**Cylindrospermopsin- and Deoxycylindrospermopsin-Producing** *Raphidiopsis raciborskii* **and Microcystin-Producing** *Microcystis* **spp. in Meiktila Lake, Myanmar**

**Andreas Ballot 1,\*, Thida Swe 1,2,3, Marit Mjelde 1, Leonardo Cerasino 4, Vladyslava Hostyeva <sup>1</sup> and Christopher O. Miles <sup>5</sup>**


Received: 12 March 2020; Accepted: 3 April 2020; Published: 7 April 2020

**Abstract:** Meiktila Lake is a shallow reservoir located close to Meiktila city in central Myanmar. Its water is used for irrigation, domestic purposes and drinking water. No detailed study of the presence of cyanobacteria and their potential toxin production has been conducted so far. To ascertain the cyanobacterial composition and presence of cyanobacterial toxins in Meiktila Lake, water samples were collected in March and November 2017 and investigated for physico-chemical and biological parameters. Phytoplankton composition and biomass determination revealed that most of the samples were dominated by the cyanobacterium *Raphidiopsis raciborskii*. In a polyphasic approach, seven isolated cyanobacterial strains were classified morphologically and phylogenetically as *R. raciborskii*, and *Microcystis* spp. and tested for microcystins (MCs), cylindrospermopsins (CYNs), saxitoxins and anatoxins by enzyme-linked immunosorbent assay (ELISA) and liquid chromatography–mass spectrometry (LC–MS). ELISA and LC–MS analyses confirmed CYNs in three of the five *Raphidiopsis* strains between 1.8 and 9.8 μg mg−<sup>1</sup> fresh weight. Both *Microcystis* strains produced MCs, one strain 52 congeners and the other strain 20 congeners, including 22 previously unreported variants. Due to the presence of CYN- and MC-producing cyanobacteria, harmful effects on humans, domestic and wild animals cannot be excluded in Meiktila Lake.

**Keywords:** Meiktila Lake; *Raphidiopsis*; *Microcystis*; cylindrospermopsin; deoxycylindrospermopsin; microcystin

**Key Contribution:** This study confirmed the production of CYN and deoxyCYN by *Raphidiopsis raciborskii* strains and numerous MCs by *Microcystis* strains isolated from Meiktila Lake in Myanmar. The MCs included many novel congeners demonstrated by LC–MS and chemical derivatization methods. Among these were the rarely reported l-Glu and l-dihydrotyrosine-containing congeners. This is the first finding of toxin-producing cyanobacteria in a Myanmar waterbody.

### **1. Introduction**

Many lakes and reservoirs worldwide are affected by periodic cyanobacterial dominance or even cyanobacterial blooms. Such mass developments of cyanobacteria are typical for eutrophic conditions and are often induced by nutrient enrichment caused by increased agricultural, urban and industrial activities and are also expected to increase due to regional and global climate change [1]. Various cyanobacterial species forming such blooms are potential producers of hepatotoxic or neurotoxic compounds and their presence is often associated with animal poisonings and a threat to human health [2].

Myanmar is characterized by the presence of several natural lakes and numerous man-made reservoirs. Meiktila Lake is one of the numerous reservoirs in Myanmar and was built in ancient times, dating from an unknown period [3] but most likely in the reign of King Narapathisithu (1173–1210) [4]. Today the lake is divided by a dam into a northern and a southern part (Figure 1) [5]. Meiktila Lake is exposed to sedimentation due to deforestation in the catchment and especially the northern part has been partially filled with sediment over a period of more than 100 years [6]. The priority use of water from Meiktila Lake is drinking water, water for domestic purposes and for irrigation, although the lake is also polluted with domestic waste water, street runoff and solid waste [6,7].

**Figure 1.** Map of Meiktila Lake. The map shows the locations of water sampling (Stations MK1-MK5). The location of Meiktila Lake in Myanmar is shown in the inset.

Only limited information is available about the limnological characteristics of Meiktila Lake and other freshwater habitats in Myanmar. In 1995, a study of the algal flora of Meiktila Lake was reported [4]. A recent study described the investigation of physical parameters, macrophyte and phytoplankton

composition in the period 2011–2014 in Meiktila Lake [5]. Twenty taxa of aquatic macrophytes including helophytes have been documented in Meiktila Lake [5]. Several heterocytous cyanobacterial taxa, e.g., *Anabaena*, *Anabaenopsis* and *Calothrix* and a few nonheterocytous cyanobacterial taxa e.g., *Aphanocapsa*, *Chroococcus*, *Microcystis*, *Arthrospira* and *Oscillatoria* have been reported but not further investigated [4,5]. Neither study mentioned the presence of the cyanobacterium *Raphidiopsis raciborskii* (formerly *Cylindrospermopsis raciborskii*) (Woloszynska) Aguilera, Berrendero Gómez, Kastovsky, Echenique & Salerno nor the presence of the microcystin (MC) and cylindropsermopsin (CYN) groups of cyanobacterial toxins, which are documented from many lakes in Asia [8–10].

We suspect that the number of cyanobacterial species documented to date in Meiktila Lake was underestimated and that various toxin-producing cyanobacteria were present in the cyanobacterial community. There is clearly a lack of information about cyanobacteria and the production of cyanobacterial toxins in Meiktila Lake and the recently described potentially toxic cyanobacterium *Microcystis*is most likely not the only potential toxin-producing cyanobacterium. The ongoing pollution of the lake suggests the potential for occurrence of more frequent severe cyanobacterial blooms, which would have a negative impact on the use of the lake for drinking water and domestic purposes by the residents. This study aimed therefore to investigate the presence of cyanobacteria and their potential toxins in Meiktila Lake, applying modern analytical methods in a polyphasic approach to elucidate in detail the cyanobacterial composition, phylogeny and toxin production and toxin profiles.

### **2. Results**

### *2.1. Physico-Chemical Parameters*

At both sampling dates in March and November 2017, all sampling points in Meiktila Lake were characterized by water temperatures of 26.2–28.0 ◦C, pH of 8.5–9.3 and conductivities of 580–729 μS cm<sup>−</sup>1. Secchi depth was between 0.8 m (at sampling point MK1) and 1.8 m (at MK3). Total phosphorus and total nitrogen concentrations were 12–23 and 360–570 μg L<sup>−</sup>1, respectively.

### *2.2. Phytoplankton Community*

At all sampling stations, and at both sampling dates, cyanobacteria were the dominant group in the phytoplankton in both parts of Meiktila lake together with diatoms (Bacillariophyceae) Cryptophyceae, Chlorophyceae and Euglenophyceae (Table 1). The most dominant cyanobacterium was *R. raciborskii*, which comprised biomasses between 0.2 and 1.9 mg L−<sup>1</sup> fresh weight (FW), or 27%–91% of the cyanobacterial biomass at the sampling points MK1, MK2, MK4 and MK5. Other cyanobacteria present in the samples belonged to the genera *Aphanocapsa*, *Aphanothece*, *Chroococcus*, *Merismopedia*, *Limnothrix*, *Microcystis*, *Planktolyngbya*, *Planktothrix*, *Sphaerospermopsis* and *Synechococcus*. They together comprised biomasses between 0.06 and 1.2 mg L−<sup>1</sup> cyanobacterial wet weight. At MK3, the biomass of *R. raciborskii* at both sampling dates (0.02–0.08 mg L−1) was lower than at the other sampling points MK1, MK2, MK4 and MK5 (0.20–1.94 mg L−1) (data not shown). The *Microcystis* biomass was lower than the *Raphidiopsis* biomass at all sampling points and sampling dates and ranged from 0.003 to 0.16 mg L−<sup>1</sup> (data not shown).


**Table 1.** Biomass (mg L−<sup>1</sup> FW) of phytoplankton groups at sampling points MK1–MK5 in Meiktila Lake in March and November of 2017.

### *2.3. Morphological and Phylogenetic Characterization*

Seven potentially toxin-producing cyanobacterial strains were isolated from Meiktila Lake (Table 2). Based on morphological features, e.g., presence and form of colonies or filaments, vegetative cells and heterocytes, five of the isolated cyanobacterial strains were identified as *R. raciborskii* and two strains as *M. aeruginosa* and *M. novacekii*, respectively (Figure 2). However, Harke et al. [11] suggested all *Microcystis* warrant placement into the same species complex. Therefore, we use "*Microcystis*" instead of species names in the following parts of the manuscript.



**Figure 2.** Micrographs of cyanobacteria investigated in this study. (**a**) *Microcystis novacekii* (AB2017/14); (**b**) *Microcystis aeruginosa* (AB2017/15); (**c**) *Raphidiopsis raciborskii* (AB2017/05); (**d**) *Raphidiopsis raciborskii* (AB2017/09); (**e**) *Raphidiopsis raciborskii* (AB2017/12); (**f**) *Raphidiopsis raciborskii* (AB2017/13); (**g**) *Raphidiopsis raciborskii* (AB2017/16). Scale bars indicate 50 μm.

The *Raphidiopsis* strains were mostly characterized by straight tapered filaments. The filament length and width varied between 8.8–90 × 1.9–5.8 μm. Heterocytes were observed in some filaments of all isolated strains. Akinetes were not observed in any of the investigated strains. As in the cultured strains, only a few of the filaments possessed heterocytes in the environmental samples. The two *Microcystis* strains were characterized by cell diameters ranging from 4.2 to 6.6 μm (strain AB2017/15) and from 3.7 to 5.8 μm (strain AB2017/14) (data not shown).

The morphological determination of the isolated strains was supported by phylogenetic analyses (Figure 3; Figure 4). Phylogenetic relationships of the investigated strains are presented in the maximum-likelihood (ML) tree of the 16S rRNA gene of *Cylindrospermopsis*/*Raphidiopsis* (Figure 3) and a separate ML tree of the *Microcystis* 16S rRNA gene (Figure 4). In the ML tree in Figure 3, the *Raphidiopsis* strains from Meiktila Lake grouped together with 16S rRNA gene sequences derived from *Cylindrospermopsis* and *Raphidiopsis* strains from Asia, Europe, Africa, Australia and North America (cluster I). The CYN-producing and nonCYN-producing *Raphidiopsis* strains from Meiktila Lake could not be distinguished phylogenetically using 16S rRNA gene and had similar 16 rRNA gene sequences (Figure 3). In cluster II, strains from North and South America, (USA, Mexico, Brazil), North

Africa (Tunisia), Southwest Europe (Spain) and New Zealand, grouped together. Both *Microcystis* strains from Meiktila Lake possessed similar 16S rRNA gene sequences and clustered together with 16S rRNA gene sequences of *Microcystis* from Europe, Asia, Africa and South America (Figure 4).

**Figure 3.** ML tree based on partial 16S rRNA gene sequences of 40 *Raphidiopsis*/*Cylindrospermospis* strains. Outgroup = *Sphaerospermopsis aphanizomenoides* (LN846954). Cluster I includes *Cylindrospermopsis* and *Raphidiopsis* strains from Asia, Europe, Africa, Australia and North America, cluster II includes strains from North and South America (USA, Mexico, Brazil), North Africa (Tunisia), Southwest Europe (Spain) and New Zealand. Strains from this study are marked in bold. Bootstrap values above 50 are included. The scale bar indicates 0.5% sequence divergence.

**Figure 4.** ML tree based on partial 16S rRNA gene sequences of 40 *Microcystis* strains. Outgroup = *Chroococcus subviolaceus* (MF072353). Strains from this study are marked in bold. Bootstrap values above 50 are included. The scale bar indicates 2% sequence divergence.

.

### *2.4. Identification of Cyanobacterial Toxins and Toxin-Producing Strains*

Three of the five investigated *Raphidiopsis* strains produced CYNs in variable amounts by either enzyme-linked immunosorbent assay (ELISA) or liquid chromatography with tandem mass spectrometry (LC-MS/MS) (Table 3). Concentrations of CYNs were 1.8–4.3 μg mg−<sup>1</sup> FW by ELISA. Using LC–MS/MS, CYN concentrations of 1.7–2.5 μg mg−<sup>1</sup> FW and deoxyCYN from 1.3 to 7.3 μg mg−<sup>1</sup> FW were detected. In the three CYN-producing strains, deoxyCYN comprised 43%–75% of the total CYNs.

**Table 3.** Concentrations (μg mg-1 FW) of CYNs by ELISA and of CYN and deoxyCYN by LC–MS/MS in cultured *R. raciborskii* strains isolated from Meiktila Lake\*.


\*- = biomass not determined; n.d. = not detected; FW = fresh weight; percentages are of total CYNs by LC–MS/MS.

All investigated *Raphidiopsis* strains tested negative for saxitoxins (STXs), anatoxins (ATXs) and MCs by ELISA. Both *Microcystis* strains tested negative for CYNs, STXs and ATXs by ELISA but were identified as MC-producers by ELISA and their MC profiles were therefore investigated by high resolution LC–MS/MS (LC–HRMS/MS).

Underivatized samples were analysed by LC-HRMS/MS in positive and negative ionisation modes as previously described [12,13] and then after reaction with mercaptoethanol (targeting Mdha7/Dha<sup>7</sup> moieties in MCs) [14] and Oxone/DMSO (targets sulfide groups in methionine and Cys/GSH conjugates of MCs) [15]. Results of these analyses are summarised in Figure 5, Table 4 and Table S1. Peaks from putative MCs were identified by their reaction with mercaptoethanol, production of characteristic product ions in data-dependent and/or data-independent acquisition (DDA and/or DIA) LC-MS/MS screens, and possessing plausible potential elemental formulae based on both positive and negative mode full scan HRMS. These peaks were then targeted by LC-HRMS/MS at suitable collision energies to obtain structurally informative HRMS/MS spectra to assist with identification, compared by LC–HRMS with samples containing some of the putative MCs, and subjected to selective oxidation to detect the presence of sulfide moieties that could be present in some of the MCs.

Peaks were only considered to be MCs if they: 1, showed apparent pseudo-molecular ions appropriate for a MC in both positive and negative ionisation modes; 2, showed one or more of the characteristic MC fragments shown in Figure 5; 3, displayed appropriate chemical reactivity for the putative structure, and; 4, displayed retention times (*t*R) and charge states (*z*) appropriate to the putative structure (e.g., based on the apparent number of polar and charged residues, such as Arg). MCs were considered "confirmed" (**1**, **3**, **13**, **14**, **17**, **18**, **21**, **25**, **26** and **41**) if they behaved identically in all respects to the standards (Table 4). Structures were considered "probable" if they behaved identically in all respects to a compound already identified with high probability in an available sample (**2**, **4**, **8**, **11**, **12**, **20**, **28**, **29**, **36**, **45** and **50**). For compounds for which standards or appropriate samples were not available, these were regarded as "probable" if, in addition to displaying the appropriate physical and chemical characteristics (Table 4 and Table S1), they also displayed interpretable MS/MS spectra that were clearly consistent with the proposed structure by comparison with related compounds (**5**–**7**, **9**, **10**, **15**, **16**, **24**, **27**, **30**–**32**, **37**–**40**, **42**–**44**, **46**–**49**, **51**, **53** and **54**). Compounds were considered tentative if there was limited MS/MS spectral evidence (**23**) or if the evidence was ambiguous (e.g., several isomers were present that showed indistinguishable MS/MS spectra, i.e., **52**, **55** and **56**). Compounds designated "unidentified" were definitively identified as MCs, but the spectral data was insufficient

to identify them (**19**, **22** and **33**–**35**). All compounds listed in Table 4 as containing Mdha<sup>7</sup> and which gave adequate signal-to-noise in their MS/MS spectra in positive mode, showed product ions at *m*/*z* 135.0804, 375.1914 and 446.2286, indicative of the presence of Adda5–d-Glu6–Mdha–d-Ala1, while in those listed as containing Mser<sup>7</sup> or Dha7 the latter two product ions were heavier, or lighter, by a mass corresponding to H2O or CH2, respectively (all with Δ*m* < 5 ppm), and the presence of these units is implicit in the discussion of the structural elucidation in Section 3.

**Figure 5.** Structures and exact negative ionisation *m*/*z* of microcystins (MCs) identified in cultures AB2017/14 and AB2017/15 in this study, showing characteristic product ions at *m*/*z* 135.0804 (positive) and 128.0353 (negative) LC–MS/MS spectra (see Table 4 and Table S1). The origins of additional positive mode product ions containing Mdha7 (R<sup>3</sup> = Me) at *m*/*z* 375.1914 and 446.2286 are also shown. Note that the corresponding product ions containing Dha7 (R3 = H) have *m*/*z* 361.1758 and 432.2129, and *m*/*z* 393.2020 and 464.2391 for Mser<sup>7</sup> (R<sup>4</sup> = OH). A full version of this table including positive ionisation data is shown in the Supporting Information (Table S1). Abbreviations: Abu, aminobutyric acid; Dha, dehydroalanine; (H2)Tyr, dihydrotyrosine; (H4)Tyr, 4,5,6,7-tetrahydrotyrosine; Kyn, kynurenine; Mdha, *N*-methyldehydroalanine; Mlan, *N*-methyllanthionine; Mser, *N*-methylserine; Oia, oxindolyalanine; Tyr(OMe), methoxytyrosine.

**Table 4.** Identities of microcystins detected by LC-HRMS/MS analysis in *Microcystis* strains AB2017/14 and /15 isolated from Meiktila Lake, their retention times (*t*R), concentrations, relative abundances (%) and observed *m*/*z* values in negative ionisation mode*a.*



**Table 4.** *Cont.*

*<sup>a</sup>* A comprehensive version of this table, including positive and negative ionisation MS data, reactivity towards thiols and mild oxidising agents, number of rings plus double-bond equivalents (RDBE) and presence of characteristic ions observed in positive and negative ionisation MS/MS spectra, is in the Supporting Information (Table S1) together with LC–HRMS/MS spectra (Figures S1–S59). *<sup>b</sup>* Concentration expressed per weight of biomass (FW) and as a percentage of total microcystins detected in each culture); ND = not detected; *<sup>c</sup>* Reacted with mercaptoethanol; *<sup>d</sup>* Oxidised by Oxone/DMSO; *<sup>e</sup>* HRMS/MS spectrum of oxidation product obtained.

In culture AB2017/14, 52 microcystin congeners were detected by LC–HRMS, with a total concentration of 1100 μg g−<sup>1</sup> FW. Twenty-one of these were unidentified or previously unreported variants. In culture AB2017/15, 20 microcystin variants (of which six were previously unreported) were detected, with a total concentration of 14000 μg g−<sup>1</sup> FW. The microcystin variants and the concentrations found in each strain are shown in Table 4 and Table S1.

### **3. Discussion**

This study clearly demonstrates for the first time the presence of CYN- and deoxyCYN-producing *R. raciborskii* and MC-producing *Microcystis* in the phytoplankton community of Meiktila Lake in Myanmar. The relatively high biomass of *R. raciborskii*, up to 1.9 mg L−<sup>1</sup> in the phytoplankton community of Meiktila Lake, is expected to cause elevated concentrations of CYNs in the lake water. The results suggest a higher risk for humans and animals to be affected by CYNs than by MCs, although this could be affected by variations in the biomass of CYN-producing *R. raciborskii* versus MC-producing *Microcystis*. Variations in cyanobacterial bloom composition and toxin production are influenced by abiotic factors such as nutrients, temperature and light and by biotic factors such as grazing, parasitism and predation [16,17]. The distribution of CYN/deoxyCYN-producing and nonproducing *Raphidiopsis* strains in Meiktila Lake is likely to vary over time, and dominance by a *Raphidiopsis* strain such as AB2017/13 would lead to CYN and deoxyCYN concentrations up to 20 μg L−<sup>1</sup> for the highest *Raphidiopsis* biomasses measured in this study. It is therefore expected that CYN/ deoxyCYN concentrations in the lake water will at times exceed the guideline value for CYN in drinking water of 1 μg L−<sup>1</sup> [18]. The tolerable daily intake value of 0.03 μg kg−<sup>1</sup> for a person of 70 kg body weight would be exceeded after the intake of slightly more than 100 mL of lake water if CYN and deoxyCYN are similarly toxic. However, the toxicity of deoxyCYN to humans is not yet clear. According to Norris et al. [19], deoxyCYN does not contribute significantly to the toxicity of *R. raciborskii*. In contrast, cell viability assays showed that deoxyCYN was only slightly less toxic than CYN and most likely operates by similar toxicological mechanisms [20]. The potential risk of deoxyCYN for humans needs therefore to be clarified [20]. The use of Meiktila Lake water for drinking water, irrigation, domestic purposes or animal consumption is complicated by the fact that an unknown proportion of CYNs can be extracellular and is therefore not eliminated by filtration. Lake water contaminated by CYN and other toxins like MC-LR can lead to morphological and physiological changes and potential loss of productivity by agricultural plants, and bioaccumulation of cyanotoxins in the tissues of edible terrestrial plants in a concentration-dependent manner has been reported [21].

Griffith and Saker [22] have shown that in stationary phase of cultures, more than 50% of CYN can be extracellular. In environmental samples, the same authors found that extracellular CYN could exceed 90%. Boiling in water does not significantly degrade CYN within 15 min [23]. The removal of extracellular CYN/deoxyCYN therefore needs other methods, like the use of activated carbon, membrane filtration or chemical inactivation (Ultraviolet (UV), or oxidants) [24]. The presence of *Raphidiopsis* strains in Meiktila Lake that do not produce CYN makes it likely that CYN concentrations in the lake will vary considerably depending on the ratio of the two chemotypes in the phytoplankton community. As Meiktila Lake water is used for domestic purposes (drinking water, irrigation, washing of clothes, and personal hygiene), regular monitoring of cyanobacterial biomass and CYNs is recommended.

*R. raciborskii* has not been described from Myanmar water bodies and was not observed in a phytoplankton community study conducted in Meiktila Lake from 2011 to 2012 [5]. *Raphidiopsis* (and *Cylindrospermopsis*) spp., however, have been described from various other Southeast Asian freshwater habitats [25–27] and other water habitats worldwide [8]. *R. raciborskii* is only known to produce CYNs in Australia and the Asian countries China, Japan, Vietnam and Thailand and to produce STXs in Brazil [26,28–32]. The prime radiation centre of *R. raciborskii* is thought to be in Africa, with a second radiation centre in Australia [8]. Our 16S rRNA gene analysis confirms the close relationship of the *Raphidiopsis* strains from Meiktila Lake to other *Raphidiopsis* strains from Asia, Europe and Australia. Our 16S rRNA gene tree also clearly supports the suggested movement of *Raphidiopsis* from the American continents to Southwest Europe and North Africa and probably further to Greece and China, as has been described [33,34]. The close relationship of *Raphidiopsis* strains from Australia, Asia and Europe does not, however, explain why CYN- and deoxyCYN-producing strains have only been found in Asia and Australia but not in Europe, Africa or the Americas. Parts of, or the whole, CYN gene cluster could have been lost during the spread from Asia westwards, or only nonCYN-producing strains may have spread to Europe. Our finding of nontoxic *Raphidiopsis* strains in Meiktila Lake supports the latter hypothesis.

Both *Microcystis* strains AB 2017/14 and AB2017/15 isolated from Meiktila Lake are confirmed microcystin producers and are closely related to *Microcystis* strains from Africa, Europe and Asia based on 16S rRNA gene phylogeny. Both strains had identical 16S rRNA gene sequences but were clearly distinguished chemically by their MC congener profiles. Fifty-six microcystin variants were found in *Microcystis* strains AB2017/14 and AB2017/15 isolated from Meiktila Lake (Figure 6).

