Phormidium is a cyanobacterium that can form expansive benthic mats on the substrata of lakes and rivers [1,2,3]. Benthic proliferations dominated by Phormidium are of concern because some strains produce the potent neurotoxic compounds anatoxin-a, dihydroanatoxin-a, homoanatoxin-a, and dihydrohomoanatoxin-a, hereafter collectively referred to as anatoxins [1,2,4]. Anatoxins are powerful neuromuscular blocking agents that act through the nicotinic acetylcholine receptor. They bind to acetylcholine receptors triggering muscle contraction or other neural signal responses. They cannot be broken down by acetylcholine esterase and cause over-stimulation. When ingested, these toxins can cause death via asphyxia from respiratory paralysis [5]. Consumption of anatoxin-containing Phormidium mats has resulted in numerous dog poisonings and fatalities worldwide [1,4,6,7].

Studies of benthic Phormidium mats in rivers across New Zealand have shown a wide variability in anatoxin concentrations [2,4,8,9]. Samples taken from individual sites (100 m² in area) on the same day can also show a wide divergence in anatoxin content. Both anatoxin and non-anatoxin containing samples were identified and these were sometimes spaced less than 1 m apart [10]. Studies have also noted large differences in toxin concentrations from week to week, and within rivers on the same day [8,9,10].

Cyanobacterial toxin dynamics have been examined in planktonic bloom forming cyanobacteria for many decades with similar extreme variability in toxin concentrations recorded. This variability has been observed spatially and temporally in both long-term and fine scale investigations [11,12]. Within toxic planktonic blooms of anatoxin-a-producing Aphanizomenon issatschenkoi [13], saxitoxin-producing Anabaena [14], and microcystin-producing Microcystis [15,16], or Planktothrix [17], it has now been demonstrated that this variability is largely due to the co-occurrence of toxic and nontoxic genotypes. Differences in toxin content can be explained by the relative dominance of toxin-producing or non-producing strains. Culture based studies of strains isolated from blooms have also shown that toxic genotypes vary in the amounts of toxin and specific congeners produced [14,18]. Despite an increasing number of reports of toxic benthic cyanobacteria [19,20,21,22,23] there is only very limited information available on the co-occurrence of toxic and non-toxic strains [2,10,24]. There is no information on variation in anatoxin content among toxic strains isolated from the same benthic mat.

In this study we investigated the toxin producing abilities of 30 Phormidium strains isolated from four anatoxin-containing mats collected from three different sites. The anatoxin production potential of each strain was assessed using a PCR targeting the anaF gene. Strains were grown to stationary phase and their anatoxin content determined using liquid chromatography-mass spectrometry (LC-MS). Each strain was characterized using morphological and molecular (16S rRNA gene sequences) techniques.

A total of 30 Phormidium strains were established from single filaments. Ten strains (CYN103–CYN112) were isolated from the Waimakariri River mat, four (CYN113, CYN115–117) and eight strains (CYN126–133) were isolated from mat 1 and mat 2 from Hutt River, Site 1 respectively, and eight strains (CYN118–125) were isolated from a mat collected at Hutt River, Site 2.

Anatoxin-a and dihydroanatoxin-a were detected in 18 of the 30 samples. One of these strains (CYN112), isolated from the Waimakariri River, also produced homoanatoxin-a (Table 1). All of the strains isolated from mat 1 from the Hutt River, Site 1, produced anatoxins. In contrast not all of the isolates from the other mats produced anatoxins; nine out of ten isolated from the Waimakariri River, one out of eight from the Hutt River, Site 2, and six out of eight from mat 2 from the Hutt River, Site 1 did not contain anatoxins (Table 1).

Among the anatoxin producing strains isolated, the concentration of anatoxins produced varied and there was no relationship between the concentration produced and the original source. For example, total anatoxins produced by the isolates from the Waimakariri River varied from 10.06 to 211.83 mg kg⁻¹ and from 4.75 to 115.42 mg kg⁻¹ dry weight for isolates from mat 1, Hutt River, Site 1. Dihydroanatoxin-a concentrations were always higher than anatoxin-a (Table 1).

All strains that were found to produce anatoxins via LC-MS also tested positive by PCR for a segment of the anaF gene (Table 1). Gene segments (404 bp) from strains CYN103–107 and 109–112 were sequenced and were identical. Using Megablast, these segments shared 99% sequence homogeneity with an anaF gene segment from strain CYN53 (JX088093), a P. autumnale strain isolated from the Rangitaiki River (North Island, New Zealand; [2]). The next closest sequence similarities were 94% with Oscillatoria PCC 6506 (FJ477836), and 91% with Anabaena sp. (JF803645) and A. flos-aquae (AY210784).

