Identification of a Biosynthetic Gene Cluster and the Six ... · mology used to synthesize these molecules is conserved. This conservation comes from NRPSs consisting of a set of

JOURNAL OF BACTERIOLOGY, Sept. 2007, p. 6312–6323 Vol. 189, No. 170021-9193/07/$08.00�0 doi:10.1128/JB.00725-07Copyright © 2007, American Society for Microbiology. All Rights Reserved.

Identification of a Biosynthetic Gene Cluster and the Six AssociatedLipopeptides Involved in Swarming Motility of Pseudomonas syringae

pv. tomato DC3000�†Andrew D. Berti,1,2 Nathan J. Greve,1 Quin H. Christensen,1‡ and Michael G. Thomas1,2*

Department of Bacteriology1 and Microbiology Doctoral Training Program,2 University ofWisconsin—Madison, Madison, Wisconsin 53706

Received 9 May 2007/Accepted 22 June 2007

Pseudomonas species are known to be prolific producers of secondary metabolites that are synthesized wholly orin part by nonribosomal peptide synthetases. In an effort to identify additional nonribosomal peptides produced bythese bacteria, a bioinformatics approach was used to “mine” the genome of Pseudomonas syringae pv. tomatoDC3000 for the metabolic potential to biosynthesize previously unknown nonribosomal peptides. Herein we describethe identification of a nonribosomal peptide biosynthetic gene cluster that codes for proteins involved in theproduction of six structurally related linear lipopeptides. Structures for each of these lipopeptides were proposedbased on amino acid analysis and mass spectrometry analyses. Mutations in this cluster resulted in the loss ofswarming motility of P. syringae pv. tomato DC3000 on medium containing a low percentage of agar. This phenotypeis consistent with the loss of the ability to produce a lipopeptide that functions as a biosurfactant. This work givesadditional evidence that mining the genomes of microorganisms followed by metabolite and phenotypic analysesleads to the identification of previously unknown secondary metabolites.

Nonribosomal peptide synthetase (NRPS) enzymology is in-volved in the biosynthesis of many natural products with di-verse biological activities, such as antifungal, antibacterial, an-ticancer, immunosuppressant, and metal chelation (reviewedin references 18 and 55). While the chemical structures of thenatural products biosynthesized by NRPSs are diverse andexplain their wide-ranging biological activities, the core enzy-mology used to synthesize these molecules is conserved. Thisconservation comes from NRPSs consisting of a set of repeat-ing core protein domains grouped into enzymatic modules,with each module typically controlling the incorporation of oneamino acid precursor into the nonribosomal peptide. Thestructural diversity of the NRPS-synthesized natural productscomes from variations in the number and order of the mod-ules, alterations in the substrate incorporated by the modules,and the potential addition of catalytic domains into modulesthat result in modifications to the growing peptide chain.

The core domains that are repeated for each NRPS moduleare the adenylation (A) domain, peptidyl carrier protein (PCP)domain, and condensation (C) domain. The A domains arecommonly referred to as the “gatekeepers” of a module, be-cause they recognize the substrate that is incorporated into thegrowing peptide chain. In a significant breakthrough in under-standing NRPS enzymology, an amino acid substrate specificitycode was identified that enables one to deduce the amino acid

that is likely recognized by an A domain, given the proteinsequence of the A domain (8, 9, 35, 60). This offers the abilityto predict which module incorporates each amino acid of anonribosomal peptide of known structure, but it also providesa means for proposing what amino acid is recognized by anNRPS module of unknown function. Once it recognizes theamino acid substrate, an A domain generates the correspond-ing aminoacyladenylate and subsequently works in conjunctionwith its partner PCP domain to tether the amino acid onto the4�-phosphopantetheinyl prosthetic group of the PCP domainvia a thioester linkage. Once the aminoacylthioester is formed,the C domain of the module catalyzes directional peptide bondformation with the amino acid tethered to the PCP domain ofthe preceding module. Thus, in a stepwise process, the nonri-bosomal peptide is biosynthesized.

The conservation of these repeating catalytic domains pro-vides the ability to identify new nonribosomal peptides usingan in silico approach, also referred to as “genome mining” (4,12, 31, 51). This approach involves the scanning of the pro-posed proteome of a sequenced genome for enzymes showingsequence similarity to the three core domains of NRPSs, fol-lowed by genetic, biochemical, and metabolite analysis of wild-type and mutant strains. We have previously used such anapproach to identify a biosynthetic gene cluster and purify itsassociated siderophore from the plant pathogen Agrobacteriumtumefaciens C58 (51), and others have taken a similar approachto identify the nonribosomal peptide siderophore coelichelinfrom Streptomyces coelicolor (7, 31).

Pseudomonas species are prolific producers of secondarymetabolites, in particular, metabolites synthesized wholly or inpart by NRPSs. These nonribosomal peptide-containing me-tabolites are used by pseudomonads in many biological con-texts, such as siderophores (e.g., pyoverdine [37]), biosurfac-tants (e.g., arthrofactin [41]), and virulence factors (e.g.,

* Corresponding author. Mailing address: Department of Bacteriol-ogy, University of Wisconsin—Madison, 150 Biochemistry, 420 HenryMall, Madison, WI 53706. Phone: (608) 263-9075. Fax: (608) 262-9865.E-mail: [email protected].

† Supplemental material for this article may be found at http://jb.asm.org/.

‡ Present address: Department of Microbiology, University of Illi-nois at Urbana-Champaign, 601 S. Goodwin Avenue, Urbana, IL61801.

� Published ahead of print on 29 June 2007.

6312

on April 3, 2021 by guest

http://jb.asm.org/

Dow

nloaded from

https://crossmark.crossref.org/dialog/?doi=10.1128/JB.00725-07&domain=pdf&date_stamp=2007-09-01http://jb.asm.org/

syringomycin [5] and coronatine [39]). These various biologicalactivities, the well-established production of nonribosomalpeptides, and the public availability of many Pseudomonasgenomes (6, 13, 23, 43, 46, 61) suggest that in silico mining ofthe genomes of Pseudomonas species will lead to the discoveryof new nonribosomal peptides. In fact, this approach was re-cently used to identify a cyclic lipopeptide from Pseudomonasfluorescens Pf0-1 (12).

Here we present the in silico identification of a biosyntheticgene cluster in Pseudomonas syringae pv. tomato DC3000 thatwe hypothesized to be involved in the production of a lipopep-tide. Metabolite profiles of wild-type strains compared to mu-tant strains containing gene disruptions in the targeted biosyn-thetic gene cluster identified a set of metabolites associatedwith this gene cluster. Purification of these metabolites fol-lowed by amino acid and mass spectrometry analyses identifiedthese metabolites as six related linear lipopeptides. Phenotypiccharacterization of the mutant strains determined these li-popeptides were essential for the swarming motility of P.syringae pv. tomato DC3000. This work demonstrated that insilico genome mining of organisms was a valuable approach toidentify previously unknown nonribosomal peptides.

MATERIALS AND METHODS

Bacterial strains, plasmids, and media. All bacterial strains and plasmids arelisted in Table 1. P. syringae pv. tomato DC3000 strains were grown at 28°C inKing’s medium B (KB) (26). Escherichia coli strains were grown in Luria-Bertani(LB) medium. As needed, strains were grown on KB or LB medium solidifiedwith 1.5% (wt/vol) agar. When required, antibiotics were added at the followingconcentrations: spectinomycin (100 �g/ml), nalidixic acid (5 �g/ml), gentamicin(5 �g/ml for E. coli, 2 �g/ml for P. syringae pv. tomato DC3000), and kanamycin(50 �g/ml).

Construction of a pspto_2829::lacZ-nptII strain (ADB1001). A 650-bp internalfragment of pspto_2829 was introduced into pVIK112 using PCR-based cloning.Briefly, the primers used for PCR amplification were 2829-EcoRI (5�-AACGAATTCTGGCTTGCCTCTTCAGACTT-3�) and 2829-SpeI (5�-CCGACTAGTTTACCTGGGCAAAGTGGC-3�). The amplicon generated by these primers was

digested with EcoRI and SpeI and ligated into the corresponding sites ofpVIK112. The resulting construct, pADB1, contained the internal pspto_2829fragment inserted in the same orientation as the promoterless lacZ. The plasmidwas subsequently introduced into E. coli S17-1 for conjugation into P. syringae pv.tomato DC3000. For conjugation, E. coli S17-1/pADB1 and P. syringae pv.tomato DC3000 were mixed in a ratio of 1:2, plated on KB agar, and grown for18 to 24 h. The mixed colony was scraped from the plate, resuspended in 1 ml of10 mM MgSO4, and plated on KB medium containing nalidixic acid (5 �g/ml)and kanamycin (50 �g/ml). Exconjugants were streaked for isolation on the samemedium and subsequently characterized for proper pspto_2829::lacZ-nptII in-sertion by PCR amplification.