**Figure 6.** LC–HRMS full scan extracted ion chromatograms (3.8–19.4 min, positive ionization mode) of extracts from (**A**) *Microcystis* culture AB2017/14 and (**B**) *Microcystis* culture AB2017/15. Chromatograms were produced by extracting at *m*/*z* (± 5 ppm) for all MCs listed in Table 4 (see Table S1 for positive ionisation *m*/*z* values). Note that some of the smaller peaks are not labelled on the chromatograms, and the peak marked with an asterisk is not from a MC.

In order to reliably estimate the quantities of the MCs in the extracts by LC-HRMS, it was necessary to characterize, and if possible, identify all of them. The reason for this is that the response in LCMS can be expected to vary from congener-to-congener, primarily due to variations in the number of easily ionisable amino acid residues, especially Arg, present in the MC's structure. Only the identification of previously unreported MCs (see Bouaïcha et al. [35]) in the cultures is discussed further (i.e., **7**, **9**, **19**, **22**, **24**, **27**, **30**–**35**, **37**, **38**, **40**, **42**, **46**, **51**–**53**, **55** and **56**) but spectra of all compounds for which adequate MS/MS spectra were obtained are available in the Supporting Information.

Three of the compounds were sulfide-containing variants (**5**, **15**, and **31**) which reacted when the extract was oxidised with Oxone/DMSO (Table 4). The first two have been reported in cultures and blooms [15,36], and their characteristics were fully consistent with those reported for **5** and **15** here, and in the case of **15** its oxidation product (MC-M(O)R) showed characteristic product ions including neutral loss of CH4OS and displayed an MS/MS spectrum (Figures S8, S11 and S14) identical to that reported previously for MC-M(O)R [15]. The third of sulfide-containing MC was identified as MC-RM (**31**) based on its physical and chemical properties (Table 4), which were essentially identical to those of **15** except for its longer *t*<sup>R</sup> and that its MS/MS spectrum closely paralleled that of MC-RA (**28**) and displayed product ions characteristic of an MC with one Arg at position-2 rather than at the more common position-4 (Figures S8, S14 and S15). For example, fragments at *m*/*z* 440.2263 (C18H30O6N7 +, Δ*m* = 2.2 ppm, from Mdha7–d-Ala1–Arg2–d-Masp3) and 731.3716 (C34H51O10N8 <sup>+</sup>, <sup>Δ</sup>*m* = <sup>−</sup>0.9 ppm, from Adda5–d-Glu6–Mdha7–d-Ala1–Arg2–d-Masp3), together with the complete absence of a product ion at *m*/*z* 599.3552 (from Arg4–Adda5–d-Glu6), confirmed Arg at position-2 and Met at position-4 (see Okello et al. [37] for assigned product ions from MC-YR) of **31**.

Eight of the compounds (**4**, **9**, **24**, **27**, **36**–**38** and **40**) showed characteristics of MCs containing one or more Glu residues at position-2 or -4. Two of these (**9** and **27**) had formulae consistent with MC-RE or MC-ER (Table 4). Compound **9** gave product ions (Figures S2 and S5–S7) typical of a MC with Arg at position-4, including *m*/*z* 599.3536 (C31H47O6N6 +, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>2.6 ppm, from Arg4–Adda5–d-Glu6), 284.1238 (C12H18O5N3 <sup>+</sup>, <sup>Δ</sup>*m* = <sup>−</sup>1.1 ppm, from Mdha7–d-Ala1–Glu2), and 286.1497 (C11H20O4N5 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>4.3 ppm, from <sup>d</sup>-Masp3–Arg4), indicating **9** to be MC-ER. The MS/MS spectrum of **27** (Figures S2 and S15) included product ions at *m*/*z* 440.2248 (C18H30O6N7 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>0.9 ppm, from Mdha7–d-Ala1–Arg2–d-Masp3), and 731.3678 (C34H51O10N8 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>6.1 ppm, from Adda5–d-Glu6–Mdha7–d-Ala1–Arg2–d-Masp3) which, together with the complete absence of a product ion at *m*/*z* 599.3552, confirmed Arg at position-2 and Glu at position-4 (see Okello et al. [37]) for assigned product ions from MC-RY), showing **27** to be MC-RE. The characteristics of **37** (Table 4) were consistent with MC-EE. In addition, **37** gave product ions (Figures S27–S33) including *m*/*z* 276.1189 (C10H18O6N3 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>0.4 ppm, from <sup>d</sup>-Masp3–Glu4), 405.1605 (C15H25O9N4 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>2.9. ppm, from Glu2–d-Masp3–Glu4) and 575.2703 (C28H39O9N4 <sup>+</sup>, Δ*m* <sup>=</sup> <sup>−</sup>1.5 ppm, from Adda5–d-Glu6–Mdha7–d-Ala1–Glu2) that confirmed **<sup>37</sup>** as MC-EE. Compound **<sup>40</sup>** was identified as MC-LE based on the characteristics presented in Table 4, as well as product ions (Figures S27–S33) observed in its MS/MS spectra, including *m*/*z* 460.2397 (C19H34N5O8 +, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>1.2, from <sup>d</sup>-Ala1–Leu2-d-Masp3–Glu4) and 397.2073 (C18H29N4O6 <sup>+</sup>, <sup>Δ</sup>*m* = <sup>−</sup>2.2, from Mdha7–dAla1–Leu2-d-Masp3), confirmed its identity as MC-LE (**40**). Earlier-eluting desmethylated d-variants of **9**, **27**, **37** and **40** were similarly identified as the corresponding d-Asp3-congeners [d-Asp3]MC-ER (**4**), [d-Asp3]MC-RE (**24**), [d-Asp3]MC-EE (**36**) and [d-Asp3]MC-LE (**38**) based on analysis of their MS/MS spectra (Figures S21–S27) and characteristics presented in Table 4. Furthermore, the **4** and **36** in this sample coeluted with, and gave identical product ion spectra to, [d-Asp3]MC-ER (**4**) and [d-Asp3]MC-EE (**36**) identified [12] in an extract of a culture of *Planktothrix prolifica* NIVA-CYA544.

Twenty-one conventional late-eluting nonArg-containing MCs (**30**, **32** and **36**–**56**) were detected. Of these, the identities of four that contained Glu2 or Glu<sup>4</sup> (**36**–**38** and **40**) were discussed above. The remaining previously unreported nonArg MCs were **30**, **32**, **42**, **46**, **51**–**53**, **55** and **56**. Compound **42** had the same characteristics as MC-LAbu (**45**) (Table 4), however, its MS/MS spectrum (Figures S34–S36) was consistent with MC-HilA. In particular, product ions at *m*/*z* 573.3270 (C30H45N4O7 +, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>2.3 ppm, from Adda5–d-Glu6–Mdha7–d-Ala1–Hil2 minus C9H10O (cf *<sup>m</sup>*/*<sup>z</sup>* 559.3126 for **<sup>41</sup>** and **45**)) and 411.2231 (C19H31N4O6 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>1.6 ppm, from Mdha7–d-Ala1–Hil2-d-Masp<sup>3</sup> (cf *<sup>m</sup>*/*<sup>z</sup>* 397.2082 for **41** and **45**)) as well as a range of other ions indicated the identity as MC-HilA (**42**), although the actual connectivity of the carbons in the amino acid side-chain at position-2 cannot be determined by mass spectrometry. A related compound (**46**) had characteristics (Table 4) and gave product ions (Figures S40–S42) that were consistent with MC-HilAbu. Product ions included *m*/*z* 573.3260 (C30H45N4O7 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>3.9, from Adda5–d-Glu6–Mdha7–d-Ala1–Hil2 minus C9H10O), 232.1291 (C9H18N3O4 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>0.4, from <sup>d</sup>-Masp3–Abu4) and 430.2651 (C19H36N5O6 <sup>+</sup>, <sup>Δ</sup>*m* = <sup>−</sup>2.1, from d-Ala1–Hil2-d-Masp3–Abu4 (cf *m*/*z* 402.2347 for **41** and 416.2504 **42**). This data unambiguously shows the presence of an extra CH2 group in both amino acid-2 and -4 in **46**, relative to MC-LA (**41**) and is consistent with MC-HilAbu (**46**). Compound **51** displayed characteristics consistent with MC-FV (Table 4), as well as product ions (Figures S49–S51) at *m*/*z* 246.1456 (C10H20N3O4 <sup>+</sup>, <sup>Δ</sup>*m* = <sup>−</sup>0.4, from d-Masp3–Val4), 593.2960 (C32H41N4O7 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>1.6, from Adda5–d-Glu6–Mdha7–d-Ala1–Phe2 minus C9H10O (cf. 559.3126 for **41**)) and 464.2494 (C22H34N5O6 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>3.1, from <sup>d</sup>-Ala1–Phe2 –d-Masp3–Val4 (cf. 402.2347 for **41**)). This establishes an extra C5H2 and 4 RDBE in amino acid-2 and C2H4 in amino acid-4, relative to MC-LA (**41**), consistent with MC-FV (**51**). Compound **53** had characteristics consistent with MC-WV (Table 4). This was supported by its MS/MS spectra (Figures S49–S51), which included product ions at *m*/*z* 246.1456 (C10H20N3O4 <sup>+</sup>, Δ*m* = 2.9, from d-Masp3–Val4), 632.3063 (C34H42N5O7 +, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>2.5, from Adda5–d-Glu6–Mdha7–d-Ala1–Trp2 minus C9H10O (cf. 559.3126 for **<sup>41</sup>**)) and 503.2605 (C24H35N6O6 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>2.6, from <sup>d</sup>-Ala1–Trp2–d-Masp3–Val4 (cf. 402.2347 for **<sup>41</sup>**)). This indicates the presence of an extra C5HN and 6 RDBE in amino acid-2 and C2H4 in amino acid-4, relative to MC-LA (**41**), consistent with MC-WV (53). Later-eluting isomers of **50**, **51** and **54** were also present (i.e., **52**, **55** and **56**), with identical characteristics (Table 5) and product ion spectra (Figures S52–S56). These compounds all contain branching amino acids at the variable position-2 (nominally Leu for **52** and **56**) or -4 (Val for **55**), and most likely the isomers present result from changes to this branching (e.g., Ile or 2-aminohexanoic acid at position-2, and 2-aminopentanoic acid or isovaline at position-4).


**Table 5.** Sampling points and sampling depth in Meiktila Lake for chemical and biological measurements.

In addition, **30** and **32** differed from each other by CH2 and had characteristics consistent with MC-(H2)YA (**32**) and MC-(H2)YG (**30**), respectively (Table 4). These fragmented somewhat differently from typical Arg-freeMCs such asMC-LA (**41**) (Figure S17). Both compounds showed weak product ions at *m*/*z* 155.0815 and 580.3017, indicating that **30** and **32** both contained Adda5–d-Glu6–Mdha7–d-Ala1. However, both **30** and **32** also gave product ions at *m*/*z* 320.1605, 611.3075 and 745.3807 (cf. 268.1650, 559.3117 and 693.3854 in MC-LA (**41**)), indicating the presence of an extra C3O and 3 RDBE at amino acid-2 relative to **41**, consistent with the presence of the unusual amino acid l-dihydrotyrosine ((H2)Y) at position-2. Compounds **30** and **32** gave product ions as *m*/*z* 449.2033 and 449.2015 (C21H29N4O7 +, <sup>Δ</sup> 0.5 and <sup>−</sup>3.6 ppm, respectively, from Mdha7–d-Ala1–(H2)Tyr2–d-Masp3; cf. *<sup>m</sup>*/*<sup>z</sup>* 397.2082 for **<sup>41</sup>**, from Mdha7–d-Ala1–Leu2–d-Masp3). Thus, the difference in mass (14.0157, i.e., CH2) between **30** and **32** lies not in residue-3 (d-Masp<sup>3</sup> vs d-Asp3) as might be expected but in residue-4. Thus, **32** is identified as MC-(H2)YA, and **30** as MC-(H2)YG, which appears to be the first MC so far reported [35] with Gly at position-4.

Compounds **6** and **7** coeluted but differed by a mass corresponding to CH2, with **6** having the same accurate mass, MS/MS spectrum (Figure S3) and retention time as MC-(H2)YR (**6**) identified in a sample from a recent study [13]. The MS of **7** was consistent with [d-Asp3]MC-(H2)YR, and the MS/MS spectra of **6** and **7** were very similar. The MS/MS spectrum of **7** included product ions (Figure S4) at *m*/*z* 155.0813 (Δ*<sup>m</sup>* <sup>=</sup> <sup>−</sup>1.6 ppm, from Mdha7–d-Ala1) and 599.3525 (C31H47O6N6 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>4.5 ppm, from Arg4–Adda5–d-Glu6), 120.0806 (C8H10N<sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>1.6 ppm, from (H2)Tyr), and 320.1611 (C16H22O4N3 +, Δ*m* = 2.0 ppm, from Mdha7–d-Ala1–(H2)Tyr2), 272.1353 (C10H18O4N5 <sup>+</sup>, Δ*m* = 2.0 ppm, from d-Asp3–Arg4) and 714.3802 (C35H52O9N7 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>2.7 ppm, from <sup>d</sup>-Asp3–Arg4–Adda5–d-Glu6), showing that **7** is [d-Asp3]MC-(H2)YR.

Five MCs were present (**19**, **22** and **33**–**35**) whose structures could not be identified from their characteristics (Table 4 and Table S1) or product ion spectra (Figures S12 and S18–S20).Compounds **33** and **34** gave apparent *m*/*z* values that did not correspond to known or plausible MC variants and differed from each other by a mass corresponding to CH2. They did not contain Arg but contained one extra nitrogen atom, 5 extra RDBE and, more surprisingly, one less oxygen atom than MC-LA (**41**). Compound **33** gave product ions at *m*/*z* 375.1911 (C20H27O5N2 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>0.9 ppm, from Adda5–d-Glu6–Mdha7 minus C9H10O), 446.2284 (C23H32N3O6 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>0.4 ppm, from Adda5–d-Glu6–Mdha7–d-Ala1 minus C9H10O) and 580.2986 (C32H42N3O7 <sup>+</sup>, <sup>Δ</sup>*<sup>m</sup>* <sup>=</sup> <sup>−</sup>5.4 ppm, from Adda5–d-Glu6–Mdha7–d-Ala1), however, fragments containing amino acids 2–4 were either shifted or absent. Although the data appear to be consistent with analogues containing Orn and Phe at positions 2 and 4 with an amide linkage to the neighbouring d-Asp3/Masp3 residue, further data are required for even a tentative structural assignment. The apparent pseudomolecular ion isotope envelopes of **33** and **34** (Figure S60) displayed unusual patterns that suggest that these compounds may be reacting during ionisation, possibly including dehydration, further complicating mass spectral analysis.

As with **30** and **32**, compound **35** showed product ions at *m*/*z* 213.0866, 320.1583, 375.1909, 446.2279, 449.2013, 509.2680 and 611.3040, consistent with the presence of Adda5–d-Glu6–Mdha7–d-Ala1–(H2)Tyr2–d-Masp3, which would require an MC with an amino acid side-chain at position-4 possessing an unprecedented C3H3 and 2 RDBE. This could possibly be due to the presence of a larger fragile amino acid at this position that undergoes ready elimination during MS, and the full structure of **35** remains undetermined.

Compounds **19** and **22** had identical product ion spectra including ions at *m*/*z* 135.0803, 213.0866, 269.1235, 375.1907, 446.2281, 599.3533, 640.2807 and 710.3886, identical to those from MC-LR, indicating the presence of d-Masp3–Arg4–Adda5–d-Glu6–Mdha7–d-Ala<sup>1</sup> and that these compounds differed from MC-LR (**17**) and MC-WR (**26**) only in the amino acid at position-2. Product ions at 469.1815 and 486.2111 (both from X2–d-Masp3–Arg4) and 173.0709 and 144.0455 (from X2 and 173.0709−CH3N) were consistent with this and indicated the presence of a side chain at position-2 containing C9H6NO and 7 RDBE (cf. C9H8N and 6 RDBE for the side chain of Trp in **26**), suggesting the presence of an unidentified oxidised variant of Trp present in the two isomeric MCs, **19** and **22**.

Between them, the two cultures contained 34 known and 22 previously unreported MCs of which 14 were assigned probable structures based on their chemical and mass spectrometric properties. In all cases, the number of Arg residues present in the MCs were reliably determined from their elemental compositions and charge-state, even for congeners for which definitive structures could not be established. Quantitation of the individual congeners was then performed from negative (Table 4) and positive mode full scan chromatograms, which gave essentially identical results, relative to 3-point calibration curves of the appropriate MC reference materials (RMs) containing no, one or two Arg residues ([d-Leu1]MC-LY [38], MC-LR (**17**) and MC-RR (**3**), respectively). AB2017/14 contained 52 MCs, of which only 31 had been previously reported, and the newly reported MCs constituted nearly 20% of the total identified MC content. AB2017/15 contained fewer MCs (20) of which only six had not been reported previously, but these constituted nearly 34% of the total MC content in this culture. Notable amongst the newly reported MCs are those containing Glu at positions 2 and 4 (**4**, **9**, **24**, **27**, **36**–**38**, and **40**), which constituted 1.7% of the MCs in AB2017/14 and 37.5% of the MCs in AB2017/15. MCs

of this type have only been reported previously from two sources [12,39]. AB2017/14 also contained a number of congeners that appeared to be derived from MC-WR (**19**, **20**, **22**, **23** and **26**) that together constituted 20.2% of its MC content, and an unusually high number (25) of nonArg-containing MCs (**30**, **32**–**35** and **37**–**56**) together constituting 20.5% of the total MC content. These results underscore both the diversity of MCs that may be present in a single sample and the potential difficulty of reliably quantitating the total MCs using traditional targeted LC–MS/MS methods. These factors may contribute to the reported apparent overestimation of MC levels when using less-targeted methods such as ELISA and PP2A inhibition, relative to highly congener-targeted LC–MS/MS approaches [40].

Although both strains produce a variety of MC variants, the risk of harmful effects caused by microcystins is likely to be low due to the low *Microcystis* biomasses observed in Meiktila Lake during this study. An increase in *Microcystis* biomass cannot, however, be excluded due to the pollution from various sources, e.g., waste water and street runoff [6]. This will most likely lead to an increase in microcystin concentrations in the lake, with potential harmful effects on humans, domestic and wild animals using the untreated lake water. The toxicity of microcystins found in the strains isolated from Meiktila Lake varies from highly toxic variants like MC-LR or MC-LA to less toxic variants such as MC-RR [41]. The toxicity of most of the MC variants found in this study has not yet been described, which makes a full risk assessment difficult. The production of a high number of MC variants (up to 47) has also been shown for two other *Microcystis* strains isolated from South African Haartbeestpoort Dam and Japanese Lake Kasumigaura [42,43].

Shallow lakes like Meiktila Lake are often characterized by competition between macrophytes and phytoplankton. High nutrient loading and phytoplankton growth lead to turbid water conditions and prevent the growth of macrophytes [44]. In Meiktila Lake the present turbid conditions are explained by deforestation and erosion in the catchment area, pollution with domestic waste water, street runoff and solid waste [6]. The relatively strong growth of *R. raciborskii* most likely is an additional reason for the observed turbidity at sampling points MK1, MK2, MK4 and MK5 in Meiktila Lake. The *Potamogeton* belt which separates MK2 and MK3 seems to act as a kind of filter or barrier because the Secchi depth at MK3 (1.8 m) was considerably higher than at MK1 (0.8 m). According to Van Donk and Van de Bund [45], macrophytes significantly modify the composition of the phytoplankton community and lead to a decrease in its abundance and biomass. It has been shown that certain macrophyte species exhibit allelopathic activity against certain phytoplankton species [46]. The obvious decrease of *Raphidiopsis* biomass from MK2 to MK3 may be therefore attributable to allelopathic effects of the macrophyte species in the belt. Allelopathy has been suggested by several authors to be responsible for observed phytoplankton patterns in whole-lake studies of vegetated, shallow lakes, but evidence for or against allelopathy has not been provided [47–49].

### **4. Conclusions**

In conclusion, this is the first report of *R. raciborskii* and *Microcystis* in Meiktila Lake, Myanmar. Three of five *Raphidiopsis* isolates produced CYN and deoxyCYN, like other *Raphidiopsis*/*Cylindrospermopsis* isolates described from other Asian countries and Australia. Both *Microcystis* strains isolated from Meiktila Lake produced at least 56 MC variants (52 for AB2017/14 and 20 for AB2017/15), including 22 previously undescribed congeners. Harmful effects on humans and animals using Meiktila Lake as a water source cannot therefore be excluded.

### **5. Materials and Methods**

### *5.1. Study Area, Measurements and Sampling*

Meiktila Lake is a shallow reservoir, located close to Meiktila city in central Myanmar in the Mandalay region at an altitude of 230 m (Figure 1). During the rainy season, from April/May to October/November, it receives water from Mondaing Dam, located ca. 15 km west of Meiktila at an altitude of 245 m. Depending on the season, Meiktila Lake covers an area of around 54 km2 with a maximum water depth of 10 m. It is divided by a dam into a northern and a southern lake [5]. Five sampling points were selected, three in the northern part of Meiktila Lake and two in the southern part (Figure 1, Table 5).

Sampling points 2 and 3 in the Northern Part of Meiktila Lake were separated by a broad *Potamogeton* belt. At all five sampling points in March and November 2017, in situ measurements of water temperature, pH, conductivity and dissolved oxygen were conducted, and integrated water samples were taken (1 m steps up to max. 3 m water depth) for the analysis of chemical parameters (ammonium, nitrate, total nitrogen, soluble reactive phosphorous, total phosphorous, Ca, phytoplankton composition and biomass and for the isolation of cyanobacterial strains. For quantitative phytoplankton analysis, a 50 mL subsample was removed from a sample taken from integrated samples and preserved with Lugol's solution and a concentrated net sample (mesh size 20 μm) was taken and preserved by addition of formaldehyde (4% final concentration) for qualitative analysis. A 50 mL water sample for isolation of cyanobacteria was taken at each sampling point and kept in a cool shady place and gently shaken twice per day before further treatment in Norway.

### *5.2. Phytoplankton Analysis*

The Lugol-fixed phytoplankton samples were counted in sedimentation chambers (Hydro-Bios Apparatebau GmbH Kiel, Germany) using an inverted microscope (Leica DMi8; Ortomedic, Oslo, Norway) according to Utermöhl [50]. Phytoplankton biomass was calculated by geometrical approximations using the computerized counting programme Opticount (SequentiX - Digital DNA Processing, Klein Raden, Germany). The specific density of phytoplankton cells was calculated as 1 g cm<sup>−</sup>3.

### *5.3. Isolation of Strains and Morphological Characterization*

Using a microcapillary, single colonies of *Microcystis* and filaments of *Raphidiopsis* were isolated. They were washed five times and placed in wells on microtiter plates containing 300 μL Z8 medium [51]. After successful growth, the samples were placed in 50 mL Erlenmeyer flasks containing 20 mL Z8 medium and maintained at 22 ◦C. Strains were classified based on morphological traits [52,53]. Morphological examination was conducted using a Leica DM2500 light microscope, Leica DFC450 camera and Leica Application Suite software (LAS) (Leica, Oslo, Norway). The morphological identification was based on the following criteria: (i) size of vegetative cells and heterocytes and (ii) nature and shape of filaments or colonies. Length and width of 50–250 vegetative cells or filaments and of 20–50 heterocytes were measured. Akinetes were not detected in the samples. All strains used in this study (Table 2) are maintained at the Norwegian Institute for Water Research, Oslo, Norway.

