|Home | About | Journals | Submit | Contact Us | Français|
Toxic Microcystis strains often produce several isoforms of the cyclic hepatotoxin microcystin, and more than 65 isoforms are known. This has been attributed to relaxed substrate specificity of the adenylation domain. Our results show that in addition to this, variability is also caused by genetic variation in the microcystin synthetase genes. Genetic characterization of a region of the adenylation domain in module mcyB1 resulted in identification of two groups of genetic variants in closely related Microcystis strains. Sequence analyses suggested that the genetic variation is due to recombination events between mcyB1 and the corresponding domains in mcyC. Each variant could be correlated to a particular microcystin isoform profile, as identified by matrix-assisted laser desorption ionization-time of flight mass spectrometry. Among the Microcystis species studied, we found 11 strains containing different variants of the mcyABC gene cluster and 7 strains lacking the genes. Furthermore, there is no concordance between the phylogenies generated with mcyB1, 16S ribosomal DNA, and DNA fingerprinting. Collectively, these results suggest that recombination between imperfect repeats, gene loss, and horizontal gene transfer can explain the distribution and variation within the mcyABC operon.
Cyanobacteria are phototrophic organisms that often form water blooms in eutrophic or estuarine waters. These water blooms undergo fluctuations and may exhibit toxic states. One common genus in such water blooms, Microcystis, produces the hepatotoxin microcystin (6). There are approximately 65 known isoforms of microcystin, representing a family of cyclic heptapeptides having the common structure cyclo(d-Ala-l-X-d-MeAsp-l-Z-Adda-d-Glu-Mdha), where l-X and l-Z are variable l amino acids, Adda is 3-amino-9-methoxy-2,6,8,-trimethyl-10-phenyl-4,6-decandienoic acid, d-MeAsp is 3-methyl-aspartic acid, and Mdha is N-methyl-dehydroalanine (Fig. (Fig.11).
Microcystin is produced nonribosomally by the microcystin synthetase enzyme complex via a thio-template mechanism (2). Nonribosomal peptide synthetase genes consist of modules that are built up of domains, and each module activates one amino acid, which is incorporated into the growing peptide chain in the order in which the modules are arranged. Most modules contain adenylation, thiolation, and condensation domains, and the adenylation domain is responsible for recognition of the specific amino acid. After the amino acid is activated to its acyladenylate, the aminoacyl adenylate is transferred to the 4′-phosphopantetheine carrier within the thiolation domain. Peptide bond formation between two activated amino acids is mediated by the condensation domain (for reviews see references 24 and 30). There are 10 conserved motifs, designated A1 to A10, within the adenylation domain. The 10 amino acid residues lining the substrate-binding pocket are also located within this region and are believed to be responsible for the substrate specificity. Nine of these amino acids are located between core motifs A4 and A5.
Microcystins mediate their toxicity through inhibition of the eukaryotic serine/threonine protein phosphatase 1 and 2A activities (13, 14, 20, 28). There have been several reports of death of livestock and humans due to massive hepatic hemorrhage (5, 6, 21). Immunoassays, phosphatase inhibition assays, and other analytical techniques (e.g., matrix-assisted laser desorption ionization-time of flight [MALDI-TOF] mass spectrometry) have been developed for microcystin assays and toxicity measurement (1, 12, 32).
Several attempts have been made to link toxin production to other molecular markers. Previous studies include, among others, studies of random amplified polymorphic DNA (RAPD) (33, 35), repetitive DNA elements (3, 39), and 16S rRNA genes (34, 37, 43, 44, 50). None of these analyses revealed any simple correlation with toxicity, although some recent studies have demonstrated that it may be feasible to generate genetic probes that are indicators of toxicity (4, 50). Although the relationship between toxicity and phylogeny within the genus Microcystis has not been resolved, phylogenetic analyses with different markers have suggested a monophyletic origin of Microcystis (50).
