Diversity of microcystin genotypes among populations of the filamentous cyanobacteria Planktothrix rubescens and Planktothrix agardhii

Abstract

Microcystins (MCs) are toxic heptapeptides that are produced by filamentous cyanobacteria Planktothrix rubescens and P. agardhii via non-ribosomal peptide synthesis. MCs share a common structure cyclo (-D-Alanine¹-L-X²- D-erythro-ß-iso-aspartic acid³-L-Z⁴-Adda⁵-D-Glutamate⁶- N-methyl-dehydroalanine⁷) where X² and Z² are variable L-amino acids in positions 2, 4 of the molecule. Part of the mcyB gene (1,451 bp) that is involved in the activation of the X² amino acid during MC synthesis was sequenced in 49 strains containing different proportions of arginine, homotyrosine, and leucine in position 2 of the MC molecule. Twenty-five genotypes were found that consisted of eight genotype groups (A-H, comprising 2-11 strains) and 17 unique genotypes. P. rubescens and P. agardhii partly consisted of the same mcyB genotypes. The occurrence of numerous putative recombination events that affected all of the genotypes can explain the conflict between taxonomy and mcyB genotype distribution. Genotypes B (homotyrosine and leucine in X²) and C (arginine in X²) showed higher nonsynonymous/synonymous (d_N/d_S) substitution ratios implying a relaxation of selective constraints. In contrast other genotypes (arginine, leucine, homotyrosine) showed lowest d_N/d_S ratios implying purifying selection. Restriction fragment length polymorphism (RFLP) revealed the unambiguous identification of mcyB genotypes, which are indicative of variable X² amino acids in eight populations of P. rubescens in the Alps (Austria, Germany, Switzerland). The populations were found to differ significantly in the proportion of specific genotypes and the number of genotypes that occurred over several years. It is concluded that spatial isolation might favour the genetic divergence of microcystin synthesis in Planktothrix spp.

Introduction

Toxic heptapeptides microcystins are produced by various genera of the phylum of cyanobacteria, i.e. the genera Microcystis, Anabaena, and Planktothrix. The filamentous cyanobacteria Planktothrix spp. are one of the most important microcystin producers and can be found in freshwater habitats in the temperate region of the Northern Hemisphere (Fastner et al. 1999). P. agardhii co-occurs with Limnothrix redekei and other filamentous cyanobacteria, e.g. Pseudoanabaena limnetica in high abundance in shallow and eutrophic lakes (Rücker et al. 1997). In contrast P. rubescens forms metalimnetic layers in deep stratified and less eutrophic lakes, and often monopolizes resources and dominates phytoplankton completely (Anneville et al. 2004).

Microcystins (MCs) are cyclic heptapeptides that are produced by cyanobacteria and share a common structure cyclo (-D-Ala¹ -L-X² -D-MeAsp³ -L-Z⁴ -Adda⁵ -D-Glu⁶ -Mdha⁷), where X and Z are variable L-amino acids (e.g. LR refers to leucine and arginine in the variable positions), D-MeAsp is D-erythro-ß-iso-aspartic acid, Adda is (2 S, 3 S, 8 S, 9 S -3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid), and Mdha is N methyl-dehydroalanine (Carmichael et al. 1988). MCs are synthesized by thiotemplate mechanism such as other non-ribosomal peptides produced by bacteria and fungi (Marahiel et al. 1997). The large enzyme complex encoded by the mcy gene cluster is composed of peptide synthetases, polyketide synthases, and tailoring enzymes (Tillett et al. 2000, Christiansen et al. 2003). It has a modular structure, each module of which contains specific functional domains for activation = aminoacyl adenylation (A) domains, and thioesterification (thiolation domains) of the amino acid substrate and for the elongation (condensation domains) of the growing peptide (Tillett et al. 2000). Adenylation domains show a high degree of conservation (core motifs) enabling the definition of general rules for the structural basis of substrate recognition of non-ribosomal peptide synthetases (Marahiel et al. 1997). The determination of substrate specificity is based on the crystallization of the adenylation domain of gramicidin synthetase (GrsA) (Conti et al. 1997). This enabled the identification of core motifs constituting a putative binding pocket for the activation of the amino acid substrate during peptide synthesis and the in silico identification of critical binding pocket residues that are located between core motifs A3-A6 (Challis et al. 2000, Stachelhaus et al. 1999).