Construction of a �pspto_2829 strain (ADB1002). To construct the plasmidused to delete pspto_2829, two PCR amplicons were cloned into the vectorpEX19-Gm, resulting in the fusion of the DNA upstream of pspto_2829 to thestart of pspto_2830, thereby generating a deletion of pspto_2829. This generateda vector, pADB2, that placed the start codon of pspto_2830 at the start site of thedeleted pspto_2829. The amplicons were generated using two primer sets: 2829-5�EcoRI (5�-ATGAATTCGACCCCCAGTTTGGCGGTGGCG-3�) and 2829-3�NdeI (5�-ATCATATGTCAAGGTCCTTCTTGGCAG-3�), 2930-5�NdeI (5�-ATCATATGAGCATCAACGAACTTCTGGC-3�), and 2930-3�KpnI (5�-ATGGTACCTCAGGCGCTGGTGGGTGGTGCG-3�). The amplicon from the firstprimer set was cloned into the corresponding EcoRI/NdeI sites of pEX19-Gm.The amplicon from the second primer set was cloned into the NdeI/KpnI sites ofthe pEX19-Gm vector containing the first amplicon. The resulting plasmid,pADB2, was transformed into E. coli S17-1, and the resulting strain was used forconjugations with P. syringae pv. tomato DC3000, selecting for gentamicin (2�g/ml) resistance. The resulting exconjugants were grown in KB agar lackinggentamicin and subsequently plated on KB agar supplemented with 5% (wt/vol)sucrose. Strains that grew were screened for gentamicin sensitivity, and thedeletion of pspto_2829 was confirmed by PCR amplification analysis. One ofthese resulting strains was used for further analysis (ADB1002).

Construction of pspto_2828::spc (ADB1003) and pspto_2833::spc (ADB1004)strains. A 2.1-kb region of the P. syringae pv. tomato DC3000 genome containingpspto_2828 was PCR amplified using the following primers: 2828-5�SpeI (5�-TTAACTAGTGAGCTTTGATAGCCGGGTATT-3�) and 2828-3�SpeI (5�-GAACTAGTAAGTCTGAAGAGGCAAGCCA-3�). The amplicon was digested withSpeI and cloned into the corresponding site of pEX19-Gm. The resulting plas-mid, pADB3, was digested with EcoNI, and the SmaI fragment of pHP45-�containing the spectinomycin resistance gene cassette was ligated into this site.This resulted in a vector construct, pADB4, which contained a spectinomycinresistance cassette inserted into the cloned pspto_2828.

For construction of a pspto_2833::spc strain, a 2.1-kb region of the P. syringae

TABLE 1. Strains used in this study

Strain or plasmid Relevant genotype or descriptiona Reference or source

E. coli strainsS17-1 recA thi pro hsdRM� RP4-2 Tcr::Mu Km::Tn7 57DH5� �(lac)U169 �80�lacZM15 hsdR17 endA1 gyrA96 recA1 relA1 supE44 thi-1 14

P. syringae pv. tomatoDC3000 strains

Wild type 11ADB1001 DC3000 pspto_2829::lacZ-nptII This studyADB1002 DC3000 �pspto_2829 This studyADB1003 DC3000 pspto_2828::spc This studyADB1004 DC3000 pspto_2833::spc This study

PlasmidspEX19-Gm Gene replacement vector, Gmr (aacC1) oriT� sacB� ColE1 replicon 20pJQ200SK� Gene replacement vector, Gmr (aacC1) oriT� sacB� p15a replicon 49pHP45-� Source of � cassette containing Spr/Smr gene (aadA) 47pVIK112 lacZ fusion vector, Kmr (nptII) oriT� lacZY R6K replicon 24pADB1 Internal pspto_2829 fragment cloned into pVIK112 This studypADB2 pspto_2828 and pspto_2830 fragments cloned into pEX19-Gm This studypADB3 Internal pspto_2828 fragment cloned into pJQ200SK� This studypADB4 � cassette from pHP45-� inserted into EcoNI site of pADB3 This studypADB5 pspto_2833 fragment cloned into pEX19-Gm This studypADB6 � cassette from pHP45-� inserted into FspI site of pADB5 This study

a Abbreviations: Gmr, gentamicin resistance; Kmr, kanamycin resistance; Spr, spectinomycin resistance; Smr, streptomycin resistance.

VOL. 189, 2007 LIPOPEPTIDES FROM P. SYRINGAE PV. TOMATO DC3000 6313


http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

pv. tomato DC3000 genome containing pspto_2833 was PCR amplified using thefollowing primers: 2833-5�SacI (5�-AAGGAGCTCGACCACAACACCGCTCAA-3�) and 2833-3�SacI (5�-CTTGAGCTCGATATTCCGGCAGCTCTGAA-3�).The amplicon was digested with SacI and cloned into the corresponding site ofpEX19-Gm. The resulting plasmid, pADB5, was digested with FspI, and theSmaI fragment of pHP45-� containing the spectinomycin resistance gene cas-sette was ligated into this site. This resulted in a vector construct, pADB6, whichcontained a spectinomycin resistance cassette inserted into the cloned pspto_2833.

To introduce the inactive gene into P. syringae pv. tomato DC3000, each ofthese plasmids was introduced into E. coli S17-1, and the resulting strains wereused in conjugation reactions with P. syringae pv. tomato DC3000. Exconjugantswere selected using gentamicin (2 �g/ml) resistance and screened for spectino-mycin resistance (100 �g/ml). Strains showing resistance to both antibiotics weregrown in KB medium containing spectinomycin (100 �g/ml) and then plated onKB agar containing sucrose (5% [wt/vol]) and spectinomycin (50 �g/ml). Thoseshowing sucrose and spectinomycin resistance were screened for gentamicinsensitivity. Insertion of the spectinomycin cassette into the genome of thesestrains was confirmed by PCR.

Swarming motility assay. Cultures of the analyzed strains were grown in KBmedium and adjusted to an optical density at 600 nm of 1.0. Five �l of thesecultures was spotted on KB agar containing 0.3, 0.5, 0.7, 0.9, and 1.5% (wt/vol)agar. The plates were incubated at 28°C for 48 to 72 h. It was determined that anagar concentration of 0.5% enabled wild-type P. syringae pv. tomato to swarmacross the KB agar surface. When analyzing expression of lacZ, the KB agar wassupplemented with 40 �g/ml of 5-bromo-4-chloro-3-indolyl-�-D-galactopyrano-side.

Droplet collapse assay. Standard droplet collapse assays were used (22).Briefly, for strains grown in KB medium, 1 ml of cells was removed from asaturated culture and the cells were removed by centrifugation. Twenty �l ofcell-free supernatant was mixed with 1 �l of methylene blue (0.1% [wt/vol] inwater). The resulting mixture was spotted onto Parafilm and allowed to settle for5 min before analyzing for droplet collapse. The methylene blue was added solelyfor visualization purposes and does not influence droplet collapse activity. Forstrains grown on KB agar, cells were removed from the plate by scraping,resuspended in liquid KB medium, and then removed by centrifugation. Twenty�l of the cell-free supernatant was spotted on Parafilm and allowed to settle for5 min before analyzing droplet collapse.

Thin-layer chromatography (TLC) and lipopeptide detection. Two �l of asample from the different steps of metabolite purification was spotted on a silicagel chromatography plate (silica gel F-254; Fisher). Metabolites were separatedusing a mobile phase of chloroform-methanol-NH4OH (5 N) in a ratio of 80:25:4. Lipopeptides were visualized by staining with bromothymol blue (0.1%[wt/vol] in 10% [vol/vol] ethanol), followed by brief heating.

Metabolite purification. The P. syringae pv. tomato DC3000 strains weregrown in 3-ml cultures of KB broth with vigorous shaking for 80 h at 28°C. Sixhundred ml of total volume was used for the purification of metabolites fromeach strain. The cells were removed by centrifugation (7,000 rpm, 10 min), andcell-free supernatant from pooled cultures was extracted with ethyl acetate con-taining 1% (vol/vol) formic acid. The culture supernatant and acidified ethylacetate were mixed in a ratio of 3:5 (vol/vol). The ethyl acetate fraction wascollected and dried to completion in a rotary evaporator. The metabolites weresuspended in deionized water and neutralized to pH 8 with dilute NaOH. Theaqueous mixture was applied to a silica chromatography column (SelectoScien-tific), and the metabolites were eluted with deionized water. The fractions activefor droplet collapse activity were dried to completion and resuspended in meth-anol. Methanol-soluble compounds were filtered through a 0.8-�m nylon filter(CoStar; Corning), dried to completion, resuspended in deionized water, andincubated at 20°C for several hours. Fortuitously, the metabolites of interestprecipitated from solution after freezing. The precipitated metabolites wererecovered by centrifugation at 13,000 rpm. The pellet was resuspended in ace-tonitrile-water (3:7).

The metabolites were separated by reverse-phase high-performance liquidchromatography (HPLC) using an analytical C18 small-pore column (Vydac) ata flow rate of 1 ml min1 and monitored at 210 nm. The solvents used were asfollows: A, deionized water with 0.1% trifluoroacetic acid; B, acetonitrile with0.1% trifluoroacetic acid. The HPLC separation profile consisted of a 5-minisocratic development at 100% A–0% B, a 7-min linear gradient from 100%A–0% B to 50% A–50% B, a 15-min linear gradient from 50% A–50% B to 0%A–100% B, and a 10-min isocratic development at 0% A–100% B. Metabolitesof interest were collected as they eluted from the HPLC column and dried tocompletion under vacuum.

When investigating surfactin for hydrolysis during ethyl acetate extraction, 1

mg of surfactin (Sigma-Aldrich) was added to cell-free supernatant of a 15-mlculture of P. syringae pv. tomato DC3000 that was grown under conditionsoptimized for lipopeptide production. The sample was then extracted with acid-ified ethyl acetate as described above, resuspended in methanol, and submittedto the University of Wisconsin Biotechnology Center for matrix-assisted laserdesorption ionization–time of flight mass spectrometry (MALDI-TOF MS) anal-ysis as described below.

