Introduction

Toxin-producing cyanobacteria are recognized as a threat to public health globally, particularly in regions without access to high quality potable water. Increased aquatic eutrophication has resulted in massive blooms of cyanobacteria both in freshwater and marine environments. The most commonly produced cyanobacterial toxins are the hepatotoxins, microcystin and nodularin (Fristachi and Sinclair, 2008), both capable of inhibiting protein phosphatase 1 and 2A to varying degrees (Mackintosh et al., 1990) and inducing protein hyperphosphorylation, cytoskeletal collapse and massive hepatic bleeding (Kuiper-Goodman et al., 1999). Both toxins are also recognized as potential tumor promoters and carcinogens, and hence it is of importance to identify the production of these toxins in the environment to eliminate even low-level exposure to humans (Humpage, 2008).

Nodularin is a cyclic pentapeptide (Figure 1) synthesized by the planktonic cyanobacterial species Nodularia spumigena and the benthic species Nodularia sphaerocarpa PCC7804 (Beattie et al., 2000; Moffitt et al., 2001). It occurs primarily in brackish water blooms worldwide, including the Baltic Sea and coastal waters off New Zealand, Australia, South Africa and the United Kingdom (Beattie et al., 2000). Variants of the nodularin toxin have been reported, namely [L-Har2] nodularin (Beattie et al., 2000; Saito et al., 2001), [6(Z)-Adda3] nodularin, [D-Asp1] nodularin and [DMAdda3] nodularin (Namikoshi et al., 1994). The toxin motuporin, isolated from the marine sponge Theonella swinhoei, is thought to be another variant. It has arginine replaced by valine at position 2 (De Silva et al., 1992). The producer of this toxin is unknown but thought to be a cyanobacterial symbiont. Nodularin synthesis has not been reported in any other cyanobacterial genus. Conversely, numerous cyanobacterial genera, including Microcystis, Anabaena, Phormidium, Nostoc and Planktothrix, produce the heptapeptide microcystin-LR, similar in structure to nodularin with an additional two amino-acid residues, D-alanine and D-leucine, and with N-methyldehydroalanine replacing N-methyldehydrobutyrine (Rinehart et al., 1988).

Figure 1
figure 1

The chemical structure of the pentapeptide nodularin and its variant [L-Har2] nodularin indicating the numbering of the amino acids. The * indicates where the additional amino acids, D-alanine and D-leucine are inserted for the heptapeptide toxin, microcystin. Adda, (2S,3S,8S,9S)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4(E),6(E)-dienoic acid; Arg, arginine; D-Glu, D-glutamate; D-MeAsp, D-erythro-β-methylaspartic acid; Har, homoarginine; Mdhb, N-methyldehydrobutyrine (Rinehart et al., 1988).

Full size image

Nodularins and microcystins are biosynthesized via modular megasynthases (Figure 2A), which consists of both non-ribosomal peptide synthetase (NRPS) and polyketide synthase (PKS) modules (Tillett et al., 2000; Moffitt and Neilan, 2004). Each module contains enzymatic domains responsible for substrate selection and activation, modification and condensation. Within NRPS modules these functions are catalyzed by adenylation, thiolation, methyltransferase, epimerase and condensation domains. The 48-kb gene cluster encoding the synthesis of nodularin (nda) (Moffitt and Neilan, 2004) resembles the microcystin synthetase cluster (mcy) (Tillett et al., 2000), with the exception of two NRPS modules that are absent in nda (Figure 2A). Christiansen et al. (2003) proposed that the mcy gene cluster arose from an insertion of two additional activation domains into the nda cluster. In contrast, investigation of the co-evolution of housekeeping genes and the toxin gene clusters (Rantala et al., 2004) and the complete gene cluster (Moffitt and Neilan, 2004) suggested that the nda cluster arose from an ancestral mcy cluster by a deletion event spanning the condensation domain of the first module of mcyA (mcyA-C1) to the condensation domain of the first module of mcyB (mcyB-C1) (Figure 2B). The plasticity of the mcy gene cluster is well known, with various deletions and recombination events allowing the synthesis of a diverse array of microcystin variants. Deletion mutants of the N-methyltransferase domain of mcyA in microcystin producing Anabaena strains were still able to synthesize microcystin (Fewer et al., 2008). Independent recombination of adenylation domains without their associated condensation domains resulted in the generation of functional new peptide synthetases (Fewer et al., 2007), thereby confirming that gene reorganization is a common means of diversification of non-ribosomal peptide synthesis in the cyanobacteria.