McyBA1 is responsible for the activation of amino acids as aminoacyl adenylate followed by peptide bond formation to the growing microcystin molecule by the condensation domain located upstream of mcyBA1 (Tillett et al. 2000). In Planktothrix spp. three different amino acids have been described in position 2 of the microcystin molecule: leucine, arginine, homotyrosine (Luukkainen et al. 1993, Henriksen & Moestrup 1997, Fastner et al. 1999, Kurmayer et al. 2005). In an earlier study, we described the occurrence of novel microcystin variants, which appeared to be unique and dominant for Planktothrix and that exclusively occurred in a specific lake, Schwarzensee in the Austrian Alps (Kurmayer et al. 2004). It has been speculated that the populations of P. rubescens may be geographically isolated, for example Schwarzensee, the lake of origin of the isolates containing those novel microcystin variants is at an altitude of 716 m above sea level, while many other lakes have a lower altitude, such as Mondsee, which is located only 11.4 km away.

Typically the distribution of microorganisms is not restricted by geographical barriers (Finlay 2002). Studies have shown that recombination and migration rates in natural populations of soil bacteria (Bacillus spp.) and aquatic cyanobacteria (Microcoleus sp., Nodularia sp.) were high, and were not influenced by geographic factors (Roberts & Cohan 1995, Barker et al. 2000, Lodders et al. 2005). On the other hand, the occurrence of physical isolation, which is defined as the spatial separation of two or more populations among geographically isolated habitats (Papke & Ward 2004) cannot a priori be excluded. Geographic isolation has been observed among isolated populations of extreme habitats, for example, Synechococcus inhabiting hot springs (Papke et al. 2003) and hyperthermophilic Archaea (Whitaker et al. 2003), but also for free living Pseudomonas spp. in pristine soil samples (Cho & Tiedje 2000). The question of isolation is of fundamental importance to understand the evolution of microcystin synthesis, i.e. populations may diverge through neutral processes (i.e. random drift, Kimura 1983), local adaptation processes, or both.

The ability to detect and interpret genotypic variation depends on the type of the gene under investigation and the resolving power of the genetic markers. Genes showing high variability with potential functional consequences are considered highly suitable to interpret genotypic variation in potentially isolated populations (Palys et al. 1997). The species P. rubescens (sensu Suda et al. 2002) only consists of microcystin genotypes while - similar to Microcystis sp. – populations assigned to P. agardhii consisted of microcystin genotypes and genotypes without the mcy genes (Kurmayer et al. 2004). It has been shown that Microcystis sp. colonies that were isolated from field samples of a eutrophic lake show a high genetic variability within mcyBA1 (Kurmayer et al. 2002), for example, due to domain duplication and recombination within the mcyB gene (Mikalsen et al. 2003).

It was the aim of the present study (i) to analyse the genetic variation within strains, (ii) identify genotypes that are indicative of the production of specific MC variants, (iii) and to directly detect and quantify those genotypes in various lakes in the Alps. This knowledge is necessary to understand MC net production in phytoplankton as well as the microevolution of microcystin synthesis in cyanobacteria. For this purpose, 49 Planktothrix strains isolated in Europe were sequenced for the first adenylation domain (A1) of the mcyB gene. In order to characterise populations directly (and to overcome a possible isolation bias), restriction fragment length polymorphism (RFLP) profiles were assigned to each mcyBA1 genotype and subsequently used to distinguish the gene pools of spatially separated populations.

Materials and Methods

Cultivation and sampling of cyanobacteria

The 49 strains that were used in this study were either isolated from several European freshwater habitats, as described in Kurmayer et al. (2004), or acquired from international culture collections (Table 1). The isolates were assigned to the genus Planktothrix (P. agardhii and P. rubescens, Suda et al. 2002, min. 99.2% similarity in 16S rDNA) according to the morphological criteria provided by Komarek (2003). In addition all of the strains were genetically tested using Planktothrix specific primers binding to the intergenic spacer region within the phycocyanin operon (PC-IGS) as described in Kurmayer et al. (2004). Clonal strains were grown in BG-11 medium (Rippka 1988) modified to contain 2 mM NaNO₃ + 10 mM NaHCO₃ and kept in batch culture at 15°C under continuous light conditions (5–10 μmol m⁻² s⁻¹, Osram Type L30W/77 Fluora).

Table 1.

