Novel Xanthomonas campestris Long-Chain-Specific 3-Oxoacyl-Acyl Carrier Protein Reductase Involved in Diffusible Signal Factor Synthesis

ABSTRACT The precursors of the diffusible signal factor (DSF) family signals of Xanthomonas campestris pv. campestris are 3-hydroxyacyl-acyl carrier protein (3-hydroxyacyl-ACP) thioesters having acyl chains of 12 to 13 carbon atoms produced by the fatty acid biosynthetic pathway. We report a novel 3-oxoacyl-ACP reductase encoded by the X. campestris pv. campestris XCC0416 gene (fabG2), which is unable to participate in the initial steps of fatty acyl synthesis. This was shown by the failure of FabG2 expression to allow growth at the nonpermissive temperature of an Escherichia coli fabG temperature-sensitive strain. However, when transformed into the E. coli strain together with a plasmid bearing the Vibrio harveyi acyl-ACP synthetase gene (aasS), growth proceeded, but only when the medium contained octanoic acid. In vitro assays showed that FabG2 catalyzes the reduction of long-chain (≥C8) 3-oxoacyl-ACPs to 3-hydroxyacyl-ACPs but is only weakly active with shorter-chain (C4, C6) substrates. FabG1, the housekeeping 3-oxoacyl-ACP reductase encoded within the fatty acid synthesis gene cluster, could be deleted in a strain that overexpressed fabG2 but only in octanoic acid-supplemented media. Growth of the X. campestris pv. campestris ΔfabG1 strain overexpressing fabG2 required fabH for growth with octanoic acid, indicating that octanoyl coenzyme A is elongated by X. campestris pv. campestris fabH. Deletion of fabG2 reduced DSF family signal production, whereas overproduction of either FabG1 or FabG2 in the ΔfabG2 strain restored DSF family signal levels.

In X. campestris pv. campestris, a cluster of genes designated rpfABCDEFG (rpf stands for regulation of pathogenicity factors) is involved in the biosynthesis, perception, transduction, and turnover of DSF family signaling molecules (2,8,9). The synthesis of DSF family signaling molecules is dependent on RpfF, an enoyl-acyl carrier protein (enoyl-ACP) hydratase/thioesterase. RpfF is a bifunctional enzyme that not only catalyzes the dehydration of 3-hydroxyacyl-ACPs to cis-2-enoyl-ACPs but also cleaves the acyl-ACP thioester bonds to produce free fatty acids (6, 10) (Fig. 1A). The complex pathway that regulates pathogenicity is beyond the scope of this report, and thus readers are referred to recent reviews (1,2).
Bacteria utilize primarily a disassociated fatty acid synthase system for de novo production of fatty acids (11,12). The flexible nature of this system allows the diversion of intermediates to other end products, including lipid A (13,14), quorum-sensing signal molecules (15,16), and vitamin cofactors (17,18). The X. campestris pv. campestris genome contains all of the genes known to be required for fatty acid synthesis, although the synthesis mechanism has received little study. The precursors of the DSF family signals are 12-or 13-carbon 3-hydroxyacyl-ACP molecules (6,10) derived by 3-oxoacyl-ACP reductase (OAR)-catalyzed reduction of 3-oxoacyl-ACPs (Fig. 1B). Overexpression of FabG1 led to a significant increase in the production of DSF family signals (6).
The X. campestris pv. campestris genome carries four putative OAR genes: fabG1 (XCC1018), fabG2 (XCC0416), fabG3 (XCC4003), and fabG4 (XCC0384) (Fig. 1C). XCC1018 (fabG1) is located within a cluster of fatty acid synthesis genes, and 69.1% of the residues of the FabG1 protein are identical to those of Escherichia coli FabG. The active-site triad (Ser, Tyr, and Lys) and the N-terminal cofactor-binding sequence defined by the E. coli FabG X-ray crystal structures are conserved in X. campestris pv. campestris FabG1 (19,20) (Fig. 1C). Therefore, given these motifs, together with its genome location, FabG1 was considered to play the major role in the reduction of 3-oxoacyl-ACPs for the synthesis of the phospholipid fatty acyl chains. In contrast, FabG3 seems involved in the biosynthesis of xanthomonadin polyketides (21). XCC0384 (fabG4) is located in a putative biotin synthesis operon but contains neither the conserved catalytically active triad nor the N-terminal cofactor-binding sequence and thus seems unlikely to have OAR activity (Fig. 1C).