Figure 2
figure 2

(A) The diagrammatic representation of the regions of the microcystin and nodularin gene clusters, of M. aeruginosa PCC7806 and Nodularia spumigena NSOR10, respectively, investigated in this study. Primers HEPF/HEPR (Jungblut and Neilan, 2006) were used to target the aminotransferase region of the mcyE and ndaF genes (diagonal hatching). Analysis of the evolutionary relationship between nda and mcy gene clusters was performed on the ndaA and mcyA/B gene regions indicated by perpendicular hatching. Analysis of the adenylation-binding domain encoded by ndaB and mcyC (horizontal hatching) was also performed. The diagram is not drawn to scale. (B) A diagram representing the nodularin biosynthetic gene, ndaA, and its proposed evolution from the microcystin biosynthetic genes mcyA and mcyB. Each gene encodes two non-ribosomal peptide synthetase (NRPS) modules. The approximate size and location of the encoded adenylation (A), condensation (C), thiolation (T), epimerase (Ep) domains within each module is shown and numbered according to their order of appearance. (a) The nodularin synthetase gene, ndaA, contains two partial condensation domains, C1a and C1b, that are proposed to have evolved from mcyA-C1 and mcyB-C2 following deletion of a fragment from the microcystin biosynthetic gene cluster. Forward and reverse primers, designed in this study for the amplification and sequencing of the Nostoc ndaA condensation domain, are indicated by the arrows. (b) The region from mcyA-C1 and mcyB-C2 (solid line) is proposed to have been deleted during the evolution of the ndaA gene.

Full size image

Nostoc strains are known to associate symbiotically with a wide range of hosts, such as diatoms, bryophytes (lichens and hornworts), pteridophytes (Azolla), gymnosperms (cycads) and angiosperms (Gunnera) (Rai et al., 2000; Bergman et al., 2007). A total of 24 endosymbiont species, mostly Nostoc spp., were recently isolated from the coralloid roots of 74 cycad plants (Gehringer et al., 2010). It has been previously demonstrated that the symbiotic cyanobacteria in Macrozamia riedlei, while photosynthetically inactive, fixed nitrogen and transferred it to the host cycad plant in the form of citrulline and glutamine (Lindblad et al., 1991). Microcystin-producing Nostoc symbionts of lichens have also been identified (Oksanen et al., 2004; Kaasalainen et al., 2009), with both strains producing several variants of microcystin. Microcystins are known to disrupt many plant physiological processes, including CO2 fixation, sucrose synthesis, starch storage (McElhiney et al., 2001) and photosynthesis (Abe et al., 1996; Pflugmacher et al., 2007) by inhibiting plant protein phosphatases (Peuthert et al., 2008). Microcystin and nodularin exposure causes oxidative stress and stunted growth in plants exposed to the toxin in the rhizosphere (Gehringer et al., 2003; Lehtimaki et al., 2011), as well as inhibiting germination and seedling growth of several agricultural plants (Saqrane et al., 2008). Accumulation of microcystin and nodularin has also been shown in plants irrigated with toxin-containing water (Mohamed-Al Shehri, 2009). The aims of the present study were to identify the potential for hepatotoxin production in commonly occurring terrestrial cyanobacterial species known to associate with gymnosperms (Cycadaceae), given the evidence that lichen symbionts are capable of synthesizing microcystin (Oksanen et al., 2004; Kaasalainen et al., 2009). In addition, we wished to determine whether this toxin was produced in planta and, if so, whether the levels of toxin production were high enough to affect plant physiology.

Materials and methods

Culture conditions

Cyanobacterial isolates (Gehringer et al., 2010) were grown on BG11 medium and incubated at 22 oC on a 16:8 h light:dark cycle of 2000 lux. Isolates used for toxin detection were grown in 500 ml flasks with gentle agitation as above. Cells were harvested in late exponential phase (day 10).

Toxin biosynthesis gene detection

DNA was extracted from fresh culture material using the XS buffer extraction method (Tillett and Neilan, 2000) and screened using the primer pair HEPF/R targeting the mcyE/ndaF gene (Jungblut and Neilan, 2006; Figure 2A). Briefly, 50 ng of DNA was used in a 20 μl reaction containing 5 pmol of each primer, 1 × F1 Taq polymerase buffer, 0.2 U F1 Taq (Fisher Biotec, Wembley, WA, USA), 200 μM of each deoxynucleotide triphosphate and 2.5 mM MgCl2. The reaction was heated to 95 oC for 2 min followed by 35 cycles of 30 s at 95 oC, 40 s at 55 oC and 50 s at 72 oC with a final extension of 5 min at 72 oC. PCR products were precipitated (Gehringer et al., 2010) and sequenced using the PRISM Big Dye Terminator V3.1 system (Applied Biosystems, Mulgrave, VIC, Australia). Total genomic plant DNA was extracted from the leaves of the cycad plants containing toxin-producing symbiont species, using the Qiagen Plant DNAEasy Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions. This DNA was used as a negative control for the HEPF/R PCR.

Characterization of the Nostoc nda cluster via gene walking

Elucidation of the unknown regions of DNA flanking the putative nodularin aminotransferase biosynthesis gene of ndaF, as well as the complete ndaA and ndaB genes, were determined by an adaptor-mediated (pan-handle) PCR method that was modified as previously described (Siebert et al., 1995). A detailed description is provided in the supporting text. The ndaA and ndaB genes were identified using universal degenerate primers (MTF2: 5′-GCNGG(C/T)GG(C/T)GCNTA(C/T)GTNCC-3′ and MTR: 5′-GCNGG(C/T)GG(C/T)GCNTA(C/T)GTNCC-3′) targeting peptide synthetases (Neilan et al., 1999). The primer pair NdaA C1F2 (5′-GACGAATCAAAATGGCGAAT-3′) and NdaA C1bR (5′-AATTGGCTGGAAGATGCAAG-3′) was designed from the ndaA sequence (Figure 2B) obtained from Nostoc sp. ‘M. serpentina 73.1’ to amplify the partial condensation domain sequences. The primers were used at 5 pmol each in a 20 μl reaction containing 100 ng of DNA, 1 × F1 Taq polymerase buffer, 0.2 U F1 Taq (Fisher Biotech), 200 μM of each dNTP and 2.5 mM MgCl2. The amplification cycle was initially 2 min at 95 oC followed by 30 cycles of 20 s at 95 oC, 20 s at 58 oC and 90 s at 72 oC with a final extension of 5 min at 72 oC. PCR products were precipitated and sequenced with both primers as described above.

