A predatory soil bacterium reprograms a quorum sensing signal system to regulate antifungal weapon production in a cyclic-di-GMP-independent manner

Soil bacteria often provide multiple weapons for eukaryotes or prokaryotes to use against predators. Diffusible signal factors (DSFs) represent a unique group of quorum sensing (QS) chemicals that modulate interspecies competition in bacteria that do not produce antibiotic-like molecules. However, the molecular mechanism by which DSF-mediated QS systems regulate weapons production for interspecies competition remains largely unknown in soil biocontrol bacteria. In this study, we found that the necessary QS system component protein RpfG from Lysobacter, in addition to being a cyclic dimeric GMP (c-di-GMP) phosphodiesterase (PDE), regulates the biosynthesis of an antifungal weapon (heat-stable antifungal factor, HSAF), which does not appear to depend on the enzymatic activity. Interestingly, we showed for the rst time that RpfG interacts with three hybrid two-component system (HyTCS) proteins, HtsH1, HtsH2, and HtsH3, to regulate HSAF production in Lysobacter. In vitro studies showed that each of these proteins interacted with RpfG, which reduced the PDE activity of RpfG. Finally, we showed that the cytoplasmic proportions of these proteins depended on their phosphorylation activity and binding to the promoter controlling the genes implicated in HSAF synthesis. These ndings reveal a new mechanism of DSF signalling in antifungal weapon production in soil bacteria.

activities of RpfG, mutations at the conserved His-190, Asp-191, Gly-253, Tyr-254, and Pro-255 of the HD-GYP signature motif with alanine (Ala) were examined by site-directed mutagenesis. We tested HSAF production in the ΔrpfG mutant strain carrying plasmids encoding these mutant proteins. The strains expressing the RpfG H190A, D191A, G253A, Y254A, and P255A mutant proteins showed increased HSAF production compared with the ΔrpfG mutant strain. These results were superior to those of the complementary strain ΔrpfG/rpfG. Importantly, the ΔrpfG mutant strain and complemented strains (ΔrpfG/rpfG, ΔrpfG/rpfG H190A, ΔrpfG/rpfG D191A, ΔrpfG/rpfG G253A, ΔrpfG/rpfG Y254A, and ΔrpfG/rpfG P255A) did not impair bacterial growth ( Figure 2B, C). As described above,  Gly-253, Tyr-254, and Pro-255 were found to be critical for the PDE activity of RpfG (Figure 1), implying that HSAF is regulated in a PDE independent manner. To test this prediction, we compared intracellular cdi-GMP concentrations in the ΔrpfG mutant and the wild type and in the HSAF-production medium (10% TSB). We found that the concentration in the ΔrpfG mutant did not signi cantly change c-di-GMP production compared with the wild-type strain ( Figure 2D). These ndings indicated that the regulatory activity of RpfG does not depend on its PDE enzymatic activity against c-di-GMPs.

