Azolla as a Safe Food: Suppression of Cyanotoxin-Related Genes and Cyanotoxin Production in Its Symbiont, Nostoc Azollae

1. Introduction

Azolla Lam. is the only plant with a permanent nitrogen-fixing cyanobacterial symbiont (cyanobiont) that has chains of cells (filaments) comprising photosynthetic vegetative cells and thicker-walled heterocysts that contain the nitrogen-fixing enzyme nitrogenase [1]. The cyanobiont has been assigned to both Anabaena azollae and Nostoc azollae because its morphology resembles free-living species of both genera, including their change into motile hormogonia and akinetes (resting cells) that ensure survival during stressed conditions [2].

Genetic and paleontological data indicate that the Azolla – N. azollae symbiosis originated 80 million years ago in North America following Whole Genome Duplication (WGD) that increased the genome of Azolla’s immediate ancestor [3]. Nostoc azollae’s subsequent coevolution with Azolla caused extensive changes in the cyanobiont’s genome compared to free-living species of Anabaena and Nostoc [3,4,5,6,7,8]. Some changes involved the upregulation of genes that enhanced N. azollae’s sequestration of atmospheric nitrogen and provision of nitrogen-based compounds to Azolla, increasing the plant’s speed of growth free-floating on fresh water. The downregulation, loss, or conversion to pseudogenes of other genes changed N. azollae’s ancestors from independent free-living organisms into obligate endosymbionts, reflecting N. azollae‘s permanent location inside the leaves and female megasporocarps of Azolla. These included genes that previously expressed proteins involved in the synthesis of carotenoid and chlorophyll pigments, so that A. azollae is reliant on Azolla’s cellular pigments for protection against photooxidative damage [6].

Colonies of N. azollae live in specialized cavities in Azolla’s dorsal floating leaves, providing nitrogen-based nutrients to the plant that enable it to double its biomass in less than two days while free-floating on fresh water [9,10]. As a result, Azolla has been used for hundreds of years in India and the Far East as a nitrogen biofertilizer for paddy rice, reducing mosquito breeding populations by 95% [11,12,13] and emissions of the potent greenhouse gas methane from paddies by 25-50% [14,15,16]. Azolla also provides livestock feed, biofuel and biofertilizer for other plants, alleviating shortages of the ‘three Fs’ that increasingly threaten food supplies globally: feed, fuel, and fertilizer. It absorbs and removes phosphates and nitrates from water contaminated by chemical fertilizers, industrial pollutants, and animal and human waste that trigger toxic cyanobacterial (aka blue-green algal) blooms in rivers and lakes. The symbionts’ combined CO₂ sequestration also increases Azolla’s carbon capture, so that it can sequester large amounts of atmospheric CO₂, with the plants being compressed and stored to reduce anthropogenic climate change through Carbon Capture and Storage (CCS). Azolla can, therefore, mitigate many of the threats arising from a Perfect Storm as our population increases by more than a million every three days. Its remarkable properties are increasingly recognised, and it has designated as a unique superorganism by [1].

Azolla has the potential to help feed millions of people because of its rapid growth, ease of outdoor cultivation in tropical and temperate regions, plus global production using the indoor Azolla Biosystem described by Bujak & Bujak (2020) in ‘The Azolla Story’ [2]. The use of Azolla for human consumption was thought to be limited by its high total polyphenolic content (TPC), but Winstead et al. (2024) [17] showed that the TPC of raw Azolla caroliniana, which is native in the eastern United States, have only 4.26 g gallic acid equivalent (GAE) kg⁻¹ DW, and that simple cooking methods can decrease TPC in all Azolla species. They also demonstrated that its protein content is 19% DW and its apparent protein digestibility is 78.45%, with a yield 173 g FW m⁻² day⁻¹ and 5.53 g DW m⁻² day⁻¹, confirming Azolla‘s potential for cultivation and domestication as a nutritious food. This raises the question of whether Azolla is safe to eat because of the presence of harmful cyanotoxins in many cyanobacteria.

Cyanotoxins are produced by cyanobacteria of the genera Anabaena and Nostoc among others and include some of the most powerful natural poisons that target the nervous system (neurotoxins: BMAA, saxitoxin and anatoxin-a), the liver (hepatotoxins: microcystins and nodularins), protein synthesis and DNA modification (cylindrospermopsin) and the skin (dermatoxins; nodularin). Cyanotoxins are alkaloids (anatoxin-a, saxitoxin, cylindrospermopsin) or peptides (pentapeptide nodularin or the heptapeptide microcystin) [18,19,20,21]. Upon their release in water, they are ingested by zooplankton and animals, or absorbed by phytoplankton and plants that can have acute or chronic effects when eaten by humans. This is a global health issue owing to bioconcentration and bioaccumulation in the food chain and poisoning through ingestion of contaminated food, so that cyanotoxins are now widely analyzed and studied to determine their effects on plants and animals [22]. For example, the World Health Organization (WHO) recommends a value of 1 μg/L for microcystin-LR in drinking water [23].

