Xanthomonas campestris FabH is required for branched-chain fatty acid and DSF-family quorum sensing signal biosynthesis

Xanthomonas campestris pv. campestris (Xcc), a Gram-negative phytopathogenic bacterium, causes black rot disease of cruciferous vegetables. Although Xcc has a complex fatty acid profile comprised of straight-chain fatty acids and branched-chain fatty acids (BCFAs), and encodes a complete set of genes required for fatty acid synthesis, there is still little known about the mechanism of BCFA synthesis. We reported that expression of Xcc fabH restores the growth of Ralstonia solanacearum fabH mutant, and this allows the R. solanacearum fabH mutant to produce BCFAs. Using in vitro assays, we demonstrated that Xcc FabH is able to condense branched-chain acyl-CoAs with malonyl-ACP to initiate BCFA synthesis. Moreover, although the fabH gene is essential for growth of Xcc, it can be replaced with Escherichia coli fabH, and Xcc mutants failed to produce BCFAs. These results suggest that Xcc does not have an obligatory requirement for BCFAs. Furthermore, Xcc mutants lost the ability to produce cis-11-methyl-2-dodecenoic acid, a diffusible signal factor (DSF) required for quorum sensing of Xcc, which confirms that the fatty acid synthetic pathway supplies the intermediates for DSF signal biosynthesis. Our study also showed that replacing Xcc fabH with E. coli fabH affected Xcc pathogenesis in host plants.

Xanthomonas campestris pv. campestris (Xcc) is the causal agent of black rot disease, one of the most destructive diseases of cruciferous vegetables worldwide 21,29 . This bacterium uses quorum sensing (QS) mechanisms mediated by molecules of the diffusible signal factor (DSF) family to regulate the expression of factors that contribute to virulence 8,21 . The DSF signal family is a novel structural class of QS signals with the cis-2-unsaturated fatty acid moiety, which includes cis-11-methyl-2-dodecenoic acid (11-Me-C 12 :Δ 2 ), cis-2-dodecenoic acid (C 12 :Δ 2 ), cis-11-methyldodeca-2,5-dienoic acid (11-Me-C 12 : Δ 2,5 ), and cis-10-methyl-2-dodecenoic acid (10-Me-C 12 :Δ 2 ) in Xcc 8,30,31 (Fig. 1B). Genetic analyses show that Xcc rpfF encodes a key enzyme for biosynthesis of the DSF family signals 32 . Furthermore, using in vitro assays, RpfF is shown to be a bifunctional enzyme that catalyzes not only dehydration of 3-hydroxydodecanoyl-ACP to cis-2-dodecenoyl-ACP, but also cleaves the thioester bond to produce C 12 :Δ 2 , a DSF family member 8,9 (Fig. 1B). This suggests that the DSF family signals synthetic precursors come from the fatty acid biosynthetic pathway 8,9 . According to this hypothesis, to produce 11-Me-C 12 :Δ 2 , another main DSF family signal, Xcc should synthesize 11-methyl-3-hydroxydodecanoyl-ACP first, which is an intermediate of the BCFA synthetic pathway (Fig. 1B). However, although Xcc encodes a complete set of genes required for fatty acid synthesis 33 , there is still little known about the mechanism of BCFA synthesis, especially the substrate specificity of Xcc FabH. In the present study, we use several molecular biological techniques, including genetic complementation, biochemical analysis, and gene knockout, to characterize the functions of Xcc FabH in BCFA synthesis and DSF family signal production in Xcc. We found that FabH is a key BCFA biosynthetic enzyme in Xcc. Moreover, we confirm that Xcc FabH, like FabH in many Gram-positive bacteria, also prefer branched-chain acyl-CoAs as primers to initiate fatty acid synthesis. We also report that replacement of Xcc fabH with E. coli fabH results in the loss of the ability of Xcc to produce 11-Me-C 12 :Δ 2 , and this further provides in vivo evidence that the precursors of DSF family signals shunt from the fatty acid biosynthetic pathway.

Results
Bioinformatics analysis of Xcc FabH. To investigate the Xcc FabH function in BCFA synthesis, alignments of Xcc FabH with FabH proteins from E. coli 34 , L. monocytogenes 17 , and B. subtilis 19 were examined (Fig. 1C). The results showed that the Cys-His-Asn catalytic triad of 3-ketoacyl-ACP synthase III 2 is conserved in Xcc FabH, and the Xcc FabH protein shares 61.2%, 47.2%, 45.3%, and 41.5% identical residues with E. coli FabH, L. monocytogenes FabH, B. subtilis FabHA, and B. subtilis FabHB, respectively (Fig. 1C). We also aligned Xcc FabH with Ralstonia solanacearum FabH 35 , and Streptomyces coelicolor FabH 16 , and the respective identity values were 56.1% and 36.3% (data not shown). Interestingly, both E. coli and R. solanacearum produce only straight-chain fatty acids 12,36 , while L. monocytogenes, B. subtilis, and S. coelicolor are able to synthesize several BCFAs 18,19,37 . Because the Xcc FabH protein has higher amino acid identities when compared to the straight-chain fatty acid producing bacterial FabHs, than to the BCFA producing bacterial FabHs, it would be expected that Xcc FabH would prefer acetyl-CoA as a primer to initiate straight-chain fatty acid synthesis, rather than it playing a role in BCFA synthesis.
