|Home | About | Journals | Submit | Contact Us | Français|
The major cyclic peptide cyanopeptolin 1138, produced by Planktothrix strain NIVA CYA 116, was characterized and shown to be structurally very close to the earlier-characterized oscillapeptin E. A cyanopeptolin gene cluster likely to encode the corresponding peptide synthetase was sequenced from the same strain. The 30-kb oci gene cluster contains two novel domains previously not detected in nonribosomal peptide synthetase gene clusters (a putative glyceric acid-activating domain and a sulfotransferase domain), in addition to seven nonribosomal peptide synthetase modules. Unlike in two previously described cyanopeptolin gene clusters from Anabaena and Microcystis, a halogenase gene is not present. The three cyanopeptolin gene clusters show similar gene and domain arrangements, while the binding pocket signatures deduced from the adenylation domain sequences and the additional tailoring domains vary. This suggests loss and gain of tailoring domains within each genus, after the diversification of the three clades, as major events leading to the present diversity. The ABC transporter genes associated with the cyanopeptolin gene clusters form a monophyletic clade and accordingly are likely to have evolved as part of the functional unit. Phylogenetic analyses of adenylation and condensation domains, including domains from cyanopeptolins and microcystins, show a closer similarity between the Planktothrix and Microcystis cyanopeptolin domains than between these and the Anabaena domain. No clear evidence of recombination between cyanopeptolins and microcystins could be detected. There were no strong indications of horizontal gene transfer of cyanopeptolin gene sequences across the three genera, supporting independent evolution within each genus.
Cyanobacteria produce a wealth of secondary metabolites, including enzymatically synthesized peptides. The participating enzymes, nonribosomal peptide synthetases (NRPSs), are often huge enzymes with modular architectures where each module is responsible for activation and modification of an amino acid to be incorporated in the peptide product. The modules consist of several domains, of which adenylation (A), condensation (C), and thiolation (T) domains are commonly present, while other types of tailoring domains frequently occur. Amino acids are covalently coupled by thioester bonds to phosphopantetheine prosthetic groups attached to the modules and transferred from one module to the next in an orderly fashion under formation of a peptide bond (38).
In most cases, the order of modules is colinear with the amino acid residue sequence of the produced peptide. Thus, the structure of the peptide to some extent can be deduced from protein sequence data by examining binding pocket signature sequences (6, 39), phylogenetic placements, and the order in which modules are arranged (21). Combined with characterization of peptides produced by the organism and experimental evidence of the relationship between gene and peptide, e.g., through knockout studies, in silico analyses of sequence data constitute powerful tools for resolving the functions of unknown tailoring domains and A domains with unknown specificities (33, 41).
More than 600 peptides can be found in genera from all sections of cyanobacteria, and Welker and von Döhren have grouped the peptides in seven peptide families (43). The major peptide families are common to several cyanobacterial genera, but large variations as to what peptides they actually produce are observed among individual strains. Some completely lack NRPSs and produce no peptides in this way, some produce one peptide, and yet others produce peptides belonging to several peptide families. Such variation often is due to a combination of loss of functional regions (e.g., nodularin synthetase  and microcystin synthetase ), gain of genes through horizontal gene transfer (HGT) (e.g., cyanopeptolin synthetase [36, 45]), and intragenomic recombinational events (e.g., microcystin synthetase [9, 21, 24, 43]). Comparisons of gene clusters within and across families may give valuable information concerning the evolutionary histories of the clusters. Microcystin synthetase gene clusters have been extensively characterized for Microcystis (26, 27, 40), Anabaena (34), and Planktothrix (8), making it possible to identify recombination sites (22), transposon-promoted inactivations (9), and other causes of genetic variation in these gene clusters among cyanobacterial genera (19, 20). Based on the phylogenetic concurrence between a few housekeeping genes and microcystin synthetase genes, Rantala et al. (29) have suggested that microcystin gene clusters are evolutionarily old and have a common ancestor. Whether this hypothesis can be extended to other NRPS gene clusters and to what extent coevolution and recombination between microcystin synthetase genes and other NRPS genes have contributed to the impressive variation found in the NRPS gene families are issues still far from elucidated.