BMAA (β-N-methylamino-L-alanine) is a non-proteinogenic amino acid is produced by free-living cyanobacteria in marine, freshwater, and terrestrial environments [24,25]. It has been detected in plants with endosymbiotic cyanobacteria including lichens, hornworts, the leaf petioles of the tropical flowering plant Gunnera, and the cycad Cycas circinalis [24,25,26] and linked to the amyotrophic lateral sclerosis/parkinson-dementia complex (ALS/PDC) detected among the Chamorro people living on the Pacific island of Guam [24,25]. BMAA, like other cyanotoxins, can be biomagnified in seafood eaten by people, including fish [27,28], shrimps [29], mussels, oysters, and crabs [30]. BMAA can also be synthetized by eukaryotes such as diatoms [31] and dinoflagellates [32] which are food sources for crustaceans, fish and shellfish [33].

These observations raise the question of whether eating Azolla may be harmful to humans due to the possible production of BMAA and other cyanotoxins by N. azollae. Unlike free-living Anabaena and Nostoc, the loss or conversion to pseudogenes of genes involved in cyanotoxin and/or BMAA production may have occurred in N. azollae because they were no longer needed by the permanently enclosed cyanobiont. The following analyses were therefore undertaken on all seven extant Azolla species and their cyanobionts to determine if they can be safely eaten by people:

• The presence of genes coding for microcystin, nodularin, cylindrospermopsin and saxitoxin.
• The presence of the anatoxin-a/homoanatoxin-a gene cluster by bioinformatic tools.
• The presence of BMAA.

The seven examined species of Azolla are A. caroliniana, A. filiculoides, A. mexicana, A. microphylla, A. nilotica, A. rubra and A. pinnata including its two subspecies A. pinnata subsp. pinnata and A. pinnata subsp. imbricata. Table 1 lists the 48 accessions that provided the Azolla species and subspecies used in this study.

2. Results

2.1. The Cyanotoxins Microcystin, Nodularin, Saxitoxin, Cylindrospermopsin and Anatoxin-a

The PCR amplification of 12 genes that encode for four cyanotoxins (cylindrospermopsin, nodularin, saxitoxin and microcystin) was determined for all seven Azolla species and two A. pinnata subspecies from 48 Azolla accessions listed in Table 1. Global distribution of the accession countries is shown in Supplementary Material Figure S25. The results indicate that cyanotoxin gene amplification on Azolla accessions were negative when matched with positive and negative controls for 12 genes: cyl, mcy A, mcy B, mcy B domain A, mcy C, mcy C domain A, mcy D domain ACP, mcy D domain KS, mcyE/ndaF, mcy E domain GSA-AMT, mcy G domain CM, sxt.

The results also showed that same 12 genes in N. azollae isolated from the 47 Azolla accessions were not amplified compared with positive and negative controls. This indicates that both Azolla and N. azollae do not have genes that biosynthesize those cyanotoxins. Photographs of all gels from the PCR amplifications are shown in Supplementary Material Figures S1 to S24.

The BLAST search for the anatoxin-a and homoanatoxin-a gene cluster showed a query cover of only 3% and percent identity of 73.04%. Most of these alignments were partial segments comprising less than 250 bp of the anaH transposase gene with identities around 70%. The aligned genes in N. azollae are only associated with pseudogenes and not with protein coding genes. All other alignments were partial, and none included any of the whole genes associated with the anatoxin-a biosynthesis gene cluster.

2.2. Detection of BMAA (β-N-methylamino-L-alanine)

The detection of the non-proteinogenic amino acid BMAA by LC-MS/MS with method 1 on all seven Azolla species (A. caroliniana, A. filiculoides, A. microphylla, A. mexicana, A. nilotica, A. rubra and the two subspecies of A. pinnata) showed that the BMAA was absent from both Azolla and its cyanobiont N. azollae since there is not found the retention time for BMAA and the isomer 2,4-DAB (2,4-diaminobutyric acid) when compared with their standards which are not shown).