The Xcc fabH gene complements the R. solanacearum fabH deletion mutant. To confirm the above hypothesis, the plasmid pYYH1, in which the Xcc fabH was expressed from the E. coli lac promoter in vector pSRK-Km 38 , was introduced into the R. solanacearum fabH mutant strain RsmH by conjugation. The R. solanacearum RsmH strain is an octanoic acid auxotrophic mutant, in which the R. solanacearum genomic fabH gene contains an in-frame deletion 35 . This strain does not grow in medium in the absence of octanoic acid ( Fig. 2A). The derivative of RsmH containing the plasmid pYYH1 grew well on BG medium in the absence of octanoic acid, whereas the strain RsmH derivative carrying the vector plasmid failed to grow under the same conditions ( Fig. 2A). This indicated that Xcc fabH encodes a 3-ketoacyl-ACP synthase III functional in R. solanacearum. To investigate the function of the Xcc FabH in fatty acid synthesis, the fatty acid composition of RsmH containing the pYYH1 plasmid was determined by gas chromatography-mass spectrometry (GC-MS). R. solanacearum GMI1000 does not produce BCFAs 36 (Fig. 2B). However, the fatty acid profile of RsmH containing pYYH1 differed from the R. solanacearum wild type GMI1000 strain, and contained more than 33.8% BCFAs, which mainly included iso-C15, iso-C16, iso-C17, and anteiso-C17 fatty acids ( Fig. 2B and Table S3). This suggested that Xcc FabH could use branched-chain acyl-CoAs, such as isovaleryl-CoA, isobutyryl-CoA, or 2-methylbutyryl-CoA, as substrates to produce BCFAs in vivo.
Xcc FabH exhibits a wide range of substrate specificities in vitro. In order to perform a direct in vitro assay of Xcc FabH substrate specificity, the initial reaction was reconstituted by addition of E. coli fatty acid synthetic enzymes (including FabD, FabG, FabZ, and FabI), Xcc FabH, and an acyl-CoA as described in "Materials and Methods". First, we tested Xcc FabH using acetyl-CoA as a primer to initiate fatty acid synthesis (Fig. 3A). As expected the reaction containing E. coli FabH produced butyryl-ACP (Fig. 3A, lane 4). However, although the Xcc FabH reaction resulted in the appearance of new bands on conformationally sensitive gels, butyryl-ACP was not a product. Surprisingly, Xcc FabH converted acetyl-CoA to hexanoyl-ACP and octanoyl-ACP (Fig. 3A, lane 3). This suggested that Xcc FabH not only uses acetyl-CoA as its primer to initiate fatty acid synthesis, but also condenses short-chain acyl-ACPs with malonyl-ACP to produce longer acyl-ACPs.
Xcc FabH also used branched-chain acyl-CoAs, isobutyryl-CoA, or isovaleryl-CoA, as a primer to initiate fatty acid synthesis (Fig. 3B, lanes 2 and 3). We also tested if Xcc FabH utilized butyryl-CoA, hexanoyl-CoA, octanoyl-CoA, or decanoyl-CoA as a primer to initiate fatty acid synthesis, and found that each of these acyl-CoAs could be converted into longer acyl-ACPs by Xcc FabH (Figs 3B and S1). Hence, Xcc FabH can use a wide range of acyl-CoAs as the primer substrate. In order to probe the substrate specificity of Xcc FabH, the decrease in absorbance of NADPH at 340 nm was monitored spectrophotometrically in reaction mixtures containing holo-ACP, malonyl-CoA, NADPH, E. coli FabD and FabG, Xcc FabH, and various acyl-CoAs (C 2 -C 12 ) (Table 1). Xcc FabH exhibits a low activity with acetyl-CoA but a high activity with isobutyryl-CoA or isovaleryl-CoA, consistent with the FabH enzyme activities of bacteria that produce BCFAs 17,19 (Table 1). However, Xcc FabH had higher activities with 4-to 8-carbon straight chain acyl-CoAs than with branched-chain acyl-CoAs ( Table 1). The best Xcc FabH substrate was butyryl-CoA. As the chain length of the acyl-CoA increased, the activities of Xcc FabH decreased. These results showed that Xcc FabH has a slight preference for short, medium-straight chain acyl-CoAs as a primer to initiate fatty acid synthesis and has the ability to use various acyl-CoAs (including short, medium-straight chain, and branched-chain acyl-CoAs) as primers. This characteristic distinguishes the Xcc FabH enzyme from E. coli and B. subtilis FabH enzymes 17,19,34 . Xcc requires fabH for growth. In order to identify the physiological function of Xcc fabH in BCFA synthesis, we attempted to disrupt the Xcc fabH gene in the Xcc genome. First, we constructed an in-frame deletion suicide plasmid, pZTT-1, according to the strategy described in Fig. S2A. After the plasmid pZTT-1 was introduced into the Xcc wild type strain 8004, Xcc fabH deletion mutants were selected according the methods described in the "Material and Methods". Although several independent experiments were done, no Xcc fabH deletion strain was isolated. Only the single crossover integrant, Xcc YH1, was isolated ( Fig. S2B,C), suggesting that Xcc fabH is an essential gene for Xcc growth.
