Identification and functional analysis of genescontrolling biosynthesis of 2-methylisoborneolMamoru Komatsu*†, Muneya Tsuda‡, Satoshi Omura†§, Hideaki Oikawa‡, and Haruo Ikeda*§

*Kitasato Institute for Life Sciences, Kitasato University, 1-15-1 Kitasato, Sagamihara, Kanagawa 228-8555, Japan; †The Kitasato Institute, 5-9-1 Shirokane,Minato-ku, Tokyo 108-8642, Japan; and ‡Division of Chemistry, Graduate School of Science, Hokkaido University, Sapporo 060-0810, Japan

Contributed by Satoshi Omura, March 12, 2008 (sent for review December 12, 2007)

To identify the genes for biosynthesis of the off-flavor terpenoidalcohol, 2-methylisoborneol (2-MIB), the key genes encodingmonoterpene cyclase were located in bacterial genome databasesby using a combination of hidden Markov models, protein–familysearch, and the sequence alignment of their gene products. Pre-dicted terpene cyclases were classified into three groups: sesquit-erpene, diterpene, and other terpene cyclases. Genes of the ter-pene cyclase group that form an operon with a gene encodingS-adenosyl-L-methionine (SAM)-dependent methyltransferase werefound in genome data of seven microorganisms belonging toactinomycetes, Streptomyces ambofaciens ISP5053, Streptomycescoelicolor A3(2), Streptomyces griseus IFO13350, Streptomyceslasaliensis NRRL3382R, Streptomyces scabies 87.22, Saccharopolys-pora erythraea NRRL2338, and Micromonospora olivasterosporaKY11048. Among six microorganisms tested, S. ambofaciens, S.coelicolor A3(2), S. griseus, and S. lasaliensis produced 2-MIB but M.olivasterospora produced 2-methylenebornane (2-MB) instead.The regions containing monoterpene cyclase and methyltrans-ferase genes were amplified by PCR from S. ambofaciens, S.lasaliensis, and Saccharopolyspora erythraea, respectively, andtheir genes were heterologously expressed in Streptomyces aver-mitilis, which was naturally deficient of 2-MIB biosynthesis byinsertion and deletion. All exoconjugants of S. avermitilis produced2-MIB. Full-length recombinant proteins, monoterpene cyclase andmethyltransferase of S. lasaliensis were expressed at high level inEscherichia coli. The recombinant methyltransferase catalyzedmethylation at the C2 position of geranyl diphosphate (GPP) in thepresence of SAM. 2-MIB was generated by incubation with GPP,SAM, recombinant methyltransferase, and terpene cyclase. Weconcluded that the biosynthetic pathway involves the methylationof GPP by GPP methyltransferase and its subsequent cyclization bymonoterpene cyclase to 2-MIB.

genome mining � methyltransferase � monoterpene cyclase

Terpenoid metabolites, monoterpenes, sesquiterpenes, and diter-penes, have been isolated from terrestrial and marine plants or

from fungi, with only a relatively minor fraction from prokaryotes.Their compounds are used as antibiotics, hormones, flavor or odorconstituents, and pigments. Some of them possess other physiolog-ically or commercially important properties (1, 2). Three terpenoidcompounds (Fig. 1), 2-methylisoborneol (2-MIB), geosmin, andalbaflavenone are known as odorous and volatile microbial me-tabolites. The former two terpenoid alcohols are the most fre-quently found secondary metabolites of actinomycetes (3–5), fila-mentous cyanobacteria (6–8), myxobacteria (9), and fungi (10, 11),and account for many odor problems encountered with freshwateror with fish (7, 12–14). Geosmin is also known to contribute to thecharacteristic earthy red beet flavor (15). 2-MIB is related to themusty-earthy notes in Brie and Camembert cheese flavor (16). An�,�-unsaturated sesquiterpene ketone, albaflavenone, was isolatedfrom highly odorous Streptomyces culture and is an unusual odorousmetabolite with antibacterial activity (17). The structures of both2-MIB and geosmin originally isolated from actinomycetes weredetermined by Gerber (18, 19) and Medsker et al. (20). Geosmin isa degraded sesquiterpenoid alcohol, and biochemical and molec-

ular genetic studies of its biosynthesis have been carried out inactinomycete strains Streptomyces coelicolor A3(2) (21) and Strep-tomyces avermitilis (22). Another sesquiterpene cyclase generatingepi-isozizaene that could be an intermediate of albaflavenonebiosynthesis, epi-isozizaene synthase (SCO5222p) of S. coelicolorA3(2), has been characterized (23).

Plant monoterpene, sesquiterpene, and diterpene synthases havestrongly conserved amino acid sequences, as well as similar intronorganization and exon sizes, suggesting a common evolutionaryorigin (24). This feature has allowed homology-based physical andbioinformatic searching methods to be very successful in identifyingnew terpene synthase genes from not only plant but also fungi (25).Conversely, the amino acid sequence of microbial sesquiterpenecyclase (pentalenene synthase, germacradienol/geosmin synthase,and epi-isozizaene synthase) has shown no significant similarity toany other known sesquiterpene synthase. Indeed, microbial terpenesynthases in general show no overall sequence similarity either toone another (except for orthologs synthesizing the same terpeneproduct) or to any other protein. This divergence in primarysequence has thwarted attempts to prospect for additional micro-bial terpene synthases by techniques such as Southern hybridizationor PCR amplification using consensus nucleotide sequences (26).Microbial sesquiterpene cyclases have been investigated, but infor-mation of microbial diterpene and monoterpene cyclases is limitedand not only biochemical but also genetic approaches of microbialmonoterpene cyclases have not been elucidated to date.