The remaining OAR candidate, FabG2, is encoded by a lone gene located far from the above-mentioned genes. Although alignments showed that FabG2 is only 32.4% identical to E. coli FabG, it contains the typical active-site triad and the N-terminal cofactor-binding sequence (Fig. 1C). Based on these data, it was reasonable to hypothesize that fabG2 encodes a functional 3-oxoacyl-ACP reductase. However, given that of FabG1, the role of FabG2 seemed unlikely to be involved in bulk fatty acid synthesis. A possible FabG2 role is DSF synthesis. We report that FabG2 is a novel OAR that specifically reduces long-chain substrates.

RESULTS
fabG2 encodes an OAR of novel specificity. To determine whether FabG2 has OAR activity, we transformed the E. coli fabG(Ts) strain CL104 with a plasmid that expressed FabG2 under arabinose control and assayed growth at 42°C. Strain CL104 lacks 3-oxoacyl-ACP reductase activity at 42°C and is unable to grow at that temperature (22). As a control, we similarly expressed the housekeeping OAR, FabG1, with the expectation that it would allow growth of strain CL104 at 42°C, and that was the case ( Fig. 2A). In contrast, strain CL104 expressing FabG2 failed to grow at 42°C either in the presence or the absence of arabinose induction ( Fig. 2A), and thus FabG2 seemed to lack OAR activity. However, because overexpression of FabG2 in the wild-type X. campestris pv. campestris strain Xc1 increased DSF production (see below), it seemed that FabG2 might specifically reduce 3-oxoacyl-ACPs to provide substrates for DSF synthesis and be unable to reduce short-chain 3-oxoacyl-ACPs. If so, when provided with a sufficiently long 3-oxoacyl-ACP substrate, FabG2 should functionally replace E. coli FabG. This hypothesis was tested by expressing both FabG2 and the AasS acyl-ACP synthetase (23) in E. coli CL104 and testing for growth at 42°C on plates containing octanoic acid. In this scenario, AasS converted exogenous octanoic acid to octanoyl-ACP, which was elongated to 3-oxodecanoyl-ACP. FabG2 then reduced this product to 3-hydroxydecanoyl-ACPs to allow synthesis of the fatty acids required for E. coli growth. Under these circumstances, and production of both enzymes was required (Fig. 2B).
In a second approach, we expressed both FabG2 and X. campestris pv. campestris FabH in E. coli CL104 and tested growth in the presence or absence of octanoic acid at 42°C. In this second scenario, X. campestris pv. campestris FabH condensed octanoyl coenzyme A (octanoyl-CoA) with malonyl-ACP to produce 3-oxodecanoyl-ACP (24), which FabG2 reduced to 3-hydroxydecanoyl-ACP, as described above. Fatty acid synthesis was thus primed, and growth at 42°C proceeded. Both enzymes were required (Fig. 2B). These results argued strongly that FabG2 is a 3-oxoacyl-ACP reductase that specifically reduces long-chain substrates.
To test whether a specific fatty acid chain length was required to support growth, plates supplemented with 100 g/ml of straight-chain saturated fatty acids with chain lengths of C 4 to C 16 were tested ( Fig. 2B and C). Only the C 6 and C 8 fatty acids supported the growth of derivatives of strain CL104 that expressed FabG2 plus either AasS or X. campestris pv. campestris FabH ( Fig. 2B and C). The failure of butyric acid to support growth can be attributed to the inability of AasS to use this substrate (16) and/or the weak activity of FabG2 with 3-oxohexanoyl-ACP observed in vitro (see below). Longer fatty acids (ϾC 8 ) failed to support the growth of ϾC 8 acids because they feed into the pathway past the branch point for unsaturated fatty acid synthesis (see below).

FabG2 preferentially reduces long-chain 3-oxoacyl-ACPs in vitro.