To confirm this possibility, a new suicide plasmid pYYH-3 was constructed, in which an inner 550 bp DNA fragment of Xcc fabH was inserted into the plasmid pK18mobsacB (Fig. S3). The plasmid pYYH-3 was introduced into Xcc wild type strain 8004 by conjugation, and the conjugants were selected on nutrient glycerol medium (NYG) plates containing kanamycin. Although several selections were performed, no kanamycin resistant conjugant was isolated (Fig. S3), indicating that Xcc fabH is essential for Xcc growth.
Xcc fabH can be replaced with E. coli fabH. Bioinformatics analyses showed that Xcc FabH is highly identical to E. coli FabH. Thus, to test whether a straight chain-specific FabH could provide Xcc FabH function, we replaced the Xcc fabH with the in-frame E. coli fabH. First, we constructed a replacement mutant plasmid, pZTT-3, a pK18mobsacB-borne plasmid that carries a long DNA fragment that includes a 500-bp upstream flanking sequence of Xcc fabH, the complete E. coli fabH gene, and a 500-bp downstream flanking sequence of Xcc fabH (Fig. S4A). After the introduction of pZTT-3 into the wild-type Xcc strain 8004 by conjugation, cultures were plated on NYG containing sucrose. Colony PCR assays using the primer pair EcfabH NdeI and EcfabH XbaI (Table S2) showed that one of the colonies, named strain Xcc T-3 (Xcc fabH::EcfabH), contained the 1.3 kb E. coli fabH-containing fragment (Fig. S4B,C). As expected no fragments were amplified from the wild type strain using the same primer pair (Fig. S4B,C). The replacement of Xcc fabH with E. coli fabH in strain Xcc T-3 was also confirmed by sequencing of the PCR fragment of the insertion allele.
We then examined the growth of Xcc T-3 in NYG or minimal medium (SXFM). The growth of Xcc T-3 was substantially slower than that of wild type strain Xcc 8004 in either medium (Fig. 4A). To investigate possible reasons for this observation, two possibilities were tested. The first possibility was that the difference of E. coli FabH substrate specificities from Xcc FabH caused Xcc T-3 strain not to produce enough BCFAs to maintain its growth. In fact, Xcc T-3 lost the ability to produce BCFAs (Please see next section). Therefore, we supplemented iso-C 14:0 , iso-C 15:0 , or anteiso-C 15:0 fatty acids into the NYG medium and tested if these fatty acids restored Xcc T-3 growth. However, all these long-chain BCFAs failed to restore the growth of strain Xcc T-3 in NYG medium (data not shown). This indicated that less production of BCFAs is not the main reason for the weak growth of Xcc T-3. The second possibility was that expression of E.coli FabH causes the low rate of fatty acid biosynthesis in Xcc T-3, which leads to its growth weak. We determined the rate of fatty acid biosynthesis in cell free extracts of Xcc T-3 using acetyl-CoA as a substrate and monitored the decrease in NADPH absorbance at 340 nm. The results showed that the rate of fatty acid biosynthesis in cell free extracts of Xcc T-3 (47.7 ± 0.37 μ mol/mg/min) was much lower than that in cell free extracts of the wild type strain 8004 (60.0 ± 7.39 μ mol/mg/min). To further test this possibility, we tried to increase the rate of fatty acid biosynthesis in the Xcc strain by overexpression of E. coli FabH. We constructed a new replacement mutant by introducing an E. coli fabH-encoded plasmid pSRK-EcfabH into the Xcc fabH single crossover integrant Xcc YH1 (Fig. S5A). After selection, we successfully obtained the Xcc fabH deletion strain Xcc EcH, in which the E. coli fabH gene was expressed from plasmid pSRK-Gm, whereas the Xcc fabH gene had been deleted from the genome (Fig. S5B,C). The data showed that the rate of fatty acid biosynthesis in cell free extracts of Xcc EcH (67.7 ± 11.36 μ mol/mg/minute) was slightly higher than that of the wild type strain. We subsequently tested the growth of strain Xcc EcH in both NYG and SXFM media (Fig. 4A). The results showed that strain Xcc EcH grew faster than strain Xcc T-3 in both media, while the growth of strain Xcc EcH was still weaker than that of wild type strain 8004. The expression level of fabH in strain Xcc 8004, Xcc T-3, and Xcc  EcH was determined by qRT-PCR and data showed that the expression level of Xcc fabH in the wild type strain was almost the same as that of E. coli fabH in Xcc T-3, while the expression level of E. coli fabH in Xcc EcH was 37-fold higher than that of E. coli fabH in Xcc T-3 (data not shown). These data suggested that overexpression of E. coli fabH in Xcc EcH could increase the rate of fatty acid biosynthesis to support growth of the mutant. However, there are other possible causes for the weak growth of the fabH mutant, and these possibilities are being further investigated.

