Nodularin toxin is commonly detected in cyanobacterial blooms containing Nodularia spp. (2, 42), and Nodularia isolates from the blooms usually produce nodularin (6, 20, 42). Nodularin is a cyclic pentapeptide hepatotoxin with a 50% lethal dose of 50 to 70 μg kg⁻¹ when tested intraperitoneally in mice (39, 42). Furthermore, nodularin is an inhibitor of protein phosphatases 1 and 2A and a potent tumor promoter (36). Toxic blooms of Nodularia have been associated with poisonings of domestic animals in different parts of the world (see, for example, references 13, 23, and 29). A few nontoxic strains of Nodularia have been isolated from the plankton of the Baltic Sea (20). In physiologic experiments, the nontoxic strains never produced nodularin, whereas the nodularin-producing strains were toxic in all test conditions (19).

Traditionally, Nodularia have been classified on the basis of the morphology of the different types of cells (vegetative cells, heterocytes, and akinetes), on their ability to produce gas vesicles (structures essential for providing buoyancy), on nodularin production, on ultrastructural features of the cells, and on ecological characteristics (17). The genus Nodularia was recently divided into seven species (17). Four species (Nodularia spumigena, N. baltica, N. litorea, and N. crassa) are planktic, with the capability to produce gas vesicles. Three of the species (N. harveyana, N. sphaerocarpa, and N. willei) lack gas vesicles and grow in benthic, periphytic, or soil habitats. Phylogenetically, on the basis of the 16S rRNA gene, the genus Nodularia is most closely related to the genera Anabaena, Nostoc, and Cylindrospermum (22, 44, 45). Methods involving the whole genome (20) and 16S rRNA sequences (20, 26) have indicated the close overall relatedness of Nodularia strains and also distinguished the nodularin-producing strains from the nontoxic ones. In addition, putative peptide synthetase and polyketide synthetase sequences distinguished the toxic from the nontoxic strains (25). Nodularia isolates from the Baltic Sea plankton were assigned to three groups based on the 16S-23S rRNA internally transcribed spacer (ITS) sequences, to three groups on the basis of the sequences from the phycocyanin encoding operon intergenic spacer and flanking regions (PC-IGS), and two groups based on the intergenic region between genes encoding gas vesicle proteins (gvpA-IGS) (3). Data from RAPD [random(ly) amplified polymorphic DNA]-PCR and PC-IGS sequences have suggested that hierarchical patterns of genetic variation within Nodularia isolates exist on both regional and global scales (6).

From both an ecological and a human point of view it would be valuable to understand which conditions in nature favor the occurrence of nodularin-producing and nontoxic strains. Analysis methods, such as high-pressure liquid chromatography (HPLC), enzyme-linked immunosorbent assay (ELISA), or the mouse bioassay enable detection of hepatotoxins from water samples. However, single strains or filaments of Nodularia spp. cannot be assessed. For the detection of individual strains of Nodularia in the field, toxicologic and molecular information derived from laboratory isolates can be extended to the natural populations by detecting predetermined alleles by PCR amplification and subsequent molecular analyses of Nodularia filaments.

In the present study, 18 cultured Nodularia strains from planktic, benthic, and soil habitats were characterized by phenotypic and molecular approaches. In terms of classical taxonomy (17), the examined strains were defined as Nodularia sp., N. spumigena, N. baltica, N. harveyana, and N. sphaerocarpa. The characterization was performed in order to find morphologic or molecular markers for discriminating nodularin-producing from nontoxic strains of Nodularia in water samples. In addition, a field survey was undertaken in the Gulf of Finland to assess the composition of the Nodularia populations.

Nodularia strains. Eighteen monospecific strains of Nodularia were used in the study (Table 1). They were grown in liquid Z8 medium without nitrogen and with added salt (43). The strains N. harveyana Hübel 1983/300 and N. baltica Hübel 1988/306a and Hübel 1988/306b had been used in the intrageneric evaluation of Nodularia taxonomy (17). The species description of N. baltica (17) is partly based on strain Hübel 1988/306, which was later on divided into strains Hübel 1988/306a and Hübel 1988/306b.

Phenotypic characteristics of the strains. The sizes of vegetative cells and heterocytes were measured from 3-week-old cultures, and the akinetes were measured from 3.5-month-old cultures. Examination was carried out with a Leica Aristoplan phase-contrast microscope. Thirty cells were measured; however, when the akinetes were rare, the measurements were made with fifteen cells. The presence or absence of gas vesicles and the shapes of terminal cells were recorded. The averages and standard deviations of the sizes of the different kinds of cells were calculated for each Nodularia strain. Photographs of 3-week-old strains were taken with a Polaroid digital microscope camera fitted to a Leitz DM IRB phase-contrast microscope.

Principal component analysis (PCA) of the mean lengths and widths of the vegetative cells, heterocytes, and akinetes of each strain and of the descriptions of Nodularia species (as in reference 17) was performed with the program Statgraphics Plus 3.0. The data was standardized prior to PCA by reducing the average of the sample from each value and by dividing the result by the standard deviation of the sample. The purpose of PCA was to relate the Nodularia strains to each other and to the descriptions of Nodularia species in terms of cell dimensions. Other characters of taxonomical importance, such as nodularin production, the isolation habitat, the shapes of terminal cells, and the presence or absence of gas vesicles, were taken into account when each Nodularia strain was assigned to a species. The cell dimensions of the descriptions of Nodularia species (N. spumigena, N. baltica, N. harveyana, N. litorea, and N. sphaerocarpa) used in the PCA were as described previously (17).

