The 2-acetoxyl-2-methyl-ethyl acetoacetate, NADPH, 2-ketoisovalerate (KIV), 2-ketoisocaproate (KIC), 2-keto-methylvalerate (KMV), [¹³C] glucose, and antibiotics were from Sigma–Aldrich (St. Louis, MO, United States). Takara Biotechnology Co. (Dalian, China) provided the molecular biology reagents. Novagen (Madison, WI, United States) provided the pET vectors. Ni-agarose columns were from Invitrogen (Carlsbad, CA, United States). Agilent Technologies (Palo Alto, CA, United States) provided HC-C18 HPLC columns and Bio-Rad (Hercules, CA, United States) provided Quick Start Bradford dye reagent. All other reagents were of the highest available quality. Sangon Biotechnology Co. (Shanghai, China) synthesized oligonucleotide primers.

Strains and plasmids used in this study are in Supplementary Table S1. E. coli strains were grown in Luria-Bertani medium at 37°C. Xcc strains were grown at 30°C in media listed in Supplementary Table S3. When required, antibiotics were added at 100 μg/mL sodium ampicillin, 30 μg/mL kanamycin sulfate, 30 μg/mL gentamicin for E. coli or 10 μg/mL for Xcc, and 50 μg/mL rifampicin. Bacterial growth in liquid medium was determined by measuring optical density at 600 nm (OD₆₀₀) using a Bioscreen-C Automated Growth Curves Analysis System (OY Growth Curves FP-1100-C, Helsinki, Finland).

To clone the Xcc ilvC gene, genomic DNA extracted from strain Xcc Xc1 was used for PCR amplification with Pfu DNA polymerase, using primers in Supplementary Table S2. PCR products were inserted into pET-28b (+) to produce plasmids pKH2. The ilvC gene was confirmed by nucleotide sequencing by Shanghai Sangon Inc. (Shanghai, China). Xcc ilvC with a vector-encoded His₆-tagged N-terminus was expressed in E. coli BL21 (DE3), and purified with Ni-NTA agarose (Qiagen, Chatsworth, CA, United States) using a nickel-ion affinity column (Qiagen). Protein purity was monitored by SDS-PAGE.

The enzymatic activity assays for IlvC were as described previously (Brinkmann-Chen et al., 2013). Briefly, reaction mixtures contained 250 mmol/L potassium phosphate (pH 7.0), 1 mmol/L DTT, 200 μmol/L NADPH, 10 mmol/L MgCl₂, and various concentrations of 2-acetolactic produced from hydrolyzation of 2-acetoxyl-2-methyl-ethyl acetoacetate by sodium hydroxide, in a final volume of 100 μL. Reactions were initiated by adding 1 μg IlvC, and ketol-acid reductoisomerase activity of IlvC determined by monitoring the rate of NADPH oxidation at 340 nm using extinction coefficient 6220 M^-1.

To disrupt the Xcc ilvC gene, the pK18mobsacB-borne in-frame deletion suicide plasmid pKH5 was constructed (Supplementary Figure S2). The 500-bp DNA fragments flanking the ilvC gene were amplified with Pfu DNA polymerase using Xcc genomic DNA as template, and either Xcc ilvC EcoRI and Xcc ilvC up1 (for Up ilvC), or Xcc ilvC down1 and Xcc ilvC HindIII (for Dn ilvC), as primers (Supplementary Table S2). Fragments were purified and joined by overlapping PCR. The fused fragment was digested with EcoRI and HindIII, and inserted in pK18mobscaB (Schafer et al., 1994) to obtain plasmid pKH5. Following mating derivatives of E. coli strain S17-1 carrying suicide plasmid pKH5 with Xcc Xc1 on NYG (Supplementary Table S3) plates for 36 h at 30°C (Supplementary Figure S2), cells were suspended in NYG medium, and appropriate dilutions were inoculated onto NYG plates containing rifampicin to select against the donor strain, plus kanamycin to select for integration of the non-replicating plasmid into the recipient chromosome. A single-crossover integrant colony was spread on NYG medium without kanamycin at 30°C for 36 h, and after appropriate dilutions, culture was spread on NYG plates containing 15% sucrose. Colonies sensitive to kanamycin were screened by PCR using primers in Supplementary Table S2, and ilvC deletion strain Xcc KH1 was obtained (Supplementary Figures S2B,C). Using a similar approach, we deleted the ilvC gene from Xcc Xc1 ΔrpfBC for the Xcc KH2 mutant. For single-copy complementation of the ilvC mutant, the coding region of ilvC plus 600 bp upstream of the ilvC translational start codon was PCR amplified and cloned into a versatile Mini-Tn7 delivery vector (mini-Tn7T-Gm). The resulting plasmid was transferred into strain Xc1 KH1 or Xc1 KH2 by electroporation, with selection as described previously (Jittawuttipoka et al., 2009). The KH3 and KH4 strains were obtained.