Replacement of Xcc fabH with E. coli fabH causes loss of Xcc ability to synthesize BCFAs. It
has been reported that replacement of S. coelicolor fabH with E. coli fabH leads to a dramatic change in the fatty acid profile of S. coelicolor 18 . Although E. coli fabH could replace Xcc fabH, it was unknown if replacement of Xcc fabH with E. coli fabH would lead Xcc to changes in the fatty acid profile. Thus, the fatty acid composition of strain Xcc T-3 grown in NYG medium was determined by gas chromatography-mass spectrometry (GC-MS) ( Fig. 4B and Table 2). The fatty acid profiles for wild type strain Xcc 8004 consisted of iso-BCFAs, anteiso-BCFAs, straight-chain saturated fatty acids, and straight-chain unsaturated fatty acids ( Fig. 4B and Table 2). The iso-BCFAs (46.66 ± 4%) were the predominant fatty acids, including iso-C 15:0 (24.2 ± 0.74%) and iso-C 17:1 (10.68 ± 2.85%). The anteiso-BCFAs comprised 17.6 ± 0.48%, mainly of anteiso-C 15:0 (16.77 ± 0.74%). The predominant straight-chain fatty acid was n-C 16:1 (15.8 ± 0.82%). These results were consistent with a previous report 24 . However, the fatty acid composition of strain Xcc T-3 was very different from Xcc 8004, being comprised of more than 99% straight-chain fatty acids with only trace amounts of BCFAs (< 1%). The predominant fatty acids were n-C 16:1 (44.44 ± 2.61%), n-C 16:0 (29.44 ± 1.27%), and n-C 18:1 (20.11 ± 1.72%) ( Fig. 4B and Table 2). We also determined the fatty acid profiles of strain Xcc EcH grown in the same medium ( Fig. 4B and Table 2). The species and amounts of fatty acids of Xcc EcH were almost the same as those of Xcc T-3, in which the predominant fatty acids (> 99%) were also straight-chain fatty acids, with only trace amounts (< 1%) of BCFAs. These data indicated that Xcc FabH was a key enzyme in the biosynthesis of BCFAs by Xcc.
To confirm that the BCFA synthesis defect in strain Xcc T-3 was due to the absence of the Xcc fabH gene, complementation analyses were performed. A plasmid, pYYH-1, in which Xcc fabH was placed under the control of the isopropyl β -D-1-thiogalactopyranoside-inducible P lac promoter was introduced into strain Xcc T-3. Expression of Xcc fabH in strain Xcc T-3 (strain Xcc YH4) reestablished the synthesis of BCFAs. However, as shown in Table 2, strain Xcc YH4 synthesized a smaller amount of BCFAs than the wild-type strain 8004, likely due to the activity of E. coli FabH for acetyl-CoA, which was higher than that of Xcc FabH.
To further investigate the function of Xcc FabH in BCFA synthesis, the fatty acid profiles of strain Xcc T-3 grown in SXFM medium were determined by GC-MS (Table S4). As seen for cultures grown in NYG medium, strain Xcc T-3 grown in SXFM medium did not produce any detectable BCFAs and contained almost the same species of fatty acids as those grown in NYG medium. Moreover, the fatty acid composition of strain Xcc EcH grown in SXFM medium was not significantly different from that grown in NYG medium (Table S4). However, although wild type strain Xcc 8004 also had a fatty acid profile similar to that grown in NYG medium, the total amount of BCFAs decreased, and there was a switch between the two species of BCFAs, with anteiso-BCFA becoming the major BCFA species (Table S4). These results indicated that the medium composition influences the fatty acid profiles of Xcc.
We also tested the effect of temperature on the fatty acid composition of Xcc strains grown in SXFM (Table  S4). As the temperature dropped from 30 to 15 °C, the wild-type strain 8004 significantly increased the production of n-C 16:1 fatty acids from 25.54 to 40.44% of the total fatty acids, which resulted in an increase in the total unsaturated fatty acid content from 36.25 to 49.75%. At the same time, the total BCFA content of strain 8004 decreased from 38.34 to 26.37%, especially the anteiso-BCFAs, which decreased from 23.34 to 14.83%. These data suggested that the wild-type Xcc strain increases the unsaturated fatty acid content, not the anteiso-BCFA content, in response to low temperatures. Low temperatures also affected the fatty acid composition of strains Xcc EcH and Xcc T-3 (Table S4). At low temperatures, the Xcc EcH strain also increased its total unsaturated fatty acid content. Although strain Xcc T-3 did not change its total unsaturated fatty acid content significantly, its n-C 16:1 fatty acid content increased.
Replacement of Xcc fabH with E. coli fabH abrogates the ability of Xcc to produce the 11-Me-C 12 :Δ 2 signal. Previous in vitro studies showed that the precursor of C 12 :Δ 2 was 3-hydroxydodecanoyl-acyl-ACP,   and that inhibition of fatty acid biosynthesis using cerulenin, an antibiotic that specifically binds to long-chain 3-keto-acyl-ACP synthases (FabF and FabB), caused Xcc to drastically reduce the production of the DSF family signals 8 . These results suggested that the precursors of DSF family signal synthesis come from the fatty acid synthetic pathway. If our hypothesis is correct, the mutant strains, Xcc T-3 and Xcc EcH, should not produce the 11-Me-C 12 :Δ 2 signal, because both strains have lost the ability to synthesize BCFAs. To test this hypothesis, we first examined the DSF family signals produced in strains Xcc T-3 and Xcc EcH by a bioassay using Xcc 8523 39 as a reporter strain. Both Xcc T-3 and Xcc EcH produced DSF signals (Fig. 5A), indicating that both Xcc T-3 and Xcc EcH possessed the ability to produce DSF signals. The active fractions of DSF signals produced by mutant strains were separated by high performance liquid chromatography (HPLC). After incubation in NYG medium for 24 hours, the wild type strain 8004 produced mainly 11-Me-C 12 :Δ 2 (0.97 μ M) and C 12 :Δ 2 (0.23 μ M), two DSF signals, while both Xcc EcH and Xcc T-3 strains only produced the C 12 :Δ 2 signal (0.83 μ M and 0.26 μ M, respectively) (Fig. 5B). When the incubation time was increased to 36 hours, both 11-Me-C 12 :Δ 2 and C 12 :Δ 2 , DSF signals, produced by the wild type strain 8004, declined to trace amounts, and the concentration of C 12 :Δ 2 signal in the culture of strain Xcc EcH decreased to 0.15 μ M (Fig. 5B). These results were consistent with a previous report that DSF signals reached a maximum at stationary phase and decreased drastically when bacterial numbers declined 32 . However, at the same time, the amount of C 12 :Δ 2 in the culture of strain Xcc T-3 was increased to 0.6 μ M (Fig. 5B), and this was probably due to the low growth rate of Xcc T-3, which caused the strain Xcc T-3 to enter the stationary phase later than did the wild type strain or Xcc EcH strain. Moreover, the data also showed that during these conditions, neither mutant strain produced the 11-Me-C 12 :Δ 2 signal. These observations confirmed that the precursors of DSF signals come from the intermediates of the fatty acid synthesis pathway.