Amino acid analysis. A portion of each sample that eluted from the HPLCcolumn was submitted to the University of California at Davis Molecular Struc-ture Facility for amino acid analysis.

Mass spectrometry analysis. A portion of each individual fraction recoveredfrom the C18 HPLC column was submitted to the UW Biotechnology Facility forMALDI-TOF MS analysis. Samples were spotted to stainless steel targets, mix-ing 1:1 analyte with saturated �-cyano-4-hydroxy cinnamic acid matrix in 70%acetonitrile–water, 0.2% trifluoroacetic acid. Spots were allowed to dry and thenanalyzed by using a 4800 TOF/TOF mass spectrometer (Applied Biosystems).For the analysis of ethyl acetate extracts from strains, MALDI-TOF MS analysiswas performed using a Voyager Biospectrometry workstation (DE Pro; AppliedBiosystems, Foster City, CA) in linear mode. Calibration was performed usingapomyoglobin, bovine insulin, and cytochrome c (Sigma).

A portion of each sample was also submitted to the UW Biotechnology Facilityfor electrospray ionization mass spectrometry/mass spectrometry (ESI-MS/MS)analysis. Samples were introduced into a 3200 Q trap (Applied Biosystems) massspectrometer by direct infusion via a 1-ml Hamilton syringe and metering pumpat 30 �l min1. The capillary was maintained at 5,500 V. Collision energy was setto 25 V for collection of doubly charged species. Spectra were collected inenhanced product ion scan mode with precursors being selected with an opened(“low”) resolution setting on the Q1 quadrupole. Linear trap functions were setto “dynamic fill” time and were optimized for peptides. The declustering poten-tial was 15 V.

RESULTS AND DISCUSSION

In silico identification of a nonribosomal peptide biosyn-thetic gene cluster and prediction of the metabolite structure.A gene cluster that may code for an NRPS that synthesizes anonribosomal peptide was identified in the P. syringae pv. to-mato DC3000 genome by searching the predicted proteomeusing the genomic BLAST program from the National Centerfor Biotechnology Information (1, 54), with the default settingsfor proteins similar to the Bacillus subtilis GrsA protein. GrsAcontains the initiation module of the NRPS that generates thenonribosomal peptide gramicidin (28). This protein was cho-sen for the BLAST search because it is biochemically andstructurally the most well studied NRPS subunit, and the Adomain of GrsA was used as the basis for the development ofthe A domain amino acid specificity codes (8, 9, 59, 60). Asshown in Fig. 1, one of the gene clusters we identified con-tained two large genes, pspto_2829 and pspto_2830, that codedfor eight NRPS modules. Immediately downstream of thesetwo genes were two open reading frames (ORFs; pspto_2831and pspto_2832) that showed sequence similarity to subunits ofa multidrug efflux pump. These two genes were potentially inan operon with pspto_2829 and pspto_2830, based on the over-lapping stop and start codons of pspto_2830 and pspto_2831and the separation of pspto_2831 and pspto_2832 by only 3nucleotides. On either side of this putative four-gene operonwere genes coding for proteins showing sequence similaritywith the LuxR family of transcriptional regulators. Thus, theproduction of the putative nonribosomal peptide potentiallyinvolved pspto_2828 through pspto_2833 (Fig. 1).

Analysis of the NRPS encoded by pspto_2829 and pspto_2830identified eight modules, suggesting an octapeptide would bebiosynthesized (Fig. 1). The likely order of function for thesetwo NRPS subunits was first Pspto_2829, consisting of the first

6314 BERTI ET AL. J. BACTERIOL.


http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

three NRPS modules, and then Pspto_2830, consisting of thefinal five modules. This hypothesis was based on the C-termi-nal thioesterase domains of Pspto_2830. These domains typi-cally catalyze the final step in nonribosomal peptide synthesisand denote the terminal module of an NRPS (reviewed inreference 25). Pspto_2830 terminated in tandem thioesterasedomains that are both likely active and may function to en-hance the rate of product release from the NRPS, as seenduring the analysis of arthrofactin biosynthesis (53). Also ofinterest was the N-terminal C domain present in the initiatingmodule of Pspto_2829. This domain organization implied thatthe nonribosomal peptide generated by this enzymology wouldcontain an N-terminal lipid. This hypothesis was based onprecedent whereby an N-terminal C domain recognizes a fattyacyl-coenzyme A or fatty acyl-acyl carrier protein derivativefrom primary metabolism and condenses the first amino acidtethered to the NRPS with the thioesterified fatty acid togenerate a lipidated nonribosomal peptide (21, 38, 63).

The eight A domains of the NRPS were analyzed for theiramino acid substrate specificity codes to predict the amino acidsequence of the octapeptide based on comparisons to knownsubstrate specificity codes (Table 2). Based on this analysis, theNRPS identified by in silico analysis was predicted to generatea lipooctapeptide with the sequence lipid-Leu-Leu-Gln-Leu-Thr-Val-Leu-Leu (Fig. 1). Bioinformatics analysis also deter-mined the C domains of modules 2, 3, 4, 6, and 8 were slightlylarger than standard C domains, having additional amino acidsat the N terminus. Work by other groups has established thatthese types of C domains in Pseudomonas species are bifunc-tional (2). These C domains catalyze not only the directionalcondensation between two neighboring PCP-tethered aminoacids but also epimerization of the amino acid associated withthat particular module. Thus, it was reasonable to hypothesizethat residues 2, 3, 4, 6, and 8 of the lipooctapeptide would beD-isomers rather than the natural L-isomers.

Finally, the N-terminal C domain of pspto_2829 showed the

highest level of amino acid sequence similarity with the N-terminal C domain of ArfA, the initiating NRPS that biosyn-thesizes the cyclic lipopeptide arthrofactin (52). However, theremainder of pspto_2829 showed a high level of sequence

FIG. 1. Schematic of the targeted biosynthetic gene cluster in P. syringae pv. tomato DC3000. (Top) Graphic representation of the targeted genecluster spanning pspto_2828 to pspto_2833. Colors of genes indicate putative functions of coded proteins: white, NRPS; gray, regulation; black,export. (Middle) Predicted NRPS domains, shown as circles, encoded by pspto_2829 and pspto_2830. Abbreviations: A, adenylation domain; C,condensation domain; Te, thioesterase domain; C*, dual-function condensation-epimerization domain. Domains have been grouped into modules,with each module identified by a bar and module number. (Bottom) Proposed lipopeptide sequence based on the order and predicted substratespecificity of A domains and the presence of an initiating C domain. The use of syf for the gene designation and Syf for the protein designationis proposed for those genes and proteins noted (see “Summary and conclusions,” below).

TABLE 2. Analysis of A domain amino acid specificity codes

NRPS component andA domaina Specificity code

b Predicted amino acidactivated (reference)

Pspto_2829 A1 DAWFLGNVVKBacA A3 DAWFLGNVVK Leu (27)

Pspto_2829 A2 DAWPLGNVVKBacA A3 DAWFLGNVVK Leu (27)

Pspto_2829 A3 DAWQVGVVDKTycC A2 DAWQFGLIDK Gln (40)


Pspto_2830 A2 DFWNIGMVHKDhbF A2 DFWNIGMVHK Thr (36)

Pspto_2830 A3 DAMFIGGTFKGrsB A2 DAFWIGGTFK Val (62)


Pspto_2830 A5 DAWAHLNVVKBacA A3 DAWFLGNVVK Leu (27)

a NRPS components are grouped according to adenylation domain homology.For example, the amino acid substrate specificity code of the A1 domain ofPspto_2829 is most similar to the A3 domain of BacA. Since some proteins consistof more than one A domain, the specific A domain is identified by a number. Forexample, the first A domain of Pspto_2829 is denoted as A1, while the second is A2.

b Specificity code, as described in references 8 and 60. The residues predicted toconstitute the amino acid binding pocket are listed. The alignment program toidentify the substrate specificity code can be found at http://www.nii.res.in/nrps-pks.html.



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

similarity to ArfB, starting with the A domain of module 4 ofthe arthrofactin NRPS. Thus, protein domains correspondingto the first three modules of the arthrofactin NRPS were ab-sent. This suggested that the NRPS system targeted here in P.syringae pv. tomato DC3000 evolved from the arthrofactin sys-tem, with one of the changes being the deletion of three mod-ules of the arthrofactin NRPS, resulting in the fusion of the Nterminus of ArfA with a portion of ArfB. Importantly, thedeleted modules include the module that incorporates thethreonyl residue that forms the ester linkage involved in cycli-zation of arthrofactin (52, 53). This suggested that the NRPSfrom P. syringae pv. tomato DC3000 would generate a linearlipopeptide rather than a cyclic lipopeptide.