RpfG binds directly to the HyTCS protein HtsHs
The above ndings con rmed that RpfG does not regulate HSAF biosynthesis through the c-di-GMP signalling pathway, indicating that RpfG might regulate HSAF biosynthesis through interactions with other proteins in L. enzymogenes. To further explore the mechanisms underlying the contribution of RpfG to HSAF production, we used a bioinformatic tool (STRING) to identify potential interactors for RpfG; these represent interactions that possibly lead to alterations in HSAF synthesis. We discovered that RpfG interacts with three HyTCS proteins through bioinformatics predictions ( Figure S1A). We designated the HyTCS protein HtsH (hybrid two-component signalling system regulating HSAF production) based on the ndings of this study. To verify the operon structure of htsHs for in-depth genetic analyses, a series of RT-PCR primers (Table S2 and Figure S1B) were designed to determine whether there are intergenic transcripts crossing the adjacent genes. As shown in Figure S1C, Le3071 (htsH1), Le3072 (htsH2), and Le3073 (htsH3) likely constitute a single transcription unit because the corresponding intergenic transcripts were successfully ampli ed. Le3071 (htsH1), Le3072 (htsH2), and Le3073 (htsH3) encode a group of typical HyTCS proteins with pfam Reg_prop, pfam Y-Y-Y, HisKA, HATPase_c, and Response_reg domains. All three HyTCS proteins contain one predicted transmembrane region ( Figure S1D). We examined the alignments of three HyTCS proteins (HtsH1, HtsH2, and HtsH3), and the results showed that the HtsH1 protein shares 50% and 53% identity with HtsH2 and HtsH3, respectively. We also aligned HtsH2 with HtsH3, and the identity values were 50% ( Figure S2).
To examine whether RpfG could directly bind to the HyTCS proteins (HtsH1, HtsH2 and HtsH3), we used a pull-down assay using E. coli-expressed proteins in vitro. We puri ed recombinant RpfG-MBP, and the cytoplasmic fragments of HtsH1, HtsH2, and HtsH3 (HtsH1C-Flag-His, HtsH2C-HA-His, and HtsH3C-Myc-His, respectively) from E. coli ( Figure 1A and Figure S3). First, we tested the ability of RpfG-MBP to pull down HtsH1C-Flag-His, HtsH2C-HA-His, and HtsH3C-Myc-His. Finally, we con rmed the interaction between RpfG and HtsH1, HtsH2 or HtsH3 using pull down assays ( Figure 3A). Conversely, we examined the ability of HtsH1C-Flag-His, HtsH2C-HA-His, and HtsH3C-Myc-His to pull down RpfG-MBP and observed a positive signal ( Figure 3B-D). Second, we used surface plasmon resonance (SPR) to measure the possible binding events between the RpfG and HtsH1, HtsH2, or HtsH3 proteins. The RpfG-MBP sensor physically bound HtsH1C-Flag-His with a binding constant (K D ) of 0.06675 μM ( Figure 3E), suggesting an intermediate level of protein-protein interaction. We also con rmed direct binding between the RpfG-MBP and HtsH2C-HA-His or HtsH3C-Myc-His proteins by SPR (K D = 0.2998 μM or K D = 0.1678 μM, respectively) ( Figure 3F-G). Additionally, the SPR assay revealed that HtsH1C-Flag-His bound to HtsH2C-HA-His or HtsH3C-Myc-His with reasonably high a nity (K D = 0.09619 μM or K D = 0.1597 μM, respectively) and revealed that HtsH2C-HA-His bound HtsH3C-Myc-His with a K D value of 0.1782 μM ( Figure S4). Taken together, these experiments demonstrate that RpfG directly interacts with HtsH1, HtsH2 or HtsH3 proteins in vitro.