Labeling experiments conducted by Bentley and Meganathan(27) supported earlier assumptions that 2-MIB is a methylatedmonoterpene alcohol, the additional methyl-group being derivedfrom S-adenosyl-L-methionine (SAM). Recent feeding experi-ments suggested that the methylation of geranyl diphosphate (GPP)

Author contributions: M.K. and M.T. contributed equally to this work; S.O� ., H.O., and H.I.designed research; M.K., M.T., H.O., and H.I. performed research; H.O. and H.I. analyzeddata; and H.I. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

§To whom correspondence may be addressed. E-mail: omura-s@kitasato.or.jp or ikeda@ls.kitasato-u.ac.jp.

This article contains supporting information online at www.pnas.org/cgi/content/full/0802312105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

Fig. 1. Structures of microbial volatile terpenoid metabolites, 2-MIB (Left;1,2,7,7-tetramethyl-exo-bicycloheptan-2-ol), geosmin (Center; 4,8a-dimethyl-octahydro-naphthalen-4a-ol) and albaflavenone (Right; 2,6,7,7-tetramethyl-tricyclo [6.2.1.01,5]undec-5-en-4-one).

7422–7427 � PNAS � May 27, 2008 � vol. 105 � no. 21 www.pnas.org�cgi�doi�10.1073�pnas.0802312105

would generate the substrate for the cyclization to form 2-MIB (9).However, no further direct evidence on the biosynthesis of 2-MIBin actinomycetes, cyanobacteria, myxobacteria, and fungi is avail-able. More than 590 kinds of microbial genome analyses have beencompleted. Although existing databases contain above microor-ganisms, many of their protein-coding genes are still annotated ashypothetical proteins. Consequently, we assumed that genes con-nected with the biosynthesis of 2-MIB would be buried in thegenome data. Because the amino acid sequences of microbialterpene cyclase bear little significant similarity, it was unlikely thatscreening biosynthetic genes for 2-MIB by sequence similaritieswould prove fruitful. However, if polypeptides have similar catalyticactivities, they would have conserved sequence signature. Wetherefore adopted a searching method based on hidden Markovmodels Pfam search (28, 29) for the primary selection of desiredterpene cyclases. Profile hidden Markov models are more suitablefor sensitive database searching using statistical descriptions of asequence family’s consensus. The terpene cyclases predicted wereclassified by phylogenetic analysis. Final candidates were harvestedas the gene formed an operon with SAM-dependent methyltrans-ferase gene, because 2-MIB is a methylated monoterpenoid alcoholand methyl residue at the C2 position of 2-MIB is derived from themethyl residue of methionine in feeding experiments (9, 27).

We have identified two genes encoding 2-MIB biosynthesis fromexisting microbial genome databases, and they have been charac-terized. Here, we elucidate the biosynthetic mechanism fromisopentenyl diphosphate (IPP) to 2-MIB in actinomycetes.

Results and DiscussionTerpene Cyclases in Bacteria. The production of monoterpenoidmetabolites, 2-MIB and so on, from microbial origin was firstreported many years ago, but the biosynthesis of the microbialmonoterpenoid metabolites is still not understood. Microbial ses-quiterpene cyclases, pentalenene synthase, germacradienol/geosmin synthase and epi-isozizaene synthase, and diterpene cy-clase were identified from actinomycete strains and have beencharacterized. The amino acid sequences of these microbial terpenecyclases have shown no significant similarities to that of plant origin.However, a significant conserved motif containing an acid-richdomain, associated with binding of catalytically essential magne-sium ions, was found in these microbial terpene cyclases.

To elucidate the biosynthesis of the microbial monoterpenoidalcohol, 2-MIB, all microbial terpene cyclases were harvested fromthe genome-sequenced bacterial database of National Center forBiotechnology Information, Streptomyces scabies and Streptomycesgriseus genome data, Micromonospora olivasterospora draft se-quence data, and limited sequence data from a linear plasmid pKSLof Streptomyces lasaliensis [putative terpene cyclase and methyl-transferase genes were found near the genes for modular polyketidesynthases (30)] by searching on the basis of the profile hiddenMarkov models using a model of PF03936 (terpene synthase family,metal binding domain). The E-value used was �10�5 for setting theparameter of hidden Markov models search because proteinsselected using �10�5 E-value were also selected by other Pfammodels with lower E-values. Of 1,922,990 proteins, 41 proteins wereselected with E-value ranges from 2.2 � 10�6 to 1.8 � 10�82. Theseproteins were classified into three major groups on the basis ofphylogenetic analysis (Fig. 2). The sesquiterpene cyclases, pen-talenene, germacradienol/geosmin and epi-isozizaene synthases,were classified in group II. Group III contained diterpene cyclasein terpentecin biosynthesis of Kitasatospora griseola (31) and unde-fined terpene cyclase of Salinispora arenicola. The proteins classi-fied into group I were uncharacterized, and the sequence similar-ities of these proteins against known sesquiterpene and diterpenecyclases were low level. Sequence alignment with sesquiterpene andditerpene cyclases revealed that the proteins of group I also had twosignificant conserved motifs containing a metal-binding domain(32–37).

