Unprecedented diversity of catalytic domains in the first four modules of the putative pederin polyketide synthase

Unprecedented Diversity of Catalytic Domains inthe First Four Modules of the Putative PederinPolyketide SynthaseJˆrn Piel,*[a] Gaiping Wen,[b] Matthias Platzer,[b] and Dequan Hui[a] Polyketides of the pederin group are highly potent antitumor region of the symbiont genome. It contains at least three novel compounds found in terrestrial beetles and marine sponges.
catalytic domains that are predicted to be involved in pederin Pederin is used by beetles of the genera Paederus and Paederidus chain initiation and the formation of an unusual exomethylene as a chemical defense. We have recently identified a group of bond. The region is bordered by mobility pseudogenes; this putative pederin biosynthesis genes and localized them to the suggests that gene transposition led to the disjointed cluster genome of an as yet unculturable Pseudomonas sp. symbiont, the organization. With this work, all putative pederin genes have been likely true pederin producer. However, this polyketide synthase identified. Their heterologous expression in a culturable bacterium cluster lacks several genes expected for pederin production. Here will provide important insights into how sustainable sources of we report an additional polyketide synthase encoded on a separate invertebrate-derived drug candidates can be created.
It has been proposed for many years that natural productsymbiosis, that is, the biosynthesis of secondary metabolites bysymbiotic organisms, plays an important role in the chemistry ofinvertebrates.[1±4] Such suspected symbiont metabolites includemany promising drug candidates isolated from marine animals.
The existence of producing bacteria could be exploited toovercome the supply problem that currently hampers animal-derived drug development and production. However, ourcurrent general inability to cultivate such symbionts representsa serious technical challenge in pinpointing the true producer.
So far, the most compelling evidence for symbiotic producers polyketide synthase (PKS) ± nonribosomal peptide synthetase exists for the complex polyketide pederin (1), a highly potent system, whose architecture perfectly matches a large portion ofthe pederin (1) structure.[7] Although these genes were isolatedfrom the total DNA of Paederus fuscipes beetles, they belong tothe genome of an as yet unculturable symbiotic bacterium witha very close relationship to Pseudomonas aeruginosa.[7, 8] Bacterial type I PKSs are giant, modular enzymes harboring a multitude of catalytic domains.[9] Each module typically consistsof at least a ketoacyl synthase (KS), an acyltransferase (AT), and Max Planck Institute for Chemical Ecology anticancer agent. It is found in apparently all species of the rove Department of Bioorganic ChemistryHans-Knˆll-Strasse 8, Beutenberg Campus beetle genera Paederus and Paederidus,[5] whose females use pederin to chemically defend their offspring against predators.[6] Remarkably, closely related compounds have also been isolated from several sponges during marine antitumor and antiviral screening programs, for example, theopederin C (2) and onna- mide A (3) from Theonella swinhoei.[5] Recently we have reported the ped biosynthesis gene cluster encoding a hybrid type I ¹ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim can be distinguished from those of cis-AT systems by phyloge-netic analysis.[18] Such an analysis placed the amplified productdesignated as PS3 close to the cis-AT, and PS1 and PS2 into thetrans-AT group. Since the missing part of the ped PKS shouldbelong to the latter group, we intended to isolate the gene(s)belonging to these PCR products. We therefore screened2000 clones of a cosmid library prepared from DNA of P. fuscipesbacteria with probes derived from PS1 and PS2. Unexpectedly,restriction mapping of the positive clones revealed that the twoPCR products did not originate from the same genomic region an acyl carrier protein (ACP) domain, and attaches one polyke- and apparently belonged to different secondary metabolites.
tide building block, usually a small malonyl-CoA derivative, to Partial sequencing of both regions identified the PS2 system as a the growing polyketide chain. During each of these elongation mixed peptide synthetase ± PKS unrelated to the pederin (1) steps, various optional additional domains present on the structure (data not shown), while the domain arrangement of module can introduce further structural modifications. The order the PS1 PKS was in very good agreement with the proposed and domain architecture of modules on the PKS enzymes usually architecture. The cosmid pPS9D2 containing the entire PS1 strictly correspond to the order and structure of each polyketide region, as judged from sequencing of insert ends, was therefore building block, a phenomenon known as the colinearity rule.