Replacement of Xcc fabH with E. coli fabH leads to reduced Xcc virulence in plants.
Previous studies confirmed that membrane BCFAs play a key role in virulence regulation of L. monocytogenes, which included anteiso-BCFA-enhanced bacterial resistance against phagosomal killing by macrophages and modulation of the production of the critical virulence factor, and listeriolysin O 40,41 . Xcc wild type strain 8004 has the ability to synthesize BCFAs, whereas replacement of Xcc fabH with E. coli fabH in the mutant strains, Xcc EcH or Xcc T-3, results in the inability to synthesize BCFAs. To investigate the role of BCFAs in virulence regulation of Xcc, the pathogenesis of these mutant strains on plants was tested. A leaf clipping virulence assay using Chinese radish was conducted. The average lesion length caused by the wild type strain 8004 on a leaf of Chinese radish was 12.4 mm after 2 weeks inoculation (Fig. 6A). Replacement of Xcc fabH with a single copy of E. coli fabH in strain Xcc T-3 resulted in a significantly reduced average lesion length (1.7 mm) (Fig. 6A). This suggested that a lack of BCFAs seems to cause a reduction of Xcc plant pathogenesis. However, the average lesion length on Chinese radish leaves caused by strain Xcc EcH, a strain that carried an E. coli fabH encoding plasmid with a chromosomal fabH in-frame deletion, was 7.1 mm (Fig. 6A), which was shorter than that caused by the wild type strain 8004, but longer than that caused by strain Xcc T-3. Although both mutant strains lost the ability to produce BCFA and had a similar fatty acid profile, the growth rate of Xcc T-3 was lower than that of strain Xcc EcH. Thus, we speculate that reduction of pathogenesis to Chinese radish leaves caused by Xcc T-3 was mainly due to the lower growth rate of Xcc T-3, and not due to the lack of BCFAs. To confirm this hypothesis, we determined the growth of these mutant strains in fully mature Chinese cabbage (Wongbok) extracts. The data showed that the growth of strain Xcc T-3 was significantly weaker than that of the wild type strain, and the growth of Xcc EcH was  faster than that of Xcc T-3, but was still slower than that of the wild type strain (Fig. S6A). We also determined the CFU in Chinese cabbage after inoculation with the mutant strains. The CFU in Chinese cabbage infected by Xcc T-3 was lower than that of those infected by the wild type strain or Xcc EcH during 6 days after inoculation (Fig. S6B). Another possible cause could be the difference of the total amount of DSF signals produced among the Xcc strains. Although it has been reported that 11-Me-C 12 :Δ 2 and C 12 :Δ 2 have almost the same effects on the induction of pathogenesis toward plants 8 , the total amount of DSF signals produced by the Xcc T-3 and Xcc EcH strains was lower than that produced by the wild-type Xcc strain 8004 (Fig. 5B). We also evaluated several pathogenicity-related virulence factors produced by mutant strains. The activity of extracellular enzymes (including cellulase, amylase, and protease) was first tested. Substitution of Xcc fabH with a single copy of E. coli fabH in strain Xcc T-3 resulted in a significant decrease in the production of extracellular cellulase, amylase, and protease (Fig. 6B). The levels of cellulase, protease, and amylase produced by Xcc T-3 were only 74%, 60%, and 49%, respectively, of that produced by the wild-type strain. Although Xcc EcH slightly reduced the production of extracellular cellulase, amylase, and protease, the observed decreases were not statistically significant (Fig. 6B). Next, extracellular polysaccharide (EPS) production by mutant strains was tested. The amounts of EPS produced by wild type 8004, mutant strain, Xcc EcH or Xcc T-3 were 7.1 mg/mL, 5.9 mg/mL, and 2.0 mg/mL, respectively. The EPS produced by Xcc T-3 was significantly lower than that produced by wild-type strain 8004, but the amount of EPS produced by Xcc EcH was not statistically different from that produced by wild-type strain 8004 (Fig. 6C). These data also support our hypothesis that although replacement of Xcc fabH with E. coli fabH affects the pathogenesis in host plants, BCFAs do not play a key role in the regulation of Xcc virulence.

