Synergistic activity of cosecreted natural products fromamoebae-associated bacteriaJohannes Arpa,1, Sebastian Götzea,1, Ruchira Mukherjia, Derek J. Matternb, María García-Altaresc, Martin Klappera,Debra A. Brockd, Axel A. Brakhageb, Joan E. Strassmannd, David C. Quellerd, Bettina Bardle, Karsten Willinge,Gundela Peschele, and Pierre Stallfortha,2

aChemistry of Microbial Communication, Independent Junior Research Group, Leibniz Institute for Natural Product Research and Infection Biology, HansKnöll Institute (HKI), 07745 Jena, Germany; bDepartment of Molecular and Applied Microbiology, Leibniz Institute for Natural Product Research andInfection Biology, HKI, 07745 Jena, Germany; cDepartment of Biomolecular Chemistry, Leibniz Institute for Natural Product Research and Infection Biology,HKI, 07745 Jena, Germany; dDepartment of Biology, Washington University in St. Louis, St. Louis, MO 63130; and eDepartment of Bio Pilot Plant, LeibnizInstitute for Natural Product Research and Infection Biology, HKI, 07745 Jena, Germany

Edited by Jerrold Meinwald, Cornell University, Ithaca, NY, and approved March 1, 2018 (received for review December 15, 2017)

Investigating microbial interactions from an ecological perspective isa particularly fruitful approach to unveil both new chemistry andbioactivity. Microbial predator–prey interactions in particular rely onnatural products as signal or defense molecules. In this context, weidentified a grazing-resistant Pseudomonas strain, isolated from thebacterivorous amoeba Dictyostelium discoideum. Genome analysisof this bacterium revealed the presence of two biosynthetic geneclusters that were found adjacent to each other on a contiguousstretch of the bacterial genome. Although one cluster codes forthe polyketide synthase producing the known antibiotic mupirocin,the other cluster encodes a nonribosomal peptide synthetase leadingto the unreported cyclic lipopeptide jessenipeptin. We describe itscomplete structure elucidation, as well as its synergistic activityagainst methicillin-resistant Staphylococcus aureus, when in combi-nation with mupirocin. Both biosynthetic gene clusters are regulatedby quorum-sensing systems, with 3-oxo-decanoyl homoserine lac-tone (3-oxo-C10-AHL) and hexanoyl homoserine lactone (C6-AHL)being the respective signal molecules. This study highlights the reg-ulation, richness, and complex interplay of bacterial natural productsthat emerge in the context of microbial competition.

nonribosomal peptides | Pseudomonas | structure elucidation |social amoebae | synergy

The endeavor to expand our access to nature’s chemical di-versity is a matter of utmost urgency as most anticanceragents or antibiotics are somehow derived from natural products(1, 2). The advent of multidrug resistance—both in infectiousdiseases and cancer—calls for new potent bioactive naturalproducts. In particular, the use of combinations of the latter hasreceived much attention due to beneficial synergistic effects, aswell as reduced resistance rates (3).By virtue of their exquisite biosynthetic machineries, microor-

ganisms are exceptionally prolific producers of bioactive and ther-apeutically useful natural products (1, 2, 4). Traditional screeningapproaches, in which bacteria are isolated and cultured under a fewdifferent conditions, have led to the identification of a large numberof bioactive natural products. Although this approach was fruitfulduring the golden age of antibiotics (1950s and 60s), it now suffersfrom a major drawback: the prohibitively large rediscovery rate ofalready known compounds (3, 5). If microorganisms are consideredwithin their ecological context, however, the identification of novelmicrobial natural products can be greatly enhanced. A detailedunderstanding of certain ecological niches can thus lead to newchemistry, when brute force screening approaches fail to unveil thetrue metabolic potential of selected bacteria. In particular, micro-organisms that are found in association with other organisms haveled to the discovery of unforeseen structural diversity and in-teresting bioactivities. Examples are numerous (reviewed in refs.6–8) and to name but a few they include: wasp–bacteria (9), ant–bacteria (10), beetle–bacteria (6, 11), bee–bacteria (12), sponge–bacteria (13), and bacteria–algae interactions (14, 15).

In the last few years, research has shown that amoebae–bacteriainteractions are particularly rich sources of natural products (16–18). Associations between amoebae and bacteria range fromhighly antagonistic, when amoebae feed on bacteria or bacteriakill amoebae (17, 18), to highly mutualistic, in the case of primitivefarming (16, 19). However, bacteria are often loosely associatedwith amoebae and little is known about this nominally commensalrelationship. Amoebae are highly bacterivorous organisms thatcan devour large amounts of bacteria, with ingestion rates up to300 h−1 for the model organism Dictyostelium discoideum (20, 21).It is not surprising that amoebae exert a tremendous evolutionaryselection pressure on bacteria that share the same habitat. Thus, itis expected that amoebae-associated bacteria may have adapted tosurviving in the vicinity of the highly bacterivorous predator. Variousmechanisms for bacterial survival have been described: the forma-tion of strong biofilms can hinder predation, enhanced swarmingmotility may allow for evasion, and some bacteria are even able tosurvive within these predators (22). Additionally, the production ofsecondary metabolites provides a very efficient means for bacteria tocombat amoebae, as well as competing bacteria, and thus surviveand even thrive in hostile environments.Herein, we provide a detailed account on one such example