Partial 16S rRNA gene sequences (1340 bp) were obtained for all strains. The strains which tested positive for anatoxins (CYN103–107, 109–113, 115–118, 126–128 and 131-133) were identical over this region (Figure 1). These strains were also identical to CYN53 (JX088083). These segments had a 99% sequence homology with Tychonema sp. K27 (GQ324965), T. bourrellyi CCAP (AB045897) and T. tenue SAG 4.82 (GQ324973). The sequences from all other strains isolated in this study, except CYN108, were identical, and were 17 nucleotides different from the anatoxin positive strains. When submitted to Megablast the closest matches (99%) for these strains were: Phormidium sp. KU003 (AB094351), Microcoleus vaginatus SEV1-KK3, CJI-U2-KK1, PCC 9802 (EF654076, EF654078, AF284803), P. autumnale SAG 35.90, Ant-Ph68, Arct-Ph5 16S, SAG 78.79 (EF654081, DQ493874, DQ493873, EF654084), Tychonema sp. K27 (GQ324965), T. bourrellyi CCAP (AB045897) and T. tenue SAG 4.82 (GQ324973). Strain CYN108 differed from the other anatoxin negative strains by four nucleotides and an insertion of 10 nucleotides. Megablast analysis returned the same results as obtained for the other anatoxin negative strains.

The 16S rRNA gene sequences were used to construct a phylogenetic tree. The anatoxin and non-anatoxin producing strains isolated in this study clustered in a single clade with 100% bootstrap support (Figure 1). The anatoxin-producing strains clustered most closely with Tychonema species, whilst the non-toxic strains (CYN108, 119–125, 129, 130) formed their own sub-clade (Figure 1).

All strains shared a similar morphology although there was considerable heterogeneity among cell lengths and widths. Filaments consisted of a single trichome within a firm colourless sheath. Trichomes were light to dark brown, motile, gradually attenuated towards apical cells, commonly forming hormogonia and not constricted at cross-walls. Cells were generally shorter than wide with some cells having densely arranged granules at cross-walls. Apical cells were rounded and occasionally capitate or calyptrate when mature. Cell dimensions among anatoxin producing strains (CYN103–107, 109–113, 115–118, 126–128 and 131–133) were 5.8–12.5 μm wide by 1.2–7.8 μm long. Among the non-anatoxin-producing strains (CYN108, 119–125, 129–130) the cell dimensions were 6.4–10.6 μm wide by 2.4–4.8 μm long.

The results of this study demonstrate, for the first time, the coexistence of anatoxin- and non-anatoxin-producing Phormidium strains within small (1 cm²) regions of benthic mats. This co-occurrence of anatoxin and non-anatoxin-producing genotypes within benthic mats has been postulated previously, but not proved conclusively. Heath et al., [2] showed evidence for their co-occurrence by isolating seven strains of P. autumnale from mats that tested positive for anatoxins using LC-MS. Five of the seven strains did not produce anatoxins. Similarly, anatoxin- and non-anatoxin-producing strains of Phormidium were isolated from floating scum samples from one site on the River Tarn, France; however these strains were isolated from samples collected on different dates [24]. Despite the notably different habitats and species involved, the data from this study concurs with planktonic cyanobacteria studies where the co-occurrence of toxic and non-toxic genotypes is now well documented [13,14,15,16,17]. Molecular studies on natural populations of planktonic cyanobacteria have shown that the proportion of toxic and non-toxic genotypes varies as blooms progress. In many instances, these shifts in genotype composition correlate with changes in toxin concentration [25]. Although PCR assays targeting a range of genes involved in the biosynthesis of anatoxin have been developed [26], the use of quantitative PCR (QPCR) to assess changes in genotype abundance has not yet been applied to the study of benthic Phormidium. To determine if there is a relationship between the relative abundances of anatoxin and non-anatoxin-producing benthic Phormidium and anatoxin concentrations, QPCR assays targeting genes involved in anatoxin biosynthesis need to be used to study environmental sample sets.

