PRODUCTION OF NEW WAP-8294A CYCLODEPSIPEPTIDES BY THE BIOLOGICAL CONTROL AGENT LYSOBACTER ENZYMOGENES OH11

ABSTRACT Naturally-occurring environmental microorganisms may provide ‘green’ and effective biocontrol tools for disease management in agricultural crops. Due to the constant threat of resistant pathogens there is a pressing and continual need to search for new biocontrol tools. This study investigated the production of new analogs of WAP-8294A compounds by the biocontrol agent Lysobacter enzymogenes OH11 through biosynthetic engineering. WAP-8294As are a family of natural cyclic lipodepsipeptides with potent activity against Gram-positive bacteria. A series of genetic manipulations was therefore conducted on the accessory genes in the WAP biosynthetic gene cluster. The resulting strains containing a single-point mutation in ORF4, which was predicted to encode a 2-ketoglutarate dependent dioxygenase, produced deoxy-WAP-8294As. This result provides evidence for the function of ORF4 in catalyzing βhydroxylation of the D-asparagine residue in WAP-8294As


INTRODUCTION
The search for safe and effective pesticides for the protection of crops has become increasingly reliant on natural products as a source for novel chemistry. Environmental microorganisms that produce bioactive natural products are important tools in pest management. Several pesticidal products such as Avermectin and Spinosad derived from microorganisms have been commercialized. The BT endotoxins have become an integral component of strategies to develop transgenic plants that express insecticidal proteins and have been very successful in the control of insect pathogens. These products are often classified by the US Environmental Protection Agency as reduced-risk pesticides that pose little risk to human health and the environment. The identification and production of natural products with new chemistry and unique mode of action are critical steps in combating the constant emergence of resistant pathogens.
Lysobacter are gliding Gram-negative bacteria ubiquitously inhabiting soils and waters [1] . Several species of Lysobacter are prolific producers of extracellular lytic enzymes and are emerging as promising sources of new bioactive natural products [2] . Of these, Lysobacter enzymogenes is a microbial species of ecological and agricultural relevance [3,4] . L. enzymogenes strain C3 exhibited field efficacy against diseases of turfgrass, Bipolaris leaf spot caused by Bipolaris sorokiniana [5] , brown patch caused by Rhizoctonia solani [3] , stem rust caused by Puccinia graminis and bean rust caused by Uromyces appendiculatus [6] . It also reduced the severity of Kentucky bluegrass summer patch caused by Magnaporthiopsis poae, suppressed damping-off of sugar beet caused by Pythium ultimum and Fusarium head blight caused by Fusarium graminearum (Gibberella zeae) [7,8] .
The biosynthetic gene cluster (BGC) for WAP-8294As from L. enzymogenes OH11 has been identified [19] . The BGC contains three structural genes, ORFs 1-3 that encode the multimodular nonribosomal peptide synthetases (NRPS) WAPS1 and WAPS2 and the NRPS-associated protein MbtH (Fig. 1). Clustered with the three structural genes are seven accessory genes, ORFs 4-10, which were predicted to function in the modification, regulation and resistance of WAP-8294As [19] . Various approaches including synthetic biology methods have been used in the genetic manipulation of WAP-8294A BGC [27,28] . However, efforts have focused on increasing the yields of the known compounds. So far, the biosynthetic production of new WAP-8294A analogs in L. enzymogenes has not been attempted. Here, we conducted a series of genetic manipulations on the accessory genes and found deoxy analogs of WAP-8294As in single-point mutants. Also, using ultra-HPLC and high-resolution tandem mass spectrometry (UHPLC-HR-MS/MS), we detected six new WAP-8294A compounds in wild type OH11. The results show that the spectrum of biocontrol compounds of L. enzymogenes OH11 may be expanded through manipulation of the biosynthetic genes.

Bacterial strains, plasmids, primers and general DNA manipulations
Two Escherichia coli strains were used, XL1-Blue as the host for general plasmid DNA propagation and S17-1 as the conjugal strain to transfer DNA into the wild type L. enzymogenes OH11. Plasmid pJQ200SK was the vector for in-frame deletion of OH11 genes and plasmid pEX18 as the vector for mutagenesis of single amino acid residues in biosynthetic genes. Kits from Qiagen (Valencia, CA) were used for general plasmid preparation and DNA gel extraction and all other DNA manipulations were conducted using standard methods. Luria-Bertani (LB) broth medium was used for OH11 growth. Primers used are shown in Table S1.