Re-analysis of six of the Azolla species (A. rubra that was not re-analyzed) using method 2, in which the samples were derivatized, corroborated the results obtained with method 1. This indicates that the Azolla species shown in Figure 1 did not show the retention times for BMAA (RT=13.92 min) and the two isomers 2,4-DAB (RT=14.93 min), and AEG (N-(2-aminoethyl)-glycine) (RT=13.22 min) when compared with their standards (Figure 2).

3. Discussion

Most genera of free-living cyanobacteria synthesize cyanotoxins and include cyanobacteria that have temporary symbiosis with some plants, so that the host plant has the potential to assimilate and bioaccumulate the cyanotoxins, discussed above. There are few published studies documenting genetic and chromatographic detection of cyanotoxins in cyanobionts. Cyanobacteria from lichens have been analysed and contain genes that encode nodularin and microcystin, and can translate the peptides nodularin and microcystins [34,35,36]. The synthesis of those two cyanotoxins may be linked to the temperature and humidity in which the lichens grow and may be important for the maintenance of lichens in diverse ecological habitats [37].

Unlike lichens, the fern Azolla has a permanent symbiosis with the cyanobacteria N. azollae giving this symbiosis a unique evolution pattern and, ultimately, the loss of genes by the cyanobiont [4,5,6,7]. Our genetic analyses show, for or the first time, that all seven Azolla species and their cyanobiont, N. azollae, do not possess genes associated with the synthesis of microcystin, nodularin, saxitoxin, cylindrospermopsin, anatoxin-a, and homoanatoxin-a. The biosynthetic pathways of microcystins [38] and anatoxin-a/homoanatoxin-a [39] are a multi-step process that requires several genes to synthesize both cyanotoxins. All genes associated with microcystin synthesis were, therefore, amplified by specific primers with Azolla and N. azollae DNA, and the complete anatoxin-a gene cluster was BLAST searched against the N. azollae genome. There were no matches, supporting the model that N. azollae lost the ability to synthesise microcystin, nodularin, saxitoxin, cylindrospermopsin, anatoxin-a, and homoanatoxin-a due to downregulation of cyanotoxin biosynthesis genes or loss of the genes during the co-evolution of Azolla and N. azollae.

BMAA was isolated in 1967 from seeds of Cycas circinalis (cycad) [24,25,26] and identified as the primary cause of amyotrophic lateral sclerosis/parkinson-dementia complex (ALS/PDC) in the Chamorro people on the Pacific island of Guam [40,41], with the high levels of the neurotoxin resulting from biomagnification through the food chain [27,28,33,42,43]. BMAA is a cyanotoxin that can cross the blood-brain barrier where it forms a reservoir [44]) and can be inserted into proteins instead of the amino acid L-serine, causing protein misfolding and aggregation [45,46,47]. BMAA can induce changes in the expression of genes in brain cells and thus resulting in a wide range of other neurodegenerative disorders [19]. Alzheimer’s, Parkinson’s and other neurological diseases including amyotrophic lateral sclerosis (ALS), progressive supranuclear palsy (PSP) and dementia with Lewy bodies (DLB) [48] may therefore be partially caused or facilitated by BMAA. However, the gene/genes for the codification of BMAA are not known in any cyanobacteria and plant, so that their presence can only be detected by analytical methods, including those used in this study.

BMAA was detected in plant-cyanobacteria symbiosis such as hornworts, liverwort, lichens, cycads and Gunnera [24] and also in A. filiculoides with 2 μg/g [42]. Some analytical methods to detect this non-proteinogenic molecule can result in erroneous interpretations due to structural isomers DAB (2,4-diaminobutyric) and AEG (N-(2-aminoethyl)-glycine which can co-elute and be mis-identified as BMAA [49,50]. For the present study, two methods were therefore used to detect BMAA, DAB and AEG, and did not detect BMAA, AEG, or DAB in any of the analyzed Azolla and N. azollae. These data indicate that the previous reported detection of 2 μg/g BMAA in Azolla [42] is incorrect.

Harmful algal blooms (HABs) of other cyanobacteria species also release cyanotoxins upon cell necrosis. The uptake and bioaccumulation of cyanotoxins from irrigated water for crop and non-crop therefore also needs to be evaluated to determine if Azolla species may bioaccumulate cyanotoxins. Previous studies show that A. filiculoides does now uptake or bioaccumulate microcystin [51] or cylindrospermospin [52], confirming that Azolla can be safely eaten.