Variation of amino acids in position 2 of microcystin molecules (Arg, arginine, Hty, homotyrosine, Leu, leucine) produced among 49 strains of P. rubescens (rub) and P. agardhii (aga). Strains are ordered according to the proportion of each of the structural variants produced (mean ± 1SE). The number of MC measurements using HPLC analysis is indicated for each strain. Strains marked by asterisks have been shown to be inactive in MC production (Christiansen et al. 2006). The corresponding AluI (I-X) and TseI (I-IV) restriction types, and the sequence accession numbers of mcyBA1 are indicated. Supercripts A-H identify identical genotypes. Strains have been isolated during the study by Kurmayer et al. (2004). Country codes (ISO format). SAG Culture Collection of Algae (Göttingen, Germany), PCC Pasteur Culture Collection (Institute of Pasteur, Paris, France), CCAP Culture Collection of Algae and Protozoa (Windermere, UK).

¹New microcystins with Hty/Tyr in position 2 of the microcystin molecule (R. Kurmayer, W. Yoshida, K. Ishida, T. Hemscheidt, unpublished)

²Accession numbers AJ749267-AJ749284, AJ863131-AJ863134 have been published by Kurmayer et al. (2005).

P. rubescens was sampled by pulling a plankton net (30 μm in mesh size) from a depth of 20 m to the surface at the deepest part of eight lakes in the Alps (Table 2). Aliquots (a few ml from net samples) were filtered onto glass fibre filters (GF/C, Whatman, Kent, Great Britain) under vacuum pressure and stored frozen (−20°C) until DNA extraction. In this study each lake was sampled a total of three times from 2001-2005.

Table 2.

Morphometric and limnological characteristics of lakes sampled for Planktothrix rubescens. Trophic data (M, mesotropic, O, oligotrophic) are yearly minimum-mean-maximum value (sample size) of the year 2001 from the Kärntner Institut für Seenforschung, Carinthia (Afritzersee, Wörthersee), Bayrisches Landesamt, Wielenbach, Germany (Ammersee), Department for Environment, Kanton Aargau (Hallwilersee), Federal Agency for Water Management, Scharfling, Upper Austria (Irrsee, Schwarzensee) and Zurich Water Supply (Zürichsee). Chlorophyll a values were integrated over the watercolumn. As inferred from microscopical inspection of net samples (30 μm mesh size) the abundance of P. rubescens is given: D, dominant; A, abundant, R, rare. M a SL, meter above sea level.

Lake	N(°)	E(°)	M a SL	Area (km2)	Tropic status	Secchi depth (m)	Chlorophyll a (μg L−1)	Abundance Planktothrix
Afritzer See, AT	46°44	13°46	750	0.5	M	3.3-4.0-5.2 (4)	2.4-4.4-6.8 (4)	A
Ammersee, DE	47°16	11°4	533	47	M	1.6-4.4-6.5 (11)	1.5-5.1-9 (11)	A
Hallwilersee, CH	47°20	8°10	449	10	M	1-2.7-5 (12)	8.8-18.9-30 (12)	D
Irrsee, AT	47°56	13°19	553	3.6	M	2.2-5.1-8.2 (12)	1.4-2.8-4.3 (7)	R
Mondsee, AT	47°48	13°22	481	14.2	M	2.4-3.8-5 (24)	1.8-4.3-9.6 (24)	A
Schwarzensee, AT	47°45	13°30	716	0.5	O	5.5-6.5-7.5 (3)	0.5-0.9-1.5 (3)	R
Wörthersee, AT	46°36	14°3	440	19.4	M	3.5-4.9-7.5 (7)	5.4-9-13.3 (7)	D
Zürichsee, CH	47°22	8°32	406	65	M	2.3-3.9-5.9(18)	5-9.5-15.7 (12)	D

Microcystin analysis

All strains were analysed for their composition of microcystin variants, i.e. amino acid composition in position 2 of the MC molecule by HPLC-DAD (high performance liquid chromatography with diode array detection) as described (Kurmayer et al. 2004, 2005). Briefly, the cells were filtered on glass fibre filters (GF/C, Comesa, Vienna, Austria) and MCs were extracted using 75% (w/v) aqueous methanol. MC variants were quantified at 240 nm by their characteristic absorption spectra (original spectrum and first order derivative) and retention times (Fastner et al. 1999) using a linear gradient of acetonitrile (0.05% TFA, Trifluoroacetic acid) against water (0.05% TFA) according to Lawton et al. (1994).