Generation of gene deletion mutants and single-point mutants
To construct in-frame deletion vectors two homologous arms of 1.5 kb in length were amplified by PCR from the upstream and downstream regions of the target gene in the OH11 genome using the primers shown in Table S1. The PCR fragments were then cloned into the conjugal vector pJQ200SK by the corresponding restriction enzymes. The constructs were confirmed by DNA sequencing. To construct the vectors for single-point mutagenesis the WAP-ORF4 gene was amplified by PCR from the OH11 genome using the primer pair ORF4-epF(BamH)/ORF4-epR(Hind) and then cloned into vector pET28a, producing plasmid orf4-pET28a. Another run of PCR was conducted using plasmid orf4-pET28a as template and orf4 H104A-F/orf4 H104A-R or orf4 H298A-F/orf4 H298A-R as primers. The PCR fragments were digested with DpnI at 37 °C overnight to specifically digest the methylated and hemimethylated DNA (the parental DNA template). The DpnIdigested products were transformed into XL1-Blue competent cells. The plasmids were extracted from the cells and singlepoint mutation at H104A or H298A was confirmed by introducing restriction enzyme site PaeI or Eco52I. The fragment between BamHI and HindIII from the single-point mutated plasmid pET28a-H104A or pET28a-H298A was released and cloned into vector pEX18.
The constructs for in-frame deletion and single-point mutagenesis were transformed into E. coli S17-1 which was grown in LB medium until OD 600 reached 0.6. L. enzymogenes OH11 was cultured in LB containing 10 mmol·L −1 MgCl 2 until OD 600 reached 0.6. After centrifugation the pellet from 1 ml OH11 and the pellet from 1 ml S17-1 were separately washed with 10 mmol·L −1 MgSO 4 and then mixed for intergeneric conjugation. The cells were spread on LB plates containing kanamycin (Km, 100 μg·mL −1 ) and gentamicin (Gm, 150 μg·mL −1 ). After growth at 30 °C for 72 h the colonies were transferred to LB plates containing 10% (w/v) sucrose and Km (100 μg·mL −1 ) at 30 °C for 72 h. The transformants were transferred to LB plates supplemented with Km (100 μg·mL −1 ) or Km (100 μg·mL −1 ) + Gm (150 μg·mL −1 ). The Km-resistant and Gm-sensitive strains were selected as putative in-frame deletion mutants and single-point mutants that were confirmed by diagnostic PCR.

Extraction of the WAP-8294A compounds from Lysobacter enzymogenes strains
Three liquid media (L1M with 2.5% glucose, 2% soybean flour, 0.4% soybean oil, 0.25% NaCl, and 0.5% CaCO 3 at pH 7.2; L2M with 0.5% bacteriological peptone, 0.3% yeast extract and 2% glycerol; and L3M with 2% glucose, 0.5% beef extract, 1.6% soybean oil, 0.1% NaCl and 0.1% CaCO 3 at pH 8.5) and three solid media (S1M as L1M with 1.5% agar; S2M as L2M with 1.5% agar; and S3M as L3M with 1.5% agar) were used for screening WAP8294A production. After growth in 3 mL LB containing 100 μg·mL −1 kanamycin for one day, a 500-μL aliquot of OH11 seed culture was transferred to a 250-mL flask containing 50 mL of liquid medium, L1M, L2M or L3M. The culture was incubated at 28 °C for 2 days with shaking at 200 r·min −1 . To extract the metabolites the 50 mL culture broth was collected and adjusted to pH 2.5 with 1 mol·L −1 HCl. The supernatant was extracted with 50 mL n-butanol-ethyl acetate (1:1 v/v), containing 0.05% trifluoroacetic acid (TFA). The organic phase was dried with a rotary evaporator (R-200; Buchi, Flawil, Switzerland) to obtain the crude extract. The extract was dissolved in 1.5 mL methanol containing 0.05% TFA. A 25-μL aliquot of each extract was analyzed by HPLC. Solid cultures were prepared by spreading an aliquot of 150 μL OH11 seed culture on a solid plate and incubating at 28 °C for 2 days. To extract the metabolites the solid culture from three plates was collected and extracted with 50 mL methanol. The methanol extract was dried and redissolved in 1.5 mL methanol containing 0.1% TFA. A 25-μL aliquot of each extract was analyzed by HPLC, or 5 μL by high-resolution UHPLC-ESI-MS/MS. After comparison of the metabolite profile from the liquid cultures and the solid cultures, S2M medium was chosen as the medium for all mutant strains because this medium consistently produced WAP-8294A compounds.

