Introduction

The genus Phormidium Kützing 1843 has been used to encompass widely diverse morphotypes. For the Oscillatoriacean cyanobacteria, including those discussed in this communication, Maurice Gomont’s Monographie des Oscillatoriées (Gomont 1892) is usually considered the earliest comprehensive source for their taxonomic classification. Gomont’s scheme was followed with minor modifications by Geitler (1932), Frémy (1934), Elenkin (1949), Desikachary (1959), Starmach (1966) and Umezaki (1961).

The criteria traditionally prescribed for classification of genera in Oscillatoriaceae relied predominantly on the quality of external sheaths and colony formation rather than on cellular features. This left considerable freedom for interpretation errors, and opened the possibility that some species may have been described separately and placed in more than one genus. The genus Phormidium Kützing is characterized by thin, hyaline, mucous, partly diffluent or completely dissolved sheaths that cause filaments to stick together in mat-like layers. It includes a large number of species from freshwater and marine environments, with different cell sizes and proportions, and different end cell morphology as well as different degrees of constrictions at cross-walls. Forms with thin, hyaline, firm but externally sticky sheaths, and with filaments, occasionally falsely branched, arranged in upright bundles were placed in the genus Symploca Kützing. The genus Lyngbya Agardhi included forms with well defined, thick sheaths and separated filaments. Difficulties in making generic distinction in view of possible environmental influences on sheath quality prompted Bourrelly (1970) to abolish the genera Phormidium and Symploca, and to transfer all species described within these genera to Lyngbya on the grounds of priority. Drouet (1968) radically revised the systematics of the Oscillatoriaceae, under the a priori assumption that differences among cyanobacterial phenotypes are mostly environmental modifications of a limited number of genotypes. The result was a drastic reduction of the number genera and species considered unacceptable by many authors (see Anagnostidis and Komarek 1985) and some of the taxonomic oversimplifications were experimentally proven to be unjustified (Stam and Venema 1977; see also Castenholz and Waterbury 1989).

Rippka et al. (1979) introduced a classification system of generic assignments for cyanobacteria by accepting or modifying generic definitions derived from the study of field populations (Geitler 1932). The resulting systematic scheme was based on axenic cultures, with limited applicability to morphological and ecological diversity of cyanobacteria observed in nature. Generic distinctions within Oscillatoriaceae were not completely resolved, but left as complexes LPP-groups A and B (for Lyngbya–Phormidium–Plectonema). This system was adopted with minor modifications and the addition of a few, well characterized generic descriptions derived from studies of natural populations in the first and second editions of Bergey’s Manual of Systematic Bacteriology (Castenholz and Waterbury 1989; Castenholz et al. 2001). A major difference between these two editions, however, was the recognition (Castenholz et al. 2001) of the genera Leptolyngbya Anagnostidis and Komarek 1988, Geitlerinema Anagnostidis 1989 and Microcoleus Desmazières 1823 for some members previously assigned to the LPP-group B.

A major revision of cyanobacteria (cyanoprokaryotes) was recently introduced by Anagnostidis and Komárek (1985, 1988) and Komarek and Anagnostidis (2005). The system of the traditional genera complex that included Phormidium, Lyngbya, Plectonema (“LPP-group” of Rippka et al. 1979) and Oscillatoria was revised by introduction of new criteria such as cell proportions and division patterns, occurrence of gas vesicle clusters (aerotopes), motility, and manner of trichome disintegration. The earlier emphasis on sheaths was deprecated. As a consequence, a large number of species (about 440) were transferred from the above four “Geitlerian” genera to 18 newly defined generic entities, most of them acquiring corresponding name changes. Notably, the genus Phormidium was defined by Komárek and Anagnostidis (2005) as having, among other characteristics, radially oriented thylakoids in transversal sections of the cells. Unfortunately, this feature has not yet been proven for the majority of the numerous species described from both nature and culture studies. Leptolyngbya and Geitlerinema were created as a new genus and subgenus of Phormidium, respectively (Anagnostidis and Komarek 1988), to include a large number of oscillatorialean species with trichomes up to 3 μm wide. Narrow-celled species exhibiting little or no motility of the Geitlerian genera: Phormidium, Plectonema and Lyngbya were placed into the newly created genus Leptolyngbya Anagnostidis and Komarek 1988. Species with similar cell dimensions that show rapid gliding motility by rotation and contain peripheral thylakoids were placed into Geitlerinema, elevated to generic status in 1989 (Anagnostidis 1989).

Another problematic genus is Microcoleus Desmazières 1823, used for the taxonomic assignment of filamentous non-heterocystous cyanobacteria that form bundles of filaments united by a common gelatinous sheath (Geitler 1932; Castenholz et al. 2001; Anagnostidis and Komarek 1988; Komarek and Anagnostidis 2005). However, their trichome morphology may correspond to many different genera of the Oscillatoriales. Furthermore, under laboratory culture conditions sheath formation may be altered or lost, as shown for Microcoleus chthonoplastes (Garcia-Pichel et al. 1996) and generic assignment will be impossible, if based on this trait.