Discussion
The 3-ketoacyl-ACP synthase III, FabH, condenses acyl-CoAs with malonyl-ACP to initiate fatty acid biosynthesis in the type II fatty acid synthase systems of bacteria 1,2,4 . Xcc produces BCFAs, which account for approximately 50% of the total fatty acids of the Xcc cell. Bioinformatics analyses showed that Xcc FabH is highly identical to the FabH proteins from straight-chain fatty acid-producing Gram-negative bacteria, such as E. coli or R. solanacearum, but has a low amino acid sequence homology with the FabH proteins from BCFA synthesizing Gram-positive bacteria, such as B. subtilis or L. monocytogenes. This suggests that Xcc FabH prefers to select acetyl-CoA as a primer to initiate straight-chain fatty acid synthesis rather than to produce BCFAs. However, in the present study, we found that the Xcc fabH gene was able to restore the growth of the R. solanacearum fabH deletion mutant strain, and it caused this strain to produce certain BCFAs. In vitro assays demonstrated that Xcc FabH condensed branched-chain acyl-CoAs with malonyl-ACP to initiate fatty acid biosynthesis. Moreover, substitution of Xcc fabH with E. coli fabH resulted in the loss of Xcc ability to synthesize BCFAs. These results demonstrated that Xcc FabH is a key enzyme required for Xcc to synthesize BCFAs and provided the first evidence that in BCFA-producing Gram-negative bacteria, FabH proteins also prefer to use branched-chain acyl-CoAs as substrates to initiate fatty acid biosynthesis (Fig. 7A).
The fabH is an essential gene for growth of Xcc and cannot be disrupted from its genome, suggesting that fatty acid biosynthesis is a key metabolic pathway for Xcc. However, Xcc does not have an obligatory growth requirement for BCFAs. First, the mutant strain Xcc T-3, in which Xcc fabH was replaced with the E. coli fabH gene, lost the ability to synthesize BCFAs, but could grow in either enriched or minimal media, even though it grew slower than the wild-type strain 8004. Second, exogenous iso-C 14:0 , iso-C 15:0 , or anteiso-C 15:0 did not restore the growth rate of this mutant. Third, the Xcc T-3 strain was still pathogenic toward Chinese radish leaves, even though the virulence of this strain was lower than that of the wild-type strain 8004. Moreover, we showed that the fatty acid profiles of another Xcc fabH mutant, Xcc EcH, which carried an E. coli fabH-encoding plasmid and a chromosomal fabH in-frame deletion, were similar to those of strain Xcc T-3. The Xcc EcH strain grew faster in either enriched or minimal media, and was more virulent toward a host plant than strain Xcc T-3. These data also demonstrate that BCFAs are not required for Xcc growth and pathogenesis.
The Xcc mutant strains Xcc T-3 and Xcc EcH produced more unsaturated fatty acids than did the wild-type strain 8004, indicating that, to some extent, unsaturated fatty acids were able to replace the function of BCFAs in maintaining the membrane fluidity of Xcc. In addition, we observed that Xcc responds to a decrease in the ambient temperature by increasing the proportion of unsaturated fatty acids, especially n-C 16:1 fatty acids, in the membrane lipids. This adaptive response markedly differs from that in L. monocytogenes, which mainly stimulates anteiso-BCFA biosynthesis to increase membrane disorder 40,41 . Although the mechanism of unsaturated fatty acid biosynthesis in Xcc has not been well studied, the Xcc genome encodes two key anaerobic, unsaturated fatty acid biosynthesis genes, fabA (XC_3651) and fabB (XC_3652), which could be co-transcribed in a fabAB operon, as well as an aerobic, unsaturated fatty acid biosynthesis gene, desA (XC_0171). These data suggest that Xcc has the ability to produce unsaturated fatty acids, and that unsaturated fatty acids can play a key role in the regulation of membrane fluidity.
Although it has been reported that the inhibition of fatty acid biosynthesis causes Xcc to drastically reduce the production of 11-Me-C 12 :Δ 2 and C 12 :Δ 2 signals 8 , there is still a lack of in vivo evidence to support the view that Xcc shunts intermediates from the fatty acid synthesis pathways to produce DSF signals. Our report provides direct in vivo evidence that the precursors of DSF signals come from the fatty acid synthesis pathway of Xcc (Fig. 7A). Substitution of Xcc fabH with E. coli fabH not only caused Xcc to fail to produce BCFAs, but it also abrogated the synthesis of the 11-Me-C 12 :Δ 2 signal.
Xcc FabH was found to utilize a wide range of substrates as the acyl-CoA primer in vitro, but it remains unclear whether Xcc FabH exhibits a wide range of substrate specificities in vivo. However, we speculate that Xcc FabH should possess this feature. In Xcc, RpfF has been reported to have a wide range of acyl-ACP thioesterase activities, with the ability to hydrolyze acyl-ACP intermediates of fatty acid synthesis pathways to produce free fatty acids 8,9 . RpfB is a fatty acyl-CoA ligase, which converts free fatty acids produced by RpfF to fatty acyl-CoAs.
It has been reported that long-chain fatty acyl-CoAs produced by RpfB are converted to phospholipids 42 . But it is unknown how Xcc utilizes the rest of the short-or medium-chain fatty acyl-CoAs. One possibility is that Xcc FabH condenses the short-or medium-chain fatty acyl-CoAs produced by RpfB with malonyl-ACP to produce new acyl-ACP intermediates, which re-enter the fatty acid synthetic cycle to synthesize long acyl-ACPs (Fig. 7B). This possibility is currently being studied in our laboratory.