It should be noted that while the manuscript was beingprepared, a paper was published that used a similar approachto identify cyclic lipopeptides in Pseudomonas species (12). Inthat study, the authors proposed a structure for the metaboliteproduced by the biosynthetic gene cluster targeted here. Thereare two important differences in their proposed structure andthe one we propose. First, they proposed the metabolite is acyclic lipopeptide, and our analysis suggested the moleculewould be linear. Second, they predicted a different amino acidsequence for the peptide. The peptide sequence they predictedis Leu-Leu-(Glu/Asp)-Leu-Dhb-Ile-Leu-Leu, with Dhb repre-senting 2,3-dihydroxy-2-aminobutyric acid. Thus, three of theeight proposed amino acids are different. It can be argued thatthe previously predicted amino acid sequence is very unlikely.First, there are no nonribosomal peptides we are aware of thatcontain the unusual amino acid 2,3-dihydroxy-2-aminobutyricacid. There is precedent for 2-aminobut-2-enoic acid, thechemically correct name for what has been called 2,3-dehydro-2-aminobutyric acid in syringomycin and syringopeptin (3, 16),but not 2,3-dihydroxy-2-aminobutyric acid. Second, 2,3-dihy-droxy-2-aminobutyric acid is a hemiaminal, which is not achemically stable compound. Therefore, it is unlikely it will befound as the free amino acid in the cell. This is exemplified bythe fact that hemiaminal amino acids in peptides are found asintermediates in enzymatic reactions that lead to peptide cleav-age (30, 56). These discrepancies in proposals highlight thatwhile bioinformatics analysis enables predictions to be made,interpretations can differ and experimental approaches mustbe used to test such hypotheses before structures are proposedon uncharacterized clusters.

Mutants disrupted in pspto_2829 have phenotypes consis-tent with the loss of lipopeptide production. To initiate theanalysis of the targeted gene cluster, two strains mutated inpspto_2829 were constructed. The first strain contained aninsertion of a pVIK112 suicide vector into the chromosomeof P. syringae pv. tomato DC3000 that resulted in a promot-erless lacZ being inserted into, and in the same orientationas, pspto_2829. This offered the ability to monitor expres-sion of pspto_2829. A second mutation was constructedwhereby the pspto_2829 gene was deleted from the chromo-some, with pspto_2830 being positioned in place of pspto_2829. This strain provided a means of phenotypic character-ization and metabolite purification that did not require theaddition of antibiotic to retain the mutation, thereby simplify-ing metabolite profile comparisons with the wild-type strain.

The bioinformatics analysis of the targeted gene cluster sug-gested it was involved in the formation of a lipopeptide. One of

the functions of lipopeptides is as a biosurfactant that enablesthe bacterium to swarm on solid medium containing a lowpercentage of agar (32–34). This offered a simple phenotype toinvestigate. The wild-type and pspto_2829::lacZ-nptII strainswere spotted on KB medium containing various concentrationsof agar (0.3, 0.5, 0.7, 0.9, and 1.5% [wt/vol]) and supplementedwith 5-bromo-4-chloro-3-indolyl-�-D-galactopyranoside to en-able the detection of lacZ expression. These plates were incu-bated at 30°C for 48 h, and the plates were analyzed for bluepigment formation and swarming motility. From this analysis itwas determined that the pspto_2829::lacZ-nptII strain turnedblue on these plates and also failed to swarm on any of the agarplates. On a control plate, the wild-type strain did not turnblue, since it does not include lacZ, but did show swarmingmotility on medium containing 0.5% (wt/vol) agar (data notshown). Importantly, when the pspto_2829::lacZ-nptII strainwas grown on LB with various concentrations of agar, thestrain did not turn blue in the presence of 5-bromo-4-chloro-3-indolyl-�-D-galactopyranoside and did not swarm. Also, thewild-type strain did not show swarming motility on LB medium(data not shown). This confirmed that on KB medium, P.syringae pv. tomato DC3000 expressed pspto_2829 and themetabolite associated with the NRPS encoded by pspto_2829was important for swarming motility of the bacterium. Retest-ing the swarming phenotype with the �pspto_2829 strain indi-cated that this strain also failed to swarm on KB mediumcontaining 0.5% agar, consistent with the loss of biosurfactantproduction. The wild-type strain was unaffected (Fig. 2A).

As a second test for the presence or absence of a biosurfac-tant, standard droplet collapse assays were performed (22).Briefly, aqueous solutions lacking a surfactant form domeddroplets on various surfaces due to surface tension. When asurfactant is present, the surface tension is reduced and thedomed droplets collapse. The cells from the swarming plates ofthe wild-type and mutant strains were scraped from the plate,resuspended in water, and removed by centrifugation. Theresulting supernatant was analyzed for whether the superna-tant stimulated droplet collapse. The supernatant of the wild-type strain caused droplet collapse, but the supernatants of themutants did not (data not shown). This same result was ob-served when the strains were grown in liquid culture, the cellswere removed by centrifugation, and the culture supernatantwas spotted on Parafilm (Fig. 2B). These data were consistentwith the targeted gene cluster being involved in the productionof a lipopeptide biosurfactant.

A strain disrupted in pspto_2828, but not pspto_2833, hasphenotypes consistent with the loss of lipopeptide production.ORFs coding for LuxR-like transcriptional regulators are lo-cated upstream of pspto_2829 and downstream of pspto_2832(Fig. 1). This raised the possibility that either one or bothgenes code for a transcriptional regulator involved in the pro-duction of the metabolite with biosurfactant activity. To inves-tigate this, spectinomycin resistance gene cassettes were intro-duced into pspto_2828 and pspto_2833 and the resultingmutant strains were analyzed for swarming motility and stim-ulation of droplet collapse. These analyses determined that aninsertion in pspto_2828 resulted in the loss of swarming mo-tility and droplet collapse, but an insertion in pspto_2833 didnot have a discernible effect (Fig. 2A and B). Therefore, onlypspto_2828 played a clear role in lipopeptide production.



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

Detection and purification of the metabolites associatedwith the pspto_2828-pspto_2832 gene cluster. The wild-type P.syringae pv. tomato DC3000 was grown under a variety ofconditions to determine the optimal conditions for biosurfac-tant activity. It was determined that small-volume (3-ml) cul-tures with vigorous shaking for 3 days resulted in the strongestdetectable biosurfactant activity based on the droplet collapseactivity of serial dilutions of cell-free culture supernatants. Itwas surprising that such high aeration conditions would en-hance biosurfactant production, due to the role the biosurfac-tant plays in swarming motility. However, high levels of bio-surfactant production in liquid culture are not uncommon(e.g., rhamnolipid [19], surfactin [10], and putisolvins I and II[29]). Furthermore, the regulation of biosurfactant productionis also influenced by the medium in which the cells are grown,

since no biosurfactant activity or pspto_2829::lacZ-nptII ex-pression was observed on LB plates. Thus, the regulation oflipopeptide production is influenced in this bacterium by manyfactors besides growth on a solid support. Further investiga-tions into the regulation of biosurfactant production by thisstrain and the role that the LuxR homolog, Pspto_2828, playsin this process are ongoing.

Analysis of ethyl acetate extracts of the supernatants of theaerated cultures using TLC followed by visualization by bro-mothymol blue staining to discern lipid-containing metabolites(64) detected a metabolite in the wild-type culture that wasabsent in the �pspto_2829 mutant strain (Fig. 2C). Thus, un-der these growth conditions, P. syringae pv. tomato DC3000produces a metabolite that contains a lipid moiety, and pro-duction of this metabolite is associated with pspto_2829. Thecombination of droplet collapse assay results and TLC sepa-ration and bromothymol blue staining were used to monitormetabolite purification.

Six-hundred-milliliter cultures (200 3-ml cultures) of the wild-type, �pspto_2829, and pspto_2828::spc strains were grown formetabolite purification and comparison. The purificationscheme involved ethyl acetate extraction, silica resin chroma-tography, precipitation, and reverse-phase HPLC as describedin Materials and Methods. The final step using HPLC puri-fication enabled a comparison of the metabolite profiles ofthe wild-type strain versus the two mutant strains. From thisanalysis, four distinct peaks were observed in the HPLCtraces of the metabolites purified from the wild-type strainthat were absent from the �pspto_2829 and pspto_2828::spcstrains (Fig. 3). Each of these purified metabolites stimu-lated droplet collapse, consistent with the purification ofmetabolites with lipopeptide characteristics. These data sug-gested the targeted biosynthetic gene cluster was involved inthe production of at least four metabolites. This is in con-trast to the bioinformatics that predicted a single lipopep-tide structure.

Amino acid analysis of purified metabolites. The metabo-lites corresponding to the four absorbance peaks observedfrom the HPLC analysis were independently collected for fur-

FIG. 2. Phenotypes associated with the pspto_2828-pspto_2833gene cluster. (A) Motility of P. syringae pv. tomato DC3000 strains onKB medium containing 0.5% agar. Five �l of cells from each strain wasseparately spotted to the center of a plate and subsequently incubatedfor 48 to 72 h at 28°C. (B) Droplet collapse results for 20-�l aliquotsof cell-free supernatants of liquid cultures. Methylene blue was addedto each sample to aid in visualization. (C) Silica gel TLC separation ofethyl acetate extractions of liquid cultures, visualized with bromothy-mol blue staining. The arrow identifies a region of the TLC plate wherea metabolite from the wild-type (WT) strain reacted with the bromo-thymol blue stain, indicating a lipid-containing metabolite. Spc, spec-tinomycin resistance gene cassette.

FIG. 3. Reverse-phase HPLC analysis of partially purified metab-olites. Trace A, wild type; trace B, pspto_2828::spc; trace C, �pspto_2829. Metabolite elution was monitored at 210 nm. The absorptionpeaks of the four samples analyzed further are highlighted with arrows,and the time of elution from the HPLC is noted.