Cyanobacterial systematics are currently in a state of confusion. It is recognized among workers in this field that the identification of any particular species based solely on morphology is highly dubious and impractical. Apart from the nomenclatural difficulties mentioned above, cyanobacteria maintained in culture may also loose or no longer express certain characteristic features observed in nature (Palinska et al. 1996; Garcia-Pichel et al. 1996). In order to overcome some of these problems and to permit identification of cyanobacteria at the genetic level, different molecular tools have been applied for their distinction at different taxonomic levels (e.g. Iteman et al. 2002; Neilan et al. 1995; Rajaniemi et al. 2005; Schönhuber et al. 1999; Stam 1980; Turner 1997; Zehr et al. 2003). With the goal of establishing in the future taxonomy of cyanobacteria supported by phylogenetic relationships, a polyphasic approach, combining phenotypic and genotypic characterizations has been recommended (Wilmotte and Herdman 2001; Stackebrandt 2001; Turner 1997; Palinska et al. 1996).

In this study, 30 strains of filamentous, non-heterocystous, cyanobacteria of the Phormidium group coming from different habitats (freshwater and marine, terrestrial crusts, mats, and turfs) and different geographical regions (Europe, Australia, South and North America, Asia, and Antarctica) have been characterized using a polyphasic approach comparing phenotypic, and molecular characteristics. The combination of modern phylogenetic research with classical taxonomic procedures was expected to improve the basis for classification of this group of microorganisms.

Materials and methods

Cyanobacterial material and culture conditions

The cyanobacterial strains listed in Table 1 were cultured in 50 ml Erlenmeyer flasks at ambient light and temperature, except strain Oscillatoria sp. CCMEE 416 which was kept at 12°C. The culture media were either BG11 or ASNIII (Rippka et al. 1979) as specified in Table 1. The strains were not bacteria-free and are available from the appropriate culture collection upon request. To test for the ability of the organisms to perform complementary chromatic adaptation, samples were kept for 6 weeks in red or green light using filters described by Tandeau de Marsac (1977).

Table 1 Compilation of the cyanobacterial strains examined, their isolation site and the medium used for their cultivation
Full size table

Light microscopy, scanning electron microscopy and ultrastructural studies

Light microscopy studies were performed with a Zeiss Axioskop 50 microscope equipped with transmitted light, phase contrast, and Nomarski interference contrast illumination. Photomicrographs were taken using a Nikon Digital Camera DXM 1200. For scanning electron microscopy (SEM), samples were fixed in 4% glutaraldehyde in 0.1 M sodium phosphate buffer (pH 7.2), and dehydrated through a series of ethanol–water solutions. The samples were then critical-point dried in a Balzers Union (CPD 010) apparatus before gold sputtering (SCD 030, Balzers Union). The samples were examined with a ZEISS DSM 940 or a Hitachi S-450 scanning electron microscope operated at 10 or 20 kV and with working distances of 7–9 mm. The light microscopy and SEM studies were used to characterize cell sizes, cell forms, and the degree of constriction at cell junctions.

For ultrastructural studies by transmission electron microscopy (TEM), samples were prepared and embedded according to the procedures described previously (Surosz and Palinska 2004) and examined with a Zeiss EM 109 or EM 902A transmission electron microscope.

Isolation of genomic DNA

Twenty-five milliliter of the cyanobacterial cultures were spun down in a tabletop centrifuge and resuspended in 1 ml TESC (10 mM Tris, 1 mM EDTA, 20 mM NaCl, 2% cetyltrimethylammonium bromide, pH 8.0). After addition of lysozyme (1% final concentration), the samples were incubated at 37°C for 1 h. After 10 cycles of freeze (in liquid nitrogen) and thaw (at 65°C), 5 μl Proteinase K (100  μg/ml) and 90 μl 10% sodium dodecyl sulphate was added and the samples were incubated at 52°C for 150 min. The samples were centrifuged in a microcentrifuge at 12,000g for 5 min and the supernatants were extracted twice with phenol, phenol/chloroform, and chloroform. The DNA was precipitated from the aqueous phase with 0.6 volumes of 2-propanol, washed with 70% ethanol, vacuum dried, and stored in 100μl TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0).

