Boronated tartrolon antibiotic produced by symbioticcellulose-degrading bacteria in shipworm gillsSherif I. Elshahawia,1, Amaro E. Trindade-Silvab, Amro Hanorac, Andrew W. Hana, Malem S. Floresd, Vinicius Vizzonib,Carlos G. Schragob, Carlos A. Soaresb, Gisela P. Concepciond, Dan L. Distele, Eric W. Schmidtf,and Margo G. Haygooda,2

aDivision of Environmental and Biomolecular Systems, Institute of Environmental Health, Oregon Health and Science University, Beaverton, OR 97006;bDepartamento de Genética, Universidade Federal do Rio de Janeiro, 21944-970, Rio de Janeiro, Brazil; cDepartment of Microbiology and Immunology,Faculty of Pharmacy, Suez Canal University, Ismailia, 41522, Egypt; dThe Marine Science Institute, University of the Philippines-Diliman, Quezon City, 1101,Philippines; eOcean Genome Legacy, Ipswich, MA 01938; and fDepartment of Medicinal Chemistry, University of Utah, Salt Lake City, UT 84112

Edited by Bonnie L. Bassler, Princeton University and Howard Hughes Medical Institute, Princeton, NJ, and approved December 3, 2012 (received for reviewAugust 22, 2012)

Shipworms are marine wood-boring bivalve mollusks (familyTeredinidae) that harbor a community of closely related Gammap-roteobacteria as intracellular endosymbionts in their gills. Thesesymbionts have been proposed to assist the shipworm host incellulose digestion and have been shown to play a role in nitrogenfixation. The genome of one strain of Teredinibacter turnerae, thefirst shipworm symbiont to be cultivated, was sequenced, reveal-ing potential as a rich source of polyketides and nonribosomalpeptides. Bioassay-guided fractionation led to the isolation andidentification of two macrodioloide polyketides belonging to thetartrolon class. Both compounds were found to possess antibacte-rial properties, and the major compound was found to inhibit othershipworm symbiont strains and various pathogenic bacteria. Thegene cluster responsible for the synthesis of these compounds wasidentified and characterized, and the ketosynthase domains wereanalyzed phylogenetically. Reverse-transcription PCR in addition toliquid chromatography and high-resolution mass spectrometry andtandem mass spectrometry revealed the transcription of thesegenes and the presence of the compounds in the shipworm, sug-gesting that the gene cluster is expressed in vivo and that thecompounds may fulfill a specific function for the shipworm host.This study reports tartrolon polyketides from a shipworm symbiontand unveils the biosynthetic gene cluster of a member of this classof compounds, which might reveal the mechanism by which thesebioactive metabolites are biosynthesized.

symbiosis | biosynthesis | natural products | acyl-transferase | cecum

Marine bivalve mollusks of the family Teredinidae (commonlyknown as shipworms) seem to rely on their gill symbionts to

survive in their unusual environment. They comprise a diverse,cosmopolitan group that is well known for the ability to burrowinto wood, causing damage to wooden ships and other manmadestructures in marine and brackish waters (1). As in other animalsthat consume wood, it is thought that the shipworm’s microbialsymbionts facilitate the degradation of lignocellulose, which oth-erwise is difficult for animals to digest (2, 3). Multiple geneticallydistinct but closely related Gammaproteobacterial symbionts existin the gill (ctenidia) of shipworm species examined to date (4).Teredinibacter turnerae, a cultivated shipworm symbiont species,has been isolated from different shipworm hosts collected aroundthe world (5). T. turnerae secretes lignocellulose-degradingenzymes thought to assist the host in wood decomposition. Thesymbiotic bacteria live in the gill of the shipworms, but cellulose isdegraded in the digestive tract in a specific organ known as thececum. Although the gut is an excellent habitat for microbes inmost xylophagous organisms, the cecum in shipworms containsvery few bacteria (6). This absence of microbes is striking, becausecellulose digestion liberates glucose, which is an excellent nutrientsource for microbes. The genome of one strain, T. turnerae T7901,was sequenced and revealed, in addition to genes that encodeenzymes specific for lignocellulose degradation and nitrogen fix-ation, at least nine regions that encode enzymes for the biosynthesis

of polyketides and nonribosomal peptides (7). We therefore hy-pothesized that some of the secondary metabolites produced bythe shipworm symbiont T. turnerae might contribute to reducingthe bacterial population in the cecum to prevent glucose scav-enging. Moreover, secondary metabolites might play a significantrole in microbial competition among symbionts in the gill. Here,we describe the polyketide tartrolons, antibiotics that are pro-duced by T. turnerae and were detected in whole shipworm ani-mals. These and other antibiotics from shipworm symbionts mayhelp structure the symbiont community, possibly even enabling theunique lignocellulose digestion strategy found in shipworms. Thisstudy reports the secondary metabolites identified from T. turneraeand their bioactivities, describes the biosynthetic gene clusterlinked to them, and presents evidence that these metabolites areproduced in the symbiotic state.

ResultsGrowth Conditions, Isolation, and Identification of the AntibacterialMetabolites. Growth of T. turnerae T7901 in modified shipwormbasal medium (SBM) (8) with sucrose as the carbon source and withreduced inorganic phosphate was found to cause the highest anti-bacterial activity against Bacillus subtilis using a disk diffusion assay.T. turnerae was grown in a medium that optimized antibiotic

production, and a combined liquid culture of 24 L was extracted(Supporting Information). Antibacterial activity-guided fractionationled to isolation of the previously reported tartrolon D (compound1) and its boronated derivative tartrolon E (compound 2) (9)(Table 1 and Fig. 1). Although in the initial report of compound 1the authors noted the presence of a boronated derivative that wasrelated to compound 2, they did not purify or fully characterize thecompound (Figs. S1–S5). Other boronated tartrolon derivativeswere present in the extracts, as determined by MS (Fig. S6), buttheir quantities were too small to be characterized.

Bioactivities of Tartrolons. Compounds 1 and 2 inhibited B. subtilisin disk diffusion assays (Table S1). Furthermore, compound 2 wasfound to have significant antibacterial activity against Pseudo-monas aeruginosa in addition to methicillin-sensitive and methi-cillin-resistant Staphylococcus aureus (Table 2). All T. turnerae

Author contributions: S.I.E., A.E.T.-S., and M.G.H. designed research; S.I.E., A.E.T.-S., A.H.,A.W.H., M.S.F., V.V., and C.G.S. performed research; D.L.D. contributed new reagents/analytic tools; S.I.E., A.E.T.-S., A.H., A.W.H., M.S.F., V.V., C.G.S., C.A.S., G.P.C., D.L.D.,E.W.S., and M.G.H. analyzed data; and S.I.E., E.W.S., and M.G.H. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1Present address: Department of Pharmaceutical Sciences, College of Pharmacy, University ofKentucky, Lexington, KY 40536.

2To whom correspondence should be addressed. E-mail: haygoodm@ebs.ogi.edu.

See Author Summary on page 1153 (volume 110, number 4).

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1213892110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1213892110 PNAS | Published online January 3, 2013 | E295–E304

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

strains tested were resistant (n = 3). Compound 2 also inhibitedthe growth of the marine organism Vibrio anguillarum in additionto one (BS02) of eight shipworm symbionts selected from ourOcean Genome Legacy (Ipswich, MA) culture collection (TableS1 and Fig. S7) representing three groups of phylogeneticallydistinct symbiont strains (6). The minimum inhibitory concen-tration (MIC) of compound 2 against B. subtilis was determinedto be 1 μg/mL (1.1 μM), but no inhibition was detected againstEscherichia coli at a concentration >32 μg/mL (39 μM). More-over, compound 2 did not show any inhibition activity against thefungus Candida albicans (Table S1). Compound 2 also possessedan IC50 of 2 μM against the breast cancer cell line MCF-7.