Nodularin production of strains Hübel 1983/300, Hübel 1988/306a, and Hübel 1987/311 was determined by HPLC as described previously (19). Strains Hübel 1988/306b and Hübel 1987/310 were tested for hepatotoxin production by using an EnviroGard ELISA kit (Strategic Diagnostics, Inc.) following the instructions of the manufacturer. Information on the nodularin production of the rest of the Nodularia strains used in this study is available elsewhere (20, 42).

Preparation of template and PCR. PCR amplifications were performed alternatively with DNA or 2 μl of culture as a template. The DNA was extracted as described by Golden et al. (12) without further purification. The 2-μl aliquot from a culture was boiled for 10 min prior to PCR, and polymerase was added after the denaturation step (hot start).

The 16S-23S rRNA ITS (ITS1-S) region was amplified with the cyanospecific primers 16CITS and 23CITS (31). The ITS1-S PCRs (50 μl) contained 2 μl of Nodularia culture, 200 μM concentrations of each deoxynucleoside triphosphate (dNTP; Finnzymes, Espoo, Finland), 250 nM concentrations of the primers, 1× DyNAzyme II polymerase buffer, and 1 U of DNA polymerase DyNAzyme II (Finnzymes). The ITS1-S PCR protocol consisted of an initial denaturation at 94°C for 3 min; 30 cycles of 94°C for 15 s, 50°C for 30 s, and 72°C for 1 min; and an extension step at 72°C for 5 min.

The PC-IGS region was amplified using the primers PCβF and PCαR specific for cyanobacteria (30). The reactions (50 μl) contained either 2 μl of culture or 20 to 30 ng of DNA as templates, 100 μM concentrations of each dNTP (Finnzymes), 100 nM concentrations of both oligonucleotide primers, 1× DyNAzyme II polymerase buffer, and 0.5 U of DNA polymerase DyNAzyme II (Finnzymes). The PCR protocol was modified from that of Hayes and Barker (14) for cultured material by raising the annealing temperature from 55 to 61°C and by using a hot start.

Electrophoresis of the PCR products was carried out in 1.5% (wt/vol) agarose gels in 0.5× TAE buffer (20 mM Tris-acetate, 0.5 mM EDTA [pH 8.0]), stained with ethidium bromide, and visualized and photographed under UV light.

Sequencing and construction of phylogenetic trees. The template for DNA sequencing of the PC-IGS region was purified from the PCR products with Wizard PCR Preps DNA purification kit (Promega, Madison, Wis.). The ITS1-S PCR products were first separated by gel electrophoresis, and the smallest band (ITS1-S) was excised from the gel and subsequently purified with the Wizard PCR Preps DNA purification kit. The ITS1-S and PC-IGS sequences were resolved on both strands by using the same primers as for PCR. For sequencing, 5 to 20 ng of DNA was applied in a PCR with the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit according to the manufacturer's instructions (PE Applied Biosystems, Foster City, Calif.). The sequencing was carried out on an ABI Prism 310 Genetic Analyzer (PE Applied Biosystems). The ITS1-S and PC-IGS sequences of each strain were checked by aligning the forward and reverse sequences with PILEUP of the GCG package version 10.1 (Genetics Computer Group, Madison, Wis.) and manual editing using the GeneDoc multiple sequence alignment editor.

Two ITS1-S (AJ224448 and AJ224449) (3) and three PC-IGS (AJ224914, AJ224915, and AJ224916) (14) Nodularia sequences reported from the Baltic Sea were used as references in the ITS1-S and PC-IGS alignments. The ITS1-S sequences were aligned with conserved domains of cyanobacterial ITS sequences (15). Phylogenetic trees were constructed by the neighbor-joining method on Jukes and Cantor distances and by the Wagner parsimony method using PHYLIP (10). The trees were statistically evaluated by 500 bootstrap resamplings. Sequences AF180969 and AF178757 from Nostoc sp. strain PCC7120 were used as outgroups in ITS1-S and PC-IGS trees, respectively.

Sampling in the Gulf of Finland and determination of the PC-IGS allele. Sampling for the determination of the PC-IGS allele of single Nodularia filaments was performed from the research vessel Aranda (Finnish Institute of Marine Research) between 7 and 11 August 2000. Plankton samples were obtained with a 100-μm-mesh-size plankton net from surface water (0 to 4 m) at six sampling stations in the Gulf of Finland (the Baltic Sea). Nodularia filaments were picked and treated as described by Barker et al. (4) with slight modifications. The washed filaments were transferred to PCR tubes containing 8 μl of 10× DyNAzyme II DNA polymerase buffer and 16 μl of sterile water. After immersion in boiling water for 5 min to lyse the cells, the 345 sample tubes were stored at −20°C until use as templates in the PCR amplifications. Positive control samples (nodularin-producing strain AV3 and nontoxic strain UP16f) were treated similarly to the field samples, and the negative controls contained no templates.