PCR amplification and sequencing

Primers PLG1.1 and PLG2.1 described by Nadeau et al. (2001) were used for amplification of partial 16S rRNA genes. The Intergenic Transcibed Spacer (ITS) between the 16S and 23S rRNA genes was amplified with primers 322 and 340 (Iteman et al. 2000). The reaction volume was 100 μl and contained: 1 × REDTaq PCR Buffer, 200 μM of each deoxynucleotide, 200 μg BSA, 500 ng of each oligonucleotide primer, 5 U of RED Taq DNA polymerase (Sigma-Aldrich), and 1 μl of DNA extract. After an initial denaturation step (4 min at 94°C), 31 incubation cycles followed, each consisting of 1 min at 94°C, 1 min at 52°C, and 1.5 min (8 min at the last cycle) at 72°C. The presence of PCR products was detected by standard agarose gel electrophoresis and ethidium bromide staining. Amplification products were purified with the QIAquick PCR Purification Kit (Qiagen, Hilden, Germany). In cases of ITS amplification the entire PCR product was put on agarose gel, electrophorized, appropriate bands (the large and small ITS bands) excised and DNA extracted using the QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany).

DNA samples were directly sequenced in both directions by a commercial sequencing laboratory. The primers were the same as for amplification combined with F primer Phor-580 and R primer Phor-710 designed in this study for 16S rRNA as well as primers 323 (Iteman et al. 2000) and ITS30o and tRNA-F designed here for ITS fragments. The sequences of the primers are given in Table 2. The GenBank accession numbers of sequences determined in this study are given in Table 3. Sequence similarities were calculated online with ClustalW (Higgins et al. 1994) at http://www.ebi.ac.uk/clustalw/index.html.

Table 2 Primers used for amplification and sequencing of 16S rDNA and ITS
Full size table
Table 3 Molecular characteristics of the examined cyanobacterial strains
Full size table

Phylogenetic analysis

Phylogenetic analysis of the 16S rRNA gene was performed with the help of the software ARB (Ludwig et al. 2004, available online at http://www.arb-home.de) using the “ssu_jan04_corr_opt.arb" database. A fragment of approximately 925 nt (corresponding to Escherichia coli K12 16S rRNA residues 269–1215 and to residues 1003153–1004077 of the genome of Anabaena variabilis ATCC 29413, GenBank acc. no. NC007413) was aligned to the 16S rRNA alignment of the ARB database with the integrated aligner of the software using other cyanobacteria as reference organisms. The alignment was checked by eye and corrected manually. The aligned sequences were included into the cyanobacterial sub-tree (containing 1315 sequences) of the Parsimony tree “tree_1000_jan05” of the ARB database using the cyanobacterial filter with the “Parsimony (Quick Add Marked)” function of the software. Additionally, our sequences and selected GenBank entries were aligned online with clustalW at default configuration and the 925 nt fragment given above was used for the calculation of a Neighbor-Joining tree (Jukes and Cantor distance estimation, all sequence position considered, but insertions and deletions not taken into account, 2000 bootstrap replicates) with the software TREECON for Windows 1.3b (Van de Peer and De Wachter 1994). E. coli was used as outgroup and positioned with the “Single sequence (forced)” method.

Pigment extraction and spectroscopy

Cyanobacterial cultures were centrifuged for 5 min at 12,000g in a microcentrifuge and resuspended in 50 mM Tris/HCl (pH 8.0) with 250 mM NaCl and 10 mM EDTA. Cells were disrupted by sonication for 5–10 min on ice with a Branson Sonifier Cell Disruptor B15. Triton X-100 was added to a final concentration of 0.5% and the samples were incubated at 28°C for 30 min. After centrifugation at 12,000g for 5 min, absorbance spectra of the supernatant were recorded with a Hitachi U-3000 spectrophotometer from 400 to 750 nm.

Results

Molecular phylogeny

We analysed a 925 bp part of the 16S rRNA gene and the entire ITS of 30 filamentous, non-heterocysteous cyanobacterial strains. PCR amplification of the internal transcribed spacer (ITS) between the 16S rRNA and 23S rRNA genes yielded one or three products upon electrophoresis in agarose gels. In the latter case, two PCR products represented true ITS regions of different sizes, while the third one was probably a heteroduplex as described by Iteman et al. (2000). When rerun on an agarose gel this band yielded also the two other bands. However, no efforts were made to examine this in more detail. There were five strains with three products: Phormidium foveolarum SAG 1462-1 (no. 14, cluster 5), Leptolyngbya sp. PCC 73110 (no. 15, cluster 5), Phormidium animale CCALA 140 (no. 24, cluster 9), Phormidium cf. nigrum CCALA 147 (no. 29, cluster 11), and Phormidium unicatum SAG 81.79 (no. 30, cluster 11). The lengths of the spacers identified and sequenced are given in Table 3. The ITS of all strains giving a single PCR band upon electrophoresis harbored tRNAIle and tRNAAla genes, except the one of Phormidium tergestinum CCALA 155 (no. 28, cluster 11), whose ITS contained only the tRNAIle gene. In the case of multiple PCR products, the longer spacers also contained the two tRNA genes but both were absent from the shorter spacers, except in strain Phormidium animale CCALA 140 (no. 24, cluster 9) where the tRNAIle gene was still present (Table 3).