Gene Disruption. Nine secondary metabolite gene clusters werefound during the analysis of the genome of T. turnerae T7901,including three encoding polyketide synthase (PKS) genes (7).Analyses of the tartrolon polyketide structure and the PKSregions led to the hypothesis that region 2 was responsible fortartrolon biosynthesis. For confirmation, mutant AH02 wasconstructed by disrupting the KS1 domain of trtD by a single-crossover recombination. High-resolution MS (HR-MS) andtandem MS (MS/MS) of the ethyl acetate fractions of both thewild type and the mutant were compared, showing absence oftartrolons in the mutant (Fig. 2). We therefore propose region 2to be the tartrolon biosynthetic gene cluster, trt.

Analysis of the trt Gene Cluster. The trt cluster was analyzed andfound to be ∼50 kb in length and to contain 20 ORFs possiblyinvolved in the tartrolon biosynthesis (Fig. 3). Of these ORFs, 10

(trtA–trtJ) seem to be the core biosynthetic genes. These ORFs areorganized in a single operon including PKS domains (Table 3).Genes in the cluster had the same orientation, except for trtA, -B,and -C at the 50 end of the cluster. More than 42.5 kb of this regionare formed by trtDEF, three large genes that encode trans acyl-transferase (AT) type I PKSs (10). These three multimodularPKS ORFs contain 11 modules in addition to the loading module.The first PKS is trtD, which is most similar (amino acid similarityof 50%) to dfnG from Bacillus amyloliquefaciens involved in thebiosynthesis of the macrolide difficidin (11, 12). trtD is followedby trtE, the largest ORF in this region, which is similar to anuncharacterized PKS from Clostridium cellulolyticum (Table 3). Italso was found to have a 44% amino acid similarity to baeN,which is involved in the biosynthesis of bacillaene from B. amy-loliquefaciens FZB42 (12, 13). The third PKS ORF, trtF, wassimilar to a PKS from Paenibacillus polymyxa.Other biosynthetic genes are also present, including two hy-

pothetical acyltransferases, trtAB, one of which could act asa proofreader (14); two putative oxidoreductases, trtGI, whichappear to encode proteins for oxygenases, and a putative pol-yketide cyclase, trtJ (15), which may be involved in the cycliza-tion of compound 1. In addition to the integrated thioesterase(TE) in trtF, which is expected to cause release of the elon-gated chain, a standalone type II TE, trtH, also is present andis proposed to cause regeneration of misprimed thiolationdomains (16).The flanking regions for the core biosynthetic genes trtA–J in-

clude 10 more genes that might be involved in the transcription orresistance of tartrolon (Fig. 3). TERTU_2194 and TERTU_2212

Table 1. Tartrolons and structurally related compounds with their bacterial sources and reported biologicalactivities

Compounds Boron Bacterial source Biological activity Source

Boromycin Present Streptomyces antibioticus Antibacterial (58, 59)(terrestrial) Anti-HIV

Aplasmomycin Present Streptomyces griseus Antiplasmodium (27, 60)(marine) Antibacterial

Borophycin Present Cyanobacteria Antibacterial (61, 62)Nostoc linckia CytotoxicNostoc spongiaeforme var. tenue (marine)

Tartrolon A Absent Myxobacterium Sorangium cellulosum (terrestrial) Antibacterial (48, 63)Tartrolon B Present Myxobacterium Sorangium cellulosum (terrestrial) Antibacterial (48, 63)Tartrolon C Present Streptomyces species (terrestrial) Insecticidal activity (64)

Inhibition of HIF-1Tartrolon D Absent Streptomyces species (marine) Cytotoxic (9)

T. turnerae (marine) Antibacterial This studyTartrolon E Present T. turnerae (marine) Antibacterial This study

Fig. 1. Chemical structure of tartrolons and structurally related compounds.

E296 | www.pnas.org/cgi/doi/10.1073/pnas.1213892110 Elshahawi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

function as potential transcription regulators upstream and down-stream, respectively. TERTU_2194 is similar to the transcriptionfactor LysR, whereas TERTU_2212 is similar to a putative Rho-independent transcription terminator (17). TERTU_2190 andTERTU_2191 encode possible aldolases that might be involved inthe synthesis of the loading substrate. A transposase TERTU_2188is encoded by a gene located upstream of the core genes, suggestinga possible lateral gene-transfer event. TERTU_2193 is a potentialacyl carrier protein (ACP) phosphodiesterase required for theturnover of the ACP prosthetic group (18). Four ORFs with nopredicted functions in the tartrolon biosynthesis are also present:TERTU_2189, TERTU_2195, TERTU_2209, and TERTU_2211.No transport-related coding sequences are present in the cluster orwithin the cluster’s neighboring genes.

Acylation of the trt trans-AT PKS Enzymes. The ACP domains ofAT-less PKS modules are suggested to be trans-acylated bymono- or bifunctional AT, where AT-AT or AT-Oxy domaincompositions are commonly seen (10). Of the two discrete ATsfound in the trt cluster trtAB, only trtB seems likely to be in-volved in loading trtDEF PKSs. The gene trtB codes for a 287-amino acid hypothetical protein with high similarity (60%similarity, 46% identity) to the malonyltransferase domain ofthe malonyltransferase/oxidoreductase didomain enzyme DszDin Sorangium cellulosum. TrtB’s hypothetical malonyltransfer-

ase functionality is indicated by the presence of (i) the catalyticdyad Ser92-His201 (numbers from E. coli FabD), (ii) the pre-scribed active binding-site N-terminal motif (P/S/T) QGQC,(iii) recently described key residues for malonyl-CoA substratespecificity in integrated and discrete AT systems, and (iv) themalonyltransferase-active site (GHSxxxR) (Fig. S8A). Addi-tionally, phylogenetic analysis has shown that trtB falls withinthe clade of discrete FAS-like PKS malonyltransferases (Fig.S8B) composed by discrete AT domains that hypothetically areinvolved or biochemically are proven to load malonyl-CoAunits in a variety of trans-AT PKS systems (19).

Evolutionary Rationale Supports the trtDEF Pathway Role in TartrolonsSynthesis.Type I trans-AT PKS ketosynthase (KS) domains usuallygroup according to their substrate specificity when analyzed phy-logenetically (20). Here this evolutionary rationale was appliedto analyze the tartrolon retro-biosynthesis and to investigate thepresence of new clades of KS functionality. The amino acid se-quence of 430 trans-AT PKS-derived KS domains, including the 11KSs from the trtDEF PKS core in addition to 301 KSs with knownsubstrate specificity, were aligned and subjected to phylogeneticanalyses through three different methods. Maximum-likelihood(ML) reconstruction produced the most robust tree architectureplacing 277 (∼92%) of the previously characterized KS domainswithin well-supported clades, with 261 (∼94%) of those KSsgrouped according to their functionality (Fig. 4). The ML treetopology was strikingly similar to the one seen in the phylogeneticanalysis recently reported by Teta et al. (21) for the analysis of theelansolids (els) gene cluster. A total of 33 clades were observed,adding five clades to the 28 previously observed by these authors.Three of these additional clades appear to resolve different sub-strate specificities than those reported previously (21). The cladenamed “double-bond 3” (DB3) grouped the domains RhiKS14and MgsKS2, both of which receive olefinic substrates that aresubject to ring formation by Michael addition-dependent pro-cesses (22). The two other clades directly reflect the addition of

XI-7A-873-CE25 #4-129 RT: 0.03-1.33 AV: 126 NL: 7.73E5T: FTMS + p ESI Full ms2 873.40@cid25.00 [240.00-879.90]

300 400 500 600 700 800m/z

0

20

40

60

80

100

665.

