Although the mean total phytoplankton biovolume per year showed little change between 1980 and 2008 (minimum 1.5 mm³ L^-1, maximum 3.0 mm³ L^-1, mean 2.3 ± 0.1 mm³ L^-1), the mean Planktothrix biovolume per year underwent pronounced fluctuation (Figure 1). During a population collapse in 1984, with a minimum of 0.001 mm³ L^-1, Planktothrix accounted for only 0.1% of the total phytoplankton and 6.0% of the total cyanobacterial biovolume. Subsequently, the population recovered and increased up to a maximum of 2.0 ± 0.2 mm³ L^-1 in 2001. Within the period between 1985 and 2008, Planktothrix contributed on average 39 ± 2.9% to the total phytoplankton (minimum 11.5%, maximum 68.4%) and 87 ± 2.8% to the total cyanobacterial biovolume (minimum 47.6%, maximum 98.2%). Excepting 2005, the share of Planktothrix of the total phytoplankton exceeded 70% in the period from August to March during the years 1995 to 2008 (n = 43).

Planktothrix was detected by means of quantitative PCR (qPCR; using a Taq nuclease assay (TNA)) in all samples except for two (in the years 2002 and 2003). The mean Planktothrix biovolume per year was 1.6 ± 0.2 mm³ L^-1. The lowest annual Planktothrix biovolume was found in 1984 (0.001 mm³ L^-1) and the maximum annual biovolume was measured in 1997 (4.2 mm³ L^-1) (Figure 2). There was a highly significant positive and linear relationship between the Planktothrix biovolume quantified by 16S rDNA and the biovolume as determined by microscopic counting, y = -0.289 + 0.737x, n = 27, R² = 0.9, P < 0.001, where x is the average log₁₀ biovolume (mm³ L^-1) per year as determined by 16S rDNA and y is the average log₁₀ biovolume as determined by microscopic counting (Figure S1A in Additional file 1). The mean Planktothrix biovolume as calculated from microscopic counting was 0.71 ± 0.1 mm³ L^-1. The 16S rDNA abundance estimates were validated using a second TNA quantifying the phycocyanin intergenic spacer region (PC-IGS). Comparison of the biovolume as estimated by 16S rDNA and by the PC-IGS locus revealed a positive linear relationship following the equation y = -0.225 + 0.980x, n = 28, R² = 0.96, P < 0.001, where x is the average log₁₀ biovolume (mm³ L^-1) per year determined by 16S rDNA and y is the respective log₁₀ biovolume as determined via the PC-IGS (Figure S1B in Additional file 1). At four dates, the TNA for PC-IGS was negative, while the TNA for 16S rDNA was positive (three samples from 1984, and one sample from 1986). In all these cases the Planktothrix biovolume was very low (< 0.0006 mm³ L^-1), and the introduced variability was negligible. It is concluded that the TNA estimates on the Planktothrix abundance constitute a reliable estimate of the Planktothrix population density observed in the lake.

The TNA targeting the mcyB gene fragment indicative of the MC-producing genotype was found positive in 100 samples (87.8%). In 11 samples showing the lowest Planktothrix biovolume (< 0.0006 mm³ L^-1), no mcyB signal was recorded. A highly significant linear relationship was found between the abundance of the genotype carrying the mcyB gene fragment and the total population, y = -0.0358 + 0.960x, n = 27, R² = 0.94, P < 0.001 (Figure 3), where x is the average annual log₁₀ biovolume (mm³ L^-1) as determined by 16S rDNA and y is the average annual log₁₀ biovolume of the mcyB genotype. The mean proportion of the mcyB genotype was 106 ± 8% (minimum 38.9%, maximum 247.3%), indicating that the Planktothrix population was constantly dominated by the mcyB genotype (minimum 0.003 mm³ L^-1, maximum 4.8 mm³ L^-1, mean 1.6 ± 0.3 mm³ L^-1, Figure 2).