DNA and amino-acid sequence analysis

The aminotransferase gene sequences and the 16S rRNA gene sequences of the isolates carrying the mcyE/ndaF genes, those of their nearest neighbors and representative reference strains were aligned using Clustal W (Thompson et al., 1997), end gaps removed and the sequences re-aligned. Pairwise percentage identities were calculated for sequences within the alignments. Maximum likelihood trees were generated using the PhyML-3 package (Guindon et al., 2009) using the HKY85 substitution model with 1000 bootstrap resampling events. No amplification products were obtained from plant DNA using the HEPF/R primer pair. The NRPS/PKS analysis tool available at http://nrps.igs.umaryland.edu/nrps/ (Bachman and Ravel, 2009) identified the location of the NRPS domain sequences. The duplicated region within the ndaA condensation domain and previously described conserved core motifs (Rausch et al., 2007) were identified manually using the Ugene program (http://ugene.unipro.ru/). The Nodularia spumigena NSOR10 and Nostoc sp. ‘M. serpentina 73.1’ nodularin synthetase protein NdaA contains a condensation domain sequence NdaA-C1, located at positions 1278–1711 and 1292–1731, respectively. A second partially duplicated condensation domain, identified downstream of NdaA-C1, NdaA-C1b was located between 1712 and 1973 and 1732 and 2017 in Nodularia spumigena NSOR10 and Nostoc sp. 73.1, respectively. The Microcystis aeruginosa PCC7806 condensation domain of McyA, McyA-C1, was located between positions 1280 and 1706, and the second condensation domain of McyB, McyB-C2, was located between positions 1062 and 1492. The binding pocket sequences for the adenylation domain binding pockets of NdaB and McyC for the strains, indicated in Table 1, were determined using the NRPS/PKS synthase analysis tool available at http://nrps.igs.umaryland.edu/nrps/ (Bachman-Ravel 2009, p189) and the NRPS predicter tool (Rausch et al., 2005). Using a statistical model of multiple protein alignments, the software predicts PKS and NRPS units from an input protein sequence. It then identifies the potential eight amino-acid specificity domain derived from a database of NRPS adenylation domains previously characterized biochemically. Sequences generated in this project have been deposited in GenBank with accession numbers listed in Table 1.

Table 1 Toxin profiles and sequence data of the cycad endosymbionts investigated in this study, as well as the reference data used for phylogenetic analysis
Full size table

Protein phosphatase 2A (PP2A) inhibition assay

A bioassay was performed on cyanobacterial extracts to determine whether the five isolates carrying the mcyE/ndaF gene regions were producing bioactive toxin or not. Approximately 500 mg of fresh culture material of these five isolates were resuspended in 500 μl of 70% methanol and lyzed by bead beating (FastPrep, Savant, Thermo Fisher Scientific, Waltham, MA, USA). After centrifugation to remove cell debris, the supernatant was extracted with an equal volume of chloroform and used in the colorimetric PP2A inhibition assay, as described in Ward et al. (1997) and detailed in the supplementary text.

In vitro toxin identification

As both microcystin and nodularin inhibit PP2A, it was necessary to identify which toxin was present in those samples showing bioactivity. Extracts shown to inhibit PP2A were desiccated and resuspended in 45% methanol for the identification using electrospray ionization mass spectrometry (ESI-MS). Samples were analyzed using an Ultima API hybrid QTof tandem mass spectrometer (Micromass, Manchester, UK) instrument via nanospray injection into the ESI source. The mass spectra between 100 and 1500 m z−1 were acquired in the positive ion mode with a capillary voltage of 3 kV. Major species were subjected to ESI-MS/MS analysis by increasing the collision energy to 60 eV in order to identify the product ions and deduce the compound structure (Mazur-Marzec et al., 2006).

New extracts of fresh culture material were made for HESI-MS/MS analysis. Approximately 500 mg of cells from the Nostoc spp. cultures were resuspended in 500 μl of 90% methanol. Cells were disrupted in a FastPrep bead beater (Savant) and stored in the dark overnight. The cell debris was pelleted by centrifugation and the OD665 reading of the cleared extract was taken to determine the chlorophyll a (Chl a) content (Meeks and Castenholz, 1971). One millitre of 70% methanol was added to the remaining cell debris and left overnight to complete the toxin extraction. Both the 70 and 90% methanol extracts were pooled, diluted 1:3 in water and concentrated on a C18 column (Spoof et al., 2003).