As shown in Fig. 3, two conserved motifs of the proteins in groupI were located near the C terminus, in comparison with those ofsesquiterpene and diterpene cyclases. The distance between thefirst and second conserved motifs of the proteins in group I (112 aa)was longer than those of sesquiterpene and diterpene cyclases(104–106 aa). The first motif in sesquiterpene cyclases was acid-richdomain with a high proportion of aromatic amino acids, –FFxxD-DxxD– (pentalenene and epi-isozizaene synthases) or –FxFD-DHFLE– (germacradienol/geosmin synthase). Although the firstmotif in diterpene cyclase (–LIVNDDRWD–) and the proteins ofgroup I (–xVDDxxx[DE]–) also possessed an acid-rich domain, thecontent of aromatic amino acids was lower than that in sesquiterpnecyclases. The second motif in all proteins were conserved, –xxNxxx-SxxxE–, in which the triad of residues in bold has also beenimplicated in the binding of magnesium ion (33, 34, 36). Because theSAV (deduced amino acid sequence of truncated terpene cyclase;1,245,680 to 1,246,412 nt of the S. avermitilis genome) was definedas a pseudogene of S. avermitilis, the annotation was not completed.The deduced polypeptide was very similar to SAML0357 andSCO7700, but the second motif sequence was lacked in this deducedpolypeptide. According to phylogenic analysis, we assumed that theproteins in group I are monoterpene cyclases and that 2-MIBsynthase has to be found in this group.

Because feeding experiments with labeled precursors suggestedthat 2-MIB was methylated monoterpenoid alcohol, the biosynthe-sis should involve at least two steps, methylation of GPP andcyclization of methylated GPP. In general, the genes involvingsecondary metabolite biosynthesis in bacteria are found as a clusterin a specific location of the genome and some of these genes forma single transcriptional unit. Consequently, it suggests that the genesencoding monoterpene cyclase for 2-MIB biosynthesis form anoperon with a specific GPP methyltransferase. The genes flankingthe monoterpene cyclase gene of group I were annotated andfunctions determined by several bioinformatics analyses. Eachmonoterpene cyclase gene from Streptomyces ambofaciens, Strep-tomyces coelicolor A3(2), S. griseus, S. lasaliensis, S. scabies(SCAB5041), Saccharopolyspora erythraea, and M. olivasterosporaformed an operon with a gene encoding SAM-dependent methyl-transferase (Fig. 4). Furthermore, genes encoding cyclic nucleotide-binding protein were also located upstream of these seven mono-terpene cyclase genes. On the other hand, methyltransferasegene(s) was/were not found in or around monoterpene cyclasegenes predicted in Pseudomonas fluorescens and S. scabies(SCAB82161). These two monoterpene cyclases would be involvedin other monoterpenoid metabolite biosynthesis. Interestingly, asimilar methyltransferase gene, SAV983, was found in S. avermitilisgenome data (http://avermitilis.ls.kitasato-u.ac.jp/). Furthermore,truncated cyclic nucleotide-binding protein and monoterpene cy-clase genes were found upstream of SAV983. The truncation willbe induced by a deletion mutation, and deduced polypeptide oftruncated gene product lacked the second conserved metal-bindingmotif (Fig. 3 Upper). Once S. avermitilis would be capable ofproducing 2-MIB, but the deletion of the region encoding thesecond conserved metal-binding motif in monoterpene cyclasegene prevented production of 2-MIB in this organism.

Production of Volatile Terpenoid Metabolites in Actinomycetes Car-rying Predicted Monoterpene Cyclase/Methyltransferase Genes. Theresults of bioinformatics indicate that seven actinomycetes possessthe ability to produce the monoterpenoid metabolite, 2-MIB.Because a plant pathogen S. scabies 87.22 was not stored in ourculture stock, six actinomycete strains were examined. All six strainsproduced terpenoid metabolites, including sesquiterpenoid alcohol,geosmin, in varying quantity (Fig. 5). Five strains, S. ambofaciens,S. coelicolor A3(2), S. griseus, S. lasaliensis, and Sa. erythraea,produced a monoterpenoid metabolite that had identical retentiontime and mass spectrum ([M�], m/z 168) to those of the authentic2-MIB. Interestingly, M. olivasterospora did not produce 2-MIB, but

Komatsu et al. PNAS � May 27, 2008 � vol. 105 � no. 21 � 7423

BIO

CHEM

ISTR

Y

homomonoterpene hydrocarbon ([M�], m/z 150) was accumulatedas a major component. The hydrocarbon was identical to 2-meth-ylenebornane (2-MB), a dehydration product of 2-MIB in compar-ison with retention time and mass spectrum of the synthetic sample.Then inspection of mass spectra obtained from the extracts of allStreptomyces strains and Sa. erythraea allowed us to detect as aminor component. Thus, the monoterpene cyclase of M. olivaster-ospora is slightly different from those of other actinomycete strains.

The maximum production of 2-MIB and the corresponding2-MB in each microorganism depended on the growth period (Fig.5 A and B). Significant increase of monoterpene production inmicroorganisms except for M. olivasterospora was accomplished bygrowth on soy flour-mannitol (SFM) medium. All Streptomyces andM. olivasterospora reached the maximal production at 5–9 days ofgrowth, but Sa. erythraea required a longer growth period formaximum production (Fig. 5A). This microorganism did not de-velop morphological differentiation until 7 days of growth, in whichthe mycelium grew as substrate mycelium, and sporulation startedafter this growth period. The morphological differentiation of otheractinomycetes except for M. olivasterospora was completed within 5days of growth. Furthermore, abundant sporulation was observedby the growth on the SFM medium rather than on inorganicsalts-starch-yeast extract (M4YE) medium. All known bacterialproducers of 2-MIB and geosmin (i.e., actinomycetes, cyanobacte-

ria, and myxobacteria) exhibit complex morphological differentia-tion, including the formation of multicellular complexes. The aerialmycelium-negative mutants of Streptomyces antibioticus and Strep-tomyces sulfurous failed to produce both 2-MIB and geosmin (27).These observations assumed that complex morphological differen-tiation in these microorganisms compromised production of vola-tile terpenoid metabolites.