To elucidate whether the changes in environmental conditions affecting the phytoplankton composition also influenced the genotypic structure of the population, two types of nontoxic genotypes were quantified. First, genotypes containing the mcy gene cluster but inactive in MC production due to insertions by mobile elements or partial deletions of the mcy gene cluster 22; and second, genotypes that lack the mcy gene cluster due to evolutionary gene loss 4.

In general, all of the four genotypes containing either insertions or a deletion within the mcy gene cluster were detected frequently, but the majority in the lowest abundance (Figure 4). Only the genotype carrying a deletion between mcyH and mcyA occurred in sufficient amounts that could be quantified by the respective TNA. The mcyHA deletion was detected in 1977, although its numbers were below the quantification limit. It was not detected between 1980 and 1986, but occurred consistently later on: from 1987 to 2008, the genotype carrying the mcyHA deletion had an annual average proportion of 3.3 ± 0.4% (minimum 0.3%, maximum 7.2%) of the total population density as determined by 16S rDNA. In total, it was detected in 74% of the samples. Comparing its proportion between three decades (1977 to 1989, 1990 to 1999, 2000 to 2008) revealed no significant increase of this genotype (P = 0.16, Kruskal-Wallis one-way analysis of variance (ANOVA) on ranks). The other three inactive genotypes containing insertions were present in minimum concentrations only and could be detected in the undiluted DNA extract. The genotypes carrying insertions within mcyD, that is, mcyDIS1 and mcyDIS2, were detected in 87% and 88% of all the samples, respectively. These genotypes were detected throughout the investigated period except for the years 1983 and 1984, the years showing the lowest Planktothrix biovolume. By contrast, the mcyAIS genotype was detected as late as in 1996 and was then continuously present until 2008 (32% of the samples). It is concluded that inactive mcy genotypes occurred consistently, but never constituted a significant part of the total population.

Genotypes that lost the mcy gene cluster were quantified by amplifying a SNP in the mcyT gene, which constitutes a remnant of the mcy gene cluster in nontoxic strains 4. In a first step, the sensitivity and efficiency of the TaqMan Genotyping Assay for mcyT was tested by the serial dilution of DNA isolated from axenic strains PCC7811 (allele 1, nontoxic genotype that lost the mcy gene cluster) and PCC7821 (allele 2, toxic genotype that still contains the mcy gene cluster) and analyzed using an allelic discrimination plot (Figure S2 in Additional file 1). The SNP assay showed a clear separation of the nontoxic versus toxic genotype and the two SNP probes were considered to be specific. The lowest DNA concentration that could be depicted was equivalent to two cells per template. Mixtures of the DNA (calculated from cell equivalents) from the two strains containing a 2% concentration of DNA from nontoxic strain PCC7811 showed a slight deviation of the delta Rn values compared with those of pure DNA from the toxic strain PCC7821. The delta Rn values of DNA mixtures with concentrations of 10% and 20% of nontoxic strain PCC7811 were clearly different from pure DNA of the toxic strain PCC7821, while the concentration of 50% of DNA of the nontoxic strain PCC7811 was displayed in the center of the allelic discrimination plot (Figure S2 in Additional file 1). It is concluded that the nontoxic genotype was detected as soon as its share exceeded 2% of the total population.