### *5.4. Genomic DNA Extraction, PCR Amplification and Sequencing*

Genomic DNA was extracted according to Ballot et al. [54]. All PCRs were performed on a Bio-Rad CFX96 Real-Time PCR Detection System (Bio-Rad Laboratories, Oslo, Norway) using the iProof High-Fidelity PCR Kit (Bio-Rad Laboratories, Oslo, Norway). The 16S rRNA gene of the isolated strains from Meiktila Lake was amplified using the primers as described by Ballot et al. [54]. PCR products were visualized by 1% agarose gel electrophoresis with GelRed staining (GelRed Nucleic Acid Gel Stain, Biotium, Fremont, CA, USA) and UV illumination.

Amplified 16S rRNA gene products were purified through Qiaquick PCR purification columns (Qiagen, Hilden, Germany). Sequencing of the purified 16S rRNA gene products was performed using the same primers as for PCR and intermediate primers as described in Ballot et al. [54]. For each PCR product, both strands were sequenced on an ABI 3730 Avant genetic analyser using the BigDye terminator V.3.1 cycle sequencing kit (Applied Biosystems, Thermo Fisher Scientific Oslo, Norway) according to the manufacturer's instructions.

### *5.5. Phylogenetic Analysis*

Sequences of the 16S rRNA gene of the cyanobacterial strains were analysed using the Seqassem software package (version 07/2008) and the Align MS Windows-based manual sequence alignment editor (version 03/2007) (SequentiX - Digital DNA Processing, Klein Raden, Germany). Segments with highly variable and ambiguous regions and gaps making proper alignment impossible were excluded from the analyses.

A 16S rRNA gene set containing 1135 positions was used in the phylogenetic tree for *Cylindrospermopsis*/*Raphidiopsis*. *Sphaerospermopsis aphanizomenoides* (LN846954) was employed as the outgroup, five *Raphidiopsis* strains from Meiktila Lake and 35 additional *Cylindrospermopsis*/*Raphidiopsis* sequences derived from GenBank were included in the analyses. A set containing 1426 positions was used for the *Microcystis* 16S rRNA gene analysis*. Chroococcus subviolaceus* (MF072353) was employed as the outgroup, two strains from Meiktila Lake and 38 additional *Microcystis* sequences derived from GenBank were included in the analysis. Phylogenetic trees for 16S rRNA genes were constructed using the ML algorithm in Mega v. 7 [55]. In the ML analyses, evolutionary substitution models were evaluated using Mega v. 7 [55]. The HKY+G+I evolutionary model was found to be the best-fitting evolutionary model for the Nostocales 16S rRNA gene tree and T92+G+I for the *Microcystis* 16S rRNA gene tree. ML analyses of both trees were performed with 1000 bootstrap replicates using Mega v. 7 [55]. The sequence data were submitted to the European Nucleotide Archive (ENA) under the accession numbers listed in Table 2.

### *5.6. Toxin Analysis*

### 5.6.1. Material

LC–MS/MS utilised standards of ATX-a (Tocris Bioscience, Bristol, UK), homoATX-a (Novakits, Nantes, France) and CYN (Vinci Biochem, Vinci, Italy) and certified reference materials (CRMs) of STX, dcSTX, NeoSTX, GTX1, GTX4, GTX5 and C1 and C2 toxins (National Research Council of Canada, Halifax, NS, Canada (NRC)). LC–HRMS utilised CRMs of MC-RR (**3**), MC-LR (**17**) and [Dha7]MC-LR (**18**) (NRC) and an RM of [d-Leu1]MC-LY [38]. Additional RMs of [d-Asp3]MC-RR (**1**), d-Asp3]MC-LR (**13**), MC-YR (**14**), MC-HilR (**21**) containing traces of MC-FR (**25**), MC-WR (**26**) and MC-LA (**41**) were prepared at NRC from commercial samples (Enzo Life Sciences, Farmingdale, NY, USA), and extracts containing an array of other identified MCs were available from previous work [12,13,15]. Standards for the Adda-ELISA and for the CYN, ATX and STX ELISAs were as provided with the kits (Abraxis LLC, Warminister, PA, USA).

### 5.6.2. ELISA for MCs, CYNs, ATXs and STXs

Fresh culture material of two *Microcystis* and five *Raphidiopsis* strains was frozen and thawed three times. The *R. raciborskii* strains were tested for CYNs using the Abraxis Cylindrospermopsin ELISA kit (Abraxis LLC, Warminister, PA, USA) following the manufacturer's instructions. The test is a direct competitive ELISA that detects cylindrospermopsin but also recognizes deoxycylindrospermopsin and 7-*epi*-cylindrospermopsin. The ELISA results do not distinguish between dissolved and cell-bound toxins. Both *Microcystis* strains were tested for microcystins using the Abraxis Microcystins/Nodularins (ADDA) ELISA kits (Abraxis LLC, Warminister, PA, USA). The test is an indirect competitive ELISA designed to detect Adda, (3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid), based on specific recognition of the Adda moiety [56]. ADDA is a nonprotein amino acid and is the most common side chain at position-5 in microcystins (Figure 5).

All strains were also tested for saxitoxins and anatoxin-a using the Abraxis Saxitoxins (PSP) and Abraxis Anatoxin (VFDF) ELISA kits (Abraxis LLC, Warminister, PA, USA). The saxitoxin ELISA is a direct competitive ELISA that detects saxitoxin based on specific antibody recognition but also recognizes other saxitoxins (e.g., dcSTX, GTXs, lyngbyatoxin, NeoSTX) to varying degrees according to the manufacturer's instructions. The test for anatoxin-a is a direct competitive ELISA that detects

anatoxin-a based on specific antibody recognition but also recognizes homoanatoxin according to the manufacturer's instructions The colour reaction of all ELISA tests was evaluated at 450 nm on a Perkin Elmer1420 Multilabel counter Victor3 (Perkin Elmer, Waltham, MA, USA), and concentrations were evaluated by manual analysis of the absorbance data as recommended by the vendor.

### 5.6.3. Microcystin Analysis by LC–HRMS

Fresh culture material of both *Microcystis* strains was prepared for LC–HRMS by freeze-thawing (3 times), diluting with an equal volume of MeOH and filtering (0.22 μm) [57]). LC–HRMS/MS analysis was performed on a Q Exactive-HF Orbitrap mass spectrometer equipped with a HESI-II heated electrospray ionization interface (ThermoFisher Scientific, Waltham, MA, USA) using an Agilent 1200 LC system including a binary pump, autosampler and column oven (Agilent, Santa Clara, CA, USA). Analyses were performed with SymmetryShield 3.5 μm C18 column (100 × 2.1 mm; Waters, Milford, MA, USA) held at 40 ◦C with mobile phases A and B of H2O and CH3CN, respectively, each of which contained formic acid (0.1% v/v). Gradient elution (0.3 mL min−1) was from 20% to 90% B over 18 min, then to 100% B over 0.1 min and a hold at 100% B (2.9 min), then returned to 20% B over 0.1 min with a hold at 20% B (3.9 min) to equilibrate the column (total run time 25 min). Injection volume was typically 1–5 μL.

The MS was operated in positive ion mode and calibrated from *m*/*z* 74 to 1622. The spray voltage was 3.7 kV, the capillary temperature was 350 ◦C and the sheath and auxiliary gas flow rates were 25 and 8 units, respectively, with MS data acquired from 2 to 20 min. Mass spectral data were collected using a combined full scan (FS) and data independent acquisition (DIA) method. FS data was collected from *m*/*z* 500 to 1400 using the 60,000-resolution setting, an AGC target of 1 <sup>×</sup> 10<sup>6</sup> and a max IT of 100 ms. DIA data was collected using the 15,000 resolution setting, an AGC target of 2 <sup>×</sup> 105, maxIT set to "auto" and a stepped collision energy of 30, 60 and 80 eV. Precursor isolation windows were 62 *m*/*z* wide with centering at *m*/*z* 530, 590, 650, 710, 770, 830, 890, 950, 1010, 1070, 1130, 1190, 1250, 1310 and 1370. Mass spectral data were also collected using a combined full scan (FS) and top-10 data-dependent acquisition (DDA) method. Data was acquired as for DIA but with an exclusion list generated from a blank injection and an inclusion list (both at ± 5 ppm) from a publicly available database of MC *m*/*z* values [58], except that maxIT was set to 100 ms and dynamic exclusion 5.0 s and "if idle pick others" were selected. Putative MCs detected using the above FS/DIA method were further probed in a targeted manner using the parallel reaction monitoring scan (PRM) mode with a 0.7 *m*/*z* precursor isolation window, typically using the 30,000-resolution setting, an AGC target of 5 <sup>×</sup> <sup>10</sup><sup>5</sup> and a max IT of 400 ms. Typical collision energies were: stepped CE at 30 and 35 eV for MCs with no Arg; stepped CE at 60, 65 and 70 eV for MCs with one Arg; and CE at 65 eV for [M+H]<sup>+</sup> and stepped CE at 20, 25 and 30 eV for [M+2H]2<sup>+</sup> of MCs with two Arg groups. Full scan chromatograms were obtained in MS-SIM mode as for DIA but with resolution 120,000 and max IT 300 ms.

In negative mode, the mass spectrometer was calibrated from *m*/*z* 69 to 1780 and the spray voltage was −3.7 kV, while the capillary temperature, sheath and auxiliary gas flow rates were the same as for positive mode. Mass spectrometry data were collected in FS/DIA scan mode as above using a scan range of *m*/*z* 750–1400, a resolution setting of 60,000, AGC target of 1 <sup>×</sup> 106 and a max IT of 100 ms. For DIA, MS/MS data was collected from *m*/*z* 93 to 1400 using a resolution setting of 15,000, AGC target of 2 <sup>×</sup> 105, max IT set to "auto" and stepped collision energy 65 and 100 eV. Isolation windows were 45 *m*/*z* wide and centered at *m*/*z* 772, 815, 858, 902, 945, 988, 1032, 1075, 1118, 1162, 1205, 1248, 1294, 1335 and 1378. Mass spectral data were also collected using a combined full scan (FS) and top-10 data dependent acquisition (DDA) method. Data were acquired as for DIA but with an exclusion list generated from a blank injection and an inclusion list from a publicly available list (both at ± 5 ppm) of MC *m*/*z* values [58], except that maxIT was set to 100 ms and dynamic exclusion 5.0 s and "if idle pick others" were selected. Full scan chromatograms were obtained over a scan range *m*/*z* 750–1400 at a resolution setting of 120,000 using an AGC target of 1 <sup>×</sup> 106 and a max IT of 300 ms.

Thiol derivatizations were performed by addition of (NH4)2CO3 (0.1 M, 200 μL) to the filtered extract (200 μL), with 200 μL transferred to two LC-MS vials. To one vial was added 1 μL of a 1:1 mixture of mercaptoethanol and *d*4-mercaptoethanol (Sigma–Aldrich, St. Louis, MO, USA), while 1 μL of water was added to the other vial as a control. Oxidations were performed by addition of DMSO (5 μL) and Oxone (10 mg/mL in water; 25 μL) to 50 μL of extract [15]. Samples and reactions were placed in the sample tray (held at 15 ◦C) for analysis, and the reactions were monitored periodically until completion and then analysed.

### 5.6.4. CYN, deoxyCYN, ATXs and STXs Analysis by LC-MS/MS

Extraction was performed on freeze-dried cultures (40 mL), according to the protocol in [59]. In brief, dry material was treated with 6 mL of 50% methanol and sonicated (Omniruptor4000 probe sonicator, Omni-Inc., Kennesaw, MA, USA) for 10 min in pulsed mode (50%) using 160 W power. An aliquot of the solution was then filtered on Phenex RC syringe filters (0.2 μm; Phenomenex, Castel Maggiore, Italy) and analyzed by LC–MS/MS.

LC-MS/MS analysis were performed using a Waters Acquity UPLC system (Waters, Milford, MA, USA) coupled to a SCIEX 4000 QTRAP mass spectrometer (AB Sciex Pte. Ltd., Singapore). Chromatographic separation of analytes was performed using a HILIC column (Ascentis Express OH5, 2.7 μm, 50 × 2.1 mm; Merck Life Science S.r.l., Milan, Italy), while MS detection was performed using positive electrospray ionization using scheduled Multiple Reaction Monitoring. Details of the experimental set up are as described by Cerasino et al. [60], and the method was suitable for the detection and quantification of the following toxins: ATX-a, homoATX-a, CYN, STX, dcSTX, NeoSTX, GTX1, GTX4, GTX5 and C1 and C2 [60]. Quantification limits were 0.2–200 μg L–1. Other toxic alkaloids not available as pure standards were also screened but only for tentative analysis (hydroxyand epoxy- and homo-ATXs, deoxyCYN, dcNeoSTX, GTX2/3, dcGTX2/3 and C3 and C4 toxins) using equivalent detection settings to their most similar analogs.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6651/12/4/232/s1, Figures S1–S59: LC-HRMS/MS spectra of **1**, **3**, **4**, **6**, **7**, **9**, **11**–**15**, **17**, **19**–**25**, and **27**–**56**, Figure S60: LC-HRMS spectra of unidentified microcystins **33**–**35,** Table S1: Identities of microcystins detected by LC-HRMS/MS analysis in *M. aeruginosa* strains AB2017/14 and /15, their retention times (*t*R), concentrations, elemental compositions, observed *m*/*z* values in positive and negative ionisation modes, whether they matched retention times in reference samples and whether characteristic microcystin product ions were observed.

**Author Contributions:** Conceptualization, A.B.; methodology, A.B., C.O.M., L.C., V.H.; validation, A.B., C.O.M., L.C.; investigation, A.B., C.O.M., L.C., M.M., T.S.; data curation, A.B., C.O.M., L.C.; writing—original draft preparation, A.B., T.S., C.O.M.; writing—review and editing, A.B., C.O.M., M.M., L.C., T.S., V.H.; supervision, A.B.; project administration A.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Project "Integrated Water Resources Management (IWRM) – Institutional Building and Training in Myanmar (funded by the Norwegian Ministry of Foreign Affairs and the Royal Norwegian Embassy in Myanmar).

**Acknowledgments:** We thank K. Thomas for preparation of RMs of [d-Asp3]MC-RR (**1**), d-Asp3]MC-LR (**13**), MC-YR (**14**), MC-HilR (**21**) containing traces of MC-FR (**25**), MC-WR (**26**) and MC-LA (**41**) at National Research Council, Halifax, Canada in partnership with the Norwegian Veterinary Institute (Oslo, Norway).

**Conflicts of Interest:** The authors declare no conflict of interest

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Determination of Cyanotoxins and Prymnesins in Water, Fish Tissue, and Other Matrices: A Review**

**Devi Sundaravadivelu 1, Toby T. Sanan 2,\*, Raghuraman Venkatapathy 1, Heath Mash 2, Dan Tettenhorst 2, Lesley DAnglada 3, Sharon Frey 3, Avery O. Tatters <sup>4</sup> and James Lazorchak 5,\***


**Abstract:** Harmful algal blooms (HABs) and their toxins are a significant and continuing threat to aquatic life in freshwater, estuarine, and coastal water ecosystems. Scientific understanding of the impacts of HABs on aquatic ecosystems has been hampered, in part, by limitations in the methodologies to measure cyanotoxins in complex matrices. This literature review discusses the methodologies currently used to measure the most commonly found freshwater cyanotoxins and prymnesins in various matrices and to assess their advantages and limitations. Identifying and quantifying cyanotoxins in surface waters, fish tissue, organs, and other matrices are crucial for risk assessment and for ensuring quality of food and water for consumption and recreational uses. This paper also summarizes currently available tissue extraction, preparation, and detection methods mentioned in previous studies that have quantified toxins in complex matrices. The structural diversity and complexity of many cyanobacterial and algal metabolites further impede accurate quantitation and structural confirmation for various cyanotoxins. Liquid chromatography–triple quadrupole mass spectrometer (LC–MS/MS) to enhance the sensitivity and selectivity of toxin analysis has become an essential tool for cyanotoxin detection and can potentially be used for the concurrent analysis of multiple toxins.

**Keywords:** harmful algal blooms; cyanotoxins; cyanobacteria; fish tissue; shellfish; detection methods

**Key Contribution:** This review article examines the available methodologies for the detection and quantification of cyanotoxins in complex matrices, such as fish tissue. An understanding of toxin concentrations using a standardized and reliable methodology is crucial to accurately assess the threat of these toxins to ecosystems and human health.

### **1. Introduction**

Harmful algal blooms (HABs) have been observed in freshwater, estuarine, and marine waters in the U.S. and around the globe. Cyanobacteria frequently contribute to HABs in freshwater systems and are able to produce highly potent toxins, known as cyanotoxins. A large number of cyanotoxins have been reported from different species of cyanobacteria, including microcystins (*Microcystis*, *Anabaena*, *Hapalosiphon*, *Dolichospermum*, *Gloeotrichia*, *Nostoc*, *Oscillatoria*, *Phormidium*/*Microcoleus*, and *Synechocystis*), nodularins (*Nodularia*, *Nostoc*, and *Iningainema*), anatoxins (*Anabaena*, *Aphanizomenon*, *Cylindrospermum*, *Dolichospermum Planktothrix*, *Oscillatoria*, *Geitlerinema*, *Phormidium*/*Microcoleus*, and *Tychonema*),

**Citation:** Sundaravadivelu, D.; Sanan, T.T.; Venkatapathy, R.; Mash, H.; Tettenhorst, D.; DAnglada, L.; Frey, S.; Tatters, A.O.; Lazorchak, J. Determination of Cyanotoxins and Prymnesins in Water, Fish Tissue, and Other Matrices: A Review. *Toxins* **2022**, *14*, 213. https://doi.org/ 10.3390/toxins14030213

Received: 25 January 2022 Accepted: 13 March 2022 Published: 16 March 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

cylindrospermopsins (*Oscillatoria* and *Raphidiopsis*), lyngbyatoxin-a (*Lyngbya*), saxitoxins (*Anabaena*, *Aphanizomenon*, *Cylindrospermum*, *Lyngbya*, *Phormidium*, and *Raphidiopsis*), and aplysiatoxins (*Lyngbya*, *Schizothrix*, and *Oscillatoria*) [1–11].

Microcystins (MCs) are the most frequently studied and the most widespread cyanotoxins [12], with approximately 50% of publications on cyanotoxins focusing on MCs, 25% on saxitoxins, and 25% on other toxins (such as nodularin, anatoxin-a, β-N-methylamino-L-alanine, cylindrospermopsins, and prymnesins). Of these publications, a majority focus on detection in water matrices. There is limited information on the presence of MCs and other cyanotoxins in matrices such as aquatic food webs, phytoplankton, zooplankton, periphyton, macroinvertebrates, forage fish, bottom feeders, and top carnivore fish. Since research on cyanotoxins has mostly focused on MCs, especially in water matrices, the development and standardization of better monitoring methods for other cyanotoxins and complex matrices, such as those mentioned above, are of utmost importance.

Various techniques have been developed to analyze cyanotoxins in water. These methods include enzyme-linked immunosorbent assay (ELISA), protein phosphatase inhibition assay (PPIA), oxidation of MCs to produce erythro-2-methyl-3-methoxy-4-phenylbutyric acid (MMPB) [13], and chromatography coupled with various detection methods [14]. A survey of chromatographic methods for the detection of cyanotoxins in water could be divided into the following main categories: liquid chromatography with ultraviolet detection (LC–UV), LC with fluorescence detection (LC–FLD), LC–photo-diode array (LC–PDA) or similar, LC with tandem mass spectrometry (LC–MS/MS), gas chromatography with mass spectrometry (GC–MS), ion mobility spectrometry (IMS), and immunological assays. LC–MS, or single quadrupole mass spectrometry, has largely been replaced by LC–MS/MS. Among these methods, a few LC–MS/MS and ELISA-based standard methods developed by the U.S. Environmental Protection Agency (EPA) and the International Organization for Standardization (ISO) are available, are considered standard, are widely used for the measurement of cyanotoxins in water, and can potentially be adapted for use with other matrices, such as fish tissue [15–18]. However, there is no standardized analytical method to detect these cyanotoxins in other matrices [19]. The surveyed studies herein show that ELISA and LC–MS/MS analysis have most frequently been used for detection/quantitation. While specific extraction procedures used vary considerably, overall method performance metrics, such as matrix spike recoveries and replicate precision, were sometimes poor or unreported. This raises questions about the quality of the data in the case of fish tissue and organs, shellfish and similar matrices exposed to HABs.

Most methods developed to date depend on extracting toxins in some form of an aqueous matrix, typically water and one or a combination of organic solvents. For example, ELISA methods should be carried out directly in water with only limited amounts of organic solvent, while LC–MS/MS methods typically require samples to be prepared with significant amounts of organic solvent in water [20]. In addition to the presence of organic solvents, some methods are sensitive to low pH, lipid content, or other matrix effects due to the properties of the target toxins or the nature of the matrix [21]. Hence, analytical methods for cyanotoxins in complex matrices, such as tissue, organs, and plants, frequently require efficient extraction, clean-up, and transfer of the toxin into a suitable solvent for analysis. Use of inefficient extraction methods may lead to poor matrix spike recoveries and poor precision among replicate samples.

The prevalence of HAB events globally and the potential for exposure and bioaccumulation of toxins necessitate monitoring and detecting cyanotoxins in tissues associated with HAB-related mortality events. There is a need for the development of reliable extraction and analytical methods capable of analyzing edible fish/shellfish, plant, and animal tissues for multiple toxins and for such methods to be broadly applicable to all inland waters and coastal food webs. The overarching goal of this study is to summarize available methodologies for the measurement of toxins, identify gaps in the detection of toxins in complex matrices, and highlight the complexity of quantifying groups of compounds with diverse chemistries.

### **2. Cyanotoxins and Prymnesins**

Cyanotoxins can be categorized based on (1) their mechanism of action on terrestrial vertebrates, especially mammals, e.g., hepatotoxins, neurotoxins, and dermatotoxins, and (2) their chemical structure, e.g., cyclic peptides, alkaloids, and lipopolysaccharides [22–24]. Table 1 lists the names, chemical groups, and mammalian targets of some common cyanotoxins. Physicochemical properties of cyanotoxins not only determine which extraction procedures or detection techniques are suitable but also impact the bioaccumulative and toxicological properties of the toxins on several taxa of aquatic invertebrates and vertebrates [25].


**Table 1.** Characteristics of the most commonly studied cyanotoxins and prymnesins.

MCs, the most studied class of cyanotoxin, are produced several cyanobacterial genera including *Microcystis*, *Anabaena*, *Hapalosiphon*, *Nostoc*, *Dolichospermum*, *Gleootrichia*, and *Oscillatoria* [12]. These compounds are short cyclic heptapeptides synthesized by nonribosomal pathways [26]. MCs are characterized by significant structural variability in amino acid composition, including residue substitutions, methylations, and demethylations. In total, more than 248 congeners/variants have been reported in the environment. Among them, MC-LR and MC-LA are among the most toxic [27,28]. Only a limited subset of MCs has commercially available analytical standards, which limits the scope of quantitative methods, such as LC–MS/MS, which rely on these for accurate quantification. MCs are cytotoxic and inhibit protein phosphatases, which leads to several subsequent harmful effects at concentrations over 10 μg/L [29].

Nodularins (NODs) are a class of cyclic pentapeptides structurally similar to MCs. Currently, 10 variants of this water-soluble and stable toxin have been identified [30,31], the most common being the variant with arginine as the variable amino acid (NOD-R). Similar to MCs, NODs are hepatotoxins acting through the inhibition of protein phosphatases and are potential tumor promoters. While NODs have historically been considered to be less common than MCs, their occurrence has been observed in bloom events corresponding to the species *Nodularia spumigena*, and animal deaths have been reported [31].