Recombinant hexahistidine-tagged FabG2 was expressed in E. coli and purified to homogeneity (see Materials and Method). Purified FabG2 had the size expected from the sequence of the tagged protein (26.4 kDa) (see Fig. S1A in the supplemental material). Size exclusion chromatography indicated that FabG2 is a multimer (trimer or tetramer) in solution (Fig. S1B). , and E. coli FabG proteins were expressed from plasmids pHZ003, pHZ004, and pTWH21, respectively, which were derived from using the compatible vectors pBAD24M and pBAD33. To allow entry of exogenous fatty acids into the E. coli fatty acid synthesis pathway, a plasmid encoding either AasS (pYFJ86) or X. campestris pv. campestris FabH (pYYH56) was used to obtain the results shown in panel B or C, respectively. "Vectors" denotes the empty vectors. No Ara, without arabinose induction; Plus Ara, arabinose induction. The medium was RB agar. (B) Growth at 42°C of E. coli strain CL104 carrying the plasmids expressing FabG2 or E. coli fabG in the presence or absence of AasS expression on RB plates supplemented with various fatty acids (see Materials and Methods). No FA, no fatty acid supplementation; C4, butyric acid supplementation; C6, hexanoic acid supplementation; C8, octanoic acid supplementation; C10, decanoic acid supplementation; C12, dodecanoic acid supplementation; C14, tetradecanoic acid supplementation; C16, hexadecanoic acid supplementation. The lack of growth on ϾC 8 fatty acids is because their chain lengths are past the C 8 to C 10 branch points for synthesis of the unsaturated fatty acids required for membrane function. (C) Growth of E. coli strain CL104 containing a plasmid expressing FabG2 or E. coli FabG as in panel B, except that the expressed octanoate entry enzyme was X. campestris pv. campestris FabH in place of AasS. Fatty acid supplementation was as described for panel B.
Hu et al.
Overexpression of FabG2 plus supplementation with octanoic acid allows deletion of the fabG1 gene. The physiological functions of FabG2 were tested by disruption of fabG1 and fabG2 using suicide plasmids carrying in-frame gene deletions (Fig. S2). A ⌬fabG2 deletion strain was readily generated (Fig. S2C), but no fabG1 deletion strain could be isolated. Only the single-crossover integrant strain HZ1 was obtained (Fig. S2), which indicated that X. campestris pv. campestris fabG1 is essential. However, since FabG2 restored E. coli CL104 growth in the presence of exogenous octanoic acid, we plated the fabG1 single-crossover integrant (strain HZ1) on medium containing octanoic acid to allow the second crossover to give a fabG1 deletion, but this failed. Arguing that differential expression levels of the two genes might explain this failure, we measured their transcription and found that fabG1 transcription was 5to 7-fold higher than that of fabG2 (data not shown). Given these data, we overexpressed FabG2 using the vector pSRK-Gm (25) in the single-crossover strain and selected for growth in the presence of octanoic acid, which produced the ΔfabG1 strain HZ6 (ΔfabG1/pfabG2) (Fig. S2F).
Deletion of fabG2 did not affect growth on NaCl-yeast extract-glycerol (NYG) plates, whereas the ΔfabG1/pfabG2 strain that overproduced FabG grew only when the plates contained octanoic acid ( Fig. 4AB; Fig. S3A). Our finding that the ⌬fabG1/pfabG2 strain (HZ6) grew when provided with octanoic acid or (less so) with hexanoic acid in the absence of AasS expression argued that X. campestris pv. campestris contained an enzyme that converted the C 6 and C 8 acids to the ACP thioesters required to enter the fatty acid synthesis pathway. In Pseudomonas aeruginosa PA3286a, novel 3-oxoacyl-ACP synthase III condenses malonyl-ACP with ␤-oxidation-derived acyl-CoAs of mediumchain lengths (C 6 to C 8 ) to produce longer-chain 3-oxoacyl-ACPs that prime fatty acid synthesis (26). Since in vitro X. campestris pv. campestris FabH uses octanoyl-CoA in place of octanoyl-ACP (24) and X. campestris pv. campestris encodes two acyl-CoA synthetases, RpfB and FadD (XCC1017), X. campestris pv. campestris may have an enzyme functionally analogous to PA3286. This was tested by use of a ⌬fabG2 derivative of a strain in which E. coli FabH replaced X. campestris pv. campestris FabH (24). E. coli FabH cannot accept octanoyl-CoA (27), and hence this strain is unable to incorporate labeled octanoate into long-chain fatty acids. In contrast, X. campestris pv. campestris strains expressing X. campestris pv. campestris FabG2 as the sole FabG elongated [1-14 C]octanoic acid, but not [1-14 C]acetic acid, whereas the X. campestris pv. campestris ΔfabH strain expressing E. coli FabH elongated only [1-14 C]acetic acid. The wild-type strain Xc1 elongated both precursors (Fig. 5). Hence, X. campestris pv. campestris FabH is responsible for the entry of octanoic acid into the long-chain fatty acid synthesis pathway.