The analysis of field samples obtained from Lake Zürich and other shallow eutrophic lakes (Table 1) by an allelic discrimination plot revealed four distinct clusters (Figure 5). The first cluster comprised samples clustering along the y-axis (allele 2, toxic genotype), indicating populations only composed of the toxic genotype. This cluster included samples dominated by the red-pigmented ecotype: Lake Zürich, Irrsee and Mondsee and one lake with a population of the red- and the green-pigmented ecotype (Steinsfjorden). Samples of the green-pigmented populations (Wannsee, Slotermeer, Tjeukemeer and Havel) formed a second cluster in the center of the allelic discrimination plot, indicating populations comprising both the toxic and the nontoxic genotype. The Planktothrix population tested from Klinkenberger Plas comprising the red- and the green-pigmented ecotype was also found in this central cluster. By contrast, samples from shallow polymictic lakes (for example, Lake Frederiksborg Slotssø, Zeegerplas and the Albufera Lagoon) with a green-pigmented Planktothrix population formed a third cluster along the x-axis (allele 1, nontoxic genotype), indicating populations composed entirely of the nontoxic genotype. It is concluded that, in Lake Zürich, the nontoxic genotype carrying the mcyT gene as a remnant of the mcy gene cluster never became abundant during the entire observation period. By contrast, Planktothrix populations that were either only green-pigmented or both red- and green-pigmented typically showed a higher proportion of the nontoxic genotype.

To find out whether the abundance of the green-pigmented (nontoxic) ecotype was influenced by the trophic change during the study period, two phylogenetic lineages differing in pigmentation and the presence of the mcy gene cluster were quantified using the P700 apoprotein subunit Ia (psaA) and Ib (psaB) intergenic spacer region (PSA): lineage 1, the green-pigmented, nontoxic genotype that lost the mcy gene cluster (PSA I), and lineage 2, both red- and green-pigmented toxic genotype that always carried the mcy gene cluster (PSA II) 4. The green-pigmented lineage 1 was detected during the entire observation period although in a smaller number of samples (43%). By contrast, lineage 2 (red- and green-pigmented ecotypes containing the mcy genes) was detected in 83% of all the samples. A few samples with low Planktothrix biovolume (< 0.09 mm³ L^-1) did not show a TNA signal indicative of lineage 2 (17% of the samples). The average proportion of the genotype of lineage 1 was generally low (minimum 0%, maximum 3.8%, mean 0.9 ± 0.2%) and no increase or decrease was observed when comparing the mean proportion of this genotype between the three decades: 1977 to 1989: 1.3 ± 0.5% (n = 9), 1990 to 1999: 0.6 ± 0.2% (n = 9) and 2000 to 2008: 0.7 ± 0.2% (n = 9); P = 0.26. Maximum abundances were measured in 1983 and 1985 (3.8% and 3.5% of the total population), coinciding with the time of the Planktothrix population breakdown and the following recovery. The mean proportion of the genotype of lineage 2 was 7.3 ± 1% of the total population (minimum 0.2%, maximum 18.5%). A significant increase of this genotype was detected when comparing the mean proportion between the three decades: 1977 to 1989: 2.3 ± 0.5% (n = 9), 1990 to 1999: 7.5 ± 1.5 (n = 9) and 2000 to 2008: 12.1 ± 1.2% (n = 9); P < 0.001.

The abundances of genotypes of lineage 1 and 2 were both linearly correlated to the total population as follows: lineage 1, y = -2.165 + 0.57x, n = 23, R² = 0.45, P < 0.001 (dotted line, Figure 3); lineage 2, y = -1.207 + 1.1x, n = 27, R² = 0.83, P < 0.001, (dashed line, Figure 3), where x is the log₁₀ biovolume of the total population determined by 16S rDNA (mm³ L^-1) and y is the log₁₀ biovolume of Planktothrix lineage 1 or 2 (mm³ L^-1), respectively. The regression curve of lineage 1 significantly differed in intercept and slope from the regression curve obtained between the mcyB genotype and the total population (P < 0.001, n = 27). By contrast, the regression curve of lineage 2 significantly differed in intercept but not in the slope (P = 0.179, n = 27) from the mcyB regression curve. Thus, where lineage 2 showed a parallel increase with the abundance of the total population, lineage 1 showed a tendency to decrease in proportion with a higher Planktothrix population density. The same results were obtained when the samples from the years with lowest Planktothrix density (1977 and 1982 to 1984) were excluded (data not shown). It is concluded that the genotype of lineage 2 consistently outgrew the genotype of lineage 1 under conditions of favorable growth for the Planktothrix population.