Anatoxin-a (ANA-a) and its structural variants have been associated with the genus *Anabaena* (now *Anabaena*/*Dolichospermum*) and also to *Aphanizomenon*; *Planktothrix*; *Cylindrospermum*; *Microcystis*; and the benthic *Oscillatoria*, *Phormidium*, *Tychonema*, and *Geitlerinema* [32]. ANA-a is an alkaloid known to exhibit acetylcholinesterase inhibition activity, which is its primary mode of toxicity. Of the structural variants of ANA-a, homoanatoxin-a, resulting from the methylation of the carbon at the extremity of the ketone function, is one of the more commonly observed [33]. Several other derivatives of ANA-a have been identified, including 2,3-epoxy-anatoxin-a; 4-hydroxy- and 4-oxo-derivatives; dihydroanatoxin-a; and dihydrohomoanatoxin-a [34,35]. Toxicity indications for these other variants are mixed, with some reports indicating lower toxicity while others suggesting higher oral toxicity from the dihydroanatoxin variant specifically, suggesting that monitoring of these variants is of interest [36].

Saxitoxins (STXs) are highly polar, nonvolatile, tricyclic perhydropurine alkaloids. While the overall molecular structure is largely conserved within the group, substitutions and variations amount toat least 57 analogs in the environment. They are also referred to as paralytic shellfish toxins (PSTs). PSTs can be non-sulfated, singly sulfated, or doubly sulfated [33,37]. These water-soluble toxins can persist for over 90 days in freshwater [38], but they are altered by high temperatures and may be degraded into more toxic variants [32]. STX is one of the most potent natural neurotoxins, and a dose of approximately 1 mg of the toxin from a single serving of contaminated shellfish is fatal to humans [39].

The cyanotoxin β-N-methylamino-L-alanine (BMAA) is a non-protein amino acid reportedly produced by the majority of cyanobacterial isolates [40]. Some literature reports have suggested that cellular exposure to BMAA may lead to neurological damage in the brain and the central nervous system of humans and animals, potentially contributing to one of several neurodegenerative diseases [41]. However, this remains uncertain as while other studies have illustrated a wide spectrum of effects to exposure, the neurodegenerative disease relationship was not universally observed [42].

Cylindrospermopsins (CYNs) are another class of alkaloid cyanotoxins produced by a variety of freshwater cyanobacteria. Production of CYNs have been reported from Nostocalean species mostly, as well as recently from one Oscillatoriale [43]. CYN and its variants are highly polar, polycyclic uracil derivatives containing guanidino and sulfate groups, and their detection has been reported worldwide. It is a potent hepatotoxin and is likely to be taken up by a variety of aquatic organisms, suggesting potential bioaccumulation risks [22,44].

The invasive algae *Prymnesium parvum* (golden algae) is a species of haptophyte (Prymnesiophyta). The species is of concern because of its ability to produce a suite of complex polyether toxins, known as prymnesins. Prymnesins may refer to Prymnesin-1, Prymnesin-2, or Prymnesin-B1 (to name a few of the presently identified species). These variants differ primarily in the length of the carbon backbone and are recognized hemolytic and ichthyotoxic agents that have been associated with massive fish kills in at least 14 countries [45]. Because of the uncertainty as to the identity of potentially toxic species, and the unavailability of analytical standards, the detection methods used for other cyanotoxins are often not suitable for the detection of prymnesins, which often require more complex analyses, such as high-resolution mass spectrometry (HRMS), quadrupole-time of flight mass spectrometry (qTOF), and nuclear magnetic resonance (NMR) [46,47].

### **3. Current Detection Methods for Cyanobacterial Toxins**

A wide range of detection methods is available for the analysis of cyanotoxins. The most commonly used methods for different toxins are listed in Table 2. These methods are primarily targeted for water matrices; however, some can potentially be adapted for use with other matrices, such as fish tissue, shellfish, or organs.

**Table 2.** Analytical methods available for the detection of cyanotoxins and prymnesins (primarily used for water but adapted for use with fish tissue and other matrices).


\* Includes LC–MS, LC–MS/MS, and high-resolution mass spectrometric technologies.

The quantification of cyanotoxins in water matrices is at this point a well-studied problem, with multiple EPA standard methods developed for drinking water and an extensive literature presence. However, studies on the extraction and quantification of cyanotoxins in more complex matrices, such as shellfish, fish tissue, and organs, are more limited in quantity and scope. This review will present a summary of studies evaluating cyanotoxins in various matrices, and their extraction, preconcentration, and clean-up stages will be discussed, along with the specific quantitation techniques employed.

### **4. Sample Preparation and Analytical Methods for Cyanotoxins and Prymnesins Detection**

Various methods are available for the quantification of cyanotoxins, as described above, and are primarily split between immunoassays and chromatographic separation (i.e., GC or LC), followed by quantification using mass spectrometry, UV, and photodiode array detectors. The matrix suitability varies by procedure, with a variety of interferences inherent to specific analysis; for example, mass spectrometric methodologies are susceptible to matrix interferences related to the ionization technique employed, with ion suppression/enhancement commonly observed in more challenging sample matrices. For each method, then, the samples often require specific preparation before analysis. This is especially common for tissue and other more complex matrices than water.

### *4.1. Sample Preparation*

Sample preparation procedures differ according to the toxin analyzed and the nature of the tissue, organ, or biomass analyzed. Typically, this includes extraction, clean-up, and any post-extraction modifications (e.g., derivatization and solvent exchange) prior to analysis. Table 3 lists a variety of sample extraction and preparation procedures from a survey of the literature. The typical workflow being followed is summarized in Figure 1.

**Figure 1.** Common sample extraction and preparation procedures for measuring cyanotoxins in tissue.

Procedures for extracting toxins from complex matrices vary widely (Table 3). In most cases, heterogeneous samples, including tissues, organs, and food supplements, are mechanically homogenized or lyophilized to break down any solid structures and produce a suspension or emulsion amenable to extraction by solvents containing various percentages of aqueous and/or organic solvents (acidified methanol, acidified water, acetone, acetonitrile, butanol, etc.). In samples containing intact cyanobacterial cells, lysis to release intracellular toxins is often facilitated by ultrasonication, extended mixing, or incubation

at a specific temperature and application of heat [48]. Following lysis/homogenization and extraction, a variety of clean-up techniques are available (Figure 1), including centrifugation, solid-phase extraction (SPE), protein precipitation, hexane washes, filtration, and immunoaffinity columns [49]. The clean-up steps attempt to remove matrix components, after which the toxin is eluted/extracted with a solvent and reduced in volume by evaporation prior to analysis. The efficiency of extraction techniques and the clean-up steps in the recovery of cyanotoxins from complex matrices vary widely based on the toxin physicochemical properties, matrix effects, holding times, collection methods, and other site-specific variables. In addition, each of these extraction techniques and clean-up steps may result in some loss of the target toxins during sample processing.

Extraction of cyanotoxins from complex matrices requires use of appropriate solvent mixtures to remove them, ideally without also extracting interferences to the analytical techniques being employed. The references cited in Table 3 are, broadly, either attempting to extract only individual types of toxins or attempting to quantify multiple classes with a single extraction workflow. The former can potentially allow for optimal conditions to be obtained, while in the case of more general extractions, compromises may result from attempting to measure chemicals less compatible with the extraction conditions. The primary differentiating step is the extraction procedure, solvents, and conditions used, which are elaborated upon below.

### 4.1.1. MCs and NODs

There are numerous literature examples of procedures for the extraction of MCs from a variety of matrices, such as tissue, listed in Table 3, including some that have systematically evaluated various extraction conditions. Because of their similar structure and chemical properties to MCs, there is little to differentiate extraction conditions for NODs, and as seen in Table 3, the conditions are comparable. The most common methods for the extraction of MCs have used methanol:water mixtures of varying ratios, in some cases with added acid, typically acetic acid. Because MCs are a diverse class of species with varied amino acid substitutions, one complicating factor is the variability in hydrophobicity/hydrophilicity of MCs that could be present in a sample. This has been found to be an issue in extractions with pure methanol, for which lower recovery of more hydrophilic congeners, such as MC-RR, has been reported. Instead, most procedures have used 75% to 90% methanol [50], the remainder being water, which has been seen to be generally appropriate for a broad suite from hydrophobic to hydrophilic congeners (See Table 3). An additional complication, however, has been observed in extractions using methanol when used in conjunction with some ELISA based toxin measurements. In at least some studies, interferences resulting in false positive detections were observed and confirmed by comparison with LC–MS/MS, where no MCs were observed [49]. While it is possible that the targeted methods were missing unknown MCs, this was further evaluated by comparison with a direct monoclonal ELISA with known higher specificity to MCs, and the absence of MCs via that method was confirmed.

One alternative to attempting to extract individual MC congeners has been developed using Lemieux oxidation, which has been demonstrated to react with a variety of MCs and convert them to a common product from all MCs with "adda", producing MMPB, which can be extracted instead. Quantification of MMPB can then serve as a proxy measurement for the total of MC congeners present in the sample. This does have limitations in that MMPB oxidation is known to generate positive results when parent MCs may be partially metabolized or otherwise transformed, as in drinking water treatment where oxidized MCs give a false positive signal, but recent publications indicate correlations seen between solution phase MC detections and detections in fish tissue [51].

### 4.1.2. ANA-a

Examples in Table 3 show three different procedures for the extraction of ANA-a from a solution. The first uses an acidic ethanol:water mixture (80:20) that is largely comparable to that for the majority of MC detecting methods above [52]. An alternative approach used immunoaffinity beads to pull the toxin from waters and slurries, but the practicality of this for heterogeneous systems was less clear [53]. Finally, for water samples, SPE is a viable avenue for extracting ANA-a, but this is again of limited utility for more heterogeneous matrices, where cartridge clogging will be more of a challenge [54]. More recent examples for ANA-a extraction in mixed methods are discussed below.

### 4.1.3. STXs

STXs are universally prepared using acidic or buffered extractions owing to their chemical instability under basic conditions. As these are a hydrophilic class of toxins, the extraction solvents used have generally been acidic, typically aqueous acetic [55–57] or hydrochloric acid [58], with buffered water being used in one instance [59]. Elevated temperatures are frequently used to extract STXs, including boiling of the solvent/sample mixture to facilitate extraction efficacy. Interestingly, in one study [59], acidic water and acidic methanol:water mixtures were evaluated and found to produce significantly lower recoveries of STX specifically, while the use of neutral pH phosphate buffer increased recoveries to >50%, which they speculated was due to reduced solubility in low pH.

### 4.1.4. CYNs

Cylindrospermopsins are a significantly more hydrophilic toxin than MCs, without as much structural diversity, and as a result extraction procedures have generally been more straightforward. In Table 3, three procedures are listed for the analysis of CYNs, and these all use mechanical homogenization or lyophilization, followed by extraction with methanol or water.

### 4.1.5. BMAA

Extraction of BMAA is typically performed under acidic conditions following lyophilization [60–62]. Because it is an amino acid, it is highly water soluble and does not require the use of organic solvents for extraction from tissue. However, removal of proteins and/or lipids is part of the process in some cases on extraction with chloroform [61].

### 4.1.6. Prymnesins

Evaluation of extraction techniques for prymnesins and other toxins associated with *P. parvum* is difficult owing to the many questions related to the toxic species, their structure, and the lack of available analytical standards. The protocols listed in Table 3 illustrate these challenges, with [46,47,63] repeatedly re-extracting the water/algal cell lysates with solvents of varying polarity in an attempt to isolate and characterize potential toxic species. In [46], cold acetone was used as a pre-extraction solvent to remove chlorophyll, followed by methanol extraction, in which prymnesins were observed by LC–HRMS. In [47], *P. parvum* lysates were extracted with dichloromethane, ethyl acetate, methanol, and water and while prymnesins were not observed under LC–HRMS analysis, cytotoxic activity was seen in the ethyl acetate and methanol extracts and a variety of fatty acid amides and one hydroxamic acid were observed. Finally, in [63], cold acetone was again used as a chlorophyll-removal step before sequential methanol and *n*-propanol extraction of the cellular material, which was then followed by solvent exchange to water. The aqueous material was then extracted with ethyl acetate four times to remove fatty material, before a final SPE step for clean-up prior to analysis. In this study, two prymnesins and a variety of related ions and fragments were characterized by LC–HRMS.

### 4.1.7. Extraction of Multiple Cyanotoxins

Workflows to extract multiple toxin classes typically attempt either to group compatible species together or to accept compromises in recovery arising from the impossibility of trying to recover diverse species simultaneously. As described above, procedures for the extraction of MCs and NODs have typically settled on 75–90% methanolic water as

an extraction solvent. In studies such as [20,64–66], extraction of both MCs and NODs is described and the solvent mixtures used were uniformly in that range, showing the ease of extracting similar classes of toxins in tandem. However, studies attempting to extract those toxins as well as the more hydrophilic species, such as STXs or CYNs, as in [67–70], typically reduced the percentage of organic solvent to ~50% in order to improve solubility of the more hydrophilic constituents of the methods. An alternative approach was followed in [71], where two separate extraction procedures were devised to recover collectively STX, ANA-a, and CYN with 25% acetonitrile in water, while MCs and NOD were extracted with 75% acetonitrile in water from a split sample. The latter approach attempts to reduce the impact of chemical incompatibility, but at a cost of doubling extraction and analysis requirements. Once the samples are prepared based on the extraction and clean-up techniques described above, the many analytical methods discussed in Section 4.2 may be used for detection and quantification.

### *4.2. Analytical Methods*

### 4.2.1. Immunological Assays

Cyanotoxins can be detected through recognition and binding to specific antibodies, either monoclonal or polyclonal. For example, ELISA kits are commercially available for the detection of MCs in water based upon either the common "adda" moiety present or specific recognition of a single MC congener [49,72–74]. Depending upon the antibody and the procedure employed, these kits can achieve a detection limit (DLs) as low as 4 ng/L, with an upper quantitation limit (due to saturation) of 5 μg/L for MC-LR [73]. While ELISA is frequently employed for the detection of MCs, particularly in drinking water, ELISA kits have also been made commercially available for the detection of ANA-a, CYN, and STX [75,76]. The most significant advantage of ELISA methods is that they do not require expensive and high-upkeep analytical instrumentation to be maintained, as they typically rely only on colorimetric assays for quantification.

However, detection methods based on ELISA have some limitations of varying severity by target compound. The measurement of a variety of MCs by ELISA is possible because the antibody assay is broadly cross-reactive over different MC congeners; however, this cross-reactivity is not uniform, and in some cases, MCs may be measured with greater or lesser responses relative to MC-LR [20]. Beyond that, only a single signal for a general MC concentration will be obtained, even for a sample that might contain a variety of MC congeners of varying potential toxicity levels. As a result, ELISA-based measurements should be considered to be semi-quantitative in that the measured concentrations are influenced by a number of variables that may or may not be known at the time of analysis. In addition to this limitation, cross-reactivity of the assay with other compounds in the sample and matrix may lead to over- or underestimation of the concentration of toxins, for both MCs and other cyanotoxins. This was observed in one paper, mentioned in Table 3, where ELISA kits targeting the "adda" moiety common to most MCs were found to be generating false positive results when compared with LC–MS/MS and an alternative, monoclonal antibody specific to MC-LR [77]. Conversely, it has also been demonstrated that ELISA-based screening may fail to measure protein-bound or glutathione-conjugated species in tissue matrices, perhaps due to the molecules being inaccessible to the antibody. When ELISA extracts were compared with a chemical oxidation/derivatization technique for measuring total MCs in tissue, a significant enhancement in the measured concentrations was observed in the latter [78]. These types of issues are not specific only to ELISA, but rather a general complication for any measurement technique. The sequestration/transformation of toxins through metabolic/removal processes may make them unavailable for many types of assays.

ELISA-based techniques can also be sensitive to the presence of solvents such as methanol and acetonitrile that are often used for tissue extraction. Commercially available ELISA kits typically recommend <5–10% solvent, but this value may be as low as <2.5% for ANA-a and as high as <20% for CYN and STX, based on standard protocols included in the

various ELISA kits; therefore, in many studies relying on ELISA to quantify toxins from tissue extracts (see Table 3), the organic extracts are evaporated to dryness under nitrogen at 30–60 ◦C and reconstituted in DI water or an appropriate diluent prior to analysis [20]. Low pH can also affect ELISA performance, which could be an issue for STX extracted under acidic conditions. In many of the reviewed studies, pH was adjusted with 0.1 N NaOH to improve compatibility with the ELISA assay.

### 4.2.2. Mass Spectrometry

Following sample extraction and clean-up, separation of compounds typically employs either liquid or gas chromatographic (LC or GC) techniques. Typically, LC methods use a reversed phase C18 or a hydrophilic interaction liquid chromatography (HILIC) column and either methanol:water or water:acetonitrile gradients for separation, as these allow for flexibility, speed, and adaptability to a wide range of detectors relying on UV absorbance, fluorescence, or mass spectrometry. GC can be used as a separation method for cyanotoxins [79]. However, many cyanotoxins, including MCs, are larger molecules and are either not volatile or do not ionize well without chemical derivatization techniques. In addition, GC-based methods might require a solvent-exchange step to a nonpolar solvent such as hexane or ethyl acetate prior to analysis, which also typically involves blowing samples down to dryness prior to reconstitution with the nonpolar solvent. As such, GCbased separation requires more complex and time-consuming sample preparation, while also presenting more avenues for toxin loss from samples. While there are examples in the literature of analytical methods for cyanotoxins using GC–MS methods, these are a clear minority [78]. The single example in Table 3 using GC–MS involved MMPB oxidation of MCs, and in this instance, using GC–MS as the detector required an added derivatization step in the workflow.

Mass spectrometric methods for quantifying cyanotoxins rely on comparison with specific analytical standards for target compounds and can provide greater specificity than is possible with ELISA. In the LC–MS/MS techniques typically employed for toxin measurement, both a starting mass and a characteristic fragment ion specific to a given chemical are used to ensure specificity of assignment, along in some cases with a separate confirming ion fragment and a unique ratio of confirming to main ion. This is advantageous in that it means that false positive identifications are less common (although not impossible; ANA-a is isobaric with the common amino acid phenylalanine and both parent and product masses are identical; confusion of these two species can occur when MS-based detection is used [53]).

An additional useful feature of LC–MS/MS methods is that surrogate and internal standards, usually consisting of isotopically labeled analogs of target compounds, can be added to samples prior to extraction and analysis. This enables the losses encountered during extraction, clean-up, and analysis steps to be accounted for through comparison with expected recoveries of surrogates, while isotopic dilution techniques can allow for compensation for some matrix interferences during all of these processes, improving overall quantitation accuracy [77].

Quantitative mass spectrometric methods have been developed for the majority of the cyanobacterial toxins, including MCs, ANA-a, CYNs, STXs, and BMAA, either as a native compound or after chemical modification [80]. This allows samples to be separated using conventional C18 phases, although with ANA-a(S), HILIC methods work well for the assessment of this very hydrophilic toxin [81]. Due to the specificity, sensitivity, and rapidity of the analysis, mass spectrometric methods are now the physicochemical method of choice for the quantitative analysis of most cyanobacterial toxins in complex matrices for labs with access to appropriate instrumentation.

Cyanotoxins can also be detected by MS without preliminary chromatographic separation, particularly with time of flight (TOF) mass spectrometers in which many compounds can be identified and quantified concurrently. For example, matrix-assisted laser desorption/ionization coupled with TOF (MALDI-TOF) can be used to perform toxin analysis in

even individual cell colonies [79]. In a typical workflow, target compounds enclosed in the dried and solid sample are ionized by a laser beam and accurately identified through the high mass resolution provided by the TOF instrument. However, TOF mass spectrometers usually tend to be less sensitive than other mass spectrometers of the same generation, and these methods are less commonly used for routine sample screening and quantitation than LC–MS/MS methods. However, TOF- and HRMS-based instruments can be used to qualitatively identify toxins that lack analytical standards, unknown toxins, and/or their metabolites or quantify using standards of other structurally similar toxins [82–84].

LC–MS/MS is the most commonly used chromatographic method for cyanotoxin detection in general, but it is limited by the need for analytical standard material to be available. For example, LC–MS/MS methods for the measurement of prymnesins and other toxins produced by *P. parvum* have not been commonly reported in the literature; these represent an emerging class of contaminants for which there is presently little information. The key issue preventing their measurement is the absence of a pure analytical reference standard for these toxins, which could be used for calibration. The ability to incorporate an internal standard would be ideal for highest precision for an analytical method, ideally a stable, isotopically labeled form of the analyte, but along with the native (unlabeled) prymnesins, these are not available. To date, only a few studies have reported isolating potential toxins, including prymnesin molecules, from *P. parvum*, [85–89], and replication of this work is yet to be reported in the peer-reviewed literature. Analytical standards for the prymnesins, either normal or isotopically labeled, along with any other structurally similar molecule are not currently available. A few studies have used UTEX strain 2797 as the source material for the preparative isolation of toxin material for analytical method development [46,63,89–92], but these preparations are not quantitatively exact enough to allow for use as true analytical standards. As a result, there are currently no published validated methods for the quantitative analysis of prymnesins [93].

### 4.2.3. LC–UV and LC–FLD

Monitoring UV absorbance was historically one of the first techniques to detect cyanotoxins after LC separation. MCs and CYNs have specific UV spectra with maximum absorbances at 240 and 262 nm, respectively [94,95]. Analytical workflows using LC–UV and/or LC–PDA allow for measuring the concentration of MCs by using these characteristic absorbances. In conjunction with good chromatographic separation, samples can not only have toxin concentrations measured but also give some information on the congeners of the MCs present. The key limitation of UV-absorbance-based techniques, however, is that the absorbance is not limited to only MCs and that background interferences may also impact the baseline signal. This can result in both lower DLs and potentially false positive signals, depending on the matrices being studied. Literature results for MC monitoring using LC–UV and LC–PDA have allowed for measurement of up to seven MC variants through comparison with analytical standards [96,97].

Detection by fluorescence is also commonly used after LC separation. However, cyanotoxins do not naturally fluoresce and, therefore, require the addition of a derivatization process during the sample preparation. For example, in the high-performance liquid chromatography with fluorescence detection (HPLC–FLD) system used by [98], post-column derivatization was performed using a solution containing 10 mmol/L of periodic acid and 550 mmol/L of ammonia in water (flow rate 0.3 mL/min). Nitric acid (0.75 mol/L in water; flow rate 0.4 mL/min) was used for reducing the pH value to 2–3. Fluorescence detection was applied for the determination of STX oxidation products at an excitation and emission wavelength of 330/395 nm.

While most studies that have used LC–UV and LC–FLD techniques have done so in water matrices, several prior studies have used fluorescence and/or UV-based methods to measure cyanotoxins (and in particular MCs) in fish, sediments, and plants [96,99]. The results of these studies show sensitivities comparable to those of LC–MS/MS methods, but with reduced specificity, due to which they can be more susceptible to matrix interferences, resulting in higher practical DLs. While LC–UV, LC–FLD, and LC–PDA methods can be simple and cost effective, misidentification can occur due to the non-specific nature of these methods. LC–MS/MS methods are becoming more prevalent in environmental laboratories now and are the preferred method for toxin identification as they can precisely and accurately identify toxins based on a specific mass/charge precursor ion that is unique to each toxin. In addition, despite typically offering only a unit mass resolution, LC–MS/MS methods can improve confidence in the identification of target compounds through screening for confirmation ions and the ratio of target and confirmation ions being produced.