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

ther characterization. The hypothesis was these metaboliteswould be structurally related lipooctapeptides produced by theenzymes encoded by the pspto_2828-2832 gene cluster. EachHPLC-purified sample was submitted for amino acid analysisto determine its amino acid content (Fig. 4). These data con-firmed that each metabolite contained eight amino acids. Themetabolites eluting at 24.1 and 26.5 min contained the aminoacids Leu, Val, Thr, and Glx in a molar ratio of 5.0:1.0:0.9:1.1and 5.0:1.0:0.9:1.1, respectively, with Glx representing eitherGln or Glu. These amino acids and their ratios were consistentwith the prediction made through bioinformatics analysis of a5:1:1:1 ratio of Leu, Val, Thr, and Gln (Fig. 1). Since these twometabolites had identical amino acid content but different re-tention times on the reverse-phase HPLC, it was predictedthese metabolites would have different N-terminal lipids.Interestingly, the metabolites eluting at 24.7 and 27.2 min fromthe HPLC had slightly altered amino acid content compared tothe metabolites eluting at 24.1 and 26.5 min. The amino acidswere Leu, Ile, Thr, and Glx in a ratio of 5.4:0.6:0.9:1.1 and5.5:0.5:0.9:1.1 for the metabolites eluting at 24.7 and 27.2 min,respectively (Fig. 4). This suggested the metabolites eluting at24.7 and 27.2 min were each a mixture of two lipopeptides thatdiffered in having either Leu or Ile at one position, most likelythe position that was Val in the metabolites eluting at 24.1 and26.5 min. Additionally, since the retention times were differentbut the amino acid content was the same, it was reasonable topropose that the metabolites eluting at 24.7 min have a differ-ent fatty acid from those eluting at 27.2 min. From this anal-ysis, it was concluded that the NRPS encoded by pspto_2829and pspto_2830 resulted in the production of six structurallyrelated lipooctapeptides. Thus, while bioinformatics analysisand HPLC separation suggested one or four metabolites wereproduced by this biosynthetic gene cluster, respectively, theamino acid analysis supports the presence of six lipopeptides.

Mass spectrometry analysis of the purified metabolites. Togain further insights into the structures of the metabolites,

each of the purified samples from the HPLC was analyzed byMALDI-TOF MS (Table 3). Interpretations of these data ben-efited from our knowing the amino acid content of each me-tabolite. MALDI-TOF MS analysis of the metabolite eluting at24.1 min identified four masses at 1,104.8, 1,120.8, 1,126.8, and1,142.7 Da. The 1,104.8-Da mass was consistent with a metab-olite consisting of a peptide with the determined amino acidcontent (Fig. 4), an N-terminal 3-hydroxydecanoyl fatty acid,and a sodium ion (910.6 � 171.1 � 23 1,104.7 Da). Thehypothesis that 3-hydroxydecanoate was the lipid attached tothe N terminus was based on the observed mass and this lipid

FIG. 4. Amino acid analysis. Data shown represent the numbers and identities of amino acids recovered after acid hydrolysis of metabolitescollected at the elution times indicated. The data were normalized to a total of eight amino acids. GLX represents Gln or Glu.

TABLE 3. MALDI-TOF MS analysis of purified metabolites

HPLC elutiontime (min) Observed mass (theoretical mass)

a

24.1 1,104.8 �M � Na�� (1,104.7)1,120.8 �M � K�� (1,120.7)1,126.8 �(M-H) � 2Na�� (1,126.7)1,142.7 �(M-H) � Na � K�� (1,142.7)

24.7 1,118.8 �M � Na�� (1,118.7)1,134.8 �M � K�� (1,134.7)1,140.8 �(M-H) � 2Na�� (1,140.7)1,156.7 �(M-H) � Na � K�� (1,156.7)

26.5 1,132.8 �M � Na�� (1,132.7)1,148.8 �M � K�� (1,148.7)1,154.8 �(M-H) � 2Na�� (1,154.7)1,170.7 �(M-H) � Na � K�� (1,170.7)

27.2 1,146.8 �M � Na�� (1,146.7)1,162.8 �M � K�� (1,162.7)1,168.8 �(M-H) � 2Na�� (1,168.7)1,184.8 �(M-H) � Na � K�� (1,184.7)

a The theoretical mass was determined by calculating the mass of a metabolitethat included the amino acids determined by amino acid analysis, an N-terminal3-hydroxydecanoyl (24.1 min and 24.7 min) or 3-hydroxydodecanoyl lipid (26.5min and 27.2 min), and the salt adduct noted.



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

being a common component of other lipidated nonribosomalpeptides from Pseudomonas spp. (17, 41). The three otherobserved masses were consistent with different adducts of thepredicted lipopeptide (Table 3). Importantly, the observedmasses suggested the lipopeptide was linear, not cyclic. If thelipopeptide were cyclic, the mass would have been 18 Dasmaller to account for the loss of two hydrogens and oneoxygen for the cyclization.

From the amino acid analysis, the metabolite eluting at 26.5min had the same amino acid content as that eluting at 24.1min; therefore, they likely differed in their lipid component.Consistent with this, the masses observed for the various ionforms of the metabolites eluting at 26.5 min were 28 Da largerthan those observed for the metabolite eluting at 24.1 min,suggesting a lipid component two carbon units larger (3-hy-droxydodecanoyl rather than 3-hydroxydecanoyl) (Table 3).3-Hydroxydodecanoyl is also found attached to lipopeptidesproduced by Pseudomonas species (15, 48), and so it was notsurprising to find a change in the mass consistent with this lipidbeing attached to the N terminus of the purified lipopeptides.

The masses of the two remaining metabolite samples wereexpected to differ from the other two samples by at least 14 Da,because a Val residue was replaced by a Leu or Ile residuebased on amino acid analysis (Fig. 4). Additionally, based onthe predicted differences in the lipid present in the prior set ofmetabolites, it was expected that the masses between the me-tabolites eluting at 24.7 and 27.2 min would differ by 28 Da.Consistent with these predictions, the metabolites eluting at24.7 and 27.2 min differed by 14 mass units from the metabo-lites eluting at 24.1 and 26.5 min, respectively (Table 3). Ad-ditionally, the metabolites at 24.7 and 27.2 min differed fromeach other by 28 Da. From these data, it was concluded that sixstructurally distinct lipooctapeptides had been identified, andthese lipooctapeptides differ either in lipid content or in havinga Val, Leu, or Ile residue at one amino acid position of thepeptide. Furthermore, the masses of all of these lipopeptidessuggested they were all linear, not cyclic. Importantly,MALDI-TOF MS analysis of the metabolites present after theinitial ethyl acetate extraction detected metabolites with thesame masses listed in Table 3 but did not detect masses con-sistent with the cyclized forms of the metabolites (data notshown). Ethyl acetate extraction is a common step in cycliclipopeptide purification (44, 45, 48, 58); therefore, it was un-likely that this purification step caused the metabolites to be-come linear.

To gain additional support for the conclusion that the me-tabolites were not hydrolyzed to the linear forms by the ethylacetate extraction, we tested whether the cyclic lipopeptidesurfactin was linearized by this purification step. Surfactin waschosen because it has a similar amino acid content as themetabolites isolated from P. syringae pv. tomato DC3000, andit is cyclized with an ester linkage (42), the linkage that wouldbe expected if our isolated metabolites were cyclic. Surfactinwas added to cell-free supernatant of wild-type P. syringae pv.tomato DC3000 after the culture had produced the metabo-lites we were characterizing. This mixture was then subjectedto ethyl acetate extraction and subsequent MALDI-TOF MSanalysis. As a control, surfactin that had not been added to thesupernatant or extracted with ethyl acetate was analyzed byMALDI-TOF MS. The mass spectrometry analyses deter-

mined that surfactin was not linearized by its addition to thecell-free supernatant and subsequent ethyl acetate extraction.In the sample containing surfactin and supernatant, the linearlipopeptides listed in Table 3 were detected (data not shown).Thus, it is unlikely that the ethyl acetate extraction methodcaused the linearization of the metabolites, suggesting P.syringae pv. tomato produces linear lipopeptides.

To more directly analyze the lipooctapeptides, each purifiedsample was analyzed by ESI-MS/MS. Using this technique,metabolite fragmentation patterns were compared with theknown amino acid content and MALDI-TOF MS results todetermine the likely structures of the lipooctapeptides. For thisanalysis, the [M � 2H]2� ion of each sample was collected andfragmented. Figure 5 shows the spectrum from the ESI-MS/MS analysis of the metabolite eluting at 24.1 min, alongwith the identification of the b and y fragments. The combina-tion of b and y fragments, along with the associated dehydrated(bo) and deaminated (b*) derivatives, supported the proposedlinear structure of 3-hydroxydecanoyl-Leu-Leu-Gln-Leu-Thr-Val-Leu-Leu for the metabolite eluting at 24.1 min (Fig. 5 and6). We propose the name syringafactin A for this metabolite.The masses of the fragments along with the MALDI-TOF datasupported the conclusion that Gln was the third amino acidrather than Glu, since the masses of these compounds wouldbe 1 Da larger if the residue were Glu. Also, since amino acidresidue six was Val, this was the amino acid that was expectedto vary in the metabolites eluting at 24.7 and 27.2 min from theHPLC.