HtsHs inhibits the PDE enzymatic activity of RpfG
We wondered whether RpfG and HtsH1, HtsH2, or HtsH3 interactions affect the PDE activity of RpfG. To test this hypothesis, we used a biochemical assay in which c-di-GMP hydrolysis by RpfG-MBP was assayed in the absence or presence of HtsH1, HtsH2, or HtsH3, ranging from 0 µM to 32 µM. The results of the assay showed that the PDE activity of RpfG-MBP was lower in the presence than in the absence of HtsH1C-Flag-His, HtsH2C-HA-His, or HtsH3C-Myc-His ( Figure 4). Therefore, the results of the assays suggested that HtsH1, HtsH2, and HtsH3 inhibited the PDE enzymatic activity of RpfG. This result further con rms that the ability of RpfG to regulate HSAF production does not depend on its PDE enzymatic activity against c-di-GMP in L. enzymogenes.
To test this hypothesis, we performed an E. coli-based one-hybrid assay. As shown in Figure 6A, HtsH1, HtsH2, and HtsH3 directly bound to the promoter of the rst gene lafB in the HSAF biosynthesis operon (plafB).
Next, we tested the ability of HtsH1, HtsH2, and HtsH3 to bind to the lafB promoter, using an electrophoretic mobility shift assay (EMSA). A PCR-ampli ed 590 bp DNA fragment from the plafB was used as a probe. The addition of puri ed HtsH1C-Flag-His, HtsH2C-HA-His, and HtsH3C-Myc-His protein, ranging from 0 µM to 10 µM, to the reaction mixtures (20 μL and at 28°C, 25 min) caused a shift in the mobility of the plafB DNA fragment, and the EMSA revealed strong HtsH1, HtsH2, and HtsH3 binding with the plafB probe in a dose-dependent manner ( Figure 6B-D). We quanti ed the binding a nity of HtsH1, HtsH2, and HtsH3 to the HSAF operon promoter. In an SPR analysis, HtsH1C-Flag-His directly bound to the promoter of the HSAF biosynthesis gene (plafB) with high a nity (K D = 0.5356 μM) ( Figure 6E). In addition, HtsH2C-HA-His and HtsH3C-Myc-His bound to plafB with K D values of 1.379 μM and 0.2491 μM, respectively ( Figure 6F-G). The results demonstrated that HtsH1, HtsH2, and HtsH3 could directly target the promoters of the HSAF biosynthesis gene.
Based on the above results, we compared the transcriptome pro les of the wild-type strain and the htsHs mutants (ΔhtsH1, ΔhtsH2, ΔhtsH3, ΔhtsH12, ΔhtsH13, ΔhtsH23, and ΔhtsH123) by RNA-Seq and observed changes in the expression levels of several hundred genes (Table S3). We then performed trend analysis of the differential gene expression and found that the amounts of HSAF biosynthesis gene cluster mRNA were constitutively decreased in the htsHs mutants (ΔhtsH1, ΔhtsH2, ΔhtsH3, ΔhtsH12, ΔhtsH13, ΔhtsH23 and ΔhtsH123) ( Figure S5). Using quantitative RT-PCR (qRT-PCR), We measured the mRNA abundance of lafB in the htsHs mutants (ΔhtsH1, ΔhtsH2, ΔhtsH3, ΔhtsH12, ΔhtsH13, ΔhtsH23 and ΔhtsH123) and found that it was reduced compared to that in the wild type ( Figure S6).
Taken together, these results suggested that HtsH1, HtsH2, and HtsH3 can directly target the promoters of the HSAF biosynthesis genes to increase their expression and HSAF production by L. enzymogenes.
Phosphorylated HtsH1, HtsH2, and HtsH3 positively regulate HSAF biosynthesis Since HtsH1, HtsH2 and HtsH3 function as a group of HyTCS proteins, we explored whether they could directly target plafB depending on its phosphorylation activity in L. enzymogenes. To achieve this goal, we assessed the phosphorylation levels of these proteins with calf intestine alkaline phosphatase (CIAP) in the reaction mixtures (20 μL, 28°C, 60 min). Using Mn 2+ -Phos-tag SDS-PAGE, we showed that the phosphorylation levels of HtsH1C-Flag-His, HtsH2C-HA-His, and HtsH3C-Myc-His decreased upon CIAP treatment ( Figure 7A-C). To test the effect of HtsH1-, HtsH2-, or HtsH3-mediated phosphorylation on the function of the target plafB, EMSAs were performed with the same experimental conditions. The EMSAs revealed that the amount of probes bound to HtsH1C-Flag-His, HtsH2C-HA-His, and HtsH3C-Myc-His decreased with increasing amounts of CIAP ( Figure 7D-F). The above ndings indicated that HtsH1, HtsH2 and HtsH3 could directly regulate HSAF biosynthesis gene expression depending on their phosphorylation activity in L. enzymogenes.
RpfG-and HtsH1-, HtsH2-, and HtsH3-dependent regulatory patterns are present in a wide range of bacterial species By BLAST analysis of the nonredundant protein sequence (Nr) database of the National Center for Biotechnology Information (NCBI), we found that RpfG, the rpf cluster, and HtsH1, HtsH2, and HtsH3 were present not only in the genomes of Lysobacter species, but also Xanthomonas species ( Figure S7). These ndings suggest that RpfG-, HtsH1-, HtsH2-, and HtsH3-dependent regulatory patterns are conserved mechanisms in Lysobacter and Xanthomonas.