Despite intensive research, the biological function(s) of microcystins has not been determined yet. Putative roles for microcystin include feeding deterrence of zooplankton grazers, siderophoric scavenging of and binding to trace metals (such as iron), and involvement in quorum sensing (9, 10, 51). Notably, synthetase gene knockout by insertional mutagenesis revealed no apparent effects on laboratory cultures (11). This and other inconclusive investigations of microcystin function may suggest that the role of this family of secondary metabolites can be addressed most efficiently by studies of the various microcystin producers and nonproducers in their natural habitats combined with the use of clearly defined mutants or genetic variants. Another key issue that needs to be addressed from a functional perspective is the evolutionary history of microcystin synthetase and the mechanisms that cause both the distribution and variation observed for microcystin biosynthesis.
The recent cloning and sequencing of the complete (and almost complete) microcystin synthetase gene cluster (mcyABC and mcyDEFGHIJ) from two different strains, PCC 7806 (49) and K-139 (36), have provided a new tool for studying microcystin variation, evolution, and function. We performed a study of sequence divergence and organization of selected regions of the mcyABC genes of closely related Microcystis strains previously characterized at the ribosomal DNA (rDNA) level (44), with particular emphasis on mcyB. Different synthetase gene sequences were correlated with synthetase mRNA transcription, as well as with the structures of the cyclic peptides produced by the strains harboring a functional gene cluster. Furthermore, comparative phylogenetic analyses with the mcyABC region and other molecular markers were carried out.
The strains used are listed in Table Table1.1. Unialgal cultures were grown at the Norwegian Institute for Water Research (NIVA) as previously described (45); the only exceptions to this were strains PCC 7806 and HUB 5-2-4, which were kindly provided by H. Utkilen (National Institute of Public Health, Oslo, Norway).
DNA from Microcystis cultures were isolated by a method designed for plant DNA (27), with the following modifications. The DNA preparations were treated with RNase (50 μg per ml of supernatant) after chloroform-isoamyl alcohol (24:1) extraction, and the first DNA precipitation was performed with two-thirds volume of isopropanol instead of cetyltrimethylammonium bromide precipitation buffer. Later, this method was further modified by using 2× sodium dodecyl sulfate lysis buffer (100 mM Tris-HCl, 50 mM EDTA, 100 mM NaCl, 2% sodium dodecyl sulfate, 0.2% mercaptoethanol) instead of cetyltrimethylammonium bromide lysis buffer and treating the preparations with 10 μl of proteinase K (10 mg/ml) prior to phenol-chloroform (25:24) and chloroform-isoamyl alcohol (24:1) extraction.
PCRs were carried out in 25-μl mixtures containing 10 mM Tris-HCl (pH 8.8), 1.5 mM MgCl2, 50 mM KCl, 0.1% Triton X-100, each deoxynucleoside triphosphate (dNTP) at a concentration of 200 μM, 0.5 U of DyNAzyme II DNA polymerase (Finnzymes OY, Espoo, Finland), 5 pmol of each primer, and 1 ng of template. The primers used are listed in Table Table2.2. The PCR programs consisted of an initial denaturation step of 94°C for 4 min, followed by 30 cycles of 95°C for 30 s, the appropriate annealing temperature (Table (Table2)2) for 30 s, and 72°C for 1 min, ending with an additional extension step of 72°C for 7 min. PCR products were verified on 1.5% agarose gels stained with ethidium bromide.
Microcystis cultures (100 ml) from NIVA, grown as previously described, were centrifuged, and each pellet was suspended in 800 μl of TE buffer (10 mM Tris-HCl [pH 8], 1 mM EDTA [pH 8]). Total RNA was isolated from 100 μl of the cell suspension with addition of 1 ml of TRIzol reagent (Gibco BRL, Life Technologies, Rockville, Md.). All other steps were performed according to the manufacturer's recommendations. The RNA pellet was dissolved in Milli-Q water to a concentration of approximately 1 μg/μl, and 2.5 μl was treated with DNase I (Gibco BRL, Life Technologies) according to the manufacturer's protocol. First-strand cDNA synthesis was performed by using 500 ng of DNase-treated total RNA, SuperScript II (Gibco BRL, Life Technologies), the gene-specific primers 135-F and 676-R (Table (Table2),2), and RNase OUT recombinant RNase inhibitor (Gibco BRL, Life Technologies) according to the manufacturer's protocol. PCR amplifications were carried out by performing standard reactions (see above) and using 2.5 μl of the first-strand reaction mixture with primers 135-F and 676-R.
