant with broad spectrum antibiotic activity, has wide application prospects in anticancer therapy, biocontrol and bioremediation. Bacillus amyloliquefaciens LL3, previously isolated by our lab, contains an intact srfA operon in the genome for surfactin biosynthesis. Results: In this study, a genome-reduced strain GR167 lacking ~4.18% of the B. amyloliquefaciens LL3 genome was constructed by deleting some unnecessary genomic regions. Compared with the strain NK-1 (LL3 derivative, ΔuppΔpMC1), GR167 harbored faster growth rate, higher transformation eciency, increased intracellular reducing power level and higher heterologous protein productivity. Furthermore, the optimal chassis GR167 was engineered for enhanced surfactin production. Firstly, the iturin and fengycin biosynthetic gene clusters were deleted from GR167 to generate GR167ID. Subsequently, two strong promoters PR suc and PR tpxi from LL3 were obtained by RNA-seq and promoter strength characterization, and then they were individually substituted for the native srfA promoter in GR167ID to generate GR167IDS and GR167IDT. The best mutant GR167IDS showed a 678-fold improvement in the transcriptional level of the srfA operon relative to GR167ID, and it produced 311.35 mg/L surfactin, with a 10.4-fold increase relative to GR167. Conclusions: The genome-reduced strain GR167 was advantageous over the parental strain in several industrially relevant physiological traits assessed and it was highlighted as an optimum chassis for enhanced surfactin production. In future studies, further reduction of the LL3 genome can be expected to create high-performance chassis for synthetic biology applications.


Introduction
With the development of systems and synthetic biology, numerous studies have focused on the design and construction of the optimal functional microbial chassis with reduced genomes and superior physiological characteristics [1,2]. Moderate genome reduction can create synthetic biology chassis with optimized genomic sequences, e cient metabolic regulatory networks and superior cellular physiological characteristics [3][4][5]. So far, several model microorganisms, such as Escherichia coli [6], Bacillus subtilis [7,8] and Pseudomonas putida [9], have been intensively researched for minimal genome construction due to their clear genetic background and e cient genome editing approaches.
In recent years, several metabolic engineering strategies have been proposed for enhancing biosurfactant production, mainly including promoter engineering [13][14][15], the reduction of by-product formation [16], the enhancement of the precursor supply [2], the improvement of biosurfactant transmembrane e ux [17], and the modi cation of global regulatory factors [15,16]. Among which, promoter engineering is highlighted as a powerful tool for enhancing the titer of biosurfactants. For example, the titer of iturin A was increased from an undetectable level to 37.35 mg/L by inserting a strong promoter C2up into upstream of the itu operon in B. amyloliquefaciens [15]. In another study, the titer of surfactin in B. subtilis was elevated from 0.07 g/L to 0.26 g/L by the replacement of the native srfA promoter with a constitutive promoter P veg [18]. In addition to the natural promoters, Jiao et al [14] developed a chimeric promoter Pg3 for driving the synthesis of surfactin, resulting in a 15.6-fold increase in the titer of surfactin relative to the wild-type B. subtilis THY-7. However, e cient promoters need to be explored for the enhancement of biosurfactants production by members of the genus Bacillus.
Currently, endogenous promoters are highlighted as promising candidates for improved production of bacterial secondary metabolites [19]. For example, 14 endogenous promoters identi ed from Streptomyces albus J1074 by RNA-seq and reporter assays were successfully used to activate a cryptic gene cluster in S. griseus [20]. In another study, four endogenous promoters identi ed from S. coelicolor M145 by RNA-seq and reporter assays were used to activate cryptic biosynthetic clusters for jadomycin B production in S. venezuelae ISP5230 [9]. B. amyloliquefaciens LL3 was isolated initially for poly-γ-glutamic acid (γ-PGA) production by our lab, and whole genome of LL3 is currently available in the GenBank database (accession no. NC_017190.1) [21]. LL3 has a genomic size of 3,995,227 bp with an average G + C content of 45.7% and a circular plasmid (pMC1) of 6,758 bp. In particular, an intact srfA operon was found in the genome of LL3, suggesting the capability for surfactin biosynthesis. The essential genes and genomic islands (GIs) in LL3 were also identi ed by the Essential Genes Database (http://tubic.tju.edu.cn/deg/) and GIs Analysis Software (http://tubic.tju.edu.cn/GC-Pro le/). Previously, a marker-free large fragments deletion method was well established in LL3 [22]. Therefore, previous studies have laid a foundation for genome reduction and enhanced surfactin production in LL3.
In this study, a genome-reduced strain GR167 was constructed from B. amyloliquefaciens NK-1 (LL3 derivative, ΔuppΔpMC1) [23] and evaluated as an optimum chassis for several physiological traits.
Furthermore, GR167 was engineered using metabolic engineering strategies for enhanced surfactin production.