where a bacterium isolated from the close vicinity of the amoebal

Significance

Bacterially produced small molecules are indispensable leads in thedevelopment of antibiotics, anticancer therapeutics, or immuno-modulators. To unveil novel aspects in the biosynthetic potential ofbacteria, a consideration of the ecological context in which theadapted producers thrive is extremely insightful. Here, we describetwo natural products produced by Pseudomonas sp. QS1027, abacterium that resides in the vicinity of the bacterial predatorDictyostelium discoideum. The twometabolites are jessenipeptin, anonribosomal cyclic lipopeptide, and mupirocin, a known polyke-tide antibiotic. Both compounds are quorum-sensing regulated anddisplay potent synergistic inhibitory activity against clinically rele-vant methicillin-resistant Staphylococcus aureus (MRSA).

Author contributions: P.S. designed research; J.A., S.G., R.M., D.J.M., M.G.-A., M.K., D.A.B.,B.B., K.W., and G.P. performed research; J.A., S.G., R.M., D.J.M., M.G.-A., M.K., A.A.B.,J.E.S., D.C.Q., and P.S. analyzed data; and J.A., S.G., M.K., and P.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.

Data deposition: The sequences reported in this paper have been deposited in the DDBJ/ENA/GenBank database (accession no. PHSU00000000, the version described in this paperis version PHSU01000000).1J.A. and S.G. contributed equally to this work.2To whom correspondence should be addressed. Email: pierre.stallforth@leibniz-hki.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplemental.

Published online March 28, 2018.

3758–3763 | PNAS | April 10, 2018 | vol. 115 | no. 15 www.pnas.org/cgi/doi/10.1073/pnas.1721790115

Dow

nloa

ded

by g

uest

on

June

25,

202

1

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1721790115&domain=pdfhttp://www.pnas.org/site/aboutpnas/licenses.xhtmlhttp://www.ncbi.nlm.nih.gov/nuccore/PHSU00000000http://www.ncbi.nlm.nih.gov/nuccore/PHSU01000000mailto:pierre.stallforth@leibniz-hki.dehttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1721790115

predatorD. discoideum, led to the identification of an antimicrobiallyactive cyclic lipopeptide (CLP). Bioinformatics and molecular bi-ological analyses of the biosynthetic gene cluster allowed inferenceabout its regulation and hinted at the cooperative biological activitywith coproduced mupirocin.

ResultsIsolation of a D. discoideum–Associated Bacterium and Genome Analysis.We purified bacteria from fruiting bodies of D. discoideum (Fig. 1)isolated from soil and deer dung collected at Mountain Lake Bi-ological Station in Virginia. We focused on bacteria, which weretransiently associated with amoebae, and tested them for grazingresistance using a D. discoideum edibility assay. Specifically, a

plaque assay was used in which amoebae are added to a lawn ofbacteria on solid growth medium (SM/5) (23). Edible bacteria wereidentified by the appearance of characteristic amoebal grazingplaques (Fig. 1). Only bacteria that proved to be inedible werefurther tested by adding various amounts of comestible Klebsi-ella pneumoniae. We identified one strain (QS1027), which pre-vented amoebal grazing, even in the presence of an excess ofK. pneumoniae—an indication for the production of amoebicidalcompounds.To infer about the metabolic potential of this strongly antago-

nistic bacterial strain, we sequenced its genomic DNA using NextGeneration Illumina sequencing. A draft genome was assembledinto 82 contigs with a total size of circa 7.48 Mbp and a GCcontent of 61%. Preliminary sequence analysis based on threegenes (16S rRNA, gyrB, rpoD) showed similarity to Pseudomonasjessenii. Subsequently, the genome of Pseudomonas sp. QS1027(hereafter referred to as QS1027) was mined for putative bio-synthetic gene clusters (BGCs) using antiSMASH (24). The latterpredicted one polyketide synthase (pks) gene cluster and severalnonribosomal peptide synthetase (nrps) gene clusters.We found a stretch of about 150 kbp that harbored two BGCs, the

pks and a truncated nrps cluster, intersected by a short stretch ofregulatory genes (Fig. 2). The pks BGCwas predicted to code for themupirocin biosynthetic machinery. The antibiotic mupirocin, alsoknown as pseudomonic acid, is produced by a type I PKS along withadditional tailoring enzymes encoded on a 75-kbp stretch (25, 26).The nrps BGC adjacent to the mupirocin cluster appeared to

span over the contig boundaries. Gap closing between nrps clustersthat ended/started at the contig boundary via PCR and Sangersequencing of the amplicons allowed us to obtain the complete nrpsBGC. In total, 19 modules were present, each containing a con-densation (C), adenylation (A), and thiolation (T) domain. Addi-tionally, the presence of a starter C domain indicated the presenceof an N-terminal fatty acid moiety, suggesting a lipopeptide con-sisting of 19 amino acids may be produced by this NRPS.