The PC-IGS regions of the filaments were amplified in 25-μl volume reactions with 8 μl of template, otherwise using conditions similar to those for the cultures. The amplified PCR products (12 μl) were digested at 37°C for 2 h with 2 U of HaeIII restriction endonuclease (Promega). The restriction fragments were visualized by running the reactions through 1.5% (wt/vol) agarose gel at 76 V for 55 min, by staining with ethidium bromide, and by photographing the results under UV illumination. The enzyme for distinguishing the nodularin-producing from the nontoxic Nodularia strains was chosen by screening both the ITS1-S and the PC-IGS sequences from each Nodularia strain by Map and Mapsort of the GCG package for appropriate recognition sites. HaeIII recognized the site 5′-GGCC-3′, which was present in the nodularin-producing planktic strains and cut the PC-IGS PCR products into three fragments of ca. 380, 190, and 100 bp. The enzyme left intact the PC-IGS allele found in the nontoxic strains (HKVV, PCC73104/1, UP16a, UP16f, and Hübel 1983/300) and in strain PCC7804. Strain PCC7804 was isolated in a French thermal spring in 1966 and mainly produces a nodularin variant (5).

The nucleotide sequences have been deposited in the GenBank-EMBL database under the accession numbers AF367159 to AF367176 (ITS1-S) and AF367141 to AF367158 (PC-IGS). In addition, the ITS1-S and PC-IGS alignments have been deposited in the GenBank-EMBL database.

Phenotypic features of the strains. Eight Nodularia strains (Hübel 1988/306a, Hübel 1988/306b, Hübel 1987/310, Hübel 1987/311, NSPI-05, NSOR-12, GR8b, and AV63) had discoid cells in unbroken, straight, or slightly curved trichomes (Fig. 1a to g and k). The trichomes of F81, AV3, and HEM were straight and fragmented, and the cells were spherical rather than discoid (Fig. 1i, j, and l). Strains BY1, UP16a, UP16f, HKVV, PCC 73104/1, Hübel 1983/300, and PCC7804 had unbroken and straight or flexuous trichomes with roundish to barrel-like cells (Fig. 1h and m to r).

FIG. 1 Photomicrographs of Nodularia strains used in the study. The heterocytes are indicated with black arrows and the akinetes with white ones. Panels: a, Hübel 1988/306a; b, Hübel 1988/306b; c, Hübel 1987/310; d, Hübel 1987/311; (more ...)

The lengths of the vegetative cells of all strains varied from 2.7 μm (Hübel 1983/300) to 6.5 μm (UP16a), and the widths varied from 5.7 μm (BY1) to 12.3 μm (NSPI-05) (Table 2). The lengths of the heterocytes varied from 4.5 μm (NSOR-12) to 7.6 μm (AV3), and the widths varied from 6.4 μm (HEM) to 13 μm (NSPI-05). The lengths of the akinetes, in turn, were between 5.0 μm (Hübel 1987/310) and 9.9 μm (UP16f), and the widths were between 6.6 μm (Hübel 1983/300) and 10.8 μm (NSPI-05). In strains UP16a, UP16f, HKVV, PCC7804, and PCC73104/1, the akinetes were spherical, 7.5 to 9.9 μm long and 8.4 to 10.5 μm wide, ochre to brown in color, and positioned in series (Fig. 1m and n). Strains PCC7804 and Hübel 1983/300 had terminal cells with a conical shape. Gas vesicles were present in all strains except BY1, HEM, UP16a, UP16f, HKVV, PCC73104/1, PCC7804, and Hübel 1983/300 (Table 2). In BY1 and HEM the vesicles had been lost during culturing. Strains Hübel 1988/306a, Hübel 1988/306b, Hübel 1987/310, and Hübel 1987/311 produced nodularin, whereas Hübel 1983/300 did not (Table 2).