Using the ARB software, we added our 16S rDNA sequences to a cyanobacterial Parsimony tree containing 1,315 entries. The strains we investigated did not form a separate cluster but were scattered over multiple branches of the resulting tree. Some strains as Phormidium fragile (no. 1 in Tables 1, 3 and 4, cluster 1 in Figs. 1 and 2) and Phormidium spp. OL S4 and OL 81 (nos. 18 and 19, cluster 7) grouped with unicellular cyanobacteria of the genus Synechococcus. Due to its large size the entire phylogenetic tree cannot be shown here. A partial tree with selected strains is given in Fig. 1. The 16S rDNA sequence data for these strains were also used to calculate a Neighbor-Joining tree with the TREECON software (Fig. 2). The topology of the inferred tree differed from that determined by maximum parsimony analysis, especially in the partitions defined by the deeper branches where the bootstrap values in the Neighbor-Joining tree are small. However, the two trees do share identical clusters that are strongly supported by high bootstrap values. The clusters containing strains that were examined in this study are indicated with brackets and numbers in Figs. 1 and 2. Although Oscillatoria sancta PCC 7515 clusters with nos. 25–27 (cluster 10) in the Neighbor-Joining tree with fairly high bootstrap support (87%), it was not included in the demarcation of cluster 10 in Figs. 1 and 2 since it was closer to cluster 8 in the Parsimony tree (Fig. 1).

Table 4 Morphological, ultrastructual and pigment characteristics of the examined cyanobacterial strains
Full size table
Fig. 1
figure 1

Maximum parsimony tree based on partial 16S rRNA gene sequences (925 bp) as inferred with the ARB software, see Material and methods for details (Ludwig et al. 2004). GenBank accession numbers are in square brackets. The strains examined in this study are marked with asterisks. Bar substitutions per site

Full size image
Fig. 2
figure 2

Neighbor-Joining tree based on partial 16S rRNA gene sequence (925 bp) as inferred with the TREECON software (Van de Peer and De Wachter 1994). GenBank accession numbers are in square brackets. The strains examined in this study are marked with asterisks. The numbers at the nodes show bootstrap support as percentages based on 2,000 resampled data matrices. Only bootstrap values greater than 50% are given. Bar substitutions per site

Full size image

Some strains showed identical 16S rDNA sequences, as in the case of Phormidium sp. strains from the North Sea OL M10, OL S3, OL 05, OL S5, and OL 32 (nos. 2–6, cluster 2), Phormidium foveolarum SAG 1462-1 and Leptolyngbya sp. PCC 73110 (nos. 14 and 15, cluster 5), Phormidium sp. OL S4 and OL 81 (nos. 18 and 19 cluster 7), Phormidium sp. strains OL S12, OL S6, OL 75, and OL “sphere" (nos. 20–23, cluster 8). In most of these cases also the ITS sequences were identical. In the case of the strains of cluster 8 the ITS sequences of Phormidium sp. OL 75 (no. 22) and Phormidium sp. OL “sphere" (no. 23) were identical, while those of Phormidium sp. OL S12 (no. 20) and Phormidium sp. OL S6 (no. 21) differed in only one nucleotide. However, there were seven, respectively eight, nucleotide differences between both groups.

At least 99% identity were found between 16S rDNA sequences of Phormidium ectocarpi PCC 7375 and Phormidium persicinum SAG 80.79 (no. 7, cluster 2), Oscillatoria priestleyi UTCC 4762 and Phormidium animale CCALA 140 (no. 24, cluster 9), Oscilatoria sp. (acc. no. AF263333) and Phormidium autumnale CCAP 1462/10 (no. 25, cluster 10), and Oscilatoria sp. PCC 6407 and Phormidium inundatum SAG 79.79 (nos. 26 and 27, cluster 10). In the latter case the strains showed identities of 99.8% for the 16S rRNA gene and 91.4% identity for the less conserved ITS sequence.

Generally, the degrees of pairwise identity within single clusters varied between 93.4 and 100% for the rDNA and between 32.9 and 100% for the ITS sequences (Table 3). The similarity between rDNA sequences of strains from different clusters was 85.6–94.2% identity. For the ITS sequences the corresponding values were between 28.7 and 70.0%. Shorter and longer spacers from the same strain differed not only in the presence or absence of tRNA genes (Table 3) but also in their nucleotide sequences. They showed a degree of identity of 47.0% for strains Phormidium foveolarum SAG 1462-1 and Leptolyngbya sp. PCC 73110 (nos. 14 and 15, cluster 5), 72.8% for Phormidium animale CCALA 140 (cluster 9), 49.6% for Phormidium cf. nigrum CCALA 147, and 43.6% for Phormidium unicatum SAG 81.79 (nos. 29 and 30, cluster 11). An example is shown in Fig. 3.

