A polyketide synthase-peptide synthetase gene cluster from an uncultured bacterial symbiont of Paederus beetles

Results

Amplification of Pederin-Specific Polyketide Synthase (PKS) PCR Products.

The structure of pederin and early labeling studies (13) indicate that the metabolite is largely synthesized from malonyl- and methylmalonyl-CoA units by a type I PKS. Such megasynthases consist of repeated modules, along which the growing polyketide chain is processed in an assembly line-like fashion (14). Each module minimally carries KS, acyltransferase (AT), and acyl carrier protein (ACP) domains to perform one chain-elongation cycle and optional additional domains to catalyze further modifications. To clone the pederin cluster, a PCR strategy was pursued involving degenerate primers based on universally conserved motifs of KS domains (15). Total DNA, isolated from different beetle specimens, was used as a PCR template. Analysis of three species of the genera Paederus and Paederidus collected at two different locations consistently revealed that only those adult beetles with high pederin content gave the PCR product expected to arise from the presence of PKS genes (Fig. 2). Amplification products were also obtained by using eggs from (+)-females of a fourth species, Paederus riparius, collected at Bayreuth, Germany (data not shown). Sequencing of the amplified DNA showed in all cases the exclusive presence of the same group of 3–4 different sequences exhibiting strong homology to KS domains of bacterial type I PKSs. The perfect correlation between pederin content and PKS sequences, independent of species and locality of collection, suggests that the amplified fragments belong to the pederin cluster. These DNA fragments therefore were used to locate the cluster.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

PCR amplification of PKS gene fragments from beetle total DNA. (a) TLC analysis of the ethanolic beetle extracts. The arrow indicates the position of pederin. Small amounts of pederin found in males and (−)-females are due to pederin transfer from (+)-females to the offspring. (b) Agarose gel of the PCR products obtained from total DNA of the same beetle specimens that were used for pederin extraction. L, 1-kb DNA ladder; B, negative PCR control without template DNA; 1, P. fuscipes, (+)-female, collected at Jena, Germany; 2, P. fuscipes, (−)-female, Jena; 3, P. fuscipes, male, Jena; 4, P. fuscipes, (+)-female, collected at Aydın, Turkey; 5, P. fuscipes, (−)-female, Aydın; 6, P. fuscipes, male, Aydın; 7, Paederidus rubrothoracicus, (+)-female, Aydın; 8, Pd. rubrothoracicus, (−)-female, Aydın; 9, Pd. rubrothoracicus, male, Aydın; 10, P. litoralis, (+)-female, Jena; 11, P. litoralis, male, Jena. No (−)-female of P. litoralis was available.

Localization of the PKS Gene Cluster on the Symbiont Genome.

A metagenomic cosmid library consisting of 80,000 clones was constructed from total DNA of P. fuscipes beetles. Initial quantitative PCR experiments had indicated that this library size was sufficient for a multiple coverage of the symbiont genome. By screening the library with specific PCR primers derived from the amplified sequences (see Materials and Methods), eight positive cosmids were identified. Sequencing of a 53.8-kb region revealed ORFs homologous to type I PKS genes, designated as ped genes (Fig. 3 and Table 1). All the amplified KS sequences could be found on these ORFs. Additional regions covering ≈65 kb outside of the cluster were obtained on two cosmids isolated by chromosome walking and subjected to extensive spot sequencing. All putative genes present on this >110-kb region exhibit typical bacterial features: they are tightly packed, free of introns and polyadenylation sites, and preceded by Shine–Dalgarno patterns in appropriate distances to the start codons. Furthermore, when subjected to database homology searches, each of the translated ORFs exhibited the highest similarity to bacterial proteins. Of the 68 deduced proteins analyzed, 27 are known exclusively from prokaryotes including enzymes used in vitamin B₆ and B₁₂ biosynthesis and regulatory proteins of the LuxR and LysR families. Interestingly, most ORFs located outside of the ped region, including almost all housekeeping genes, exhibit an extraordinary similarity to genes from P. aeruginosa, with an identity of >80% at the nucleotide level. This result is in perfect agreement with the 16S rRNA data (11) and provides clear evidence that the ped cluster belongs to a bacterial symbiont of the genus Pseudomonas.

• Download figure
• Open in new tab
• Download powerpoint

Fig 3.

Map of the sequenced ped genes and proposed pederin biosynthesis pathway. The product of PedF could be directly cleaved off the enzyme (route A) or elongated further by PedH to yield an onnamide-type intermediate and subsequently cleaved (route B). TP, transposase pseudogene; MT, methyltransferase; OR, oxidoreductase; OXY, oxygenase; C, NRPS condensation domain; A, NRPS adenylation domain; T, NRPS thiolation domain; DH, putative nonfunctional dehydratase domain.