Strain	Mean dimensions (SD) of:						Gas vesicles	Nodularin productionb
	Vegetative cells (n = 30)		Heterocytes (n = 30)		Akinetes (n = 15)
	Length	Width	Length	Width	Length	Width
Hübel 1988/306a	3.6(0.7)	7.7(0.5)	5.1(0.8)	8.6(0.9)	8.3(0.8)	10.2(0.9)	+	+
Hübel 1988/306b	3.3(0.6)	10.2(1.2)	4.6(0.8)	11.0(0.8)	6.9(1.4)	9.5(0.8)	+	+
Hübel 1987/310	3.6(0.7)	9.3(1.3)	5.1(0.7)	10.4(1.5)	5.0(0.9)	8.5(1.0)	+	+
Hübel 1987/311	3.4(0.9)	6.8(0.9)	5.2(0.8)	8.6(0.8)	5.3(1.8)	10.1(0.9)	+	+
NSPI-05	3.4(0.7)	12.3(1.0)	5.1(0.7)	13.0(1.3)	5.9(1.2)	10.8(0.8)	+	+
NSOR-12	3.4(0.4)	7.8(0.8)	4.5(0.8)	8.1(0.7)	6.8(1.3)	8.9(0.6)	+	+
GR8b	3.4(0.6)	9.5(1.1)	4.9(0.6)	10.5(0.9)	6.9(1.8)	9.8(0.5)	+	+
BY1	3.9(0.7)	5.7(0.2)	4.8(0.9)	7.0(0.8)	5.7(1.4)	8.0(0.5)	−(+)	+
F81	3.9(0.8)	6.8(0.6)	6.6(0.9)	7.1(1.0)	7.8(0.8)	8.7(1.4)	+	+
AV3	5.0(1.3)	7.3(0.8)	7.6(0.8)	7.9(1.0)	8.1(0.7)	8.4(0.9)	+	+
AV63	3.6(0.8)	8.1(0.5)	5.1(0.8)	8.4(0.7)	8.4(2.5)	9.7(3.3)	+	+
HEM	4.4(1.0)	6.5(0.7)	6.4(0.9)	6.4(0.8)	7.3(1.0)	7.8(0.6)	−(+)	+
UP16a	6.5(2.0)	8.1(0.6)	6.4(0.9)	8.3(0.6)	9.2(1.9)	10.1(1.8)	−	−
UP16f	5.2(1.3)	7.8(0.6)	5.9(0.5)	8.4(0.3)	9.9(1.4)	9.9(1.4)	−	−
HKVV	5.6(1.1)	7.7(0.8)	6.3(0.9)	8.1(0.5)	9.0(1.5)	10.5(1.7)	−	−
PCC73104/1	6.2(0.9)	7.8(1.8)	5.7(0.9)	8.1(0.8)	7.5(1.0)	8.4(0.7)	−	−
Hübel 1983/300	2.7(0.2)	6.4(1.0)	6.4(1.1)	7.4(1.2)	6.1(0.7)	6.6(0.6)	−	−
PCC7804	4.4(1.1)	7.5(0.9)	4.8(0.9)	7.6(0.8)	8.7(2.0)	8.9(0.7)	−	+

Based on the PCA of dimensions of the different kinds of cells, it was not possible to distinguish the nodularin-producing strains from the nontoxic ones (Fig. 2). PCA of cell dimensions of the Nodularia strains, together with the mean cell dimensions from the descriptions of Nodularia species (17), placed the strains NSOR-12, Hübel 1987/311, Hübel 1987/310, GR8b, Hübel 1988/306b, and NSPI-05, and slightly more distantly also strains Hübel 1988/306a and AV63, in a group positioned near the description of N. spumigena (Fig. 2). All of these strains have gas vesicles, produce nodularin, and were isolated from plankton and so were were identified as N. spumigena. Strains Hübel 1983/300 and BY1, on the other hand, were located close to the reference species N. baltica. BY1 fits the description of N. baltica, whereas a lack of gas vesicles, the conical shape of terminal cells, and the benthic origin of Hübel 1983/300 characterized the strain as N. harveyana. Strains UP16a, HKVV, AV3, UP16f, and PCC73104/1 and, more loosely, HEM, F81, and PCC7804 were associated with N. sphaerocarpa (Fig. 2). However, only UP16a, HKVV, UP16f, and PCC 73104/1 carried the distinctive large brownish akinetes in the series typical of N. sphaerocarpa (Fig. 1m and n). Furthermore, UP16a, HKVV, UP16f, and PCC 73104/1 had no gas vesicles and did not produce nodularin and so were determined to be N. sphaerocarpa. Strain PCC7804 also carried the distinctive large brownish akinetes in series and did not have gas vesicles, but it produced nodularin and had smaller vegetative cells and heterocytes than N. sphaerocarpa (Fig. 1r, Table 2). PCC7804 had conical terminal cells and was isolated in a thermal spring. Due to these highly intermediate features between N. sphaerocarpa and N. harveyana, PCC7804 was not identified at the species level. Strains AV3, HEM, and F81 were positioned closest to N. sphaerocarpa by PCA (Fig. 2). However, they are all nodularin producing, and AV3 and F81 had gas vesicles (strain HEM had lost them) and had been isolated in plankton. Once these features are taken into account, strains AV3, HEM, and F81 were determined to be representatives of either N. spumigena or N. baltica with altered cell sizes but will be referred to as Nodularia sp. below.

FIG. 2 First two components of the PCA on the average dimensions of the vegetative cells, heterocytes, and akinetes of Nodularia strains and those of the species descriptions of N. spumigena, N. baltica, N. litorea, N. harveyana, and N. sphaerocarpa as given (more ...)

The ITS1 region of Nodularia and its phylogeny. Amplification with primers 16CITS and 23CITS yielded two ITS1 amplification products from each Nodularia strain. In most strains, the length of the larger product (ITS1-L) was ca. 900 bp and that of the smaller product (ITS1-S) ca. 550 bp (data not shown). Strain Hübel 1983/300 had shorter ITS1-L and ITS1-S fragments of 870 and 470 bp, respectively. Without the flanking 16S and 23S rRNA regions, the ITS1-S sequence lengths varied from 277 bp (strain Hübel 1983/300) to 356 bp (NSPI-05). The alignment contained 360 positions.

The parsimony method gave a tree congruent with the neighbor-joining tree. The neighbor-joining tree consisted of clusters A and B and branches C and D (Fig. 3). Cluster A consisted of nodularin-producing strains and the reference strains from the Baltic Sea plankton and the Australian strain NSPI-05. The strains, which in the present study were determined to be N. spumigena and N. baltica and the reference N. litorea strains, were not distinguished according to the species. Cluster B of the nontoxic strains (HKVV, PCC73104/1, UP16f, and UP16a), which were identified as N. sphaerocarpa, was distant from cluster A containing most nodularin-producing strains and separated with 100% bootstrap support. The ITS1-S sequences of Australian N. spumigena NSOR-12 and the French Nodularia sp. strain PCC7804 were highly divergent from the other strains and also rather distant from one another. A separate branch D carried only the N. harveyana strain Hübel 1983/300 (Fig. 3).