Materials and Methods
Materials. Malonyl-CoA, acetyl-CoA, acyl-CoAs, fatty acids, NADH, NADPH, and antibiotics were purchased from Sigma-Aldrich. Takara Biotechnology Co. provided molecular biology reagents. Novagen provided the pET vectors. Ni-agarose columns were obtained from Invitrogen. GE healthcare provided the HiTrap Q strong anion-exchange column and Bio-Rad provided the Quick Start Bradford dye reagent. All other reagents were of the highest available quality. Takara Biotechnology Co. synthesized the oligonucleotide primers.
We also tested the growth of Xcc strains in cabbage extracts. The cabbage juice was prepared as follows. Mature Chinese cabbage (Wongbok) (0.5 kg) was minced using an electric juicer and the supernatant was collected by centrifugation at room temperature. Then, the supernatant was filtered using a membrane filtration system (0.22 μ m) to obtain the cabbage extracts subsequently used for growth analyses.
Complementation of the R. solanacearum fabH deletion strain. The Xcc fabH gene was amplified from genomic DNA of Xcc wild type strain 8004 using primers listed in Table S2. Then the amplified fragment was purified, digested with NdeI and HindIII, and cloned into the same sites of pSRK-Km 38 to get plasmid pYYH-1. Following the mating of derivatives of E. coli strain S17-1 carrying the pYYH-1 or empty vector, with R. solanacearum fabH deletion strain RsmH 35 on BG plates with octanoic acid (0.1%) for 48 hours at 30 °C, the cells were suspended in BG medium, and appropriate dilutions were inoculated onto BG plates (with octanoic acid) containing chloramphenicol (to select against the donor strain) plus kanamycin. The transformed strains were inoculated onto the BG plates with or without octanoic acid, and growth was determined after 2 days incubation at 30 °C.
Protein expression and purification. The Xcc fabH gene was cloned into pET-28b to yield plasmid pYYH-2. Xcc FabH with a vector-encoded His 6 -tagged N-terminus was expressed in E. coli BL21 (DE3), and purified with Ni-NTA agarose (Qiagen) using a nickel-ion affinity column (Qiagen). The purities of proteins were monitored by SDS-PAGE. The E. coli FabD, FabH, FabG, FabZ, and FabI, and E. coli holo-ACP proteins were purified as described previously 43

Spectrophotometric assay of 3-ketoacyl-ACP synthase activity. A continuous assay format was
used to monitor Xcc FabH activity with straight-and branched-chain acyl-CoA by coupling the condensing activity of Xcc FabH to 3-ketoacyl-ACP reductase (FabG) of E. coli as described previously 45 . The reaction mixture contained 100 μ M ACP, 0.5 mM acyl-CoA, 0.5 mM malonyl-CoA, 0.2 mM NADPH, 2 μ g of FabG, 2 μ g of purified E. coli FabD, and 0.1 M sodium phosphate buffer, pH 7.4, in a final volume of 100 μ L. The reaction was initiated by adding 1 μ g Xcc FabH to the mixture, and the 3-ketoacyl-ACP synthase III activities of Xcc FabH were determined by monitoring the rate of oxidation of NADPH at 340 nm using an extinction coefficient of 6,220 M −1 .

Disruption of the Xcc fabH gene.
To disrupt the Xcc fabH gene, a pK18mobsacB-borne in-frame deletion suicide plasmid, pZTT-1, was constructed according to Fig. S2A. The 500 bp DNA fragments flanking up or down the Xcc fabH gene were amplified with Pfu DNA polymerase using Xcc genomic DNA as the template, and either XcfabH1 EcoRI and XcfabH2 (for up fabH), or XcfabH3 and XcfabH4 HindIII (for down fabH), were used as primers (Table S2). The two fragments were purified and joined by overlapping PCR. The fused fragment was then digested with EcoRI and HindIII, and inserted between the same sites of pK18mobscaB 46 to give the plasmid pZTT-1. We also constructed a suicide plasmid, pYYH-3, in which a 550 bp inner DNA fragment of Xcc fabH was inserted into plasmid pK18mobsacB (Fig. S3). Following the mating of derivatives of E. coli strain S17-1 carrying suicide plasmid (pZTT-1 or pYYH-3) with Xcc 8004 on NYG plates for 36 hours at 30 °C, the 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 the integration of the non-replicating plasmid into the chromosome of the recipient. Single crossover integrants of pZTT-1 (i.e., Xcc YH1) (Fig. S2) were selected for further study, while no integrant of pYYH-3 was obtained (Fig. S3). Then the culture of Xcc YH1 was incubated at 30 °C for 36 hours, and after appropriate dilutions, the culture was inoculated onto NYG plates containing 10% sucrose. Colonies sensitive to kanamycin were screened by colony PCR with primers listed in Table S2. NYG medium with fatty acid supplementation 47 was also used to select Xcc fabH deletion mutants.

Replacement of Xcc fabH with E. coli fabH.