Demethylated variants containing arginine in position 2 of the molecule, i.e. [D-Asp³, Mdha⁷]-RR and [D-Asp³, Dhb⁷]-RR eluted from 13.4-13.9 min and 14.3-14.8 min, each, those containing homotyrosine (Hty) in position 2, [Asp³]-MC-HtyR eluted from 18.3-18.5 min, and those containing leucine in position 2, [Asp³]-MC-LR eluted 19.3-19.6 min (Kurmayer et al. 2005). All strains from Lake Schwarzensee were found to produce new microcystin variants eluting from 23.5-24.0 min (Kurmayer et al. 2004). Amino acid analysis, one-dimensional, and two-dimensional NMR revealed that the two novel variants contained homotyrosine/tyrosine in position 2 of the molecule: [D-Asp³, (E)-Dhb⁷]-MC-HtyY and [D-Asp³, (E)-Dhb⁷]-MC-HtyHty (R. Kurmayer, W. Yoshida, K. Ishida, T. Hemscheidt, unpublished).

Genetic analysis

Two millilitres of each strain culture were incubated 1h on ice and centrifuged at 16,000 g for 10 min. The pellet was lyophilised in a vacuum centrifuge at 30°C. DNA extraction from the strains or field samples was performed as described in Kurmayer et al. (2003).

PCR amplifications were performed in a volume of 20 μl, containing 2 μl of Qiagen PCR buffer (Qiagen, VWR, Vienna, Austria), 1.2 μl MgCl₂ (25 mM, Qiagen), 0.6 μl deoxynucleotide triphosphates (10 μM each, MBI Fermentas, St Leon-Rot, Germany), 1 μl of each primer (10 pmol μl⁻¹), 0.1 μl Taq DNA polymerase (5 u μl⁻¹, Qiagen), 13.1 μl sterile Millipore water and 1.0 μl of the diluted DNA extract (1:100). The PCR thermal cycling protocol included an initial denaturation at 94°C for 3 min, followed by 35 cycles at 94°C for 30 s, with an annealing temperature of 52°C for 30 s, and elongation at 72°C for 2 min, followed by 72°C for 5 min.

Oligonucleotide primers for mcyBA1 amplification and sequencing are published in Kurmayer et al. (2005). McyBA1totfwd and McyBA1totrev were used to amplify the total fragment of the first adenylation domain (A1) of mcyB in strains and field samples (1,693 bp, including the core motifs A1-A10; Marahiel et al. 1997). The primers were specific for Planktothrix. Amplification products for mcyB were purified using the Qiagen QIAquick PCR Purification Kit (Qiagen, VWR, Vienna, Austria) and sequenced directly by standard automated fluorescence techniques (Applied Biosystems, Weiterstadt, Germany). These sequence data have been submitted to the DDBJ/EMBL/GenBank databases under the accession numbers: AJ890255-AJ890282 (Table 1).

In addition mcyBA1 PCR products were digested using AluI and TseI. The restriction profiles were compared with the mcyBA1 sequences. Restriction analysis was performed directly from the PCR amplification product using AluI (MBI Fermentas, St Leon-Rot, Germany) and TseI (New England Biolabs, Frankfurt am Main, Germany) according to the manufacturer's instructions. Digestions were performed in a volume of 20 μl, containing 0.2 μl of the restriction enzyme (5 u μl⁻¹) and 7 μl of the PCR products and incubated at 37°C (AluI) and 65°C (TseI) for 3h. The PCR products for mcyB were visualized in 1% agarose in 0.5 × TBE (Tris-borate-EDTA buffer) and ethidium-bromide staining performed via agarose gel electrophoresis according to standard procedures (Sambrook et al. 1989). For the electrophoresis of restriction fragments, 2% agarose was used.

PCR products for mcyBA1 obtained from field samples were cloned using the pDrive Cloning Vector system (Qiagen, VWR, Vienna, Austria) according to the manufacturer's instructions. The transformation efficiency was 2.1 × 10⁴ – 7.0 × 10⁴ colonies per μg of DNA. After the transformed colonies were grown on an agar plate overnight, white colonies were randomly picked using the tip of a pipette, and re-dissolved in 10 μl of Millipore water each. Re-PCR amplification from the white clones was always successful and the subsequent restriction analysis of colony specific PCR products revealed the unambiguous identification of a specific restriction type. The sequencing of mcyBA1 primer binding sites performed for all of the different AluI restriction types observed among strains (Table 3) revealed high similarity, i.e. identical nucleotides for the binding site of the mcyBA1totrev primer (5′-AGA CTT GTT TAA TAG CAA AGG C–3′). One substitution of adenine by cytosine in position 6 of the mcyBA1totfwd primer for restriction types I, IV (5′-CAC CTA GTT GAA GAA CAA GTT CT-3′) was found. Consequently the PCR amplification of the restriction types from the field samples was considered unbiased.

Table 3.