Bacterial cultures were grown aerobically for 2–4 days. Cells were harvested and washed three times with sterile water. Fatty acid methyl esters were synthesized and extracted as described previously (Stead, 1989). Cellular lipids were saponified by addition of 2 mL sodium hydroxide/methanol solution at 100°C for 40 min with shaking (800 rpm). Fatty acids were methylated by addition of 4 mL hydrochloric acid/methanol solution at 80°C for 30 min, and cooled to below 20°C. Fatty acid methyl esters were obtained by three extractions with 1 mL petroleum ether. Solvent was removed under a stream of nitrogen, and residue dissolved in 100 μL of hexane. The crude extract was filtered with 0.22-μm Mini-star units and 2 μL of extract was analyzed by gas chromatography-mass spectrometry (GC-MS). Samples were analyzed with a GC-MS system (Agilent 5975c) with chromatographic column DB 5MS. The oven temperature was held at 100°C for 5 min, changed at 10°C/min to 200°C and held for 5 min, and changed at 10°C/min to 250°C and held for 5 min. Electron impact ionization (EI⁺, 70 eV) was used for all samples. Mass spectrometry was carried out at 1 sec/scan, m/z 35-500, 1 kV, and data were analyzed by the NIST 08 database.

For analysis of ¹³C-labeled fatty acids, 10 mL cultures of Xc1 or KH1 grown in XOLN medium for 1 day were harvested and washed three times in PBS (Supplementary Table S3). Bacterial cells were re-suspended in 10 mL XO medium (Supplementary Table S3) and 500 μL transferred into 50 mL XO medium supplemented with 1 mg/mL ¹³C-glucose (for Xc1) or 1 mg/mL ¹³C-glucose plus 100 μmol/L valine and isoleucine mixtures (for KH1). Cells were grown at 30°C for 2 days. Fatty acids were extracted and analyzed as described above.

The protocol for extraction and purification of DSF family components was described previously (Deng et al., 2015; Zhou et al., 2015a,b). Xcc strains were cultured in liquid medium for 36 h and 50 mL bacterial supernatant was collected by centrifugation at 4000 × g for 15 min at 4°C. The pH of supernatants was adjusted to 4.0 by adding hydrochloric acid prior to two extractions with an equal volume of ethyl acetate. Ethyl acetate fractions were collected and solvent removed by rotary evaporation to dryness at 40°C. Residue was dissolved in 100 μL methanol. Crude extract was subjected to 0.22-μm Mini-star filtration and filtrate was concentrated to 100 μL. Extract (10 μL) was injected into a C18 reverse-phase HPLC column (4.6 mm × 250 mm, Agilent Technologies), eluted with water in methanol (23:77 v/v, 0.1% formic acid) at a flow rate of 1 mL/minute in a Waters HPLC E2695 system (Waters, Milford, MA, United States) with a UV220 detector.

For detection of DSF signals produced by resting Xcc cells, 50 mL KH2 cultures grown in XOLN medium for 24 h were harvested and washed three times using PBS buffer. Cells were re-suspended in 50 mL PBS buffer and grown for 2 h. Cultures were harvested and washed three times using PBS buffer, and transferred to PBS medium supplemented with nutrients, and cells were cultivated at 30°C for 6 h. DSF family signals in supernatants were detected as described above.

Virulence was tested using potted Chinese radishes (Raphanus sativus L. var. radiculus Pers.) as described previously (Dow et al., 2003). Bacteria grown overnight in NYG medium were washed and re-suspended in PBS buffer to OD₆₀₀ 0.1. For leaf clippings, the last completely expanded leaf was cut and dipped in bacterial suspensions at OD₆₀₀ 0.1. Lesion length was measured 14 days after inoculation on 30 leaves for each strain tested.