The observed increase of the Planktothrix population in Lake Zürich over the last 30 years can be explained by the improved light regime in the water column that occurred following re-oligotrophication, through the disappearance of eukaryotic algae blooms. This phenomenon has already been documented for other lakes 2. It is well known that P. rubescens stratifies in the metalimnion of thermally stratified lakes at approximately 12 m depth, and grows at the lowest irradiance 23. Furthermore, Planktothrix can take advantage of higher concentrations of the limiting macronutrient (soluble reactive phosphorus) in the metalimnion as well as intracellularly stored phosphate for three or four further cell divisions 124. The population breakdown in 1984 was probably the consequence of various factors that weakened the population. First, higher biomasses of eukaryotic algae occurred during the years 1983 and 1984 and might have reduced deep light penetration and forced Planktothrix to enter the mixed epilimnion 25. Mur 26 reported a higher growth rate of the green algae (Scenedesmus) at high light intensities whereas Planktothrix grew more efficiently at the lowest light intensities, thereby outcompeting Scenedesmus. Second, Posch et al. 27 reported that holomixis occurred in Lake Zürich from 1980 to 1987, probably leading to major losses in the P. rubescens population. Walsby and co-workers 17 reported a major decrease in population biomass in spring after a strong winter 1996 to 1997) when compared with population biomass in spring after a warmer winter (1994 to 1995). Posch et al. 27 further suggested that the increased epilimnetic water temperatures since 1988 led to a more stabilized stratification of the water column and allowed the survival of a bigger share of the population over winter due to reduced loss processes during lake mixing. However, it is unclear whether changes in environmental parameters (for example, depth of the euphotic zone or higher stability of the water column due to rising water temperatures) may not only have an impact on the growth of the total population but also on the subpopulation dynamics of specific ecotypes.

Strikingly, the genotypic composition of Planktothrix was found to be rather stable over the last 30 years. The population was constantly dominated by the toxic mcy-containing genotype, while the share of the nontoxic (mcy-lacking) genotype was low. Sabart et al. 28 found the proportions of toxic and nontoxic genotypes in the populations of Microcystis along the catchment of River Loire to be generally stable over time but variable between sampling sites. Okello et al. 29 reported differing proportions of toxic and nontoxic genotypes from spatially isolated Microcystis populations in Uganda, and these differences in proportions were much more pronounced than the observed seasonal variation within lakes. These findings are in accordance with the present study and imply the role of both abiotic and biotic parameters of the habitat (for example, light regime, lake morphometry, zooplankton grazers, viruses) that regulate genotype composition 12. It is important to note that these factors are not directly related to MC production and that effects of re-oligotrophication on the Microcystis genotype structure will differ considerably from those on the Planktothrix genotypic composition because of the overall difference between the ecological niches of the two taxa 14.

Probably the best example for an ecological trait not directly related to MC production is green versus red pigmentation. In Lake Zürich, the green-pigmented nontoxic Planktothrix lineage 1 was detected throughout the study period, but in very low proportions only. Interestingly, in the period before and after the population breakdown, a maximum proportion of the green-pigmented lineage 1 was observed. During this eutrophic period, blooms formed by non-stratifying phytoplankton still occurred (Figure 1). Under these conditions, the green-pigmented ecotype might have been at a selective advantage because of the variable irradiances, as it is found to be characteristic of shallow polymictic lakes 14. By contrast, the red-pigmented ecotype was shown to be a better competitor for light only under mesotrophic and stratified conditions as its compensation depth for growth was lower compared with that of the green-pigmented ecotype 13. Consequently, under eutrophic strong light shading conditions, the red-pigmented Planktothrix are seldom abundant 1430.