Schematic representation of AluI and TseI restriction profiles within mcyBA1 (1,693 bp) of Planktothrix spp. The size of the restriction fragments in bp was calculated from the sequences.

Phylogeny and statistical analysis

Sequences (1,451 bp) were aligned using multiple sequence alignment (Clustal W 1.8) and similarities between nucleotide sequences were calculated using DNADIST in the PHYLIP package (Version 3.6 alpha; Felsenstein 1989). Maximum likelihood analysis was used to estimate nucleotide substitution parameters under a general time-reversible nucleotide substitution model by estimating the gamma distribution for variable rates among the sites. Ambiguous sites at which at least one sequence showed an undetermined nucleotide were removed (two sites) and the discrete gamma algorithm was used to approximate a continuous gamma distribution using five categories of rates (ncatG = 5) in the programme BASEML of the PAML package (Version 3.14, Yang 1997). The best fit of the model showed (log-) likelihood lnL = −2871.96 with parameter estimates for the transition/transversion ratio κ = 2.07 and the shape parameter α = 0.006. Statistical significance of the branches was estimated by bootstrap analysis generating 100 replicates of the original data set using the PHYLIP package. Finally consensus trees following the 50% majority rule were computed.

The ratio of nonsynonymous (d_N) and synonymous (d_S) substitution rates per site was determined using the likelihood approach, which was implemented in the program CODEML of the PAML package (Yang 1998). The “one ratio” model, assuming the same ratio for all of the branches in the phylogeny and the “free ratio” model, assuming as many ratios as the number of branches in the phylogenetic tree were used. Branch-site models (Zhang et al. 2005) were employed to test for positive selection acting on specific branches in the phylogenetic tree showing increased d_N/d_S ratios. Branches of the tree were divided a priori into foreground and background lineages, and a likelihood ratio test was constructed by comparing a model that allows positive selection on the foreground lineages (the alternative model) with a model that does not allow such positive selection (the null model). The improved branch-site likelihood method for detecting positive selection implemented in the PAML package was used (Version 3.15, Anisimova et al. 2003, Zhang et al. 2005).

Following Tanabe et al. (2004) the runs test implemented in the programme GENECONV (v1.81, Sawyer 1999) was used to investigate whether substitutions were significantly clustered, and whether gene conversion (recombination) events occurred within mcyBA1. The settings used were the default (/g0 = mismatches within fragments were not allowed). The Global P-value calculated from 10,000 random permutations of the alignment was used to assess the significance of any unusually long fragments that were sufficiently similar to be suggestive of past gene conversion. Recombination events were independently quantified using a recombination detection programme (RDP) developed by Martin & Rybicki (2000) using the default settings (window size 10, highest acceptable p-value 0.05).

The frequency of the occurrence of genotypes and restriction types between the Planktothrix species was compared using the Yates-corrected chi-square statistic in the analyses of 2 × 2 contingency tables (Zöfel 1992). The occurrence of restriction types between lakes was statistically compared using chi-square statistics for 8 × 2 contingency tables. Because the test can produce false results when the expected frequencies are small, proportions with an expected frequency <1 were not tested (Jongman et al. 1995). The differences in the total number of restriction types found in a population were tested using one-way ANOVA followed by the Tukey test for pairwise comparison (overall significance p=0.01).

Results

Genetic variation within the microcystin mcyB gene

Among the mcyBA1 sequences obtained from the 49 Planktothrix strains, 25 different genotypes were detected, which differed in at least one base pair. We found eight mcyBA1 genotypes (A-H) consisting of 2-11 strains each and 17 unique genotypes (Table 1). The genetic variability was between zero and 3.6%. Genotype B (11 strains) consisted of both species P. agardhii and P. rubescens while A (8), H (2), G (2), D (2), F (3) consisted of P. rubescens only. In contrast C (2) and E (2) occurred significantly in P. agardhii only (Chi-square test, χ² (df=1), p < 0.05).

The ratio of nonsynonymous (d_N) and synonymous (d_S) substitution rates for the entire phylogenetic tree was d_N/d_S = 0.19. Genotypes B, C, and the clade consisting of genotypes A (D, E, H, strains No34, 64, 91/1) showed a relatively high frequency of nonsynonymous substitutions resulting in d_N/d_S = 0.73, 0.87, 0.23, respectively. The branch-site model was used to detect a possible positive selection acting on genotypes B, C and the clade consisting of genotypes A, D, E, H, No34, 64, 91/1. Although the log-likelihood value of the alternative model was found improved when compared with the log-likelihood value of the null model the difference between the two models was not found to be significant (2Δl < 3.84, df = 1, p > 0.05).