Control reactions were carried out by using the same reaction conditions and primers with DNase-treated RNA which had not been subjected to the reverse transcription step. No PCR products were detected by agarose electrophoresis (data not shown).
DNA amplified with primers 135-F and 676-R, primers 2156-F and 3111-R, and primers CC and CD (Table (Table2)2) was sequenced manually by the cyclic dideoxy chain termination method by using a Thermo Sequenase radiolabeled terminator cycle sequencing kit (U.S. Biochemical Corp., Cleveland, Ohio) according to the protocol supplied by the manufacturer.
The nucleotide sequences were aligned with the ClustalX multiple-alignment software (48), and then the alignments were manually edited with MacClade (29). A phylogenetic tree was constructed by the minimum-evolution method by using the Phylogenetic Analysis Using Parsimony package (47). The Kimura two-parameter model (23) was used to compute the distance matrix.
The amino acid sequences were aligned with the multisequence alignment algorithm PILEUP in the Wisconsin Package, version 10.0 for Unix (Genetics Computer Group, Madison, Wis.). A Dayhoff PAM matrix was computed with the Protdist program (Genetics Computer Group), and then a neighbor-joining tree was created. To infer confidence in the branch points in both trees, bootstrap analyses (17) were performed. The consensus trees were constructed from 500 bootstrap replicates.
Approximately 1 μg of genomic DNA was digested with 15 U of HindIII overnight (37°C). Restricted DNA was separated on a 1% agarose gel and then transferred to an Amersham Hybond-N membrane by Southern dry blotting overnight. Hybridization was performed at 68°C by using standard procedures (18). The probe (Fig. (Fig.2)2) was generated from the probe-F-probe-R PCR product from strain PCC 7806 by standard MSLP (magnetic solid-phase labeling) random primed synthesis (15) by using a random primed labeling kit (Boehringer GmbH, Mannheim, Germany).
Portions (100 ml) of unialgal Microcystis cultures were harvested by centrifugation and freeze-dried. Lyophilized samples were dissolved in 50% methanol and sonicated for 15 min. A mixture of 1 μl of sample and 1 μl of matrix (saturated α-cyano-4-hydroxycinnamic acid matrix solubilized in 50% acetonitrile-0.03% trifluoroacetic acid) was prepared directly on the template. The samples were analyzed by using a MALDI-TOF mass spectrometer (Voyager elite; PerSeptive BioSystems, Framingham, Mass.) with a nitrogen laser having a 337-nm output. The ions were accelerated with a voltage of 20 kV. Measurements were obtained in the delayed extraction mode, which allowed determination of monoisotopic mass values. A low mass gate of 500 improved the measurement by filtering out the most intense matrix ions. The mass spectrometer was used in the positive-ion detection and reflector mode. Post Source decay (PSD) measurements were obtained after peptide mass determination with the same samples on the template. The operating voltages of the reflectron were reduced stepwise to record 12 spectral segments sequentially (12, 16).
The following 7 primers (Operon Technologies Inc., Alameda, Calif.) of the 41 primers tested were used for RAPD analyses of the strains: A04 (5′-AATCGGGCTG-3′), A18 (5′-AGGTGACCGT-3′), C08 (5′-TGGACCGGTG-3′), C11 (5′-AAAGCTGCGG-3′), C20 (5′-ACTTCGCCAC-3′), D03 (5′-GTCGCCGTCA-3′), and D18 (5′-GAGAGCCAAC-3′). These primers were selected on the basis of band quality, band number, and PCR performance. PCRs were carried out in 25-μl mixtures containing 10 mM Tris-HCl (pH 9.0), 1.5 mM MgCl2, 50 mM KCl, 0.1% Triton X-100, each dNTP at a concentration of 100 μM, 0.75 U of Super Taq (HC) polymerase (HT Biotechnology Ltd., Cambridge, England), 6 pmol of primer, and 1 ng of template. The PCR program consisted of an initial denaturation step of 94°C for 3 min, followed by 35 cycles of 94°C for 15 s, 40°C for 30 s, and 72°C for 1 min and then an additional extension step of 72°C for 5 min. Fingerprint patterns were visualized on 1.4% agarose gels stained with ethidium bromide.