### 4.2.4. Biochemical Assays

MCs and NODs are potent inhibitors of protein phosphatases (PPs), and in addition to antibody screening, these toxins can also be detected and measured using a protein phosphatase inhibition assay (PPIA) [74,100–103]. This assay measures the rate of formation of *p*-nitrophenol (pNP; yellow color) through hydrolysis of *p*-nitrophenol phosphate by PPs over time and measures all PP inhibitors present in a sample. The colorimetric PPIA has been optimized for MC and NOD detection in cyanobacteria extracts using 96-well plates and has shown acceptable correlation with HPLC data [104]. However, PPIA cannot distinguish co-occurring variants of MCs and cannot distinguish MCs from NODs. Therefore, results are often expressed as equivalent dosages relative to MC-LR, which is used as a reference standard. In addition, when analyzing bloom-containing water, interferences with unknown compounds can occur, leading to overestimation or underestimation of toxin concentration. Few studies have documented the use of PPIA for toxin measurement in tissue because of the complexity of tissue extracts and potential for interferences [64,105]. Note that because PPIA detects only inhibitors of that enzyme, within the realm of cyanotoxins it is known to quantify only MCs and NODs and alternative analysis should be undertaken to detect other cyanotoxins if required.

### **5. Discussion**

Methods for the extraction and measurement of toxins from complex matrices are influenced by the chemistry of both the toxins and the composition of various matrices. MCs are a particularly diverse class of cyanotoxins, with more than 248 structurally diverse congeners identified in the environment, with chemical properties (including hydrophobicity) varying accordingly. Analysis of MCs in tissue and organs primarily relies on the extraction of homogenized tissue with various fractions of methanol:water, typically 75–80%, or conversion of MCs to MMPB to enable measurements of a single species rather than an ensemble [78,106–108]. In both pathways, the extracts are often cleaned via centrifugation [106], hexane wash [108], or SPE [107]. The methanol:water extracts are compatible with LC–MS/MS or other chromatographic methods directly or with ELISA analysis after dilution or solvent evaporation and reconstitution (to reduce the organics in the final sample) to avoid matrix effects [109]. While only a limited number of individual congeners can be quantified using LC–MS/MS methods due to the need for matching analytical standards (in limited availability) and difficulty with isomer separation, both ELISA and the MMPB method can be used to measure total MCs, with the latter potentially also including bound, conjugated, or partially degraded MCs. DLs for ELISA and LC–MS/MS methods for tissue matrices in the literature are generally as low as a few nanograms/gram. DLs for PPIA [106] in water matrices were higher, on the order of 1 μg/L in water. LC–PDA methods also generally have higher detections for MCs in water, with 2.9 μg/L as the lowest level detected, in [97]. Total MCs quantified using the MMPB method, as a proxy for total MCs, have been shown to have DLs of 2.18 ng/g wet weight [51]. One concern for anti-adda-based ELISA methods in the measurement of MCs is the potential for false positives, discussed above; confirmation of ELISA results with LC–MS/MS or direct monoclonal antibodies has been found to be effective at confirming or refuting such results [49].

Due to their structural similarity to MCs, NODs are also frequently extracted using methanol or methanol:water mixtures. If only pure methanol is used, as in [110], if employed for a simultaneous analysis for MCs, it would be expected to be susceptible to the same under-extraction of more hydrophilic MCs, as seen in [72]. More broadly, there is cross-reactivity in ELISA analysis between NODs and MCs, which can lead to difficulties differentiating the specific toxins at a site. LC–MS/MS methods should be used to confirm the specific toxins present if this information is needed. In the reports discussed herein, detection of NODs in tissue varied significantly from organ to organ, from 21 μg/kg to 1.4 mg/kg, with ELISA being the predominant detection method used for tissue samples.

In this review, methods for the analysis of solely ANA-a or CYN were limited and some are more than a decade old. These studies employed extraction of ANA-a using solvents (ethanol:acetic acid, 80:20), immunoaffinity beads, or SPE, followed by analysis using ELISA, IMS, or HPLC–UV [52–54]. While DLs for these methods were comparable to extraction/analysis procedures reviewed for other toxins (25 ng/L in water and 4 μg/g in tissue), they are rarely used. Similarly, the extraction of only CYN was achieved using methanol or water, followed by analysis using ELISA, LC–MS/MS, or LC–UV methods, achieving detection varying from 2.7 ng/g for the highly sensitive MS/MS methods to 2.5 μg/g using LC–UV [111–113]. LC–MS/MS is the most common methodology currently employed to detect ANA-a and CYN concurrently in the studies reviewed. Since most recent studies optimize the extraction of multiple toxins in one workflow, the multi-toxin methods reviewed herein provide the path forward and best practices for ANA-a and CYN detection in complex matrices. For example, the use of 75:25 water:acetonitrile for the extraction and use of chemically similar isotopically labeled surrogates, such as ANA-a-13C4 or CYN-15N5, which have similar extraction recovery, column retention, and ionization efficiency, can help with improved compensation for recovery bias and matrix effects [71,114]. These highly sensitive methods provide method detection limits (MDLs) as low as 0.14 ng/g for ANA-a and 0.12 ng/g for CYN [71].

The methods for the analysis of only BMAA in tissue were also limited in scope. Because it is a small amino acid and highly hydrophilic, studies that concurrently analyze BMAA with these other toxins are limited because of the need for either chemical derivatization or specialized HILIC LC–MS/MS methods, as seen in [60–62]. In general, the extraction of BMAA from tissue has been accomplished with an acidic solvent, but the need for hydrolysis, derivatization, and clean-up steps to reduce matrix interferences have made the process lengthy and potentially incompatible with other toxins. Following extraction, analysis was performed using HPLC–FLD or LC–MS/MS, with measurements ranging from micrograms/gram to milligrams/gram. The data from both analysis methods are generally comparable, indicating that the use of HPLC–FLD could be a beneficial first step or screening method. Studies support the use of LC–MS/MS as a necessary confirmation tool [115,116].

Most methods for STX detection in complex matrices usually involve extraction under acidic conditions (with hydrochloric acid or acetic acid, for instance) and often at an elevated temperature [55–58]. Detection was accomplished with ELISA, HPLC–FLD, LC– MS/MS, or LC–qTOF MS, with DLs mostly in the micrograms/kilogram range. The ELISA method is sensitive and allows for rapid screening of a large number of samples, and LC– MS/MS serves as a useful confirmation tool. STX is stable at high temperatures, which is an important consideration for food safety; it also plays a role in many extraction procedures that require the sample to be boiled [55–59,117]. STX is known to persist for over 90 days and is considered extremely stable even at a high temperature and low pH [38]. Although it can be stored in acid without loss for many decades, some studies suggest that it may not be stable at pH > 8, even at ambient temperature [118]. This suggests that STX may not be stable in water and bivalves unless it is stabilized with acid. The stability also depends on the chemical structure of specific compounds; GTX1, GTX4, and NEO variants are less stable at acidic pH than GTX2, GTX3, and STX [119]. Overall, the stability of various STX variants is an unsolved analytical problem that is in need of continued study.

Significant questions and uncertainties remain with regard to the toxins associated with *P*. *parvum* at this time. Sequential extraction of cultures with various solvents in [47] identified material with cytotoxic properties, including fatty acids and fatty acid amides, but no prymnesins. However, in [63], two toxins classified as prymnesins were observed by LC–MS/MS, following multiple preparative and clean-up steps, including chlorophyll removal and post-extraction solid-phase extraction. At this point, it remains unclear which toxins are the specific causative agents for *P. parvum* toxicity, a problem exacerbated by a lack of commercially available standards from which to perform toxicity assessments.

One of the emerging needs in analytical laboratories is for methods suitable for quantifying multiple cyanotoxins in a single workflow. As cyanobacterial blooms become more common and severe, the need for rapid analysis is only going to increase and methods suitable for screening and quantifying multiple classes of toxin at once will greatly improve sample throughput. In Table 3 are listed 20 studies relying on multi-toxin methods.

One approach for extracting multiple toxins is to attempt a compromise extraction mixture that is suitable for both hydrophilic and more hydrophobic toxins, as described in [68–70]. In these studies, the percentage of organic solvent in the extraction mixture was reduced to ~50%, rather than the 75–90% range typical for MCs, to improve the recovery of the more hydrophilic classes of contaminants, including CYNs, ANA-a, and STXs. In [68], an acidic methanol:water (50:50) mixture was used to extract benthic algae and analysis was performed using LC–MS/MS and LC–qTOF. In this study, two STX variants were observed in field samples at 209–270 ng/g, but no detections were seen for MCs, NODs, CYN, or ANA-a. The matrix fortification of dried tissue prior to extraction found that all five classes of contaminant were recovered with 80–105% efficiency, with the caveat that the only MC congeners used were MC-LR and MC-RR, two of the most hydrophilic MCs. In [69], in contrast to a single ~50% organic extraction, sequential extractions were performed in fish and shellfish, first with 100% methanol, followed by an acidic water:acetonitrile (55:45) mixture, with the two pooled before further processing. This study was designed to be as broadly suitable as possible for screening purposes, with two chromatographic methods used, HILIC for hydrophilic toxins and reverse phase for the more lipophilic toxins. As a part of method validation, an extensive matrix spike evaluation was performed for numerous MCs, STX, and other toxins outside the scope of this review, such as domoic acid, and okadaic acid. The authors found that recoveries ranged significantly, with many within an 80–120% window but others, particularly ANA-a, showing around 200% recovery of the spiked amount. According to them, this was mostly likely the result of matrix interferences in the analysis procedure, which is not surprising given the exhaustive extraction procedure and tissue matrix, but without isotopically labeled materials, confirmation of this hypothesis is impossible. In addition, the DLs in this study were somewhat higher than typical, with DLs of 150 ng/g for MCs and 600 ng/g for the more hydrophilic toxins.

In another multi-toxin study, [71], a method to measure MCs, NOD, ANA-a, CYN, and STX in tissue using LC–MS/MS was reported. Because of differences in chemical properties, the study relied on two separate extraction procedures, one for ANA-a, CYN, and STX and the other for NOD and MCs (Table 3). Once prepared, the samples were subject to one of two LC–MS/MS methods to separate and quantify the compounds, with C18-based chromatography for NOD/MCs and HILIC for ANA/CYN/STX. As described above, the technique of isotope dilution was used in this procedure to track recovery and quantification of ANA-a, CYN, and a subset of MCs through extraction and analysis. Matrix effects for MCs were seen in fish tissue, with magnitudes ranging from −26% to −58%, which were corrected through the internal standards to −16% to 10%. Comparable matrix suppressions of −44% to −50%, corrected to 1.4 to 3.4%, were observed for ANA-a and CYN, respectively. Similar magnitudes were seen in water extraction. Extensive method validation was performed using matrix spikes to determine DLs, from which MDLs were determined to be 0.12 to 0.70 μg/kg in tissue and 4 to 80 ng/L in water. However, validation of this procedure in fish exposed to exogenous MCs in [71] was not successful in producing positive results, which the authors attributed to the short duration of exposure, but MCs were detected in environmental samples. Overall, this study illustrates the utility of the use

of isotopically labeled materials in these multi-toxin methods to improve the tracking of interferences and extraction efficiencies.

Method development and optimization of techniques to extract and analyze toxins from different matrices rely on spiking the matrix with the target analytes, processing these samples through the optimized workflow, and reporting recoveries of the spiked compounds and precision as a metric of method performance. While this process may work in the case of simple matrices, such as water or algal cells, it may only partially mimic toxin-containing environmental samples from more complex matrices, such as zooplankton or tissue, as noted in [71]. For example, in spike recovery studies involving tissue, either the sample is spiked with the target toxins using a syringe needle inserted into the tissue prior to homogenization or the homogenate is spiked with the target toxins, followed by mixing prior to the processing of these samples through the optimized workflow. However, these spiked toxins may not entirely mimic the way that natural toxins are bound in tissue samples in the environment. This may lead to biases in estimated recoveries, especially if the spiked toxins are more readily amenable to the extraction process than the natural toxins, leading to incorrect quantifications of these toxins in complex matrices. While spike recovery studies are a good first step for developing and optimizing methods for toxin detection, method validation with different types of samples collected from HAB sites will provide an insight into the reliability of the methods.


**Table 3.**

Selected list of

publications

summarizing

 methods for toxin detection

 in various matrices.











(LC–QqQMS),

 and

quadrupole–time

 of flight mass

spectrometry

(LC–QqTOFMS).

### **6. Conclusions and Recommendations**

This review documents the methodologies currently used to measure cyanotoxins and prymnesins in complex matrices and to assess the advantages and limitations of the various techniques summarized in this paper.

Due to the diversity in chemical structures and properties among cyanotoxins, it is a significant challenge to develop and validate procedures for consistent and high-yield extraction from various matrices, including tissues. In individual toxin methods, optimum conditions can typically be achieved through validation experiments, such as the consensus for extractions using 75–90% methanol:water for MCs/NODs consistently seen in the literature, although variations in matrix constituents such as lipid or protein content could potentially result in varied outcomes. The use of matrix spikes, where known amounts of toxins are fortified to sample matrices, can help further quantify the various matrix effects and interferences in extraction and analysis of cyanotoxins and add confidence to an analytical workflow. This is particularly important in cases where ambient detection of toxins was negative, as in studies such as [55], where only spiked samples showed detectable levels of toxins to allow for evaluation of recovery efficiencies.

As the breadth of cyanotoxins known to be present in the environment continues to increase, it is clear that methods for quantifying single classes of toxin may eventually overwhelm analytical facilities. A single method capable of extracting and detecting many or all cyanotoxins would be an ideal goal for the future. Several publications describe such approaches using mass spectrometric detectors for a subset of contaminants of interest, but as expected from a group of compounds with diverse chemistry, there are significant limitations in recoveries during sample processing, chromatographic performance, and sensitivity [104–113]. For example, in [69], MCs, NOD, ANA-a, CYN, and STXs were all analyzed using LC–MS/MS following extraction, with sequential extraction first with methanol, then a mixture of acetonitrile:water (45:55), which were pooled for analysis. Method recoveries in this study varied between 80 and 200% and the reported detection limits were high, most likely due to matrix interferences. In other cases, multiple complementary workflows for hydrophilic and hydrophobic toxins were developed, as in [71], and showed improved performance.

To control for variation in extraction efficiencies, one approach is to add isotopically labeled internal standards to the sample prior to extraction, allowing for quantification of recovery percentages through the extraction and analysis procedures. This can help compensate for recovery bias and matrix suppression/enhancement and is a recommended best practice but is limited in scope to those compounds for which isotopically labeled materials are readily available and when analytical methods can differentiate the native and labeled material (e.g., mass spectrometric methods). Recently, labeled materials have become available for a broader set of cyanotoxins, including MCs, STX, ANA-a, and CYN, which has made this approach feasible. Examples of this as applied to toxin measurement can be found in reference [71], where recoveries were assessed and corrected for both extraction efficiency and matrix effects using labeled toxin analogs.

Analysis with LC–MS/MS has become an essential tool for cyanotoxin detection and it can potentially be used for the concurrent analysis of multiple classes of toxins due to its rapid scan rate and ability to cycle polarity from positive to negative ion modes. LC–MS/MS is best employed for targeted screening for toxins, particularly where both native and isotopically labeled standards are available. In contrast, LC coupled with HRMS or TOF is better suited to detect unknown toxins or those for which standards are not available, but a more complete discussion is beyond the scope of this article [82].

An alternative to LC–MS/MS methods could be an ELISA microplate test strip that contains antibodies for multiple toxins. This would require design in a way that similar incubation periods for the binding step could be achieved, as the typical ELISA workflow has specific duration of each step of binding and rinsing of the plate. If this was achievable, it would allow simultaneous testing for multiple toxins without requiring expensive and bulky experimental apparatus, such as mass spectrometers. Because ELISA methods tend

to be semi-quantitative for reasons described above, it is a best practice that detections be confirmed by LC–MS/MS or some other more specific technique where possible. One potential alternative in the future could be electrochemical biosensors that contain a biological recognition element that specifically reacts with the target of interest; these are in active development for use with water samples. The suitability of these biosensors to fish tissue and other matrices still needs to be determined.

Another complicating factor for assessment of toxins in tissue matrices is common to any analytical workflow, namely, the amenability of these bound toxins to be extracted into a solution for measurement. Literature results suggest a significant fraction of cyanotoxins in tissue samples could be bound to organs or otherwise unavailable through the extraction procedure [106]. Ref. [78] showed significant differences in MC measurements in tissue using ELISA and GC–MS/MS following Lemieux oxidation, which they attributed to the oxidation technique freeing bound analytes for measurement. Ref. [106] and references therein suggest that the total concentration could potentially be an order of magnitude higher. The level of underestimation and impact on risk assessments and health outcomes should be investigated in future studies.

An additional challenge is that many emerging cyanotoxins do not have commercially available standards, and, in other cases, toxin standards are prepared from minute amounts of natural sources or unidentified sources of unknown purity, making it difficult to accurately quantify toxin concentrations. In this review, prymnesins and other potential toxins produced by *P. parvum* are one such class of toxins, which at present are only available through laborious culturing, extraction, and isolation, as described above, and no commercial sources exist. Even in cases where materials can be procured, the use of certified reference standards (CRMs) with exact concentrations is recommended to improve confidence in the absolute concentrations; however, these are not available for many contaminants, and even for those with CRMs available, such as MCs, only a subset of variants may be covered. It is recommended that laboratories monitor their cyanotoxin standards over time for variations in purity/concentrations and where possible obtain standards from multiple vendors if certified materials are not available.

There is a great deal of variability in the analytical procedures presently being used to prepare, extract, and analyze for cyanotoxins associated with harmful algal blooms in diverse sample matrices. In this review, the most common procedures were highlighted, and best practices were identified. It is clear that there is a compelling need for more standardized, reliable, and affordable screening methods compatible with tissue and similar matrices exposed to cyanotoxins, particularly as new toxins are continually being identified in the environment. For LC–MS/MS methods, the best approach is to ensure the use of extensive quality control procedures, including evaluating matrix interferences, though matrix spikes where possible and using labeled surrogate and internal standards to monitor method performance across both the extraction and analysis phases (e.g., [71]). For the commonly used immunological methods, because they are incompatible with the use of labeled standards, researchers should instead ensure they perform similar studies of matrix performance to determine cross-reactivity parameters and potential interferences as part of their method validation procedures.

**Author Contributions:** Conceptualization, D.S., T.T.S., R.V., and J.L.; resources, J.L.; data curation, D.S. and J.L.; writing—original draft preparation, D.S., T.T.S., R.V., L.D., A.O.T., and J.L.; writing review and editing, D.S., T.T.S., R.V., H.M., D.T., L.D., S.F., A.O.T., and J.L.; visualization, D.S.; supervision, R.V. and J.L.; project administration, J.L.; funding acquisition, J.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.**Disclaimer:** The views expressed in this article are those of the author(s) and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. This research was supported in part by U.S. EPA's Office of Research and Development, Cincinnati, OH, under Contract No. 68HERC20D0029. The use of trade names or commercial products is for identification only and does not imply endorsement by any agency of the U.S. government.

### **References**


### *Review* **Anabaenopeptins: What We Know So Far**

**Patrick Romano Monteiro 1,2,\*, Samuel Cavalcante do Amaral 1, Andrei Santos Siqueira 2, Luciana Pereira Xavier <sup>1</sup> and Agenor Valadares Santos 1,\***


**Abstract:** Cyanobacteria are microorganisms with photosynthetic mechanisms capable of colonizing several distinct environments worldwide. They can produce a vast spectrum of bioactive compounds with different properties, resulting in an improved adaptative capacity. Their richness in secondary metabolites is related to their unique and diverse metabolic apparatus, such as Non-Ribosomal Peptide Synthetases (NRPSs). One important class of peptides produced by the non-ribosomal pathway is anabaenopeptins. These cyclic hexapeptides demonstrated inhibitory activity towards phosphatases and proteases, which could be related to their toxicity and adaptiveness against zooplankters and crustaceans. Thus, this review aims to identify key features related to anabaenopeptins, including the diversity of their structure, occurrence, the biosynthetic steps for their production, ecological roles, and biotechnological applications.

**Keywords:** cyanobacteria; peptide; NRPS; anabaenopeptin

**Key Contribution:** The present work approach various features of Anabaenopeptins, including structural, biosynthetic and regulatory aspects as well as their ecological and biotechnological importance.

### **1. Introduction**

Cyanobacteria are photosynthetic microorganisms widely distributed in the world. They can inhabit several types of ecosystems, including aquatic and terrestrial. These microorganisms produce a great variety of bioactive compounds, which have been investigated mainly due to their biotechnological potential and environmental relevance [1–3]. Cyanotoxins are among the most studied compounds originated from cyanobacteria since they are capable of negatively affecting human and animal health [4,5]. These metabolites can vary drastically concerning their action mechanism and chemical structure, which include peptides, alkaloids, and lipopolysaccharides [6–8]. The majority of publications related to peptides from cyanobacteria have mainly focused on the class of microcystins with over 300 characterized variants [9,10]. However, cyanobacteria usually do not exclusively produce a single class of compounds, given that specific strains are co-producing different groups of secondary metabolites [11].

Other peptides beyond microcystins have been poorly explored, lacking information mainly in the environmental sciences [11]. These several metabolites are known for their potent inhibitory properties against several enzymes in nanomolar concentrations, resulting in toxic effects [12,13]. Moreover, similar to microcystins, they have been regularly detected in diverse environments [14]. In certain regions, their occurrence is more pronounceable than microcystins themselves [15]. However, information about the concentrations which are encountered is rarely reported [11]. Cyanobacteria have developed different peptides

**Citation:** Monteiro, P.R.; do Amaral, S.C.; Siqueira, A.S.; Xavier, L.P.; Santos, A.V. Anabaenopeptins: What We Know So Far. *Toxins* **2021**, *13*, 522. https://doi.org/10.3390/ toxins13080522

Received: 29 March 2021 Accepted: 25 May 2021 Published: 27 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

as a protection mechanism against parasites [16]. Concerning their origin, some peptides as microviridins and cyanobactins are produced via ribosomal whereas others as microginins and aeruginosins are synthesized by non-ribosomal pathways [13,17,18].

Among the most recurrent peptides encountered in the environment are anabaenopeptins (APs), a family of cyclic peptides containing six amino acid residues [19]. They have been found in an enormous variety of cyanobacteria isolated from both the aquatic and terrestrial environments, including *Anabaena*, *Nostoc*, *Microcystis*, *Planktothrix*, *Lyngbya,* and *Brasilonema* [12,20–24]. In their general structure is a well-conserved Lysine (Lys) residue in D-configuration, which is responsible for the ring formation and five additional variable amino acids, either proteinogenic or non-proteinogenic, resulting in 124 described AP variants from cyanobacteria (Supplementary Table S1). [19]. Besides their structural variety, molecules belonged to this group exhibit an impressive functional diversity, which includes inhibitory activity for proteases, phosphatases, and carboxypeptidases [22,25,26].

The enormous structural diversification of anabaenopeptins can be attributed to the low substrate specificity of some enzymes involved in their synthesis as well as the presence of alternative starter modules [16]. Their production is strongly influenced by environmental factors [27]. Besides that, because of their diversified bioactive properties, they exhibit an elevated biotechnological potential. This review aims at presenting the main researches on anabaenopeptins, emphasizing their general characteristics, biosynthesis as well as ecological and biotechnological relevance.