Results And Discussion
Construction of a genome-reduced B. amyloliquefaciens strain GR167 To adapt to the adverse environmental conditions, there is a common mechanism horizontal gene transfer (HGT) among microorganisms, enabling host bacteria to acquire larger DNA segments, i.e., GIs, the G+C contents of which are signi cantly different from that of the core genome [24]. GIs usually carry some functional genes related to pathogenicity and antibiotic resistance, leading to the emergence of multiple resistant bacteria by HGT [25]. In addition, there are latent secondary metabolic biosynthesis gene clusters scattered across the LL3 genome, which may increase the metabolic burden on cells and the puri cation cost of target products [26]. Consequently, to streamline the genome of LL3, the GIs containing putative protein genes, antibiotic biosynthesis genes and prophage protein genes, where the G+C contents deviate signi cantly from 45.7%, were selected as the knockout targets. Besides, the gene clusters eps, bae and pgsBCA responsible for the biosynthesis of extracellular polysaccharides, bacillaene and γ-PGA, respectively, which consume more energy and substrates, were also deleted from the LL3 genome. The detailed information on the deleted regions is summarized in Table S1. The schematic diagram for deletion of large genomic segments in LL3 is presented in Figure S1. Overall, a genome-reduced strain GR167 lacking ~4.18% of the LL3 genome was generated from NK-1 via a markerless deletion method [22]. The exact coordinates (G1 to G6) of the deleted regions on chromosome and the physical map of the endogenous plasmid pMC1 are shown in Fig. 1a and b, respectively.
Deleting redundant genes from a bacterial genome is expected to create superior chassis cells for the industrial production of valuable bio-based chemicals. Due to the existence of unannotated genes in the LL3 genome and lack of insight into the interactions among known genes, several industrially-relevant physiological traits were evaluated to determine whether GR167 is an ideal chassis for enhanced production of surfactin.
Genome reduction can improve the growth rate of LL3 To evaluate the effect of non-essential genomic sequences on cell growth, the growth pro les of GR167 and its parental strain NK-1 were detected by following the optical density (OD 600 ) of cells cultured in both poor (M9 medium) and rich (LB medium) conditions. As shown in Fig. 2a, obviously, whether incubated in LB or M9 medium, GR167 grew faster and yielded higher biomass with approximately 1.5 and 1.2-fold higher at the plateau phase than that of NK-1, respectively. To further quantify the growth parameters, the maximum speci c growth rates (μ max ) of both strains were determined during exponential growth (Fig. 2b). The μ max values of GR167 were 23.7% and 67% higher than that of NK-1 when cultured in LB and M9 medium, respectively. Due to the block of secondary metabolite biosynthesis pathways, more energy and substrates were used for basal metabolism and cell proliferation in GR167.
When cultured in M9 medium, the μ max of NK-1 was only 0.185 ± 0.004 h -1 , with a 30.7% decrease relative to that measured in LB, while GR167 showed a similar growth behavior in both media, suggesting that nutrition may be one of the main growth-limiting factors for NK-1 but not for GR167.
Deletion of non-essential genes may perturb cellular metabolism and thus impair cell growth [27,28]. On the contrary, the genome-reduced strain GR167 acquired bene cial growth tness, which was in agreement with previous studies [29,30]. Overall, in this study, there was a positive correlation between cell growth and cumulative deletions, and deleting ~4.18% of the LL3 genome did not affect cellular viability of GR167. Moreover, the growth rates of GR167 outcompeted the parental strain under the tested culture conditions, making it a promising chassis for further genetic engineering.