The REP (repetitive extragenic palindromic element) PCR was performed with primers REP1P-I (5′-IIIICGICGICATCIGGC-3′) and REP2-I (5′-ICGICTTATCIGGCCTAC-3′) (52). PCRs were carried out in 25-μl mixtures containing 50 mM Tris-HCl (pH 9.0), 15 mM (NH4)2SO4, 0.1% Triton X-100, 1.7 mM MgCl2, each dNTP at a concentration of 360 μM, 1 U of Dynazyme EXT DNA polymerase (Finnzymes OY), 10 pmol of each primer, and 1 ng of template. The PCR program consisted of an initial denaturation step of 95°C for 6 min, followed by 40 cycles of 94°C for 1 min, 40°C for 1 min, and 65°C for 8 min and then an additional extension step of 65°C for 16 min. PCR products were verified on 1.5% agarose gels stained with ethidium bromide.
All REP and RAPD PCRs were carried out in parallel with DNA isolated from two independently grown cultures of the same strain. Direct sequencing of the 16S rDNA variable region in the different cultures confirmed that all strains had undetectable levels of contaminating bacteria, as judged by the absence of double bands in the sequencing ladder. The gels were scored conservatively, meaning that only reproducible and strong bands were scored (1, present; 0, absent).
The results were analyzed together by using the Jaccard similarity coefficient, which excludes shared absence of bands. The results were then subjected to unweighted pair group method with arithmetic averages (UPGMA) clustering by using the program NTSYS-PC (38).
In this paper we present the results of molecular analyses of the mcyABC gene cluster in 18 closely related Microcystis strains. The strains were investigated at the DNA organization, sequence, and functional levels by using Southern hybridization, PCR, fingerprinting, DNA sequencing, mRNA RT-PCR, and microcystin structure analysis by MALDI-TOF mass spectrometry. The data generated were used to study the relationship between gene variants and the microcystin isoforms produced (i.e., whether the isoforms are a result of partial relaxation of substrate specificity or are produced by genetically distinct variants of the enzyme complex) and the evolutionary forces creating the distribution and variation of the mcyABC gene cluster.
Southern hybridization was performed by using a probe derived from the first condensation domain of mcyB (positions 46174 to 46754 in PCC 7806 [accession no. AF183408]) (Fig. (Fig.2)2) with strains collected from various locations. Altogether, 16 different NIVA-CYA (N-C) strains, including 10 strains originating from Norway, 2 strains from Sweden, 2 strains from Denmark, 1 strain from the United States, and 1 strain from Canada (Table (Table1),1), were used. In addition, the previously characterized strains HUB 5-2-4 (31) from Germany and PCC 7806 (49) from The Netherlands were investigated. As shown by the results of the Southern analysis in Fig. Fig.3,3, 11 of the 18 strains tested gave a positive hybridization signal, demonstrating the presence of the mcyB gene. The hybridizing N-C strains fell into two classes based on their HindIII restriction patterns; one class included strains N-C 31, N-C 118/2, and N-C 161/1 with a fragment length of 2,400 bp, which closely resembled the fragment length determined for strain PCC 7806 (2,390 bp), and the other class included strains N-C 57, N-C 143, N-C 169/7, N-C 228/1, N-C 264, and N-C 324/1 with a fragment length of 5,900 bp, which was similar to the fragment length determined for HUB 5-2-4 (5,840 bp). This polymorphism may indicate that there are differences in the adenylation domain in the first module of mcyB (mcyB1), as displayed by a PCC 7806-like variant and a HUB 5-2-4-like variant differing in at least one HindIII site.