Fig. 3
figure 3

Pairwise alignment with ClustalW of longer and shorter ITS and adjacent rDNA regions of Phormidium cf. nigrum CCALA 147. rDNA and tRNA genes are in shaded areas. The ITS regions show 49.6% identity

Full size image

Morphology

The morphology of the strains was analyzed by light microscopy and scanning electron microscopy (SEM). Some examples of typical morphologies are shown in Fig. 4, and the results of these investigations are compiled in Table 4. The cell sizes were quite variable. Filament diameters ranged from less than 1 μm to more than 6 μm. Thin filaments were typical for organisms of clusters 1–5 and 7 (cf. Fig. 4a). All had cell diameters of less than 2 μm. Strains of clusters 6 and 8–11 had filament diameters of more than 2 μm. However, for clusters 1, 4, 6 and 9 only one strain was examined. Therefore, it remains unclear if this size range is typical for all strains in these clusters. The broadest filaments (4.5–6.3 μm) were found in cluster 10 (cf. Fig. 4c).

Fig. 4
figure 4

Scanning electron micrographs of four Phormidium strains representing different trichome diameters, cell shapes, and degrees of cross-wall constriction between the cells. a Phormidium sp. OL 05 as an example of a strain with thin filaments (less then 2 μm and cylindrical cells with pronounced constrictions between them, b Phormidium foveolarum SAG 1462-1 with filaments of intermediate diameter (2–4 μm) and approximately isodiametric cells, c Phormidium autumnale CCAP 1462/10 as an example of broad trichomes (>4.5 μm) with disc-shaped cells lacking cross-wall constrictions, and d Phormidium sp. OL 75 as an example of a strain with shallow cross-wall constrictions. The arrow in (c) indicates the calyptra

Full size image

As summarized in Table 4, the cells of the filaments were either cylindrical (Fig. 4a, a cell length/diameter ratio >1), slightly longer than wide to isodiametric (length/diameter ≥ 1) isodiametric (Fig. 4b, length/diameter ≈1), or slightly shorter than wide to disc-shaped (Fig. 4C, length/diameter <1). The cell shape usually was correlated with the filament diameter. Narrow filaments (diameter <2 μm) contained isodiametric to cylindrical cells, while broad filaments (diameter >3 μm) contained isodiametric to disc-shaped cells. In filaments with a diameter between 2 and 3 μm either of the two cell types occurred. While cells of strains in cluster 8 were isodiametric to cylindrical, those of Leptolyngbya sp. CCALA 094 (no. 16, cluster 6), Phormidium animale CCALA 140 (no. 24, cluster 9) and Phormidium unicatum SAG 81.79 (no. 30, cluster 11) were shorter than wide to disc-shaped (Table 4).

Moreover, in one strain, Phormidium sp. AA (no. 8, cluster 3), cell shape was quite variable (length/diameter <1 to>1, Table 4). Constrictions between the cells might be missing or be present to different degrees. Distinct constrictions were found in the strains of clusters 2, 3, 5, and 7 and in Leptolyngbya sp. PCC 8936 (no. 17). Strains of clusters 1 and 4, and three of the four strains of cluster 8, had very shallow constrictions, only clearly visible in some filaments or parts of the same filament (cf. Fig. 4d). Strains of clusters 9–11 had no constrictions at all (cf. Fig. 4c). A calyptra, a thick-ended cap on the outer cell wall of an apical cell, was only found with one strain, Phormidium autumnale CCAP 1462/10 (no. 25, cluster 10) (Figs. 4c, 5h).

Fig. 5
figure 5

Transmission electron micrographs of some of the examined cyanobacteria. Leptolyngbya foveolarum CCALA 081 (a) and Leptolyngbya sp. CCALA 094 (b) in longitudinal sections, Phormidium inundatum SAG 79.79 (c), Phormidium tergestinum CCALA 155 (d) and Phormidium animale CCALA 140 (e) in cross-section (in Phormidium autumnale not exactly perpendicular to the longitudinal axis), Phormidium cf. nigrum CCALA 147 (f) and Phormidium autumnale CCAP 1462/10 (g, h) in longitudinal section. a and c show two views of peripheral thylakoids, in b the thylakoids are peripheral, but undulating. d and f show fascicular thylakoids in cross and longitudinal section. g demonstrate the stacked thylakoids and h the calyptra of Phormidium autumnale. C carboxysome, CW cell wall, L lipid granulum, MS mucilaginous sheath, OM outer membrane, PG peptidoglycan layer, PM plasma membrane, Pp polyphosphate granulum, T thylakoid