Genome reduction can broaden the range of carbon sources utilized by LL3
To further evaluate the changes in the metabolic potential of GR167 and NK-1, their ability to utilize various substrates was analyzed by a GEN III MicroPlate containing 23 carbon sources tested. As shown in Table 1, the substrates utilized by GR167 and NK-1 were signi cantly different with each other. Eight carbon sources could be e ciently metabolized by GR167, especially L-aspartate and methyl pyruvate, with a 30% and 43% increase in the utilization ratio compared to NK-1, respectively, suggesting that genome reduction may improve the capacity of LL3 to utilize certain substrates.

Genome reduction can improve transformation e ciency
An ideal chassis cell is expected to possess the excellent capacity to take up exogenous plasmids. As shown in Fig. 2c, when transformed with plasmid pHT01, GR167 surpassed the transformation e ciency of the parental strain NK-1 by about 137%. The GIs and BGCs deleted in this study may contain negative regulators related to transformation e ciency, making competent cells in the optimal DNA uptake state during electroporation. Li et al [4] found that the transformation e ciency of genome-reduced strains decreased with cumulative genomic deletions. In addition, similar to the results of the transformation e ciency, the growth rates of all mutants were inferior to the parental strain [4]. In this study, on the contrary, both the growth rate and transformation e ciency of GR167 were obviously higher than that of . Similarly, in another study, an E. coli mutant MDS12 lacking 8.1% of the genome of the parental strain also displayed a positive correlation between the growth parameters and transformation e ciency [26]. We therefore speculate that higher transformation e ciency may be associated with the improved growth tness of GR167, notwithstanding which may be a synergistic effect caused by many physiological characteristics [31].
Genome reduction can increase intracellular reducing power and the productivity of heterologous proteins The intracellular reducing power (NADPH/NADP + ), which is indispensable for basic anabolic processes [32], was measured in this study. The intracellular NADPH/NADP + ratio of GR167 increased by 21.4% compared to the parental strain NK-1 (Fig. 2d), which may be attributed to the deletion of some NADPHconsuming biosynthesis pathways such as γ-PGA biosynthesis [33]. The improvement of intracellular reducing power level may be bene cial for GR167 to act as an ideal chassis for enhanced production of secondary metabolites.
Also, an optimal chassis is expected to possess high heterologous protein productivity. In a previous study, prophage and hypothetical proteins accounting for 45.6% and 54.4% of the genome of Lactococcus lactis NZ9000, respectively, were deleted, resulting in a signi cant increase in the production capability of red uorescent protein [5]. In another study, a genome-reduced strain EM383 was constructed from P. putida KT2440 by deleting agellar operon and prophage protein genes, leading to a 40% increase in the production capability of foreign proteins [34].
In this study, GFP was selected as a model protein to determine the heterologous protein productivity. As shown in Fig. 2e, when transformed with plasmid pHT-P 43 -gfp, the relative uorescence intensity of GR167 was 50.4% higher than that of NK-1, indicating that the productivity of heterologous proteins was signi cantly improved by genome reduction.
Use of genome-reduced strain GR167 as an optimal chassis for surfactin production For surafctin, it can hardly achieve a signi cant breakthrough in production only through traditional fermentation optimization because of its low yield in wild strains [14,35]. Strategies for surfactin overproduction were focused on strain modi cation recent years, such as substitution of the native promoter P srf of srfA operon [13,14], overexpressing transporters to enhance surfactin e ux [17], and modifying the regulators ComX and PhrC [35]. However, most modi cations were performed in existing strains. In our study, a genome-reduced strain GR167 with intact surfactin synthase operon was evaluated as an ideal chassis for its superior physiological characteristics. Engineering and modifying microbial chassis may maximize its practical application ranges and obtain maximum theoretical yields of bioproducts of interests. In a previous study, by deleting and co-overexpressing speci c genes conducive to guanosine accumulation in a genome-reduced strain B. subtilis BSK814, the guanosine titer in the nal strain was 4.4-fold higher than that in the control strain bearing the same genetic modi cations [4]. In another study, BSK814 was also endowed with the ability to produce acetoin using xylose as carbon source by modifying xylose utilization related pathways [36]. Therefore, genome reduction may provide a desirable chassis for further strain modi cation, and metabolic engineering of genome-reduced strains may be more bene cial to the development of microbial cell factories.
As shown in Fig. 3, surfactin production by GR167 was demonstrated by high-performance liquid chromatography (HPLC). Compared with NK-ΔLP (NK-1 derivative, ΔpgsBCA) [37], a slight increase in the surfaction titer was observed with GR167. Consequently, it is interesting and necessary to explore whether microbial cell factories with high surfactin production capabilities can be constructed by further modi cation of GR167.
Enhancing surfactin production by blocking the potential competitive pathways A transcriptional comparison between B. amyloliquefaciens LL3 and NK-ΔLP using RNA-seq revealed that the transcriptional levels of the gene clusters srfA, itu and fen, responsible for surfactin, iturin A and fengycin biosynthesis were all up-regulated (unpublished data). Iturin A and fengycin belonging to CLP antibiotics are structural analogues of surfactin [38], possibly leading to the reduction of the purity of the extracted surfactin from the culture supernatant. Iturin A and fengycin are synthesized by NRPSs like surfactin [11]; thus, they may share similar biosynthesis mechanisms with surfactin and their biosynthesis may compete for NADPH, energy and direct precursors with surfactin biosynthesis. In this study, the gene clusters itu (37.2 kb) and fen (11.5 kb) were deleted to enhance surfactin production. The resulting mutants were designated as GR167I (Δitu), GR167D (Δfen) and GR167ID (Δitu, Δfen). The titer of surfactin was increased to 32.88 mg/L in GR167ID, with a 10% and 56% improvement in the titer and speci c productivity of surfactin compared to GR167, respectively (Fig. 4). We speculate that blocking the potential competitive pathways may eliminate the competition for the same amino acid precursors, allowing for the redistribution of substrates towards surfactin biosynthesis.
Construction of endogenous promoter library of B. amyloliquefaciens LL3 Promoter engineering is considered as a promising approach for enhanced production of bacterial secondary metabolites [9,19,20]. FPKM (fragements per kilobase million) value is positively correlated with the transcriptional activity of a gene [39], which therefore can be regarded as an indicator for initial screening of promoters. Through RNA-seq analysis of LL3, all genes were ranked and classi ed into three groups based on their FPKM values, i.e., lower than 1,250, 1,250-4,000 and higher than 4,000. Then, the rst six genes with higher FPKM values in each group were selected, and their upstream regions were predicted and cloned as described in Methods, named PR x [x: the name of various related genes; PR: the sequences of predicted promoters with their ribosomal binding sites (RBSs)] and represented weak, moderate and strong promoters, respectively (Table 2). Subsequently, various reporter gene vectors derived from pHT01 containing fused fragments of the predicted promoters and gfp gene were used to assess the strengths of the tested promoters in LL3.