### **2. Structures of Anabaenopeptins**

Being non-ribosomally synthesized, anabaenopeptin structures comprise a ring of five amino acids connected through an ureido linkage to an exocyclic amino acid. Thus, its general structure is represented by X1-CO-[Lys2-X3-X4-MeX5-X6], where the bracket represents the cyclic region of this peptide and X are variable amino acids according to their positions represented by the superscript numbers (Figure 1). Its ring is formed by cyclization of the C-terminal carboxyl of the amino acid at position 6 to the ε-NH3 of the well-conserved D-lysine at position 2. Furthermore, the α-amino group of Lys is connected to the exocyclic amino acid X1 via an ureido bridge. Due to its non-ribosomal nature, proteinogenic and non-proteinogenic amino acids are usually detected in this hexapeptide [19].

**Figure 1.** The general structure of the class of Anabaenopeptins. X corresponds to different amino acids in their respective positions represented by the superscript numbers.

This family of peptides is predominantly found in cyanobacteria, but they were also detected in some sponges [28–33]. However, Anabaenopeptins from cyanobacterial origins demonstrate a well-conserved D-Lys, while the other amino acids are in L configuration and vary in residues and modifications (e.g., acetylation, methylation) [19,34]. In comparison, Anabaenopeptins derived from sponges can have both D- and L-configuration of Lys residues at the second position. Besides these differences, some features in common are also encountered, such as the frequently N-methylated amino acids at position 5 and homoamino acids at position 4. However, exceptions are also found, for example, Paltolides A–C, a subgroup of anabaenopeptin-like peptides that have in common a tryptophan residue at the C-terminus linked to ε-amine of the N-terminal Lys, and Leucine (Leu) in L-configuration at positions 4 and 5, and an L-Alanine (Ala) residue at position 3. Furthermore, Paltolide A is the first example of this class of peptides where the amino acid at position 5 lacks an N-methyl group [28–33].

The first Anabaenopeptins detected were Anabaenopeptins A and B (Figure 2; Supplementary Table S1) by Harada and co-workers in 1995 [20]. These peptides were isolated from *Anabaena flos-aquae* NRC 525-17, where they were co-produced with Microcystins (MCs) and the neurotoxic alkaloid anatoxin-A. Both peptides share the same cyclic sequence and structure, differing only at the exocyclic position: (Arginine/Tyrosine)-Lys-Valine-Homotyrosine-NMethylAla-Phenylalanine. Due to their origin, these peptides were then named after their producer. Following the first detection of this new peptide, Fujii and co-workers [35] identified Anabaenopeptins A-D from different *Anabaena* and *Oscillatoria* strains, as well as from *Nodularia spumigena*. APs C and D differ solely by the exocyclic amino acid, harboring a Lys and a Phenylalanine (Phe), respectively, and sharing the same pentapeptide ring with APs A and B. Furthermore, in the same year, Sano and Kaya [36] identified a peptide named Oscillamide Y, which was obtained from *Oscillatoria agardhii* NIES-610. Following the same nomenclature, Sano and colleagues [25] further characterized both Oscillamide B and C (known as Anabaenopeptin F) from *Planktothrix agardhii* CCAP 1459/11A and *P. rubescens* CCAP 1459/14. Besides their different nomenclature, Oscillamides peptides also possess the same common features of anabaenopeptin-peptides.

The cyanobacteria *Oscillatoria agardhii* NIES-204 had been assessed regarding AP production by two different research studies. During the first approach, only Anabaenopeptin B had been detected [37]. Later, Shin and co-workers [38] were able to characterize two new structures from the same organism: Anabaenopeptins E and F, which differ at those residues in positions 3 and 4. Later, two new AP structures were identified in *O*. *agardhii* NIES-595, then named Anabaenopeptins G and H, diverging by Tyrosine (Tyr) and Arginine (Arg) in position 1, respectively: (Tyr/Arg)-Lys-Isoleucine-Homotyrosine-NMethylhomotyrosine-Isoleucine [26].

The first unicellular cyanobacterium strain to be identified as an Anabaenopeptin producer was *Microcystis aeruginosa* Kutz. This freshwater strain was able to biosynthesize the anabaenopeptin-type Ferintoic Acids A and B [39]. Another *M*. *aeruginosa* strain and an environmental sample containing mostly *Microcystis* cells demonstrated to contain the Non-Ribosomal Peptide Synthetase (NRPS) apparatus for Anabaenopeptins B and F production. Also, the same work concluded that the filamentous cyanobacteria *Planktothrix agardhii* HUB011 produced the Anabaenopeptin G [40].

Kodani and co-workers [41] evaluated the presence of anabaenopeptins in an environmental sample from Lake Teganuma. Besides the identification of microginins and micropeptins, a newly found AP was characterized: Anabaenopeptin T. However, this nomenclature did not follow any specific order, as Anabaenopeptin I and J had been only used for the new peptides obtained from *Aphanizomenon flos-aquae* NIES-81 and identified by Muramaki and co-workers [42], one year later from this previous work.

**Figure 2.** Structures of anabaenopeptins A–J [20,26,35,38,42] and T [41].

Some of those conserved features from APs can also be visualized in other cyanopeptides. Veraguamides A-G are cyclic hexadepsipeptides, and they do not possess any exocyclic residue. Lyngbyastatin peptides demonstrate elastase, trypsin, and chymotrypsin inhibitory properties. Their structures consist of a 6-member ring coupled to a chain of 2 exocyclic residues and can bear modified and unusual residues. Also possessing a 6-member ring structure and 2 exocyclic amino acids, Tiglicamides A-C were obtained from *Lyngbya confervoides* [43].

In addition, there are several classes of toxic peptides frequently detected in cyanobacteria, each one presents the main structure codified by a set of NRPS genes. Besides Anabaenopeptins, Microcystins, Cyanopeptolins, Aerucyclamides, Aeruginosis and Microginins are examples of well-characterized cyanopeptides. Microcystins also share some resemblances to APs as the former possesses a ring structure, but it is comprised of 7 residues and no exocyclic amino acid. Also produced by the NRPSs apparatus, MCs bear D-amino acids, and the unusual residue (2S,3S,8S,9S)-3-amino-9-methoxy-2,6,8-trimethyl-10-phenyldeca-4,6-dienoic acid, also known as Adda, which is present in all MCs. One example of MC variant is Microcystin-LR, which is composed by (1)-Ala, (2)-Leu, (3)-Nmethyl-Asp, (4)-Arg, (5)-Adda, (6)-Glu, and (7)-N-methyl-dehydro-Ala, where LR refers to 2 and 4 variable positions among MCs. It was observed in different cyanopeptides inside this class that there are different variants according to these positions, contributing to their structural diversity [10,11].

Cyanopeptolins are depsipeptides containing a 6-amino acid ring bearing a side chain with 1–2 residues and modified residues, such as 3-amino-6-hydroxy-2-piperidone. Cyanopeptolin A is one example of this class of cyanopeptides and is composed by (1)-fatty acid, (2)-Arg, (3)-Ahp, (4)-Leu, (5)-methyl-Phe, (6)-Val, and (7)-Thr, in this case, the β-lacton ring is formed between Arg and Thr residues and positions 2, 4, 5 and 6 are variable. Using Anabaenopeptin A as reference (Figure 2), its structure is (1)-Tyr, (2)-D-Lys, (3)-Valine, (4)-Homotyrosine, (5)N-methyl-Alanine, (6)-Phenylalanine [44]. Positions 1, 3, 4, 5, and 6 are variable concerning APs (Figure 1) and the ureido bond is formed between 1 and 2 residues. Aerucyclamides are entirely cyclic peptides, Aerucyclamide A is composed by (1)-dehydro-Thr, (2)-Gly, (3)-thiozole, (4)-Ile, (5)-dhCys, and (6)-Ile, in this case, variations were reported in positions 2, 3, 4 and 6. Different from the cyanopeptides listed until now, Aeruginosins and Microginins are linear peptides. Aeruginosin KB 676 is formed by (1)-Hpla, (2)-Ile, (3)-Choi and (4)-Arg, only position 2 presents variation with amino acid substitution, and radical changes occur in positions 3 and 4. Finally, Microginin 713 is formed by (1)-Ahda, (2)-Ala, (3)-Val, (4)-N-methyl-Tyr, and (5)-Tyr, in this case, positions 2, 3, 4, and 5 had substitutions reported [11].

Structurally, despite the major amino acid variability, Microcystins, Cyanopeptolins, and Anabaenopeptins are most similar. Microcystins and Cyanopeptolins are heptapeptides and Anabaenopeptins are hexapeptides and comparting these structures, it is possible to distinguish a ring core and a linear region. Although Microcystins are technically cyclic peptides, the Adda moiety projected outside the ring may act like the fatty acid in Cyanopeptolin A or Tyrosine in Anabaenopeptin A. The Adda moiety is crucial for MCs inhibition towards phosphatases, as its long linear chain can penetrate the enzyme active site together with other side chains, having a similar role as the exocyclic residue of APs (Tyrosine from Anabaenopeptin A), as it will be further discussed in Section 7 [11,45]. This exocyclic or even protuberant residue was not observed in Aerucyclamides that only present a cyclic structure or Aeruginosins and Microginins, which are linear structures [11]. Therefore, cyclic peptides bearing exocyclic residues and unusual and D-configuration amino acids are also found in cyanobacteria, however, the ureido linkage in cyanopeptides is, so far, an exclusive characteristic of Anabaenopeptins.

All these AP variants named here so far are structurally related, differing only by amino acid substitutions responsible for their diversity (Supplementary Table S1). However, these peptides do not possess a fully systematic nomenclature, which can make it difficult to identify them as a member of a certain group of oligopeptides with similar structure. This fact is not specific to Anabaenopeptins, but cyanopeptides in general, as their denominations are frequently referring to the taxon or geographic locality from which the oligopeptide had been isolated, and also information regarding molecular weight, specific residues, or even the strain number can be used as a suffix, and some example can be seen applied to APs [11]. One example of a variant with a distinct name is the Schizopeptin 791 (Figure 3), which was named after the terrestrial cyanobacteria *Schizothrix* sp. IL-208- 2-2 (Schizo-), its peptide nature (-peptin) and its molecular weight of 791 Da (791) [46]. Lyngbyaureidamides A and B are Anabaenopeptins named after their isolation from the filamentous freshwater cyanobacterium *Lyngbya* sp. SAG 36.91. These anabaenopeptin-like peptides also have an uncommon feature due to the presence of a D-Phenylalanine in the exocyclic position, being the only APs bearing an amino acid in D-configuration in this position [47]. Obtained from the marine *Lyngbya confervoides*, Pompanopeptin B is an anabaenopeptin-type peptide bearing in the fifth position the *N*-methyl-2-amino-6-(4 hydroxyphenyl)hexanoic acid (N-Me-Ahpha), a methylated form of a residue found in Largamide C [23]. Nodulapeptins are also anabaenopeptin-like peptides and they were first identified by Fujii and co-workers [48] in the toxic *Nodularia spumigena* AV1. Among the different nomenclature of this class of cyclic hexapeptide, Nodulapeptin is one of the most used and it is often associated with the presence of Methionine (Met) or Serine (Ser) residues in position 6 of anabaenopeptin-like structures [49]. Isolated from the cyanobacteria *Tychonema* sp., Brunsvicamides A-C share a high resemblance to anabaenopeptin-like peptides obtained from sponges, thus indicating their possible cyanobacterial origin. These peptides obtained from a *Tychonema* sp. strain did not possess any homoamino acid and have a L-Lys besides D-Lys, in addition, Brunsvicamide C has an *N*-methyl-N'-formyl-Dkynurenine unit in position 5 [50].

Besides these distinct nomenclatures and structures for Anabaenopeptins obtained from cyanobacteria, this class of peptides can also be found in sponges, which were the initial organisms to be identified the first anabaenopeptin-related compound, not in a cyanobacterium [31,32]. Konbamide and Keramide A (Table 1 and Figure 4) were isolated from the marine sponge *Theonella* sp., which showed distinct features from cyanobacterial anabaenopeptins having a cyclic hexapeptide structure and the presence of an ureido bond. Both variants have L-Lys residue and also they contain a modified Tryptophan (Trp) residue at position 6. Konbamide had 2-bromo-5-hydroxytryptophan (2'Br-Trp) in position 6; in comparison, Keramide A possessed a 6-chloro-5-hydroxy-*N*-methyltryptophan (5'OH-6'ClTrp) in position 5 [31,32]. Keramide L was detected in *Theonella* sp. SS-342 together with Keramide K (a thiazole-containing cyclic peptide not belonging to anabaenopeptin-class). Keramide L shared similar features to Konbamide and Keramide A, having a modified Trp residue in position 5: a 6-chloro-*N*-methyltryptophan (NMe-6'ClTrp) residue [30]. Besides, the marine sponge *Theonella swinhoei* demonstrated to produce a similar class of peptides, named Paltolides A-C, and 3 additional peptides which were previously detected in the Australian sponge *Melophlus* sp. [51]. Unlike Konbamide and Keramide A, Paltolides A-C and the 3 unnamed peptides have a D-Lys residue in position 2, but 4 of these structures had modified Trp residues, and 3 of these modifications were also related to the presence of a halogen element (Bromo or Chlorine) similar to Konbamide and Keramides A and L [28]. Other examples from sponges are Mozamides A and B produced by *Theonella* sp. from Mozambique. Similar to other anabaenopeptin-like peptides from sponges, both Mozamides have the same modified Trp (*N*-methyl-L-5 -hydroxytryptophan) residue in position 5 and an L-Lysine at position 2. However, both have an amino acid in D-configuration in position 3: D-valine in Mozamide A; and D-Isoleucine in Mozamide B [29].

**Figure 3.** Example of different nomenclatures to anabaenopeptin-like structures [20,23,39,46–48,50].

**Figure 4.** Structures of anabaenopeptin-like peptides obtained from sponges [28–33].

**Table 1.** Amino acid composition of anabaenopeptin-like peptides obtained from sponges. Amino acids are considered in L-configuration unless otherwise defined. Ala: Alanine; Arg: Arginine; Ile: Isoleucine; Leu: Leucine; Lys: Lysine; MeLeu: N-methyl-Leucine; Phe: Phenylalanine; Trp: Tryptophan; Allo-Ile: Allo-Isoleucine; 2'BrTrp: 2-bromo-5-hydroxytryptophan; NMe-6'ClTrp: 6-chloro-N-methyltryptophan; 5'OHTrp: 5'-hydroxytryptophan; 6'BrTrp: 6'-bromotryptophan; 5'OH-6'Cl Trp: 6'-chloro-5'-hydroxytryptophan; 6'ClTrp: 6'-chloro-tryptophan; NMe-5OHTrp: N-methyl-5'-hydroxytryptophan; NMe-5'BrTrp: 5'-Bromo-N-methyltryptophan.


To this date, 124 Anabaenopeptins variants have been identified (Supplementary Table S1 and Table 1), making it difficult to compile a systematic nomenclature and to evaluate every individual feature. To overcome this problem, several works have named new Anabaenopeptins according to their mass or even mass and specific information regarding the strain or location in which it was isolated [34]. However, some Anabaenopeptins have unusual features, typically not found in most of these peptides belonging to this class. The previously cited Ferintoic acid B has an allo-isoleucine (Allo-Ile) residue in position 6, similar to both Mozamides, which have the same residue in position 1 [29,39]. None of the Brunsvicamides have a homoamino acid and they are also the only examples of this class of peptide with L-Lys in position 2 from cyanobacteria [50]. A group of Anabaenopeptins named SA demonstrated to possess some uncommon features. Anabaenopeptin SA1, SA4, SA5, and SA7 (Figure 5) have a 5-phenylnorvaline (PNV) in position 4, which has not been previously found in any other peptide of this class. Also, Anabaenopeptin SA8 was the only one to have a 6-phenylnorleucine (PNL) residue in position 4. Additionally, APs SA2 and SA13 have a Ser residue in position 5, which has not been previously detected in the same position, but in position 6, as in Nodulapeptins [12]. Anabaenopeptins 877B, 905, 862, and 896 are some of the few examples of N-ethylated peptides [24]. The only example of homoarginine in Anabaenopeptins is from AP KT864, which has this residue in position 1 [52]. A residue of glutamate at the exocyclic position has been only described in one Anabaenopeptin: the variant MM823 [22]. Besides its common presence in position 3, Valine (Val) has been only detected in position 4 in Nodulapeptin 855C [34]. Thus, demonstrating that Anabaenopeptin peptides have huge structural diversity.

**Figure 5.** Examples of untypical features of anabaenopeptins from cyanobacteria [12,22,24,34,52,53].

Unusual Anabaenopeptins lacking residues in their structures are also visualized. Anabaenopeptin 679 is the only example of an anabaenopeptin-like peptide where the exocyclic residue is absent (Figure 6). This anabaenopeptin possesses solely the ring structure, which shares the same amino acid sequence as anabaenopeptins A, B, C, D, and

J [53]. Additionally, Namalides are anabaenopeptins with an atypical structure lacking two amino acids from the macrocycle. They are cyclic tetrapeptides firstly identified in the marine sponge *Siliquariaspongia mirabalis* [54] and then detected in cyanobacteria, such as *Sphaerospermopsis torques-reginae* ITEP-024 [55] and *Nostoc* sp. CENA543 [56], producing namalides B and C, and namalides B, D, E, and F, respectively.


**Figure 6.** Example of Anabaenopeptins with unusual structures lacking one amino acid (Anabaenopeptin 679) and two amino acids (Namalide B) residues [53,55,56].

### **3. Occurrence of Anabaenopeptins and factors involved in their expression**

Besides its great structural diversity, it appears that those peptides are usually detected in some specific genera of cyanobacteria. As can be seen in Table 2, the majority of cyanobacteria able to biosynthesize anabaenopeptins belong to genera such as *Anabaena, Microcystis, Nodularia, Oscillatoria,* and *Planktothrix*. Except for *Microcystis*, those genera are filamentous cyanobacteria belonging to the order Nostocales and Oscillatoriales. Regarding the unicellular genus, as will be discussed later (Section 4), the Anabaenopeptin NRPS cluster seems to be horizontally transferred to *Microcystis* [57]. Also, Anabaenopeptins have been detected in genera *Aphanizomenon*, *Brasilonema,* and *Desmonostoc* belonging to Nostocales order. Similar to *Oscillatoria* and *Planktothrix*, the genus *Lyngbya* belonging to Oscillatoriales demonstrated to produce anabaenopeptins. In addition, two strains of unicellular genera of cyanobacteria that belonged to Synechococcales, *Schizothrix* and *Woronichinia*, proved to have the ability to produce Anabaenopeptins. Strains belonging to filamentous cyanobacteria tend to present a higher quantity of gene clusters than the unicellular strains [58]. The heterocyst presence in some members of the order of Nostocales can also confer some advantages in the Anabaenopeptin production since this differentiated cell provides the propitious microenvironment for the nitrogen fixation, which is an element required in large amount for the production of cyanopeptides [59].


**Table 2.** Occurrence of anabaenopeptins in different cyanobacteria genera and species.


**Table 2.** *Cont.*

*Anabaena flos-aquae* NRC 525-17 Anabaenopeptins A and B [20] *Brasilonema* sp. 360 Anabaenopeptin 802A [24] *Brasilonema* sp. 382 Anabaenopeptin 802A [24] *Brasilonema* sp. CT11 Anabaenopeptins 788, 802A, 802B and 816 [74] *Desmonostoc* sp. 386 Anabaenopeptins 848, 849, 862, 863, 877A, 877B, 891 and 905 [24] *Nostoc* sp. 352 Anabaenopeptins 841B, 855, 857 and 871 [24] *Nostoc* sp. 358 Anabaenopeptins 882 and 896 [24] *Nostoc* sp. ASN\_M Anabaenopeptins 808 \*, 828, 842 \*, 844 \* and 858 \*, [75] *Nostoc* sp. ATCC 53789 Anabaenopeptin SA9, SA10, SA11 and SA12 [12] *Nostoc* sp. KVJ2 Anabaenopeptins KVJ827, KVJ841, and KVJ811 [21] *Schizothrix sp.* IL-208-2-2 Schizopeptin 791 [46]

\* Anabaenopeptin variants with non-elucidated sequence.

A total of 45, 29, and 12 cyanobacteria strains from freshwater, marine and terrestrial environment have been analyzed for AP production, respectively. As seen in Table 2 and according to the literature [34,40,41,53,76–85], marine strains produced a total of 50 different variants of APs, in comparison to 43 and 34 variants from freshwater, and terrestrial strains, respectively (Figure 7). Thus, marine cyanobacteria demonstrate to produce a higher number of distinct APs variants in comparison to the remaining strains from different sources. However, APs from freshwater environments have the greatest diversity of amino acids in the majority of positions (Figure 8). Thus, these features could be associated to different obstacles faced in their respective environments as well as the fact that both belong to aquatic environments [86], however, this hypothesis requires further studies. Some of those APs are shared among different strains isolated from distinct environments: 2 anabaenopeptins (A and B) variants were detected in all ecosystems; in comparison, strains from both aquatic habitats had 13 APs variants in common (D, F, J, 807, NZ841, Oscillamide Y, and Nodulapeptins B, C, 855B, 871, 879, 897 and 915A); in contrast, only anabaenopeptin C were produced by both terrestrial and freshwater, and none Anabaenopeptin variant was shared by both terrestrial and marine strains.

According to Table 2 and Figure 7, there are AP variants shared among cyanobacteria strains from different environments according to the previous discussion. Anabaenopeptins A and B are the only variants detected in all habitats analyzed, and the only difference between those variants reside at the exocyclic residue. AP B is still the most recurrent among these oligopeptides in cyanobacteria (Table 2), corroborating with the previously raised hypothesis that this variant was the first cyanotoxin of this class to be emerged. [57]. Furthermore, the number of common anabaenopeptins variants increases when a comparison is made among strains only from aquatic habitats (freshwater and marine): Anabaenopeptins D, F, J, 807, NZ841, Oscillamide Y, and Nodulapeptins B, C, 855B, 871, 879, 897 and 915A. Besides their production by both freshwater and marine cyanobacteria, these prevalent oligopeptides seem to be more recurrent in marine environments, given that a higher number of cyanobacteria strains from this habitat are able to produce these APs comparing to freshwater, except for Oscillamide Y, which is more recurring in the latter. Among those variants, Nodulapeptin B is the most frequent in marine microorganisms. Besides, the only difference between the AP C (produced by freshwater and terrestrial strains) and both A and B variants is the exocyclic amino acid, and the former was not detected in marine cyanobacteria.

**Figure 7.** The number of Anabaenopeptins variants detected and shared among strains of cyanobacteria from different environments, including environmental samples.

**Figure 8.** Relative frequency (%) of amino acids in positions 1 and 3–6 of variants of anabaenopeptins characterized according to their environment (freshwater, marine and terrestrial). The total number of variants with elucidated sequences were 42, 47 and 29 for freshwater, marine, and terrestrial environments, respectively. Position 2 was omitted as the D-Lys residue being conservated among AP variants.