HPLC and UHPLC-MS/MS analysis of the WAP-8294A extracts
A 25-μL aliquot of each extract was analyzed by HPLC (1220 Infinity LC, Agilent Technologies, Santa Clara, CA) using a reverse-phase column (Cosmosil 5C18-AR-II, 4.6 ID × 250 mm). Water-0.1% TFA (solvent A) and acetonitrile-0.1% TFA (solvent B) were used as the mobile phases with a flow rate of 1.0 mL·min −1 . The HPLC program was as follows: 5% to 25% solvent B in solvent A for the first 0 to 10 min, with an increase to 80% B at 25 min, to 100% at 26 min, and back to 5% B at 30 min. The metabolites were detected at 220 nm on a UV detector. The extracts were analyzed by UHPLC-HR-MS by the Proteomics and Metabolomics Facility at the Nebraska Center for Biotechnology at the University of Nebraska-Lincoln, USA. An aliquot of each extract was injected into the column (ACCQ-TAG ULTRA C18 1.7 μm; 2.1 mm × 100 mm, Waters, Milford, MA) on a Vanquish (Thermo Fisher Scientific, Waltham, MA) HPLC at 40 °C at a flow rate of 300 μL·min −1 . The solvents were A (0.1% formic acid in 100% LC-MS grade water) and B (0.1% formic acid in 100% acetonitrile) with a gradient as follows: 10% B for 2 min, 10% to 90% B in 12 min, hold at 90% B for 2 min, then back to 10% in 0.5 min. The data were acquired on a QE-HF (Thermo Fisher Scientific) mass spectrometer using a mass range of 375 to 1200 m/z on single charge ions at 60,000 resolution using an AGC target of 3e6 and a maximum ion time of 50 ms. The isolated ions were further fragmented by HCD in data dependent mode (loop count 10, intensity threshold 1e5, dynamic exclusion 10 s) using an isolation window of 1.6 m/z and scanned using a mass range of 120-2000 m/z at a resolution of 15,000.

Feature-based molecular networking of WAP-8294A extracts
Feature-based molecular networking was conducted using the GNPS platform [29,30] . The MS/MS raw data were converted to the mzML format using ProteoWizard [31] before importing into MZmine2.5.3 software [32] for LC-MS feature detection and alignment. The noise levels of MS1 and MS2 were set to 2E5 and 1E2 according to the low abundance of deoxy-WAPs and the unknown WAPs compounds. The mascot generic format file was uploaded to GNPS and then used to generate a featurebased molecular network. The precursor ion mass tolerance and product ion tolerance were set to 0.01 Da and 0.02 Da. A minimum of eight peaks and a cosine score of 0.6 were allowed. Data were opened and visualized using Cytoscape 3.8.0 software [33] . A ClassDefault layout was used for data visualizations. The nodes, the sizes of which indicate the parent ion abundance, are presented as pie charts with different colors representing the presence of each metabolite in the wild type (OH11) and the ORF4 mutants.

In-frame deletion of the accessory genes in the WAP gene cluster in Lysobacter enzymogenes OH11 and analysis of WAP-8294A production in the mutants
To test the idea that new WAP-8294A compounds may be generated through manipulation of the accessory genes clustered with the structural genes (ORFs 1-3 for MbtH and NRPS) we conducted in-frame deletion of the genes in the WAP gene cluster (Fig. 1). First, we deleted each of the four downstream genes, ORFs 6, 7, 9 and 10. ORF8 was omitted because it codes for a TonB-dependent receptor that is unlikely to be related to structural modifications of the compounds [19] . The in-frame deletion mutants, OH11∆WAP-ORFs 6, 7, 9 and 10, were confirmed by diagnostic PCR using primers designed to distinguish the wild type and the mutants (Fig. S1). The wild type and the mutants were cultured in various media to test that the yield of metabolites with S2M was most suitable for WAP-8294A production under the test conditions. The cultures grew on S2M plates for 2 days and the metabolites were extracted from the whole plates. HPLC shows that the wild type produced WAP8294A1, A2, and A4, with A2 being the predominant WAP8294A compound (Fig. 2(a) Fig. 2(b,c)). Each of the gene deletion mutants, OH11∆WAP-ORF6, 7, 9, and 10, produced a similar WAP8294A profile to the wild type. The results indicate that none of the four accessory genes is essential for the WAP-8294A production in OH11. To further confirm this we generated a quintet mutant by in-frame deletion of all five downstream genes, ORF6 through ORF10 (including ORF8).
shows that this quintet mutant still produced the WAP-8294A compounds ( Fig. 2(a)). The data verify that these five downstream genes are not essential for WAP-8294A biosynthesis.
Next, we examined the remaining two accessory genes in the WAP gene cluster, ORF4 and ORF5, which appear to form an operon with the structural genes (ORFs 1-3) because the five open reading frames (ORFs 1-5) are tightly coupled in the same transcriptional direction (Fig. 1) [19] . To avoid potential polar effects we generated the mutants by in-frame deleting ORF4 or ORF5 individually. The identities of the mutants, OH11∆WAP-ORF4 and OH11∆WAP-ORF5, were confirmed by diagnostic PCR (Fig. S1). HPLC shows that the WAP-8294A compounds were not produced in OH11∆WAP-ORF4 or OH11∆WAP-ORF5 (Fig. 3). The results show that both ORF4 and ORF5 are essential for the biosynthesis of WAP-8294A.