Identification, Isolation, and Structure Elucidation of Mupirocin andJessenipeptin. To identify the natural products associated withthe two adjacent BGCs, we combined analytical chemistry meth-ods with molecular biology and in silico approaches. Since oneBGC was predicted to code for mupirocin biosynthetic genes (27),we used commercially available mupirocin to identify whetherQS1027 produces this polyketide, when cultured under differentconditions. We identified a peak in the HPLC profile of crude

Fig. 1. Bacterial strain Pseudomonas sp. QS1027 was isolated from Dic-tyostelium discoideum fruiting bodies. QS1027 is not a food source to theamoeba from which it was isolated.

Fig. 2. (A) Gene architecture of the jessenipeptin and mupirocin BGCs, which are found adjacent to each other on the genome of QS1027. (B) Domainorganization of the jessenipeptin BGC based on the antiSMASH prediction. The predicted A-domain substrates as well as the C-domain specificities are shown.CS, starter C domain; C/E, dual condensation–epimerization domain:

LCL, C domains that condense two L-amino acids.

Arp et al. PNAS | April 10, 2018 | vol. 115 | no. 15 | 3759

CHEM

ISTR

YMICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

1

QS1027 extract with both the same retention time and the sameMS profile as bona fide mupirocin. Generation of a gene dis-ruption mutant (Δmup) by deletion of a gene fragment coding fora keto synthase domain in the presumed mupirocin BGC led tothe disappearance of the mupirocin peak in the HPLC profile ofΔmup culture extract (SI Appendix, Figs. S19 and S22). Thus, weunambiguously linked the mupirocin BGC to the production ofmupirocin in QS1027 (Fig. 2).To identify the product of the putative NRPS, we proceeded

to inactivate the respective BGC by generating in-frame deletionof the gene fragment coding for the A domain of the first module(Δjes). Comparison of HPLC profiles of culture extracts ofQS1027 and the deletion mutant led to the identification of apeak absent in the deletion mutant (SI Appendix, Fig. S22) thatwas present in the wild type. To isolate the corresponding naturalproduct, culture conditions were optimized to increase the yieldof the presumed lipopeptide. The producing strain was fermentedon a 30-L scale in HL/5 medium. Extraction of the conditionedmedium with ethyl acetate and chromatographic purification of thecrude extract led to the isolation of 150-mg homogeneous material.High-resolution mass spectrometry (HR-MS) measurements revealeda pseudomolecular ion peak with m/z 1906.1073 [M + H]+, which isconsistent with a molecular formula of C91H148N20O24.By comparing databases, including SciFinder, Reaxys, and

Norine (28), we found no previously reported molecule with thisformula; thus we termed this compound jessenipeptin. A bio-informatics prediction of the A-domain specificities (presumingthat the colinearity rule is obeyed) provided a first assessment ofthe primary sequence of jessenipeptin and indicated the presenceof an N-terminal fatty acid moiety (24, 29). An A-domain analysis,however, did not allow us to distinguish between threonine or 2,3-

dehydro-2-aminobutyric acid (Dhb) as threonine can be dehy-drated to yield Dhb (30). Predicting the presence of functionalepimerases also failed, hence the absolute configuration of theamino acids of jessenipeptin could not be predicted. The sameproblem was previously described for the structure prediction ofsyringopeptin, a different but related CLP (31).A combination of MS/MS studies and chemical degradation

followed by derivatizations allowed us to determine the aminoacid sequence of jessenipeptin (SI Appendix, Figs. S4–S8), as wellas the nature of the N-terminal fatty acid. The latter was iden-tified as (3R)-hydroxy decanoic acid by total hydrolysis of jes-senipeptin followed by conversion of the fatty acid in its methyland Mosher ester. Comparison of the resulting ester with cor-responding synthetic standards using GC-MS allowed for anunambiguous characterization (Fig. 3 and SI Appendix, Fig. S17)(32). In addition, derivatization of the hydrolyzed amino acidswith Marfey’s reagent (33) confirmed the presence of thefollowing amino acids: 2× L-Ala, 4× D-Ala, 1× L-Dab (2,4-Dia-minobutyric acid), 1× D-Ile, 1× D-Leu, 1× L-Phe, 1× D-Pro, 2× D-Ser, 1× D-allo-Thr, and 3× D-Val. A domain analysis, however,predicted the presence of three Thr residues. This hinted at thefact that two Thr were further modified, e.g., to yield Dhb.Although MS/MS data allowed the placement of most of the

residues in the correct order, assigning their absolute configurationproved difficult. In particular, the positions of the four D-alanineand two L-alanine residues in jessenipeptin could not be resolved.We addressed this problem using a combination of partial hydrolysisand stable isotope labeling experiments.A recent publication by Ni et al. (34) describes a scandium(III)

triflate-mediated peptide hydrolysis method, which displays selec-tivity for cleavage of amide bonds adjacent to serine or threonine