It has been reported that telomeric regions in S. ambofaciens arefrequently deleted by induction of mutagenesis (38). Analysisrevealed that monoterpene cyclase (SAML0357) and methyltrans-ferase (SAML0358) genes were located near the left telomere.Once we had isolated mutants of S. ambofaciens (39), all mutantsinduced by UV irradiation were screened by PCR using primers forSAML0357 and SAML0358. One of the mutants, U1717R, wasdeficient of both genes, although the complex morphological dif-ferentiation was not affected. As expected, the mutant strainU1717R was unable to produce 2-MIB but produced another typeof sesquiterpenoid metabolite, geosmin (Fig. 5). This feature wasthe same as that of S. avermitilis that was naturally deficient of agene encoding monoterpene cyclase (Fig. 5). Absence of 2-MIBproduction in both S. ambofaciens mutant U1717R and S. aver-mitilis was presumed to be due to impact of monoterpene cyclasesin group I (Figs. 2 and 3), which are involved in the biosynthesis of2-MIB.

Fig. 2. Phylogenetic analysis of terpene cyclases from bacterial databases. Abbreviations: alr4685 (322 aa; NP�488725), Nostoc sp. PCC 7120; Ava�1982 (322 aa;YP�322499), Anabaena variabilis ATCC 29413; BURPS1106A�A1634 (370 aa; YP�001075668), B. pseudomallei 1106a; BURPS1710b�A0219 (394 aa; YP�335378), B.pseudomallei 1710b; BURPS668�A0947 (339 aa; YP�001061946) and BURPS668�A1721 (370 aa; YP�001062716), B. pseudomallei 668; FRAAL�1336 (338 aa;YP�711586) and FRAAL�6507 (758 aa; YP�716636), Frankia alni ACN14a; Francci3�4231 (751 aa; YP�483306), Frankia sp. CcI3; Franean1�5559 (750 aa;YP�001509819), Frankia sp. EAN1pec; MOL (400 aa), M. olivasterospora KY11048; MXAN�6247 (755 aa; YP�634376), Myxococcus xanthus DK 1622; Pfl�1841 (335aa; YP�347573), Pseudomonas fluorescens PfO-1; Rcas�0622 (326 aa; YP�001430766), Roseiflexus castenholzii DSM 13941; RoseRS�3509 (326 aa; YP�001277817),Roseiflexus sp. RS-1; Rxyl�0493 (324 aa; YP�643279), R. xylanophilus DSM 9941; SACE�3187 (758 aa; YP�001105388), SACE�3722 (354 aa; YP�001105919), SACE�3977(732 aa; YP�001106173) and SACE�4907 (763 aa; YP�001107098), Sa. erythraea NRRL2338; SAML0357 (440 aa; CAJ89344) and SAMR0831 (343 aa; CAJ88540), S.ambofaciens ISP5053; SAV76 (335 aa; NP�821250), SAV2163 (725 aa; NP�823339), SAV2998 (336 aa; NP�824174) and SAV3032 (363 aa; NP�824208), S. avermitilisMA-4680; SCAB20121 (735 aa), SCAB5041 (440 aa), SCAB73741 (338 aa) and SCAB82161 (349 aa), S. scabies 87.22; SCO5222 (361 aa; NP�629369), SCO6073 (726aa; NP�630182) and SCO7700 (440 aa; NP�733742), S. coelicolor A3(2); SGR1269 (437 aa), SGR2079 (335 aa), SGR6065 (339 aa) and SGR6839 (737 aa), S. griseusIF013350; SLA (481 aa), S. lasaliensis NRRL3382R; BAB39207 (311 aa; BAB39207), Kitasatospora griseola diterpene cyclase; Q55012 (337 aa; Q55012), Streptomycessp. UC5319 pentalenene synthase; ZP�01648856 (298 aa; YP�001536181), Salinispora arenicola CNS-205 terpene synthase.

7424 � www.pnas.org�cgi�doi�10.1073�pnas.0802312105 Komatsu et al.

Heterologous Expression of Monoterpene Cyclase/MethyltransferaseGenes. Because S. avermitilis was deficient of a gene encodingmonoterpene cyclase and its operon was naturally disrupted by anIS-insertion and deletions, so the microorganism is suitable for theheterologous expression of the foreign monoterpene cyclase genes.One of the large-deletion mutants, SUKA16, was used as a hoststrain for heterologous expression because the mutant lacks bio-synthetic gene clusters for the main products of S. avermitilis and forterpenoid metabolites. Segments carrying predicted monoterpenecyclase and methyltransferase genes were amplified from twoStreptomyces strains, S. ambofaciens and S. lasaliensis, and twonon-streptomycete strains, Sa. erythraea and M. olivasterospora. Thetwo Streptomyces produced large amounts of 2-MIB. The 2-MB-generating monoterpene cyclase of M. olivasterospora also provedinteresting, as did the smallest size of monoterpene cyclase genefound in Sa. erythraea. These amplified segments were joined justdownstream of the rpsJ promoter, which was constitutively ex-