Deoxy-WAP-8294A analogs produced in mutants derived from single-point mutagenesis of ORF4 in the WAP gene cluster
Given that ORF4 was predicted to encode a 2-ketoglutarate dependent dioxygenase that catalyzes either the βhydroxylation of the D-Asn residue or the β-hydroxylation of the fatty acid moiety in WAP-8294A compounds [19] , we postulated that the genetic manipulation of this gene may lead to new WAP-8294A analogs. However, even an in-frame deletion of ORF4 did not result in HPLC detectable WAP-8294A analogs. To further test the function of ORF4 we generated two single-point mutants at the active site residues, His104A or His298A, which maintain ORF4 intact in the mutants with only a single amino acid change. ORF4 is most similar to E. coli taurine dioxygenase TauD and related enzymes which contain a conserved HX(D/E)-X 23−26 (T/S)X 114−183 HX 10−13 R motif (the two highly conserved histidine residues are underlined) [34] . His104 and His298 of ORF4 correspond to the two highly conserved histidine residues in this motif. The single-point mutants, OH11WAP-ORF4-H104A and OH11WAP-ORF4-H298A, were confirmed by diagnostic PCR and digestion of a special restriction enzyme site which was introduced at the point mutation site (Fig. S2).
Using HPLC with a UV detector no apparent WAP-8294A compound was detected from either OH11WAP-ORF4-H104A or OH11WAP-ORF4-H298A (Fig. 3). To further examine the metabolites from the single-point mutants we conducted highresolution LC-MS analysis with a comparison to the metabolites from the wild type. The total ion chromatogram (TIC) of the wild type gave the expected WAP-8294A1, A2 and A4. In addition, the TIC of the wild type showed complexes of peaks (such as Ax) mixed with or flanking the three known compounds (Fig. 4(a)). In contrast, the TIC of the single-point mutant OH11-WAP-ORF4-H104A did not appear to give the main peaks or the peak complexes (such as the Ax complex in Fig. 4(a)). Instead, OH11-WAP-ORF4-H104A produced several minor peaks at the WAP-8294A region on the LC-MS. We examined the three peaks (8.7, 9.1 and 9.4 min) in the mutant because these retention times fall into the region of WAP-8294A compounds. HR-MS shows that the peak at 8.  (Fig. 4(b)). Moreover, the extracted ion chromatogram (EIC) of deoxy-WAP-8294A1 mass gave a peak at 8.7 min in mutant OH11-WAP-ORF4-H104A (Fig. 4(c)). Notably, the wild type EIC also contained two peaks with a similar extracted mass to deoxy-WAP-8294A1. As shown below these are new analogs of WAP-8294As (AZ1 and AZ2, see Table 1, Table S2) which have a similar mass (m/z 766.9090) to deoxy-WAP-8294A1. The EIC of deoxy-WAP-8294A2 mass showed a clear peak at 9.1 min in mutant OH11-WAP-ORF4-H104A but the wild type gave only noise signals in this region (Fig. 4(d)). Similarly, the EIC of deoxy-WAP-8294A4 mass showed a clear peak at 9.4 min in mutant OH11-WAP-ORF4-H104A but the wild type gave only noise signals in this region (Fig. 4(e)). Notably, a small peak at 9.0 min was also shown in the EIC of deoxy-WAP-8294A4 mass which might be an unidentified deoxy-WAP isomer with m/z in the range 780.9217-780.9237. Taken together, the data demonstrate that the ORF4 single-point mutant was able to produce new WAP-8294A analogs which lack a hydroxyl group of the parent WAP-8294A compounds.