The total anatoxins concentration of the strains isolated in this study varied from 2.21 to 211.83 mg kg⁻¹ (freeze dried weight). This represents a 100 fold variation in toxin content among toxic strains. Thus in addition to the relative abundance of anatoxin and non-anatoxin-producing genotypes, variations in toxin producing capability among toxic genotypes can influence the overall toxin concentration of a benthic mat sample. The differences in anatoxin production among strains are not apparent from the presence or absence of anaF or other anatoxin biosynthetic genes. These data highlight the utility of isolating strains and using chemical analysis to increase knowledge on variations in toxin production. Anatoxin concentrations from this study (maximum 211.83 mg kg⁻¹), were markedly lower than those measured for P. autumnale strain CYN53 (Rangitaiki River, North Island, New Zealand; [2]), which consistently produces approximately 1000 mg kg⁻¹ of anatoxins [2]. This finding demonstrates that the difference between the most and least toxic strains is even greater than measured in this study. Long-term culturing of cyanobacteria can result in changes in toxin production [27]. In this study, all strains were isolated at the same time and maintained under identical culture conditions and therefore the differences in the quantitative amounts of anatoxins are likely to represent significant differences between strains.

A further important consideration is the relative abundances of the different anatoxin congeners and the effect of this on toxicity. Anatoxin-a and homoanatoxin-a have a mouse LD₅₀ (lethal dose resulting in 50 percent deaths) of 200–250 µg kg⁻¹ body weight [28,29]. Dihydroanatoxin-a has an LD₅₀ of 2 mg kg⁻¹ [30]. Among the strains in this study that contained both anatoxin-a and dihydroanatoxin-a, there was between 10 to 1600 times more dihydroanatoxin-a produced (Table 1). In contrast CYN53 produces predominately anatoxin-a [2], thus as well as producing high levels of anatoxin-a it is also producing the congener variant with the highest toxicity.

All of the strains isolated in this study fell into a single clade supported with 100% bootstrap support within the well-defined 16S rRNA gene sequence P. autumnale clade (Figure 1). The P. autumnale clade has recently been confirmed by morphological and ultrastructure data [31,32]. In the current study all of the strains shared close homology (>99%) with species of Microcoleus and Tychonema. This observation has previously been observed for 16S rRNA sequences of P. autumnale strains [2,31,32]. Previous studies on planktonic cyanobacteria have shown that there is no, or very limited, difference in 16S rRNA gene sequences between toxic and non-toxic strains [33,34]. In contrast, in this study the anatoxin and non-anatoxin-producing strains clearly fell into two separate sub-clades. Based on analysis of a 1340 base pair region of the 16S rRNA gene sequence, there was at least a 17 nucleotide differences between the genotypes. Cadel-Six [24] studied benthic cyanobacteria using a different phylogenetic marker and also observed similar results. They showed that the internal transcribed spacer regions (ITS), ITS1 and ITS2, and the 5.8S regions of rRNA operon could provide sufficient resolution to demonstrate that each neurotoxic phenotype corresponded to a distinct genotype. Based on the observed differences in the 16S rRNA gene sequences it may be possible to design a PCR-based assay to separate these genotypes. However, for distinguishing toxic and non-toxic genotypes, PCR-based assays targeting genes involved in anatoxin biosynthesis [26] would be more conclusive in identifying anatoxin potential than using an assay based on 16S rRNA gene sequences.

It is well known from research on planktonic cyanobacteria that morphology alone cannot be used to distinguish toxic and non-toxic genotypes [35,36]. A similar result has been shown for some benthic species [2] and despite variability in the 16S rRNA gene sequence phylogeny, difference in morphology could not be used to separate anatoxin and non-anatoxin-producing genotypes. This finding is consistent with Comte et al., [31] who were unable to separate P. autumnale strains Arct-Ph5 and Ant-Ph68 based on morphological characters, but identified them as two subspecies by genetic analysis (16S rRNA gene sequences). The general morphology of all strains in this study was similar and there was considerable variability in cell dimension within both toxic and non-toxic strains. Phormidium autumnale displays considerable plasticity, with strains varying from 1.6 to 13 μm wide by 1 to 13 μm long [37].

Spatial and temporal variability in anatoxin concentration occurs among benthic Phormidium mats [2,8,9]. In this study, 30 strains of Phormidium were isolated from four small sections of benthic mats. Both anatoxin and non-anatoxin-producing genotypes were found to co-occur. There were also 100 fold differences in the amount of anatoxin produced by the toxic strains isolated from a single mat. We speculate that the ratio of anatoxin to non-anatoxin-producing strains, and the varying toxin producing capabilities of each toxic strain within a benthic mat are responsible for the dramatic shifts in anatoxin concentration observed in our previous studies (for example, from 558 to 35 mg kg⁻¹ over 7 days [47]). In-depth studies on environmental samples using quantitative molecular techniques will be required to confirm these hypotheses.