Because Lake Zürich is a deep lake, the resistance of gas vesicles to hydrostatic pressure may be a critical factor for survival during conditions of lake mixing, as gas vesicles may collapse and filaments cannot return to the euphotic zone. It has previously been shown that the red- and green-pigmented ecotypes of Planktothrix vary in the presence of gas vesicle proteins, which could be attributed to lake depths 18. The filaments remaining buoyant after mixing will, therefore, form the basis of the population for the next season 17. Gas vesicles resisting strong hydrostatic pressures are more commonly found among the red-pigmented strains of Planktothrix, as in the population of Lake Zürich, whereas gas vesicles that collapse at a relatively low critical pressure are more common within green-pigmented populations 18. Additionally, the genotype that produces gas vesicles that resist the highest hydrostatic pressures, which is the dominant genotype in Lake Zürich, was not detected in isolates obtained from 21 Norwegian lakes with a maximum depth of 67 m 18. Accordingly, in red-pigmented populations, stronger gas vesicle types were found to be positively related to lake depth 31.

It is likely that the green-pigmented Planktothrix ecotype of Lake Zürich produces gas vesicles that do not resist the critical pressure during lake mixing and that only a small percentage of filaments survive entrainment in the hypolimnion. Therefore, in spring, the basis for the population development of the green-pigmented ecotype would be low when compared with the red-pigmented ecotype, of which a larger proportion can survive mixing during the winter (> 90% of the filaments 17). Our own preliminary results showed that strains of nontoxic lineage 1 almost exclusively contained the weakest gas vesicle genotype (gvpC²⁸) while strains of toxic lineage 2 typically contained two gas vesicle genotypes (gvpC²⁸ and gvpC²⁰). Some of the strains of lineage 2 also showed the genotype encoding the smallest gas vesicle (gvpC¹⁶), which is known to resist high hydrostatic pressure 18.

The inactive genotypes formed a stable but very small subpopulation throughout the study period. By contrast, the genotype carrying an insertion within mcyA (mcyAIS) was detected for the first time in 1996. It is conceivable that this genotype arrived in the lake at a later time than the other inactive genotypes, given that all of these four inactive genotypes occurred regularly in different populations in the Alps 22. It is concluded that these inactive mcy-containing genotypes grew very slowly and still unknown factors keep their abundance low.

The role of MC as a feeding deterrent has been investigated and evidence has been gathered that MCs are indeed toxins resulting in a significantly reduced survival rate of herbivorous crustaceans 73233. The frequently cited hypothesis that MC-producing cyanobacteria evolved before their potential predators (the metazoans), which is based on phylogenetic analyses 34, ignores the fact that chemical defense still could be an additional (secondary) function simply because MCs are very effective toxins (for example, see 35). In addition, the evolutionary age of eukaryotes has been a matter of debate, and the coexistence of eukaryotes with prokaryotes has been inferred from biomolecules in fossil oil droplets with an age of 2.4 billion years 36. Thus, the possible co-occurrence of cyanobacteria and eukaryotes before snowball earth (2 billion years ago) is a matter of ongoing research. We suggest that, in nontoxic P. agardhii, other bioactive but nontoxic peptides functionally replace MC. For example, both lysogenic bacteria 37 and parasitic fungi have been described; for example, Sønstebø and Rohrlack 38 reported a relationship between chytridiomycete infectability of strains and the presence of certain peptides, such as cyanopeptolins and anabaenopeptins. The higher variability in toxic and bioactive peptide genotype proportion among nontoxic green-pigmented populations could be because of an ongoing co-evolutionary arms race between host and parasite 38.