To further investigate the sequence differences among the various strains, PCR primers were designed based on an alignment of the PCC 7806 and HUB 5-2-4 sequences. The forward primer 2156-F anneals to both sequence variants, and the two reverse primers, mcyB-R and mcyC-R, are specific for the PCC 7806-like sequence and the HUB 5-2-4-like sequence, respectively. The results confirmed the existence of the two classes of sequence variants among the previously uncharacterized strains (Table (Table3).3). The hybridization-positive strain N-C 31 did not produce a PCR product with either of the two reverse primers. However, the reason for this was clarified by sequencing (see below). Additional PCRs with primers 135-F and 676-R (Table (Table2)2) amplifying a region from the carboxy-terminal end of mcyA to the N-terminal end of mcyB (Fig. (Fig.2)2) showed that only the strains that produced a positive hybridization signal and an mcyB PCR product gave an mcyAB band (Fig. (Fig.4A).4A). This was confirmed by using primers 2156-F and 3111-R, which amplified both mcyB1 variants (Fig. (Fig.4B).4B). Thus, we concluded that these 11 hybridization-positive strains contain both the mcyA and mcyB genes.
In order to unambiguously confirm the identities of the strains, we sequenced a 560-bp highly variable region of 16S rDNA using the eubacterium-specific primers CC and CD (Table (Table2).2). The results, which showed very little variation among the N-C strains, are in full agreement with previous results obtained by Rudi et al. (44). In this region there are only six point mutations, including a single informative site that groups strains N-C 118/2, N-C 144, N-C 161/1, N-C 169/7, N-C 264, and N-C 324/1 together and strains N-C 31, N-C 43, N-C 57, N-C 122/2, N-C 123/1, N-C 143, N-C 166, N-C 172/5, N-C 228/1, N-C 279, and PCC 7806 together.
Sequencing of the region from 135 to 676 (Fig. (Fig.2)2) of mcyAB also revealed few differences between the strains. Only 11 point mutations were found in the 500-bp region. Notably, strain N-C 118/2 had more point mutations (7 of the 11 mutations) than the other strains, including a change in the mcyB start codon from AUG to GUG. Two informative sites were found in this region; one site grouped N-C 57, N-C 169/7, N-C 264, and N-C 324/1 together, and the other informative site grouped N-C 57, N-C 118/2, N-C 161/1, N-C 264, N-C 324/1, and PCC 7806 together.
Sequencing of the adenylation domain in the first module of mcyB (mcyB1) was accomplished by using PCR products generated with primers 2156-F and 3111-R (Table (Table2).2). This primer pair amplifies both the PCC 7806-like and the HUB 5-2-4-like sequence variants, and positive PCR signals were obtained for the 11 strains shown to contain the mcyB gene. In contrast to the mcyAB region, the mcyB1 module showed considerable variation among the strains. Both nucleotide and amino acid alignments from the region from 2156 to 3111 (Fig. (Fig.2)2) indeed confirmed that the strains are divided into two main groups (the same PCC 7806-like and HUB 5-2-4-like groups described above) (Fig. (Fig.55).
The nucleotide and amino acid sequence alignments were used to construct phylogenetic trees. The mcyC modules in PCC 7806 (PCC 7806c) and K-139 (K-139c) were included for comparison, and the corresponding sequence in grsA (25) was used as an outgroup. Both the amino acid and the nucleotide trees clustered the strains in two main groups (Fig. 6A and B), in agreement with the Southern analysis. The PCC 7806-like sequences comprised a B-type group [with greatest affinity to the mcyB1 module of PCC 7806; designated mcyB1(B)], while the other group displayed a C type of mcyB1 module [i.e., the greatest affinity was to the mcyC region; designated mcyB1(C)]. These two main clades were also formed when phylogenetic analyses were performed with conserved regions A3 to A8 alone (data not shown). The PCC 7806-like group, comprised of strains N-C 118/2, K-139, N-C 161/1, and PCC 7806, clustered together with 100% bootstrap support. The N-C 31 strain created a side branch with a strong affinity to the PCC 7806-like group (Fig. (Fig.6).6). The amino acid sequence alignment showed that this strain has sequence affiliation with the HUB 5-2-4-like group after core motif A8 (Fig. (Fig.55).