3075

829.

4277

607.

2652

703.

3229

815.

3754

495.

2127

521.

2282

463.

1865

855.

4068

666.

3107

XI-7C-873-CE25 #2-190 RT: 0.01-1.96 AV: 189 NL: 2.99E4T: FTMS + p ESI Full ms2 873.40@cid25.00 [240.00-879.90]

300 400 500 600 700 800m/z

0

20

40

60

80

100

873.

5530

465.

6718

773.

4279

670.

5674

469.

3713

342.

1264

675.

8954

437.

9414

304.

6628

344.

8446

595.

3285

XI-7A-873 #3-216 RT: 0.02-1.95 AV: 214 NL: 3.96E6T: FTMS + p ESI SIM ms [871.90-879.81]

872 874 876 878m/z

0

10

20

30

40

50

60

70

80

90

100 873.4

174

874.4

199

872.4

204

873.5

549

XI-7C-873 #1-216 RT: 0.01-1.95 AV: 216 NL: 9.65E4T: FTMS + p ESI SIM ms [871.90-879.81]

872 874 876 878m/z

0

10

20

30

40

50

60

70

80

90

100 873.

5529

874.

5562

873.

6217

MS/MST. turnerae wild type

AH02 mutant

HR-MS

Fig. 2. HR-MS and tandem MS spectra ofthe crude extracts of each of T. turneraewild type (Upper) and region 2 mutant,AH02 (Lower). The high-resolution signalcorresponding to compound 2 is present inthe wild type but absent from AH02. Thehigh-resolution mass of the main peak inthe wild-type spectrum is 873.4174; thismass is absent in AH02. The mass of themain peak in the mutant AH02 is 873.5529,a mass which also is present in the wildtype as a minor compound, In addition,the MS/MS fragmentation patterns ofthese peaks were different, confirming theabsence of compound 2 in the disruptedmutant.

Table 2. MIC antibacterial activities of tartrolon E againstpathogenic bacteria

Strain tested MIC in microgram per milliliter (μM)

P. aeruginosa 0.31 (0.36)Methicillin-sensitive S. aureus 0.08* (0.095)Methicillin-resistant S. aureus 1.25 (1.14)

*Lowest concentration tested.

Elshahawi et al. PNAS | Published online January 3, 2013 | E297

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

TrtKS domains into the phylogeny. The domain TrtKS1 groupedconsistently with BryKS1 from the bryostatin biosynthetic genecluster (bry) and with BT2KS1 from a Burkholderia thailandensisputative trans-AT PKS (GenBank accession no. ZP_02468762).This clade groups KSs receiving a D-lactate starter unit. Thedomains TrtKS2 and TrtKS11 grouped together with CorKS6,forming a clade of KSs for reduced substrates. This clade seems toresolve KSs specific for rare α-methylated reduced moieties, butnew KS sequences with such specificity need to be analyzed infuture studies to confirm this branching. All other TrtKSs, exceptfor TrtKS6 and TrtKS7, could be associated unequivocally withclades. TrtKS6 and TrtKS7 grouped with KSs for various sub-strates in a phylogenetically unresolved area, also observed pre-

viously (21) and labeled as “mixed” clade (Fig. 4). In fact, TrtKS6was most similar to RhiKS12 (58%), the phylogenetic grouping ofwhich has been reported as inconsistent. The substrate of TrtKS6also is unpredictable by colinearity, because of the nature of KS5,which is a nonextending KS that would deliver only the product ofmodule four. In fact, KS5 grouped with nonextending KSs fromaberrant bimodules denominated as type A (23) common in trans-AT PKSs. TrtKS7, on other hand, follows module 6 containinga ketoreductase as the only β-keto reduction domain and there-fore can be associated with a β-hydroxylated substrate.

Biosynthesis of Tartrolons. Correlation of the bioinformatic anal-ysis of the domains in each module of region 2 with that of the

Fig. 3. Biosynthesis scheme of tartrolons from the trt cluster. Compounds 1 and 2 are color coded based on the corresponding domains. Plasmid pDMrg2KSwas used to disrupt KS1. ACP, acyl carrier protein; AT, acyltransferase; cmr, chloramphenicol resistance gene; DH, dehydratase; GNAT, Gcn 5-N-acetyltransferase; KR, ketoreductase; KS, ketosynthase; M, module; MT, methyl transferase; PKS, polyketide synthase; R, enoyl reductase; TE, thioesterase.

E298 | www.pnas.org/cgi/doi/10.1073/pnas.1213892110 Elshahawi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

tartrolon chemical substructure, in addition to mutational anal-ysis, confirmed that region 2 is the gene cluster responsible forthe biosynthesis of tartrolons. We propose a route to the bio-synthesis of tartrolons (Fig. 3). Tartrolon biosynthesis starts bythe loading of the three-carbon unit, D-lactate. A conjugateddiene formed at C14–C17 is predicted to be formed by the KS2-KR-ACP-KS3-DH-ACP of modules 2 and 3 through a stutteringmechanism reported previously in the biosynthesis of otherconjugated dienes such as kalimantacin C (24), chivosazole (25),macrolactin (26), and difficidin (11). The bimodule M4-M5 loadsa saturated intermediate that is transferred by the nonextendingTrtKS5. Modules 6, 7, and 8 are responsible for the formation ofthe β-dihydroxy ketone region at C7-C11, and module 9 formsa saturated derivative. In addition, a C-MT domain in module 9appears to load the methyl group at C4. KS10, similar to KS5appears to be nonfunctional, as predicted from the KS phylog-eny, whereas KS11 forms the ketone group at C1. We proposethat the two putative oxygenases trtG and I located downstreamof the KSs are responsible for the formation of the acidic hy-droxyl group at C2. The ketone group at C3 is reduced to a hy-droxyl group by the oxygen at C7 to form a pyran ring andleading to the formation of an α,β-dihydroxy acid moiety at C2and C3 that is followed by the dimerization of two identicalmonomers to form a molecule with four hydroxyl groups capableof forming borate ester. Compound 1 was easily transformed toits boron ester derivative, compound 2, simply by the addition ofboric acid, as evidenced by TLC and MS, suggesting that theunboronated dimer binds boron in a Boesken complex form (27)without the need of an enzymatic reaction. This notion also issupported by the conversion of deboronated derivatives ofstructurally related compounds to the boronated derivativethrough the addition of boric acid (28).

Prevalence of Tartrolons in Different Strains of T. turnerae. To de-termine whether compound 2, the major tartrolon in T. turnerae,

is produced only in T. turnerae T7901 or is widespread in theTeredinibacter clade, we examined 11 additional T. turneraestrains that were reported previously (5) from different Ter-edinidae host species from different environments.Cultures of the 12 T. turnerae strains were grown under the

same conditions and extracted followed by HR-MS and MS/MSanalyses. T. turnerae T7901, in addition to the other 11 strains,was analyzed under the same conditions. The analysis showedthat compound 2 is present in at least 8 of the 12 strains tested(Table 4). However, strains that did not show the presence oftartrolons might contain a trt gene cluster within the genome thatis silent (not expressed) under these growth conditions. Thus, wesurveyed all the strains for the trt cluster using PCR.DNA extraction of each of the T. turnerae strains followed by

PCR amplification of eight locations in the PKS coding ORFstrtDEF yielded the expected products in most of the reactions forall tested strains (Table 4). The full set of expected ampliconswas obtained for nine strains, reinforcing the idea that the trtgene cluster is prevalent in the Teredinibacter clade. However,the presence of these genes does not confirm the structural in-tegrity of the cluster within the genome or the chemical identityof the product.All but one of the positive MS samples showed strong PCR

evidence of the trt cluster. The one exception, T8203, failed toamplify one fragment in trtF but was positive for all the others.In this case, a slight difference in the gene sequence may bepresent. Two samples (T8602 and CS30) that showed positivePCR amplification for all fragments in the three ORFs did notshow the compound using MS analysis. These two strains, inaddition to the two other strains that did not show the presenceof compound 2 may contain either nonfunctional or silent trtclusters under the specified growth conditions. All the strainsshowed at least some evidence of genes related to trt. Overall,tartrolon production appears to be common among the exam-ined T. turnerae strains.