This rule can be used to predict whether an isolated PKS genecluster encodes the biosynthesis of a specific metabolite. It also permits a rational design of novel polyketides by geneticmodification of the domain architecture.[10, 11] A map of the completely sequenced insert of cosmid pPS9D2 is An analysis of the domain organization of the ped system shown in Figure 1, and the deduced functions of the identified revealed a number of surprises. The AT domains, usually found open reading frames (ORFs) are listed in Table 1. As with the within each module, were instead encoded by separate genes.[7] previously sequenced ped region, several genes with remarkable Although this was the first reported example of monofunctional homology to Ps. aeruginosa ORFs were found in proximity to the ATs encoded by type I PKS clusters, a number of similar systems PKS; this indicates that the two systems are located on the same have been discovered recently; this suggests that such enzymes genome. These homologues comprise a cluster involved in the might be more widespread among bacteria.[12±14] In addition, uptake of branched-chain amino acids, a putative heat-shock several partial gene clusters have been published, whose gene, and a probable hemin receptor gene. In addition, the modules lack ATs and that probably also belong to this cosmid harbors various pseudogenes, including 11 decayed IS- group.[15±17] Work on the leinamycin biosynthesis gene cluster like elements (IS ˆ insert sequence). Framed by two of these ISs revealed that these ATs acylate each module in trans,[12] and we is a small gene cluster highly similar to PKS systems. Its three have shown that such ™trans-AT∫ PKSs are phylogenetically genes, designated pedI, pedJ, and pedK, do not contain frame- distinct from regular ™cis-AT∫ systems.[18] The second deviation shifts or stop codons and therefore seem to be intact.
from the general PKS rules was that the genes encoding the first An inspection of the domain arrangement of PedI revealed elongation steps were not found within the ped cluster.
several unprecedented features among type I PKSs. Neverthe- Unclustered genes of a bacterial PKS pathway are very rare: to less, its unusual architecture clearly mirrors the structure of the the best of our knowledge, the ansamitocin system is the only pederin (1) starter moiety. The deduced product of pedI other published example of a disconnected PKS.[19] To isolate the resembles a type I PKS consisting of four modules. By phyloge- remaining pederin genes from the bacterial consortium of netic analysis of a PCR product covering a KS domain of this P. fuscipes, we used a phylogeny-based PCR strategy that gene, we had previously predicted that the PKS lacks AT specifically targets trans-AT PKS systems.[18] This approach led domains. In accordance with this, no homology to such domains to the isolation of additional PKS gene fragments from was found within PedI. The first module lacks a KS domain and metagenomic DNA of beetle bacteria. In this paper we present should therefore catalyze chain initiation. However, instead of a evidence that one of these fragments belongs to the missing PKS conventional AT, the module harbors an unusual N-terminal catalyzing the first three rounds of elongation.
domain of about 240 aa similar to proteins of the GCN5-relatedN-acetyltransferase (GNAT) family[20] and sequence patternsdiagnostic [RQ]XXGX[GA] motif involved in binding of the acetyl-CoAsubstrate[20] was identified at aa 135 ± 140 (QQYNLG). GNATs Screening of a bacterial cosmid library for additional trans-AT catalyze the transfer of an acetyl group from acetyl-CoA to various heteroatoms,[20] which suggests that this domain primes We have previously reported the PCR amplification from the loading module with acetate. GNAT domains as part of PKSs metagenomic bacterial DNA of three PKS gene fragments have not been reported before.[9, 21] All published AT-less PKS (PS1 ± PS3) that do not belong to the already sequenced ped systems are loaded with a starter other than acetate. In these cluster.[18] PCR products covering the KS domain of trans-AT PKSs cases, the absence of a GNAT domain, which usually accepts only ¹ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Figure 1. Map of the genomic region belonging to the PCR product PS1 and proposed biosynthesis of pederin (1). Dotted boxes and arrows indicate degraded ORFs.