Another, less extensively characterized family of nonribosomally produced peptides, the cyanopeptolins, is also produced by strains from several genera of cyanobacteria, including several microcystin-producing strains. Chemically, cyanopeptolins are quite variable, both among strains of the same genus and among different genera of cyanobacteria (43). The general structure includes seven amino acid residues, of which six form a ring. All cyanopeptolins contain the residue 3-amino-6-hydroxy-2-piperidone (Ahp) and an ester linkage between the β-OH group of l-threonine and the carboxyl group of the C-terminal amino acid (24, 42). Cyanopeptolin biosynthesis has been elucidated for Anabaena, where a cyanopeptolin knockout has been created (33), and for Microcystis (41). Within the two cyanopeptolin gene clusters characterized to date, the arrangement of modules, the presence of ABC transporters, and some of the tailoring domains are common to the encoded synthetases, while A-domain specificities and the presence of additional tailoring domains vary (33, 40). It is not clear to what extent such similarities are conserved in all cyanopeptolin synthetases (41). Characterization of additional cyanopeptolin synthetase genes will provide an insight into the variability and evolutionary history of a nonmicrocystin peptide synthetase gene family. Further, comparisons of cyanopeptolin and microcystin synthetase genes across genera may demonstrate differences in the modes and rates of evolution of these genes. Since in many strains microcystin and cyanopeptolin synthetase genes coexist in the same genome, such studies may help decide to what extent exchange of genetic information between these gene clusters has contributed to the creation of diversity.
Here, we analyze the sequence of a cyanopeptolin gene cluster from the genus Planktothrix (Planktothrix agardhii NIVA CYA 116). Based on matrix-assisted laser desorption ionization-time of flight mass spectrometry (MS) screening, the strain produces no microcystins and a single major peptide. We show that this peptide is a close relative of a previously characterized cyanopeptolin (16). In accordance with this, a single NRPS gene cluster was found by using PCR with consensus primers combined with primer walking. Comparison of the new gene cluster to those previously described showed several conserved features but also the presence of novel tailoring domains. By performing separate A- and C-domain phylogenetic analyses on this novel gene cluster and the two previously known cyanopeptolin gene clusters, as well as microcystin, nostopeptolide (13), and nostocyclopeptide (2) gene clusters, we have phylogenetically addressed the evolution of the individual domains and associations between gene clusters to assess the contribution of HGT events (intergenomic recombination), intragenomic recombination, and other mutational events to cyanopeptolin evolution.
Planktothrix agardhii NIVA CYA 116 was isolated in 1983 from Lake Årungen, Norway (NIVA culture collection). For these experiments, it was cultured in Z8 medium (17) at about 18°C, with 12 h of illumination with 15 μmol photons m−2 s−1 in a Sanyo environmental test chamber (FG-4P 36-40).
DNA from Planktothrix strain NIVA CYA 116 was isolated with Dynabeads (Invitrogen, Carlsbad, CA) according to the protocol of Rudi et al. (36). Using BD Advantage 2 DNA polymerase (BD Biosciences, Mountain View, CA), we amplified A domains with the degenerated primers MTF2 (GCNGGYGGYGCNTAYGTNCC) and MTR (CCNCGDATYTTNACYTG), targeting conserved regions in the NRPS A domains (25), and regions between the obtained A domains with specific primers designed from sequences obtained by MTR and MTF2. The fragments were cloned by utilizing a TA TOPO cloning kit (Invitrogen), and several overlapping fragments were sequenced by primer walking with an ABI 3730 sequencer and BigDye v3.1 solution. Amplification of unknown flanking regions by linear PCR were followed by oligo(dC) tailing with terminal transferase and PCR with oligo(dG) and a specific primer (35), used to amplify additional NRPS regions (see Table S1 in the supplemental material). Fragments were assembled, and open reading frames (ORFs) were translated using Vector NTI (Invitrogen). Total RNA was isolated using an RNeasy kit (QIAGEN, Crawley, United Kingdom) and tested for DNA contamination by PCR. mRNA transcribed from the oci operon was detected by real-time PCR after cDNA synthesis with Trancriptor reverse transcriptase (Roche, Mannheim, Germany).
