Mutations in genes encoding antibiotic substances increase the synthesis of poly‐γ‐glutamic acid in Bacillus amyloliquefaciens LL3

Abstract Poly‐γ‐glutamic acid (γ‐PGA) is an important natural biopolymer that is used widely in fields of foods, medicine, cosmetics, and agriculture. Several B. amyloliquefaciens LL3 mutants were constructed to improve γ‐PGA synthesis via single or multiple marker‐less in‐frame deletions of four gene clusters (itu, bae, srf, and fen) encoding antibiotic substances. γ‐PGA synthesis by the Δsrf mutant showed a slight increase (4.1 g/L) compared with that of the wild‐type strain (3.3 g/L). The ΔituΔsrf mutant showed increased γ‐PGA yield from 3.3 to 4.5 g/L, with an increase of 36.4%. The γ‐PGA yield of the ΔituΔsrfΔfen and ΔituΔsrfΔfenΔbae mutants did not show a further increase. The four gene clusters’ roles in swarming motility and biofilm formation were also studied. The Δsrf and Δbae mutant strains were both significantly defective in swarming, indicating that bacillaene and surfactin are involved in swarming motility of B. amyloliquefaciens LL3. Furthermore, Δsrf and Δitu mutant strains were obviously defective in biofilm formation; therefore, iturin and surfactin must play important roles in biofilm formation in B. amyloliquefaciens LL3.

clusters are also targets of genome reduction applications because they are dispensable for the cell's growth in rich media.
B. amyloliquefaciens LL3 is a glutamic acid-independent poly-γglutamic acid (γ-PGA)-producing strain that was isolated from traditional fermented food. γ-PGA is a promising biomaterial that is nonribosomally synthesized by the PgsBCA synthetase complex using l-and d-glutamic acids as substrates, which exhibits outstanding qualities, such as good water solubility, biocompatibility, and biodegradability (Shih & Van, 2001). It is widely used in hydrogels, flocculants, drug delivery, cosmetics, and feed additives (Sung et al., 2005). B. amyloliquefaciens LL3 produces γ-PGA without additional glutamic acid in the fermentation medium and thus has great potential in industrial production systems because of the lower cost and simplified process (Cao, et al., 2011). The practical use of γ-PGA is still largely hindered by its low yield, and thus, intensive investigations have been launched to enhance its production, including optimization of fermentation conditions, modification of existing producers, and identification of new wild producers. With the availability of more and more gene manipulation methods, genome-scale modification of existing producers becomes affordable.
In the past decades, strategies to improve γ-PGA production were limited to optimization of medium and fermentation conditions. In the 21st century, there have been some attempts to improve the γ-PGA yield using metabolic engineering strategies. Yeh, Wang, Lo, Chan, and Lin (2010) integrated an efficient synthetic expression control sequence (SECS) into the upstream region of the silent pgsBCA gene cluster of B. subtilis DB430 to produce γ-PGA. Liu et al. (2011) enhanced γ-PGA productivity by depressing exopolysaccharides production. VHb (Vitreoscilla hemoglobin) alleviates the oxygen limitation at the later stage of fermentation. The encoding gene, vgb, has been successfully expressed in a γ-PGA-producing strain to improve γ-PGA production, especially under oxygen-limited conditions (Richard & Margaritis, 2003;Zhang et al. 2013;Su et al., 2010).
Heterologous expression of the γ-PGA synthetase complex (pgsBCA) is another strategy for γ-PGA production improvement, which has been carried out in coryneform bacteria. Corynebacterium glutamicum E12 harboring vector pMT-HCE-pgsBCA could express γ-PGA synthetase genes from B. subtilis and could be considered as a host for γ-PGA synthesis (Sung et al., 2005). Feng et al. (2015) and Feng, Gu, Sun, Han, Yang et al. (2014) enhanced γ-PGA production of B. amyloliquefaciens LL3 by metabolically engineering its γ-PGA synthesis-related metabolic networks: by-products synthesis, γ-PGA degradation, glutamate precursor synthesis, γ-PGA synthesis, and autoinducer synthesis. However, few reports have focused on the antibiotic substances, which may compete with γ-PGA for similar synthesis machinery or substrates.
The whole genome of B. amyloliquefaciens LL3 was sequenced (Geng et al. 2011), and several gene clusters responsible for the synthesis of antibiotic substances were found, including the bae, srf, fen, and itu clusters, as shown in Figure S1. The bae cluster (annotated as pks in B. subtilis 168) encodes bacillaene, which was originally discovered as a bacteriostatic agent that inhibited prokaryotic protein synthesis.
A transcriptional comparison between B. amyloliquefaciens LL3 (γ-PGA + ) and LL3ΔpgsBCA (γ-PGA − ) was performed using RNA-seq (unpublished data). Interestingly, the transcript levels of the bae, srf, itu, and fen clusters experienced a sharp increase in B. amyloliquefaciens LL3 ΔpgsBCA (Table 1). Specifically, the expression levels of the T A B L E 1 Comparison of expression level of the genes which encode the four antibiotic substance between the B. amyloliquefaciens LL3 (γ-PGA + ) and LL3 ΔpgsBCA (γ-PGA − ) first genes of the four aforementioned clusters in B. amyloliquefaciens LL3ΔpgsBCA were 14.42-, 8.08-, 12-, and 9.93-fold higher than that in B. amyloliquefaciens LL3. This suggested that the synthesis of γ-P-GA may suppress the transcription of the above four clusters. Hence, we proposed that the synthesis of the four antibiotic substances may consequently suppress the γ-PGA synthesis directly or indirectly in turn. In addition, bacillaene, surfactin, iturin A, and fengycin as well as γ-PGA are all nonribosomally produced. Moreover, surfactin, iturin A, and fengycin contain several glutamates or glutamines, which are the precursors of γ-PGA (Fig. 1B). Besides, acetyl-CoA, main precursor of bacillaene, plays important role in TCA cycle, which offers glutamate as precursor for γ-PGA synthesis. What is more, lipopeptides or polyketides may be viewed as costly for the cells from a metabolic point of view given the big size of the corresponding operons.