Pseudogenes in black are involved in horizontal gene transfer. Ped genes are marked in gray. GNAT, GCN5-related N-acetyltransferase domain; ACP, acyl carrier proteindomain; KS, ketosynthase domain; KR, ketoreductase domain; MT, methyltransferase domain; CR, crotonase domain; OXY, oxygenase; T, transposase; INT, phageintegrase. Numbered genes are highly similar to genes from Pseudomonas aeruginosa. The numbers refer to the PA number of the closest homologue. PedA ± PedH areencoded on a different part of the symbiont genome.[7] acetyl-CoA as substrate, is therefore not surprising. The that were proposed to catalyze the addition of acetyl- or sequencing of additional AT-less PKSs will reveal if the PedI malonyl-CoA to a keto group, followed by decarboxylation, GNAT domain is an exception or a common feature to acetate- dehydration, and/or reduction steps. We had initially expected a similar enzyme in the biosynthesis of pederin (1), but such a In the second module, signature motifs of a KS, a ketoreduc- homologue is not among the genes on pPS9D2 and could tase (KR), a methyltransferase (MT), and an ACP domain were neither be detected in the symbiont DNA by hybridization with a found, indicating that the corresponding polyketide extension probe derived from the HMG-CoA synthase homologue mupH unit bears a methyl group in an a-position to a hydroxy function.
nor by PCR with degenerate primers based on conserved motifs The KR domain contains a characteristic Asp residue at (data not shown). Instead, two unprecedented domains were position 1270, which has recently been shown to be conserved found in the third module of PedI. In addition to a KS and an ACP in domains generating alcohols with D stereochemistry[22] (B-type domain, two domains were identified that show homology to alcohols, as defined in ref. [23]). This precisely corresponds to the enzymes of the crotonase (CR) superfamily. These enzymes first extension unit of the pederin molecule, which bears a B-type catalyze a wide range of diverse reactions all involving a hydroxyl function adjacent to a methyl group. Cardani et al.
stabilized CoA thioester enolate intermediate.[28] The CR domain attempted to elucidate the biosynthesis of 1 by feeding adjacent to the KS exhibits relatively low homology to known radioactively labeled precursors to P. fuscipes.[24] They reported enzymes (42 % similarity to the next homologue). A PSI-BLAST the presence of 40 % activity of administered labeled propionate search revealed that it is related to naphthoate synthases.[29] The in the N-acyl moiety constituting the western half of pederin (1).
second CR domain is highly similar to discrete enzymes found in This seemingly suggests that the first extension unit is derived trans-AT systems, such as PksI (66 %) from Bacillus subtilis[15] and from methylmalonyl-CoA instead of malonyl-CoA and SAM- MupK from Pseudomonas fluorescens.[13] These enzymes are dependent methylation by an MT domain. However, propionate annotated as putative enoyl-CoA hydratases, but their exact is catabolized in many bacteria[25] and insects[26] to acetate by functions are presently unknown. If the colinearity rules apply, several different routes and can then enter polyketide biosyn- this module could catalyze the attachment of the exomethylene thesis. Since the radioactive label was located on the propionate group, possibly by action of the two unusual CR domains.
C2 position, it would be retained after chain shortening and be Indeed, members of the CR superfamily perform reactions very incorporated into malonyl-CoA-derived moieties. The fact that similar to those proposed for exomethylene group formation all known methyl groups generated by AT-less PKSs are attached (Scheme 1). Among the enzymes acting on CÀC bonds, naph- by MT domains[12, 13, 15] is consistent with this hypothesis.
thoate synthase from menaquinone metabolism catalyzes a The second extension unit bears a rare exomethylene group at Claisen condensation,[29] and hydroxycinnamate lyase from its C1 position. Carbon atoms that branch off such positions are Ps. fluorescens is responsible for a retro-aldol reaction.[30] Decar- known from a number of other complex polyketides, for boxylation and dehydration steps are known from the E. coli example, myxovirescin (ˆ TA),[27] mupirocin,[13] and leinamycin.[12] methylmalonyl-CoA decarboxylase and crotonase,[28] respective- The PKS clusters of these compounds encode 3-hydroxy-3- ly. Moreover, homologues of the second CR domain are located methylglutaryl coenzyme A (HMG-CoA) synthase homologues directly adjacent to HMG-CoA synthase-like genes in the mup ¹ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Table 1. Genes and pseudogenes identified in the PS1 region.