pressed and had very strong transcriptional activity in S. avermitilis.Exoconjugants of S. avermitilis SUKA16 carrying monoterpenecyclase/methyltransferase genes of S. ambofaciens and S. lasaliensis,respectively, were grown in agar medium, and the extracts of wholeculture were directly analyzed by GC-MS analyzer. Unfortunately,exoconjugant carrying the genes of M. olivasterospora did not exhibitmonoterpenoid metabolites, but three exoconjugants carrying thegenes of S. ambofaciens, S. lasaliensis, and Sa. erythraea, respectively,gave a single major product, 2-MIB (2.8 �g per plate, 1.1 �g perplate, and 65 �g per plate, respectively; supporting information (SI)Fig. S1). The exoconjugant carrying the genes from Sa. erythraea didproduce a very large amount of 2-MIB accompanied with smallamounts of three corresponding hydrocarbons, 2-MB, 2-methyl-2-bornene (2-M2B), and 1-methylcamphene (1-MC), were also pro-duced (Fig. S2). When only the monoterpene cyclase gene of Sa.erythraea was introduced into S. avermitilis, resultant exoconjugantsproduced none of the monoterpenoid metabolites, indicating thatthe methyltransferase is essential for the biosynthesis of 2-MIB.

Fig. 3. Alignment of amino acid sequences of bacterial terpene cyclases with predicted cyclases. Shadow boxes indicate metal (Mg2�)-binding motif of terpenecyclase. The strain name of each protein is described in Fig. 2. SAV indicates deduced amino acid sequence of the truncated terpene cyclase gene of S. avermitilis.SCO5222 and SAV3032, Q55012 and SAV2998, SCO6073 and SAV2163, and BAB39207 were characterized as epi-isozizaene synthases, pentalenene synthases,germacradienol/geosmin synthases, and diterpene cyclase for terpentecin biosynthesis, respectively.

Fig. 4. Organization of genes encoding predicted monoterpene cyclases and flanking genes. All predicted monoterpene cyclase genes are located in thechromosomes, but the gene of S. lasaliensis resided in a giant linear plasmid pKSL. The grayed, filled, and oblique-lined arrows (or boxes) indicate cyclicnucleotide-binding protein, monoterpene cyclase, and methyltransferase genes, respectively. The opened arrows are hypothetical or other function protein-coding genes. S. scabies 87.22(1) and -(2) indicate regions around SCAB5041 and SCAB82161, respectively.

Komatsu et al. PNAS � May 27, 2008 � vol. 105 � no. 21 � 7425

BIO

CHEM

ISTR

Y

The results of heterologous expression of monoterpene cyclase/methyltransferase genes strongly suggest that S. avermitilis has theability to supply GPP for the biosynthesis of monoterpenoid me-tabolites. Moreover, two genes encoding monoterpene cyclase andmethyltransferase are necessary for the biosynthesis of 2-MIB.

Enzymatic Reaction of Recombinant Proteins. Because exoconjugantsproduced 2-MIB, enzymatic reactions were examined by usingrecombinant proteins prepared in Escherichia coli. Two genes of S.lasaliensis were amplified and joined with pET22b(�) orpET28a(�) vectors. After four combinations of His6 residue at theN- or C-terminal ends were examined, soluble proteins wereobtained from His6-tagged monoterpene cyclase at the C terminus(52.4 kDa) and methyltransferase at the N terminus (35.0 kDa), asjudged by SDS/PAGE, respectively (Fig. S3). To test the predicted2-MIB synthesis activities of the S. lasaliensis gene products, bothsoluble recombinant cyclase and methyltransferase were incubatedwith GPP and SAM in the presence of MgCl2. The organic-solubleproduct was analyzed by GC-MS analyzer. The enzymatic reactiongave a single major product with mass, [M�] m/z 168, consistentwith formation of a monoterpenoid alcohol, and the product wasidentical to 2-MIB [by comparison with the synthetic sample of2-MIB (Fig. S4)]. When SAM or the recombinant methyltrans-ferase was removed from the reaction mixture, 2-MIB was notproduced. Furthermore, incubation with recombinant methyltrans-ferase, GPP, and SAM followed by alkaline phosphatase treatmentgave a single major product, which was identified as 2-methylge-raniol by comparison of its retention time and mass spectrum([M�], m/z 168) to those of the authentic 2-methygeraniol synthe-sized (Fig. S5). This observation strongly indicated that the meth-yltransferase catalyzes conversion of GPP to 2-methyl GPP. Theseresults were also supported with those of heterologous expressionof cloned genes. The enzymatic reaction suggests that a substratefor monoterpene cyclase predicted is a methylated GPP.

Three types of terpene cyclases, monoterpene, sesquiterpene,

and diterpene, of microbial origin have been identified. The com-parative analysis of monoterpene cyclases presented here, alongwith other microbial terpene cyclases, promises to illustrate thereaction mechanism of cyclization in detail. Experimental resultshave unambiguously elucidated the biosynthetic mechanism of themonoterpenoid odorous compound, 2-MIB, as follows: (i) GPPgenerated from isopentenyl diphosphate (IPP) and dimethylallyldiphosphate (DMAPP) by GPP synthase was converted to meth-ylated GPP by a specific methyltransferase (GPP methyltransfera-se)l; (ii) 2-methyl GPP was cyclized to generate 2-MIB by amonoterpene cyclase (2-MIB synthase) in the presence of Mg2�

(Fig. 6). This finding is the first reported observation of thebiochemical features of microbial monoterpene cyclase and of themethylation of GPP by a specific methyltransferase.