Full size image

Ultrastructure

In our TEM investigation we focused on the number and orientation of thylakoid membranes, which are the most conspicuous structures within the cells. The results are compiled in Table 4. In the majority of strains (clusters 1–8), the thylakoids were arranged peripherally, running parallel to the cytoplasmic membrane (Fig. 5a, c). In Leptolyngbya sp. CCALA 094 (no. 16, cluster 6) and Phormidium uncinatum SAG 81.79 (no. 30, cluster 11), the thylakoids were undulating at the periphery of the cells, giving rise to a more irregular pattern (Fig. 5b). The number of thylakoids in strains where they were arranged peripherally was usually no more than six. Only in Phormidium sp. OL S6 and OL 75 (no. 21 and 22, cluster 8), Phormidium inundatum SAG 79.79 (no. 27, cluster 10) and Phormidium unicatum SAG 81.79 (no. 30, cluster 11) thylakoid number was higher, between 8 and 12 (Table 4).

Highly irregular thylakoid arrangements were found only in a few strains. Phormidium tergestinum CCALA 155 (no. 28, cluster 11) and Phormidium cf. nigrum CCALA 147 (no. 29, cluster 11) showed a fascicular thylakoid pattern (sensu Casamatta et al. 2005) where the thylakoids are arranged in fascicles running parallel to the longitudinal axis of the cell (Fig. 5d, f). In Phormidium animale CCALA 140 (no. 24, cluster 9) the thylakoids also ran parallel to the longitudinal cell axis but were arranged perpendicularly to the cell wall, giving rise to a radial pattern in transversal sections (Fig. 5e). In Phormidium autumnale CCAP 1462/10 (no. 25, cluster 10) the thylakoid membranes formed stacks lying irregularly within the cells (Fig. 5g). The number of thylakoids in these strains could not be determined exactly, but it was significantly higher than in cyanobacteria with peripheral thylakoids. All strains with irregular thylakoid patterns are members of clusters 9, 10 and 11.

Presence of phycoerythrin

Phycoerythrin was found only in five strains investigated (Table 4). It was a dominating pigment in Phormidium persicinum SAG 80.79 (no. 7, cluster 2), a strain that is colored bright red. In the other four strains it was a minor pigment only detectable in the absorbance spectrum. In two strains, Leptolyngbya sp. CCALA 094 (no, 16, cluster 6) and Phormidium autumnale CCAP 1462/10 (no. 25, cluster 10), the phycoerythrin/phycocyanin ratio was enhanced under green light (Table 4) which enhances the expression of phycoerythrin in the course of complementary chromatic adaptation (Tandeau de Marsac 1977).

Discussion

The strains here classified as Phormidium sensu Geitler 1932 were found to be members of several different phylogenetic clusters, which also contain strains, assigned to the genera Leptolyngbya, Oscillatoria, Geitlerinema, Microcoleus, Pseudanabaena or Plectonema (Figs. 1, 2). These seemingly surprising relationships are largely the result of the divergent generic nomenclature employed by different authors, or culture collection curators, even for morphologically similar filamentous non-heterocystous cyanobacteria. This is the consequence of the simultaneous use of different taxonomic guides for identification of the organisms. Taxonomic treatment of cyanobacteria, including formal description of new taxa, can presently be carried out under the aegis of either the Botanical or Bacteriological Codes of Nomenclature (Castenholz 1989; Palinska et al. 2006). However, the rules of these two codes are quite different and, when applied to the same group of organisms, confusions are unavoidable (Oren 2004). Furthermore, choice of inappropriate generic or specific assignments may result from naming or renaming organisms after they have been maintained in culture, where certain properties of determinative value may no longer be expressed, or have been lost due to mutations (Palinska et al. 1996; Otsuka et al. 2001; Lyra et al. 2001). Two strains assigned to the same nomenspecies, Phormidium autumnale CCAP 1462/10 (no. 25, cluster 10) and Phormidium autumnale UTEX 1580 (cluster 11), were found in different phylogenetic clusters (clusters 10 and 11, respectively). This suggests that morphologically similar strains may differ genetically. Alternatively, a genuine misidentification or strain-mixup cannot be excluded. Such a possibility was discovered for Pseudanabaena tremula UTCC 471 (cluster 4) which previously was classified as Phormidium autumnale (and is still found under this name in the UTCC catalog). The incorrect classification of the strain was noticed by Casamatta et al. (2005), who renamed the strain based on morphological characteristics and 16S rDNA sequence analyses. Further investigations with additional Phormidium autumnale strains are required to explore the genetic diversity of this taxon in more detail.

Phormidium fragile OL O3, and Phormidium sp. OL S4 and OL 81 (clusters 1 and 7, respectively) were found to group with two different strains of Synechococcus (Figs. 1, 2). The affiliation of these, and some other, unicellular cyanobacteria with thin oscillatorian representatives was already noticed by other authors (e.g. Turner 1997; Honda et al. 1999; Robertson et al. 2001) Ishida et al. 2001; Wilmotte and Herdman 2001; Taton et al. 2003; Casamatta et al. 2005). These relationships, as suggested by Honda et al. (1999), probably reflect the convergent evolution of cellular organization. Furthermore, they demonstrate that neither the Chroococcales nor the Oscillatoriales are monophyletic.