Growth of the ΔfabG1/pfabG2 strain on plates supplemented with other fatty acids (Fig. 4B) was also tested. As seen when E. coli CL104 complemented with FabG2 was tested (Fig. 2), fatty acids (ϾC 8 ) were unable to support growth. Presumably, this was due to a lack of unsaturated fatty acid synthesis. Indeed, when the ΔfabG1/pfabG2 strain was plated with decanoic acid and oleic acid supplementation, the strain grew (Fig. S4B) and both fatty acids were required. Hence, the lack of unsaturated fatty acid synthesis is indeed responsible for the inability of fatty acids (ϾC 8 ) to support the growth of the ΔfabG1/pfabG2 strain.
To test whether FabG2 has long-chain 3-oxoacyl-ACP reductase activity in its native bacterium, we analyzed the fatty acid compositions of strain HZ6 (ΔfabG1/pfabG2) and   (Table S3). The species of fatty acids produced by strain HZ6 (ΔfabG1/pfabG2) were essentially the same as those produced by the wild-type strain Xc1 grown in NYG liquid medium.
Deletion of fabG2 resulted in reduced production of DSF family signaling molecules. 3-Hydroxyacyl-ACPs of 12 or 13 carbon atoms are the precursors of the DSF family signals (Fig. 1) (6, 10). To determine whether FabG2 preferentially produces such substrates, we assayed the production of DSF family signals in the ΔfabG2 mutant stain grown to stationary phase in NYG medium using high-performance lipid chromatography. The production of both DSF and BDSF by the ΔfabG2 mutant stain (HZ3) was Ͻ50% of the production of wild-type strain Xc1 (Fig. 6A). Upon complementation with a plasmid expressing wild-type FabG2, the ΔfabG2 strain increased its production of both DSF family signals (Fig. 6A), implying that FabG2 is involved in DSF family signal production. However, complementation with a plasmid overexpressing FabG1 also restored DSF family signal production to the ΔfabG2 strain to levels similar to those produced by the strain overexpressing FabG2 (Fig. 6A). Hence, although FabG2 has a significant role in DSF family signal synthesis, it is not the sole source of 3-hydroxydodecanoyl-ACPs. Indeed, overexpression of each FabG in wild-type strain Xc1 gave DSF family signals levels 50% higher than the Xc1levels (Fig. 6B). Hence, the level of 3-oxoacyl-ACP reductase activity rather than of the specific OAR is the important parameter in the production of DSF family signals.

DISCUSSION
In the present study, we identified FabG2 as a novel OAR that specifically reduces long-chain 3-oxoacyl-ACPs to 3-hydroxyacyl-ACPs. Unlike FabG1, FabG2 cannot replace black columns, production in the ΔfabG2 strain; gray columns, production in the ΔfabG2 strain carrying a plasmid encoding ΔFabG2 (pHZ009); stippled columns, production in the ΔfabG2 strain carrying a plasmid encoding FabG1 (pHZ013). (B) Molecules produced by strain Xc1derivatives. White columns, production in the strain carrying the vector pSRK-Gm (25); black columns, production in the strain overexpressing FabG2 (pHZ009); gray columns, production in the strain overexpressing FabG1 (pHZ013). Error bars, means Ϯ standard deviations (n ϭ 3). *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001, assessed with one-way analysis of variance (ANOVA). All experiments were repeated three times with similar results. Relative amounts of signal molecules were calculated based on peak areas of the detector response. Note that the amounts and relative levels of the various DSF signaling molecules are greatly affected by the medium and culture conditions (24), and hence only comparative levels are given.
E. coli FabG in the general fatty acid synthesis pathway. However, in the presence of an enzyme that allows exogenous fatty acids to enter the fatty acid synthetic pathway (either AasS or X. campestris pv. campestris FabH) and exogenous octanoic acid, growth was allowed at the nonpermissive temperature. Moreover, in the native bacterium, fabG1 could be deleted only when FabG2 was overexpressed and the medium contained octanoic acid. These observations argued that the failure of FabG2 to perform all of the 3-oxoacyl-ACP reductions required for general fatty acid synthesis was due to the strain's absent or weak ability to reduce the first 3-oxoacyl-ACP of the pathway, 3-oxobutyryl-ACP. Indeed, in vitro FabG2 was only weakly active with short-chain 3-oxobutyryl-ACP and 3-oxohexanoyl-ACP but readily reduced long-chain 3-oxoacyl-ACPs, with the 10-carbon substrate being the most active.
Although X. campestris pv. campestris FabG2 preferentially reduces long-chain 3-oxoacyl-ACP substrates and deletion of fabG2 decreases the ability of X. campestris pv. campestris to produce DSF family signals, FabG2 is not required for the production of DSF family signals. Indeed, overexpression of either FabG2 or FabG1 in the wild-type strain Xc1 significantly increased the production of the DSF family signaling molecules. FabG1 is the housekeeping X. campestris pv. campestris OAR and is required for normal X. campestris pv. campestris growth, although fabG1 can be deleted from the X. campestris pv. campestris genome provided that FabG2 is overexpressed in the presence of exogenous octanoic acid. Therefore, it seems that the role of FabG2 is to maintain a sufficiently high level of OAR activity for DSF family signal production.