FIG. 3 Neighbor-joining tree on the ITS1-S sequences of the Nodularia isolates with Baltic Sea reference sequences N. spumigena BC Nod-9427 and N. litorea BC Nod-9408 (3) and Nostoc sp. strain PCC7120 as an outgroup. Species designations are based on morphologic (more ...)

The divergence of N. harveyana Hübel 1983/300 was partly due to three separate deletions at positions 127 to 142, 257 to 277, and 285 to 324 of the ITS1-S alignment. At positions 252 to 357 of the ITS1-S alignment, N. spumigena NSOR-12 and Nodularia sp. strain PCC7804 contained a region wherein they were highly similar to one another and distinctly different from the other strains. Sequences from the nontoxic N. sphaerocarpa strains HKVV, PCC73104/1, UP16a, and UP16f were 100% identical. Also, the sequences from the strains N. spumigena GR8b, N. spumigena AV63, and N. baltica BY1 and the reference sequence (AJ224449) from N. litorea BC Nod-9408 were 100% identical. Sequences of the strains AV3, HEM, and F81 had ambiguous nucleotides at positions 196 (G or A), 198 and 199 (A or T), and 280 (G or C) of the alignment, indicating that two ITS1-S copies of equal length, diverging at these positions, were probably present in the genome.

PC-IGS region in Nodularia strains and its phylogeny. The primers PCβF and PCαR gave a single amplification product of ca. 670 bp from each Nodularia strain, and 506 to 508 nucleotides were unambiguously resolved from each strain. The alignment of the Nodularia PC-IGS sequences with the reference sequences and the outgroup contained 512 positions.

The neighbor-joining and parsimony methods produced similar tree topologies. The neighbor-joining tree consisted of clusters A, B, and C and branch D, which carried only N. harveyana Hübel 1983/300 (Fig. 4). The separation of cluster A from clusters B and C was sustained statistically by 86% of bootstrap resamplings. Cluster A contained seven identical sequences from the Baltic Sea strains and two slightly differentiated sequences from Australian N. spumigena strains NSOR-12 and NSPI-05 (Fig. 4). Cluster B contained five strains from the Baltic Sea assigned to three species (N. baltica, N. spumigena, and N. litorea) and the divergent French Nodularia sp. strain PCC7804. Cluster C, on the other hand, consisted of the nontoxic N. sphaerocarpa strains (HKVV, PCC73104/1, UP16f, and UP16a). Strain Hübel 1983/300 N. harveyana in branch D was distinct from all of the other strains.

FIG. 4 Neighbor-joining tree on the phycocyanin operon (PC-IGS) sequences of the Nodularia isolates with Baltic Sea reference sequences BC Nod-9401, BC Nod-9427, and BC Nod-9402 (14) and Nostoc sp. PCC7120 as outgroup. Species designations are based on morphologic (more ...)

There was no separation based on the PC-IGS between the three planktic Nodularia species. The sequences from GR8b N. spumigena, BY1 N. baltica, and reference BC Nod-9401 N. litorea were 100% identical. Strains Hübel 1988/306a and Hübel 1988/306b, which were used as material for the description of species N. baltica, were 97.8% similar to one another. Hübel 1988/306b was 99.8% similar to GR8b N. spumigena and reference BC Nod-9401 N. litorea. Between clusters A and B, e.g., between strains Hübel 1988/306b and Hübel 1987/311, the similarity was 95.4%.

Discrimination of PC-IGS alleles and their detection in the Gulf of Finland of the Baltic Sea. Based on both the ITS1-S and the PC-IGS sequences, the nontoxic N. sphaerocarpa and N. harveyana strains were different from the nodularin-producing strains (Fig. 3 and 4). Consequently, the sequence information was used for discrimination of potentially toxic Nodularia strains from the nontoxic ones. The PC-IGS sequences of the planktic nodularin-producing strains contained a recognition site (5′-GGCC-3′) for HaeIII at positions 118 to 121 and 528 to 531. The enzyme thus produced three fragments (100, 190, and 380 bp) from the PC-IGS amplicons of the planktic nodularin-producing strains. HaeIII digestion of the PC-IGS PCR products from the nontoxic strains HKVV, UP16a, UP16f, PCC73104/1, and Hübel 1983/300 and the toxic strain PCC7804 produced no restriction fragments (data not shown).

During the field survey between 7 and 11 August 2000 in the Gulf of Finland (Table 3) the temperature of the surface water varied between 15 and 18°C. Rough weather conditions, with wind speeds up to 19 m s⁻¹, prevailed between 8 and 9 August, mixing the water masses. The dominating organisms in the net plankton samples were cyanobacteria from the genera Anabaena, Aphanizomenon, and Nodularia; however, a bloom was not observed. Aphanizomenon was dominant at most stations, but Nodularia was always found in samples. At station Jussarö, after the water had been mixed by the winds, most Nodularia trichomes were tightly coiled, tangled, and in a senescent state.