The HUB 5-2-4-like group (the C type of mcyB1) could be further divided into subgroups. Although the exact topology of the subgroups could not be unambiguously determined due to differences in topology between the amino acid and nucleotide trees and low bootstrap support for some branches, the analysis showed that there were two strongly supported subgroups, one that included the mcyC modules from PCC 7806 and K-139 in addition to the mcyB1 module of N-C 264 and one that included N-C 169/7 and N-C 324/1. Notably, strains N-C 169/7 and N-C 324/1 showed different affinities in the nucleotide and amino acid trees. These two strains showed affinities to HUB 5-2-4, N-C 228/1, N-C 143, and N-C 57, as well as to the true C-type subgroup (N-C 264, K-139c, and PCC 7806c) (Fig. 6A and B).
Taken together, our results revealed the existence of chimeric mcyB1 modules that show the greatest sequence similarity to the downstream mcyC module. It seems likely that these variants arose through recombination of the repeated sequences (A1 to A10) in the different modules of mcyB and mcyC. Analyses based on the program LARD (version 2.2; Likelihood Analysis of Recombination in DNA) (19) indicated that there is a recombination breakpoint at the end of motif A3 in mcyB1 (data not shown).
The next question which we addressed was the transcriptional functionality of the mcyABC gene cluster in the different strains in terms of mRNA synthesis. An RT-PCR analysis of total RNA from the 16 N-C strains was performed with primers 135-F and 676-R (Table (Table2).2). Of the nine strains shown to contain the mcyA and mcyB genes, eight (strains N-C 31, N-C 57, N-C 118/2, N-C 143, N-C 161/1, N-C 169/7, N-C 228/1, and N-C 264) gave clear positive RT-PCR results (Fig. (Fig.7).7). These results show that the different strains containing the B or C type of the mcyB1 module are capable of producing microcystin synthetase mRNA. None of the strains lacking the mcyABC gene cluster produced detectable levels of mRNA.
The final aspect studied was determination of the microcystin isoforms synthesized by the various strains. All 16 N-C strains were analyzed by MALDI-TOF mass spectrometry, and the results are shown in Table Table3.3. Only the strains found to contain the mcyABC cluster produced detectable levels of microcystin. The B-type variants of mcyB1 produced various microcystin-LR isoforms, and one strain (N-C 161/1) produced microcystin-YR in addition to microcystin-LR. Strain N-C 31, which formed a basal branch of the mcyB1(B) strains (Fig. 6A and B), produced only microcystin-LR. N-C 118/2 and PCC 7806 both produced large amounts of [Asp3]microcystin-LR and microcystin-LR. All strains with a C-type variant of the mcyB1 module synthesized microcystin-RR, and the members of some subgroups produced microcystin-LR in addition to microcystin-RR. Strain N-C 264, which possesses the true C type (i.e., its mcyB1 module groups together with the mcyC modules of PCC 7806 and K-139), produced only a microcystin-RR isoform. This was also the case for N-C 57. The mcyB1(C) strains (N-C 169/7, N-C 228/1, and N-C 324/1) belonging to other C-type subgroups synthesized both microcystin-RR and microcystin-LR isoforms. Strain N-C 324/1, which did not give a detectable RT-PCR signal, produced small amounts of microcystin-LR and demethylated isoforms of microcystin-RR and microcystin-LR. HUB 5-2-4, belonging to the C-type subgroup, produced both microcystin-RR and microcystin-LR, while N-C 228/1 produced demethylated isoforms of these two microcystins. Our data show that there is variation in the methylation of the microcystin isoforms in the different strains, but since the enzymatic activity of methylation lies outside the mcyB1 module, the genetic basis for this variation could not be addressed in this study. Notably, strain N-C 143 did not produce detectable levels of microcystin, even though it contained the mcyA and mcyB genes and synthesized the mRNA.