To replace Xcc fabH in the chromosome, the E. coli fabH gene (EcfabH) was amplified from E. coli MG1655 genomic DNA with Pfu DNA polymerase and the primers listed in Table 2S. The EcfabH gene was ligated into the T-vector plasmid pMD19 to yield pZTT-2. The NdeI-BamHI fragment of EcfabH was cloned into the same sites of pZTT-1 to yield suicide plasmid pZTT-3. E. coli strain S17-1 carrying plasmid pZTT-3 was conjugated with Xcc 8004 on NYG plates for 24 hours at 30 °C (Fig. S4A). After appropriate dilutions, the colonies, which were resistant to rifampicin and kanamycin, were selected, and one of these colonies was named Xcc YH2. Xcc YH2 was inoculated into NYG medium at 30 °C for 48 hours, and the cultures were spread onto NYG plates containing 10% sucrose. Colony PCR utilizing the primers listed in Table S2 was performed to screen colonies sensitive to kanamycin. We successfully obtained strain Xcc T-3 (Fig. S4B,C), in which Xcc fabH was replaced with E. coli fabH in the chromosome.
Another replacement of Xcc fabH with E. coli fabH strain, Xcc EcH, was also constructed as follows (Fig. S5A). First, EcfabH encoding plasmid pSRH-EcfabH was introduced into the Xcc fabH single crossover integrant Xcc YH1 to obtain Xcc YH5 by conjugation with E. coli S17-1 (Fig. S5A). The culture of Xcc YH5 was inoculated onto NYG plates containing 10% sucrose. Colonies sensitive to kanamycin were screened by colony PCR using the primers listed in Table S2. One colony, in which plasmid pSRK-EcfabH expressed E. coli fabH and Xcc fabH had been deleted from the chromosome, was obtained, and it was named Xcc EcH (Fig. S5B,C).

Analysis of fatty acid composition.
The cultures of bacteria were grown aerobically in different media for 2-4 days. Cells were harvested and washed three times with water. Fatty acid methyl esters were synthesized and extracted as described previously 24 . Briefly, cellular lipids were saponified by addition of 1 mL of sodium hydroxide/methanol solution at 100 °C for 40 minutes with shaking (800 rpm). The fatty acids were then methylated by addition of 2 mL of hydrochloric acid/methanol solution at 80 °C for 30 minutes, and immediately cooled to below 20 °C. Fatty acid methyl esters were obtained by three extractions each with 1 mL of petroleum ether. The solvent was removed under a stream of nitrogen, and the residue was analyzed by GC-MS.
Extraction and purification of DSF-family signal components from the Xcc culture supernatant.
The protocol for extraction and purification of DSF family components was described previously 48 . In brief, Xcc strains were cultured in liquid medium for 24-48 hours and 50 mL of bacterial supernatant was collected by centrifugation at 3,800 × g for 30 minutes at 4 °C. The pH of the supernatants was adjusted to 4.0 by adding hydrochloric acid prior to two extractions with an equal volume of ethyl acetate. The ethyl acetate fractions were collected and the solvent was removed by rotary evaporation to dryness at 40 °C. The residue was dissolved in 1 mL of methanol. The crude extract was subjected to a 0.45 μ m Minisart filter unit and the collected filtrate was concentrated to 0.5 mL. Three microliters of the extract was injected into a C18 reverse-phase HPLC column (4.6 × 150 mm, Agilent Technologies), eluted with water in methanol (23:77, v/v, respectively; 0.1% formic acid) at a flow rate of 1 mL/minute in an Agilent Technologies 1260 Infinity system with a DAD G1315D VL detector.
Pathogenicity tests. Virulence was tested on potted Chinese radish (Raphanus sativus L. var. radiculus Pers.) as described previously 49 . Bacterial growth in mature Chinese cabbage (Wongbok) was tested as follows. Bacterial cells were grown in NYG medium at 30 °C for 24 h, then collected by centrifugation and washed twice with sterile saline. The cell pellets were suspended in sterile saline to a final OD 600 of 0.05. Cabbage stem discs were sampled using a cork borer (diameter, 14.25 mm), and the discs were immersed in bacterial suspension for 2 h. After washing twice with sterile water, the discs were kept in empty plates at 25 °C, and ground with a pestle in 1 ml sterile water at day 0, 3, 6, and 9 after inoculation. The homogenates were serially diluted, and dilutions were plated on NYGA supplemented with rifampicin. Bacterial colony-forming units were counted after incubation at 30 °C for 3 days.

Measurement of extracellular enzymatic activity and EPS production.
Relative activities of extracellular enzymes were assayed as described previously 50 . Two μ L of each Xcc strain culture (OD 600 ≈ 1.0) was spotted onto NYG agar plates containing 1% (w/v) skim milk (for protease), 0.5% (w/v) carboxymethylcellulose (for cellulase), or 0.1% (w/v) starch (for amylase) and incubated at 30 °C for 24-48 hours. Plates were stained where necessary as previously described 50 . Zones of clearance around the spot due to the degradation of the substrate were photographed. Three plates were inoculated in each experiment, and each experiment was repeated three times. The relative activity of the enzyme was indicated by the diameter of the clear zone.
The EPS production was measured as described previously 51 . Each Xcc strain culture (2 mL, OD 600 ≈ 1.0) was used to inoculate 100 mL of NYG containing 4% glucose in a 250 mL flask and kept at 30 °C with shaking at 180 rpm for 4 days. The EPS was precipitated from the culture supernatant by addition of 4 volumes of ethanol. The pelleted EPS was washed with 70% ethanol, air dried, and weighed. Three flasks were inoculated in each experiment and each experiment was repeated three times.