Characterization of the selected promoters via qPCR (quantitative real-time PCR) and GFP uorescence measurement
As shown in Fig. 5a, the relative transcriptional levels of the candidate promoters measured with reporter gene vectors were PR ldh , PR ahp , PR hem , PR tpxi , PR clp , PR suc , PR accD , PR gltA , PR rpsu , PR nfrA , PR gltX , PR ydh , PR ugt , PR arg , PR nad , PR lac , PR alsD , PR hom and PR pgmi in a descending order, which were inconsistent with the strengths of the promoters shown by the FPKM values (Table 2), with similar results reported in a previous study [19]. We speculate that the transcription of a gene on chromosome may be affected and regulated by anking genes and regulatory sequences. However, this interference can be eliminated if a promoter is inserted into a plasmid.
To further determine the production capabilities of GFP, in this study, the relative uorescence intensities of GFP were also measured in LL3. Among the 18 endogenous promoters, PR ahp showed the strongest production capacity of GFP, followed by PR suc , PR tpxi , PR rpsU , PR hem and PR ydh (Fig. 5b). However, the rst six promoters were PR ldh , PR ahp , PR hem , PR tpxi , PR clp and PR suc from high to low at the transcriptional levels (Fig. 5a). Considering the different RBSs located upstream of the promoters evaluated in this study, we speculate that the different RBSs may affect the translational initiation e ciencies of mRNA corresponding to GFP, leading to the different trends between the transcriptional level and production capacity of GFP.
Substitution of the native srfA promoter enhanced surfactin production Considering the heterologous expression of srfA is challenging for which large genetic sequence (over 25 kb) [40], substitution of the native srfA promoter by strong promoters is considered more bene cial for enhanced transcription of srfA operon [13,14,18]. For example, Sun et al [13] replaced the native P srf promoter of B. subtilis, resulting in a 10-fold improvement in the titer of surfactin. In this study, two strong promoters PR suc and PR tpxi , of which nucleotide sequences are shown in supplementary material, derived from endogenous promoter library of LL3, were integrated into upstream of the srfA operon in GR167ID to construct surfactin hyperproducers GR167IDS and GR167IDT. As expected, both the surfactin production and speci c productivity exhibited a signi cant elevation ( Fig. 6a and b). In particular, the PR suc promotersubstituted strain GR167IDS produced 311.35 mg/L surfactin, which was about 9.5-fold higher than that of GR167ID (Fig. 6a). Meanwhile, the transcriptional level of srfA operon in GR167IDS was 678-fold higher than that in GR167ID (Fig. 6c), indicating that the endogenous promoter PR suc could signi cantly improve surfactin production by enhancing the transcription of srfA operon in B. amyloliquefaciens LL3.