Identification of domains and binding pocket signatures and putative assignments of A-domain substrate specificities were performed using the NRPS database (http://www.nii.res.in/nrps-pks.html). A-domain specificities were also substantiated using phylogenetic analysis. A- and C-domain protein sequences were aligned using MEGA 3.1 (18), and the model of protein evolution that best fit the given set of sequences was found using ProtTest (1). Phylogenetic trees were constructed utilizing MrBayes (14) 3.0 and 3.1 (31) at the University of Oslo Bioportal (http://www.bioportal.uio.no). Analyses were performed using the optimal-protein-evolution model, and variable substitution rates across sites were accounted for by using gamma-shaped distribution. The MCMC chains were carried out for 4 million generations, with sampling of trees every 100 generations and removal of 3,000 trees sampled before the MCMC chains reached convergence. The consensus of the remaining trees was used to calculate the posterior probabilities of the clades. Neighbor-joining (NJ) trees were constructed using MEGA 3.1 with default settings (Poisson correction as the amino acid substitution model) (18). Split decomposition analyses were performed using SplitsTree4 with default settings (the uncorrected P method) and 1,000 bootstrap replicas (15) and a Phi test for recombination (4). Similarity and identity calculations were done at the amino acid level, using Vector NTI.
For structural analysis of the major peptide produced by the present strain, a methanolic extract of lyophilized NIVA CYA 116 cells was used.
Liquid chromatography (LC)-MS instrumentation included a Waters Acquity UPLC system equipped with an Atlantis column (C18, 2.1 by 150 mm, 5-μm particle size; Waters) and set to run a gradient starting with 80% solvent A (10 mM ammonium acetate, 0.1% acetic acid) and ending with 60% solvent A after 15 min. Solvent B was methanol with 0.1% acetic acid. The flow rate was 0.2 ml min−1. The LC system was connected to a Waters Quattro Premier XE tandem quadropole MS equipped with an electrospray probe. The detector was run in the positive-ion mode at a cone voltage of 50 V. A total ion scan from 600 to 1,400 Da was performed during the entire length of the LC gradient.
The structure of the major compound detected was analyzed by MS fragmentation studies. MS fragments hold valuable structural information and have earlier been successfully used for identification and structural elucidation of cyanobacterial oligopeptides, including cyanopeptolins (10, 11). Fragmentation experiments were carried out with the hardware configuration described above. The MS was run in product/fragment ion scanning mode, and all settings were automatically optimized for the fragmentation of the compound to be studied at a collision energy of 30 eV. Fragments were recorded during the entire length of the LC gradient. The identification of fragments was aided by comparison to fragment patterns of already known cyanopeptolins and by HighChemMassFrontier software version 3.0. Since the genetic analysis and the molecular mass suggested a composition similar to that of oscillapeptin E from Planktothrix strain NIES 205, all fragmentation experiments were repeated with extracts of NIES 205.
The oci gene cluster sequence has been deposited in GenBank (accession no. DQ837301).
The initial screening of NIVA CYA 116 for low-molecular-weight compounds showed one large LC peak with an MS signal at 1,138 Da, [M + H]+. The same peak also included signals at 1,120 and 1,058 Da, along with a strong signal at 1,040 Da. A fragmentation experiment with the 1,138-Da ion as the parent identified the other three signals as product/fragment ions of the original compound. The same pattern has been observed for a cyanopeptolin with a known structure, oscillapeptin E from Planktothrix strain NIES 205 (16), and is consistent with a cyanopeptolin with a sulfate group bound via glyceric acid to the N-terminal side of the molecule (1,120-Da signal, [M + H − H2O]+; 1,058-Da signal, [M + H − SO3]+; 1,040-Da signal, [M + H − H2O − SO3]+) (3).