Sampling station	Location (°N, °E)	Sampling date in 2000	No. of sampled filaments	PC-IGS PCR producta (%)	PC-IGS alleleb (%)
					“Nodularin producing”	Unresolved
LL7	59.51, 24.50	Aug 7	75	56(75)	53(95)	3(5)
XV1	60.15, 27.15	Aug 8	60	43(72)	41(95)	2(5)
LL3a	60.04, 26.21	Aug 8	30	22(73)	22(100)
Kasuuni	59.57, 24.56	Aug 9	75	56(75)	56(100)
Jussarö	59.48, 23.33	Aug 9	45	6(13)	6(100)
LL13	59.22, 23.27	Aug 10	60	8(13)	8(100)
Total		All	345	191(55)	186(97)	5(3)

Altogether, 345 Nodularia filaments were picked for PC-IGS PCR and HaeIII digestion analysis (Table 3). The PC-IGS region was successfully amplified from 72 to 75% of the filaments collected at the stations LL7, XV1, LL3a, and Kasuuni, while only 13% of the filaments collected at the stations Jussarö and LL13 provided sufficient PCR yield (Table 3). Altogether, 191 Nodularia filaments were analyzable for the PC-IGS allele. PC-IGS PCR products from those filaments were ca. 670 bp long (Fig. 5A) and similar in size to those from the Nodularia cultures. Some of the filaments failed to provide sufficient PCR yield for the digestion (e.g., Fig. 5A, lane 1). From 186 individual Nodularia filaments (97% of all analyzed filaments), HaeIII digestion of the PC-IGS produced a restriction fragment length pattern similar to that found in toxic cultures of Baltic Sea Nodularia (Fig. 5B, Table 3). The PC-IGS allele present in the nontoxic cultures and PCC7804, which did not contain the restriction site for HaeIII, was not found in any of the analyzed Nodularia filaments. Partly digested PCR products were produced from three filaments from stations LL7 and from two filaments of station XV1, and these samples were considered unresolved (Table 3).

FIG. 5 Steps of the analysis of the PC-IGS allele of Nodularia filaments collected at station Kasuuni in the Gulf of Finland. (A) Agarose gel electrophoresis of PC-IGS PCR products of Nodularia filaments 1 to 15 collected at station Kasuuni (+, positive (more ...)

The nodularin-producing and nontoxic Nodularia strains in culture had different ITS1-S and PC-IGS alleles. Consequently, PC-IGS PCR and HaeIII digestion was used to assign Nodularia filaments collected from plankton in the Gulf of Finland to one of the two types of PC-IGS alleles: the one found in the nodularin-producing cultures or the other present in nontoxic cultures of Nodularia. All 186 Nodularia filaments from the Gulf of Finland contained the PC-IGS allele found in nodularin-producing cultures. The result suggests that all of the examined planktic Nodularia were nodularin producing. At the time of our study, 0.2 to 6.0 mg of nodularin g (dry weight) of net plankton⁻¹ was measured by HPLC at 16 stations in the Gulf of Finland and the northern Baltic Proper (H. Kankaanpää et al., unpublished data), supporting our observations.

Morphologic characteristics were ambiguous in terms of distinguishing the nontoxic Nodularia strains from the nodularin-producing strains. The two most distinct characteristics of the nontoxic strains were the lack of gas vesicles and the large spherical akinetes in series in N. sphaerocarpa strains. However, the presence of gas vesicles is a facultative feature in planktic Nodularia (17), as was demonstrated by two isolates in the present study, and therefore cannot be considered as a feasible marker. The akinetes, although distinct in N. sphaerocarpa strains, are rarely observed in natural populations of Nodularia in Baltic Sea plankton (see, for example, reference 1). An experienced microscopist is likely to distinguish the planktic Nodularia strains from the benthic, periphytic, or soil Nodularia strains. However, PC-IGS PCR from single filaments and subsequent digestion with HaeIII provided an alternative and straightforward method for distinguishing between the planktic and the benthic strains. Furthermore, it allowed the detection in nature of the PC-IGS alleles, which are found in nodularin-producing isolates.

Nontoxic Nodularia strains originally isolated from plankton of the Baltic Sea were identified as benthic species N. sphaerocarpa according to the classical taxonomy (20; this study). The nontoxic N. sphaerocarpa- or N. harveyana-type Nodularia strains were not found in plankton of the Gulf of Finland by the PC-IGS PCR and HaeIII digestion method. These observations imply that the nontoxic N. sphaerocarpa strains, which in three cases were isolated in cultures from the Baltic Sea plankton, occur there only sporadically. Water masses transport loose periphytic and benthic microorganisms from coastal areas to the open sea, and occasionally these organisms can be found in plankton (see, for example, reference 9). Two of the nontoxic N. sphaerocarpa-resembling strains (UP16a and UP16f) were isolated from an open sea population on the same day, while the exact origin of HKVV is uncertain (“Stockholm archipelago water” [40]). Strain PCC73104/1, on the other hand, originates in soil, which is mentioned as a potential habitat for N. sphaerocarpa (17). It is possible that the filaments collected in the Baltic Sea plankton and isolated in culture (HKVV, UP16a, and UP16f) originally grew in the drainage area of the Baltic Sea, either in freshwater or in flooded soil, and that they were transported to the open sea by water. However, the true habitat of N. sphaerocarpa in the Baltic Sea area remains to be determined.