Table 3. Genes and functions of region 2

Protein Gene name

No. ofaminoacids Proposed function Organism

Identity/similarity(I/S) (%)

Accessionnumber

TERTU_2188 TERTU_2188 519 Transposase Cellvibrio japonicus 68/81 YP_001983638.1TERTU_2189 TERTU_2189 62 Hypothetical proteinTERTU_2190 TERTU_2190 68 Hypothetical protein Saccharophagus degradans 49/63 YP_526685.1TERTU_2191 TERTU_2191 171 Hypothetical protein Saccharophagus degradans 66/78 YP_526685.1TERTU_2193 TERTU_2193 234 ACP phosphodiesterase Idiomarina loihiensis 86/94 YP_155995.1TERTU_2194 TERTU_2194 290 Transcription regulator LysR Marinobacter adhaerens 80/91 ADP97201.1TERTU_2195 TERTU_2195 549 Peptidase Plesiocystis pacifica 39/55 ZP_01910797.1TERTU_2198 trtA 196 Acyltransferase (GNAT) Burkholderia thailandensis 46/69 ZP_05587728.1TERTU_2199 trtB 287 Acyltransferase Streptomyces cattleya 47/62 CCB76280.1TERTU_2200 trtC 236 Lipoprotein/enoyl CoA isomerase

hydrataseActinosynnema mirum 43/61 YP_003099559.1

TERTU_2202 trtD 4,539 PKS-Hyd-KR-FkbH-ACP-KS1-DH-KR-ACP-KS2-KR-ACP-KS3

Bacillus amyloliquefaciens 35/50 YP_001421792.1

TERTU_2203 trtE 4,947 PKS-DH-ACP-KS4-KR-ACP-KS5-ACP-KS6-KR-ACP-KS7-ACP-KS8

Clostridium cellulolyticum 29/47 YP_002505214.1

TERTU_2204 trtF 4,663 PKS-KR-ACP-KS9-DH-KR-MT-ACP-KS10-ACP-KS11-ACP-TE

Paenibacillus polymyxa 30/47 YP_003871371

TERTU_2205 trtG 377 Oxygenase Dickeya dadantii 44/64 YP_002988660.1TERTU_2206 trtH 268 Thioesterase Xenorhabdus bovienii 31/51 YP_003468018.1TERTU_2207 trtI 455 Dioxygenase Pseudomonas fluorescens 56/72 AAM12912.1TERTU_2208 trtJ 151 Polyketide cyclase Acidithiobacillus ferrivorans 35/55 ZP_08490092.1TERTU_2209 TERTU_2209 51 HypotheticalTERTU_2211 TERTU_2211 486 Glycoside hydrolase Sorangium cellulosum 61/75 YP_001617495.1TERTU_2212 TERTU_2212 378 Transcription regulator AraC Saccharophagus degradans 33/54 YP_527963.1

Size, proposed functions, source of closest homologs, and their accession number and percentage of identity and similarity are determined. AT, acyl-transferase; DH, dehydratase; GNAT, Gcn 5-N-acetyl transferase; ACP, acyl carrier protein; KS, ketosynthase; KR, ketoreductase; MT, methyl transferase; PKS,polyketide synthase; R, enoyl reductase; TE, thioesterase.

Elshahawi et al. PNAS | Published online January 3, 2013 | E299

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

Expression Analysis of trtD in Culture. Quantitative PCR (qPCR)was used to study the expression analysis of trtD in T. turnerae inmedium with normal and low phosphate levels or with iron-starvedconditions. The gene trtD was found to be overexpressed relativeto the housekeeping gene ftsZ when T. turnerae was grown underthe low inorganic phosphate condition (Fig. 5). This result is inagreement with previous reports (29) that low phosphate inducesthe production of secondary metabolites in microorganisms.

Detection of Tartrolons in the Shipworm Host Using Expression Analysisand LC/MS and MS/MS. To determine whether the trt cluster isexpressed in vivo, reverse transcription PCR of the trtDEF PKScore was carried out on gill RNA from the teredinid shipwormLyrodus pedicellatus and produced the expected amplificationproduct (Fig. 6A). In addition, qPCR was performed using specificprimers targeting the trtD mRNA to study the expression of thisgene in three L. pedicellatus shipworm individuals. This analysisrevealed that trtD is expressed in the shipworm gills relative to theprokaryotic cell-division gene ftsZ (Fig. 6B).To determine whether tartrolons can be detected in ship-

worms using HR-MS and MS/MS, whole L. pedicellatus ship-worms were pooled and the organic fraction was extracted.This fraction was analyzed using HR-MS and MS/MS. The char-acteristic peaks of tartrolons were detected (Fig. 7), strongly sug-gesting the expression of tartrolons in vivo.

DiscussionBioactive metabolite symbiosis (30) is a term used to describea symbiotic relationship between organisms based on chemicalcompounds. Usually one of the organisms produces one or moresecondary metabolites that provide a benefit to the host or havethe potential of protecting the host or the rest of the communityfrom environmental threats. Several examples have been repor-ted recently. For example, a wide spectrum of nine antibioticsproduced from a group of symbiotic actinobacteria seems toprotect the host insect from fungal and bacterial pathogens (31).Another example comes from the leaf-cutting ants that protecttheir fungal food by a group of antibacterial and antifungalcompounds produced by actinobacterial symbionts (32). Finally,the bryozoan, Bugula neritina, was found to harbor a Gammap-roteobacterial endosymbiont that was proposed to be the trueproducer of bryostatins that protect the larvae against predators(33, 34). These examples provide evidence that secondarymetabolites sometimes are important in symbiosis.Although shipworm symbionts previously have been shown to

contribute to nitrogen metabolism (35) in the host and have beenproposed to contribute to lignocellulose digestion (7, 8), theirpotential function as producers of secondary metabolites hasnot been proposed or explored. Compound 1 and its boronatedderivative compound 2 (Fig. 1) were isolated from one of theseshipworm symbionts, T. turnerae T7901. Perez et al. (9) identifiedcompound 1 from a marine actinomycete species that is phylo-genetically distant from the Gammaproteobacteria T. turnerae.These authors also reported the boronated derivative of thiscompound as a minor contaminant in their original sample,which we have dubbed “tartrolon E” (compound 2).Tartrolons belong to a group of macrodiolides with well-

known pharmacological activities (Fig. 1 and Table 1). They aredimers or pseudodimers consisting of two polyketide chainsjoined as diesters. Members of this group are nearly identical intheir C-terminal regions, differing primarily in oxidation state

**

*

** *

***

**

**

* *

**

**

**

**

**

**

**

* ***

*

**

*

* *

*

**

**

* **

* *

*

*

***

*

***

*

***

*

*

*

*** *

*

*

**

*

*

*

**

*

** *

** **

**

*

*

*

*

* ***

* *

* *

TrtKS1

TrtKS10

TrtKS9

TrtKS4

TrtKS8

TrtKS11TrtKS2

TrtKS3

TrtKS5

TrtKS6TrtKS7

DB1 following type A bimodule (11, 6/5/0)DB2 (3, 3/0/0)