Analysis of variance for experimental datasets used Graphpad prism 5.0. Significant effects of treatment were determined using F-values (P = 0.05). When a significant F test was obtained, separation of means was accomplished by Fisher’s protected least significant difference at P ≤ 0.05.

The genome of Xcc strain ATCC 33913 possesses an entire gene cluster required for the biosynthetic pathway of BCAAs (valine, leucine and isoleucine) (Figure 1B). In this gene cluster, XCC3323 (ilvC) is annotated to encode ketol-acid reductoisomerase (Figure 1C). Protein sequence alignments indicated that Xcc IlvC shared 56.9% identical residues with Pseudomonas aeruginosa IlvC (Eom et al., 2002; Ahn et al., 2003), and conserved regions I–V defined by the X-ray crystallography structure of P. aeruginosa IlvC were present in Xcc IlvC (Dumas et al., 2001; Tyagi et al., 2005) (Supplementary Figure S1). The residues in ketol-acid reductoisomerase that are involved in binding metal ions, an inhibitor and NADPH were conserved in Xcc IlvC (Dumas et al., 2001; Tyagi et al., 2005) (Supplementary Figure S1). To identify if Xcc ilvC encoded ketol-acid reductoisomerase, recombinant N-terminal hexahistidine-tagged Xcc IlvC was produced. The protein had a monomeric molecular weight of 37 kDa and was purified by nickel chelate chromatography to obtain preparations that gave single bands on SDS-gel electrophoresis (Figure 2A). To probe the ketol-acid reductoisomerase activity of Xcc IlvC, decreased absorbance of NADPH at 340 nm was monitored spectrophotometrically in reaction mixtures containing 2-acetolactate as a substrate. Along with increased 2-acetolactate concentrations, the rate of NADPH oxidation increased, indicating that Xcc IlvC possessed ketol-acid reductoisomerase activity (Figure 2B).

Ketol-acid reductoisomerase (IlvC) catalyzes the conversion of acetohydroxy acids into dihydroxy valerates and is a key enzyme in the second step of the BCAAs biosynthetic pathway (Dumas et al., 2001; Bastian et al., 2011; Brinkmann-Chen et al., 2013) (Figure 1A). To identify the physiological functions of IlvC in Xcc, knockout strain KH1, in which the ilvC (XCC3323) gene was deleted, was constructed by allelic replacement (Supplementary Figure S2). Growth of the Xcc KH1 (ΔilvC) strain was tested in minimal liquid medium XOS (Supplementary Table S3). The Xcc KH1 strain failed to grow in XOS medium, but the complementary strain KH3, in which KH1 carried a single copy of wild type ilvC at the glmS site, had fully restored growth in XOS medium (Figure 3A). However, in XOLN medium (Supplementary Table S3), Xcc KH1 grew as well as the wild type strain Xcc Xc1 (Figure 3B). These results indicated that deletion of ilvC caused Xcc to fail to produce some nutrients from carbohydrate metabolic pathways to support Xcc growth. Given IlvC is the key enzyme of the BCAA biosynthetic pathway, we tested the growth of strain Xcc KH1 in XOS medium supplemented with valine, leucine, or isoleucine. None of the single BCAAs was able to restore Xcc KH1 growth on XOS plates (data not shown). However, supplementing of XOS medium with 100 μmol/L mixtures of valine and isoleucine restored Xcc KH1 strain growth on XOS plates (Figure 3C). While the same concentration of valine and leucine or leucine and isoleucine mixtures did not restore strain Xcc KH1 growth. These results suggested that deletion of ilvC caused Xcc to fail to produce valine and isoleucine. Generally, in cells, production of BCAAs comes from transamination of branched-chain 2-keto acids KIV, KIC and KMV, respectively. Thus, to determine if deletion of ilvC led Xcc to fail to produce branched-chain 2-keto acids from the carbohydrate metabolic pathway, we supplemented XOS medium with KIV, KIC, and KMV and tested if strain Xcc KH1 grew. Xcc KH1 grew as well as the wild type strain Xcc Xc1 when XOS medium was supplemented with KIV and KMV mixtures (Figure 3C). However, single branched-chain 2-keto acids, or KIV and KIC, or KIC and KMV mixtures did not restore strain KH1 growth (data not shown). This result confirmed that Xcc ilvC encoded ketol-acid reductoisomerase and deletion of ilvC caused Xcc to fail to produce KIV and KMV from carbohydrate metabolism, and caused Xcc auxotrophy for valine and isoleucine.