The runs test implemented in GENECONV revealed 146 possible recombination events, i.e. gene fragments that were sufficiently similar to be suggestive of past gene conversion within mcyBA1 (123 – 1,379 bp, 12-68 polymorphic sites). All of the genotypes showed at least one putative recombination event with another genotype, in total 8, 85, and 53 events occurred within P. agardhii, P. rubescens, and between both species, respectively. The most frequent breaking points were from Pos. 526-883 (27), 707-890 (28) and 707-883 (10) of strain No10 (AJ890255). A high number of recombinations (250) was also detected using RDP.

Microcystin genotypes and microcystin variants

Genotypes A, C, D, H, No34, 64, 72 were found to contain exclusively arginine in position 2 of the molecule (Table 1, Fig. 1). Notably the gene region coding for the putative binding pocket, which is responsible for substrate activation during microcystin synthesis (A4-A5, Stachelhaus et al. 1999, position 424-735 of strain No10, AJ890255, 312 bp), showed two binding pockets only: The first arginine binding pocket genotype comprised genotype C, while all other strains (A, D, H, No34, 64, 72) containing arginine only formed another binding pocket genotype differing by 4.1% from C, d_N/d_S = 0.704, estimated through pairwise comparison by the method of Yang & Nielsen (2000). Strains of genotype B contained microcystin with homotyrosine (tyrosine) and leucine, but never arginine in position 2. Its putative binding pocket differed from the arginine binding pocket of genotype C by 3.2% (d_N/d_S = 3.34) and from the other arginine binding pocket of genotype A (D, H, strains No34, 64, 72) by 2.9% (d_N/d_S = 0.61). All other genotypes contained arginine and leucine, and in some cases additional low levels of homotyrosine (G, No31/1, 75, 82). Strains No40, 62, 65, 67, 91/1, 119, 120 were found inactive in MC production, which was partly caused by transposases or gene deletions (Christiansen et al. 2006), however, those strains did not differ in sequence compared with active mcy genotypes (Table 1).

Fig. 1.

Phylogenetic tree based on ML analysis from mcyBA1 sequences (1,451 bp) from 49 Planktothrix spp. strains. Strains assigned to P. agardhii are underlined. Capitals A-H in the superscript indicate the genotype groups (consisting of 2-11 strains) as indicated in Table 1. Genotype B consisted of both P. rubescens and P. agardhii. For each strain the corresponding restriction profile (AluI, TseI) and the occurrence of amino acids in position 2 of the microcystin molecule is indicated. The significant bootstrap percentages were obtained from 100 pseudo replicates. The two numbers shown in boxes are the maximum-likelihood estimates of the numbers of nonsynonymous/synonymous substitutions for the entire mcyBA1 along that branch (only ratios higher than the background ratio of the “one-ratio” model (d_N/d_S = 0.19) are shown, see text).

Distribution of microcystin gene restriction types among strains

Out of the 25 genotypes, 10 restriction types were obtained with AluI and 4 restriction types were obtained with TseI (Table 3, Fig. 1). In total 11 AluI+TseI combinations were observed. While some restriction types were found in both strains of P. rubescens and P. agardhii (VI+II) others were found only in P. agardhii (V+IV, VIII+III) or in P. rubescens (I+I, II+I, III+I, IX+I). The occurrence of I+I in P. rubescens and V+IV in P. agardhii was statistically significant (Chi-square test, χ² (df=1), p < 0.05). The restriction types I+I, II+I, III+I comprised a group of most closely related genotypes of P. rubescens, which differed in 0.5% of the base pairs only (15 strains, A, D, H, No34, 64, 91/1).

Restriction types I+I, II+I, III+I, V+IV identified those genotypes with arginine only in position 2 of the molecule. Genotype B lacking arginine was identified by restriction type VI+II. Those genotypes containing arginine/leucine (and homotyrosine) were identified by VI+I, VII+III, VIII+III, IX+I, X+I.

Distribution of microcystin gene restriction types among populations

In order to test the reproducibility of the results on the restriction type proportions from cloning libraries, mcyBA1 genes from field samples (Mondsee, 9 Dec 03, Wörthersee, 28 Aug 03) were independently amplified by PCR threefold, cloned, and subsequently digested using AluI after the re-amplification of 40 clones. The results were reproducible for the dominant restriction types (I, II, IV, VI) as well as the absent restriction types (VII, VIII). The results for the subdominant genotypes were more variable (Table 4).

Table 4.