BryKS1BT2KS1 D-lactate starter (3, 2/1/0)

DB3 ring formation (3,2/1/0)BCERKS2BTPKS2

DB + α-ME (25, 14/8/3)MlnKS2

CCKS2CCKS9

DB4 (52, 28/20/4)

ChiKS10LnmKS1_X

DszKS9_XRhiKS2

VirKS7

KS0 (26, 14/11/1)

D-OH some with β-L-methyl (27, 21/5/1)Keto with or without α-Me (3, 2/1/0)

VirKS4

SGKS11OzmKS4

oxazol

DB or oxazol (25, 15/8/2)

LkcKS3LkcKS2_VII

CorKS8DszKS2_XIII

DifKS1_XIII DB5, mostly short olefinic starters (4, 4/0/0)

acetyl starter (3, 2/1/0)EtnKS15

ElsKS1 aroyl (?)AlbKS1

GUKS1_VIMmpKS5_VI

TaiKS1_VIacetate introduced by GNAT AT

unusual startersEtnKS1

Various starters (16, 8/7/1)

Mixed 1 (8)

Mixed 2 (8)Glycine (3, 2/1/0)MlnKS3_V

BCERKS3BTPKS3

BCERKS1BTPKS1pyran ring (6, 6/0/0)

CorKS3reduced (2, 2/0/0)

DszKS4_IVCorKS1

reduced, acetyl β-branch (3, 2/1/0)reduced or β-γ -double bond (21, 18/3/0)

CorKS5reduced with or without α-Me (3, 3/0/0)

KS0 (10, 8/2/0)MSPKS2

PelKS7RhiKS12

β-MeOH (5, 3/2/0)KS from Clostridium (11)

α-Me, mostly β-OH (36, 24/7/5)

β-branch (12, 9/3/0)β-Me E double bond (8, 7/1/0)

L-OH (18, 16/2/0)amino acid (18, 10/8/0)

cis-AT PKS (8, 7/0/1)PsyKS11

ChiKS1MSPKS6

MSPKS1

KirKS11

ElsKS7DifKS10_XV

EtnKS14

KirKS2

DB, KS2 in type B bimodule (16, 11/4/1)

KS0 (16, 11/4/1)

0.3

KirKS1KirKS6

LkcKS4_VILkcKS1_IV

LkcKS5

β-OH (4, 3/1/0)

Fig. 4. ML-reconstituted tree of full-length unedited KS domains fromtrans-AT PKS enzymes. For clarity, known clades and clades not relevant toTrtKS were collapsed. The KS domains are numbered according to the oc-currence in the gene cluster starting from the 59 end, as previously estab-lished (20). The numbers in parenthesis indicate total number of KS domainsin the clade, number of KS with known function and matching specificity/number of KS with unknown function/number of KS with known functionand mismatching specificity, exactly as previously described (21). Alb, albi-cidin; Bae, bacillaene (Bacillus amyloliquefaciens); Bry and BryX, bryostatin;BBR and BBR2, Brevibacillus brevis NBRC 100599 clusters BBR47_31930-32020and BBR47_39780-39920; BCER, Bacillus cereus BSGC 6E1 cluster; BT2, Bur-kholderia thailandensis MSMB43 cluster; Bat, batumin; BATR, Bacillus atro-phaeus 1942 cluster; BTP, Bacillus thuringiensis pondicheriensis BGSC 4BA1cluster; CACI, Catenulispora acidiphila DSM 44928; CC, CC2, and CC3, Clos-tridium cellulolyticum H10 clusters Ccel_0858-0868, Ccel_2373-2386, andCcel_0965-0980; Chi, chivosazol; Cor, corallopyronin; Dif, difficidin; Dsz, dis-orazol; Els, elansolid; Etn, etnangien; GU, Geobacter uraniireducens Rf4; Kir,

kirromycin; Lkc, lankacidin; Lnm, leinamycin; Mgs, migrastatin; MICAU,Micromonospora aurantiaca ATCC 27029 cluster; Mmp, mupirocin; Mln,macrolactin; MSP, Micromonospora sp. ATCC 39149; Onn, onnamide; Ozm,oxazolomycin; Ped, pederin; Pel, Peltigera membranaceae cluster; PPA, Ple-siocystis pacifica SIR-1 cluster; Psy, psymberin; Rhi, rhizoxin; SBI, Streptomy-ces bingchenggensis BCW-1 cluster; Sor, sorangicin; SG, Streptomyces griseusNPRC 13350 cluster; Ta, myxovirescin; Tai, thailandamide; Trt, tartrolon (T.turnerae T7901); Vir, virginiamycin M.

E300 | www.pnas.org/cgi/doi/10.1073/pnas.1213892110 Elshahawi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

and in chain length. The production of this class of compounds isnot restricted to marine organisms; soil bacteria are capable ofproducing them as well (Table 1). Their biosynthesis is notphylogenetically restricted, because their production was foundin actinobacteria, cyanobacteria, and Deltaproteobacteria, inaddition to the Gammaproteobacterium T. turnerae (Table 1).The prevalence of tartrolons among different classes of bacteriasuggests convergent evolution, a common ancestor, or lateralgene transfer among these species reflecting an importantfunction that this class fulfills. Because of the close structuralrelationship, it is possible that the members of this family arebiosynthetically related as well.Despite the isolation of several members of compounds that

belong to the tartrolon class and the fact that they are producedfrom diverse bacteria, no genes involved in their biosynthesishave been reported. The potential starter unit, D-lactate, mostprobably is derived from pyruvate which originates from glycerolbased on feeding studies of structurally related compounds (28,36, 37). Analyses of domain composition of modular PKSsshowed that TrtKS1, in addition to BT2KS1 from B. thai-landensis and BryKS1 from Candidatus E. sertula (38), also ispreceded by the same catalytic organization, making this KSgroup a new clade. The analysis of the amino acid alignmentshad shown that the motif AxAVI/LAN, presented in KSs rec-

ognizing acetyl and others nonacetyl starters (39), is replaced byDY/LYQIAN in the D-lactate–specific KSs. Two internaldomains, TrtKS3 and TrtKS8, grouped within clades for starterunits (Fig. 4). TrtKS3 appears as an out-group of the clade forolefenic starters, and TrtKS8 grouped into the main clade foracetyl- and nonacetyl-derived starters. Such grouping patternssuggest the possibility that the tartrolon biosynthetic route alsoreleases other tartrolon derivatives as suggested by the detectionof other boronated tartrolons in the crude extract of T. turneraeT7901 (Fig. S6).T. turnerae is thought to play a major role in the shipworm

symbiosis. It is cultivable in relatively simple conditions, and itsgenome contains information enabling this bacterium to havea facultative endosymbiotic or even a free-living lifestyle (7). Todate, however, T. turnerae has been found exclusively in in-tracellular symbiotic association with its teredinid mollusk coun-terparts. The trt gene cluster was found to be expressed in theshipworms as shown by qPCR and reverse transcription PCR,(Fig. 6) and as confirmed by the detection of the compound inthe shipworm by HR-MS (Fig. 7). These results, together with theproduction of compound 2 by other T. turnerae strains isolatedfrom different shipworms that live in various environmental con-ditions (Table 4), strongly suggests that the trt cluster is not silent invivo and has a potential role in the shipworm–microbial symbiosis.Most members of this group of boronated polyketides are

reported to have antibacterial activity. Both compounds 1 and 2inhibited B. subtilis. Moreover, compound 2 inhibited the growthof the marine pathogen V. anguillarum and a shipworm Bankiasetacea isolate, BS02, but not the eukaryote C. albicans (TableS1). This effect suggests that tartrolons could play a role in mi-crobial competition in the shipworm system, possibly targetingopportunistic bacteria while sparing the host (Fig. 8). Compound2 seems to possess a deterrent activity against certain membersof the symbiont community but not others, possibly thus main-taining a population of similar strains within distinct bacterio-cytes. Fluorescence in situ hybridization analysis of shipwormsections showed that similar bacterial phylotypes seem to belocated in specific bacteriocytes separate from other phylotypes(6). Another possibility arises from recent results (6) that showthat the cecum, the wood-digesting organ of shipworms, unlikethat of other most xylophagous animals, has few microbes. An-tibacterial tartrolons produced by the shipworm symbionts in thegills might contribute to bacterial suppression in the cecum. Thissuppression could allow the host to maximize efficient uptake ofthe glucose liberated by the breakdown of lignocellulose (Fig. 8).The mechanism by which products of the symbionts in the gillcould be translocated to the cecum is unknown.