The phylogenetic clustering of strains assigned to the genus Phormidium with other members of the Oscillatoriales, including strains assigned to the genera Leptolyngbya and Oscillatoria has also been shown previously (Turner 1997; Ishida et al. 2001; Lee and Bae 2001; Litvaitis 2002; Ceschi-Berrini et al. 2004). Similarly, the high level of genetic diversity within the genus Phormidium was demonstrated (Baker et al. 2001) by analysis of the rpoC1 gene, in agreement with Teneva et al. (2005), who examined the cpB-IGS-cpA locus. In the phylogenetic tree generated by the latter study, Phormidium strains even clustered together with members of the Nostocales. However, this conclusion is contradicted by all other reports demonstrating that heterocystous cyanobacteria are monophyletic (e.g. Wilmotte 1994; Wilmotte and Herdman 2001; Lyra et al. 2001; Henson et al. 2004). In our study, the heterocystous species Anabaena cylindrica, Nostoc commune and Fischerella muscicola form a well-supported (89% bootstrap support) separate cluster.

A 16S rRNA sequence similarity of 95% has been suggested as a threshold for a congeneric bacterial genus (Stackebrandt and Goebel 1994). In this study, this criterion was met for strains of clusters 3, 4, 5, 7, 8 and 9 (Table 3) and the subcluster Plectonema sp.-Synechococcus sp. PCC 7335-Phormidium fragile OL 03 of cluster 1, the subclusters Phormidium sp. strains: OL M10, OL S3, OL 05, OL S5, OL 32-Phormidium sp. [BD061304] and Leptolyngbya ectocarpi PCC 7375-Phormidium persicinum SAG 80.79 of cluster 2, the subclusters Microcoleus vaginatus PCC 9802-Phormidium autumnale CCAP 1462/10-Oscillatoria sp. [AF263333] and Oscillatoria sp. PCC 6407-Phormidium inundatum SAG 79.79 of cluster 10, and the subcluster Phormidium autumnale UTEX 1580-Phormidium tergestinum CCALA 155 of cluster 11. Therefore the Phormidium strains analyzed are representatives of at least 10 different genera.

Cyanobacterial intergenic transcribed spacer (ITS) regions investigated earlier vary in size from 354 to 545 nucleotides (Iteman et al. 2000; Otsuka et al. 2001; Boyer et al. 2002; Laamanen et al. 2002). However, Laloui et al (2002) and Rocap et al. (2003) reported longer ITS, up to more than 1,000 bp. The 30 cyanobacteria examined here contain ITS regions of different length (442–694 nt, Table 3), but length distinctions do not correlate with specific clusters or subclusters inferred from 16S rDNA sequences. For example, the ITS regions of strain Leptolyngbya PCC 73110 (no. 15, cluster 5) and most of the strains of cluster 2 (Table 3) shared the same length. Some of the strains (see clusters 5, 9 and 11, Table 3) were shown to have at least two ITS that differed in length and tRNA content, demonstrating the heterogeneity of their rrn operons. Multiple ITS regions with different lengths were also found in Nostoc PCC 7120, where the longer and shorter ITS regions differed in the presence and absence, respectively, of the two tRNA genes (Iteman et al. 2000). In contrast, the two rrn operons of Synechocystis PCC 6803 contain ITS regions of identical length and tRNA content, as can be deduced from the genome sequence (Kaneko et al. 1996). Although in parts conserved, some regions in the ITS are highly variable. This makes a proper alignment and the deduction of phylogenetic relationships difficult for strains that are not closely related, and thus was not been attempted here. However, for some strains with identical or nearly identical 16S rDNA sequences, a higher resolution of discrimination within some of the clusters (see Table 3) was achieved based on ITS sequence analyses.

Our results show that the presence of phycoerythrin is not correlated with the strains’ position in the phylogenetic trees. Phycoerythrin-producing representatives occur in different clusters together with strains that lack phycoerythrin (Table 3). The deeply red-colored Phormidium persicinum SAG 80.79 (no. 7, cluster 2) and the phycoerythrin-rich Leptolyngbya ectocarpi PCC 7375 are on the same branch (Figs. 1, 2) and share a very high 16S rDNA similarity (99.7% identity). Consequently, they may in fact represent the same taxon kept in different culture collections under different names. With the latter exception where the high phycoerythrin content may be a characteristic feature of a specific genotype (Table 3), the presence of phycoerythrin appears unsuitable as a systematic marker. Similar conclusions were drawn by Otsuka et al. (2001) for phycoerythrin-containing strains of Microcystis. As for phycoerythrin, there is no correlation between the clustering of the strains and their geographic origin (see Table 1). The strains of cluster 3, for instance, came from Israel and Antarctica, those of cluster 5 from the European Alps, Australia or India, and Oscillatoria sp. PCC 6407 and Phormidium inundatum SAG 79.79 (nos. 26 and 27, cluster 10) with almost identical 16S rDNA sequence came from California and France, respectively. This is in agreement with findings of a cosmopolitan distribution of many cyanobacterial species (e.g. Mullins et al. 1995; Wilmotte et al. 1997; Garcia-Pichel et al. 1996).