In general, nodularin production seems to be inherent for planktic Nodularia strains, whereas nontoxicity may be perceived as an exception (2, 6, 20, 25, 26, 42; this study). To date, only two nontoxic planktic N. spumigena strains (NSBL-03 and NSBL-05), which originated in an Australian inland lake, have been reported (6). Accordingly, strains with benthic or soil origin are in principle nontoxic (6, 20, 25, 26; this study). N. sphaerocarpa and N. harveyana resembling strain PCC7804, which was isolated in a thermal spring in France in 1966, produces mainly a nodularin variant (5) and is an exception among the benthic strains. To date, however, the number of examined benthic strains is limited and new isolates from benthic, periphytic, and soil habitats should be examined in order to draw further conclusions about their toxicologic and ecological properties.

Hayes and Barker (14) suspected the poor condition of Nodularia filaments to be the reason for ambiguous PCR results in an allele-specific PCR. In our study, the Nodularia population appeared to be senescent at station Jussarö, and only 13% of the samples were successfully amplified. At station LL13, the PCR yield was also poor,but the Nodularia filaments appeared to be in good condition. Therefore, poor quality or possibly a low quantity of the template, as implied by the condition of the filaments, does not seem to explain the variability in PCR success. Drastic sequence differences in the template at the different stations also seem unlikely. A more plausible explanation for the low PCR success could be the effect of storage time since the samples with low PCR yield were the last ones to be processed after almost 2 months of storage at −20°C.

When isolated in culture, cyanobacteria often change morphologically and lose features, which are typical for organisms occurring in natural habitats (see, for example, references 8 and 38). Morphologic features can vary remarkably even among genotypically (16S rRNA) identical cyanobacterial strains (38). The planktic Nodularia usually lose the coiling of the filaments and sometimes also gas vesicles, as do strains BY1 and HEM. The distinct mucilaginous sheaths covering the filaments may be greatly diminished or lost (35). Moreover, the cell dimensions seem to change with time and prevailing growth conditions. The same strains growing under different conditions have differed remarkably in cell dimensions (6, 20; this study). The average lengths and widths of the vegetative cells and heterocytes, respectively, of the strain BY1 varied between the three studies: 3.9 by 5.7 μm and 4.8 by 7.0 μm (this study), 4.4 by 5.1 μm and 5.1 by 6.1 μm (20), and 4.1 by 8.2 μm and 4.5 by 9.1 μm (6). The strains in the present study and that by Lehtimäki et al. (20) were grown in different light conditions. In general, the differences in cell dimensions are likely to be due to differences in growth media, ambient temperatures, and light environments. This plasticity suggests that the growth environment has a pronounced effect on the cell dimensions. Contrasting results have also been reported, since only little variation in cell and heterocyte characteristics under different light, temperature, pH, and salinity conditions was noted by Nordin and Stein (35). However, the changes in the cell dimensions, together with the loss of diacritical features such as gas vesicles, renders species identification of the strains difficult and unreliable. It also impedes the use of morphologic features for the detection of strains with defined physiological characteristics, such as toxin production, in nature.

In the present study, the shift in cell widths was notable, especially in the case of strains Hübel 1988/306a and Hübel 1988/306b, which were used as material for the description of the species N. baltica (17). According to our results, these strains were morphologically closest to the species N. spumigena, indicating rather significant changes in cell dimensions. A shift in the trichome widths of cultured N. baltica and N. litorea toward the intermediate size of N. spumigena was also observed by Komárek et al. (17).

The ecological background of Nodularia strains seemed to define their phylogenetic clustering in the present study, a finding which is in accordance with previous results (20, 26). Based on both the ITS1 and the PC-IGS regions, strain Hübel 1983/300 N. harveyana was clearly different from all of the other strains, thus supporting its genetic divergence from the other Nodularia isolates. N. harveyana inhabits benthic and periphytic environments of saline and brackish waters (17), and consequently its genetic differentiation from ecologically different Nodularia is reasonable. The nontoxic N. sphaerocarpa strains, on the other hand, were distinctly differentiated from the nodularin-producing planktic strains on the basis of the ITS1-S. Based on the PC-IGS region, the N. sphaerocarpa strains clustered separately but were closely related to the planktic strains. The divergence of nontoxic N. sphaerocarpa from the nodularin-producing strains has been shown by 16S rRNA sequences (20, 26) and whole-genome techniques (20). In addition, physiological experiments have indicated differences between nodularin-producing planktic Nodularia and nontoxic N. sphaerocarpa strains (19, 27).