Table 4. Prevalence study of tartrolons in different T. turnerae strains

Strain Host Host sourceHR-MS,MS/MS

trtD trtE trtF

3/4 5/6 7/8 9/10 11/12 13/14 15/16 17/18

T7901 Bankia gouldi North Carolina + + + + + + + + +T7902 Lyrodus pedicellatus California + + + + + + + + +T7903 Teredo navalis Massachusetts + + + + + + + + +T8201 Psiloteredo healdi Maracaibo, Venezuela + + + + + + + + +T8202 Teredo furcifera Hamilton Island, Bermuda + + + + + + + + +T8203 Nototeredo edax (Clapp) Hamilton Island, Bermuda + + + + + + + + —

T8304 Nototeredo edax Andhra Pradesh, India — + + + + + + + —

T8401 Bankia rochi Pakistan — + + + + + + + —

T8402 Teredora malleolus Massachusetts + + + + + + + + +T8506 Teredo triangularis Hawaii + + + + + + + + +T8602 Dicyathifer manni Australia — + + + + + + + +CS30 Neoteredo reynei Brazil — + + + + + + + +

The prevalence of tartrolons was studied using HR-MS and MS/MS fragmentation in addition to PCR. The host shipworm and habitats are reportedelsewhere (5). Primer numbers refer to those in Table S2. Positive data indicate that tartrolons are produced by the strain; negative data indicate that thecompound was not observed, but because of possible confounding factors, they do not imply lack of production.

0

10

20

30

MMS Low Pi Low Fe

% R

elat

ive

Exp

ress

ion

Fig. 5. qPCR of trtF amplified from T. turnerae T7901 SBM broth cultures,the same medium under low inorganic phosphate (low Pi), and iron-starved(low Fe) conditions. Expression values were normalized to ftsZ.

Elshahawi et al. PNAS | Published online January 3, 2013 | E301

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

A common feature of this group of compounds is their abilityto bind boron. Boron exists in the form of borate or orthoborateand is known to play important roles in living organisms (40, 41)but is toxic at high levels (42, 43). Boronated tartrolons havea decreased permeability relative to unchelated borate and thuscould play an important role in the transport of boron. Borontransporters have been reported from other living systems, sug-gesting that its transfer across the cell membrane is regulated byactive transport (44, 45). No homologs for borate transporters

were detected in the genome of T. turnerae T7901. Just as somemicroorganisms have evolved biosynthetic pathways to acquireiron in the form of siderophores, others that lack boron trans-porters might have evolved molecules to facilitate boron trans-port or even to exclude toxic levels of boron. Harris et al. (41)reported siderophores isolated from marine bacteria that bindboron more strongly than iron and suggested that these side-rophores have a role in the detoxification of boron in the ocean.Given that the concentration of boron in the ocean is estimatedto be 400 μM (46), organisms might have evolved a controlmechanism either to decrease its toxicity or to make use of itsabundance. In fact, the crystal structure of the universal quorum-sensing molecule AI-2 bound to its receptor was found to be inthe deboronated form in terrestrial bacteria but in the boronatedform in marine ones (47). Although boron can result from glasscontamination in some cultures (48), such contamination is lesslikely to be the source of boron in marine organisms because ofthe high concentration of boron in the ocean and marine growthmedia. It also is possible that tartrolons serve multiple functions,acting as both as an antibacterial and as a boron transporter inthe shipworm system.In summary, we have isolated and identified two macro-

diolides from the marine shipworm symbiont T. turnerae T7901,compound 1 and its boronated derivative compound 2, that werefound to act as antibacterials. We identified a biosynthetic genecluster that will shed more light on the biosynthesis of otheractive natural products in this class. The biosynthetic scheme waspredicted based on phylogenetic analysis of the KS domains.Moreover, tartrolons were detected in the shipworm host and inother T. turnerae strains, strongly suggesting that it plays a role inthe bioactive metabolite symbiosis of the shipworm.

Materials and MethodsAlignment and Phylogenetic Analyses of KSs. Sequence alignments wereconducted in Clustal W (49). ML tree topology was inferred in PhyML 3.0 (50)using the LG model of sequence evolution (51) with Gamma-distributedrates across sites. Model choice was conducted in ProtTest 2.4 by the likeli-hood ratio test (52). Node support was accessed by the approximate likeli-hood ratio test (aLRT) statistic (53). The Bayesian tree was estimated inMrBayes 3.1 (54) with the WAG + G model of sequence evolution (55), whichwas applied previously (20). Bayesian inference in MrBayes is performed viathe Markov chain Monte Carlo (MCMC) algorithm. Two independent MCMC

B

0

50

100

150

200

250

1 2 3

% R

elat

ive

Expr

essi

on

A

Fig. 6. In vivo expression analysis of the trt cluster in shipworms. (A) Re-verse-transcriptase PCR products with Lyrodus pedicellatus gills total RNAand primer pairs specific for the trtD, trtE, and trtF PKS-coding genes and forthe constitutively expressed ftsZ as positive control (Left). Control PCRreactions with Taq polymerase, primer pairs for ftsZ, and Lyrodus ped-icellatus gills total RNA or DNA (Right). (B) Analysis of trtD in vivo expressionby qPCR using bulk RNA from three different L. pedicellatus shipworms.

HPLC trace of a representative shipworm crude extract

HR-MS of the region at 43 min

MS/MS of the signal at 843

Fig. 7. HPLC, HR-MS, and MS/MS analysis of ship-worm crude extract. HPLC chromatogram of a rep-resentative crude extract of a L. pedicellatusshipworm (Top), HR-MS (Middle), and MS/MS frag-mentation pattern (Bottom) of the peak corre-sponding to that of compound 1 (Fig. S2).

E302 | www.pnas.org/cgi/doi/10.1073/pnas.1213892110 Elshahawi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

runs with four chains each were used to check convergence of parametricestimates. Chains were sampled every 100th cycle for 5,000,000 generations,yielding 50,000 analyzable samples in each run, of which 25%were discardedas burn in. Convergence was checked by the average SD of split frequenciesand by the phylogenetic species recognition (PSR) factor in MrBayes. Addi-

tionally, we used the R package CODA (56) to test convergence using theHeidelberger–Welch test to estimate the effective sample sizes.