Unfortunately, monoterpene cyclase genes encoding 2-MIB syn-thase were not found in genome-sequenced cyanobateria(Anabaena, Gloeobacter, Nostoc, Prochlorococcus, Synechococcus,Synecocyctis, Thermosynechococcus, and Trichodesmium) and myx-obacteria (Myxococcus). Consequently, the productivity of monot-erpenoid metabolites from these bacteria was not examined. In thenear future, genome analysis of 2-MIB-producing cyanobacteria(Oscillatoria curviceps, Oscillatoria tenuis, and Lyngbya sp.) (7, 8)and myxobacteria (Nannocystis exedens) (9) will be completed, andtheir monoterpene cyclases will be characterized. Recently, we havefound two genes from an actinomycete strain, Kitasatospora setae(produced 2-MIB), that formed an operon and bore significantsimilarities to 2-MIB synthase and GPP methyltransferase genes,respectively. Our work has confirmed that many genes encoding for2-MIB synthase and GPP methyltransferase can be identified andcharacterized from a range of actinomycetes.

Materials and MethodsBioinformatics. Genome-sequenced bacterial genome data (total 1,906,368 pro-teins from 595 strains) was obtained from the National Center for BiotechnologyInformation (ftp://ftp.ncbi.nih.gov/genomes/Bacteria/) and the Pfam databaseversion 22.0 (July 2007, 9,318 families) from the Howard Hughes Medical Institute(ftp://selab.janelia.org/pub/Pfam/). Streptomyces scabies 87.22 sequence data(8,984 protein-coding sequences annotated) was produced by the Streptomycesscabies Sequencing Group at the Sanger Institute U.K. and was obtained fromftp://ftp.sanger.ac.uk/pub/pathogens/ssc/. Other unpublished actinomycete ge-nome data, S. griseus IFO13350 (40), Micromonospora olivasterospora KY11048(draft sequence data), and some sequence data of a giant linear plasmid pKSL ofS. lasaliensisNRRL3382Rwerealsoused.TheextractionofPfammodelofPF03936(terpene synthase family, metal-binding domain) and searching proteins fromdatabase used binary programs made by the compilation of a source package ofhidden Markov models for biological sequence analysis, hmmer-2.3.2 (ftp://selab.janelia.org/pub/software/hmmer/2.3.2/). Alignment of sequences was ana-lyzed by using MAFFT (41) version 6.24 obtained from Kyushu University (http://align.bmr.kyushu-u.ac.jp/mafft/software/source.html). Phylogenetic analysis ofaligned sequences was done by the bootstrap method (bootstrap number, 1,000;seed number, 111) of CLUSTALW (42) version 1.83 (ftp://ftp.ebi.ac.uk/pub/

Fig. 5. Volatile monoterpenoid metabolites, 2-MIB (A) and 2-MB (B), andsesquiterpenoid metabolite, geosmin (C) produced by actinomycete strainscarrying predicted monoterpene cyclase and methyltransferase genes. Allmicroorganisms were grown on SFM (left half from dashed line) and M4YE(right half from dashed line) agar plates. The strain names are as follows: SCO,S. coelicolor A3(2); SGR, S. griseus; SACE, Sa. erythraea; SAM, S. ambofaciens;SLA, S. lasaliensis; MOL, M. olivasterospora; SAV, S. avermitilis; U1717R, S.ambofaciens deletion mutant.

Fig. 6. Biosynthetic mechanism from GPP to 2-MIB (or 2-MB).

7426 � www.pnas.org�cgi�doi�10.1073�pnas.0802312105 Komatsu et al.

software/unix/clustalw/). Drawing of the bootstrap tree was done by njplot(http://pbil.univ-lyon1.fr/software/njplot.html).

Bacterial Strains and Plasmid Vectors. S. avermitilis K139 (isogenic to MA-4680),S. ambofaciens ISP5053, S. ambofaciens U1717R (39), S. coelicolor A3(2), S. griseusIFO13350, S. lasaliensis NRRL3382R, and Saccharopolyspora erythraea NRRL2338were obtained from the culture collection of the Kitasato Institute. M. olivaster-ospora KY11048 was provided by Kyowa Hakko Co. Ltd Japan. S. avermitilis largedeletion mutant SUKA16 (�79,460-1,595,563 nt �ptl::ermE �geoA::aadA �olm�8,892,894-8,917,256 nt) was used as a host strain for the heterologous expres-sion of the operon carrying genes encoding monoterpene cyclase/methyltrans-ferase predicted. E. coli DH5�, E. coli F� dcm �(srl-recA)306::Tn10 carryingpUB307-aph::Tn7, and E. coli BL21(DE3) were used for general DNA manipula-tion, for E. coli/Streptomyces conjugation, and for the preparation of recombi-nant monoterpene cyclase and methyltransferase, respectively. Integrating vec-tor pKU469 was used for the heterologous expression of actinomycetemonoterpene cyclase/methyltransferase genes in S. avermitilis SUKA16. E. coliexpression vectors pET22b(�) and pET28a(�) were used for preparation of re-combinant monoterpene cyclase and methyltransferase, respectively. Culturingof E. coli used Luria–Bertani broth (containing 10.0 g/liter tryptone, 5.0 g/literyeast extract, and 5.0 g/liter NaCl; pH 7.4). Agar was supplied at 15.0 g/liter forsolid media (LA).