To obtain reliable fragmentation data, the abundant ion [M + H − H2O − SO3]+ was used for the main fragmentation experiment. Three findings from the MS analyses indicate that the NIVA CYA 116 cyanopeptolin is very similar or identical to oscillapeptin E: (i) NIVA CYA 116 cyanopeptolin and oscillapeptin E have the same molecular mass, (ii) the initial screening for low-molecular-mass compounds in both cases gave a cluster of signals consisting of the molecule peak and the [M + H − H2O]+, [M + H − SO3]+, and [M + H − H2O − SO3]+ product/fragment ions, and (iii) the two compounds showed very similar patterns in the fragmentation experiments (see Fig. S1 in the supplemental material). However, spiking experiments showed that the NIVA CYA 116 cyanopeptolin was slightly less polar than oscillapeptin E (with a 13-second difference in retention time under the condition used here). Therefore, the two cyanopeptolins cannot be completely identical. Possible differences include a replacement of Ile with Leu or allo-Ile (at positions 5 and/or 7) and a different combination of d and l amino acids that may lead to differences in solution conformation. Our data thus suggest that the cyanopeptolin produced by NIVA CYA 116 is a previously nondescribed compound with a molecular mass of 1,138 Da ([M + H]+) and most likely the structure shown in Fig. Fig.1.1. We named the compound cyanopeptolin 1138.
Using a PCR/primer walking strategy, we obtained the sequence of a complete cyanopeptolin gene cluster, oci. The identified gene cluster, approximately 30 kb in size, contains four ORFs. Three ORFs containing the NRPS modules, ociA (10.4 kb), ociB (14.2 kb), and ociC (4.3 kb), are transcribed in the same direction. The fourth ORF, ociD, encodes a putative ABC transporter. Real-time PCR on total RNA indicated the presence of a cyanopeptolin synthetase mRNA (data not shown). The structural organization of the cyanopeptolin oci gene cluster is shown in Fig. Fig.22.
The oci operon encodes totals of seven A domains, seven C domains, and eight T domains (also called PCP domains). In addition to the two tailoring domains described below, ociA includes two NRPS modules containing C, A, and T domains. ociB consists of four modules and ociC a single module and a termination domain. In its last module, ociB encodes a methyltransferase domain that putatively methylates the amino acid activated by the sixth A domain. The seven A-domain amino acid sequences and the GrsA-Phe A domain (6) were aligned, and the signature residues (6, 39) in the binding pocket and the amino acid most likely to be activated were identified (see Table S2 in the supplemental material).
Although database searches gave inconclusive results for the A1 domain, both binding pocket residues and phylogenetic analyses (Fig. (Fig.3A)3A) suggest activation of homotyrosine (Htyr), due to the similarities to the A3 domain in the apd gene cluster (33). A2 has a characteristic Thr-binding pocket signature sequence. The signature sequence derived for the A3 domain corresponds to none of the defined signature sequences, and the domain sequence clustered apart from the other A-domain sequences in the phylogenetic analysis (Fig. (Fig.3A),3A), making it impossible to predict the activated amino acid based on sequence alone. If the gene cluster encodes the synthetase polypeptide complex that produces the peptide characterized above, the A3 domain must activate Htyr. The binding pocket signature sequence of the A4 domain is identical to the A4-domain signatures in the apd and mcn operons. In addition, the high degree of identity between oci A4 and the A4 domains in mcn and apd (96 and 84%, respectively) and the phylogenetic grouping of the A4 domain (Fig. (Fig.3A)3A) strongly suggest that the Oci A4 domain incorporates 3-amino-6-hydroxy-2-piperidone (Ahp) in the peptide. The A5, A6, and A7 domains have binding pocket signatures that occur in the NRPS database and are presumed to activate Ile, Phe, and Ile, respectively. This was also supported by their positions in the phylogenetic tree (Fig. (Fig.3A3A).