We analyzed mutant strain Xcc KH1 reaction to addition of various concentrations of valine and isoleucine mixtures. Addition of 100 μmol/L mixtures supported Xcc KH1 growth, although growth was weaker than the wild type strain Xc1 in XOS medium. However, when mixtures were increased to 300 μmol/L, Xcc KH1 growth reached the level of wild strain Xc1 growth in XOS medium. Addition of 500 μmol/L mixtures did not improve growth of mutant strain Xcc KH1 (Figure 3D). The effects of KIV and KMV mixtures on Xcc KH1 growth were tested and similar results were obtained: 300 μmol/L KIV and KMV mixtures best supported Xcc KH1 growth (Supplementary Figure S3A).

The effects of various concentrations of valine and isoleucine mixtures on Xcc KH1 growth at low temperature were also determined. Higher concentrations of valine and isoleucine mixtures did not increase Xcc KH1 growth at low temperature compared with growth of the wild type strain Xc1 in XOS medium at the same temperature (Figure 3E). Higher concentrations of KIV and KMV mixtures did not increase Xcc KH1 growth at low temperature either (Supplementary Figure S3B).

Deletion of ilvC caused Xcc auxotrophy for valine and isoleucine. Therefore, we hypothesized that the fatty acid profile of mutant strain Xcc KH1, especially for BCFAs, would be distinct from wild type strain Xc1. We analyzed the fatty acid composition of total lipid extracts from Xcc KH1 grown in XOLN medium at 30°C by GC-MS. The species and amounts of fatty acids of Xcc KH1, especially BCFAs, were almost the same as that in Xcc Xc1 (Table 1). To test if temperature affected the fatty acid profile of strain Xcc KH1, the fatty acid composition of Xcc KH1 grown at 15°C was determined. The fatty acid profile of Xcc KH1 did not differ from Xcc Xc1 under this temperature condition (Table 1). These results suggested that deletion of ilvC did not affect BCFA synthesis in strain Xcc.

Then, we determined why deletion of Xcc ilvC did not lead to changes in Xcc BCFAs composition with an isotope-labeling method. We fed the wild type strain Xcc Xc1 and mutant strain Xcc KH1 with [¹³C] glucose, which as all six carbons involved in ¹³C labeling, and we analyzed the fatty acid composition as fatty acid methyl esters (FAME) using GC-MS. The fatty acid profiles of the two strains obtained were almost same as the wild type strain fed unlabeled glucose (data not shown). Next, the structure of each ¹³C-labeled FAME peak was assigned using mass spectrometry. Compared to the corresponding unlabeled FAME peak, all ¹³C-labeled FAME molecules from both strains were heavier. For the same species n-C_m:0 straight-chain fatty acids ester, the m/z of ¹³C-labeled FAME was the same for the wild type and mutant strain, and was m Da heavier than unlabeled FAME. For example, the m/z of unlabeled n-C_16:0 methyl ester was 270 (Figure 4A), and the m/z of ¹³C-labeled n-C_16:0 methyl ester in both strains was 286 (Figures 4B,C). This measurement was 16 Da higher than the unlabeled n-C_16:0 methyl ester, which indicated that all carbons in n-C_16:0 fatty acid in both the wild type and mutant strains were from catabolism of ¹³C-labeled glucose. However, for BCFAs, the m/z of ¹³C-labeled FAME from wild type strain Xc1 differed from mutant strain KH1. The m/z of ¹³C-labeled anteiso-C_15:0 methyl ester in the wild type strain was 271 (Figure 4E), 15 Da higher than the unlabeled anteiso-C_15:0 methyl ester (Figure 4D), indicating that all 15 carbons of anteiso-C_15:0 in wild type strain Xc1 were labeled by ¹³C. While the m/z of ¹³C-labeled anteiso-C_15:0 methyl ester in mutant strain KH1 was 266 (Figure 4F), which was 10 Da higher than unlabeled anteiso-C_15:0 methyl ester and 5 Da less than the ¹³C-labeled anteiso-C_15:0 methyl ester in the wild type strain. These results suggested that only 10 carbons in anteiso-C_15:0 in mutant strain KH1 were from the catabolism of ¹³C-labeled glucose, while five unlabeled carbons came from the exogenous isoleucine. We also analyzed the ¹³C-labeled iso-C_16:0 methyl ester in the mutant strain, and concluded that four unlabeled carbons in the iso-C_16:0 methyl ester of the mutant strain came from exogenous supplement with valine (Supplementary Figure S4). All these results confirmed that deletion of ilvC resulted in Xcc not synthesizing BCFAs de novo from the carbohydrate metabolic pathway and the mutant strain used exogenous BCAAs to produce BCFAs and support strain growth.