Proportion of restriction types (I-X, new, new restriction types) of mcyBA1 obtained through AluI digestion after three independent PCR amplifications from field samples, cloning of the PCR product, and PCR-reamplification of forty randomly selected white clones from the respective clone library (1-3).

Clone library	I	II	III	IV	V	VI	VII	VIII	IX	X	new
Mondsee, 9 Dec 03
1	35	8	8	10	3	18	0	0	10	3	8
2	16	22	8	22	3	5	0	0	3	3	19
3	5	33	0	23	5	10	0	0	8	3	13
Mean ± 1SE	19±9	21±7	5±3	18±4	3±1	11±4	0	0	7±2	3±0	13±3
Wörthersee, 28 Aug 03
1	30	5	3	20	5	13	0	0	3	8	15
2	26	31	10	7	0	17	0	0	0	0	10
3	26	5	5	14	0	19	0	0	12	2	19
Mean ± 1SE	27±1	14±9	6±2	14±4	2±2	16±2	0	0	5±4	3±2	14±3

In total 24 AluI+TseI combinations were observed. Out of a total of 985 clones, the five most frequently detected restriction types were IV+I (23.8%), I+I (20%), VI+II (13%), X+I (10%), and VI+I (5.1%). All of the other restriction types occurred <5% (Fig. 2). Nine restriction types occurred only once (I+III, I+IV, II+III, III+II, III+III, IX+III, IV+III) or twice (VII+III, X+III). Other restriction types showed new AluI or TseI profiles that were derived from unknown organisms and that occurred at all of the sampling sites (11.9%). Two restriction types that were observed among the strains of P. agardhii (V+IV, VIII+III) were not detected in the field samples.

Fig. 2.

Proportion of AluI+TseI restriction types obtained from amplified and cloned mcyBA1 PCR products from eight different Planktothrix spp. populations for three sampling dates. For each sample, 40 clones of mcyBA1 PCR products were analysed and the percentage of a specific restriction type was calculated. The proportion for the total population and for the 49 sequenced strains is provided in the lower two graphs. Note that the scales of the y-axes are different. Significant differences of the proportions of restriction profiles between the eight lakes are marked by crosses at the top (xxx, p<0.001, xx, p<0.01, ns, not significant). The mean +/− 1SE number of the total number of restriction types found in a population is given in parentheses. Superscripts a, ab, b indicate homogeneous subsets whose highest and lowest means in the number of different restriction types are not significantly different (p > 0.05).

The distribution of restriction types differed significantly between sites, for example I+I and IV+I occurred frequently in many lakes and were not found in Schwarzensee. In contrast Schwarzensee was dominated by VI+II or X+I. Restriction type VI+II occurred in all of the lakes except for Irrsee and X+I was not found in Afritzersee but showed high frequency in Irrsee and Schwarzensee. In addition Schwarzensee had a significantly lower total number of restriction types when compared to the other lakes (one-way ANOVA, df = 23, p = 0.01).

In summary, many populations were found to be genetically heterogeneous. Closely located and spatially separated populations were found to differ significantly in mcyBA1 restriction type composition. Most populations contained mcyBA1 genotypes indicative of the production of arginine, homotyrosine, and leucine. In contrast genotypes indicative of arginine were absent in Schwarzensee, and that of homotyrosine were absent in Irrsee during the study period.

Discussion

Taxonomic distribution of microcystin genes in Planktothrix spp

In this study certain genotypes (C, E) and restriction types (V+IV, I+I) were found to occur exclusively either in one or the other species. On the other hand genotype B occurred in both species and genotypes E, No32, and No39 of P. agardhii were found to be the most closely related to the genotypes of P. rubescens. This conflict between taxonomy and the phylogenetic assignment of mcyBA1 genotypes can be best explained by relatively recent recombination events within mcyBA1. Homologous recombination has been increasingly recognized as an important force in prokaryotic evolution (Feil 2004). For example, in pathogenic bacteria, recombination has been estimated to change alleles 5-15 fold more frequently than mutation (e.g. Feil et al. 2003). For cyanobacteria, Rudi et al. (1998) demonstrated the occurrence of recombination for rbcL (D-ribulose 1,5-bisphosphate carboxylase-oxygenase) and the less conserved rbcX gene with a possible chaperonin-like function. Based on DNA-DNA hybridization studies, Suda et al. (2002) concluded that P. rubescens and P. agardhii differentiated relatively recently and the relatively high genetic similarity may favour the occurrence of recombination (Roberts & Cohan 1993). Typically both species do not occur in the same freshwater system, however, in some lakes both species have been found to co-occur over the course of years (e.g. Blelham Tarn, Davis et al. 2003). It would be interesting to know whether this co-occurence favours the occurrence of interspecific recombination events.