GNAT, GCN5-related histone acetyltransferase family branched-chain amino acid transporter permease protein branched-chain amino acid transporter permease protein branched-chain amino acid transporter ATP-binding protein branched-chain amino acid transporter ATP-binding protein [a] Entries in parentheses belong to pseudogenes.
and pksX clusters, suggesting that these genes participate in N-terminal KS domain of PedF lacks the conserved His residue methyl(ene) group formation. The first CR domain without close necessary for the decarboxylation of the extender unit.[31] This homologues might substitute the HMG-CoA synthase, catalyzing module should therefore be unable to perform a polyketide the aldol addition of acetyl- or malonyl-CoA (Scheme 1). The CR elongation, which in fact necessitates an additional module on related to PksI and MupK could then be involved in the further PedI. The same feature is known from the mup system, in which a processing of the intermediate and perform the final decarbox- seemingly superfluous module lacking a decarboxylation motif ylation and/or dehydration steps. If the two CR domains are is located at the N-terminus of MmpA.[13] Thomas and co-workers indeed sufficient for exomethylene group formation, this proposed that such modules could have a pseudo-loading module could represent a valuable tool in engineered biosyn- function and be involved in the proper channeling of polyketide thesis that might dramatically increase the possible structural chains between two PKS proteins.[13] Alignment of the last PedI range of recombinant polyketides. However, at present it cannot module with other PKSs (data not shown) revealed that the KS be excluded that PedK or additional enzymes encoded else- and KR domains are separated by an unusually large region of where on the symbiont genome participate in this reaction.
approximately 250 aa without convincing homology to known Module 4 represents the last module of this PKS and contains sequences or patterns. Further functional studies are needed to a KS, a KR, and an ACP domain. The presence of this module was determine whether it represents a novel domain, perhaps unexpected, since the corresponding polyketide extension step catalyzing the stereospecific intramolecular acetalization that was initially assigned to the first module of PedF. However, the likely takes place at this point in the biosynthesis of pederin (1), ¹ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim needs additional genes to be functional. These featuresstrongly suggest that PedIJK and the previously identi-fied ped genes complement each other and jointlyencode pederin (1) biosynthesis. The current uncultur-ability of the Paederus spp. symbiont prohibits standardexperiments to verify the function of ped genes, such asknockout. Present studies in our laboratory thereforeaim at expressing the ped genes in a heterologous host.
Although the ped system is to the best of our knowledge only the second example of PKS genesdistributed among two genomic regions,[19] the numer-ous IS elements located on both ends of the two pedregions readily explain the disjointed structure. Theirpresence indicates that the entire PKS cluster wasoriginally acquired as a genomic island by horizontalgene transfer. By a second IS-mediated rearrangementprocess, the pedIJK region could then have transposedto a different genome position. Ultimately, the regionwould have been stabilized by decay of all adjacentmobility genes. In free-living bacteria, frequent horizon-tal gene transfer selects for the clustering of genes fromsecondary metabolism.[33] In contrast, gene exchangewould be a rare event in the Paederus spp. symbiont,which has lived in a shielded environment for millions ofyears. These conditions could have resulted in theobserved gene scattering.
Scheme 1. Reactions catalyzed by members of the crotonase family and hypothetical The pederin system represents an ideal model to pathway leading to the formation of the pederin exomethylene group. A) napthoate understand the factors governing natural product synthase (MenB); B) 4-hydroxycinnamoyl-CoA hydratase/lyase (HCHL) from Ps. fluorescensAN103 (last step of a two-step reaction); C) methylmalonyl-CoA decarboxylase (MMCD) symbioses and how sustainable sources of drug candi- from E. coli; D) proposed pathway catalyzed by the PedI CRs. Acetyl-CoA could be used dates from invertebrates can be created. In this work, instead of malonyl-CoA in this route.
the availability of all gene candidates necessary forpederin (1) biosynthesis now sets the stage for theproduction of pederin-type antitumor compounds in a or whether it is only an unusually long spacer region between culturable bacterium. In addition, the small repertoire of known the KS and the KR. The KR domain present in this module has no PKS domains has been expanded by at least three new members.
structural counterpart in pederin (1). It could be an inactive Their further study could significantly improve the versatility of evolutionary remnant, similar to many other known PKS domains engineered biosynthesis and our understanding of polyketide with no apparent function.[9] Alternatively, it could aid in the acetalization by polarizing the carbonyl bond.