Prevalence of Tartrolons in Different T. turnerae Strains. Each strain wasstreaked on agar plates and then inoculated in 50-mL liquid cultures. The sametemperature and SBM medium were used to grow all 12 strains to test for theprevalence of compound 2. For MS analysis; cells were centrifuged at 5,000 × gfor 20 min, collected, and sonicated in 50% (vol/vol) chloroform in methanol.All dry extracts were dissolved in the same volume of methanol and analyzedby HR-MS and MS/MS, and the region between m/z 800–930 was analyzed. Inaddition, MS/MS was used to fragment the characteristic peaks for compound2. For PCR analysis, DNA was isolated from 12 T. turnerae strains using phenoland chloroform as reported previously (57). PCR reactions were performedwith 10–100 ng of T. turnerae genomic DNA, using Platinum Taq DNA poly-merase (Invitrogen). PCR was used to check the prevalence using primerstargeting each of the three biosynthetic ORFs trtD, -E, and -F (Table S2).

Analysis of Different Shipworms for the Presence of Tartrolons Using MS. In-dividual shipworms of the species L. pedicellatus were extracted from thewood, on ice, using appropriate tools and sterile seawater. Shipworms werewashed with sterile seawater and frozen rapidly at −80 °C. The samples werelyophilized and sonicated three times in 50% (vol/vol) chloroform in meth-anol using a Branson digital Sonifier. The combined organic fraction wasconcentrated under vacuum and partitioned between ethyl acetate andwater. The ethyl acetate fraction was filtered over anhydrous sodium sulfateand dried under vacuum. This fraction then was dissolved in methanol be-fore being injected into LC/MS with a reverse-phase column and mobile-phase water and methanol. Proper controls containing solvents were used todetect any contamination from solvents or instruments.

Other Experiments. Additional experiments are available in the Materials andMethods section of the supporting information.

ACKNOWLEDGMENTS. We thank Dr. David Peyton and the NMR facility atPortland State University for acquiring the NMR data. This work was sup-ported by Grant 1U01 TW008163 from the Philippine Mollusk Symbiont,International Cooperative Biodiversity Group project and the Oregon Op-portunity Fund (to M.G.H.), and by National Science Foundation Grant0920540 (to D.L.D.).

1. Distel DL (2003) The biology of marine wood boring bivalves and their bacterialendosymbionts. Wood Deterioration and Preservation, eds Goodell B, Nicholas D,Schultz T (American Chemical Society, Washington, DC), pp 253–271.

2. Ohkuma M (2003) Termite symbiotic systems: Efficient bio-recycling of lignocellulose.Appl Microbiol Biotechnol 61(1):1–9.

3. Hess M, et al. (2011) Metagenomic discovery of biomass-degrading genes andgenomes from cow rumen. Science 331(6016):463–467.

4. Distel DL, Beaudoin DJ, Morrill W (2002) Coexistence of multiple proteobacterialendosymbionts in the gills of the wood-boring Bivalve Lyrodus pedicellatus (Bivalvia:Teredinidae). Appl Environ Microbiol 68(12):6292–6299.

5. Distel DL, Morrill W, MacLaren-Toussaint N, Franks D, Waterbury J (2002) Teredinibacterturnerae gen. nov., sp. nov., a dinitrogen-fixing, cellulolytic, endosymbiotic gamma-proteobacterium isolated from the gills of wood-boring molluscs (Bivalvia: Teredinidae).Int J Syst Evol Microbiol 52(Pt 6):2261–2269.

6. Betcher MA, et al. (2012) Microbial distribution and abundance in the digestivesystem of five shipworm species (Bivalvia: Teredinidae). PLoS ONE 7(9):e45309.

7. Yang JC, et al. (2009) The complete genome of Teredinibacter turnerae T7901: Anintracellular endosymbiont of marine wood-boring bivalves (shipworms). PLoS ONE 4(7):e6085.

8. Waterbury JB, Calloway CB, Turner RD (1983) A cellulolytic nitrogen-fixing bacteriumcultured from the gland of deshayes in shipworms (Bivalvia: Teredinidae). Science 221(4618):1401–1403.

9. Pérez M, et al. (2009) Tartrolon D, a cytotoxic macrodiolide from the marine-derivedactinomycete Streptomyces sp. MDG-04-17-069. J Nat Prod 72(12):2192–2194.

10. Cheng Y-Q, Tang G-L, Shen B (2003) Type I polyketide synthase requiring a discreteacyltransferase for polyketide biosynthesis. Proc Natl Acad Sci USA 100(6):3149–3154.

11. Chen X-H, et al. (2006) Structural and functional characterization of three polyketidesynthase gene clusters in Bacillus amyloliquefaciens FZB 42. J Bacteriol 188(11):4024–4036.

12. Chen XH, et al. (2009) Genome analysis of Bacillus amyloliquefaciens FZB42 reveals itspotential for biocontrol of plant pathogens. J Biotechnol 140(1-2):27–37.

13. Butcher RA, et al. (2007) The identification of bacillaene, the product of the PksXmegacomplex in Bacillus subtilis. Proc Natl Acad Sci USA 104(5):1506–1509.

14. Jensen K, et al. (2012) Polyketide proofreading by an acyltransferase-like enzyme.Chem Biol 19(3):329–339.

15. Kallio P, Sultana A, Niemi J, Mäntsälä P, Schneider G (2006) Crystal structure of thepolyketide cyclase AknH with bound substrate and product analogue: Implications forcatalytic mechanism and product stereoselectivity. J Mol Biol 357(1):210–220.

16. Xue Y, Zhao L, Liu HW, Sherman DH (1998) A gene cluster for macrolide antibioticbiosynthesis in Streptomyces venezuelae: Architecture of metabolic diversity. ProcNatl Acad Sci USA 95(21):12111–12116.

17. Lesnik EA, et al. (2001) Prediction of rho-independent transcriptional terminators inEscherichia coli. Nucleic Acids Res 29(17):3583–3594.

18. Thomas J, Cronan JE (2005) The enigmatic acyl carrier protein phosphodiesterase ofEscherichia coli: Genetic and enzymological characterization. J Biol Chem 280(41):34675–34683.

19. Lopanik NB, et al. (2008) In vivo and in vitro trans-acylation by BryP, the putativebryostatin pathway acyltransferase derived from an uncultured marine symbiont.Chem Biol 15(11):1175–1186.

20. Nguyen T, et al. (2008) Exploiting the mosaic structure of trans-acyltransferasepolyketide synthases for natural product discovery and pathway dissection. NatBiotechnol 26(2):225–233.

21. Teta R, et al. (2010) Genome mining reveals trans-AT polyketide synthase directedantibiotic biosynthesis in the bacterial phylum bacteroidetes. ChemBioChem 11(18):2506–2512.

22. Lim S-K, et al. (2009) iso-Migrastatin, migrastatin, and dorrigocin production inStreptomyces platensis NRRL 18993 is governed by a single biosynthetic machineryfeaturing an acyltransferase-less type I polyketide synthase. J Biol Chem 284(43):29746–29756.

23. Piel J (2010) Biosynthesis of polyketides by trans-AT polyketide synthases. Nat ProdRep 27(7):996–1047.

24. Mattheus W, et al. (2010) Isolation and purification of a new kalimantacin/batumin-related polyketide antibiotic and elucidation of its biosynthesis gene cluster. ChemBiol 17(2):149–159.

25. Perlova O, Gerth K, Kaiser O, Hans A, Müller R (2006) Identification and analysis of thechivosazol biosynthetic gene cluster from the myxobacterial model strain Sorangiumcellulosum So ce56. J Biotechnol 121(2):174–191.

26. Schneider K, et al. (2007) Macrolactin is the polyketide biosynthesis product of thepks2 cluster of Bacillus amyloliquefaciens FZB42. J Nat Prod 70(9):1417–1423.

27. Nakamura H, Iitaka Y, Kitahara T, Okazaki T, Okami Y (1977) Structure ofaplasmomycin. J Antibiot (Tokyo) 30(9):714–719.