On the other hand, it is noticeable that strains identical at the 16S rRNA loci were repeatedly isolated from the same geographic region: of the 12 strains from the Oldenburg collection isolated at the North Sea coast of northwestern Germany, 11 were assignable to only three genotypes, two of which (clusters 2 and 8, Table 3) are represented by 4–5 independent isolates. These Phormidium types might be numerically dominant in this area or, more likely, are easily selected for by the growth conditions employed.

In contrast to geographic origin, there is a good correlation between the original environment of the strains and their groupings/subgroupings in the phylogenetic trees, though organisms from similar habitats may also occur in different phylogenetic clusters. For example, all strains of cluster 5 are from terrestrial origin, and the OL strains of cluster 1, 2, 7 and 8, as well as their relatives, originate from marine or highly saline habitats. The OL strains show a wide range of salt tolerance with a growth optimum between 16 and 34 psu, i.e. in brackish water. The other clusters contain either terrestrial organisms or a mixture of terrestrial and freshwater strains. Terrestrial habitats can vary greatly in the amount of available water; e.g. a rock might be moistened regularly by stream or rainwater. An example of a strain from an extreme terrestrial habitat is Phormidium sp. AA (no. 8, cluster 3) that comes from the particularly dry Nizzara desert. At first glance, it may seem surprising that this strain clusters together with Oscillatoria sp. CCMEE 416 (no. 9) from Antarctica. However, the ice-free regions of Antarctica are deserts as well. A similar close phylogenetic relationship was found by Casamatta et al. (2005) for Antarctic Microcoleus acremannii and Microcoleus vaginatus from desert soil. These authors point out that both organisms share a similar habitat that is characterized by long periods of desiccation and high levels of ultraviolet radiation.

In agreement with their 16S rDNA sequence similarity, strains within a given phylogenetic cluster are morphologically rather homogeneous. However, similar morpho- and ecotypes may also occur in different clusters (e.g. clusters 2 and 7), demonstrating that even the combination of morphological and ecological data does not allow precise identification.

Although the number of thylakoids varied (Table 4), all strains of clusters 1–8 share peripheral thylakoids and thus can not be distinguished based on their cellular ultrastructure. A radial thylakoid arrangement, considered typical for members of the genus Phormidium sensu Anagnostidis and Komarek (1988), was observed in only one strain, Phormidium animale CCALA 140 (no. 24, cluster 9). If this type of thylakoid arrangement is a reliable taxonomic feature, one would predict that the close relative of this strain, Oscillatoria pristleyi UTCC 4762 (Figs. 1, 2), should also have radial thylakoids. Casamatta et al. (2005) found congruence between inferred phylogenetic relationships among strains and their thylakoid arrangement. In our study this is only partly true, since clusters 10 and 11 contain members with both peripheral and irregularly arranged thylakoids (Table 4). However, these clusters are not very well supported (75 and 71% bootstrap support, respectively).

Our results reemphasize the polyphyletic nature of the Oscillatoriales. They show in particular that, even if ultrastructural features are included, the exclusive use of phenotypic traits does not permit confident identification of oscillatorian cyanobacteria and their arrangements in a hierarchical order. Based on our phylogenetic analyses, the strains of Phormidium examined are representatives of more than 10 generic entities. Furthermore, some of the phylogenetic subclusters undoubtedly represent distinct species, for which we could propose new specific epithets. However, Fox et al. (1992) found that Bacillus strains with more than 99% 16S rRNA similarity had less than 70% DNA–DNA hybridization values, the most commonly accepted benchmark used to distinguish between bacterial species (Stackebrandt and Goebel 1994). Furthermore, the nomenclatural problems highlighted in this study evidently diminish the value of the phylogenetic trees and the possibilities of deducing meaningful systematic and evolutionary relationships. Therefore, we feel that nomenclatural changes and drastic taxonomic revisions should await acquisition of further knowledge. More phenotypic and molecular data from additional organisms and multiple genes are needed to confirm and refine the systematic relationships so far revealed within this group of cyanobacteria. Finally, strain discrimination should be attempted using sequence data of genes other than that of the small ribosomal subunit, which would provide greater phylogenetic resolution among closely related species.

Our phylogenetic trees emphasize the polyphyletic nature of the Oscillatoriales and the doubtful identification of many strains existing in culture collections.