Discussion
In previous studies, we and our collaborators have shown that quorum sensing (QS) was employed by L.
enzymogenes to affect production of the antifungal weapon HSAF 6,25 . However, the mechanism through which QS coordinates the synthesis of HSAF remains unknown. RpfG, as a necessary component protein of the DSF mediated QS signal transduction system, contains a C-terminal HD-GYP domain that can affect production of the antifungal weapon HSAF in L. enzymogenes 11,25 . However, how RpfG regulates the synthesis of HSAF remains incompletely studied. The results of the present study provide biochemical, genetic, and biophysical evidence to demonstrate for the rst time that L. enzymogenes reprograms the QS signal system that RpfG interacts with HtsHs to regulate the biosynthesis of the antifungal antibiotic HSAF. HtsH1, HtsH2, and HtsH3 regulate the expression of the synthetic genes of the antifungal weapon HSAF depending on their phosphorylation ( Figure 8).
HD-GYP domain proteins are c-di-GMP PDEs that can degrade c-di-GMP 12 . However, the role of the HD-GYP domain of RpfG in the degradation of c-di-GMP in L. enzymogenes has remained unelucidated. Therefore, we tested the PDE activity of RpfG and successfully showed that it was able to degrade the model substrate c-di-GMP to 5′-pGpG. To test whether the HD-GYP motif was important for catalytic activity in RpfG, we constructed RpfG mutant proteins (RpfG-H190A-MBP, RpfG-D191A-MBP, RpfG-G253A-MBP, and RpfG-P255A-MBP). We tested the c-di-GMP PDE activity of these mutant proteins and suggested that the HD-GYP domain was required for full PDE activity of RpfG in vivo. It is generally speculated that the PDE activity of HD-GYP domain proteins is to degrade c-di-GMP to GMP 11,26,27 . However, we showed that the activity of the HD-GYP domain of RpfG is involved in the degradation of cdi-GMP to 5′-pGpG.
Intriguingly, we found that the strains expressing the RpfG H190A, D191A, G253A, Y254A, and P255A mutant proteins resulted in increased HSAF production compared with the ΔrpfG mutant strain. Importantly, we found that the concentration of the ΔrpfG mutant did not signi cantly change c-di-GMP production compared with of the wild-type strain in the antifungal weapon HSAF-production medium (10% TSB). These data demonstrated that the regulatory activity of RpfG does not depend on its PDE enzymatic activity. This is the rst report showing that a PDE does not depend on its c-di-GMP-degrading activity to regulate a downstream pathway. Thus, we wondered whether RpfG regulates HSAF synthesis through interactions with other proteins in L. enzymogenes.
Bioinformatics predictions have shown that RpfG may interact with three HyTCS proteins (HtsH1, HtsH2, and HtsH3). Then, we used pull-down and SPR to demonstrate the binding events between the RpfG and HtsH1, HtsH2, or HtsH3 proteins. Notably, RpfG and HtsH1, HtsH2, or HtsH3 interactions affect the PDE activity of RpfG. However, how RpfG affects HtsH1, HtsH2, or HtsH3 remains unknown. We speculate that RpfG may affect HtsH1, HtsH2, or HtsH3 autophosphorylation. However, we could not obtain the fulllength HtsH1, HtsH2, and HtsH3 proteins, so further clari cation of these possible mechanisms will help elucidate the mechanism underlying the RpfG interaction with HtsH1, HtsH2, or HtsH3. Moreover, we found that htsH1, htsH2, and htsH3 likely constitute a single transcription unit. HyTCS-based regulation may be crucial for responding to environmental changes and nely tuning gene expression [28][29][30][31] . However, the biological function of three consecutive HyTCS proteins has not been reported in bacteria.
In this study, we found that the in-frame deletion of the htsH1, htsH2, and htsH3 coding sequences signi cantly decreased HSAF production. Thus, we speculate that RpfG may interact with three HyTCS proteins to coordinate HSAF production in L. enzymogenes. To test this hypothesis, we performed an E. coli-based one-hybrid assay and EMSA. The results demonstrated that HtsH1, HtsH2, and HtsH3 could directly target the promoters of HSAF biosynthesis genes. We further analysed the transcription level of HSAF biosynthesis-related genes in htsH1, htsH2, and htsH3 mutants. Knockout of htsH1, htsH2, and htsH3 signi cantly reduced the transcription level of the antifungal weapon HSAF biosynthesis genes. These results suggest that HtsH1, HtsH2, and HtsH3 can directly regulate HSAF biosynthesis gene expression and increase production of the antifungal weapon HSAF in L. enzymogenes.
Phosphorylation of TCS is critical for regulating the expression of downstream genes [32][33][34] . Phosphorylated HtsH1, HtsH2, and HtsH3 positively regulate HSAF biosynthesis and argue with the existing literature concerning whether this is a common route of gene expression regulation. We used Mn 2+ -Phos-tag SDS-PAGE and EMSA to show that HtsH1, HtsH2, and HtsH3 target plafB depending on their phosphorylation activity. In this study, we report for the rst time the biological functions of the three HyTCS proteins HtsH1, HtsH2, and HtsH3 in the regulation of antibiotic biosynthesis.
One of the notable results of this study is that RpfG, and HtsH1, HtsH2 and HtsH3 regulatory patterns seem to be conserved mechanisms in Lysobacter and Xanthomonas. To our knowledge, RpfG represents a unique example of a c-di-GMP metabolic enzyme that directly interacts with three HyTCS proteins (HtsH1, HtsH2 and HtsH3) to regulate HSAF biosynthesis.