Conclusions
In summary, a genome-reduced strain GR167 was constructed by deleting some non-essential genes accounting for ~ 4.18% of the LL3 genome and outcompeted the parental strain in several physiological traits assessed. GR167IDS, obtained from GR167 by promoter substitution, showed a 10.4-fold improvement in the titer of surfactin compared to GR167. The current results suggest that genome reduction in combination with promoter engineering may be a feasible strategy for the development of microbial cell factories capable of e ciently producing bacterial secondary metabolites.

Methods
Bacterial strains, media, and culture conditions Escherichia coli DH5α was employed for plasmid construction and propagation. For the subsequent successful electroporation of B. amyloliquefaciens strains, the E. coli JM110 was used as intermediate host to demethylate the desirable plasmids from E. coli DH5α. E. coli strains were incubated at 37 °C in Luria-Bertani (LB) broth. B. amyloliquefaciens LL3 was deposited in the China Center for Type Culture Collection (CCTCC) (accession number: CCTCC M 208109). B. amyloliquefaciens NK-1 was employed as the parental strain for genome reduction. GR167 was used as the starting strain for engineered highyielding surfactin producing mutants. All B. amyloliquefaciens strains were cultured in M9 mineral medium [41] and LB medium at 37 °C and 42 °C. For lipopeptide surfactin production, B. amyloliquefaciens was incubated at 30 °C and 180 rpm for 48 h in Landy medium [42]. When appropriate, media were supplemented with ampicillin (Ap; 100 μg/mL), chloramphenicol (Cm; 5 μg/mL) or 5-uorouracil (5-FU; 1.3 mM).

Plasmid and strain construction
To construct the gene deletion vectors, the temperature-sensitive plasmid pKSU with an upp expression cassette was used [23]. The upstream and downstream fragments of the deleted genomic regions were ampli ed by PCR and then the two fragments were joined by overlap PCR. The generated fragment was ligated into pKSU via homologous recombination, to generate the gene deletion vectors. Introduction of plasmid into B. amyloliquefaciens was carried out using an optimized high osmolarity electroporation method [43]. To carry out multiple gene deletions on a single strain, a marker-less gene deletion method was used to construct the gene knockout mutants [22]. All the constructed plasmids and mutant strains were validated by PCR detection and DNA sequencing. All plasmids, strains, and primers used in this study are listed in Table 3, Table 4, and Table S2, respectively.