Similar analyses were done on the three remaining metab-olite samples. The ESI-MS/MS spectra for these metabolitesare included online as supplementary figures (see Fig. S1 to S3in the supplemental material), while the mass fragments for allfour samples are summarized in Fig. 6A. Based on these frag-mentation patterns, the remaining metabolites were identifiedas five lipooctapeptides related in structure to syringafactin A.The metabolites eluting at 24.7 min had masses that wereconsistent with a mixture of lipooctapeptides with N-terminal3-hydroxydecanoyl lipids and with residue six of the peptidesbeing either Leu or Ile, as evident in the increase of 14 Da inthe mass of the b7, b

o7, b8, bo8, y3-y7, y

o6, and yo7 fragments (Fig.6A). These metabolites are referred to as syringafactins B(Leu) and C (Ile) (Fig. 6B). The remaining metabolite ESI-MS/MS fragmentation products were identical to those seenfor syringafactins A, B, and C, with the only difference being a28-Da increase in mass of the fragments containing the Nterminus of the metabolite (Fig. 6A). The observed increase inmass was consistent with a 3-hydroxydodecanoyl N-terminallipid compared to the 3-hydroxydecanoyl lipid seen in the threeprior syringafactins. These metabolites are referred to as sy-ringafactins D, E, and F for the metabolites containing Val,Leu, or Ile at the sixth position, respectively (Fig. 6B). Impor-tantly, bioinformatics predicted the production of a single li-popeptide; however, experimentally it was determined that sixlipopeptides were produced.

Finally, it was important to show that the reason the mutantstrains failed to swarm on low-agar KB was that they lacked theability to produce these six linear lipopeptides. Therefore, thepurified metabolites were pooled and then spread on KB agarcontaining 0.5% agar. The wild-type and �pspto_2829 strainswere spotted on these plates, and the plates were analyzed for



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

swarming motility after 48 h of incubation. Both the wild-typeand the mutant strains swarmed on these plates, showing thatthese six metabolites complemented the swarming motilityphenotype of the �pspto_2829 strain (data not shown). Impor-tantly, when the �pspto_2829 strain was grown under identicalconditions and cellular metabolites purified using the sameprotocol, samples collected from the HPLC at the same reten-tion times as the lipopeptides from the wild-type strain failedto complement swarming motility of the �pspto_2829 strain(data not shown). Thus, the restoration of swarming motility tothe mutant strain was due to the presence of the six linearlipopeptides.

Summary and conclusions. We have presented bioinfor-matic, phenotypic, chemical, and mass spectrometry data thatsix structurally related linear lipopeptides were produced by asingle biosynthetic gene cluster in P. syringae pv. tomatoDC3000. We propose that these metabolites be called syrin-gafactins A to F (Fig. 6B) and the associated genes be desig-nated as syfR (pspto_2828), syfA (pspto_2829), syfB (pspto_2830), syfC (pspto_2831), and syfD (pspto_2832) (Fig. 1). Themetabolites produced by the syf biosynthetic gene cluster wereshown to be involved in the swarming motility of P. syringae pv.tomato DC3000.

One surprising finding from this work was that the isolated

FIG. 5. Analysis of ESI-MS/MS fragmentation for the metabolite eluting at 24.1 min. (Top) A schematic is shown representing sites of cleavagein the proposed metabolite structure that would yield b and y ions. The theoretical masses of these ions are noted above the b notation and belowthe y notation. (Bottom) Ion masses obtained from ESI-MS/MS fragmentation of the [M � 2H]2� form of the metabolite are shown. The observedESI-MS/MS fragmentation ion masses are consistent with the theoretical values derived from the proposed lipopeptide structure as denoted at top.b° and y° denote dehydrated fragmentation products; b* denotes deaminated fragmentation product. The ion at 256.2 amu is consistent with the“a2” ion fragment [3-hydroxydecanoyl � Leu CHO].



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

metabolites were linear rather than cyclic. Prior to this work,lipopeptides identified from Pseudomonas species that func-tion as biosurfactants were found to be cyclic (reviewed inreference 50). Importantly, we were able to show that thepurified linear lipopeptides restored swarming motility to amutant strain. Thus, while many Pseudomonas species use cy-clic lipopeptides for a variety of biological purposes, linearlipopeptides can have equivalent functions.

We, and others (12), noticed a biosynthetic gene cluster thatcodes for an NRPS very similar to the syringafactin NRPS in P.syringae pv. syringae B728a (ORFs psyr_2576-psyr_2577).While it was proposed that this cluster would also produce acyclic lipopeptide with a different amino acid sequence than wedetermined for the syringafactins (12), the work presentedhere on the syringafactins suggests this NRPS generates linearlipopeptides with a similar amino acid sequence to the syrin-

FIG. 6. Proposed structures of lipopeptide metabolites. (A) Summary of ESI-MS/MS data obtained from metabolites eluting from the HPLCcolumn at the times indicated. Shown are the observed masses of b and y and dehydration (b°, y°) and deamination (b*) fragments. The observedmasses are consistent with the theoretical values. ND, not detected. The R1 and R2 groups are explained in panel B. (B) Proposed structures ofthe six metabolites, syringafactin A to F, produced by P. syringae pv. tomato DC3000 based on amino acid analysis and mass spectrometry.



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

gafactins. Consistent with this proposal, preliminary analysis ofthe metabolites isolated from P. syringae pv. syringae B728aidentified metabolites with the same HPLC retention time andthe same masses, based on MALDI-TOF MS, as syringafactinsA, B, and C (data not shown). Interestingly, metabolites elut-ing with the same retention time as syringafactins D, E, and Fwere not observed, suggesting only a single lipid will be foundat the N terminus of these metabolites. Thus, even NRPSsystems showing extensive amino acid similarity may not pro-duce the same metabolites. Further analyses of the metabolitesfrom P. syringae pv. syringae B728a are ongoing.

The work presented here highlights the importance of cou-pling bioinformatics analysis with experimental and metaboliteanalyses to test the hypothetical structure developed by bioin-formatics. We have shown that a single biosynthetic gene clus-ter produced not a single metabolite but six related metabolitesdue to the presumed substrate recognition flexibility of some ofthe NRPS components. The observed structural differenceslikely came from the substrate flexibility of the initiating Cdomain of Pspto_2829 (SyfA) and the A domain of module sixof the NRPS. Bioinformatics, at least at this time, is not accu-rate enough to make these distinctions. Regardless, the datapresented here show that combining genome mining and me-tabolite purification enables the identification of previouslyunknown secondary metabolites and, based on the determinedstructure, the biological role of the metabolite can be deci-phered. This approach has enormous potential in uncoveringnew metabolites due to the ever-expanding pool of sequencedmicrobial genomes, many of which have gene clusters thatpotentially generate previously unknown secondary metabo-lites.

ACKNOWLEDGMENTS

This work was supported by the Wisconsin Agricultural Experimen-tation Station project WIS04976 from the U.S. Department of Agri-culture and the Alfred Toepfer Faculty Fellow award from the Collegeof Agriculture and Life Sciences at the University of Wisconsin—Madison (M.G.T.). Q.H.C. was supported by an Ira L. Baldwin un-dergraduate scholarship.

We thank Herbert P. Schweizer (Colorado State University) forgenerously providing plasmids and Michelle R. Rondon for criticalreading of the manuscript. We thank Amy C. Harms and James F.Brown from the University of Wisconsin Biotechnology Center fortheir expert assistance in the mass spectrometry analyses. We thankMichael Burkart (University of California at San Diego) for discus-sions concerning amino acid chemistry.

REFERENCES

1. Altschul, S. F., T. L. Madden, A. A. Schäffer, J. Zhang, Z. Zhang, W. Miller,and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: a new generationof protein database search programs. Nucleic Acids Res. 25:3389–3402.

2. Balibar, C. J., F. H. Vaillancourt, and C. T. Walsh. 2005. Generation of Damino acid residues in assembly of arthrofactin by dual condensation/epi-merization domains. Chem. Biol. 12:1189–1200.

3. Ballio, A., F. Bossa, D. Di Giorgio, A. Di Nola, C. Manetti, M. Paci, A.Scaloni, and A. L. Segre. 1995. Solution conformation of the Pseudomonassyringae pv. syringae phytotoxic lipodepsipeptide syringopeptin 25-A. Two-dimensional NMR, distance geometry and molecular dynamics. Eur. J. Bio-chem. 234:747–758.

4. Bergmann, S., J. Schumann, K. Scherlach, C. Lange, A. A. Brakhage, and C.Hertweck. 2007. Genomics-driven discovery of PKS-NRPS hybrid metabo-lites from Aspergillus nidulans. Nat. Chem. Biol. 3:213–217.

5. Bidwai, A. P., and J. Y. Takemoto. 1987. Bacterial phytotoxin, syringomycin,induces a protein kinase-mediated phosphorylation of red beet plasma mem-brane polypeptides. Proc. Natl. Acad. Sci. USA 84:6755–6759.

6. Buell, C. R., V. Joardar, M. Lindeberg, J. Selengut, I. T. Paulsen, M. L.Gwinn, R. J. Dodson, R. T. Deboy, A. S. Durkin, J. F. Kolonay, R. Madupu,

S. Daugherty, L. Brinkac, M. J. Beanan, D. H. Haft, W. C. Nelson, T.Davidsen, N. Zafar, L. Zhou, J. Liu, Q. Yuan, H. Khouri, N. Fedorova, B.Tran, D. Russell, K. Berry, T. Utterback, S. E. Van Aken, T. V. Feldblyum,M. D’Ascenzo, W. L. Deng, A. R. Ramos, J. R. Alfano, S. Cartinhour, A. K.Chatterjee, T. P. Delaney, S. G. Lazarowitz, G. B. Martin, D. J. Schneider,X. Tang, C. L. Bender, O. White, C. M. Fraser, and A. Collmer. 2003. Thecomplete genome sequence of the Arabidopsis and tomato pathogenPseudomonas syringae pv. tomato DC3000. Proc. Natl. Acad. Sci. USA 100:10181–10186.