Experimental Procedures
Bacterial strains, plasmids, and growth conditions The strains and plasmids used in this study are shown in Table S1. E. coli strains were grown in Luria-Bertani medium (10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl, pH 7.0) at 37°C. L. enzymogenes strains were grown at 28°C in Luria-Bertani medium and 10% TSB. For the preparation of culture media, tryptone, peptone, beef extract, and yeast extract were purchased from Sangon Biotech (Shanghai, China). When required, antibiotics were added (30 μg/mL kanamycin sulphate, 50 μg/mL gentamycin) to the E. coli or L. enzymogenes cultures. The bacterial growth in liquid medium was determined by measuring the optical density at 600 nm (OD600) using a Bioscreen-C Automated Growth Curves Analysis System (OY Growth Curves FP-1100-C, Helsinki, Finland).

Site-directed mutagenesis
Site-directed mutagenesis and essentiality testing were performed as described previously 35 . To obtain the RpfG mutant proteins and rpfG site-directed mutant strains, plasmids harbouring mutations in rpfG were constructed. For example, to obtain the H190A mutation in RpfG, approximately 500-bp DNA fragments anking the rpfG gene were ampli ed with Pfu DNA polymerase using L. enzymogenes genomic DNA as template and either MBP-rpfG P1 and rpfG H190A P1 (for the Up rpfG H190A mutant), or rpfG H190A P1 and MBP-rpfG P2 (for the Down rpfG H190A mutant) as the primers (Table S2). The fragments were connected by overlap PCR using the primers MBP-rpfG P1 and MBP-rpfG P2. The fused fragment was digested with BamH I and HindIII and inserted into pMAl-p2x to obtain the plasmid pMAl-rpfG H190A. The other four site-directed mutant plasmids (D191A, G253A, Y254A, and P255A) were constructed using a similar method.

Protein expression and puri cation
Protein expression and puri cation were performed as described previously 36 . To clone the rpfG gene, genomic DNA extracted from L. enzymogenes was used for PCR ampli cation using Pfu DNA polymerase, and the primers are listed in Table S2. The PCR products were inserted into pMAl-p2x to produce the plasmids pMAl-rpfG. The rpfG gene was veri ed by nucleotide sequencing by Genscript (Nanjing, Jiangsu, China). Le rpfG and rpfG site-directed mutants with a vector-encoded maltose binding protein in the N-terminus were expressed in E. coli BL21 (DE3) and puri ed with dextrin sepharose high performance (Qiagen, Chatsworth, CA, USA) using an a nity column (Qiagen). The protein purity was monitored by SDS-PAGE. His 6 -tagged protein expression and puri cation were performed as described previously [35][36][37] .

PDE activity assays in vitro
The PDE activity assay was performed essentially as described earlier 24 . Brie y, 2 μM MBP-RpfG or its derivatives were tested in buffer containing 60 mM Tris-HCl (pH 7.6), 50 mM NaCl, 10 mM MnCl 2 , and 10 mM MgCl 2 . The reaction was started by the addition of 100 μM c-di-GMP. All reaction mixtures were incubated at 28°C for 5 to 60 min, followed by boiling for 10 min to stop the reaction. The samples were ltered through a 0.2 μM pore size cellulose-acetate lter, and 20 μl of each sample was loaded onto a reverse-phase C18 column and separated by HPLC. The separation protocol involved two mobile phases, 100 mM KH 2 PO 4 plus 4 mM tetrabutylammonium sulphate (A) and 75% A + 25% methanol (B).
C-di-GMP extraction and quanti cation C-di-GMP extraction and quanti cation were performed as described previously 24 . Cultures were grown in 10% TSB at 28°C until the OD600 reached 1.5 based on the growth curve. Cells from 2 mL of the culture were harvested for protein quanti cation by BCA (TransGen, China). Cells from 8 mL of culture were used for c-di-GMP extraction using 0.6 M HClO 4 and 2.5 M K 2 CO 3 . The samples were subjected to 0.22 µm Mini-Star ltration, and the ltrate was concentrated to 100 µL for liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis on an AB SCIEX QTRAP 6500 LC-MS/MS system (AB SCIEX, USA). The separation protocol involved two mobile phases, buffer A: 100 mM ammonium acetate plus 0.1% acetic acid; buffer B: 100% methanol. The gradient system was from 90% buffer A and 10% buffer B to 20% buffer A and 80% buffer B. The running time was 9 min, and the ow rate was 0.3 mL/min.