Fig. 3. (A) Our approach for the structure elucidation of jessenipeptin. (B) GC-MS traces of derivatized fatty acid moiety (top trace) and traces of thecorresponding synthetic (R)- and (S)-configured 3-hydroxy decanoic acid derivatives (bottom trace). The fatty acid derivative obtained from jessenipeptincoelutes with the respective (R)-configured synthetic counterpart. (C) Scandium triflate-mediated peptide cleavage yields fragments F1 and F2. A singleL-alanine is present in F1. (D) Stable-isotope labeling results in shifts of one mass unit for fragments containing L-alanine in MS/MS experiments. The MSspectra of unlabeled jessenipeptin are depicted in green; the corresponding spectra of the labeled peptide are depicted in red. The mass shifts observed intwo indicative fragments (m/z 776 and 507) are shown.

3760 | www.pnas.org/cgi/doi/10.1073/pnas.1721790115 Arp et al.

Dow

nloa

ded

by g

uest

on

June

25,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1721790115

residues. Hence, we were able to obtain fragment F1, which waspurified by HPLC. To our surprise, we were able to obtain anotherfragment, F2, in which cleavage occurred between the two alanineresidues at position six and seven (Fig. 3). Total hydrolysis offragment F1 and subsequent derivatization using Marfey’s methodclearly indicated the presence of L- and D-alanine in a 1:2 ratio. Infragment F2, however, we could only identify the presence ofD-alanine. This allowed us to assign the configuration of alanine atposition seven as L (Fig. 3 and SI Appendix, Fig. S16).To locate the remaining L-alanine, we made use of stable

isotope labeling. Feeding deuterium labeled alanine (L-alanine-2-d1) allows one to determine whether epimerization of L-Ala toD-Ala has occurred (35). This method relies on a characteristicfeature of bacterial nonribosomal peptides, whereby D-aminoacids are typically installed by first incorporating the respectiveL-amino acid, followed by an epimerization step. A deuterium atthe α-position of the amino acid will be exchanged for a proton, ifepimerization takes place. The mass difference between a protonand a deuterium can easily be monitored by MS/MS measure-ments. By cultivating QS1027 in minimal medium, supplementedwith L-alanine-2-d1, we were able to observe virtually quantitativeincorporation of two deuterium labels into jessenipeptin. MS/MSmeasurements of the respective labeled jessenipeptin showed thetwo L-alanine residues to be located at positions 7 and 16 (Fig. 3and SI Appendix, Figs. S13–S15), which is in accordance with theresults from Marfey’s analysis of fragment F1 and F2.Since the theoretical mass for a linear peptide, based on amino

acids and the fatty acid found by MS/MS experiments, differed by18 mass units from the observed mass, the presence of a lactonering was assumed. Treatment of jessenipeptin with aqueous so-dium hydroxide yielded a compound with a mass matching thecalculated mass for the linear peptide. To determine the size andposition of the lactone ring, we masked the free hydroxyl groups injessenipeptin as silyl ethers, using tert-butyldimethylsilyl triflate(TBS-OTf). The derivatized peptide was analyzed by MS/MS,revealing that either threonine-15 or serine-18 lacks TBS-groups,suggesting lactone formation between one of these residues andthe C terminus. The 1H,1H-correlation spectroscopy experi-ments showed a strong downfield shift for the β-proton of Thr-15, which provides evidence for the lactone ring closure asshown in Fig. 3 (SI Appendix, Fig. S9) (36). Thus, jessenipeptinfalls into the class of cyclic lipodepsipeptides, most of whichbear intriguing bioactivities (37). Cyclic lipopeptides producedby Pseudomonas species are typically classified into six groups(38). Jessenipeptin can be placed in the tolaasin group. Mem-bers of this group are CLPs with 19–25 amino acids, and theycontain lactone rings consisting of five amino acids. Further-more, they often bear unusual amino acids such as Dab andDhb. All these features are met by jessenipeptin.

Biological Activity and Synergy of Mupirocin and Jessenipeptin. SinceQS1027 was isolated from the fruiting bodies of D. discoideum, weevaluated the amoebicidal activities of jessenipeptin and mupir-ocin. Growth inhibition or killing ofD. discoideum was determinedin a liquid culture assay. Although mupirocin was inactive againstD. discoideum, jessenipeptin exhibited an IC50 (D. discoideum) of4 μg/mL (2 μM). We then tested whether D. discoideum couldgraze on Δmup, Δjes, or ΔmupΔjes. None of these strains wereedible to D. discoideum. It is conceivable that other secondarymetabolites of QS1027 may constitute additional virulence factors.Interestingly, however, a broad antimicrobial screen revealed jes-senipeptin to be highly active against MRSA as well as otherpathogenic Gram-positive bacteria (SI Appendix, Table S1). Asboth mupirocin and jessenipeptin appear to have a similar rangeof activity and are produced by the same organism, we testedfor synergistic antimicrobial effects against Enterococcus faecalis,Bacillus subtilis, MRSA, and D. discoideum. Indeed, we couldshow strong synergy between mupirocin and jessenipeptin againstB. subtilis and most importantly against MRSA. We determinedthe fractional inhibitory concentrations of a variety of mupirocin/

jessenipeptin mixtures. Fig. 4 displays the isobolograms that showsynergistic activity for mupirocin and jessenipeptin.