7. Challis, G. L., and J. Ravel. 2000. Coelichelin, a new peptide siderophoreencoded by the Streptomyces coelicolor genome: structure prediction from thesequence of its non-ribosomal peptide synthetase. FEMS Microbiol. Lett.187:111–114.

8. Challis, G. L., J. Ravel, and C. A. Townsend. 2000. Predictive, structure-based model of amino acid recognition by nonribosomal peptide synthetaseadenylation domains. Chem. Biol. 7:211–224.

9. Conti, E., T. Stachelhaus, M. A. Marahiel, and P. Brick. 1997. Structuralbasis for the activation of phenylalanine in the non-ribosomal biosynthesis ofgramicidin S. EMBO J. 16:4174–4183.

10. Cooper, D. G., C. R. Macdonald, S. J. Duff, and N. Kosaric. 1981. Enhancedproduction of surfactin from Bacillus subtilis by continuous product removaland metal cation additions. Appl. Environ. Microbiol. 42:408–412.

11. Cuppels, D. A., and T. Ainsworth. 1995. Molecular and physiological char-acterization of Pseudomonas syringae pv. tomato and Pseudomonas syringaepv. maculicola strains that produce the phytotoxin coronatine. Appl. Envi-ron. Microbiol. 61:3530–3536.

12. de Bruijn, I., M. J. de Kock, M. Yang, P. de Waard, T. A. van Beek, and J. M.Raaijmakers. 2007. Genome-based discovery, structure prediction and func-tional analysis of cyclic lipopeptide antibiotics in Pseudomonas species. Mol.Microbiol. 63:417–428.

13. Feil, H., W. S. Feil, P. Chain, F. Larimer, G. DiBartolo, A. Copeland, A.Lykidis, S. Trong, M. Nolan, E. Goltsman, J. Thiel, S. Malfatti, J. E. Loper,A. Lapidus, J. C. Detter, M. Land, P. M. Richardson, N. C. Kyrpides, N.Ivanova, and S. E. Lindow. 2005. Comparison of the complete genomesequences of Pseudomonas syringae pv. syringae B728a and pv. tomatoDC3000. Proc. Natl. Acad. Sci. USA 102:11064–11069.

14. Grant, S. G., J. Jessee, F. R. Bloom, and D. Hanahan. 1990. Differentialplasmid rescue from transgenic mouse DNAs into Escherichia coli methyla-tion-restriction mutants. Proc. Natl. Acad. Sci. USA 87:4645–4649.

15. Grgurina, I., M. Bensaci, G. Pocsfalvi, L. Mannina, O. Cruciani, A. Fiore, V.Fogliano, K. N. Sorensen, and J. Y. Takemoto. 2005. Novel cyclic lipodep-sipeptide from Pseudomonas syringae pv. lachrymans strain 508 and syringo-peptin antimicrobial activities. Antimicrob. Agents Chemother. 49:5037–5045.

16. Grgurina, I., and F. Mariotti. 1999. Biosynthetic origin of syringomycin andsyringopeptin 22, toxic secondary metabolites of the phytopathogenic bacte-rium Pseudomonas syringae pv. syringae. FEBS Lett. 462:151–154.

17. Grgurina, I., F. Mariotti, V. Fogliano, M. Gallo, A. Scaloni, N. S. Iacobellis,P. Lo Cantore, L. Mannina, V. van Axel Castelli, M. L. Greco, and A.Graniti. 2002. A new syringopeptin produced by bean strains of Pseudomo-nas syringae pv. syringae. Biochim. Biophys. Acta 1597:81–89.

18. Grunewald, J., and M. A. Marahiel. 2006. Chemoenzymatic and template-directed synthesis of bioactive macrocyclic peptides. Microbiol. Mol. Biol.Rev. 70:121–146.

19. Guerra-Santos, L., O. Käppeli, and A. Fiechter. 1984. Pseudomonas aerugi-nosa biosurfactant production in continuous culture with glucose as carbonsource. Appl. Environ. Microbiol. 48:301–305.

20. Hoang, T. T., R. R. Karkhoff-Schweizer, A. J. Kutchma, and H. P. Schweizer.1998. A broad-host-range Flp-FRT recombination system for site-specificexcision of chromosomally-located DNA sequences: application for isolationof unmarked Pseudomonas aeruginosa mutants. Gene 212:77–86.

21. Hojati, Z., C. Milne, B. Harvey, L. Gordon, M. Borg, F. Flett, B. Wilkinson,P. J. Sidebottom, B. A. Rudd, M. A. Hayes, C. P. Smith, and J. Micklefield.2002. Structure, biosynthetic origin, and engineered biosynthesis of calcium-dependent antibiotics from Streptomyces coelicolor. Chem. Biol. 9:1175–1187.

22. Jain, D. K., D. K. C. Thompson, H. Lee, and J. T. Trevors. 1991. A drop-collapsing test for screening surfactant producing microorganisms. J. Micro-biol. Methods 13:271–279.

23. Joardar, V., M. Lindeberg, R. W. Jackson, J. Selengut, R. Dodson, L. M.Brinkac, S. C. Daugherty, R. Deboy, A. S. Durkin, M. G. Giglio, R. Madupu,W. C. Nelson, M. J. Rosovitz, S. Sullivan, J. Crabtree, T. Creasy, T. Davidsen,D. H. Haft, N. Zafar, L. Zhou, R. Halpin, T. Holley, H. Khouri, T. Feldblyum,O. White, C. M. Fraser, A. K. Chatterjee, S. Cartinhour, D. J. Schneider, J.Mansfield, A. Collmer, and C. R. Buell. 2005. Whole-genome sequence analysisof Pseudomonas syringae pv. phaseolicola 1448A reveals divergence amongpathovars in genes involved in virulence and transposition. J. Bacteriol. 187:6488–6498.

24. Kalogeraki, V. S., and S. C. Winans. 1997. Suicide plasmids containingpromoterless reporter genes can simultaneously disrupt and create fusions totarget genes of diverse bacteria. Gene 188:69–75.

25. Keating, T. A., D. E. Ehmann, R. M. Kohli, C. G. Marshall, J. W. Trauger,and C. T. Walsh. 2001. Chain termination steps in nonribosomal peptide



http://jb.asm.org/

Dow

nloaded from

http://jb.asm.org/

synthetase assembly lines: directed acyl-S-enzyme breakdown in antibioticand siderophore biosynthesis. ChemBioChem 2:99–107.

26. King, E. O., M. K. Ward, and D. E. Raney. 1954. Two simple media for thedemonstration of pyocyanin and fluorescin. J. Lab. Clin. Med. 44:301–307.

27. Konz, D., A. Klens, K. Schorgendorfer, and M. A. Marahiel. 1997. Thebacitracin biosynthesis operon of Bacillus licheniformis ATCC 10716: molec-ular characterization of three multi-modular peptide synthetases. Chem.Biol. 4:927–937.

28. Krause, M., and M. A. Marahiel. 1988. Organization of the biosynthesisgenes for the peptide antibiotic gramicidin S. J. Bacteriol. 170:4669–4674.

29. Kuiper, I., E. L. Lagendijk, R. Pickford, J. P. Derrick, G. E. Lamers, J. E.Thomas-Oates, B. J. Lugtenberg, and G. V. Bloemberg. 2004. Characteriza-tion of two Pseudomonas putida lipopeptide biosurfactants, putisolvin I andII, which inhibit biofilm formation and break down existing biofilms. Mol.Microbiol. 51:97–113.

30. Kulathila, R., K. A. Merkler, and D. J. Merkler. 1999. Enzymatic formationof C-terminal amides. Nat. Prod. Rep. 16:145–154.

31. Lautru, S., R. J. Deeth, L. M. Bailey, and G. L. Challis. 2005. Discovery ofa new peptide natural product by Streptomyces coelicolor genome mining.Nat. Chem. Biol. 1:265–269.

32. Leclere, V., R. Marti, M. Bechet, P. Fickers, and P. Jacques. 2006. Thelipopeptides mycosubtilin and surfactin enhance spreading of Bacillus subtilisstrains by their surface-active properties. Arch. Microbiol. 186:475–483.

33. Lindum, P. W., U. Anthoni, C. Christophersen, L. Eberl, S. Molin, and M.Givskov. 1998. N-Acyl-L-homoserine lactone autoinducers control produc-tion of an extracellular lipopeptide biosurfactant required for swarmingmotility of Serratia liquefaciens MG1. J. Bacteriol. 180:6384–6388.

34. Matsuyama, T., A. Bhasin, and R. M. Harshey. 1995. Mutational analysis offlagellum-independent surface spreading of Serratia marcescens 274 on alow-agar medium. J. Bacteriol. 177:987–991.

35. May, J. J., N. Kessler, M. A. Marahiel, and M. T. Stubbs. 2002. Crystalstructure of DhbE, an archetype for aryl acid activating domains of modularnonribosomal peptide synthetases. Proc. Natl. Acad. Sci. USA 99:12120–12125.