PedJ, encoded by a gene downstream of pedI, is highly similar to several oxidoreductases of the bacterial luciferase type and contains the characteristic Pfam pattern (PF00296). The twoclosest database matches are MupA with unknown function[13] Isolation of bacteria from Paederus fuscipes: One hundred and MtaG from the myxothiazol pathway.[32] The latter enzyme P. fuscipes beetles collected at Aydın, Turkey were ground in liquid was proposed by M¸ller and coworkers to catalyze the a- nitrogen and resuspended in LB medium (5 mL). The mixture was hydroxylation of an amino acid residue. Mupirocin contains no kept for 10 min on ice to let beetle residues settle. The supernatant amino acid but features an a-hydroxylated polyketide extender was centrifuged at 100 g for 10 min to sediment eukaryotic cells. The position. PedJ could therefore be responsible for the a- supernatant was then passed through a 40-mm nylon filter (Milli- hydroxylation of either the glycine residue or the third extension pore), and the bacteria were pelleted by centrifugation at 5000 g for10 min. The purity of the bacterial fraction was checked by DAPI unit. Adjacent to pedJ, the ORF pedK encoding a putative protein staining and microscopic inspection.
of 318 aa was identified. However, homology and patternanalysis provided no insight into the possible function of this Preparation and screening of a P. fuscipes symbiont genomiclibrary: Genomic DNA from P. fuscipes bacteria was prepared by standard SDS lysis and used for the construction of a cosmid library In contrast to all other known PKS gene clusters, no domain or in the pWEB vector (Epicentre) according to the manufacturer's gene was identified on the cosmid that would be involved in instructions. Two thousand clones were deposited in 96-well plate release of the assembled chain from the PKS or in the transfer of format and transferred to Hybond N‡ membranes. Membranes were acyl-CoA units to the ACP domains. Therefore, this PKS evidently screened by using a fluorescein-labeled probe prepared from a 1.5- ¹ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim kb fragment of the ped cluster and the ECL Direct Labeling and [4] P. Proksch, R. A. Edrada, R. Ebel, Appl. Microbiol. Biotechnol. 2002, 59, 125.
Detection kit (Amersham Biosciences) according to the manufac- [5] R. Narquizian, P. J. Kocienski in The Role of Natural Products in Drug turer's protocol, except that Church buffer was used during Discovery, Vol. 32 (Eds.: R. Mulzer, R. Bohlmann), Springer, Heidelberg, prehybridization and hybridization. The probes were prepared from [6] R. L. L. Kellner, K. Dettner, Oecologia 1996, 107, 293.
PCR fragments corresponding to the KS domain as described [7] J. Piel, Proc. Natl. Acad. Sci. USA 2002, 99, 14 002.
elsewhere.[7, 18] Hybridization was performed at 60 8C, followed by [8] R. L. L. Kellner, Insect Biochem. Mol. Biol. 2002, 32, 389.
low-stringency washes for 2 Â 7 min at 25 8C and 2 Â 15 min at 60 8C.
[9] B. J. Rawlings, Nat. Prod. Rep. 2001, 18, 231.
Sequence analysis of positive cosmids: Inserts of positive cosmids [10] E. Rodriguez, R. McDaniel, Curr. Opin. Microbiol. 2001, 4, 526.
were end-sequenced to gain information about the location of PKS [11] C. D. Reeves, Crit. Rev. Biotechnol. 2003, 23, 95.
genes. The cosmid pPS9D2 with genes unrelated to PKSs at its [12] Y. Q. Cheng, G. L. Tang, B. Shen, Proc. Natl. Acad. Sci. USA 2003, 100, 3149.
[13] A. K. El-Sayed, J. Hothersall, S. M. Cooper, E. Stephens, T. J. Simpson, C. M.
termini was chosen for complete sequencing. After shearing of the DNA by using a Standard Nebulizer (Octurno), the fragment ends [14] S. Mochizuki, K. Hiratsu, M. Suwa, T. Ishii, F. Sugino, K. Yamada, H. Kinashi, were repaired with T4 DNA polymerase and Klenow fragment.
Fragments of 1 ± 1.5 kb were isolated by agarose electrophoresis, [15] A. M. Albertini, T. Caramori, F. Scoffone, C. Scotti, A. Galizzi, Microbiology cloned into the pUC18 vector, and end-sequenced by using the BigDye Terminator Ready Mix (Applied Biosystems) and an ABI 3700 [16] Y. Paitan, G. Alon, E. Orr, E. Z. Ron, E. Rosenberg, J. Mol. Biol. 1999, 286, 465.
sequencer (Applied Biosystems). Sequence data were assembled by [17] G. Z. Huang, L. H. Zhang, R. G. Birch, Microbiology 2001, 147, 631.
using GAP4 software[34] and analyzed by using the BLASTX, PSI- [18] J. Piel, D. Hui, N. Fusetani, S. Matsunaga, Environ. Microbiol., in press.