To further investigate use of exogenous BCAAs to produce BCFAs by the ilvC deletion mutant, different concentrations of valine and isoleucine mixtures were added to XOS medium and the fatty acid composition of strain KH1 grown at 30°C or 15°C was determined by GC-MS. Compared to the fatty acid composition of wild type strain Xc1 grown in XOS medium at 30°C, supplement with valine and isoleucine mixtures increased the amount of BCFAs in mutant strain KH1, especially the amount of anteiso-C_15:0 fatty acid (Table 2). As the concentrations of valine and isoleucine mixtures increased, the total BCFAs from mutant strain KH1 also increased (Table 2). Under these conditions, the total unsaturated fatty acids (UFAs) of strain KH1 were decreased. This result suggested that to a certain degree, BCFAs, especially anteiso-C_15:0, replaced the function of UFAs to maintain membrane fluidity in Xcc. However, although supplementing XOS medium with valine and isoleucine mixtures also increased BCFAs in strain KH1 grown at 15°C, BCFAs were not enhanced in strain KH1 with higher concentrations of valine and isoleucine supplements (Table 2). This result indicated that Xcc did not mainly rely on modulation of BCFAs to respond to low temperature. To confirm this finding, we determined the fatty acid composition of the wild type strain grown at 15°C in XOS medium supplemented with valine and isoleucine mixture and obtained similar results (Supplementary Table S4) that supplementing with higher concentrations of valine and isoleucine mixtures did not significantly increase BCFAs in Xcc cells at low temperature.

Xcc is the causal agent of black rot disease of cruciferous vegetables (He and Zhang, 2008). To investigate the effects of ilvC deletion on virulence toward host plants, the pathogenesis of mutant strain KH1 was tested. Leaf-clipping virulence assays using Chinese radishes were conducted. The average lesion length caused by wild type strain Xc1 on a Chinese radish leaf was 13.1 mm at 2 weeks after inoculation (Figure 5A). Deletion of ilvC in strain KH1 significantly reduced average lesion length (4.5 mm) (Figure 5A), but complementation with a single copy of wild type ilvC fully restored Xcc KH1 virulence against host plants; the average lesion length with the complementary strain was 14.5 mm (Figure 5A). This result indicated that deletion of ilvC led Xcc to have reduced virulence against host plants. Deletion of ilvC prevents Xcc from de novo synthesis of BCAAs. Thus, reduced virulence of mutant strain KH1 might have been due to weak growth of host plants caused by ilvC deletion. To test this possibility, the growth of strain KH1 in Chinese cabbage extracts was determined. Under these conditions, the growth pattern of mutant strain KH1 was similar to the wild type strain Xc1 (Figure 5B), suggesting that ilvC deletion did not weaken growth of Xcc in host plants significantly, and host plants had enough BCAAs to support mutant growth. Next, we evaluated several pathogenicity-related virulence factors produced by the mutant strain. The activity of extracellular enzymes cellulase, amylase, and protease, and the amount of EPS produced by mutant strain KH1 were not significantly different from the wild type strain Xc1 (data not shown). These results indicated that reduction of Xcc pathogenesis by deletion of ilvC was not caused by reduction of these pathogenicity-related virulence factors. We also tested the production of DSF-family signals by mutant strain KH1 grown in NYG medium using the biosensor strain. The data showed that during the KH1 growth, the amount of DSF-family signals was lower than that produced by wild type strain Xc1 grown in the same medium (data not shown). However, whether reduction of Xcc virulence was due to low-level DSF family signals produced in the ilvC mutant needs to be further investigated.