The uncertainty of detection is reported as a 95% confidence interval, based on the standard deviation of the mean. A calibration equation was derived from the linear regression of five nodularin standard solutions. The standard errors of the estimated concentrations from the calibration equation were between 5 and 9%.

In planta toxin identification

In order to determine if toxin was produced in planta, approximately 500 mg of the green cyanobiont ring, containing both plant root material and cyanobionts, was extracted with 1 ml of 70% methanol, diluted 1:3 in water, loaded onto pre-conditioned Sep-Pak C18 classic cartridges (Waters Corporation, Milford, MA, USA) and eluted with 90% (v/v) methanol. Following the solid phase extraction clean up, the eluents were filtered through Millex-HV 0.45 μm PVDF filters (Millipore, Billerica, MA, USA) (Spoof et al., 2003). Toxic compounds were separated using reversed phase high performance liquid chromatography (HPLC) and identified using an Orbitrap LTQ XL (Thermo Fisher Scientific) ion trap mass spectrometer using a HESI source. The multiple reaction monitoring product ions used were determined from a commercial nodularin standard and from the ESI-MS/MS data for [L-Har2] nodularin.

Results

Detection of mcyE/ndaF and phylogenetic analysis

A total of 47 surviving cyanobacterial isolates including 44 Nostoc (19 strains) and 3 Calothrix isolates obtained during a previous cyanobiont diversity study (Table 2 in Gehringer et al., 2010) were screened for the presence of the mcyE/ndaF toxin biosynthesis gene essential for microcystin or nodularin synthesis, respectively. Five of the isolates were found to contain regions indicative of the presence of the mcy or nda gene clusters responsible for the non-ribosomal synthesis of these toxins (Table 1). Neither an amplification product was obtained for the remaining 42 isolates nor for plant DNA controls. Phylogenetic analysis of the DNA sequences was performed to obtain an indication of the nature of potential toxin biosynthesis in these isolates. Three of the isolates clustered with the mcyE gene sequences of Anabaena sp. 90 and Anabaena sp. 318, both known to produce microcystin (Figure 3a, Table 1), with a high degree of statistical support (69%). This cluster was linked to the nodularin-producing cluster of ndaF sequences from Nodularia species and another symbiont Nostoc sp. (52% support at node). The sequence of the aminotransferase domain from Nostoc sp. ‘Macrozamia macleayi 74.2’ was placed within the phylogenetic branch of mcyE genes from microcystin-producing Microcystis species (Figure 3a). None of the cyanobiont species grouped within the cluster of nodularin producers. Of interest is the positioning of the AMT region of Nostoc sp. 152, a known microcystin producer, with the nodularin producing Nodularia spp. Four of the sequences obtained from the endosymbiont Nostoc spp., namely JF342710, JF342713, JF342714 and JF342715 (Table 1), had deletions within the sequenced regions of the aminotransferase domains that could result in frameshifts and potentially inactivate biosynthesis.

Table 2 Summary of the quantification of nodularin production in methanol extracts of toxin producing endosymbionts and the coralloid roots from which they were isolated
Full size table
Figure 3
figure 3

Maximum likelihood phylogenetic tree representing genetic similarities between potential toxin-producing Nostoc spp. and reference strains. (a) mcyE/ndaF sequence analysis of a 440-bp fragment of the aminotransferase region; (b) Partial 16S rRNA gene sequence comparison of a 770-bp fragment. Support values greater than 50% (1000 bootstrap events) are presented. Sequences in bold were generated during this study.

Full size image

Phylogenetic analysis of 16S rRNA genes from hepatotoxic genera (Figure 3b) clearly separated the filamentous genera of cyanobacteria, namely Phormidium, Planktothrix, Nostoc, Nodularia and Anabaena, from the unicellular microcystin-producing genus Microcystis. Within the filamentous cluster each species grouped coherently. The two microcystin-producing Anabaena species grouped with the nodularin-producing Nodularia species, followed by the Nostoc clade comprised of non-toxic (Nostoc PCC73102), microcystin-producing (Nostoc sp. IO-102-I and Nostoc sp. 152) and nodularin-producing Nostoc species (Nostoc sp. ‘Macrozamia serpentina 73.1’ and Nostoc sp. ‘Macrozamia riedlei 65.1’).

Molecular analysis of the Nostoc nodularin cluster

Sequencing of key genes in the non-ribosomal gene cluster would provide information regarding the biosynthetic product of the gene cluster. Pan-handle PCR and sequencing of the genetic region surrounding the AMT domain analyzed above and the NRPS-encoding genes of Nostoc sp. ‘M. serpentina 73.1’ (Table 1), indicated the presence of three open reading frames, encoding three NRPS modules and one hybrid NRPS/PKS module. The domain structure of the modules was identical with the corresponding modules in N. spumigena NSOR10 within proteins NdaA and NdaB. This analysis implied that Nostoc sp. ‘M. serpentina 73.1’ produced nodularin and that additional chemical analysis was necessary to confirm this.