Cultivation and Extraction of Volatile Terpenoid Metabolites from ActinomyceteStrains. For the production of volatile terpenoid metabolites, spores were spreadonagarmedia:M4YE(containing10.0g/liter solublestarch,1.0g/literK2HPO4,1.0g/liter MgSO4�7H2O, 1.0 g/liter NaCl, 2.0 g/liter (NH4) 2SO4, 2.0 g/liter CaCO3, 0.001g/liter FeSO4�7H2O, 0.001 g/liter MnSO4�4H2O, 0.001 g/liter ZnSO4�7H2O, 2.0 g/literyeast extract, and 15.0 g/liter agar; pH 7.0); SFM (containing 20.0 g/liter defattedsoy flour, 20.0 g/liter mannitol, and 15.0 g/liter agar; not adjusted pH); and YMG

(containing 4.0 g/liter yeast extract, 10.0 g/liter malts extract, 10.0 g/liter glucose,and 20.0 g/liter agar; pH 7.4 for exoconjugants). SFM or YMS (43) medium wasused for the sporulation of Streptomyces strains. The actinomycete strains spreadon the agar plates (diameter 90 mm) were grown at 30°C. After cultivation, 5 mlof methanol was directly added to each plate, and the plates were allowed tostand at room temperature for 30 min. The methanol extract was collected intoglass tubes. Volatile metabolites were extracted with 0.5 ml of n-hexane, and theupper n-hexane layer was carefully collected. The n-hexane extract was dehy-drated over anhydrous Na2SO4, and a portion of extract was directly evaluated ina capillary GC-MS analyzer.

GC-MS Analysis of Volatile Metabolites. A 1- to 5-�l portion of the extract wasanalyzedbycapillaryGC-MS[ShimadzuGC-17A,70eV,EI,positive ionmode;30m� 0.25-mm neutral bond-5 capillary column (5% phenylmethylsilicon), using atemperature program of 50°C–280°C, temperature gradient of 20°C/minute or60°C for 3 min, 60°C–160°C, temperature gradient of 10°C/min, then 160°C–240°C, temperature gradient 40°C/min]. 2-MIB, 2-MB, geosmin, and germacradi-enol were identified by comparison with the spectra of the corresponding ref-erence compounds in the database. Furthermore, chemically synthesized 2-MIB,2-MB, 2-M2B, and 1-MC (SI Materials and Methods), and authentic sample ofgeosmin purified from S. avermitilis were also compared with each extract ofactinomycete strain.

ACKNOWLEDGMENTS. We thank Kyowa Hakko Co. Ltd. Japan for kindly distri-butionofMicromonosporaolivasterosporaKY11048.WealsothankAndyCrumpforacritical readingofthemanuscript.ThisworkwassupportedbyaGrant-in-Aidfor Scientific Research on Priority Areas ‘‘Applied Genomics’’ from the Ministry ofEducation, Culture, Sports, Science and Technology, Japan (MEXT) (H.I.) and byGrants-in-Aid for Scientific Research from the Japan Society for the Promotion ofScience (JSPS) (A) 17208010 (H.O.) and 20310122 (H.I.).

1. Glasby JS (1982) Encyclopedia of the Terpenoids (John Wiley, Chichester, U.K.).2. Sacchettini JC, Poulter CD (1997) Biochemistry: Creating isoprenoid diversity. Science

277:1788–1789.3. Scholler CEG, Gurtler H, Pedersen R, Molin S, Wilkins K (2002) Volatile metabolites from

actinomycetes. J Agric Food Chem 50:2615–2621.4. Pollak FC, Berger RG (1996) Geosmin and related volatiles in bioreactor-cultured

Streptomyces citreus CBS 109.60. Appl Environ Microbiol 62:1295–1299.5. Gerber NN (1979) Volatile substances from actinomycetes. Crit Rev Microbiol 7:191–214.6. Juttner, F (1987) Volatile organic substances. The Cyanobacteria, eds Fay B, Baalen CV

(Elsevier, Amsterdam), pp 453–469.7. Izaguirre G, Hwang CJ, Krasner SW, McGuire MJ (1982) Geosmin and 2-methyli-

soborneol from cyanobacteria in three water-supply systems. Appl Environ Microbiol43:708–714.

8. Wu J-T, Juttner F (1988) Differential partitioning of geosmin and 2-methylisoborneolbetween cellular constituents in Oscillatoria tenuis. Arch Microbiol 150:580–583.

9. Dickschat JS, et al. (2007) Biosynthesis of the off-flavor 2-methylisoborneol by themyxobacterium Nannocystis exedens. Angew Chem Int Ed 46:8287–8290.

10. Borjesson T, Stollman U, Schnurer J (1993) Off odors compounds produced by molds onoatmeal agar: Identification and relation to other growth characteristics. J Agric FoodChem 41:2104–2111.

11. Mattheis JP, Roberts RG (1992) Identification of geosmin as a volatile of Penicilliumexpansum. Appl Environ Microbiol 58:3170–3172.

12. Tabachek JL, Yurkowski M (1976) Isolation and identification of blue-green algaeproducing muddy odor metabolites, geosmin and 2-methylisoborneol, in saline lakesin Manitoba J. Fish Res Board Can 33:25–35.

13. Persson PE (1983) Off-flavours in aquatic ecosystems. Wat Sci Technol 15:1–11.14. Yagi M, et al. (1983) Odour problems in Lake Biwa. Wat Sci Technol 15:311–321.15. Lu G, Fellman JK, Edwards CG, Mattinson DS, Navazio J (2003) Quantitative determi-

nation of geosmin in red beets (Beta Vulgaris L.) using headspace solid phase micro-extraction. J Agric Food Chem 51:1021–1025.