| Plasmid construction and bacterial transformation
The plasmids and primers used in this study are listed in Table 2 and

| Markerless deletion of the four gene clusters
Gene deletions in this study were carried out adapting a previously reported markerless gene replacement method based on upp and will be described briefly below . B. amyloliquefaciens LL3Δupp carrying an in-frame deletion of upp and that is resistant to 1.3 mmol/L 5-fluorouracil (5-FU) was used as the parental strain for subsequent mutants construction. Introduction of the deletion plasmid pKSU would restore sensitivity to 5-FU for B. amyloliquefaciens LL3 Δupp and its derivatives.
Deletion of the srf cluster will be used as an example to explain the method. The up-and downstream homologous arms (~1 kb each) used for srf deletion were obtained using primer pairs SrfUP-F/  B. amyloliquefaciens LL3Δitu was transformed with pKSU-Δsrf to obtain LL-UIS (ΔsrfΔitu), which carries double deletions of the itu and srf clusters. The fen cluster was then deleted in strain LL-IS to obtain LL-ISF (ΔsrfΔituΔfen), which carries triple deletions. Finally, the bae cluster was accumulated in strain LL-UISF to yield LL-ISFB (ΔsrfΔituΔfenΔbae), which is deficient in all four antibiotic substances. to determine the DCW. The supernatant was used to extract γ-PGA, using an ethanol precipitation method, as previously described (Zhang et al., 2013). Experiments were independently repeated at least three times, and the means and standard deviations were calculated.

| qRT-PCR analysis of the pgsB gene
The wild-type B. amyloliquefaciens LL3 and its derivatives were grown to mid-log phase (approximately 20 hr) in fermentation medium. The cells were collected at 4°C and RNA was isolated using TransZolTM Up (TransGen, Beijing, China), according to the manufacturer's instructions. cDNA was reverse transcribed using a GoScriptTM Reverse Transcription System (Promega, WI, USA). Real-time PCR analysis for the target genes was performed using the SYBR ® PremixEx Taq TM II (Takara, Dalian, China). Transcript levels of the target genes were normalized against the levels of rspU (Feng, Goa, Gu, Zang, Cao et al., 2014).

| Isolation of cyclic lipopeptides and HPLC-MS analysis
Isolation of surfactins, fengycins, and iturin A and HPLC-MS analysis were carried out using a method described previously (Luo, Liu, Zhou, Wang, & Chen, 2015). All the samples were further analyzed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MS) with a Shimadzu 2020 series HPLC-MS/MS system (Shimadzu, Japan).

| Swarming and biofilm formation experiments
Fresh colonies of strains to be tested were inoculated and cultivated overnight in LB medium. Then, 10-μL culture was spotted in the middle of plates containing different agarose concentrations (0.25%, 0.5%, and 0.7%) and incubated for 24 hr. The biofilm formation experiment was performed as described by . Overnight cultures of the wild-type and its derivatives were diluted to an OD 600 of 1.0 in fresh LB medium.