The results presented here indicate no consistent genetic differentiation between the three planktic species of N. baltica, N. spumigena, and N. litorea, which have been described on the basis of morphologic and ultrastructural (i.e., the size and density of the gas vesicles) criteria (17). The strains BY1 N. baltica and GR8b N. spumigena had 100% identical ITS1-S and the PC-IGS sequences (this study), as well as 16S rRNA sequences (20). Clustering of the strains Hübel 1987/306a and Hübel 1987/306b, which were used as material for the description of the species N. baltica (17), with N. spumigena strains was consistent. In addition, reference N. litorea ITS1-S and PC-IGS sequences (3) were identical to sequences from N. spumigena and N. baltica strains. It is highly unlikely that 16S rRNA, which is the most relevant region of the genome in terms of taxonomy and more conserved than the ITS1-S and PC-IGS, would reveal differentiation between the three species if the more variable ITS1-S and PC-IGS regions are identical. Molecular and phenotypic characterizations of Nodularia strains from the Baltic Sea plankton have been unable to reveal a consistent grouping of Nodularia in three different types, which could be considered as three different species (3, 4, 20; this study). Moreover, the most important phenotypic characteristic underlying the species descriptions of N. baltica, N. spumigena, and N. litorea, i.e., the trichome width, seems to be unstable in cultures (3, 20; this study), and intermediate forms have been found in nature as well (7). The view that there is only one planktic Nodularia species in the Baltic Sea has been presented (3, 4). The data shown here, which covers also the morphologically defined species N. baltica, is in agreement with the above view that in the Baltic only one planktic species, N. spumigena, instead of three species, is genetically justified (3, 4). However, it is worth noting that one genotype (16S rRNA) may produce different morphotypes (38), which probably is the case with the planktic Nodularia of the Baltic Sea.

The importance of comparing several regions of the genome in the study of organismal relationships has been stressed (37). The results obtained from the ITS1-S region in the present study were in good agreement with those for the 16S rRNA except for strain NSOR-12 (20, 26). With both regions of the genome, the nontoxic strains were separated from the nodularin-producing planktic strains (20, 26; this study). Grouping of the strains on the basis of the PC-IGS region was slightly different from the grouping based on the ITS1-S (this study) and the 16S rRNA (20). Especially, the planktic strains were differently positioned since they were separated into two clusters in the PC-IGS tree. Of the three regions of the genome, the ITS1-S was most polymorphic, e.g., similarities between the strains PCC73104/1 and GR8b were 91.8% (ITS1-S), 96.7% (PC-IGS), and 99.4% (16S rRNA [20]). However, among the strains from the plankton of the Baltic Sea, the PC-IGS region contained more variation than the ITS1-S.

Within the planktic Nodularia populations of the Baltic Sea, genetic exchange has been shown to most likely occur by allele-specific PCR on ITS1, PC-IGS, and gas vacuole protein A intergenic spacer regions (gvpA-IGS) (4). Already, Barker et al. (3) noted different grouping of the studied Nodularia strains on the basis of the different regions of the genome (ITS1-S, PC-IGS, and gvpA-IGS), and these authors suspected genetic exchange. In the present study, grouping of the planktic strains from the Baltic Sea was also different on the basis of the ITS1-S and the PC-IGS regions. The most striking discrepancy in the clustering on the basis of the ITS1-S and PC-IGS was noted in strains NSOR-12 and PCC7804. Most of the variation accounting for the divergence of NSOR-12 and PCC7804 from one another and the other strains was found in the 3′ end of the ITS1-S, in a region spanning ca. 100 nucleotides. In general, the 3′ end of the ITS1 region is highly variable in cyanobacteria, and it has been shown that much of the ITS1 length differences between cyanobacteria are due to variability in that region (15).

Within the genus Nodularia, toxin production has been restricted mainly to one phylogenetic cluster containing planktic strains (20, 26). In the genus Anabaena, the neurotoxic and the hepatotoxic strains were differentiated on the basis of the 16S rRNA (22). In addition, other physiological traits, such as extreme halotolerance in Halothece spp. (11), differential light adaptations in Prochlorococcus spp. (28), and thermotolerance in Synechococcus spp. (24), have been confined to phylogenetically defined groups. However, in other cyanobacterial genera, e.g., Microcystis (22, 32) and Planktothrix (22), no correlation between phylogenetic clustering and toxin production capability has been found. Nodularin is produced nonribosomally by a complex multienzyme system (33). In the future, detection of the genes responsible for nodularin synthesis (see, for example, reference 25), and studies on their expression are likely to shed more light on the features of nodularin production of Nodularia strains in nature.

In this study, we investigated phenotypic features and two variable regions of the genome, the ITS1-S and the PC-IGS, in 13 nodularin-producing and 5 nontoxic Nodularia strains. We found that the nodularin-producing and nontoxic Nodularia strains carried distinct ITS1-S and PC-IGS alleles. Consequently, we detected the two types of PC-IGS alleles in Nodularia filaments collected from plankton in the Gulf of Finland. Our results suggested that all of the analyzed Nodularia filaments from the Baltic Sea represented the nodularin-producing type. Furthermore, our results from the ITS1-S and PC-IGS regions supported a previous suggestion by Barker et al. (3, 4) that only one planktic Nodularia species is genetically justifiable in the Baltic Sea.

This work was supported by Finnish Academy grants to M.L. (FIBRE, 48008), J.L. (46812), K.H. (FIBRE 39590; 48827), and K.S. (46812) and by Centre for International Mobility and the University of Helsinki funds for M.G.

We are grateful to the donors of the strains, M. Hübel (E.-M.-Arndt University Greifswald, Biological Station Hiddensee, Kloster, Germany) and Susan Blackburn (CSIRO, Australia). We thank Leo Rouhiainen and Anne Rantala for providing intellectual help in the laboratory and Laura Forsström for technical assistance.