Further analysis of amino-acid residues that line the binding pocket of the adenylation domains and contribute to substrate specificity was performed (Challis et al., 2000). The NdaB adenylation domain carries binding pocket residues that are important in incorporating L-Arg or L-Har residues into nodularin. The binding pocket residues in Nostoc sp. ‘M. serpentina 73.1’ were identical to those found in N. spumigena NSOR10 and N. spumigena CCY9414 (Table 1), further supporting the hypothesis that this strain produces nodularin, rather than microcystin.

Previously, we proposed that the nda gene cluster evolved from the mcy cluster following deletion of a gene fragment encoding two NRPS modules from within the mcyA and mcyB genes (Figure 2A) (Moffitt and Neilan, 2004). This hypothesis is based on the identification that within the condensation domain of the NdaA NRPS module (NdaA-C), highly conserved motifs are repeated. The protein sequence before and after the domain repeat of Nostoc sp. ‘Macrozamia serpentina 73.1’ (referred to as NdaA-C1 and NdaA-C1b) was homologous to McyA-C and McyB-C2, respectively. As previous analyses have been restricted to the nda cluster from N. spumigena, we used this as an opportunity to further investigate this theory using the nda genes of the Nostoc strains. Newly designed primers targeting the condensation domains within the ndaA gene (Supplementary Figure S1B) were used to determine the partial condensation domain sequence in the four ndaF carrying endosymbiont species of Nostoc. Alignment of the DNA sequences for the region spanning the condensation motif repeat for Nostoc spp. ‘Macrozamia serpentina 73.1’, ‘M. mountperriensis 62.1’ and ‘Bowenia serrulata 1.3’ show high identities (98–99%) with one another and an 84% identity with the similar sequence found in N. spumigena NSOR10 (Supplementary Fig. S1A). All three endosymbionts have a 21 bp insertion (highlighted red in Supplementary Fig. S1) when compared to the N. spumigena NSOR10 sequence. There is a three base AAC deletion, a two base TT and a single T deletion across all the three endosymbiont sequences when compared to the sequences from N. spumigena NSOR10 (highlighted in blue in Supplementary Figure S1A).

Careful analysis of the Nostoc NdaA-C domain sequence located a region that appeared to be duplicated in a similar manner to that described in the N. spumigena NdaA-C domain sequence (Figure 4b). Each of the Nostoc NdaA-C domain sequences were divided into two sections, NdaA-C1 and its duplicate NdaA-C1b (Figure 2B), and phylogenetic analysis (Figure 4a) revealed a similar clustering of Nostoc and N. spumigena NdaA-C domain sequences. The NdaA-C1 domain of three endosymbionts grouped with the McyA-C1 domains of the microcystin producers and NdaA-C1 domains of the nodularin-producing strains (Figure 4a). The NdaA-C1 region of Nostoc sp. ‘Macrozamia riedlei 65.1’ could not be amplified and was excluded from analysis. The NdaA-C1 clade was distinct from the tightly grouped clusters observed for the McyB-C1 and McyC/NdaB condensation domains, with high support of the nodes (71% and 81%, respectively). The NdaA-C1b domains of all four ndaF carrying endosymbiont species were most similar to the NdaA-C1b homologs of N. spumigena NSOR10 and N. spumigena CCY9414, followed by the McyB-C2 domains of the microcystin-producing cyanobacterial species. To further demonstrate the similarity of the duplicated NdaA-C domain sequences to the respective Mcy domains, the conserved motifs within the condensation domains of NdaA of Nodularia spumigena NSOR10 and Nostoc sp. ‘Macrozamia serpentine 73.1’ and McyA and McyB of Microcystins aeruginosa PCC7806 were aligned (Figure 4b) (Rausch et al., 2007). The motifs C1–3 and C5, 6 and 7 are highlighted in grey with residues corresponding to the conserved motifs bolded. Due to low sequence homology, conserved motif C4 could not be easily identified. The alignment demonstrated that NdaA-C1 is homologous to McyA-C, whereas NdaA-C1b is homologous to McyB-C2. A seven amino-acid insertion was observed within the NdaA-C1 region of Nostoc sp. ‘Macrozamia serpentine 73.1’ just downstream of conserved domain C3 (Figure 4b). The nucleotide deletions resulted in two single amino-acid deletions in the NdaA-C1 and NdaA-C1b regions outside of the conserved domains when compared to N. spumigena NSOR10-derived sequence. These analyses indicated the synthesis of nodularin by some endosymbiont Nostoc spp.

Figure 4
figure 4

(a): Maximum likelihood phylogeny comparing the sequence similarities in the condensation domains (C) of the newly described Nostoc sp. ‘Macrozamia serpentina 73.1’ and representative condensation domains of NdaA, NdaB, McyA, McyB and McyC reference protein sequences. Bootstrap values greater than 50% after 1000 resampling events are indicated at the relevant nodes. Sequences indicated by a dot were from nodularin producers. (b) Protein sequence alignment of the conserved motifs C1-3 and C5, 6 and 7 (shaded in grey) of the condensation domains NdaA-C1 and NdaA-C1b from Nodularia spumigena NSOR10 (NSOR10) and Nostoc sp. ‘M. serpentine 73.1’ (73.1), and the corresponding McyA-C1 and McyB-C2 domains of M. aeruginosa PCC7806 (PCC7806). Residues corresponding to the conserved motifs are in bold. A seven amino-acid insertion in the protein sequence of Nostoc sp. ‘Macrozamia serpentina 73.1’ when compared to N. spumigena NSOR10 is boxed.