Materials.
Moravek supplied the radioactive precursors. Sigma-Aldrich provided cis-11-methyl-2-dodecenoic acid and cyclic-di-GMP. Ni-agarose columns were from Invitrogen. Agilent Technologies provided HC-C18 high-performance liquid chromatography (HPLC) columns. All other reagents were of the highest available quality. Sangon Biotechnology Co. synthesized the oligonucleotide primers.
Bacterial strains, plasmids, and growth conditions. The strains, plasmids, and primers used in this study are listed in Table S1 in the supplemental material. Luria-Bertani (LB) medium was used as the rich medium for E. coli growth at 37°C. Escherichia coli fabG(Ts) mutant strain CL104 was grown in RB medium (10 g/liter tryptone, 10 g/liter NaCl, and 1 g/liter yeast extract) (LB with one-fifth yeast extract) at 30°C (22). The X. campestris pv. campestris strains were grown in NYG medium (in grams per liter, peptone, 5; yeast extract, 3; and glycerol, 20 [pH 7.0]). Where required, antibiotics were added at the following concentrations: 100 g/ml sodium ampicillin, 30 g/ml kanamycin sulfate, 30 g/ml (for E. coli) or 10 g/ml (for X. campestris pv. campestris) gentamicin sulfate, and 50 g/ml rifampin. L-Arabinose was used at a final concentration of 0.01%. Isopropyl-␤-D-thiogalactoside (IPTG) was used at a final concentration of 1 mM. Bacterial growth in liquid medium was determined by measuring the optical density at 600 nm (OD 600 ) using a Bioscreen-C automated growth curve analysis system (OY Growth Curves).
Recombinant DNA techniques and construction of plasmids. The fabG1 and fabG2 PCR products were amplified from X. campestris pv. campestris strain Xc1 genomic DNA using Pfu DNA polymerase, and the primers given in Table S2 and were inserted into the T-vector plasmid pMD19 to produce plasmids pHZ001 (fabG1) and pHZ002 (fabG2). To produce plasmids pHZ003 (fabG1), pHZ004 (fabG2), pHZ005 (fabG1), and pHZ006 (fabG2), the T-vector pMD19 fab gene plasmids were digested with NdeI and HindIII and ligated with pBAD24M (28) or pET-28(b) digested with the same enzymes. The ΔfabG1 and fabG2 deletion mutant strains were constructed essentially as described previously (29). Construction details are given in the legend to Fig. S3.
Expression and purification of plasmid-encoded proteins. The pET28b(ϩ)-derived plasmids carrying the various fabG genes were introduced into E. coli strain BL21(DE3), and the encoded proteins were expressed at high levels and purified as described previously. The enzymes were confirmed to be homogeneous using SDS-PAGE. E. coli FabD, FabH, FabZ, and FabI, Vibrio harveyi AasS, and E. coli holo-ACP proteins were purified as described previously (28). The solution structures of FabG1 and FabG2 were analyzed with size exclusion chromatography on a Superdex 200 10/300 GL column (GE Healthcare) using a model 10 AKTA purifier at 0.45 ml/min in phosphate running buffer (135 mM NaCl, 2.7 mM KCl, 1.5 mM Na 2 HPO 4 , 8 mM K 2 HPO 4 , 10% glycerol, pH 7.4), and the standards used have been described previously (30).
Assay of FabG1 and FabG2 activities in vitro. Malonyl-ACP was synthesized from holo-ACP and malonyl-CoA with E. coli FabD. Acyl-ACPs (C 6 ACP to C 14 ACP) were synthesized from fatty acids, ATP, and E. coli holo-ACP with AasS, as described previously (23). The reaction products were resolved with conformationally sensitive gel electrophoresis on 20% or 17.5% polyacrylamide gel containing a urea concentration optimized for the separation. The gel was stained with Coomassie brilliant blue R250.
To verify the products of the FabG2-catalyzed reaction, the acyl-ACP derivatives were purified from 500 l of the above-described reaction mixture by the method of Zhao et al. (31). Their molecular masses Novel Xanthomonas campestris 3-Oxoacyl-ACP Reductase ® were determined with matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) MS (Bruker Autoflex III) as previously described (32).