Physiological traits assessment
Growth pro les of GR167 and NK-1 were measured in both M9 mineral medium and LB medium. Overnight cultures (1 mL) were inoculated into 100 mL LB or M9 medium in 500-mL asks and then incubated for 20 h at 37 °C and 180 rpm. To determine the bacterial growth status, the OD 600 was monitored every 2 h using a UV-1800 spectrophotometer (Shimadzu, Kyoto, Japan).
The metabolic phenotypic analyses were performed with a Biolog GEN Ш MicroPlate™ using a phenotype microarray system (Biolog Inc., California, USA) according to the manufacturer's instructions. The bacterial cells on the solid medium surface were collected by cotton swab, dissolved into the inoculating uid IF-B, and then the cell density was adjusted to a range of 80-86% Turbidity. Subsequently, 100 μl of bacterial suspensions were pipetted into the Biolog GEN Ш plates with different substrates. After the samples were incubated at 33 °C for 48 h, the absorbance at 590 nm was measured with the Biolog reader and the test data were analyzed by the Biolog system.
Electro-competent cells of GR167 and NK-1 (2 × 10 10 CFU/mL) were prepared according to previous methods [15]. Subsequently, approximately 100 ng of plasmid pHT01 was absorbed by 100 μL of electrocompetent cells via electroporation. After 3 h of incubation at 30 °C and 180 rpm, the mixture was spread on LB agar plates supplemented with 5 μg/mL Cm. The number of colonies was calculated to evaluate the transformation e ciency.
Cells were cultured in LB medium at 37 °C for 18 h. The intracellular cofactors NADPH and NADP + were extracted and quanti ed by enzymatic methods [44] using an EnzyChrom™ assay kit (BioAssay Systems, USA) according to the manufacturer's protocols.
The heterologous protein productivity was determined by introducing plasmid pHT-P 43 -gfp into GR167 and NK-1. The detailed protocols for strain cultivation and uorescence intensity measurement refer to our previous work [9]. The relative uorescence intensity was normalized against per OD 600 of whole cells. The uorescence signal of NK-1 harboring pHT01 was set as background and was subtracted from overall uorescence.
RNA-seq, promoter prediction, and construction of reporter gene vectors RNA-seq analyses of LL3 were carried out according to our previous methods [33]. The expression levels of the predicted genes were quanti ed as the FPKM value [45]. The upstream regions of genes with different FPKM values were submitted online (http://www.fruit y.org/seq_tools/promoter) for promoter prediction.
Furthermore, each promoter sequence plus its native RBS and gfp gene were ampli ed by PCR from the LL3 genome and pHT-P 43 -gfp, respectively. Subsequently, 3′-end of a promoter sequence was fused with 5′-end of gfp gene and the fusion fragment was inserted into plasmid pHT01, to generate reporter gene vector pHT-PR x -gfp for promoter strength characterization ( Figure S2). Moreover, a control vector pHT-PR lac -gfp with gfp expression driven by lac promoter was similarly constructed with pHT01.     Table 4 Strains used in this study Figure 1 The construction of genome-reduced strain GR167. a The exact coordinates of the deleted regions (G1 to G6) on the chromosome of B. amyloliquefaciens LL3; b the physical map of the cured endogenous plasmid pMC1 Figure 2 Physiological characteristics assessment of genome-reduced strain GR167 and the parental strain NK-1.

Figures
a Growth curves; b maximum speci c growth rate (μmax); c transformation e ciency; d intracellular reducing power level (NADPH/NADP+ molar ratio); e heterologous protein productivity (GFP, green uorescent protein; FI, uorescence intensity). Values denote mean ± SD of triplicates (*P < 0.05, **P < 0.01)  Surfactin production by GR167 and its derivatives (Δitu, Δfen). a Surfactin production; b speci c productivity of surfactin (mg/g, the ratio of surfactin to CDW). To accumulate surfactin, the strains were incubated in Landy medium for 48 h at 30 °C and 180 rpm. Values denote mean ± SD of triplicates (**P < 0.01) Figure 5 Characterization of the strengths of the selected endogenous promoters using reporter gene assays in LL3. a Transcriptional levels of gfp gene quanti ed via qPCR under the control of different promoters (rpsU gene was used as internal standard; the transcriptional level of gfp gene controlled by lac promoter was set as 1); b the relative uorescence intensity of GFP (FI/OD600) under the control of different promoters. Values denote mean ± SD of triplicates Surfactin production by GR167ID, GR167IDS and GR167IDT, and transcriptional levels of srfA operon in the strains. a Surfactin production; b speci c productivity of surfactin (mg/g, the ratio of surfactin to CDW); c transcriptional levels of srfA operon quanti ed via qPCR (the transcriptional level of srfA operon in GR167ID was set as 1). Values denote mean ± SD of triplicates (*P < 0.05, **P < 0.01)

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.