As seen in Figure 7, the environment can exert a crucial role in the biosynthesis of different APs, justifying their distribution in certain locations. The presence and frequency of certain amino acids in Anabaenopeptin structures can vary according to their respective source environment. Anabaenopeptins from both aquatic environments demonstrate to have Isoleucine as the most recurrent amino acid in position 1, while this same amino acid

was detected in only one AP variant in terrestrial strains (Figure 8). Phenylalanine was highly detected in position 1 of Anabaenopeptins isolated from terrestrial strains. Then, freshwater cyanobacteria may be promising biotechnological targets due to its highest diversity of amino acids in position 1, as the exocyclic residue is crucial for its inhibitory activity [12,34,45]. Regarding the variable position 3, Anabaenopeptins from freshwater and marine environments displayed a similar pattern of amino acid frequencies, Valine (Val) being the most frequent, followed by Ile and L-Methionine sulfone (MetO2). In contrast, terrestrial strains produce several AP variants with Ile in position 3, followed by Val and Leu, the latter being absent in this position on APs detected in aquatic environments. Homotyrosine (Hty) and Homophenylalanine (Hph) are the most found residues in position 4 among APs from all habitats analyzed, however, among terrestrial and marine strains Hph is more predominantly, while Hty is commonly observed in APs from freshwater strains. Except for Glycine (Gly) in some Anabaenopeptins from terrestrial strains, all the other residues in position 5 are N-methylated. APs from non-aquatic cyanobacteria do not harbor homoamino acids in the fifth position and, in addition, Asparagine is only detected in some of those variants in the respective position. Besides their detection in position 5, homoamino acids seem to be more persistent in position 4 from those APs analyzed. Position 6 has the highest richness of amino acids among AP variants obtained from marine environments, having incorporated 7 different residues, while this position in variants from freshwater habitats have assimilated 9 different amino acids, being the second most diverse site. Such heterogeneity in the last position in APs from aquatic strains is not clear, as the first amino acid residue demonstrated to be important in Anabaenopeptin interaction towards its enzyme target [12,34,45]. This array of several amino acids detected in position 6 is not visualized in Anabaenopeptins from terrestrial strains, where Phe was the amino acid more detected, similar to those APs from freshwater microorganisms.

The identification of the external physicochemical parameters involved in the regulation of these molecules can assist in controlling and assessing their risks [87]. Furthermore, this type of information can enable a better comprehension of their functions in producing organisms [27,88–91].

Microcystins, nodularins, and saxitoxins are among the most studied toxins from cyanobacteria, in which their gene clusters can operate independently, being, therefore, able to react oppositely when exposed to the same conditions [92,93]. The relationship between the gene cluster of Anabaenopeptins with these clusters within an individual strain is not well explored, demanding more detailed studies by a holistic approach, since enabling the study of various peptides at the same time [89,90].

Anabaenopeptins content per cell is strongly affected by environmental factors (Figure 9). Tonk and co-workers [27] investigated the effect of light intensity, temperature, and phosphorous concentration on the growth of the cyanobacterium *Anabaena* sp. 90 as well as its production of Anabaenopeptins A and C among some MC variants and with the anabaenopeptilide 90 B. This later belongs to an underexplored group of depsipeptides, which similar to the APs has the structure of a ring with a side chain, but without the ureido linkage. In the phosphorus-limited condition, all peptides were detected in a higher amount. These data match the result of Teikari and colleagues [94], who studied the same *Anabaena* strain and encountered higher quantities of transcripts belonged to anabaenopeptins, anabaenopeptilide, and microcystins gene clusters under low-phosphate conditions. Phosphate limitation also increases the content of protease inhibitors of the cyanobacterium *M. aeruginosa* NIVA CYA 43 [95].

**Figure 9.** Major factors involved in anabaenopeptin regulation in cyanobacteria.

In contrast, anabaenopeptilide 90B responded in a different way of Anabaenopeptins A and C to light exposure. The former increases considerably with light intensity while the others had their production reduced. These two peptide groups exhibit a compensatory dynamic, where the reduction of one is accompanied by the increase of the other. This strategy employed by cyanobacteria ensures the constitutive production of peptides with similar functions in an unstable environment, increasing its survival change. For example, anabaenopeptilides are also serine protease inhibitors, having near functions to APs in the cell. In respect to the temperature, its increase favored anabaenopeptins production but resulted in a slower growth rate [27]. Such observations demonstrate that APs production is a constitutive process and is not always positively related to the growth of *Anabaena* sp. 90, contrasting, therefore, with the argument that the best growth conditions are more favorable to toxins production.

A compensatory mechanism has been described by other authors in cyanobacteria [88–90,96]. A comparison of the oligopeptide profile of *M. aeruginosa* PCC 7806 and its microcystin-deficient mutant revealed that the loss of this toxin has as consequence the increased production of cyanopeptolins and aeruginosins [90]. Pereira and co-workers [96] observed similar behavior for different variants of MCs in *Radiocystis fernandoii* 28. Microcystin-RR exhibited an opposite response to those observed for microcystins YR, FR, and WR under distinct light conditions in this strain. This type of modification can strongly affect its final toxicity since MC variants display different toxicity degrees. Another possible alteration consists of a change of antifeedant potential. APs significantly vary as substrate specificity, exhibiting different bioactivity. In a previous study, the inactivation of the genes involved in the synthesis of anabaenopeptilides in *Anabaena* sp. 90 resulted in a considerable increase in the production of anabaenopeptins [88]. One plausible explanation for this phenomenon is that the knockout of genes involved in the biosynthesis of some oligopeptides can lead to a credit of energy, which is allocated to the production of the remaining peptides [90].

The impact of the cell density on anabaenopeptin production and other oligopeptides have been investigated since it has been related to the increase in the production of some antibiotics. Moreover, this phenomenon can provide valuable data about probable alterations during bloom processes and mat formation [97,98]. High-density cultivation of photosynthetic microorganisms can be a challenge as the light availability decreases with cell density. Guljamow and co-workers [21] utilized a two-tier vessel developed by Bähr and co-workers [99] to cultivate *N. punctiforme* PCC 73102 and *Nostoc* sp. strain KVJ2 at high cell density and observed a higher diversity of secondary metabolites. Anabaenopeptin was absent in *Nostoc* sp. strain KVJ2 biomass obtained under conventional cultivation. High-density cultivation of this strain revealed the presence of three novel anabaenopeptins (KVJ827, KVJ841, and KVJ811). The increase of the content of these APs in the strain KVJ2 is attributed to a higher number of transcriptions among the cells. In the conventional cultivation, the distribution of the *aptA* transcripts (an NRPS gene related to AP production) was restricted only to a cell at (pre-)akinete state while in high-cell density culture, this transcript was widely distributed among the vegetative cells [21].

The interaction between different chemotypes of cyanobacteria in a water body can provoke significant alterations in their secondary metabolites profile. Consequently, differences are observed between laboratory culture and natural environments. In co-culture with *M. aeruginosa* PCC 7806, the non-microcystin-producing strain *M. aeruginosa* PCC 9432 enhanced its bioactive peptide content, including Ferintoic acids A and B [90]. These findings suggest the release of diffusible signals by cyanobacteria with the capacity of regulating the production of APs. The chemical nature of such metabolites was not determined in this study. However, certain oligopeptides can fill the signaling function since they are occasionally found in the extracellular compartment, acting as infochemicals. In addition to peptides, cyanobacterial exudate has also some nutrients, which affect the production of certain toxins and can be, consequently, responsible for the increase of Ferintoic acids A and B in *M. aeruginosa* PCC 9432 [100]. In a later study, the supplementation of the culture medium of a *P. agardhii* with two oligopeptides extracts from samples of *P. agardhii* as the predominant cyanobacterial species had different effects on the synthesis of the peptides of this strain. Both extracts showed a positive impact on biomass accumulation and chlorophyll-a production, being attributed to those nutrients and oligopeptides now present. The high nutritional content of the extracts is associated with the ability of cyanobacterial in fixing nitrogen and producing vitamins, phytohormones, and polysaccharides. Three out of four anabaenopeptins maintained constant (m/z 851, 844, and 837) while the variant with m/z 828 was substituted by other with m/z 923. One of the extracts increased the anabaenopeptin content of variants m/z 844, 851, and 837 while the other diminished the quantity of these last two [101]. The opposite responses to these extracts may be assigned to the content differences observed between them. The extract responsible for reducing the APs expression exhibits a superior concentration of nitrate and phosphate, which, as was previously mentioned has a negative effect on the production of anabaenopeptin [27].

In addition to interaction with other cyanobacteria, these microorganisms are capable to establish symbiotic associations with invertebrates, such as corals, mollusks, and sponges. Both organisms can be benefited during this consortium through secondary metabolite production, for example [102]. Sponges host an enormous quantity of microorganisms belonging to diverse phyla, where cyanobacteria are mainly represented by genera *Aphanocapsa*, *Synechocystis*, *Phormidium*, and *Oscillatoria* [103]. These photosynthetic microorganisms can occupy either extra- or intracellular spaces, aiding the host in the control of the redox potential, supplying pigments and energy through carbon fixation, and in the defense mechanism by the production of secondary metabolites. Published reports have demonstrated that as a consequence of these processes, cyanobacteria have their metabolic profile altered, resulting in the production of distinct variants of natural products. The compound 2-(2',4'-dibromophenyl)-4,6-dibromophenol is solely biosynthesized by a cyanobacterium belonging to genus *Oscillatoria* in association with the sponge

*Dysidea herbacea* [104]. These factors corroborate with the hypothesis that anabaenopeptins primarily observed in sponges could be of cyanobacterial origin, as brominated APs variants were isolated only from sponges [28,31,33] and the *Oscillatoria* genus is known for APs production. For instance, the polyketide nosperin and some variants of oligopeptide nostopeptolide are encountered exclusively during symbiosis, which may be the same mechanism for anabaenopeptin variants production found in sponges.

### **4. Biosynthesis**

The features of Anabaenopeptins are related to Non-Ribosomal Peptide Synthetases (NRPSs), which operate with a nucleic acid-free mechanism at the protein level and are structured as multifunctional proteins. NRPSs are organized as gene clusters in bacteria, usually possessing all the proteins required for proper biosynthesis of the secondary metabolites, from the generation of building blocks to product transport [105–107].

The variability of NRP structures, both cyclic and linear, reflects the concept of the complex modular system of NRPSs organized as an assembly line. Each module is responsible for the activation and coupling of an amino acid to the respective oligopeptide being synthesized. The principle known as the collinearity rule dictates that, for example, a hexapeptide requires six modules to be produced. Those modules are composed of enzymatic domains present in an NRPS, which are responsible for specific biosynthetic steps, as amino acid activation, bond formation, and oligopeptide liberation. Besides the initiation module, an elongation module from an NRPS requires, at least, an Adenylation-domain (A-domain) for amino acid recognition and activation; the Thiolation-domain (T-domain), required to carry the synthesized peptide; and a Condensation-domain (C-domain), responsible for the peptide bond formation. The last module of this assembly line requires the Thioesterase-domain (Te-domain) for the proper maturation of the peptide, also responsible for the cyclization step [18,105–108].

Similar to other peptides produced by NRPS, the biosynthesis of APs requires all the specific steps of the assembly line. Besides, due to some specific characteristics present in this cyclic hexapeptide and its variants, other proteins and domains can also be related to its synthesis, as the biosynthetic apparatus for homoamino acid production and domains for D-Lys formation (Epimerization-domain; E-domain) and N-methylation of specific residues (Methylation-domain; M-domain) [18,19,105,106,108,109].

Besides the fact that the anabaenopeptin structure's first detection in cyanobacteria occurred in 1995 [20], its gene cluster was only described ten years later in a *Planktothrix rubescens* strain [18]. The gene cluster detected in this cyanobacterium comprised of 5 genes (*ana*ABCDE): 4 NRPSs, and an ATP-Binding Cassette-transporter (ABC-transporter) protein. It was also visualized NRPSs possessing an epimerase domain (AnaA) and a methyltransferase domain (AnaC), which could be related to typical features encountered in APs, such as D-Lys and N-methylated amino acids, respectively. Only one cluster was detected in this organism, and it was attributed for the biosynthesis of all four peptides produced: Anabaenopeptin A, B, F, and Oscillamide Y, which differ by the combinatory of two residues in two distinct positions: (Tyr/Arg)-Lys-(Val/Ile)-Hty-MeAla-Phe. Thus, this phenomenon indicates that these NRPSs demonstrated a certain degree of promiscuity regarding their substrates and A-domains, as different amino acids can interact with the same catalytic site [18].

Rouhiainen and co-workers [110] detected gene clusters related to the production of APs in *Anabaena* sp. 90, *Nodularia spumigena* CCY9414, and *Nostoc punctiforme* PCC3102. In *Anabaena* sp. 90, five Open Reading Frame (ORF) were identified to be encoding NRPSs (*aptA1, aptA2, aptB, aptC,* and *aptD)* and two additional genes to be encoding proteins with similarity to HMGL-family (*aptE*) and ABC-transporter protein (*aptF*). When compared to the clusters identified in *N. spumigena* and *N. punctiforme,* 4 NRPS and two homolog proteins to AptE and -F were also detected, indicating that *Anabaena* sp. had an additional NRPS gene (*aptA1* and *aptA2*). Similar to AnaA from *Planktothrix rubescens* NIVA-CYA 98, AptA1 and AptA2 also have an epimerase domain indicating their role as

initial enzymes, and AptC possessing the N-methyltransferase domain as AnaC [110]. The proteins AptA1/AptA2, AptB, AptC, and AptD are homologs to the NRPS proteins AnaA, AnaB, AnaC, and AnaD, sharing the same functions, respectively.

A genomic analysis of *Sphaerospermopsis torques-reginae* ITEP-024 accomplished by Lima and colleagues [107] demonstrated that the *apt* gene cluster is close to the spumigin cluster. Both AP and spumigin are peptides with protease inhibitory activity which usually possess Homophenylalanine and Homotyrosine residues, then indicating that both NRPS apparatus share a biosynthetic cluster related to the production of these nonproteinogenic residues. The *apt* gene cluster of *S. torques-reginae* strain has a similar organization to the anabaenopeptin clusters from *Anabaena, Nodularia, Nostoc,* and *Plaktothrix* [18,110]. Thus, its cluster also holds four genes encoding a six-module NRPS (*apt*ABCD), where the Te-domain is present at the last module, then being responsible for the final step of AP production, similarly to other NRPS products [107].

Entfellner and co-workers [57] suggested that the AP cluster could be transferred among cyanobacterial species due to horizontal gene transfer (HGT). This hypothesis is supported by the high similarity visualized between the *apnA-E* cluster from *Planktothrix* and *Microcystis* composed by *apnA*, *apnB, apnC, apnD* and *apnE*, which genes codified proteins homologs to AnaA/AptA, AnaB/AptB, AnaC/AptC, AnaD/AptD, and AnaE/AptF, respectively. Some strains belonging to the *Planktothrix* genus demonstrated to possess the same AP cluster, but not all of them, thus suggesting that the common ancestors of these organisms did not have the NRPS apparatus for AP biosynthesis, which could be visualized by a phylogenetic analysis using *apnA-E* clusters as biological markers. By phylogenetic analysis of different sequences of anabaenopeptin cluster, it could be inferred that an ancestral cluster was introduced into the chromosome of a *Planktothrix* strain and diversified into different variants, which could be grouped according to *apnA* sequences. Thus, the high frequency of AP producers belonging to *Microcystis* and *Planktothrix* in nature could be an indication of this mechanism of genetic transference by the AP cluster and its wide distribution among those genera, requiring further analysis of the same mechanism in other AP producers, such as *Anabaena, Aphanizomenon, Nodularia, Nostoc, Oscillatoria,* and *Lyngbya* (Table 2) [57].

It had been detected in *Nostoc* sp. CENA543 six variants of APs. Through genomic analysis, a gene cluster of 26 kb containing four NRPS and additional enzymes was visualized related to AP production. Following the same pattern, the NRPS proteins were AptABCD, and the additional enzymes were an ABC-transporter, 2-isopropylmalate synthase (HphA), and an ORF similar to Nuclear Transport Factor-2 (NFT2) proteins [56].

Thus, as discussed, several AP clusters have been identified (Figure 10) and their nomenclatures are not standardized, which usually are assigned according to the strains detected. For example: *ana* and *apn* for *Planktothrix* [18,57,111,112]; *apt* for *Anabaena, Microcystis, Nodularia, Nostoc* and *Sphaerospermosis* [56,107,110]; and even *kon* from *Candidatus Entotheonella* sp. TSY referencing the konbamide biosynthetic gene cluster [113]. Among these nomenclatures, *apt* is the most recurring, being applied to refer the AP gene cluster along this manuscript. However, all anabaenopeptin gene clusters from these different strains of cyanobacteria share common features. The first NRPS, AptA, is a bimodular initiation enzyme containing two A-domains, two T-domains, one C-domain, and one Edomain. The second NRPS enzyme, AptB, contains one elongation module (condensation, adenylation, and thiolation domains), followed by the third enzyme, AptC, which is an NRPS enzyme with two elongation modules, which commonly contains distinct domains related to peptide modification, such as N-methyltransferases. Finally, the termination module from AptD comprises an elongation module which also includes a Te-domain (Figures 10 and 11). Then, it totalizes 6 modules, following the collinearity rule and confirmed by bioinformatic analyses regarding the specificity of each module with its amino acid [18,56,57,107,110–112].

**Figure 10.** Anabaenopeptin cluster (*apt*) organization from different cyanobacteria strains. The genes *aptA1*, *aptA2*, *aptA*, *aptB*, *aptC*, *aptD* and *aptE* are Non-Ribosomal Peptide Synthetases (NRPSs) related to Anabaenopeptin production; *hphA* gene belongs to homoamino acid biosynthetic pathway and *hph*ABCD cluster. These clusters were obtained according to their accession codes (AC) from National Center for Biotechnology Information (NCBI): *Anabaena* sp. 90 (AC: GU174493), *Nodularia spumigena* CCY9414 (AC: CP007203), *Nostoc punctiforme* PCC 73102 (AC: CP001037), *Sphaerospermopsis torquesreginae*ITEP-024 (AC: KX788858), *Planktothrix agardhii* NIVA-CYA 126/8 (AC: EF672686), *Snowella* sp. 249/25 (AC: MF741700), *Microcystis aeruginosa* PCC 9701 (AC: HE974200), *Microcystis aeruginosa* SPC777 (AC: PRJNA205171), *Microcystis aeruginosa* PCC 9432 (AC: HE972547). This information is available on the public database NCBI (https://www.ncbi.nlm.nih.gov/; accessed on 16 March 2021).

> The first adenylation domain from the NRPS apparatus belonging to the first module of AptA (Figures 10 and 11) had been analyzed by several works due to the inhibitory role of the first amino acid residue towards specific enzymes [57,111]. Evolutionary analysis coupled to molecular biology demonstrated that one of the first anabaenopeptin to be produced possessed Arg at position 1, such as anabaenopeptin B (Figure 2). This data corroborates with the inhibitory activity of carboxypeptidase B of AP variants bearing Arg at the exocyclic position, which is greater than Tyr, Phe, and Ile. In addition, analysis of *Planktothrix* producers strains demonstrated a high frequency of AP B producers (83 out of 89 strains), followed by AP A, AP F, and Oscillamide Y (55%, 45%, and 33% of the strains), corroborating with Table 2 [57]. Some wild-type adenylation domains from the first module of AptA demonstrated to be highly specific for arginine and tyrosine, and single point mutations within this domain can result in significant substrate promiscuity [57,110–112]. Due to its high frequency, inhibition towards carboxypeptidase B, and the possibility to be the first oligopeptide of its group to be originated, the biosynthesis of Anabaenopeptin B is outlined in Figure 11 and will be used as a standard for APs production.

> Through a search of APs biosynthetic clusters in several cyanobacteria, Shishido and colleagues [56] detected that the majority of strains of cyanobacteria contained only one *apt*A gene. However, ten cyanobacteria and the tectomicrobia *Candidatus Entotheonella* sp. TSY1 possessed two alternative *apt*A genes. Thus, under other works [18,56,57,107,110,111, 113], the biosynthesis initiation of APs has two different approaches. The first one is the NRPS with the presence of two starter modules with distinct substrate specificities that can produce different variants of APs. The second mechanism is due to the promiscuity of the first adenylation domain of AptA, producing different variants at position one [112]. Both mechanisms can increase the chemical diversity of Anabaenopeptins produced.

**Figure 11.** Scheme of biosynthesis of anabaenopeptin B in *Anabaena* sp. 90 by NRPS apparatus [107,110]. A: adenylation domain; T: thiolation domain; C: condensation domain; E: epimerization domain; M: N-methylation domain; Te: thioesterase domain.

As discussed previously, Rouhiainen and co-workers [110] identified an anabaenopeptin cluster from *Anabaena* sp. 90, possessing one additional NRPS enzyme with two modules (AptA1 and/or AptA2). This cyanobacterium was able to produce 3 different AP variants differing at position one. Through sequence comparison and substrate specificity analysis, it had been demonstrated that the first adenylation domain of AptA1 had an affinity to L-Lys and L-Arginine (Arg), while AptA2 demonstrated to interact with L-Tyr. Both adenylation domains from the second module of AptA1 and AptA2 incorporated D-Lys. Thus, demonstrating that *Anabaena* sp. 90 carried two distinct initiations NRPS producing different variants of anabaenopeptin, which a similar mechanism could also be visualized for puwainaphycins and minutissamides [110,114]. However, in Figure 11, only AptA1 is represented due to its specificity towards Arg.

Regarding the promiscuity of the adenylation domains aiming to understand the production of distinct AP variants, the adenylation domain of AptA from *Plaktothrix agardhii* PCC 7821 had been evaluated and concluded that it demonstrated to be bispecific for two different amino acids: Arg and Tyr. This feature corroborates with the variants produced by this strain of *P*. *agardhii*: Anabaenopeptins 908A and 915, which differs solely in the exocyclic residue (Arg or Tyr) [111,112]. A similar pattern had been visualized in *Planktothrix rubescens* NIVA-CYA 98, which possesses only one AP cluster, but it was able to biosynthesize different variants of anabaenopeptin differing at the exocyclic position (Tyr and Arg) and the third position (Val and Ile) [18].

One important feature encountered only in Anabaenopeptin among cyanobacterial peptides is the ureido linkage between the first and second residues [34,49]. However, this linkage can also be found in other natural products, including pacidomycins, mureidomycins, napsmycins, and syringolin A. This configuration is not common due to the mechanism present in NRPSs, which assembles amide bonds in an approach where the chain polarity remains unidirectional. The presence of ureido linkage alters this polarity due to the presence of N-to-N terminal condensation. Then, a specific enzyme and/or domain must be present in NRPSs involved at the ureido linkage formation, suggesting a possible role of the first elongation module in their formation [115].

When comparing the initial NRPSs genes encoding both modules of AptA from *Anabaena* sp. 90, *Nodularia spumigena* CCY9414, and *Nostoc punctiforme* PCC 73102, they all contained typical adenylation and condensation domains, also demonstrating highly conserved motifs. Besides their conservation, one hypothesis was that both modules of initiation and elongation of AptA would be related to ureido bond formation, similar to the SylC protein, from *Pseudomonas syringae*, which role is the catalysis of the ureido linkage between two Val residues from syringolin A [110,115]. Though the SylC protein possesses a domain with structural similarity to acetyltransferase between the A- and C-domains from the NRPSs, which is responsible for the ureido linkage formation and no homologous is present in the anabaenopeptins synthetases, suggesting a different mechanism for this step during AP biosynthesis [107,115].

Besides the initiation step and the formation of the ureido bond between the first residue and the conserved Lys, several steps of elongation of the peptide are required to produce a fully mature peptide. The signature sequences analyzed of the A-domains of these NRPS enzymes, such as AptB and AptC, are consistent with the respective amino acid residue of the final product and confirmed in vitro by biochemical methods. Also, usually, the fifth module bears an *N*-methyltransferase domain, as seen in AptC and their homologs, responsible for the *N*-methyl in Ala in position 5 of Anabaenopeptin B, as seen in Figure 11 [110].

Unlike the initiation enzyme related to residues at position 1 and 2, clusters related to AP production has not been shown to possess more than one NRPS for each residue. Thus, the variants produced by the cyanobacterial differing at positions 3–6 are biosynthesized due to the promiscuity of the adenylation domains of AptBCD. This phenomenon can be visualized by innumerous AP variants differing at those positions with only one correspondent gene cluster in the genome, for example, *Nostoc* sp. CENA 543 producing six variants [56].