As the Xc1 rpfBC double-deletion strain produced higher levels of DSF family signals (Zhou et al., 2015a), to obtain enough DSF family signals for analysis, we constructed mutant strain KH2 by deleting of ilvC from the parent strain Xc1 ΔrpfBC mutant, and the complementary strain KH4, in which a single copy of wild type ilvC was inserted into the glmS site of strain KH2 by mini-Tn7 delivery vector. Production of DSF-family signals of the mutant strain KH2 grown in NA medium (Supplementary Table S3) was first tested. All strains produced DSF-family signals: DSF, BDSF, and IDSF, but DSF was the main DSF-family signal under these conditions (Figure 6A), consistent with a previous report (Zhou et al., 2015b). However, the amount of each species of DSF-family signals produced by mutant strain KH2 was significantly lower than the Xc1 ΔrpfBC strain (P < 0.01). The complementary strain KH4 yielded DSF-family signals at the level of the Xc1 ΔrpfBC strain. We also examined DSF-family signal production of the mutant strain grown in XLON medium and obtained similar results (Supplementary Figure S5A). These findings indicated that ilvC deletion reduced Xcc synthesis of DSF-family signals.

Given that ilvC deletion caused Xcc auxotrophy for valine and isoleucine, we hypothesized it would prevent Xcc from producing branched-chain DSF-family signals from carbohydrate metabolism, and branched-chain DSF-family signals from mutant strain KH2 would come from conversion of exogenous BCAAs. To test this hypothesis, we determined DSF-family signals produced by resting cells of mutant strain KH2 in PBSS (Supplementary Table S3). Under these conditions, strain Xc1 ΔrpfBC produced three kinds of DSF-family signals: DSF, BDSF, and IDSF. While mutant strain KH2 produced only the BDSF signal. DSF and IDSF, both branched-chain DSF-family signals, were not detected in cultures of strain KH2 (Figures 6B,C). The complementary strain KH4 produced three kinds of DSF-family signals at the level of strain Xc1 ΔrpfBC (Figures 6B,C). To determine if other carbon sources promoted KH2 resting cells to synthesize branched-chain DSF-family signals, we supplemented PBS (Supplementary Table S3) medium with sodium acetate, glucose, mannitol, and glycerol, and DSF-family signal production profiles were determined. None of these candidates promoted KH2 production of branched-chain DSF-family signals (Figure 6D). These results indicated that strain KH2 failed to synthesize branch-chain DSF-family signals from carbohydrate metabolism. To determine if branched-chain DSF-family signals produced by mutant strain KH2 were from conversion of exogenous BCAAs, we tested DSF-family signal production by resting cells of mutant strain KH2 in PBSS medium supplemented with valine, leucine or isoleucine. Resting KH2 cells produced DSF when valine or leucine was added to PBSS medium, while addition of isoleucine to PBSS medium resulted in IDSF production (Figures 6E,F). We also tested DSF-family signal production by KH2 resting cells when KIV, KIC, or KMV were added to PBSS medium (Supplementary Figures S5B,C). Addition of KIV and KIC resulted in KH2 production of DSF, while addition of KMV resulted in production of the IDSF signal. These results confirmed that the branched-chain DSF-family signals produced by KH2 were from conversion of exogenous BCAAs and that KIV and KIC were the precursors of the DSF signal, and KMV was the precursor of the IDSF signal.

To determine why ilvC deletion reduced Xcc pathogenesis against host plants, we determined DSF-family signal production by the KH2 strain grown in Chinese cabbage extracts. All strains (Xc1 ΔrpfBC, KH2 and KH4) mainly produced DSF and BDSF signals. However, the amount of the DSF-family signals produced by strain KH2 was less than the Xc1 ΔrpfBC strain. Although the amount of DSF produced by complementary strain KH4 was almost the same as the amount produced by strain Xc1 ΔrpfBC, the KH4 strain did not produce BDSF at the level of the Xc1 ΔrpfBC strain (Figure 6G). Given that the complementary strain restored virulence against host plants, we hypothesized that the DSF signal was critical for Xcc infection of host plants, even though Xcc produces more BDSF signal under this condition.