28. Chen TSS, Chang C-J, Floss HG (1981) Biosynthesis of the boron-containing macrolideantibiotic aplasmomycin by Streptomyces griseus. J Am Chem Soc 103(15):4565–4568.

29. Martín JF (2004) Phosphate control of the biosynthesis of antibiotics and othersecondary metabolites is mediated by the PhoR-PhoP system: An unfinished story.J Bacteriol 186(16):5197–5201.

Competing Gill Symbionts

Tartrolon ET. turnerae

Cecum Gill

Transient Microbes

Glucose

B

B

Cellulose

Fig. 8. Potential functions of tartrolon in the shipworm–microbial symbio-sis. Tartrolon is proposed to participate in bacterial inhibition, either in thegills (Right) by inhibiting certain bacterial phylotypes or in the cecum (Left)by preventing microorganisms from scavenging glucose.

Elshahawi et al. PNAS | Published online January 3, 2013 | E303

MICRO

BIOLO

GY

PNASPL

US

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0

30. Hildebrand M, et al. (2004) Approaches to identify, clone, and express symbiont

bioactive metabolite genes. Nat Prod Rep 21(1):122–142.31. Kroiss J, et al. (2010) Symbiotic Streptomycetes provide antibiotic combination

prophylaxis for wasp offspring. Nat Chem Biol 6(4):261–263.32. Schoenian I, et al. (2011) Chemical basis of the synergism and antagonism in microbial

communities in the nests of leaf-cutting ants. Proc Natl Acad Sci USA 108(5):1955–1960.33. Davidson SK, Allen SW, Lim GE, Anderson CM, Haygood MG (2001) Evidence for the

biosynthesis of bryostatins by the bacterial symbiont “Candidatus Endobugula

sertula” of the bryozoan Bugula neritina. Appl Environ Microbiol 67(10):4531–4537.34. Lopanik N, Lindquist N, Targett N (2004) Potent cytotoxins produced by a microbial

symbiont protect host larvae from predation. Oecologia 139(1):131–139.35. Lechene CP, Luyten Y, McMahon G, Distel DL (2007) Quantitative imaging of nitrogen

fixation by individual bacteria within animal cells. Science 317(5844):1563–1566.36. Chen TS, Chang CJ, Floss HG (1980) Biosynthesis of the boron-containing antibiotic

aplasmomycin. Nuclear magnetic resonance analysis of aplasmomycin and

desboroaplasmomycin. J Antibiot (Tokyo) 33(11):1316–1322.37. Lee JJ, et al. (1987) Further studies on the biosynthesis of the boron-containing

antibiotic aplasmomycin. J Am Chem Soc 109(18):5426–5432.38. Sudek S, et al. (2007) Identification of the putative bryostatin polyketide synthase

gene cluster from “Candidatus Endobugula sertula”, the uncultivated microbial

symbiont of the marine bryozoan Bugula neritina. J Nat Prod 70(1):67–74.39. Fisch KM, et al. (2009) Polyketide assembly lines of uncultivated sponge symbionts

from structure-based gene targeting. Nat Chem Biol 5(7):494–501.40. Bennett A, Rowe RI, Soch N, Eckhert CD (1999) Boron stimulates yeast (Saccharomyces

cerevisiae) growth. J Nutr 129(12):2236–2238.41. Harris WR, Amin SA, Küpper FC, Green DH, Carrano CJ (2007) Borate binding to

siderophores: Structure and stability. J Am Chem Soc 129(40):12263–12271.42. Carrano CJ, Schellenberg S, Amin SA, Green DH, Küpper FC (2009) Boron and marine

life: A new look at an enigmatic bioelement. Mar Biotechnol (NY) 11(4):431–440.43. Masumi Y, Yukio Y (1994) Effects of boron on nodule development and symbiotic

nitrogen fixation in soybean plants. Soil Sci Plant Nutr 40(2):265–274.44. Takano J, et al. (2002) Arabidopsis boron transporter for xylem loading. Nature 420

(6913):337–340.45. Takano J, et al. (2010) Polar localization and degradation of Arabidopsis boron

transporters through distinct trafficking pathways. Proc Natl Acad Sci USA 107(11):

5220–5225.46. Bowen HJM (1966) Trace Elements in Biochemistry (Academic Press, London).47. Miller ST, et al. (2004) Salmonella typhimurium recognizes a chemically distinct form

of the bacterial quorum-sensing signal AI-2. Mol Cell 15(5):677–687.

48. Schummer D, Irschik H, Reichenbach H, Höfle G (1994) Antibiotics from glidingbacteria, LVII. Tartrolons: New boron-containing macrodiolides from Sorangiumcellulosum. Liebigs Ann Chem 1994(3):283–289.

49. Thompson JD, Higgins DG, Gibson TJ (1994) CLUSTAL W: Improving the sensitivity ofprogressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res 22(22):4673–4680.

50. Guindon S, Gascuel O (2003) A simple, fast, and accurate algorithm to estimate largephylogenies by maximum likelihood. Syst Biol 52(5):696–704.

51. Le SQ, Gascuel O (2008) An improved general amino acid replacement matrix. MolBiol Evol 25(7):1307–1320.

52. Abascal F, Zardoya R, Posada D (2005) ProtTest: Selection of best-fit models of proteinevolution. Bioinformatics 21(9):2104–2105.

53. Anisimova M, Gascuel O (2006) Approximate likelihood-ratio test for branches: A fast,accurate, and powerful alternative. Syst Biol 55(4):539–552.

54. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference undermixed models. Bioinformatics 19(12):1572–1574.

55. Whelan S, Goldman N (2001) A general empirical model of protein evolution derivedfrom multiple protein families using a maximum-likelihood approach. Mol Biol Evol18(5):691–699.

56. Plummer M, Best N, Cowles K, Vines K (2006) CODA: Convergence diagnosis andoutput analysis for MCMC. R News 6(1):7–11.

57. Sambrook J, Russell DW (2001) Molecular Cloning: A Laboratory Manual (Cold SpringHarbor Lab Press, Cold Spring Harbor,NY).

58. Hütter R, et al. (1967) [The metabolic products of microorganisms. Boromycin]. HelvChim Acta 50(6):1533–1539.

59. Kohno J, et al. (1996) Boromycin, an anti-HIV antibiotic. Biosci Biotechnol Biochem 60(6):1036–1037.

60. Okami Y, Okazaki T, Kitahara T, Umezawa H (1976) Studies on marinemicroorganisms. V. A new antibiotic, aplasmomycin, produced by a streptomyceteisolated from shallow sea mud. J Antibiot (Tokyo) 29(10):1019–1025.

61. Hemscheidt T, et al. (1994) Structure and biosynthesis of borophycin, a new boesekencomplex of boric acid from a marine strain of the blue-green alga Nostoc linckia. J OrgChem 59(12):3467–3471.

62. Banker R, Carmeli S (1998) Tenuecyclamides A-D, cyclic hexapeptides from thecyanobacterium Nostoc spongiaeforme var. tenue. J Nat Prod 61(10):1248–1251.

63. Irschik H, Schummer D, Gerth K, Höfle G, Reichenbach H (1995) The tartrolons, newboron-containing antibiotics from a myxobacterium, Sorangium cellulosum. J Antibiot(Tokyo) 48(1):26–30.

64. Lewer P, Chapin EL, Graupner PR, Gilbert JR, Peacock C (2003) Tartrolone C: A novelinsecticidal macrodiolide produced by Streptomyces sp. CP1130. J Nat Prod 66(1):143–145.

E304 | www.pnas.org/cgi/doi/10.1073/pnas.1213892110 Elshahawi et al.

Dow

nloa

ded

by g

uest

on

Dec

embe

r 22

, 202

0