Gene deletion and complementation
The in-frame deletions in L. enzymogenes OH11 were generated via double-crossover homologous recombination as described previously 24,35 using the primers listed in Table S2. In brief, the anking regions of each gene were PCR-ampli ed and cloned into the suicide vector pEX18Gm (Table S1). The deletion constructs were transformed into the wild-type strain by electroporation, and gentamycin was used to select for integration of the nonreplicating plasmid into the recipient chromosome. A singlecrossover integrant colony was spread on LB medium without gentamycin and incubated at 28°C for 3 days, and after appropriate dilution, the culture was spread on LB plates containing 15% sucrose. Colonies sensitive to gentamycin were screened by PCR using the primers listed in Table S2, and the gene deletion strains were obtained.
For gene complementation constructs, DNA fragments containing the full-length genes along with their promoters were PCR ampli ed and cloned into the versatile plasmid pBBR1MCS5 38 . The resulting plasmids were transferred into the L. enzymogenes strain by electroporation, and the transformants were selected on LB plates containing Gm.

RNA-Seq
The RNA-Seq assay was performed as described previously 39,40 . Brie y, the wild-type, ΔhtsH1, ΔhtsH2, ΔhtsH3, ΔhtsH12, ΔhtsH13, ΔhtsH23, and ΔhtsH123 mutant strains were grown in 10% TSB medium at 28°C, and their cells were collected when the OD600 reached 1.0 based on the growth curve. The collected cells were used for RNA extraction by the TRIzol-based method (Life Technologies, CA, USA), and RNA degradation and contamination were monitored on 1% agarose gels. Then, clustering and sequencing were performed by Genedenovo Biotechnology Co., Ltd (Guangzhou, Guangdong, China). To analyse the DEGs between the wild-type, ΔhtsH1, ΔhtsH2, ΔhtsH3, ΔhtsH12, ΔhtsH13, ΔhtsH23, and ΔhtsH123 mutant strains, the gene expression levels were further normalized using the fragments per kilobase of transcript per million (FPKM) mapped reads method to eliminate the in uence of different gene lengths and amounts of sequencing data on the calculation of gene expression. The edgeR package (http://www.r-project.org/) was used to determine DEGs across samples with fold changes ≥ 2 and a false discovery rate-adjusted P (q value) < 0.05. DEGs were then subjected to enrichment analysis of GO functions and KEGG pathways, and q values were corrected using < 0.05 as the threshold.
Quantitative real-time PCR Quantitative real-time PCR was carried out according to previous studies 41 . The bacterial cells were collected when the cellular optical density (OD600) reached 1.0 in 10% TSB. Total RNA was extracted using a TRIzol-based method (Life Technologies, CA, USA). RNA quality control was performed via several steps: (1) the degree of RNA degradation and potential contamination were monitored on 1% agarose gels; (2) the RNA purity (OD260/OD280, OD260/OD230) was checked using a NanoPhotometer® spectrophotometer (IMPLEN, CA, USA); and (3) the RNA integrity was measured using a Bioanalyser 2100 (Agilent, Santa Clara, CA, USA). The primers used in this assay are listed in Table S2. cDNA was then synthesized from each RNA sample (400 ng) using the TransScript ® All-in-One First-Strand cDNA Synthesis SuperMix for qPCR (One-Step gDNA Removal) Kit (TransGen Biotech, Beijing, China) according to the manufacturer's instructions. qRT-PCR was performed using TransStart Top Green qPCR SuperMix (TransGen Biotech) on a QuantStudio TM 6 Flex Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) with the following thermal cycling parameters: denaturation at 94°C for 30 s, followed by 40 cycles of 94°C for 5 s and 60°C for 34 s. Gene expression analyses were performed using the 2 -∆∆CT method with 16S rRNA as the endogenous control, and the expression level in the wild type was set to a value of 1. The experiments were performed three times, and three replicates were examined in each run.