Microevolution of microcystin genes

In recent years a number of recombination events within microcystin genes were reported (Mikalsen et al. 2003, Tanabe et al. 2004, Kurmayer et al. 2005) suggesting that microcystin genes are under continuous modification and re-organisation. While Tanabe et al. (2004) observed the recombination of shorter DNA fragments (< 1,000 bp) within mcyA (but not within mcyD, mcyG, mcyJ), Kurmayer et al. (2005) reported the replacement of whole domains, i.e. the typical N-methyl-dehydroalanine adenylation domain (2,854 bp) of mcyAA1 was replaced by an adenylation domain without the N-methyl transferase (1,692 bp) resulting in dehydrobutyrine in position 7 instead of the common N-methyl-dehydroalanine. The flanking regions as well as the site of recombination in mcyAA1 were found to be identical within twelve strains containing the replaced domain implying that this recombination event happened only once (R. Kurmayer, C. Molitor, unpublished). In this study a larger number of short fragments that are indicative of recombination were observed implying that parts of the mcyB gene were frequently transferred between lineages of Planktothrix spp. It is concluded that the mcyA and mcyB genes generally show a mosaic structure rather than a bifurcating phylogenetic tree.

The “one ratio” maximum likelihood model estimated d_N/d_S = 0.19, which is similar to the ratios calculated by Tanabe et al. (2004), i.e. d_N/d_S = 0.2 (mcyA), 0.14 (mcyD), 0.17 (mcyG), 0.11 (mcyJ). Rantala et al. (2004) reported d_N/d_S <1 for mcyA, mcyD, mcyE. An excess of synonymous substitutions over non-synonymous ones indicates that mcy genes are subject to purifying selection and mutations affecting the protein sequence are in general deleterious. Notably the genotypes B (indicative of homotyrosine, leucine but no arginine) and C (indicative of arginine only) showed higher d_N/d_S ratios implying either relaxation of purifying selection or positive selection. Averaging d_N/d_S rates over all of the sites of the protein typically underestimates positive selection because the ratio is overwhelmed by the ubiquitous purifying selection. To increase the power of detection branch-site models have been invented to test for positive selection acting on specific sites of specific genotypes (Yang & Nielsen 2002, Zhang et al. 2005) or on a subset of sites in the whole phylogenetic tree (Yang et al. 2000). Indeed employing site models (Yang et al. 2000) 3% of the sites within the whole mcyBA1 phylogeny were found positively selected (d_N/d_S ~ 5.0, unpublished data). However, no statistically significant support for positive selection in specific branches was found implying that some sites in genotypes B, C rather experienced a relaxation of selective constraints. In the future quantifying the fitness of specific microcystin ecotypes (i.e. Kurmayer et al. 2005) under various environmental conditions will help to elucidate whether the observed differences in d_N/d_S ratios resulted from possible adaptive speciation events or from a relaxation of selective constraints.

Divergence of microcystin genotype composition between populations

In this study genotype C (2 strains isolated from the English Lake District, UK) was never detected in field samples in the Alpine lakes. In addition populations were found to differ with regard to the occurrence of specific genotypes as well as to the number of genotypes occurring in total in a specific population. This study is the first that documents homogeneous and more heterogeneous populations of aquatic cyanobacteria existing in closely located yet spatially isolated habitats over the course of several years. According to Horner-Devine (2004) three fundamental processes, the rates of dispersal, speciation, and extinction all contribute to the geographic patterns. There is no reason to assume that P. rubescens has fewer dispersal capabilities than other aquatic prokaryotes. Typically P. rubescens occurs in depths of 9-12 m (in Mondsee) with low light intensity (i.e. neutral buoyancy in Lake Zürich occurred between 4-8 μmol m⁻² s⁻¹, Walsby et al. 2004). Notably the populations in Schwarzensee and Irrsee showing the lowest genetic diversity observed in this study, also had the lowest numbers of individuals when compared with the populations of high genetic diversity (Mondsee, Wörthersee) as reported previously (Christiansen et al. 2006, Table 4). Re-colonization of those habitats by single genotypes after the extinction of the population during winter or other catastrophic events may contribute to the low homogeneity and the high degree of clonality observed in sparse populations. Consequently isolation combined with low numbers of individuals may lead to divergence in microcystin synthesis through genetic drift.