4. Materials and Methods

4.1. Azolla Accessions and Culturing

The seven Azolla species including two A. pinnata subspecies from the germplasm collection at IRRI (International Rice Research Institute) and two A. filiculoides accessions from Portugal (FI-BGLU and FI-BGM) were used to detect the cyanotoxin genes of microcystin, nodularin, cylindrospermospin and saxitoxin by PCR, and BMAA by LC-MS/MS (Table 1.) The 48 Azolla accessions have a global distribution from 33 countries (Supplementary Material, Figure S25). The Azolla species were cultured in Hoagland medium (H-40), pH 6.1-6.2, at controlled temperature, photoperiod, and light intensity [53]. The biomass was collected, washed in distilled water, frozen at -80°C, lyophilized, and then weighed.

4.2. Isolation of Nostoc azollae from Azolla accessions

Nostoc azollae cyanobionts were isolated from the dorsal foliar cavities of 48 Azolla accessions (Table 1) using the gentle roller method [54,55] with the following modifications. Roots were cut-off and sporophytes were disinfected in aqueous sodium hypochlorite (1 ml NaClO:10 ml distilled water, v:v) for 20 minutes, followed by three washes in ultrapure water (Millipore, Madrid, Spain). Sporophytes were sectioned and squeezed with a roller to separate the cyanobiont from Azolla cavities. The extract (Azolla+water+N. azollae) was centrifuged twice at 3000 g for 3 minutes to settle fern debris. The recovered supernatant with N. azollae filaments was centrifuged twice at 1000 g for 1 minute to free cyanobionts from the cellular debris. The recovered dark-green pellet was centrifuged at 11000 g for 10 minutes, stored at -20ºC, frozen at -80°C, lyophilized and weighted.

4.3. Detection and Analysis of BMAA (β-N-Methylamino-L-Alanine)

4.3.1. Method 1

The methodology, including reagents and materials, described by Baptista et al. (2015) [56] was used., with extraction of BMAA and quantification by LC-MS/MS using validated analytical methods [50,56]. Lyophilized Azolla and N. azollae (10 mg each sample) were acid-digested in 6 M HCl at 90°C for 20 minutes, using a high-pressure microwave system (Milestone-Ethos 1). After evaporation with nitrogen, 20 mM of HCl was added to samples and filtered (0.22 μm Millipore).

Analyses of BMAA by LC-MS/MS were performed in a Thermo LCQ Fleet Ion Trap LC/MSn system (Thermo Scientific) using a 2.1×100 mm, 5-μm diameter ZIC-HILIC column (SeQuant) and a 14×1 mm, 5-μm guard column (SeQuant). The mobile phase was acetonitrile (0.1% formic acid) and deionized water (0.1% formic acid). A linear gradient of 90% acetonitrile for 20 minutes was followed by 60% acetonitrile for 15 minutes and 90% acetonitrile for 5 minutes. The flow rate was 0.5 ml min^-1, injection volume was 10 μl, column temperature at 40°C and the positive mode on the electrospray ionization (ESI). Nitrogen was the sheath gas at a rate of 45 (unitless), and auxiliary gas at a rate of 20 (unitless). The capillary temperature was held at 250°C. Mass-to-charge ratio (m/z) scan was performed from 50 to 150 and the ion m/z 119 to assess 2,4-DAB (2,4-diaminobutyric acid) was monitored. The occurrence of the product ions m/z 102, 88 and 76 was verified at collision energy of 14 V for the presence of BMAA.

4.3.2. Method 2

Method 2 followed that described by Pravadali-Cekic S. et al. (2023) [49] with some modifications for the amount of starting material, chromatographic column, eluents and mode of mass detection. Lyophilized Azolla biomass, (100 mg) was dissolved in 3 ml of trichloacetic acid (TCA) 10% (v/v) and sonicated on ice (5 min, 70% amplitude, 20Hz), followed by overnight precipitation at 4°C. The mixture was then centrifuged (5000 g, 15 min, 4°C), the supernatant reserved and the pellet submitted to a second extraction cycle. The third extraction step used 10% TCA/acetone. The pellet as the bound fraction was transferred to a glass vial with acetone (100 %), centrifuged, and the supernatant added to the_free fraction. The pooled free fraction was then evaporated to dry in a SpeedVac and kept at -80°C. Pellets were also dried using the SpeedVac and acid hydrolysis by adding 3 ml of 6 M HCl overnight at 110°C. The hydrolyzed pellet was re-suspended in 1 ml ultrapure water and added to the free fraction. Samples were then filtered through a This was carried out with a 20 µl standard mix solution or sample extract, 20 µl of derivatizing reagent, and 60 µl of borate buffer. Following AccQ-Tag Ultra Derivatisation Kit in accordance with the manufacturer’s guidelines, the mixture was vortexed for several seconds and placed in a thermocycler at 55°C for 10 minutes. The final extract was then transferred to a 1.5mL vial for LC/MS/MS analysis.