Full size image

Bioactivity, identification and quantification of toxin from symbionts

In order to identify cyanobiont isolates actively producing toxin, crude methanol extracts of culture material were analyzed qualitatively in the PP2A inhibition assay. Strong PP2A inhibition corresponding to approximately 10 μM equivalent microcystin-LR (approximately 8 μg nodularin per ml or 8 μg nodularin per g ww) was identified in the methanol extracts of Nostoc spp. ‘Macrozamia riedlei 65.1’ and ‘Macrozamia serpentina 73.1’, confirming the presence of bioactive toxins. No bioactivity was detected in extracts from Nostoc sp. ‘Bowenia serrulata 1.3’, Nostoc sp. ‘M. mountperriensis 62.1’ or Nostoc sp. ‘Macrozamia macleayi 74.2’.

These samples were subsequently analyzed using HPLC, with an ESI-MS/MS, and were found to contain nodularin, as indicated by the m z−1 of 825 Da. MS/MS analysis (Supplementary Figure S2A–D) confirmed the identity of nodularin by the dissociation of indicative fragmentation patterns at high collision energy (Mazur-Marzec et al., 2006). The toxin extract of Nostoc sp. ‘M. serpentina 73.1’ was also found to contain the nodularin variant, [L-Har2] nodularin (Supplementary Figure S2E–H). The structure of this variant was confirmed by the presence of fragmentation peaks with an increased m z−1 of 14 Da when compared to the data for nodularin, that were only detected when arginine residues were present. Adda was ruled out as the variable residue as neither the PheCH2CH(OCH3) nor C11H15O fragments varied from the nodularin variant fragmentation patterns (Supplementary Figure S2). Other supporting evidence included the presence of L-Har in [L-Har+H+NH3]+ (Supplementary Figure S2, peak 5), which has a m z−1 of 188 instead of 174 [Arg+H+NH3]+, the difference of a 14 Da moiety corresponding to the mass of the additional CH2 unit in [L-Har2] nodularin.

HPLC, with a heated electrospray ionization source (HESI) coupled to tandem mass spectrometry (HPLC HESI-MS/MS), was used to quantify the amount of nodularin in newly obtained extracts obtained from the Nostoc sp. ‘M. riedlei 65.1’ and Nostoc sp. ‘M. serpentina 73.1’ against Chl a content. A high-resolution molecular ion scan and multiple reaction monitoring scans from the precursor molecular ion of m z−1 825.45 Da were used to quantify the mass concentration of nodularin (Supplementary Figures S3–S5). From triplicate measurements of Nostoc sp. ‘M. riedlei’ 65.1 and duplicate measurements of Nostoc sp. ‘M. serpentina 73.1’, the mass concentrations of nodularin were 9.55±2.4 ng μg−1Chl a and 12.5±8.4 ng μg−1Chl a, respectively (Table 2). Further two scans, one for the precursor molecular ion m z−1 of 839.5 Da and one for the MS/MS product ion, m z−1 767.0 Da (Supplementary Figures S5A and S5B, spectra e–f), indicated the likely presence of [L-Har2] nodularin in both cultures. The absolute mass concentrations for [L-Har2] nodularin are not presented as a standard for this isoform was not available.

Analysis of nodularin in cycad roots

The availability of small amounts of frozen coralloid root samples from which the toxin-producing endosymbionts had been isolated allowed us to address the question: Are these toxins produced in planta? Methanol extracts of the green symbiotic ring within the remaining coralloid roots were obtained. HPLC HESI/MS/MS analysis identified the presence of nodularin in the coralloid root extract obtained from Macrozamia riedlei MZ65 (Gehringer et al., 2010) (Supplementary Figure S6), the same plant from which Nostoc sp. ‘M. riedlei 65.1’ was isolated. HPLC HESI-MS/MS also indicated the presence of [L-Har2] nodularin in the root extract of Macrozamia serpentina MZ73 (Gehringer et al., 2010) (Supplementary Figure S7, spectra e–f) from which the endosymbiont Nostoc sp. ‘M. serpentina 73.1’ was obtained.

Data from the M. riedlei 65.1 roots, using both the nodularin molecular ion and multiple reaction monitoring peak intensities, were mean centered and combined in a principal components analysis. Three replicate measurements of the root extract revealed a mean concentration of 1.34±0.74 ng ml−1 (Table 2). Accurate quantitation of toxin content against Chl a was not feasible as the samples had been stored at –20 oC for 2 years. Sample size prevented quantitation of toxin against protein content.