The REP and RAPD fingerprinting matrices, calculated by using REP and seven different RAPD primers, were analyzed together by using UPGMA. This analysis showed that the strains formed two clusters (Fig. (Fig.6C).6C). One of the clusters contained all of the strains characterized as M. aeruginosa. This is in agreement with 16S rDNA results (44). However, the clustering shown by fingerprinting markers did not resemble the major groups revealed by the mcyABC region (compare Fig. 6A and C).
The different analyses performed in this study showed that Microcystis strains can have different adenylation domains in the mcyB1 module of the microcystin synthetase genes. Our initial studies (Southern blotting and PCR) showed that the strains could be divided into two groups, the PCC 7806-like variants (B type) and the HUB 5-2-4-like variants (C type). With this information in hand, we performed a BLAST search using the partially sequenced module of HUB 5-2-4 deposited in the GenBank database (accession no. Z28338), which showed that the ends of the HUB 5-2-4 sequence are most similar to the first module in mcyB (96 and 95% identity to strain PCC 7806) and that the middle (A3 to A9) is most similar to the module in mcyC (91% identity to PCC 7806). In other words, the well-characterized strain HUB 5-2-4 displays unmistakable features of genetic recombination. Our detailed studies at the sequence level, including phylogenetic analyses, supported this interpretation and suggested that there are several phylogenetic subgroups of the adenylation domain in mcyB1. We designated these two main variants mcyB1(B) (the PCC 7806-like variants) and mcyB1(C) (the HUB 5-2-4-like variants). In contrast, the upstream mcyAB region displayed extensive sequence homology in all the strains, in agreement with the variable region of the 16S rRNA gene (44).
The variation in the mcyB1 module does not seem to influence the functionality of the genes as both variants produce mRNA (Fig. (Fig.7)7) and microcystins (Table (Table3).3). The investigation of the microcystin synthetase transcripts revealed potential regulation at the translational (or posttranslational) level, as exemplified by one of the strains which synthesized the mRNA but no detectable microcystin. However, in the absence of real quantitative measures of both transcript and microcystin levels, transcriptional regulation was not addressed in this study. The few differences observed in the mcyAB region do not influence functionality either, as strain N-C 118/2, which contains the highest relative frequency of point mutations, including a change in the start codon from AUG to GUG, synthesizes both mRNA and normal amounts of microcystin. It should be noted that the unaffected transcription could be explained by recent findings indicating that mcyA and mcyD have two transcriptional start points (22).
The different microcystin isoforms synthesized by each strain were determined by MALDI-TOF mass spectrometry, and a good correlation was found between the genetic variants and the microcystin isoforms produced by the strains. Most importantly, the B-type variants produce various microcystin-LR isoforms (one strain produces microcystin-YR in addition to microcystin-LR), and the C-type variants synthesize either microcystin-RR (the true C type) or microcystin-RR in combination with microcystin-LR (the other subgroups of the C type). The observed substrate specificity follows the predicted amino acid sequence variation that influences the substrate-binding pocket of the module (Table (Table4),4), which consists of 10 amino acids. From the deduced amino acid sequences we could identify nine of these amino acids in the different variants (Table (Table4).4). The substrate-binding pocket of the mcyB1(C) variant is very similar to the binding pocket of the mcyC gene (Table (Table4),4), containing arginine (R) in the strains used in this study, thus explaining the presence of various microcystin-RR isoforms. However, some of these strains also produce microcystin microcystin-LR; this may be a position-dependent property or due to residue Gly236 (Table (Table4).4). This result is particularly interesting in light of the finding that the C-type subgroups possess a genetic recombinant mcyB1 module (a chimeric mcyB1/C module). The mcyB1(B) variants, which usually activate leucine (Table (Table3),3), show a high degree of similarity to other leucine-activating domains (7, 8, 46) (Table (Table44).