As suggested for Microcystis 39, MC is considered to have an intracellular impact on the stability and activity of proteins involved in carbon-nitrogen metabolism by interfering with their redox state control. The enhanced binding of MC to proteins is thought to be part of a general response to oxidative stress 3940. The binding of MC to cysteine residues of proteins might be stimulated by conformational changes, which delays the degradation of redox-sensitive proteins. Zilliges et al. 39 showed that a MC-deficient mutant of Microcystis was more susceptible to high irradiance compared with the MC-producing wild type. It is conceivable that, in particular for red-pigmented Planktothrix, conditions of enhanced oxidative stress occur, for example when the low-light-adapted buoyant Planktothrix population accumulates at the surface in the autumn 41. It is known that, among all phytoplankton species, red-pigmented Planktothrix is most efficient in light harvesting but this capability increases the chance of light damage to the cells under high irradiance, such as during calm days at the water surface 41. Green-pigmented P. agardhii has been found more resistant to high light intensities 15. In light of these findings, it can be speculated that MC production would be of selective advantage in red-pigmented Planktothrix populations, whereas MC production might be of less selective value in green-pigmented populations due to the general lower sensitivity to high light damage. Further, mcy genotypes inactive in MC production in red-pigmented populations would be selectively reduced during surface blooms under high irradiance conditions. It is possible that mcy genotypes inactive in MC synthesis grow under physically stable stratified conditions of the water column, as losses due to accidental accumulation at the surface will be low.

Lake Zürich in Switzerland (47°15'N, 8°38'E, 406 m above sea level) is a deep (mean depth 51 m, maximum depth 136 m), monomictic lake with a surface area of 68 km². The first signs of Lake Zürich eutrophication were observed before 1890 (problems with the clogging of drinking water filters due to high zooplankton biovolumes). At the peak of eutrophication in 1965, the total phosphorus concentration reached a maximum of 100 μg L^-1 at spring overturn 50 and the Planktothrix abundance was close to zero from 1965 to 1975. Since 1975, the Planktothrix population has recovered and has an average share of 33% per year (minimum 0.1%, maximum 68%) of the total phytoplankton biovolume between 1980 and 2008. Today, Lake Zürich is classified as mesotrophic with average total phosphorus concentrations of 23 μg L^-1 (minimum 3.9 μg L^-1, maximum 78.1 μg L^-1) between 1980 and 2008. For the analysis of both phytoplankton and total phosphorus concentrations (Figures 1 and 2 and Figure S1A in Additional file 1), the average of the countings between the surface and 20 m in depth (0, 1, 2.5, 5, 7.5, 10, 12.5, 15 and 20 m) of the monthly sampling data was used.

A total of 111 samples from 1977 to 2008 were analyzed (in 1981 and 1998, no sampling was performed). The samples originated from a biweekly monitoring program targeting the euphotic zone by separate filtration of two liters from depths of 0, 5, 10 and 20 m through glass fiber filters, which were subsequently dried at 110°C for 2 h and stored in a desiccator until the next day. The filters were glued on paper and stored at room temperature (Figure S3 in Additional file 1). Filters were inspected and four filters from each year that showed maximum biomass were used for DNA extraction. Thus, the sampling depth varied according to the occurrence of the maximum population density. From the majority of the samples, one half of the filter with the preserved phytoplankton was used for DNA extraction, from ten samples only one quarter of the filter was used, and from three samples (600 mL from different depths, 7 December 1977) the phytoplankton of the complete filter was used for DNA extraction.

DNA was extracted from filters using the chloroform-phenol method as previously described 51. The extracted DNA was precipitated from the aqueous phase by adding half the sample volume of ammonium acetate (7.5 M), one tenth of the sample volume of magnesium chloride (0.1 M), 1 μL of glycogen (Fermentas, St. Leon-Rot, Germany) and one sample volume of isopropanol before overnight incubation (-20°C). After centrifugation (16,000 g, 4°C, 1 h) the supernatant was discarded and samples were washed twice with 70% ethanol by centrifugation (20 min, 4°C). The DNA pellets were dried in a sterile hood and resuspended in 50 μL of sterile Millipore water.