Deoxy-WAP-8294A analogs missing the βhydroxyl group on the D-asparagine residue of the parent WAP-8294A compounds
As mentioned above, the 2-ketoglutarate dependent dioxygenase encoded by ORF4 may be responsible for the βhydroxylation of the D-Asn residue (Fig. 4(f-h)) or the βhydroxylation of the fatty acid chain in WAP-8294A compounds [19] . The deoxy-WAP-8294A compounds produced in OH11-WAP-ORF4-H104A are most likely due to the lack of the β-hydroxyl group on the D-Asn residue rather than a lack of the β-hydroxyl group on the fatty acid moiety. This is because the β-hydroxyl group on the fatty acid is involved in cyclization of the cyclic lipodepsipeptides, and the lack of this Jing ZHU et al. New WAP-8294As produced by Lysobacter β-hydroxyl group would lead to a drastic structural change in the WAP-8294A analogs which would be unlikely to show similar retention times on LC-MS to the known WAP-8294As. To obtain experimental data we analyzed the deoxy-WAP-8294A compounds using UHPLC-HR-MS/MS in comparison with the fragmentation pattern of the parent WAP-8294A compounds (Figs. S3-S4).
The HR-MS/MS fragments of the parent compounds, WAP-8294A1, A2 and A4, from the wild type are summarized in Table S2 which cover the 12 amino acid residues and D-3-OHfatty acid moiety. The HR-MS/MS profile of deoxy-WAP-8294A1 shared a large number of fragment ions with WAP-8294A1 except at the position of D-threo-β-OH-Asn. The deoxy compound showed a reduced mass of 16 at the Asn position, supporting the position of the deoxy at D-threo-β-OH-Asn rather than at the D-3-OH-fatty acid moiety (Fig. S4). Similarly, the comparison of the MS/MS fragments from deoxy-WAP-8294A2 with that from WAP-8294A2, and deoxy-WAP-8294A4 with that from WAP-8294A4, further confirm that the β-hydroxyl of the D-Asn residue was missing while the β-hydroxyl of the fatty acid was retained. In addition, we conducted molecular networking analysis of the HR-MS/MS fragments of the WAP metabolites. The results show the presence of deoxy-A1 (m/z 766.9064), deoxy-A2 (m/z 773.9146), and deoxy-A4 (m/z 780.9223) (Fig. 5). In addition, an unknown deoxy-Ax (m/z 794.9248) was also detected by the molecular networking analysis. In the networking the ions of deoxy-A1, deoxy-A2 and deoxy-A4 clearly shared MS/MS fragments with the respective parent ions from the wild type, A1 (m/z 774.9044), A2 (m/z 781.9119), and A4 (m/z 788.9197). The new deoxy-Ax (m/z 794.9248) was also connected to a small group of ions that appeared to be new analogs produced in the wild type. Taken together the data obtained from HR-MS/MS confirm the identity of the deoxy-WAP-8294A compounds as shown in Fig. 4(f-h). In addition, the results provide evidence for the function of ORF4 in β-hydroxylation of the D-Asn residue in WAP-8294A biosynthesis.