BLAST, FramePlot, and InterProScan algorithms.
[19] T. W. Yu, L. Q. Bai, D. Clade, D. Hoffmann, S. Toelzer, K. Q. Trinh, J. Xu, S. J.
Moss, E. Leistner, H. G. Floss, Proc. Natl. Acad. Sci. USA 2002, 99, 7968.
The nucleotide sequence has been deposited at GenBank under [20] A. F. Neuwald, D. Landsman, Trends Biochem. Sci. 1997, 22, 154.
[21] B. S. Moore, C. Hertweck, Nat. Prod. Rep. 2002, 19, 70.
[22] R. Reid, M. Piagentini, E. Rodriguez, G. Ashley, N. Viswanathan, J. Carney, D. V. Santi, C. R. Hutchinson, R. McDaniel, Biochemistry 2003, 42, 72.
[23] P. Caffrey, ChemBioChem 2003, 4, 654.
[24] C. Cardani, C. Fuganti, D. Ghiringh, P. Grassell, M. Pavan, M. Valcuron, We thank H¸seyin BasÀpınar and Ekin SÀavk for helping with the [25] S. Textor, V. F. Wendisch, A. DeGraaf, U. M¸ller, M. I. Linder, D. Linder, W.
collections of Turkish beetle specimens, Elke Meier, Kathleen Seitz, Buckel, Arch. Microbiol. 1997, 168, 428.
[26] P. P. Halarnkar, J. D. Chambers, G. J. Blomquist, Comp. Biochem. Physiol.
Ivonne Hˆfer, and Domenica Schnabelrauch for technical assis- Part B: Biochem. Mol. Biol. 1986, 84, 469.
tance, and Konrad Dettner for valuable discussions. We are [27] Y. Paitan, E. Orr, E. Z. Ron, E. Rosenberg, Microbiology 1999, 145, 3059.
indebted to Chris M. Thomas for providing the mupirocin producer [28] H. M. Holden, M. M. Benning, T. Haller, J. A. Gerlt, Acc. Chem. Res. 2001, 34, Ps. fluorescens NCIMB 10586, and to Wilhelm Boland for support.
[29] V. Sharma, K. Suvarna, R. Meganathan, M. E. S. Hudspeth, J. Bacteriol.
This work was supported by Research Grants from the Deutsche Forschungsgemeinschaft (PI 430/1-1 and PI 430/2-1) and the Max [30] M. J. Gasson, Y. Kitamura, W. R. McLauchlan, A. Narbad, A. J. Parr, E.
Lindsay, H. Parsons, J. Payne, M. J. C. Rhodes, N. J. Walton, J. Biol. Chem.
1998, 273, 4163.
[31] A. Witkowski, A. K. Joshi, S. Smith, Biochemistry 2002, 41, 10 877.
Keywords: biosynthesis ¥ natural products ¥ polyketides ¥ [32] B. Silakowski, H. U. Schairer, H. Ehret, B. Kunze, S. Weinig, G. Nordsiek, P.
symbiosis ¥ synthases ¥ uncultivated bacteria Brandt, H. Blocker, G. Hˆfle, S. Beyer, R. M¸ller, J. Biol. Chem. 1999, 274,37 391.
[33] J. G. Lawrence, J. R. Roth, Genetics 1996, 143, 1843.
[34] J. K. Bonfield, K. F. Smith, R. Staden, Nucl. Acids Res. 1995, 23, 4992.
[1] J. Kobayashi, M. Ishibashi, Chem. Rev. 1993, 93, 1753.
[2] J. Faulkner, M. D. Unson, C. A. Bewley, Pure Appl. Chem. 1994, 66, 1983.
[3] M. G. Haygood, E. W. Schmidt, S. K. Davidson, J. D. Faulkner, J. Mol.
¹ 2004 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Source: http://sarton.imb-jena.de/publications/download/free/Piel_ChemBioChem_2004.pdf