Anabaenopeptins usually have homoamino acids at positions 4 and 5, which are added by AptC during elongation steps, as visualized in Figure 11 by the additional Hty added in position 4. The AptE, now known as HphA, was first suggested to be responsible for ureido linkage formation and is related to homoamino acid synthesis [110]. Succeeding previous works, it has been elucidated that AptE belongs to a biosynthetic cluster named *hphABCD*. Genes from *hph* cluster are frequently detected in the same genomic region as *apt* and *spu* clusters, which both products, Anabaenopeptins and Spumigins, are peptides displaying protease inhibitory activity and homoamino acids. A genomic analysis of *Sphaerospermopsis torques-reginae* ITEP-024 demonstrated that both Spumigin and Anabaenopeptin clusters were present in proximity in the genome. In between both clusters, the *hphABCD* biosynthetic cluster and additional genes were detected in this region, which a similar organization was also visualized in *Nodularia spumigena* CCY9414 [107]. The *hph* genes are responsible for the biosynthesis of Hph and Hty, nonproteinogenic amino acids commonly found in both anabaenopeptin and spumigin [116]. Thus, indicating that HphA is not responsible for ureido linkage formation but behind the supply of both Hph and Hty. In addition, the presence of the homophenylalanine and homotyrosine biosynthetic enzymes in this region could suggest that this cluster is supplying both homoamino acids for APs and Spumigins [107]. In accordance with Lima and co-workers [107], Shishido and colleagues [56] also visualized that from 56 genomes analyzed containing the *apt* cluster all demonstrated to possess the *hph* biosynthetic cluster, except for *Scytonema hofmanii* PCC 7110 and *Candidatus Entotheonella* sp. TSY. The genes encoding the proteins HphABCD were frequently found upstream or downstream of the AP cluster, supporting the hypothesis about their roles in providing homoamino acids to APs [107].

Thus, homoamino acids are produced by the HphABCD enzymes and then incorporated by the NRPS apparatus. In addition, these non-proteinogenic amino acids can also be further modified by the NRPS enzymes, considering that residues at position 5 are mostly methylated by the N-methylation domain in the second module of AptC. However, methylation of residues at position 4 was also visualized, such in Ferintoic acids A and B [39], Anabaenopeptin E [37], 863, 891, 848, and 882 [24].

The final step for Anabaenopeptin production is mediated by a Te-domain, which is commonly associated with the termination process of the biosynthesis of NRPS peptides. Thus, after the incorporation of the last residue, for example, L-Phenylalanine in AP B (Figure 11), these domains can be involved with the release of the peptide by hydrolysis, or even cyclization involving peptidic or ester bonds [19,106]. The last NRPS enzymes AptD and its homologs [18,111] bear the thioesterase domain, suggesting then their role as the termination step.

Besides those typical alterations to the amino acid residues discussed, several variants of APs have been found with different modifications, such as ethylated (Figure 2, Figure 3, and Figure 5), acetylated, and oxidized residues [22,24,34]. In addition to such modifications during the elongation steps by the NRPS, an analysis of cytochrome P450 monooxygenases from cyanobacteria revealed that some proteins of this class may be related to anabaenopeptin modifications. In *Synechococcus* sp. PCC 7502, it had been suggested that a P450 belonging to CYP110 is involved in the production of Anabaenopeptin NZ857. *Anabaena* sp. TAU NZ-3-1 was capable to coproduce this anabaenopeptin and APs NZ 825 and NZ841. Anabaenopeptin NZ857 differs from AP NZ825 and AP NZ841 by the number of oxidized residues at positions 4 and 6. Anabaenopeptin NZ857 has in both positions 4 and 6 the homotyrosine residue, while the other peptides have at least one homophenylalanine. Besides the possible relation of cytochrome P450 in anabaenopeptin production, its possible catalytic role has not been demonstrated [117].

Regarding the unusual anabaenopeptins lacking residues in their structure, the biosynthesis of Anabaenopeptin 679 (Figure 6) has not been described so far [53], requiring further analysis of its production. Due to Namalide similarity to APs, it has been suggested that the biosynthesis of this tetrapeptide is realized by the *apt* cluster, as during a genomic screening of both namalides-producing cyanobacteria no exclusive cluster related to the production of these peptides have been found. The prediction of amino acids incorporation of adenylation domains of AptABCD is in accordance with both AP and Namalides. Thus, the preliminary results obtained by Shishido and co-workers [56] strongly suggested that Namalides are biosynthesized by *apt* cluster through a module skipping event. During synthesis, the second domain of AptC and the C-domain of AptD (but not the thioesterase domain) are ignored resulting in the production of namalides, similar to the module-skipping process of Myxochromide from myxobacteria [56].

### **5. Ecology**

Cyanopeptides confer a competitive advantage for their producing organisms due to their toxicity, which effect has been examined against parasites and grazers (Figure 12) [118,119]. Other strategies, such as colony formation and filaments aggregation with low nutrition content have also been documented as a defensive mechanism [120]. However, they cannot, on some occasions, be sufficient to explain the different susceptibility levels encountered among cyanobacterial populations [121].

**Figure 12.** Ecological relevance of anabaenopeptins.

Anabaenopeptin presence in the cyanobacterial extract can confer a certain level of protection against some predators but is not a determining factor in the process as illustrated by the work developed by Urrutia-Cordero and coworkers [122]. These authors attested anti-amoeba activity against *Acanthamoeba castellanii* by *Microcystis* strains capable of producing either APs or MCs. Among the tested strains, the anabaenopeptin-producing was the one that caused the highest mortality rate. In contrast, the existence of the same APs in the extract of *A. lemmermannii* NIVA-CYA 426 did not result in any type of activity for the protozoan. Due to APs and MCs inhibitory activities against phosphatase, the loss of cytoskeleton integrity of *A. castellanii* was associated with the action of these cyanopeptides, which led to impairment of crucial functions associated with cytoplasmic projections, including motility and feeding.

Deleterious effects in organisms belonging to aquatic fauna were also linked to APs production and other cyanopeptides [123–125]. The negative impact of these metabolites can partially justify the substitution of large-bodied zooplankton by small-bodied species during the blooming process since they affect differently these living beings [126,127]. The absorption of such molecules can occur by ingestion of cyanobacteria or through uptake of water. Like the filtration system of large-bodied zooplankton has a greater tendency to absorb these microorganisms, they are more susceptible to the effect of toxins [128].

Some published reports have focused solely on the effect on a determined organism by an individual oligopeptide, especially MCs [92,129]. However, this type of investigation is

not sufficient to verify the real impact of cyanobacterial bloom in the environment. Studies that employ only one type of solvent to obtain the cyanobacterial extract can also provide limited information since these microorganisms harbor an enormous variety of metabolites with distinct polarity, which are not, therefore, totally isolated and investigated during this type of analyses. Even though APs concentration in the aquatic environment can exceed 1 g.L−1, they pose unknown consequences for human health [15]. Furthermore, their full effects on other animals are largely unknown. In Zebrafish (*Danio rerio*), an animal model very close to the human being, APs B and F as well as Oscillamide Y do not have any significant effect on the mortality of their embryos [130]. Otherwise, another study demonstrated that the APs A, B, and F exhibited the greatest toxicity as compared to other cyanopeptides, such as microcystin-RR, microginin 690, and cyanopeptolins CYP-1007, CYP-1020, and CYP-1041 to the nematode *Caenorhabditis elegans*. The exposure to these understudied toxins was responsible for diminishing the reproduction potential of this worm, affecting the brood size, the hatching time of eggs and vulvar integrity. Moreover, lifespan was also reduced by nearly 5 days [131].

Concerning APs action in other animals, Pawlik-Skowro ´nska and colleagues [127] demonstrated that extracts containing Anabaenopeptins originated from bloom samples, where the predominant species were *P. agardhii* or *Microcystis* sp., caused different responses in the behavior of planktonic species *Daphnia pulex* and *Brachionus calyciflorus*. The unicellular cyanobacterium extract did not cause acute toxicity to any of the investigated zooplanktons. Otherwise, the *Planktothrix* extract strongly reduced the survivorship rate of *D. pulex*. This difference was attributed to the oligopeptide profile in the extract, which considerably varies as to their quantity and structure. Anabaenopeptin and Aeruginoside had a superior contribution in *P. agardhii* extract as compared to *Microcystis* one, suggesting that those oligopeptides act in synergism. A similar analysis harboring a larger number of organisms, verified the toxicity of *Nodularia spumigena* extracts against the crustaceans *Thamnocephalus platyurus* and *Artemia franciscana* and also the bioaccumulation of their oligopeptides in these invertebrates and some blue mussels [118]. Nine APs and nodularin were encountered in the mussels collected from a bloom formed by this cyanobacterium. In *T. platyurus and A. franciscana*, the exposure to *Nodularia spumigena* extract results in the accumulation of various Anabaenopeptins, one aeruginosin, and one spumigin. The cyclic structure of APs confers them chemical stability preventing their degradation by the mussels tested, as linear peptides were not detected. Moreover, it also led to an increased mortality rate for both organisms. Among the fractions obtained from *N. spumigena* biomass extract, that with APs and a demethylated form of nodularin exerted the highest acute toxicity effect.

Anabaenopeptins also participate in the defensive mechanism of *Planktothrix* allowing, therefore, its dominance towards pathogens in the same environment. A comparison between a wild-type strain of *P. agardhii* NIVA CYA 126/8 and their mutants with dysfunctions in the production of APs, microviridins or MCs, indicated that these oligopeptides reduce the virulence of the fungi belonged to the division of Chytridiomycota, known as chytrids. This contrast may partially explain the dominance of a cyanobacterial chemotype in a determined environment, serving as a great defensive strategy to retard the parasite adaptation and to increase its diversity [16].

Anabaenopeptins and microviridins are most likely involved with the inhibition of the protease released by rhizoids whereas MCs are most probably related to the inactivation of their phosphatases [16]. Both groups of enzymes occupy a significant position in cell metabolism, participating in regulatory processes and signaling [132]. Strategies utilized by chytrids during infection are still an open question, but it is known that they can infect akinetes, vegetative cells, and/or heterocyst. Resource available inside the host can be one of the reasons for variations encountered among the infectivity methods. Akinetes offer higher energy and organic material content than vegetative cells. Oligopeptides distribution among the vegetative cells and those that are differentiated, such as akinetes and heterocysts

should also be investigated since these secondary metabolites can sometimes be restricted to a cellular subgroup [133].

Cyanopeptides could also act as carbon and nitrogen source for various heterotrophic bacteria. These microorganisms are capable of degrading an enormous quantity of molecules with variable structures [134,135]. Besides the oligopeptides, the phycosphere is rich in carbohydrates, proteins, and lipids that originated from the exudate of cyanobacteria or its cell lysis. The mineralization of these organic compounds leads to CO2 production, which may contribute to the growth of cyanobacteria [136]. Briand and colleagues [137] observed that the supplementation of an axenic culture of *M. aeruginosa* PCC7806 with bacteria associated with the mucilage of *M. aeruginosa* colonies collected during a bloom eliminated all oligopeptide encountered in extracellular fraction, including MC, cyanopeptolin, and cyanobactin. In a previous investigation, Kato and co-workers [138] identified in the cell extract of the bacterium *Sphingomonas* sp. B-9 hydrolytic activity for AP A, microcyclamide, nostophycin, aeruginopeptin 95-A, and microviridin I. Anabaenopeptin A degradation was gradual and subproducts were not observed.

APs and planktopeptin BL1125, both isolated from the bloom-forming *Planktothrix rubescens*, were associated with the collapse of cyanobacterial populations during the bloom termination. These oligopeptides act as triggers, inducing cellular lysis by virus-like particles, most likely cyanophages. Such propriety can explain in part the dominance and the high invasive potential of this species [139]. Sedmak and colleagues [67] attested it when verified that the cell growth of a non-xenic culture of *M. aeruginosa* MA2-NIB was inhibited when treated with these oligopeptides whereas the axenic were not affected. Such activity was not attributed to the known property of these peptides since the serine protease inhibitor 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride failed in inducing any type of effect in the growth of these strains. Enrichment of the medium with the bacteria isolated from the non-axenic culture also did not produce any type of alteration. In contrast, the addition of the planktopeptin in an axenic culture of *Microcystis* previously supplemented with particulate materials obtained from cell lysate of the non-axenic culture provokes cell disintegration.

According to the hypothesis raised by these authors [67], the release of the cyclic peptides, mediated by cell lysis, signalizes the presence of a determined host and consequently activates the lytic cycle. A small concentration of these oligopeptides in the environment causes limited lysis, confined solely to a specific region. Cyanobacterial blooms offer the ideal condition for collapse since their cells are exposed to an elevated quantity of infection agents and cyclic peptides [99]. Cell-lysis provoked by cyanophages can promote the release of oligopeptide to the extracellular matrix, feeding positively the cycle [140]. In a subsequent study, algaecide property was reported for Anabaenopeptin KVJ811, which was capable of jeopardizing the growth of the strain *Nostoc* sp. KVJ11, pointing out the importance of these oligopeptides in populational control [21].

Given the above importance of these agents in the environment, the ecological functions of APs in the aquatic environment are much more extensive than we have already known. They are remarkably diverse in structural and functional terms. Novel Anabaenopeptins have been constantly isolated and identified. From an ecological and evolutionary perspective, these cyclopeptides allow communication with different organisms and are decisive elements in natural selection.

### **6. Applications of Anabaenopeptins**

Cyanopeptides such as APs have a well-demonstrated capacity of protease inhibition [141]. Protein Phosphatase 1 (PP1), Protein Phosphatase 2A, Carboxypeptidase-A (CPA), Human Serine Protease, Leucine Aminopeptidase, Trypsin, and Thrombin have already been tested against several cyanobacterial extracts and confirmed the catalysis blockage [11].

Among the enzymes listed above, APs were more effective inhibitors against CPA, PP1, and elastase. In a cyanobacteria bloom, it was isolated eight different Anabaenopeptins

which showed activity towards CPA and protein phosphatase 1 [34,142]. PP1 inhibition may be influencing the HIV-1 transcription, cancer, or cardiac hypertrophy, for example [143,144]. Some APs half-maximal inhibitory concentration (IC50) values are presented in Table 3.

**Table 3.** Detected IC50 values of some Anabaenopeptins in nM. (TAFIa: Thrombin Activatable Fibrinolysis Inhibitor; PP1: Protein Phosphatase 1).


Serine/threonine protein phosphatases inhibition was also reported [22,25]. Nevertheless, several other cyanopeptides presented more effective IC50 levels against elastase, such as some variants of lyngbyastatins, symplostatins, microvirins, and others. Concerning PP1, MCs remain the best inhibitor among all cyanopeptides [11]. IC50 reported values to MCs and nodularins are from 1.1 to 1.9 nM as PP1 inhibitors [147]. In this case, APs remain promising candidates in Carboxypeptidase inhibition.

Cyanopeptides blooms events may present the production of different classes of cyanopeptides like MCs, APs, and cyanopeptolins. A few studies quantified cyanopeptides beyond Microcystins, even so, in 10 eutrophic lakes in the United States and Europe the cyanopeptides concentration including these 3 types of cyanopeptides were from <4 μg/L to >40 μg/L [11]. In wet weight, 2.1 mg of AP and 7.4 mg of Microcystin-LR were obtained from 1.7 kg of biomass in a water bloom of lake Teganuma (Japan) [41].

In a study conducted by Spoof and coworkers [34], the production range of the APs measured in extracts from cyanobacteria sampled by plankton net was from 1.7 to 181.9 μg/mL in 22 isolated Anabaenopeptins. Bioactivity assays identified IC50 values from 16 to 435 ng/mL (Nodulapeptin 933 and Anabaenopeptin 813, respectively) against PP1 and from below 3 to 45 μg/mL against CPA (Anabaenopeptins A, D and Nodulapeptin 883C and 917: <3 μg/mL; Nodulapeptin 867: 45 μg/mL). The inhibition of elastase, trypsin, or thrombin does not occur independently of the exocyclic residue (Phe, Ile, and Tyr). The residues adjacent to the ureido bond have a major influence on CPA inhibition. Therefore, APs with Ile and Tyr in the exocyclic position presented the best IC50 values

against this enzyme. Thus, hydrophobic aromatic or linear sidechain next to the ureido moiety presents more favorable interactions with CPA while positive amino acids such as Arg are unfavorable. It explains why Anabaenopeptin B presents IC50: >20 μg/mL and Anabaenopeptin 679 (different only in this position) had an improved inhibitory activity IC50: 4.6 μg/mL [53].

Anabaenopeptins B and F presented activity against human leukocyte elastase (HLE) and porcine pancreatic elastase (PE). Ki values of HLE inhibition were in the 0.1–1 μM range in a linear competitive model [148]. In another study, APs A and B were capable of relaxing rat aortic preparations in a concentration-dependent form using 10–400 μg/mL [20].

Some studies have been explored APs bioactive properties in a pharmaceutical/ biotechnological way. Despite APs ability to inhibit diverse proteases, other cyanopeptides present the best IC50 values than them in most cases. However, one application shows more promising results using APs: the inhibition of the Thrombin Activatable Fibrinolysis Inhibitor (TAFIa), which is a proteolytic enzyme that cleaves Arg and Lys residues on fibrin and may be a novel antithrombotic mechanism [149]. Anabaenopeptins B, C, and F, isolated from *Planktothrix rubescens*, presented high promising results inhibiting TAFIa selectively over other coagulation enzymes as Carboxypeptidases A, B and N, FXa, FVIIa, FIIa, and FXIa [12,145]. In this sense, Anabaenopeptin B showed the best values of IC50 (1.5 nM, in different studies, similar to PP1 inhibition by microcystins) on a screening performed with 20 APs isolated from *Nostoc* and *Planktothrix* strains. It was elucidated that Lys and Arg residues in the R1 position (considering Anabaenopeptin B as reference: Arg-Lys-Val-Hty-MeAla-Phe) are associated with high activity (IC50 values of 2.1 and 1.5 nM, respectively) since those structures presenting Tyr residue in this position showed a significant decline of activity by two orders of magnitude (IC50 of 400 nM). In the R3 position, it was observed that the substitution of Ile by Val does not affect activity. R5 position also presents a loss of potency when the residue Ala is replaced by Ser. It is also observed a high tolerance towards substitutions in the pentacyclic region [12]. The Val to Ile difference among APs B and F does not implicate in gain or loss of activity against TAFIa. The mechanism involved in TAFIa inhibition depends on the linear part of APs mimics the carboxy−terminus of fibrin which is able of penetrating the active site pocket. Hence, the circular fraction of APs blocks the channel's entrance, preventing the interaction with other molecules [12]. To compare the interactions of the Anabaenopeptin B-TAFIa complex with Microcystin LR-PP1 complex, TAFIa structure was obtained from Protein Data Bank (PDB), it was resolved by X-ray diffraction presenting 2,5 Å resolution (PDB code: 3LMS) (Figure 13). Also, 1.84 Å resolution PP1 in complex with Microcystin-LR was used to represent the binding mechanism (PDB code: 6OBQ). Microcystin MeAsp residue blocks the access to the PP1 active site, the long hydrophobic tail composed of Adda residues plays an important role in this inhibition due to its interaction with the hydrophobic groove region, adjacent to the catalytic site [45]. Yet, different from the linear part of Anabaenopeptin B which accesses a protein channel, Adda residue in Microcystin-LR makes contact with superficial residues of PP1 (Figure 13).

Besides some cyanopeptides presented anticancer activity, APs have been presented poor results in cytotoxic tests [150]. Anabaenopeptin B had been tested about its anticancer potential and did not demonstrate cytotoxic effects against N2a, MCF−7, and GH4 cells even at the 500 μg/mL concentration [151]. Despite anticancer activity was detected in *Aliinostoc* sp. CENA543 extract containing AP, it was not possible to attribute this effect exclusively to this class of oligopeptides because there were other cyanopeptides in the extract, and the exact AP was not identified [152]. No cytotoxic activity was presented by Nodulapeptins 883C, 869, 867, 865, and Anabaenopeptin 813 as well [34].

**Figure 13.** Interaction between (**A**) Anabaenopeptin B (AnaB) with Thrombin Activatable Fibrinolysis Inhibitor (TAFIa) and (**B**) the complex between Microcystin-LR and Protein Phosphatase 1 (PP1).

### **7. Final Considerations**

Anabaenopeptins are structurally diverse molecules widely distributed in distinct ecosystems. Some structural features of these oligopeptides are shared with other cyanotoxins, such as the presence of modified residues, exocyclic amino acids, circular structure, and amino acids in D-configuration. However, among the cyanopeptides, the ureido linkage is exclusively found in APs. Despite their elevated occurrence and structural diversity, the majority of this group of peptides has been isolated from filamentous cyanobacteria, being commonly associated with specific genera. In freshwater environments, APs B and F are the most recurrent whereas the marine strains normally display a higher number of exclusive APs. Terrestrial cyanobacteria possess in common few APs with those from the aquatic environment.

The production of these toxins is influenced by environmental factors, which include nutrient concentration, temperature, light intensity, and association with other organisms. NRPS apparatus mediates the APs biosynthesis, following the collinearity rule of the modules. The low specificity of adenylation domains toward their substrate and the presence of additional modules are responsible for the production of various variants by a single strain. Homoamino acids present in the APs structure are supplied by HphABCD biosynthetic pathway. Some modifications are catalyzed by specific domains encountered in the NRPS modules. However, the mechanism of ureido bond formation is still unknown.

APs have been increasingly detected in reservoirs, lakes, and oceans in very elevated concentrations. Their toxicity for human beings has not yet been determined. However, assays employing animal models as well as other organisms have demonstrated their deleterious actions. In face of this fact, there is a dire need to further investigate the real impact of these oligopeptides on human health. The inhibitory activity of these molecules against proteases, phosphatases, and carboxylases makes them very promising for biotechnological use, but their mechanisms of action need to be investigated in detail to be properly applied.

However, Anabaenopeptins still require further studies to comprehend their behaviors in nature. Among AP producers, it must be evaluated the evolutionary relationship between terrestrial and aquatic strains as they do not share a high number of AP variants, similar to freshwater and marine. Besides, specific residues are more predominantly found in some environments, requiring additional analysis to comprehend the relationship between their frequency and habitats. Moreover, just a few variants have been analyzed

regarding their inhibitory properties, then demanding more tests to discern the role of specific amino acids during the interaction with their targets. Another issue that must be investigated deeply is the APs amount produced by cyanobacteria. Their low yield can be a limited factor for industrial purposes. Such bottlenecks can be minimized through the use of heterologous expression, which has been well established for other cyanopeptides.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/toxins13080522/s1, Table S1: Sequence composition of Anabaenopeptins.

**Author Contributions:** Conceptualization, P.R.M., S.C.d.A., L.P.X., and A.V.S. Investigation, P.R.M. and S.C.d.A. Writing—original draft preparation, P.R.M., S.C.d.A. and A.S.S. Writing—review and editing, P.R.M., S.C.d.A., A.V.S. and L.P.X. Supervision, A.V.S. and L.P.X. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was financed in part by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior—Brasil (CAPES)—Finance Code 001 and Fundação Amazônia Paraense de Amparo a Estudos e Pesquisas (FAPESPA)—03/2019.

**Institutional Review Board Statement:** Not applied.

**Informed Consent Statement:** Not applied.

**Data Availability Statement:** Not applied.

**Acknowledgments:** The authors would like to thank Pró−Reitoria de Pesquisa e Pós−Graduação da Universidade Federal do Pará (PROPESP/UFPA).

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


### *Review*