Regulation of Mupirocin and Jessenipeptin Production. Since luxI/luxR-type regulatory genes intersect both the mupirocin and jessenipeptinBGC, we investigated their role in the production of mupirocin andjessenipeptin. These luxR/luxI-type genes typically code for an acylhomoserine lactone (AHL)-based quorum-sensing (QS) system,with the luxI gene coding for an autoinducer synthase and the luxRcoding for the cognate AHL-binding response regulator (39–41).Furthermore, we could also identify a set of luxR- and luxI-typegenes within the mupirocin BGC. Hence, we investigated whetherthese luxI/luxR genes code for QS systems. In particular, weaddressed the question of the nature of the respective AHLs andwhether the systems display any cross talk.To identify the nature of the QS signals as well as their role in the

regulation of mupirocin and jessenipeptin production, we generatedin-frame deletion mutants of these five regulatory genes, as well ascombinations thereof in QS1027. Deletions of any of the threeregulatory genes luxR1, luxR2, and luxI led to a complete suppressionof jessenipeptin production, yet mupirocin production was left un-altered. Thus, we termed these regulatory genes jesI, jesR1, andjesR2. JesR2 has a predicted N-terminal autoinducer binding domainas well as a C-terminal DNA-binding helix-turn-helix (HTH) motif,whereas JesR1 only bears the C-terminal DNA-binding domain. Thefact that both JesRs are required for the production of jessenipeptinhints at the formation of a possible heterodimeric JesR1/JesR2complex, which upon binding of the cognate AHL activates thejessenipeptin BGC. Although the formation of LuxR homodimersis documented (42, 43), heterodimer formation is typically asso-ciated with inactivation rather than activation. Episomal comple-mentation of ΔjesR1 and ΔjesR2 with the respective genes led to arestoration of jessenipeptin production. Jessenipeptin productionin the ΔjesI mutant could also be restored by supplementation ofthe growth medium with C6-AHL (a variety of AHLs were tested,and the response to C6-AHL was the largest, followed by that toC8-AHL; SI Appendix, Fig. S23). This QS signal is also identifiablein the WT supernatant by liquid chromatography (LC)-MS anal-ysis and was found to be absent in the ΔjesI mutant (SI Appendix,Fig. S21).Analogously, we identified the QS signal associated with the

luxI/luxR-type genes in the mupirocin BGC (from here on forwardreferred to as mupI/mupR). Both the ΔmupR and the ΔmupImutant did not produce any mupirocin and jessenipeptin pro-duction remained unaltered. Episomal complementation ofΔmupR with the respective gene restored mupirocin production(SI Appendix, Fig. S28). Supplementation of the growth mediumof ΔmupI mutant with 3-oxo-C10 AHL led to the production ofmupirocin (SI Appendix, Fig. S26), and this signal molecule was

Fig. 4. Mupirocin and jessenipeptin act synergistically against MRSA (A) andB. subtilis (B). Two compounds are considered to act synergistically, if thesum of their fractional inhibitory concentrations (FIC) is below 0.5 (light bluearea). Strong synergy is present if the sum of the FICs is below 0.25 (darkblue area). Compound ratios are shown in brackets (wt/wt, jessenipeptin:mupirocin).

Arp et al. PNAS | April 10, 2018 | vol. 115 | no. 15 | 3761

CHEM

ISTR

YMICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplemental

also identified in the supernatant of WT, but was absent in ΔmupImutant strain (SI Appendix, Fig. S20). These data are in accor-dance with previous reports that mupirocin is under QS control by3-oxo-C10 AHL (44, 45).These results indicate the existence of two independent QS sys-

tems, which regulate the production of jessenipeptin and mupirocin,respectively (Fig. 5). To investigate whether the production of eitherof the two secondary metabolites affects production of the othermetabolite, we compared the production of mupirocin in the Δjesmutant and vice versa. However, no change was observed in theproduction of either mupirocin or jessenipeptin (SI Appendix, Fig.S22). We can conclude that production of either secondary me-tabolite does not affect production of the other one.