16. Karahadian C, Josephson DB, Lindsay RC (1985) Volatile compounds from Penicilliumsp. contributing musty-earthy notes to Brie and Camembert cheese flavors. J AgricFood Chem 33:339–343.

17. Gurtler H, et al. (1994) Albaflavenone, a sesquiterpene ketone with a zizaene skeletonproduced by a streptomycete with a new rope morphology. J Antibiot 47:434–439.

18. Gerber NN (1968) Geosmin, from microorganisms, is trans-l,10-dimethyl-trans-9-decalol. Tetrahedron Lett 25:2971–2974.

19. Gerber NN (1969) A volatile metabolite of actinomycetes, 2-methylisoborneol. J An-tibiot 22:508–509.

20. Medsker LL, Jenkins D, Thomas JF, Koch C (1969) Odorous compounds in naturalwaters. 2-Exo-hydroxy-2-methylbornane, the major odorous compound produced byseveral actinomycetes. Environ Sci Technol 3:476–477.

21. Cane DE, Watt RM (2003) Expression and mechanistic analysis of a germacradienolsynthase from Streptomyces coelicolor implicated in geosmin biosynthesis. Proc NatlAcad Sci USA 100:1547–1551.

22. Cane DE, He X, Kobayashi S, Omura S, Ikeda H (2006) Geosmin biosynthesis in Strep-tomyces avermitilis. J Antibiot 59:471–479.

23. Lin X, Hopson R, Cane DE (2006) Genome mining in Streptomyces coelicolor: Molecularcloning and characterization of a new sesquiterpene synthase. J Am Chem Soc128:6022–6023.

24. Trapp SC, Croteau RB (2001) Genomic organization of plant terpene synthases andmolecular evolutionary implications. Genetics 158:811–832.

25. Oikawa H, et al. (2001) Cloning and functional expression of cDNA encoding aphidi-colan-16-�ol synthase: A key enzyme responsible for formation of an unusual diter-pene skeleton in biosynthesis of aphidicolin. J Am Chem Soc 123:5154–5155.

26. Hohn TM (1999) Cloning and expression of terpene synthase genes. ComprehensiveNatural Products Chemistry: Isoprenoids Including Carotenoids and Steroids, ed CaneDE (Elsevier, Oxford), pp 201–215.

27. Bentley R, Meganathan R (1982) Geosmin and methylisoborneol biosynthesis in strep-tomycetes. FEBS Lett 125:220–222.

28. Eddy SR (1998) Profile hidden Markov models. Bioinformatics 14:755–763.29. Bateman R, et al. (2002) The Pfam protein families database. Nucleic Acids Res

30:276–280.30. Watanabe K, et al. (2006) Total biosynthesis of antitumor nonribosomal peptides in

Escherichia coli. Nat Chem Biol 2:423–428.31. Dairi T, et al. (2001) Eubacterial diterpene cyclase genes essential for production of the

isoprenoid antibiotic terpentecin. J Bacteriol 183:6085–6094.32. Lesburg CA, Zhai G, Cane DE, Christianson DW (1997) Crystal structure of pentalenene

synthase: Mechanistic insights on terpenoid cyclization reactions in biology. Science277:1820–1824.

33. Seemann M, et al. (2002) Pentalenene synthase. Analysis of active site residues bysite-directed mutagenesis. J Am Chem Soc 124:7681–7689.

34. Starks CM, Back K, Chappell J, Noel JP (1997) Structural basis for cyclic terpenebiosynthesis by tobacco 5-epi-aristolochene synthase. Science 277:1815–1820.

35. Caruthers JM, Kang I, Rynkiewicz MJ, Cane DE, Christianson DW (2000) Crystal structuredetermination of aristolochene synthase from the blue cheese mold, Penicilliumroqueforti. J Biol Chem 275:25533–25539.

36. Rynkiewicz MJ, Cane DE, Christianson DW (2001) Structure of trichodiene synthasefrom Fusarium sporotrichioides provides mechanistic inferences on the terpene cy-clization cascade. Proc Natl Acad Sci USA 98:13543–13548.

37. Peters RJ, Croteau RB (2002) Abietadiene synthase catalysis: Mutational analysis of aprenyl diphosphate ionization-initiated cyclization and rearrangement. Proc NatlAcad Sci USA 99:580–584.

38. Choulet F, et al. (2006) Evolution of the terminal regions of the Streptomyces linearchromosome. Mol Biol Evol 23:2361–2369.

39. Ikeda H, Tanaka H, Omura S (1982) Genetic and biochemical features of spiramycinbiosynthesis in Streptomyces ambofaciens. J Antibiot 35:507–516.

40. Ohnishi Y, et al. (March 28, 2008) The genome sequence of the streptomycin-producingmicroorganism Streptomyces griseus IFO13350. J Bacteriol, 10.1128/JB.00204-08.

41. Katoh K, Misawa K, Kuma K, Miyata T (2002) MAFFT: A novel method for rapidmultiple sequence alignment based on fast Fourier transform. Nucleic Acids Res30:3059 –3066.

42. 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 22:4673–4680.

43. Ikeda H, Kotaki H, Omura S (1987) Genetic studies of avermectin biosynthesis inStreptomyces avermitilis. J Bacteriol 169:5615–5621.

Komatsu et al. PNAS � May 27, 2008 � vol. 105 � no. 21 � 7427

BIO

CHEM

ISTR

Y