View this table:

• View inline
• View popup

Table 1.

Deduced functions of the ped ORFs shown in Fig. 3

Domain Analysis of the ped Cluster.

Fig. 3 summarizes the results of the completely sequenced 53.8-kb ped region. The predicted gene products of the ORFs pedF and pedH are giant 8,601- and 6,266-aa proteins resembling mixed modular PKSs/nonribosomal peptide synthetases (NRPSs) (16, 17). In striking contrast to nearly all known PKSs, AT homologies are completely absent on these proteins. Their alignment with other known type I PKSs revealed that a continuous ≈300-aa region of each AT domain, including the active-site GHS motif, are deleted in PedF and PedG, with no other homologies replacing them (Fig. 4). This architecture results in a considerable shortening of each module. The missing AT domains are possibly complemented by the two isolated genes pedC and pedD upstream of pedF, which exhibit high similarity to monodomain ATs. Apart from the lacking ATs, however, PedF is perfectly colinear with pederin structure. A particularly characteristic feature is the second module (PedF2) with homology to NRPSs, catalytic entities involved in the attachment of a wide range of proteinogenic and nonproteinogenic amino acids. Previous studies have identified sequence patterns of NRPS adenylation (A) domains that allow a prediction of the incorporated amino acid (18, 19). The structure of pederin suggests that glycine is selected by PedF2. In accordance, a pattern analysis of its A domain revealed 100% identity to the known consensus nonribosomal code for this amino acid (ref. 20; Table 2). A second distinctive property of PedF is the rare methyltransferase (MT) domain of the PKS module PedF4. MTs integrated into PKS modules are known from yersiniabactin (21) and epothilone (22) biosynthesis, where they generate geminal dimethyl groups. Such a moiety is present also at the corresponding position of the pederin structure. A third notable feature is an unprecedented repeated dehydratase (DH) domain in the next module, PedF5. This module should be responsible for the closure of the dimethylated tetrahydropyrane ring, most likely via dehydration and subsequent Michael addition of the C-11 hydroxyl group, which is the reverse reaction. The catalysis of the identical transformation sequence by two DH copies is known also from nonactin biosynthesis, although the DH performing the addition step is encoded on a separate gene, and the product is a five-membered ring (23). Taken together, the architecture of PedF clearly indicates that this protein assembles the largest part of the pederin molecule except for the six-membered ring bearing the exomethylene group. A corresponding gene candidate responsible for the biosynthesis of this moiety could not be identified on the cluster.

• Download figure
• Open in new tab
• Download powerpoint

Fig 4.

Comparison of the first modules of PedF and the avermectin AVES2. The shaded area in AVES2-1 indicates the AT region that is deleted in PedF1. Conserved motifs are shown as vertical bars. Motifs are: 1, EPIAIV; 2, DPQQRL; 3, CSSS; 4, HGTGTxLGD; 5, GxGGxNAHVILEE; 6, AYTLQxGR; 7, GHSxG; 8, YPF; 9, GxDS. ACP, acyl carrier protein.

View this table:

• View inline
• View popup

Table 2.

Prediction of the amino acid specificity of the PedF2 and PedH6 NRPS modules based on known nonribosomal codes

Downstream of pedF lies pedH, encoding four additional PKS modules and one NRPS module. The presence of this gene was surprising at first, because it does not correspond to any part of the pederin structure. However, a detailed module analysis of pedH in fact revealed further substantial evidence for an involvement of the cluster in pederin biosynthesis. The terminal NRPS module PedH6 features a nonribosomal code most similar to the patterns of the arginine-specific syringomycin (24) and microcystin adenylation domains (ref. 25; Table 2) and contains a thioesterase domain invariably involved in release of assembled polyketide and peptide chains from the enzymatic machinery. PedH therefore should theoretically further elongate the pederin-type intermediate generated by PedF and release a compound having the hypothetical extended structure shown in Fig. 3 (route B). The remarkable similarity to most members of the pederin group isolated from sponges is obvious. These compounds include the 13 known onnamides and icadamide A, which all feature a variable polyketide side chain and a terminal arginine residue.