Samples were analyzed by TNA for the total population of Planktothrix by the 16S rDNA locus 11 and the PC-IGS 46; for the abundance of the mcyB genotype encoding MC synthesis 12; for the abundance of various mutant mcy genotypes carrying either a deletion or insertion inactivating MC biosynthesis 22, such as mcyDIS1, mcyDIS2, mcyAIS (insertion of a transposable element into the mcy gene cluster) and mcyHA (deletion of 1.8 kbp between mcyH and mcyA); for the abundance of the nontoxic genotype that lost the mcy gene cluster 4; and for the abundance of red-pigmented versus green-pigmented ecotypes via the IGS between psaA and psaB (PSA-IGS; see below). The nontoxic genotype that lost the mcy gene cluster, except for mcyT, was differentiated from the genotype still containing the mcy gene cluster using a SNP. The primers and TaqMan probes for the 16S, PC-IGS, the mcyB gene and the four inactive mcy genotypes have been published previously (Table 2). For the design of TNAs targeting the PSA-IGS locus, sequences from 62 Planktothrix strains (GenBank:EU258202 to EU2582634) were aligned together with 78 sequences of Planktothrix strains isolated from Europe, North America and East Africa (R. Kurmayer, unpublished) by means of Clustal W 2.0. A target region was identified that discriminates all the strains of lineage 1 (n = 56; containing strains that are green-pigmented and lost the mcy gene cluster) from strains of lineage 2 (n = 84; containing both red-pigmented and green-pigmented strains that always retained the full mcy gene cluster). The following forward primers, TaqMan probes and reverse primers were designed: lineage 1 (167 bp): forward primer PSA I + (5'-CCAGCAATTCAACCTCGC-3'), TaqMan probe PSA I (5'-TGGTGTAGCTCACTACCTCTTAGGAGGCAT-3') and reverse primer PSA I- (5'-AAAGTAGATTAGATTTCCTCCCACCT-3'); lineage 2 (158 bp): forward primer PSA II + (5'-CCAGCAATTCAACCTCGC-3'), TaqMan probe PSA II (5'-CGTGCGGTTGGTGTAGCTCACTACC-3') and reverse primer PSA II- (5'-ATTCAGCCTTTCCCAGTCCC-3'). Both probes were labeled with 6-carboxyfluorescein (FAM) at the 5'-end and BlackBerry Non-Fluorescent Quencher (NFQ) at the 3'-end.

In total, 113 mcyT sequences (GenBank: EU266304 to EU266364, and RK, unpublished; Table S1 in Additional file 1) of both toxic and nontoxic strains were aligned. A SNP was found to discriminate all the strains that lost the mcy gene cluster from strains that still contain the full mcy gene cluster. A custom TaqMan SNP genotyping assay based on this SNP was created using the Custom TaqMan Assay Design Tool (Applied Biosystems, ABI, Vienna, Austria) to differentiate the two genotypes. The following forward and reverse primers and two TaqMan probes labeled with 4,7,2'-trichloro-7'-phenyl-6-carboxyfluorescein (VIC) or FAM at the 5'-end and a NFQ at the 3'-end were obtained: mcyT SNP + (5'-ACAGAGAAAGCCGAGTTGGTT-3'), mcyT SNP-(5'-AGATTTGAAACCTAACGCCTTGGA-3'), TaqMan Probe allele 1 (VIC 5'- TGTTCCCACCAAGCTT-3' NFQ) and TaqMan Probe allele 2 (FAM 5'- CCCGCCAAGCTT-3' NFQ) (66 bp). The specificity and sensitivity of the assay were tested by measuring the mixtures of DNA from axenic strains PCC7811 (loss of the mcy gene cluster except the mcyT gene) and PCC7821 (containing the mcy gene cluster).