T A B L E 3 Oligonucleotide primers used in this study
Samples of 10 μl of the diluted cells were then added to 10 ml of MSgg broth in six-well microtiter dish. The dish was incubated for 72 hr at 30°C without stir.

| Construction of recombinant strains carrying single-or multiple-gene deletions
The marker-less gene knockout method was used to construct the gene deletion mutants, using the upp cassette and 5-fluorouracil (5-FU) selection . The primers BaeOUT-F/R, SrfOUT-F/R, ItuOUT-F/R, and FenOUT-F/R were used to confirm the construction of gene deletion mutants. As shown in Figure 2, the single, double, triple, and quadruple mutants of the bae, srf, itu, and fen genes were successfully generated, and the single-gene deletion mutants were designated as B. amyloliquefaciens LL3Δbae, LL3Δsrf, LL3Δitu, and LL3Δfen, respectively. The multiple-gene deletion mutants were named LL-IS (ΔsrfΔitu), LL-ISF (ΔsrfΔituΔfen), and LL-ISFB (ΔsrfΔituΔfenΔbae), respectively.

| Swarming ability and biofilm formation of the mutant strains
Swarming is a social motility behavior found in Bacillus strains and is associated with biofilm development. As previously reported, surfactin plays important roles in the swarming of B. subtilis strains (Kearns et al. 2004). In the "swim plates" (LB plate with 0.25% agar), all the mutants generated colonies that spread over the plate and showed efficient swimming motility (data not shown). However, as shown in Figure 3, B. amyloliquefaciens LL3Δsrf and LL3Δbae showed obvious defects in swarming motility in the "swarm plates" (LB plate with 0.5% agar), while B. amyloliquefaciens LL3Δfen and LL3Δitu showed slight defects in swarming. This agreed with the previous reports that the srf gene cluster is involved in swarming motility.
All the multiple-gene mutants, B. amyloliquefaciens LL-IS (ΔituΔsrf), LL-ISF (ΔituΔsrfΔfen), and LL-ISFB (ΔituΔsrfΔfenΔbae), had significant defects in swarming motility (Fig. 3). LL3Δitu had a slight defect in swarming; however, when srf was deleted to construct LL-IS (ΔituΔsrf), its swarming motility was significantly weakened, which further proved surfactin's effect on swarming. Rahman et al. (2007)reported that the biofilm formation of transformant B. subtilis RM/iSd16 containing wild sfp, itu operon, and degQ was better than the wild-type strain. Zeriouh, de Vicente, Perez-Garcia, and Romero (2014) found that surfactin triggered biofilm formation of B. subtilis. In this study, the biofilm formation ability of all the mutant strains was compared with that of the wild-type strain.
B. amyloliquefaciens LL3Δbae and LL3Δfen showed similar biofilm F I G U R E 2 Confirmation of the deletion of the genes by agarose gel electrophoresis of PCR products with primers BaeOUT-F/R (lanes 1 and 2), SrfOUT-F/R (lanes 3 and 4), ItuOUT-F/R (lanes 5 and 6), and FenOUT-F/R (lanes 7 and 8). Chromosomal DNA of the mutant strains served as the template for PCR. Fragments of the wild-type strain were too long to obtain PCR products F I G U R E 3 Swarming experiments of the wild-type strain, B. amyloliquefaciens LL3Δsrf, LL3Δfen, LL3Δbae, LL3Δitu, LL-IS (ΔituΔsrf), LL-ISF (ΔituΔsrfΔfen), and LL-ISFB (ΔituΔsrfΔfenΔbae). Strains were observed after 24-hr cultivation on LB medium with 0.5% agar formation compared with the wild-type strain (Fig. 4). B. amyloliquefaciens LL3Δsrf, LL3Δitu, LL-IS, LL-ISF, and LL-ISFB were significantly defective in biofilm formation (Fig. 4). This could be inferred that iturin A and surfactin play important roles in biofilm formation.