DiscussionThe coevolution of bacteria and bacterivorous amoebae in a con-fined space imposes strong selection pressures upon both organisms.Hence, some bacteria have evolved efficient defense strategies toevade amoebal grazing (17, 18, 22). In particular, the productionof amoebicidal secondary metabolites by the bacteria can ensurethe survival of the bacteria in the vicinity of its predator. More-over, it is advantageous for bacteria to live in the vicinity of abacterivorous organism due to the reduction in bacterial com-petitors for the grazing-resistant bacteria. The latter gain greateraccess to resources that need not be shared between other bac-terial species. Bacteria isolated from the habitat of amoebae arebelieved to display chemical diversity that may be larger than inthe absence of bacterivores. Hence, we specifically interrogatedthe biosynthetic capacities of amoeba-associated bacteria. Wewere particularly interested in bacteria that did not display a closemutualistic relationship with amoebae [such as farmed bacteria(19)] but rather those transiently associated with amoebae. Weisolated bacteria from the fruiting bodies of D. discoideum, whichclearly indicated that a direct amoebae–bacteria contact had oc-curred and that these bacteria were able to evade amoeabal pre-dation. A particular strain was further analyzed due to its stronglyamoebicidal phenotype. Genome sequencing and subsequentbioinformatics analysis revealed the strain to belong to the genusPseudomonas. A large number of BGCs could be identified in thegenome of the Gram-negative γ-proteobacterium, and the unusualpresence of two BGCs intersected by a few regulatory genes caughtour attention. We identified one BGC cluster to code for the bio-synthetic machinery of the antibiotic mupirocin. Interestingly, not

many strains that produce mupirocin have been described to date(46). The other BGC coded for a nonribosomal peptide synthetase,which was found to produce the previously unreported cyclic lip-opeptide jessenipeptin. The structure of the latter was fully eluci-dated by a combination of chemical, analytical, and bioinformaticsmethods. The large CLP consists of 19 amino acids including D- andL-amino acids as well as nonproteinogenic amino acids and a (3R)-hydroxy decanoic acid residue. The absolute configuration of mostamino acids in jessenipeptin could be determined via Marfey’smethod (33). The presence of multiple D- and L-alanines, however,rendered its structure elucidation difficult. A recent study showedthat scandium(III) triflate promotes selective cleavage of a peptideadjacent to a serine or threonine residue by means of an N,O-acylrearrangement (34). This was the key to fragment jessenipeptin,which tremendously facilitated subsequent structure elucidation.Eventually, feeding experiments with α-deutero L-alanine (35)combined with MS/MS analyses enabled us to determine the posi-tion of all D- and L-alanine residues within jessenipeptin. Hence, wefully elucidated the structure of jessenipeptin. Since QS1027 wasisolated from the habitat of social amoebae, we tested jessenipeptinand mupirocin with respect to amoebicidal activity against thepredator D. discoideum. Although mupirocin was inactive againstD.discoideum, jessenipeptin exhibited a strong amoebicidal activity(IC50 = 4 μg/mL). Importantly, however, both mupirocin and jes-senipeptin were highly active against MRSA with minimal inhibitoryconcentration values of 0.2 and 3.12 μg/mL, respectively.With both the jessenipeptin and the mupirocin BGC being on a

contiguous stretch on the genome we wondered if their bio-synthetic products acted synergistically. Indeed, mupirocin andjessenipeptin displayed strong synergistic effects against MRSA.Mupirocin is an isoleucyl-tRNA synthetase inhibitor (47), andmany CLP interact with the cell membrane to form pores (48). Itis possible that in this case synergy is caused by jessenipeptin in-creasing the cytosolic concentration of mupirocin.Both BGCs were only intersected by a few regulatory genes of

the luxI/R family. This led to the question of whether both BGCswere coregulated. Generation of in-frame gene deletions of ei-ther of the two jesR or the jesI gene led to the suppression ofjessenipeptin production, yet mupirocin was still produced. LuxI/luxR systems typically code for QS systems. In this case C6-AHLwas identified as the QS signal associated with jessenipeptinproduction. The BGC of mupirocin also revealed luxI/R genes.Deletion of either of these genes led to the suppression of

Fig. 5. Two independent QS systems regulate the production of jessenipeptin and mupirocin, which synergistically inhibit the growth of MRSA.

3762 | www.pnas.org/cgi/doi/10.1073/pnas.1721790115 Arp et al.

Dow

nloa

ded

by g

uest

on

June

25,

202

1

http://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalhttp://www.pnas.org/lookup/suppl/doi:10.1073/pnas.1721790115/-/DCSupplementalwww.pnas.org/cgi/doi/10.1073/pnas.1721790115

mupirocin production, while jessenipeptin production was unal-tered. We identified 3-oxo-C10-AHL as the QS signal. This wasalso reported as the QS signal in the only well-characterizedmupirocin producer (26). Thus, two independent QS systemsare involved in the production of mupirocin and jessenipeptin.Interestingly, in P. aeruginosa two similar QS systems with C4-AHL and 3-oxo-C12 form a hierarchical QS system, which is notthe case for the jessenipeptin–mupirocin QS system.

ConclusionIn summary, our study highlights the structural and functionaldiversity of natural products that can be obtained from mi-crobial associations. Genome mining and detailed structuralanalysis led to the identification and structure determination ofthe nonribosomal peptide jessenipeptin, whose BGC is adja-cent to that of mupirocin. Both secondary metabolites are QSregulated and display synergistic activity against MRSA. As thenature of amoebae–bacteria interactions is multifaceted, we can

expect this diversity to be reflected in an equally rich bacterialsecondary metabolome, which enables bacteria to adapt to coex-isting with amoebae.