All the samples were measured in triplicate in a total volume of 25 μL including 5 μL of template DNA. The initial denaturation of 10 min at 95°C was followed by 50 cycles of a two-step PCR with an annealing and elongation temperature of 60°C or 55°C on an Eppendorf Master Cycler Ep Realplex system (Eppendorf, Vienna, Austria), as described 11. To increase the DNA concentration in the template as much as possible, the inactive mcy genotypes were measured in a total volume of 12.5 μL including 1 μL of the pure DNA extract and 1 μL (50 μg μL^-1) of bovine serum albumin. Because these results were frequently below the limit of quantification, only a presence/absence analysis was performed.

TaqMan assays were tested for specificity and sensitivity using dilution series of DNA isolated from Planktothrix strains 110, 139, 40 and 62, as described (Table 2) 22. Calibration curves were also prepared from a dilution series of DNA from strains PCC7811 and PCC7821 for the TNA PSA I and PSA II, representing lineage 1 or lineage 2 4. Strains PCC7811 and PCC7821 were grown at 20°C in a BG11 medium under continuous light (5 to 15 μmol photons m^-2 s^-1) and the cells were harvested during logarithmic growth. For each TNA calibration curve, the predetermined DNA concentrations in the template (expressed as biovolume per template) were diluted and related to the measured C_t values 22. The limit of quantification was defined as the lowest concentration of the calibration curve, which corresponded to 3 to 18 cells per template for the individual assays (Table 2). In the case that the C_t values of a particular gene locus were below the limit of quantification, only the results from the 16S rDNA locus were included in the analysis (n = 8).

All the samples were measured in triplicate on a ViiA7 Real-Time PCR System (ABI, Vienna, Austria), using the ViiA7 software (version 1.2.1) in a total volume of 12.5 μL including 2.5 μL of template DNA, ABI TaqMan Universal PCR Master Mix, the SNP Genotyping Assay Mix and sterile water according to the manufacturer's recommendations. A two-step PCR of 55 cycles was run with denaturation (92°C, 15 s) and annealing and extension (60°C, 1 min). The protocol included an initial denaturation step at 92°C for 10 min and a pre-read and post-read step at 60°C for 30 s, each. The relative threshold cycle was automatically set to 0.04 for all reads. The results were analyzed based on the delta Rn values (with Rn = fluorescence signal of reporter dye normalized to the fluorescence signal of the passive reference ROX) for allele 1 and allele 2, which represent the difference of the normalized fluorescence signal (VIC, FAM) between the pre-PCR and the post-PCR read.

To test the sensitivity and the specificity of the SNP assay, DNA of the axenic nontoxic strain PCC7811 was mixed down to 0.2, 0.5, 1, 2, 10, 20 and 50% of DNA from the toxic strain PCC7821 per template (calculated as cell equivalents). Dilution series from DNA of axenic strains PCC7811 and PCC7821 reaching from 2 to 180,000 and 2 to 230,000 cells per template, respectively, were established. At least one sample per year (1977 to 2008, n = 35), as well as depth integrated and net samples (0 to 20 m) of Lake Zürich from the years 2005, 2006 and 2007 (n = 16 11) were analyzed. In addition, samples of various European lakes composed of either green- or mixed-pigmented Planktothrix populations 12 were included (Table 1).

To calculate regression curves, the raw data were log₁₀-transformed and tested for normal distribution (Shapiro-Wilks, P > 0.05) and for constant variance by computing Spearman's rank correlation between the absolute values of the residuals and the observed value of the dependent variable (P > 0.05). The residuals were tested for their independence from each other by the Durbin-Watson statistic.

The linear regressions between the total population density (as estimated from 16S rDNA) and the abundance of the mcyB gene and the two PSA-IGS genotypes (representing lineage 1 and lineage 2) were compared in terms of the slope and intercept using a general factorial model of ANOVA 52. A SPSS statistical package (V 19.0 for Windows) was used. To compare the genotype proportions from samples between the three decades (1977 to 1989, 1990 to 1999, 2000 to 2008), the nonparametric Kruskal-Wallis one-way ANOVA on ranks was used.