The C-domain phylogeny (Fig. (Fig.4A)4A) shows that none of the C domains clustered with C domains that utilize d amino acids as donors (Fig. (Fig.4A),4A), suggesting that the amino acids in the peptide all are in the l configuration (32). The C-domain phylogeny essentially showed clustering according to gene cluster affiliation and position in the gene cluster, as expected from previous studies (41).
The overall structure of the Planktothrix cyanopeptolin gene cluster (oci) is similar to those of the cyanopeptolin gene clusters in Microcystis (mcn) (41) and Anabaena (apd) (33), and the numbers of modules and ORF orientations are the same (Fig. (Fig.2).2). The most apparent differences concern the presence or absence of tailoring domains (halogenases and sulfotransferases) and the binding pocket signatures of some of the equivalent A domains. ociB corresponds to mcnC and apdB (Fig. (Fig.2),2), and after removal of the apd methyltransferase-encoding sequence, ociB displays 81% nucleotide sequence identity to mcnC and 67% identity to apdB. ociC has 83% and 74% identity to mcnE and apdD, respectively. ociA is substantially longer than the corresponding genes in the other strains and encodes other tailoring domains.
Phylogenetic analyses were performed on A domains from the cyanopeptolin genes and microcystin genes (8, 26, 27, 34, 40), and the results are shown in Fig. Fig.3A.3A. The A domains cluster according to the apparent amino acid specificity, except for McnC A3 and McnE A7 (supposedly activating Ile and Gln, respectively, based on the binding pocket signatures) (41). Generally, the oci and mcn A domains appear to be more closely related to each other than to the apd A domains within clades containing domains activating the same or similar classes of amino acids (e.g., the Ahp, Ile, Thr, and aromatic clades). Notably, the various microcystin A domains give no such consistent pattern. The topology of the A-domain split tree suggests two possible recombination events (Fig. (Fig.3B);3B); however, the Phi test did not find statistically significant evidence for recombination (P = 0.99) (4).
A C-domain alignment was constructed using the amino acid sequences from cyanopeptolin, microcystin, nostocyclopeptide (2), and nostopeptolide (13) domains. The phylogenetic tree indicated separate clusters for positions C6 and C7 for mcn, oci, and apd domains and for positions C3, C4, and C5 for oci and mcn domains. Within clades where the C domains of all three cyanopeptolin synthetases were represented, the oci C domains again seemed more closely related to their mcn counterparts than to the apd counterparts, as demonstrated for the A domains. Clustering according to C-domain position or function was also observed for the microcystin C domains, but no consistent phylogeny was found. In accordance with Roongsawang et al. (32), our results indicate that C domains that add non-amino acid residues (OciA C1 and McnA C1) are different from the regular C domains. C domains from the nostopeptolide (nos) and nostocyclopeptide (ncp) gene clusters (from the Nostoc clade) and C1, C2, C3, and C4 domains from the anabaenopeptilide (apd) gene cluster (from Anabaena) group according to species or NRPS system. SplitsTree analysis (Fig. (Fig.4B)4B) suggested two possible recombination events in modules C6 and C7 in the C domains. Again, the Phi test did not find statistically significant evidence for recombination between C domains (P = 0.82) (4). Generally, microcystin C domains form longer branches than the cyanopeptolin C domains do.
ociA begins with an approximately 2,700-bp segment encoding a non-NRPS domain in module 1 (Fig. (Fig.2).2). At the amino acid level, the 200- to 1,000-bp region was 37% identical to fkbM, which encodes a polyketide synthase subunit assumed to have O-methyltransferase activity in Streptomyces hygroscopicus var. ascomyceticus (44). The remaining portion of the segment showed 41% identity to fkbH, a gene involved in ascomycin biosynthesis with a possible role in the formation of glyceryl-acyl carrier protein, with a three-carbon glycolysis intermediate as the substrate (44). The loading domain in OciA thus appears to be a combination of FkbH and FkbM homologs (5, 12, 30, 44). This domain is likely to catalyze the incorporation of a methylated glyceric acid extender unit. Such a function is consistent with its similarity to FkbH and FkbM, and we have therefore tentatively named it the GA domain (Fig. (Fig.2).2). A T domain and a sulfotransferase domain follow this domain in the OciA polypeptide.