Discussion

Nodularin synthesis has previously been reported only in the aquatic Nodularia spumigena species and the benthic, mat-forming Nodularia sphaerocarpa strain PCC7804 (Beatti et al., 2000; Saito et al., 2001; Moffitt and Neilan, 2004; Fristachi and Sinclair, 2008). This study reports the synthesis of nodularin and [L-Har2] nodularin synthesis by terrestrial Nostoc spp. Two of the terrestrial symbiont isolates investigated in this study, Nostoc sp. ‘M. serpentina 73.1’ and Nostoc sp. ‘M. riedlei 65.1’, were confirmed genetically and chemically as nodularin producers (Tables 1 and 2, Supplementary Figures S2 and S5). The only other terrestrial symbiotic cyanobacteria reported to produce hepatotoxins are Nostoc IO-102, isolated from a lichen, and Nostoc sp. strain UK18, isolated from P. leucophlebia cephalodia, both of which produce microcystin (Oksanen et al., 2004; Kaasalainen et al., 2009).

Phylogenetic analysis of the AMT domain (Figure 2A) indicated that five endosymbiont species may produce a hepatotoxin. It was unable to identify whether they were capable of producing either nodularin or microcystin, as illustrated with the positioning of the AMT sequence from the microcystin producer, Nostoc 152 with the nodularin-producing Nodularia species. The positioning of the AMT domain sequence of Nostoc sp. Macrozamia macleayi 74.2 within the microcystin producing Microcystis cluster would suggest that this endosymbiont carries at least part of the mcy cluster. This positioning on the AMT tree would be consistent with the studies of Oksanen et al. (2004) and Kaasalainen et al. (2009) who demonstrated microcystin production in Nostoc symbionts found in lichens. A bioassay utilizing PP2A indicated that only extracts from two cyanobiont isolates inhibited the eukaryotic protein phosphatase, namely Nostoc sp. ‘M. serpentina 73.1’ and Nostoc sp. ‘M. riedlei 65.1’. The lack of bioactivity in the remaining three AMT-carrying symbiont isolates would suggest that they may carry non-functional mcy/nda gene clusters, express toxin at levels below assay detection limits or produce a novel microcystin or nodularin congener that does not inhibit PP2A. In addition, two of the Nostoc isolates demonstrating no bioactivity in the PP2A assay, Nostoc spp. ‘Bowenia serrulata 1.3’ and ‘Macrozamia mountperriensis 62.1’, also had frame-shift mutations in their aminotransferase sequences that may have contributed to their non-toxic phenotype.

The synthesis of both nodularin and [L-Har2] nodularin in these cyanobacterial isolates presented the opportunity to study the mechanistics of toxin production. Sequencing of the partial nda cluster in Nostoc sp. ‘Macrozamia serpentina 73.1’ revealed identical domain structures for NdaA and NdaB as seen in N. spumigena NSOR10. However, additional sequencing is required to confirm if the remaining PKS and NRPS domains are conserved in Nostoc sp. ‘Macrozamia serpentina 73.1’. It has been proposed that changes in the substrate-binding pocket of the terminal adenylation domain of NdaB may allow for more flexibility of binding for the activation and incorporation of L-Har into nodularin during biosynthesis (Moffitt and Neilan, 2004). Nostoc sp. ‘M. serpentina 73.1’, that produces both nodularin and [L-Har2] nodularin, has an NdaB-binding pocket sequence identical to that found in nodularin producers but differs from the other [L-Har2] nodularin producer, Nodularia sphaerocarpa PCC7804 (Table 1). Our analysis indicated that the asparagine to serine substitution is not essential for [L-Har2] nodularin synthesis, and that the binding pocket sequence DVWNFGFV provides sufficient molecular space for the incorporation of the larger L-Har residue.

The identification of a novel nodularin-synthesizing cyanobacterial species other than Nodularia spp. raises questions as to the evolutionary origin of the nda cluster. This cluster differs from the mcy gene cluster as it does not carry the NRPS modules responsible for the incorporation of the amino acids D-Ala and L-Leu (Moffitt and Neilan, 2004). Analysis of the partial condensation domain DNA sequences of the ndaA-C1 and ndaA-C1b genes of the toxin-producing Nostoc sp. ‘M. serpentina 73.1’, as well as the AMT carrying endosymbiont strains Nostoc sp. ‘Bowenia serrulata 1.3’ and Nostoc sp. ‘Macrozamia riedlei 62.1’, revealed the identical condensation motif repeat present in N. spumigena NSOR10 (Figure 4b and Supplementary Figure S1A). This suggested that the genetic event generating the nda cluster was the same for all nodularin-producing strains investigated to date. An identical insertion of seven amino acids was observed in the NdaA-C1 domains of all nda-carrying endosymbiont strains sequenced and was not present in the NdaA-C1 domain of N. spumigena NSOR10. This seven amino-acid insertion was also not observed in the McyA-C1 protein sequence derived from M. aeruginosa PCC7806 (Figure 4b) from which the NdaA-C1 domain is thought to have originated. This would suggest that the seven amino-acid insertion probably occurred after the proposed mcy truncation event and after transfer of the nda cluster into Nostoc and Nodularia species. The protein alignment in Figure 4b and DNA alignment of Supplementary Fig S1A would suggest that the nda cluster was first transferred into Nodularia spumigena species, as all the endosymbionts carry the identical 21 bp insertion and 6 nucleotide deletions. Analysis of the condensation domains of NdaA, McyA and McyB (Figure 4) demonstrate high similarity between the endosymbiont-derived protein sequence and that of N. spumigena NSOR10. This suggests that the duplicated sequence within the NdaA-C domain of Nostoc sp. ‘Macrozamia serpentina 73.1’ may have evolved from the fusion of a McyA-C domain and the McyB-C2 domain following deletion of two NRPS modules from mcy, in the same manner as proposed for the nodularin-producing N. spumigena (Moffitt and Neilan, 2004). This study thereby provides support for the common origin of the nda cluster, however, it cannot offer support as to the order of recombination and transfer of the nda cluster to Nodularia spp.