Materials and MethodsWe propagated Pseudomonas sp. QS1027 in Luria Bertani (LB) liquid or onsolid (supplemented with 1.5%, wt/wt, agar) medium (Carl Roth) or SM/5 medium (Formedium) (both solid and liquid) at 28 or 22 °C depending onthe assay.

ACKNOWLEDGMENTS. We thank A. Perner and H. Heinecke for MS and NMRmeasurements. We also thank C. Weigel for performing antimicrobial assaysand F. Kloss for useful discussions. We are grateful for financial support fromthe Leibniz Association. This work was supported by Deutsche Forschungsge-meinschaft Grants STA1431/2-1 and SFB1127. An Aventis Foundation Ph.D.fellowship (to M.K.) is acknowledged. M.G.-A. acknowledges financial supportfrom the ERC for a Marie Skłodowska-Curie Individual Fellowship (IF-EF) Proj-ect reference 700036.

1. Newman DJ, Cragg GM (2012) Natural products as sources of new drugs over the30 years from 1981 to 2010. J Nat Prod 75:311–335.

2. Clardy J, Fischbach MA, Walsh CT (2006) New antibiotics from bacterial naturalproducts. Nat Biotechnol 24:1541–1550.

3. Brown ED, Wright GD (2016) Antibacterial drug discovery in the resistance era. Nature529:336–343.

4. Clardy J, Walsh C (2004) Lessons from natural molecules. Nature 432:829–837.5. Lewis K (2013) Platforms for antibiotic discovery. Nat Rev Drug Discov 12:371–387.6. Scott JJ, et al. (2008) Bacterial protection of beetle-fungus mutualism. Science 322:63.7. Beemelmanns C, Guo H, Rischer M, Poulsen M (2016) Natural products from microbes

associated with insects. Beilstein J Org Chem 12:314–327.8. Piel J (2009) Metabolites from symbiotic bacteria. Nat Prod Rep 26:338–362.9. Kroiss J, et al. (2010) Symbiotic Streptomycetes provide antibiotic combination pro-

phylaxis for wasp offspring. Nat Chem Biol 6:261–263.10. Oh D-C, Poulsen M, Currie CR, Clardy J (2009) Dentigerumycin: a bacterial mediator of

an ant-fungus symbiosis. Nat Chem Biol 5:391–393.11. Piel J (2002) A polyketide synthase-peptide synthetase gene cluster from an un-

cultured bacterial symbiont of Paederus beetles. Proc Natl Acad Sci USA 99:14002–14007.

12. Müller S, et al. (2014) Paenilamicin: structure and biosynthesis of a hybrid non-ribosomal peptide/polyketide antibiotic from the bee pathogen Paenibacillus larvae.Angew Chem Int Ed Engl 53:10821–10825.

13. Wilson MC, et al. (2014) An environmental bacterial taxon with a large and distinctmetabolic repertoire. Nature 506:58–62.

14. Seyedsayamdost MR, Wang R, Kolter R, Clardy J (2014) Hybrid biosynthesis of rose-obacticides from algal and bacterial precursor molecules. J Am Chem Soc 136:15150–15153.

15. Seyedsayamdost MR, Case RJ, Kolter R, Clardy J (2011) The Jekyll-and-Hyde chemistryof Phaeobacter gallaeciensis. Nat Chem 3:331–335.

16. Stallforth P, et al. (2013) A bacterial symbiont is converted from an inedible producerof beneficial molecules into food by a single mutation in the gacA gene. Proc NatlAcad Sci USA 110:14528–14533.

17. Klapper M, Götze S, Barnett R, Willing K, Stallforth P (2016) Bacterial alkaloids pre-vent amoebal predation. Angew Chem Int Ed Engl 55:8944–8947.

18. Götze S, et al. (2017) Structure, biosynthesis, and biological activity of the cyclic lip-opeptide anikasin. ACS Chem Biol 12:2498–2502.

19. Brock DA, Douglas TE, Queller DC, Strassmann JE (2011) Primitive agriculture in asocial amoeba. Nature 469:393–396.

20. Steinert M, Heuner K (2005) Dictyostelium as host model for pathogenesis. CellMicrobiol 7:307–314.

21. Chisholm RL, Firtel RA (2004) Insights into morphogenesis from a simple de-velopmental system. Nat Rev Mol Cell Biol 5:531–541.

22. Matz C, Kjelleberg S (2005) Off the hook–how bacteria survive protozoan grazing.Trends Microbiol 13:302–307.

23. Froquet R, Lelong E, Marchetti A, Cosson P (2009) Dictyostelium discoideum: a modelhost to measure bacterial virulence. Nat Protoc 4:25–30.

24. Weber T, et al. (2015) antiSMASH 3.0-a comprehensive resource for the genomemining of biosynthetic gene clusters. Nucleic Acids Res 43:W237–W243.