New WAP-8294A compounds identified by UHPLC-HR-MS/MS analyses
Encouraged by the identification of the deoxy-WAP8294A analogs in the ORF4 mutant we set out to examine the minor peaks in the wild type by UHPLC-HR-MS/MS. As shown in Fig. 4(a) the TIC revealed multiple peak complexes present in the metabolites from the wild type. These minor peaks were mixed with or flanking the main peaks (A1, A2, and A4). We detected at least six new minor compounds, WAP-8294AZ1-AZ6, with m/z in the range 766.9078 to 797.9075 for [M + 2H] 2+ (Table 1 and Table S2). UHPLC-HR-MS/MS analyses of each of the six compounds indicate that these belong to the WAP-8294A family, with variations in the amino acid residues (AZ1, AZ2, AZ3 and AZ6) or in D-3-OH fatty acid (AZ4) (Fig. 6(a), Table S2). New compounds were also identified in the molecular networking analysis (Fig. 5). Interestingly, AZ5 appears to be the same as the previously reported minor compound WAP-8294Ax8 [25] .
In addition to the six compounds we detected an unusual compound, WAP-8294AZ7, which gave m/z 538.7794 for [M + 2H] 2+ . This mass is significantly smaller than those of the other family members (Table 1). UHPLC-HR-MS/MS analysis of this compound reveals that AZ7 contains only nine amino acid residues, with the same sequence and composition as the fourth residue (L-Ser) through to the twelfth residue (D-Orn) of WAP-8294A2 (Fig. 6(b)). In the structure of AZ7 the D-3-OH-fatty acid moiety, the first two amino acid residues, L-Ser and D-threo-β-OH-Asn, and the last amino acid residue, L-N-MeVal, are totally missing. Overall, this study has detected six previously unknown WAP-8294A analogs through UHPLC-HR-MS/MS analysis of the metabolites from the wild type L. enzymogenes OH11. (c) Extracted ion chromatogram (EIC) of deoxy-WAP-8294A1 at 8.7 min. Note that the wild type contains two analogs (AZ1 and AZ2) with the similar mass (m/z 766.9090, see Table 1