Taken together, our results demonstrate that there is a strong correlation between the microcystin isoforms produced and the genetic variants of the mcyB1 module; the genetic variants are the result of recombinations between modules (i.e., between mcyB1 and mcyC). The presence of multiple genetic variants of the mcyBC region is in agreement with the findings of Kurmayer et al. (26). However, Kurmayer et al. (26) did not perform an extensive sequence and phylogenetic analysis, and for that reason the distribution of specific microcystin isoforms could not be linked to the diversity of genotypes. According to our results, identification of isoform structure by analyses such as MALDI-TOF mass spectrometry should enable deduction of the type of genetic organization of the microcystin synthetase genes.
Of the 18 strains used in this study, 7 were shown not to contain the mcyA and mcyB genes (Fig. (Fig.33 and and4).4). Our results, together with data from a recent study by Tillett et al. (50) showing conservation of the chromosomal location of the mcyABC gene cluster in all Microcystis strains, as well as the same gene arrangement in the synthetase-negative strains, suggest that the mcyA and mcyB genes (or possibly the whole operon) have been lost in some strains. Thus, microcystin synthesis is probably an ancestral feature of the genus Microcystis.
Phylogenetic studies of prokaryotic adenylation domains have shown that they normally cluster according to function, making it unlikely that they arose by convergent evolution (7). Obviously, the selection forces acting upon these genetic modules have been significant during evolution. The amino acid sequences in the two main variants of mcyB1 are highly divergent from motif A3 (amino acid 20 in Fig. Fig.5)5) onward. The ratio of nonsynonymous to synonymous mutations is elevated in this region compared to amino acids 1 to 20 (Fig. (Fig.5)5) and to the mcyAB (135-676) region. This could be seen as evidence of diversifying selection, which would be the case if the mcyB1(C) variants originated from recombination between mcyB1 and mcyC modules. Accordingly, selection is not sufficient to explain the distribution of the mcyABC region, as well as the phylogenetic relationships between the module variants as opposed to other markers, such as 16S rDNA and DNA fingerprinting (RAPD, REP). Thus, the peptide synthetases appear to be the result of an evolutionary mechanism involving horizontal transfer, since structurally related peptides occur in diverse microorganisms (24). Notably, horizontal transfer processes have recently been suggested to be important evolutionary mechanisms involved in genomic stability and variation in several closely related cyanobacterial strains (44), as manifested in the rbcLX (43) and trnl(UAA) genes (40-42). The evolutionary processes leading to the distribution and genetic variation of the microcystin synthetase gene cluster, including the two major genetic variants of the mcyB1 module, are likely to be caused by horizontal gene transfer, lateral recombination, and gene loss, particularly as suggested by the chimeric mcyB1 modules and the nonconcordance between the microcystin phylogenies and the fingerprinting results. In this context it is interesting that uma4, a gene downstream of mcyC, encodes a peptide with 45% identity to TnpA, a transposase from Anabaena sp. strain PCC 7120 (49).
The few differences in the 16S rDNA or mcyAB regions did not exhibit any correspondence to the variants found in the mcyB gene or to properties of toxicity or nontoxicity. This is in agreement with other studies that were also unable to identify differences in the 16S rDNA genes that were correlated with toxic or nontoxic strains (34, 37, 50).
The biological function or functions of microcystin are currently unknown, although there are many theories, ranging from iron binding (51) to involvement in quorum sensing (10). The diversity of microcystin isoforms is unlikely to have a restraining effect on the function of microcystin in the cyanobacterial cell. Rather, it seems more likely that the diversity of isoforms is crucial to the various strains in their natural habitats. Furthermore, the genetic processes generating the genetic diversity, particularly in the substrate-binding pockets, are also expected to be of crucial biological importance. Based on these results, it should now be possible to design genetic constructs of the mcyB1 variants and introduce these into identical genetic backgrounds (strains) in order to test for phenotypic effects in complex laboratory-designed ecosystems.
We thank Randi Skulberg for providing the N-C strains, Kamran Shalchian-Tabrizi for help with the phylogenetic analyses, and Tonje Fossheim for modifications of the DNA isolation protocol. Thanks are also due to Hans Utkilen and Arne Mikalsen for providing the PCC 7806 and HUB 5-2-4 strains and for fruitful discussions.
This work was supported by grant 107622/420 from the Norwegian Research Council to K.S.J.