25. Fuller AT, et al. (1971) Pseudomonic acid: an antibiotic produced by Pseudomonasfluorescens. Nature 234:416–417.

26. Thomas CM, Hothersall J, Willis CL, Simpson TJ (2010) Resistance to and synthesis ofthe antibiotic mupirocin. Nat Rev Microbiol 8:281–289.

27. Shanks RMQ, Caiazza NC, Hinsa SM, Toutain CM, O’Toole GA (2006) Saccharomycescerevisiae-based molecular tool kit for manipulation of genes from gram-negativebacteria. Appl Environ Microbiol 72:5027–5036.

28. Pupin M, et al. (2016) Norine: A powerful resource for novel nonribosomal peptidediscovery. Synth Syst Biotechnol 1:89–94.

29. Ziemert N, et al. (2012) The natural product domain seeker NaPDoS: a phylogenybased bioinformatic tool to classify secondary metabolite gene diversity. PLoS One 7:e34064.

30. Grgurina I, Mariotti F (1999) Biosynthetic origin of syringomycin and syringopeptin22, toxic secondary metabolites of the phytopathogenic bacterium Pseudomonassyringae pv. syringae. FEBS Lett 462:151–154.

31. Scholz-Schroeder BK, Soule JD, Gross DC (2003) The sypA, sypS, and sypC synthetasegenes encode twenty-two modules involved in the nonribosomal peptide synthesis ofsyringopeptin by Pseudomonas syringae pv. syringae B301D. Mol Plant MicrobeInteract 16:271–280.

32. Jenske R, Vetter W (2007) Highly selective and sensitive gas chromatography-elec-tron-capture negative-ion mass spectrometry method for the indirect enantiose-lective identification of 2- and 3-hydroxy fatty acids in food and biological samples.J Chromatogr A 1146:225–231.

33. Marfey P (1984) Determination of D-amino acids. II. Use of a bifunctional reagent, 1,5-difluoro-2,4-dinitrobenzene. Carslberg Res Commun 49:591–596.

34. Ni J, Sohma Y, Kanai M (2017) Scandium(iii) triflate-promoted serine/threonine-selective peptide bond cleavage. Chem Commun (Camb) 53:3311–3314.

35. Bode HB, et al. (2012) Determination of the absolute configuration of peptide naturalproducts by using stable isotope labeling and mass spectrometry. Chemistry 18:2342–2348.

36. Nutkins JC, et al. (1991) Structure determination of tolaasin, an extracellular lip-odepsipeptide produced by the mushroom pathogen Pseudomonas tolaasii paine.J Am Chem Soc 113:2621–2627.

37. Süssmuth RD, Mainz A (2017) Nonribosomal peptide synthesis-principles and pros-pects. Angew Chem Int Ed Engl 56:3770–3821.

38. Gross H, Loper JE (2009) Genomics of secondary metabolite production by Pseudo-monas spp. Nat Prod Rep 26:1408–1446.

39. Keller L, Surette MG (2006) Communication in bacteria: an ecological and evolu-tionary perspective. Nat Rev Microbiol 4:249–258.

40. Lazdunski AM, Ventre I, Sturgis JN (2004) Regulatory circuits and communication inGram-negative bacteria. Nat Rev Microbiol 2:581–592.

41. Venturi V (2006) Regulation of quorum sensing in Pseudomonas. FEMS Microbiol Rev30:274–291.

42. Fuqua C, Greenberg EP (2002) Listening in on bacteria: acyl-homoserine lactone sig-nalling. Nat Rev Mol Cell Biol 3:685–695.

43. Zhu J, Winans SC (2001) The quorum-sensing transcriptional regulator TraR requiresits cognate signaling ligand for protein folding, protease resistance, and di-merization. Proc Natl Acad Sci USA 98:1507–1512.

44. El-Sayed AK, Hothersall J, Thomas CM (2001) Quorum-sensing-dependent regulationof biosynthesis of the polyketide antibiotic mupirocin in Pseudomonas fluorescensNCIMB 10586. Microbiology 147:2127–2139.

45. Hothersall J, et al. (2011) Manipulation of quorum sensing regulation in Pseudomonasfluorescens NCIMB 10586 to increase mupirocin production. Appl MicrobiolBiotechnol 90:1017–1026.

46. Matthijs S, et al. (2014) Antimicrobial properties of Pseudomonas strains producingthe antibiotic mupirocin. Res Microbiol 165:695–704.

47. Hughes J, Mellows G (1978) Inhibition of isoleucyl-transfer ribonucleic acid synthetasein Escherichia coli by pseudomonic acid. Biochem J 176:305–318.

48. Schneider T, Müller A, Miess H, Gross H (2014) Cyclic lipopeptides as antibacterialagents - potent antibiotic activity mediated by intriguing mode of actions. IntJ Med Microbiol 304:37–43.

Arp et al. PNAS | April 10, 2018 | vol. 115 | no. 15 | 3763

CHEM

ISTR

YMICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

June

25,

202

1