The observation that the toxin-producing endosymbiont AMT regions do not show good sequence similarity to the AMT sequences found in nodularin-producing Nodularia spp. (Figure 2A), would suggest that the evolution of the nda cluster in Nostoc spp. is not straightforward. Additional sequencing of the nda clusters is required before further sequence and phylogenetic analysis can be undertaken. Whether the genetic event generating the nda cluster occurred in a Nodularia, Microcystis or another species of cyanobacteria, remains an enigma. What is evident from the alignment of the DNA sequences spanning the potential deletion site is that all nda carrying endosymbiont Nostoc species and the nda carrying N. spumigena NSOR10 show high similarity, thereby suggesting a similar evolutionary origin.

Tandem MS analysis indicated the production of nodularin and not microcystin in the two strains showing bioactivity in the PP2A assay. Data on product ion spectra (S2) confirmed that the synthesis of [L-Har2] nodularin as well. Nodularin was further confirmed by HPLC HESI-MS/MS at low concentrations (Table 2) in extracts of the cyanobionts cultures. The levels of toxin produced by these endosymbionts are low, with a μg:μg ratio of nodularin:Chl a of 0.01:1, when compared to the nodularin producing abilities of Nodularia KAC66, an aquatic isolate, with ratios ranging between 0.4:1 and 0.7:1 (Stolte et al., 2002). The discrepancy between nodularin levels detected by the PP2A assay (about 8 μg ml−1) versus the HPLC HESI-MS/MS (over 100 ng ml−1) could be ascribed to differences in the assay procedures. The PP2A would reflect the influence of all the hepatotoxins in an extract on the enzyme, that is, nodularin and the potential inhibition by [L-Har2] nodularin, whereas the HESI-MS/MS reflects the nodularin concentration only. The PP2A assay was standardized against microcystin-LR, whereas nodularin was used to establish a standard curve for HESI-MS/MS. In addition, the HESI-MS/MS nodularin quantitation was completed 2 years after the PP2A quantitation and the cultures may have altered their toxin profile and levels of production.

Nodularin was detected in planta in extracts of the cycad roots that yielded the endosymbiont Nostoc sp. ‘M. riedlei 65.1’, while [L-Har2] nodularin was only present in root extracts that contained Nostoc sp. ‘M. serpentina 73.1’ (Supplementary Figures S6 and S7). Nodularin production was not observed in the root extracts of Macrozamia serpentina (MZ73) (Supplementary Figure S7, spectra a–d). Although the symbiont Nostoc sp. ‘M. serpentina 73.1’ synthesized nodularin, the quantities were significantly lower than that seen for [L-Har2] nodularin (Supplementary Figure S5, Table 2). The lack of amplification of mcyE/ndaF genes from plant material not containing cyanobacterial symbionts indicates that this toxin cannot be of plant origin. The concentration of nodularin in the coralloid root tissue of Macrozamia riedlei (MZ65) can be approximated as 3 pmol g−1 (for a single sample of approximately 500 mg). Lehtimäki et al. (2011) demonstrated signs of oxidative stress in young spinach plants irrigated with nodularin-containing water and detected toxin in young roots in the range of 0–24 pmol g−1. This would suggest that, at least locally, nodularin would induce oxidative stress in the coralloid root tissue. Whether the toxin is transported to the rest of the plant or restricted to the coralloid root, remains to be elucidated.

Research into the toxic effects of nodularin has been largely over-shadowed by studies on the biological effects of microcystin (Humpage, 2008). Given that both toxins inhibit protein phosphatases, it is thought that they exhibit similar biological activities (Mackintosh et al., 1990; Honkanen et al., 1991). The different structures of nodularin has leant itself to the theory that it can target different tissues than those effected by microcystin (Ohta et al., 1994) and therefore induce different toxicological effects in exposed animals and potentially plants. This study has demonstrated the production of nodularin in planta at levels capable of inducing oxidative stress in spinach plant root tissue (Lehtimaki et al., 2011). Whether these effects can be extended to include cycad roots remains to be determined. The nodularin-producing Nostoc spp. produce toxin at levels over 100-fold lower than that observed for the aquatic, nodularin-producing Nodularia sp. Toxin released into an aquatic environment would be immediately diluted, thereby reducing the immediate toxic effects on the aquatic biota. In contrast, toxin released into a drier environment, and possibly in planta, would not be diluted and may accumulate. Whether these levels of hepatotoxin production are significant in a terrestrial environment remains to be investigated. Further research into the ancient relationship between cyanobacteria and the Cycadacea should reveal the molecular interactions between toxin and host plant, and also highlight the evolutionary importance of secondary metabolism in symbioses and the terrestrial environment.