Pull-down assay
The assay was performed as described previously 24 . Brie y, the puri ed proteins were used to perform the pull-down assay in a reaction system comprising 800 µL PBS buffer, 5 µM ( nal concentration) MBP-RpfG and HtsH1C-Flag-His, HtsH2C-HA-His or HtsH3C-Myc-His proteins, and 50 µL Dextrin Sepharose High Performance agarose (Sigma-Aldrich, St. Louis, MO, USA). All samples were incubated at 4°C overnight. The agarose was collected by centrifugation and washed 10 times with PBS containing 1% Triton X-100 to remove non-speci cally bound proteins. The MBP-bead-captured proteins were eluted by boiling in 6× SDS loading dye for 10 min. These samples were subjected to SDS-PAGE and Western blotting. Protein detection involved the use of MBP-(ab49923), Flag-(ab1162), HA-(ab187915), Myc-(ab32072), and His-(ab18184) speci c antibodies obtained from Abcam, UK.

Phosphorylation analysis through Phos-tag gel
The puri ed HtsH1C-Flag-His, HtsH2C-HA-His, or HtsH3C-Myc-His proteins (100 ng) were incubated with CIAP (Solarbio, Beijing, China) at 28°C for 60 min and resolved by 8% SDS-PAGE prepared with 50 µM acrylamide-dependent Phos-tag ligand and 100 µM MnCl 2 as previously described 42 . Gel electrophoresis was performed with a constant voltage of 80 V for 3-6 h. Before transfer, the gels were equilibrated in transfer buffer with 5 mM EDTA for 20 min two times, followed by transfer buffer without EDTA for another 20 min. Protein transfer from the Mn 2+ phos-tag acrylamide (APExBIO, Houston, USA) gel to the PVDF membrane (Millipore, Massachusetts, USA) was performed for ~24 h at 80 V at 4°C, and then the membrane was analysed by Western blotting using the anti-His antibody.
Bacterial one-hybrid assays Bacterial one-hybrid assays were performed as previously described 43,44 . In brief, the bacterial one-hybrid reporter system contains three components: the plasmids pBXcmT and pTRG, which are used to clone the target DNA and to express the target protein, respectively, and the E. coli XL1-Blue MRF kan strain, which is the host strain for the propagation of the pBXcmT and pTRG recombinants 45 . In this study, the promoter of the HSAF biosynthesis gene (plafB) was cloned into pBXcmT to generate the recombinant vector pBXcmT-plafB. Similarly, the coding regions of Le htsH1, Le htsH2, and Le htsH3 were cloned into pTRG to create the nal constructs pTRG-htsH1, pTRG-htsH2, and pTRG-htsH3, respectively. The two recombinant vectors were transformed into the XL1-Blue MRF kan strain. If direct physical binding occurred between HtsH1, HtsH2, or HtsH3 and plafB, the positive-transformant E. coli strain containing both pBXcmT-plafB and pTRG-HtsHs would grow well on selective medium, that is, minimal medium containing 5 mM 3-amino-1,2,4-triazole, 8 μg/mL streptomycin, 12.5 μg/mL tetracycline, 34 μg/mL chloramphenicol, and 30 μg/mL kanamycin. Furthermore, cotransformants containing pBX-R2031/pTRG-R3133 served as a positive control 45 , and cotransformants containing either empty pTRG or pBXcmT-plafB were used as negative controls. All cotransformants were spotted onto selective medium, grown at 28°C for 3-4 days, and then photographed.
Electrophoretic mobility gel shift assays (EMSAs) EMSA was performed as previously described 46,47 . For HtsH1, HtsH2, or HtsH3 gel shift assays, we used DNA fragments that included plafB as a probe. The probe DNA (50 ng) was mixed with protein in a 20 μL reaction mixture containing 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 1 mM dithiothreitol, and 0.4% glycerol. After incubation for 30 min at 28°C, samples were electrophoresed on a 5% nondenaturing acrylamide gel in 0.5× TBE buffer at 4°C. The gel was soaked in 10,000-fold-diluted SYBR Green I nucleic acid dye (Sangon Biotech, Shanghai, China), and the DNA was visualized at 300 nm.
HSAF extraction and quanti cation HSAF was extracted from 50 mL L. enzymogenes cultures grown in 10% TSB for 48 h at 28°C with shaking (at 180 rpm). HSAF was detected via HPLC and quanti ed per unit of OD600 as described previously 6,24,48 . Three biological replicates were used, and each was examined with three technical replicates.

Statistical analyses
The experimental datasets were subjected to analyses of variance using GraphPad Prism 7.0. The signi cance of the treatment effects was determined by the F value (P = 0.05). If a signi cant F value was obtained, separation of means was accomplished by Fisher's protected least signi cant difference at P ≤ 0.05. F.L. revised the manuscript. All authors read and approved the nal manuscript.