Samples were injected in a Liquid Chromatograph Thermo Finnigan Surveyor HPLC System (Thermo Scientific, MA, USA), coupled with a Mass Spectrometry LCQ Fleet™ Ion Trap Mass Spectrometer (Thermo Scientific, MA, USA). XcaliburTM version 2. Mass Spectrometer Tune Method parameters optimization was used for data acquisition and processing using direct injection of BMMA and co-occurring isomers in a solution of 1 ppm in LCMS grade water (Table 2). The Mass Spectrometer operated in electrospray positive polarity mode using Collision Ionisation Dissociation (CID) corresponding to the [M+H]⁺ BMAA, AEG (N-(2-aminoethyl)-glycine), and 2,4-DAB molecules ion precursors and respective diagnostic fragments. The spray voltage was maintained at 3.5 kV, capillary temperature at 350°C, and capillary voltage at 20 kV and tube lens at 120 kV. Nitrogen was used as the sheath and auxiliary gas, with collision energy at 20 eV in Colission Induced Dissociation Mode. Separation was achieved on an ACE Excel C18 (50 × 2.1 mm I.D., 1.7 μm, Batch: V17-1253, Avantor^® ACE ^®, VWR, PT) at 18°C, with a flow rate of 0.3 ml/min injected at a volume of 10 μL in a no-waste mode. The eluents used were methanol (A) and water (B), both acidified with formic acid at 0.1% (v/v). The gradient program started at 13% A, increasing to 90% A in 20 minutes, turning back to initial conditions in 5 minutes, equilibrating for an additional 10 minutes with 20% A. See Table 2 for the chromatographic and mass parameters.

4.4. Cyanotoxin Genes at Azolla Accessions and Nostoc Azollae

4.4.1. DNA Extraction

DNA from N. azollae was extracted with PureLink^® Genomic DNA MiniKit (Invitrogen, Carlsbad, California, USA) and DNA from Azolla accessions was extracted with Genomic DNA from Plant NucleoSpin^® Plant II (Macherey-Nagel, Düren, Germany) according to the manufacturer instructions. DNA was stored at -20°C. The DNA was quantified in a Qubit fluorometer (Invitrogen) using the Quant-iT^® dsDNA HS assay following the manufacturer instructions. A working DNA concentration of 0.1 μg/μl was made with sterile ultrapure water, and saxitoxin, nodularin, and microcystin were assessed by specific primers (Table 3). A Biometra TProfessional (Goettingen) thermocycler was used for PCR amplification using the conditions listed in Table 4 for each gene, with a hold at 4°C for all the programs. Each 20 μl reaction contained 1 μl of 0.5 µM of each primer (Invitrogen), 2 μl of 0.1 μg/μl DNA, 9 μl Supreme NZYTaq 2x Green Master Mix (NZYTech) and 7 μl of ultrapure sterile water. Negative (with sterile ultrapure water) and positive (Microcystis aeruginosa LEGE91094 for microcystin and microcystin/nodularin genes, Aphanizomenon ovalisporum for cylindrospermopsin gene and Aphanizomenon gracillaris LMECYA 40 from INSA for saxitoxin gene) controls were included. The amplification products were separated in 1.5% agarose gel electrophoresis running in TAE 1x at 150 V, 25-30 min and stained with 0.2 μg/ml ethidium bromide (BioRad). The 1 Kb Plus DNA ladder (Invitrogen) was used as molecular size marker.

4.4.3. BLAST of Anatoxin-a Genes against Nostoc azollae

To determine if N. azollae produces anatoxin-a, a nucleotide BLAST (BLASTN) search for anatoxin-a coding genes was performed. Since the anatoxin-a gene cluster was discovered after the PCR and gel analysis of the other cyanotoxins performed in this study, analysis of its presence was made separately through BLAST rather than as the query sequence [57]. This was a 34682 bp sequence encoding for proteins associated with the biosynthesis of these toxins. The nucleotide query was applied to the full genome of Nostoc azollae 0708 (taxid: 551115). Matches with E-values from the BLASTN less than 0.01 were investigated and analyzed.

5. Conclusions

Our LC-MS/MS results show the Azolla-Nostoc azollae superorganism does not contain BMAA or their isomers DAB and AEG, and that Azolla and N. azollae do not synthesize other common cyanotoxins, indicating that Azolla is a nutritious food that can be safely eaten.