DISCUSSION
We previously identified the biosynthetic gene cluster for WAP-8294As from L. enzymogenes OH11 and proposed a biosynthetic pathway for WAP-8294As, which are assembled on the giant NRPS enzyme complex encoded by ORFs 1-3 [19] . We investigated the mechanisms by which the D-3-OH-fatty acids are incorporated into the amino acid scaffold [35] . We also attempted to increase the yield of WAP-8294As using promoter replacement, CRISPR/dCas9 and gene refactoring [28] . However, new WAP-8294A analogs have not been produced using biosynthetic approaches. Here, we first conducted a series of in-frame deletion of the accessory genes downstream of the three structure genes, ORFs 1-3. This strategy would minimize potential polar effects on the flanking genes. Several gene disruption mutants had previously been generated through inserting a plasmid construct containing the homologous region of ORFs 5, 6 or 8 [19] . Our results from inframe gene deletion of a single gene, ORFs 6, 7, 9 and 10, or all five genes, ORFs 6-10, show that these accessory genes are not essential for biosynthesis. In contrast, in-frame deletion of ORF5 or ORF4 resulted in no WAP-8294A production (Fig. 3), suggesting that the efflux pump encoded by ORF5 and the dioxygenase encoded by ORF4 are essential. These findings are consistent with our recent study in which an overexpression of the biosynthetic genes, including ORF5 and ORF4, by CRISPR/dCas9 and gene refactoring, may increase the yields of WAP-8294A1, A2 and A4 four to nine times [28] . The previous ORF5 disruption mutant, which was generated by a single crossover of the plasmid construct, may undergo a second crossover, leading to a revertant (wild type) that would produce WAP-8294As [19] . On the other hand, the previous ORF6 disruption mutant led to loss of the WAP-8294A compounds, suggesting that the insertion of the whole plasmid construct into ORF6 could also have disrupted the potential promoter of the essential genes (ORFs 1-5, Fig. 1). This is likely because ORF6 lies directly upstream of ORFs 1-5 that are organized as an operon and transcriptionally coupled.
ORF4 is homologous to taurine catabolism dioxygenase, a nonheme iron 2-ketoglutarate-dependent oxygenase. We have tried to demonstrate its function through heterologous gene expression in E. coli and in vitro enzyme assays. We were able to obtain purified ORF4 protein. However, in vitro enzyme  Table 1). Note that a new deoxy-WAP-8294Ax (m/z 794.9248, in magenta) was also identified in the networking analysis.
assays indicate that the enzyme would not use free asparagine (either D-or L-) as substrate, as seen in β-hydroxylation in calcium-dependent antibiotics [36] . This result indicates that the β-hydroxylation of the asparagine residue may take place after the amino acid has been covalently loaded to the NPRS assembly line, as seen in the β-hydroxylation of histidine in the biosynthesis of nikkomycin antibiotics [37] . We subsequently expressed the A (adenylation) domain and PCP (peptidyl carrier protein) domain of WAPS1 (see Fig. 1) that are responsible for the activation and incorporation of asparagine during NRPS-catalyzed peptide chain biosynthesis [19] . However, the heterologously expressed NRPS A-PCP proteins were insoluble, despite multiple attempts with various expression systems. As an alternative, we took the in vivo approach to tackle the function of ORF4 and meantime to generate new Lysobacter strains that could remain biosynthetically viable. We generated single-point mutants at the active sites of ORF4, which had only one amino acid change and contained no foreign DNA. LC-MS analysis indicates that WAP-8294A-like compounds were produced by  Table S2. Note that compound AZ5 is the same as Ax8 [25] , and the structural variations in the new compounds versus WAP-8294A2 are highlighted. In the MS/MS of AZ7, the numbers inside the parentheses indicate the position of the building blocks, the amino acid residues #4-12 of WAP-8294A2.
Jing ZHU et al. New WAP-8294As produced by Lysobacter the new Lysobacter strains (Fig. 4). The high-resolution MS of the minor compounds confirms the deoxy analogs of parent WAP-8294A compounds. Further analysis using UHPLC-HR-MS/MS shows that the deoxy analogs lack the β-hydroxy group of the D-Asn residue, rather than on the β-hydroxy group of the fatty acid chain (Fig. 5, Figs. S3-S4, Table S2). After determining that the point mutants produced the deoxy analogs we returned to the in-frame deletion mutant and also detected the deoxy analogs using high-resolution technology (Fig. 5). The results suggest that, while ORF4 is required for the β-hydroxylation, inactivation of this gene led to extremely poor yields, indicating that a functional ORF4 is important to the overall operon to function properly. Nevertheless, our finding indicates that deoxy-WAP-8294A compounds may be produced by biosynthetic gene manipulations in Lysobacter, albeit in a rather low yield currently. Previously, WAP-8294A2 and its deoxy analog deoxy-WAP-8294A2 have been chemically synthesized [18] . The work shows that deoxy-WAP-8294A2 is also a potent antibacterial compound. However, the chemical approach is unlikely to be practical because it required 27 steps of highly elaborated chemical reactions. The biosynthetic approach may provide a viable alternative if the yield can be increased in the future.
Although previous studies indicate that the WAP-8294A family may contain > 19 closely related compounds, only WAP-8294A1, A2, A4, Ax8, Ax9, and Ax13 had been identified [24][25][26] . Here, we used UHPLC-HR-MS/MS to analyze the metabolites from L. enzymogenes OH11 and identified six new compounds, WAP-8294AZ1, AZ2, AZ3, AZ4, AZ6, and AZ7, in addition to four known compounds, A1, A2, A4, and Ax8 (AZ5) (Fig. 6, Table 1). The new compounds contain either a change in amino acid residues (AZ1, AZ2, AZ3, and AZ6) or a varied D-3-OH fatty acid (AZ4) (Fig. 1). Most notably, AZ7 contains only nine amino acid residues (#4-12) of A2 (Fig. 6). The composition and sequence of the nine amino acid residues of AZ7 are identical to that of the fourth through twelfth residues of WAP-8294A2. This truncated structure suggests that AZ7 may be a degradation product of the parent compounds, as seen in the recently reported lysocin-7, which is likely derived from a degradation of the parent lysocins, a family of cyclic lipodepsipeptides produced by Lysobacter sp. strain 3655 [38] . Alternatively, AZ7 might be an aberrant biosynthetic product of the WAP biosynthetic enzymes, due to an unusual start (from #4 L-Ser) and premature release of the peptide from the NRPS assembly line, through a cyclization between #4 L-Ser and #12 D-Orn. Natural products resulting from an unusual start and release of a peptide precursor from the NRPS assembly line have been observed in other Lysobacter NPRS [39] . Further investigations are needed to decipher the mechanism of AZ7 formation.

CONCLUSIONS
With the constant emergence of resistant pathogens and the recognized risks of toxic pesticides to human health and the environment, the search for safe and effective alternative strategies for pest control has become increasingly important. Biological control measures can provide environmentallysustainable and effective management tools for crop diseases.
One key factor in the success of this biological approach is the discovery of new microbial strains that can produce potent natural products with novel chemistry and mode of action. Lysobacter represents a group of underexplored biocontrol agents with demonstrated efficacy and proficiency in producing bioactive natural products. The present study shows that Lysobacter strains can be generated to produce new antibacterial WAP-8294A compounds through minimum manipulation of the biosynthetic genes, without introducing foreign DNA. The work lays the foundation for future efforts to engineer the WAP biosynthetic pathway for yield increase of the new compounds and to develop Lysobacter strains as a safe and effective tool for management of crop diseases.