| DCW, γ-PGA synthesis and culture viscosity of the mutant strains
B. amyloliquefaciens LL3Δbae, LL3Δsrf, LL3Δitu, LL3Δfen, LL-IS (ΔituΔsrf), LL-ISF (ΔituΔsrfΔfen), and LL-ISFB (ΔituΔsrfΔfenΔbae) were compared with the wild-type strain for culture viscosity, DCW, and γ-PGA synthesis. At the end of the fermentation, surprisingly, the culture viscosity of the Δsrf, Δitu, and Δfen mutants was decreased by 46%, 20.5%, and 29%, respectively, while the Δbae mutant showed no apparent changes compared with the wild-type strain (Fig. 5A). The DCW of the Δfen mutant experienced a slight decrease, while that of the other three mutants resembled the wild-type strain. γ-PGA synthesis of the Δfen, Δitu, or Δbae mutants showed no obvious changes; however, in the Δsrf mutant, the synthesis of γ-PGA showed a slight increase (4.1 g/L) compared with that in the wild-type strain (3.3 g/L).
As shown in Figure

| DISCUSSION
γ-PGA-producing strains are generally divided into two groups according to their nutritional requirements: glutamic acid-dependent bacteria and glutamic acid-independent bacteria. The latter does not require additional l-glutamate in the medium to stimulate γ-PGA, so that their production costs are lower than the former.
B. amyloliquefaciens LL3 is a naturally isolated, Gram-positive strain that can produce γ-PGA without the addition of glutamic acid in the medium. It secretes various antibiotic substances to adapt to the environment, such as surfactin, iturin A, fengycin, and bacillaene. Except for bacillaene, all of them are lipopeptides. Many reports have shown that the biological control exerted by B. subtilis and related species could be attributed to nonribosomally produced cyclic lipopeptides (Ongena & Jacques, 2008;Romero, de Vicente, Rakotoaly, Dufour, Veeing, Arrebola, 2007;Zeriouh et al., 2011). Lipopeptides interact with the biological membranes of microbial pathogens, inducing cell leakage and death (Romero, de Vicente, Olmos, Davila, & Pérez-García, 2007;Zeriouh et al., 2011).  baeC, baeD, baeE, baeG, baeH, baeI, baeJ, baeL, baeM, baeR, and baeS. The total lengths of the four gene clusters are 28.3, 37.2, 11.5, and 72.5 kb, respectively (Fig. S1). They were all predicted as nonessential using a comparative genomics approach and comparing the Glutamine synthetase is an enzyme that catalyzes L-glutamate to glutamine and plays important roles in glutamate consumption. This is good evidence that surfactin shares same substrates with γ-PGA.
Besides, as mentioned above, surfactin upregulates the transcription of the flagellin gene (Ghelardi et al., 2012). Chan, Guttenplan, and Kearns (2014) found that defects in the flagellar motor increase synthesis of poly-γ-Glutamate in B. subtilis. Therefore, srf mutant may also enhance γ-PGA synthesis indirectly.
However, B. amyloliquefaciens LL3Δbae, LL3Δfen, and LL3Δitu did not show significant increases in γ-PGA yield. For further explanation, HPLC-MS was used to detect whether all the strains could produce the antibiotic substances or not. As shown in Figure 6, B. amyloliquefaciens LL3Δupp could synthesize surfactin, while LL3Δsrf cannot produce surfactin anymore. It may be the main reason for the increase of γ-PGA synthesis that surfactin competes for same substrates with γ-PGA. In addition, iturin A and fengycin were undiscovered in the culture of LL3Δupp (Fig. S3). This can explain why LL3Δitu and LL3Δfen showed no increases in γ-PGA synthesis.
It was further examined whether the disruption of the four gene clusters affected the expression of pgs operon. As shown in Figure S2, the pgsB expression levels of these mutant strains were LL-IS showed slight increase in γ-PGA titer compared with LL3Δsrf although we did not discover iturin A in the culture of the wild strain.
However, as mentioned above, LL3Δitu was significantly defective in biofilm formation. It is speculated that LL3Δupp might synthesize other iturin derivatives, which may also contributes to biofilm formation. Accumulation of multiple-gene cluster deletions in one strain, LL-ISF and LL-ISFB did not give rise to a continuous increase in γ-PGA yield. This may be attributed to that the wild strain might not produce fengycin or bacillaene and that the secondary metabolites might act not only as antibiotic substances but also as signal molecules affecting various cellular activities.