The sulfotransferase domain, identified by searching the domain database Pfam, belongs to the Pfam sulfotransferase domain family (Pfam accession number PF00685), and a protein BLAST search showed 40% identity to the CurM sulfotransferase domain from the marine cyanobacterium Lyngbya majuscula (7) (data not shown). The presence of this domain in ociA adds further credibility to the notion that the oci gene cluster is responsible for the synthesis of the sulfated cyanopeptolin 1138 compound characterized above.
A 379-bp intergenic region separates the putative ABC transporter gene (ociD) from ociA. ociD is transcribed in the opposite direction relative to the main gene cluster. The gene is 2,064 bp long, and the encoded protein shows high similarity to other ABC transporters linked to NRPSs. A sequence alignment of the ociD protein and the ABC transporter encoded by the Microcystis cyanopeptolin gene cluster (mcnF) showed 86% identity.
In the phylogenetic analysis, NRPS-associated ABC transporters cluster according to the peptides produced by the corresponding synthetases and not according to species (Fig. (Fig.5).5). An NJ tree shows the same topology (data not shown).
Chemical analysis and screening for A domains by using degenerate primers indicated that Planktothrix strain NIVA CYA 116 produces a single major cyclic peptide and contains a single NRPS gene cluster. The main features of the peptide structure were elucidated by LC-MS-MS analysis and comparison to a previously characterized peptide, ocillapeptin E, with a known structure (16). The A-domain order deduced from the operon sequence matched the peptide structure and strongly indicates that the oci-encoded NRPS produces cyanopeptolin 1138. Since the in silico analyses of binding pockets clearly suggest Ile at positions 5 and 7, the difference in retention time on the LC column between ocillapeptin E and cyanopeptolin 1138 is most likely due to a different combination of d and l amino acids.
The gene cluster described here constitutes a third characterized member of the cyanopeptolin gene family. The cluster includes sequences coding for two unique domains not previously associated with cyanopeptolin synthetases and an A domain with a unique binding pocket. Given that cyanopeptolin 1138 is produced by the encoded NRPS, the novel A domain activates Htyr.
No known amino acid or nucleotide sequence shows similarity to the entire GA-domain sequence. The domain appears to be a combination of homologs of fkbH and fkbM, both part of a putative methoxymalonyl coenzyme A synthetase gene cluster (44). A bryA domain sequenced from a marine bacterial symbiont (12), gdmHIJK, involved in geldanamycin biosynthesis (30), and asm16, involved in ansamitocin biosynthesis (5), also show some similarity to parts of the GA domain (data not shown). Based on the mosaic pattern observed in the BLAST search (data not shown), one might speculate that the GA domain is a product of a deletion or recombination event.
The sulfotransferase domain is also a novel feature of NRPS genes. A BLAST search of the NCBI database gave a number of significant hits and showed the highest similarity to the sulfotransferase domain in CurM from the marine cyanobacterium Lyngbya majuscula (data not shown). The CurM sulfotransferase domain has an unknown function in the polyketide synthetase-catalyzed synthesis of curacin A, which contains no sulfate (7). Based on the module sequence encoded by ociA, a possible scheme for the biosynthesis of the cyanopeptolin 1138 compound is as follows. The GA domain loads a glyceric acid residue to the T domain, and the enzyme-bound glyceric acid then is methylated and sulfated. This initial unit and the first amino acid in the peptide chain are joined together by the first C domain in ociA, after which transfer of the intermediates from module to module and formation of new peptide bonds proceed in an orderly fashion. A rigorous biochemical analysis is needed to confirm this proposed sequence of events.