Discussion

Although the existence of bacterial symbionts as producers of invertebrate drug candidates has been proposed for many years, the general inability to grow such bacteria outside the host has prevented a direct proof of this hypothesis thus far. Our current view on such symbioses has been shaped largely by cell-separation studies conducted by Bewley and Faulkner (3), who showed that various sponge natural products in fact are localized in associated bacterial cells. However, these results do not rule out the possibility that bacteria are just importing metabolites synthesized by their hosts. To unequivocally identify an as-yet-uncultured bacterium as producer, the biosynthetic capabilities have to be examined at the molecular level. Gene cloning and heterologous expression studies, besides providing a glimpse into the metabolism of an otherwise unaccessible microorganism, should also unveil mechanisms of symbiosis and provide long-term supplies of rare invertebrate drug candidates (28). With this intention pederin biosynthesis in Paederus beetles was selected as a model for chemically mediated symbiosis. This singular terrestrial system has several advantages over the many suspected marine associations: an unparalleled amount of evidence that symbionts indeed exist (10), the likely presence of an almost pure bacterial culture as revealed by 16S rRNA analysis (11), and the unique opportunity to subsequently access highly complex and otherwise hard-to-study sponge systems by direct comparison of terrestrial and marine pederin-type gene clusters.

These cloning efforts led to the isolation of a contiguous 110-kb genomic region, which also was sequenced outside of the ped cluster to gain insight into the biology of the pederin producer. Because most ORFs, significantly those from primary metabolism, remarkably resemble genes of P. aeruginosa, the symbiont should be a very close relative of this species. This classification is supported also by recently conducted 16S rRNA gene studies (11) that revealed the presence of a bacterium with closest similarity to P. aeruginosa only in the “pederin-producing” (+)-females. P. aeruginosa is an opportunistic pathogen that causes serious and often fatal infections in immunocompromised and cystic fibrosis patients. It is virulent in a remarkably large variety of organisms including insects, which are used therefore as model infection hosts (29). An ancient infection event followed by genetic adaptation of the partners also may have resulted in the establishment of pederin symbiosis. Regardless of the evolutionary events bringing forth this association, the close relationship between the two bacteria now allows a direct comparison of symbiont genes with the completely sequenced P. aeruginosa genome. This makes the pederin symbiont a unique model organism not only to identify candidate genes governing symbiosis but also to help pinpointing pathogenic determinants in P. aeruginosa. Horizontal gene transfer allows bacteria to evolve in quantum leaps and is therefore the major driving force of bacterial adaptation. The importance of horizontally transferred genomic islands for the development of virulent or symbiotic traits in bacteria is revealed by an increasing number of studies (30). One cluster of genes that has no counterpart in P. aeruginosa and seems to have been horizontally acquired is the ped region. It is bordered on both sides by transposase pseudogenes, indicating that it has been inserted into the genome during a transposition event. This region indeed bears all the architectural and functional hallmarks of mobile symbiosis islands, which could explain the widespread occurrence of pederin-type metabolites in diverse organisms.

An unexpected finding was the complete absence of AT domains in the ped cluster. In modular PKSs, these domains are crucial for the selection of the correct acyl-CoA unit in each extension cycle and their attachment to the acyl carrier protein domain (31), raising the question of how this process is catalyzed during the assembly of pederin. PedC and PedD, encoded upstream of pedF and resembling monodomain ATs, could perform this function in trans. The ped cluster then would encode a type I protein complemented by repetitive type II enzymes. Recent studies indicate that such unusual mixed type I/type II PKS systems are more widespread in nature. A similar PKS containing such putative “super ATs” is the only partly described pksX cluster (32) on the genome of Bacillus subtilis. Its gene products consist of a large number of PKS modules without AT domains and, encoded at the upstream end of the cluster, three isolated ATs. The secondary metabolite generated by these proteins is not known. Another related system is the mupirocin PKS from Pseudomonas fluorescens recently submitted to GenBank (accession no. ; A. K. El-Sayed, J. Hothersall, S. M. Cooper, E. Stephens, and C. M. Thomas, unpublished data). This cluster contains a single gene with two AT domains. Apart from these two systems, a small number of further PKS modules lacking AT domains are known from the albicidin (33) and TA biosynthesis gene clusters (34). Both systems have only been analyzed in part thus far, but it is very likely that they also contain discrete AT genes. Sequence comparison with other known PKSs (data not shown) reveals that all enzymes lacking AT domains are related more closely to each other than to any other known modular PKS, independent of bacterial phylogeny. They therefore could belong to a phylogenetically distinct subgroup of functionally novel PKS systems.