36. May, J. J., T. M. Wendrich, and M. A. Marahiel. 2001. The dhb operon ofBacillus subtilis encodes the biosynthetic template for the catecholic sid-erophore 2,3-dihydroxybenzoate-glycine-threonine trimeric ester bacillibac-tin. J. Biol. Chem. 276:7209–7217.

37. Meyer, J. M., A. Neely, A. Stintzi, C. Georges, and I. A. Holder. 1996.Pyoverdin is essential for virulence of Pseudomonas aeruginosa. Infect. Im-mun. 64:518–523.

38. Miao, V., R. Brost, J. Chapple, K. She, M. F. Gal, and R. H. Baltz. 2006. Thelipopeptide antibiotic A54145 biosynthetic gene cluster from Streptomycesfradiae. J. Ind. Microbiol. Biotechnol. 33:129–140.

39. Mittal, S., and K. R. Davis. 1995. Role of the phytotoxin coronatine in theinfection of Arabidopsis thaliana by Pseudomonas syringae pv. tomato. Mol.Plant-Microbe Interact. 8:165–171.

40. Mootz, H. D., and M. A. Marahiel. 1997. The tyrocidine biosynthesis operonof Bacillus brevis: complete nucleotide sequence and biochemical character-ization of functional internal adenylation domains. J. Bacteriol. 179:6843–6850.

41. Morikawa, M., H. Daido, T. Takao, S. Murata, Y. Shimonishi, and T.Imanaka. 1993. A new lipopeptide biosurfactant produced by Arthrobactersp. strain MIS38. J. Bacteriol. 175:6459–6466.

42. Nakano, M. M., M. A. Marahiel, and P. Zuber. 1988. Identification of agenetic locus required for biosynthesis of the lipopeptide antibiotic surfactinin Bacillus subtilis. J. Bacteriol. 170:5662–5668.

43. Nelson, K. E., C. Weinel, I. T. Paulsen, R. J. Dodson, H. Hilbert, V. A.Martins dos Santos, D. E. Fouts, S. R. Gill, M. Pop, M. Holmes, L. Brinkac,M. Beanan, R. T. DeBoy, S. Daugherty, J. Kolonay, R. Madupu, W. Nelson,O. White, J. Peterson, H. Khouri, I. Hance, P. Chris Lee, E. Holtzapple, D.Scanlan, K. Tran, A. Moazzez, T. Utterback, M. Rizzo, K. Lee, D. Kosack, D.Moestl, H. Wedler, J. Lauber, D. Stjepandic, J. Hoheisel, M. Straetz, S.Heim, C. Kiewitz, J. A. Eisen, K. N. Timmis, A. Dusterhoft, B. Tummler, andC. M. Fraser. 2002. Complete genome sequence and comparative analysis ofthe metabolically versatile Pseudomonas putida KT2440. Environ. Microbiol.4:799–808.

44. Nielsen, T. H., C. Christophersen, U. Anthoni, and J. Sorensen. 1999. Vis-cosinamide, a new cyclic depsipeptide with surfactant and antifungal prop-erties produced by Pseudomonas fluorescens DR54. J. Appl. Microbiol. 87:80–90.

45. Nielsen, T. H., C. Thrane, C. Christophersen, U. Anthoni, and J. Sorensen.2000. Structure, production characteristics and fungal antagonism of ten-sin—a new antifungal cyclic lipopeptide from Pseudomonas fluorescens strain96.578. J. Appl. Microbiol. 89:992–1001.

46. Paulsen, I. T., C. M. Press, J. Ravel, D. Y. Kobayashi, G. S. Myers, D. V.Mavrodi, R. T. DeBoy, R. Seshadri, Q. Ren, R. Madupu, R. J. Dodson, A. S.Durkin, L. M. Brinkac, S. C. Daugherty, S. A. Sullivan, M. J. Rosovitz, M. L.Gwinn, L. Zhou, D. J. Schneider, S. W. Cartinhour, W. C. Nelson, J.Weidman, K. Watkins, K. Tran, H. Khouri, E. A. Pierson, L. S. Pierson III,L. S. Thomashow, and J. E. Loper. 2005. Complete genome sequence of theplant commensal Pseudomonas fluorescens Pf-5. Nat. Biotechnol. 23:873–878.

47. Prentki, P., and H. M. Krisch. 1984. In vitro insertional mutagenesis with aselectable DNA fragment. Gene 29:303–313.

48. Quail, J. W., N. Ismail, M. S. Pedras, and S. M. Boyetchko. 2002. Pseu-dophomins A and B, a class of cyclic lipodepsipeptides isolated from aPseudomonas species. Acta Crystallogr. C 58:o268–o271.

49. Quandt, J., and M. F. Hynes. 1993. Versatile suicide vectors which allowdirect selection for gene replacement in gram-negative bacteria. Gene 127:15–21.

50. Raaijmakers, J. M., I. de Bruijn, and M. J. de Kock. 2006. Cyclic lipopeptideproduction by plant-associated Pseudomonas spp.: diversity, activity, biosyn-thesis, and regulation. Mol. Plant-Microbe Interact. 19:699–710.

51. Rondon, M. R., K. S. Ballering, and M. G. Thomas. 2004. Identification andanalysis of a siderophore biosynthetic gene cluster from Agrobacterium tu-mefaciens C58. Microbiology 150:3857–3866.

52. Roongsawang, N., K. Hase, M. Haruki, T. Imanaka, M. Morikawa, and S.Kanaya. 2003. Cloning and characterization of the gene cluster encodingarthrofactin synthetase from Pseudomonas sp. MIS38. Chem. Biol. 10:869–880.

53. Roongsawang, N., K. Washio, and M. Morikawa. 2007. In vivo characteriza-tion of tandem C-terminal thioesterase domains in arthrofactin synthetase.ChemBioChem 8:501–512.

54. Schäffer, A. A., L. Aravind, T. L. Madden, S. Shavirin, J. L. Spouge, Y. I.Wolf, E. V. Koonin, and S. F. Altschul. 2001. Improving the accuracy ofPSI-BLAST protein database searches with composition-based statistics andother refinements. Nucleic Acids Res. 29:2994–3005.

55. Schwarzer, D., R. Finking, and M. A. Marahiel. 2003. Nonribosomal pep-tides: from genes to products. Nat. Prod. Rep. 20:275–287.

56. Silakowski, B., H. U. Schairer, H. Ehret, B. Kunze, S. Weinig, G. Nordsiek,P. Brandt, H. Blöcker, G. Höfle, S. Beyer, and R. Müller. 1999. New lessonsfor combinatorial biosynthesis from myxobacteria. The myxothiazol biosyn-thetic gene cluster of Stigmatella aurantiaca DW4/3-1. J. Biol. Chem. 274:37391–37399.

57. Simon, R., U. Priefer, and A. Pühler. 1983. A broad host range mobilizationsystem for in vivo genetic engineering: transposon mutagenesis in gram-negative bacteria. Bio/Technology 1:784–791.

58. Sorensen, D., T. H. Nielsen, C. Christophersen, J. Sorensen, and M.Gajhede. 2001. Cyclic lipoundecapeptide amphisin from Pseudomonas sp.strain DSS73. Acta Crystallogr. C 57:1123–1124.

59. Stachelhaus, T., and M. A. Marahiel. 1995. Modular structure of peptidesynthetases revealed by dissection of the multifunctional enzyme GrsA.J. Biol. Chem. 270:6163–6169.

60. Stachelhaus, T., H. D. Mootz, and M. A. Marahiel. 1999. The specificity-conferring code of adenylation domains in nonribosomal peptide syntheta-ses. Chem. Biol. 6:493–505.

61. Stover, C. K., X. Q. Pham, A. L. Erwin, S. D. Mizoguchi, P. Warrener, M. J.Hickey, F. S. Brinkman, W. O. Hufnagle, D. J. Kowalik, M. Lagrou, R. L.Garber, L. Goltry, E. Tolentino, S. Westbrock-Wadman, Y. Yuan, L. L. Brody,S. N. Coulter, K. R. Folger, A. Kas, K. Larbig, R. Lim, K. Smith, D. Spencer,G. K. Wong, Z. Wu, I. T. Paulsen, J. Reizer, M. H. Saier, R. E. Hancock, S. Lory,and M. V. Olson. 2000. Complete genome sequence of Pseudomonas aeruginosaPA01, an opportunistic pathogen. Nature 406:959–964.

62. Turgay, K., M. Krause, and M. A. Marahiel. 1992. Four homologous do-mains in the primary structure of GrsB are related to domains in a super-family of adenylate-forming enzymes. Mol. Microbiol. 6:2743–2744.

63. Vollenbroich, D., N. Mehta, P. Zuber, J. Vater, and R. M. Kamp. 1994.Analysis of surfactin synthetase subunits in srfA mutants of Bacillus subtilisOKB105. J. Bacteriol. 176:395–400.

64. Yarmosh, D. F., and R. C. Ferniany. 1977. Modified bromthymol blue stainfor chromatographic detection of phospholipids. Clin. Chem. 23:614–615.



http://jb.asm.org/

Dow

nloaded from
http://jb.asm.org/

Documents

Identification of a Biosynthetic Gene Cluster and the Six ... · mology used to synthesize these molecules is conserved. This conservation comes from NRPSs consisting of a set of