The general arrangements of the cyanopeptolin gene clusters in Anabaena, Microcystis, and Planktothrix are similar. The transcription directions are the same, the sizes of several of the ORFs are similar, and all three clusters have associated ABC transporters. Although the cyanopeptolin gene clusters are similar, some striking differences can be observed. For instance, a halogenase gene is present in apd and mcn, leading to the addition of chloride to the Tyr introduced in the peptide by the apdB A6 domain, but is absent in oci, where the corresponding amino acid introduced by oci A6 appears to be Phe. Differences in tailoring domains and A-domain specificities among the three cyanopeptolin gene clusters suggest that after spreading to different lineages, the genes independently have been subjected to several intragenomic and/or intergenomic recombination events resulting in acquisition and loss of various tailoring genes and changes in amino acid specificity. Tailoring gene recombinations could not be tested with split decomposition and Phi tests, since each gene cluster contains different tailoring genes. However, recombinations resulting in new gene cluster variants are likely, since intragenomic recombinations and HGT generally seem to play an important role in the evolution of cyanobacteria (45).
In the oci gene cluster, the position of the ABC transporter (ociD) corresponds to that of mcnF, but with a longer intergenic region, while the gene for the predicted apd ABC transporter is located on the opposite side of the NRPS genes. This corroborates a closer relationship between the oci and mcn genes than between these and apd, in line with the phylogenetic trees. Although the function of the ABC transporters that are linked to most NRPSs is unknown, the transporters seem to be essential for the function of the peptide (28). The phylogenetic analyses strongly suggest that the ABC transporters have evolved together with their respective groups of NRPS genes (Fig. (Fig.5).5). However, the “deepest” branches are unfortunately unresolved and therefore the relationship between the groups of NRPS-associated ABC transporters (i.e., those associated with mcy, nda, nos, ncp, and cyanopeptolins) cannot be elucidated from our data.
Cyanopeptolins and microcystins have coexisted in the Microcystis, Planktothrix, and Anabaena (Nostoc) groups most likely for a long evolutionary period (29). Thus, a possible scenario is that the two types of gene clusters containing the same basic building blocks (i.e., domain types) have recombined frequently during the evolutionary history. Although the splits in Fig. Fig.3B3B and and4B4B suggested recombination between the three gene clusters, the Phi tests did not lend statistical support. Thus, the simplest explanation is that the ancestral gene cluster closely resembled what exists today, and subsequent to the diversification of the three lineages Microcystis, Planktothrix, and Anabaena, the tailoring domain coding regions of the gene clusters were introduced by independent recombination events within each lineage. These recombinations could have been intergenomic or intragenomic. Such an evolutionary history is similar to what has been suggested for microcystin synthetases (29).
In the A- and C-domain trees, branches leading to domains derived from microcystin synthetases generally appear longer than the cyanopeptolin synthetase branches. This suggests either higher evolutionary rates within the microcystin gene cluster or a more recent origin (i.e., a more recent divergence of the lineages) for the cyanopeptolin gene cluster. Interestingly, the consistent phylogenies obtained from sequences of corresponding domains in the cyanopeptolin synthetases, with Microcystis and Planktothrix domains being more closely related to each other than to the Anabaena domains, were not observed for the microcystin domain groups. Again, this suggests that microcystin synthetase gene clusters may have a more complex evolutionary history than cyanopeptolin gene clusters. For microcystin gene clusters, various recombination events and inactivations have been identified (19, 20, 22), but with the exception of nodularin (23), these have taken place between members of the same genera. According to the original suggestion by Rudi et al. (37) and recently shown by whole-genome analyses (45), intergenomic recombination is expected to occur far more frequently between closely related strains, such as strains within the same genera. Thus, investigations of inter- and intragenomic recombinations within the cyanopeptolin gene cluster between closely related species are likely to provide the crucial information.
We are grateful to Ann-Kristin Hansen and Beatriz Decenciere for excellent technical assistance, Peder Magnus Haugen for numerous helpful discussions, and Randi Skulberg at NIVA for providing the strain NIVA CYA 116. This paper is dedicated to the memory of P. M. Haugen.
The work was supported by a grant to project 157338/140 from the Norwegian Research Council.
Published ahead of print on 5 October 2007.
†Supplemental material for this article may be found at http://aem.asm.org/.