To ultimately prove the involvement of the ped genes in pederin biosynthesis, loss-of-function studies after gene inactivation are needed. Naturally, the inability to culture the Paederus symbionts forbids such an approach. However, several lines of evidence provide compelling evidence that the cluster indeed encodes the pederin pathway. First, there is a strict correlation between the ability to synthesize pederin and the presence of identical PKS PCR products. These products are present exclusively in (+)-females and their eggs and are amplified from a wide range of species regardless of the locality of collection. Second, the fact that all PCR products belong to a single cluster argues against the possibility that pederin biosynthesis is encoded on another, undetected PKS cluster. The degenerate PCR primers were not designed originally for the amplification of sequences from aberrant PKS clusters lacking AT domains. Nevertheless, 4 of a possible 10 PCR products from the ped system were obtained, documenting the broad specificity of these primers for modular PKSs. It seems unlikely that a second large cluster comprised of at least eight PKS modules would not produce any amplification products. Third and most important, the architecture of the ped cluster perfectly matches the hypothetical biosynthetic pathway of pederin-type compounds (Fig. 3). Every single domain of PedF agrees with the corresponding pederin moiety, which includes diagnostically valuable and rare features such as the NRPS module with glycine nonribosomal code, the methyltransferase domain presumably attaching the geminal dimethyl group, and the repeated dehydratase domain most likely responsible for the pyran ring formation. Also in agreement with pederin biosynthesis are the post-PKS tailoring genes. Pederin contains four O-methyl groups, of which at least two could be attached by the methyltransferases PedA and PedE. One of the most distinctive features of pederin, however, is the oxidized single-carbon terminus that is extremely rare among nonrelated polyketides. This moiety should result from an oxidative cleavage of a two-carbon extension unit. Strikingly, pedG encoding an FAD-dependent monooxygenase is inserted precisely behind the last complete module corresponding to this unit. Members of this enzyme family are known to catalyze oxidative cleavages in a Baeyer–Villiger fashion and include MtmOIV, which cleaves a precursor of the aromatic polyketide mithramycin (35), and cyclohexanone monooxygenase (36). PedG therefore seems to be a good candidate to perform the proposed pederin cleavage.

An interesting issue is the function of the additional PKS-NRPS gene pedH. Although the lack of a structural counterpart to this gene at first glance seems puzzling, a more detailed analysis of its architecture strongly suggests its involvement in the synthesis of a pederin-type metabolite. The striking and unexpected similarity of the deduced full-length biosynthetic product of PedF and PedH to the large series of marine pederin analogues with terminal arginine residues raises fascinating new questions about the evolutionary relationship between marine and terrestrial pederin members. Fig. 3 shows two possible mechanistic hypotheses for pederin assembly. It could be envisioned that the ped cluster originally generated a sponge-type compound but that in the course of evolution the oxygenase gene pedG inserted into the cluster, thereby destroying one module (route A). Consequently, this would lead to a premature termination of the PKS elongation process at the module PedF6, which is exactly the module corresponding to the terminal pederin extension unit. Because a thioesterase domain is lacking on this gene, the chain would remain attached to this module until it is cleaved off oxidatively by PedG. This scenario would not only explain the module halves on both sides of pedG but also why pedG is located at exactly this position of the cluster. Alternatively, the ped system could first synthesize an extended sponge-type intermediate, which then is cleaved by PedG in a Baeyer–Villiger oxidation to give an appropriately shortened pederin precursor and a second fragment (route B). If this mechanism is correct, then the fragmented module should be functional. Such subdivided but functional PKS modules are unusual but indeed exist, as a single known example from the myxalamid cluster documents (37).

A number of genes required for pederin biosynthesis were not found on the cluster. Missing are a small PKS comprising the first three modules, a protein necessary for the attachment of the exomethylene group, and the enzyme catalyzing the formation of the aminal moiety. Although in bacteria genes from secondary metabolism are typically clustered together, there are a few known exceptions to this rule. Examples of incomplete clusters include the ansamitocin PKS genes, which are located on two regions separated by 30 kb of nonrelated genes (38), and the enterocin cluster, which lacks a dehydrogenase participating in the formation of the polyketide starter unit (39). The presence of transposase pseudogenes at both sides of the ped region strongly suggests that the cluster has been fragmented during genomic rearrangement processes.

This work provides insight into the biosynthesis of a structurally unique insect defense compound and the evolutionary relationship to the related sponge metabolites. It also shows for the first time that genes responsible for the biosynthesis of rare invertebrate drug candidates can be cloned from bacterial symbionts without the need for culturing. Because whole sets of type I PKS genes have been functionally expressed even in Escherichia coli (40), a similar production of “invertebrate” natural products in a suitable host now has become a realistic scenario.