Diversity and Toxicity of the Genus Coolia Meunier in Brazil, and Detection of 44-methyl Gambierone in Coolia tropicalis

Coolia is a genus of marine benthic dinoflagellates which is widely distributed in tropical and temperate zones. Toxicity has been reported in selected Coolia species, although the identity of causative compounds is still controversial. In this study, we investigated the taxonomical and toxicological aspects of Coolia species from Brazil. Since light- and electron microscopy-based morphology was not enough to distinguish small-celled species, ITS and LSU D1-D3 phylogenetic analyses were used for species definition. Cultures of Coolia palmyrensis and Coolia santacroce were established from samples collected along the northeastern Brazilian coast, the first record of both species in South Atlantic waters. Cultures of Coolia malayensis and Coolia tropicalis were also established and exhibited acute in vivo toxicity to adults of Artemia salina, while C. palmyrensis and C. santacroce were non-toxic. The presence of 30 yessotoxin analogues, 7 metabolites of Coolia and 44 Gambierdiscus metabolites was screened in 14 strains of Coolia. 44-methyl gambierone (formerly referred to as MTX3) and a new isomer of this compound were detected only in C. tropicalis, using both low- and high-resolution LC-MS/MS. To our knowledge, this is the first report of gambierone analogues in dinoflagellates other than Gambierdiscus; the role of C. tropicalis in ciguatera poisoning thus deserves to be considered in further investigations.


Introduction
The marine benthic dinoflagellates are recognized for producing a multitude of toxic compounds [1]. Studies concerning these microalgae have increased since the 1970s, after their identification as causative agents of ciguatera poisoning (CP), the most common non-bacterial intoxication affecting human

Phylogenetics
Species identification was initially evaluated by phylogenetic analyses based on ITS region (ITS 1, 5.8S rDNA and ITS 2) and partial LSU rDNA (D1-D3 domains). The ITS phylogenetic analyses comprised 53 sequences, including 11 sequences from our monoclonal cultures, three outgroup sequences, and sequences retrieved from GenBank. The final ITS alignment was 336-base pairs long. The best-fit model was a TrN (Tamura-Nei model), with base frequencies of A = 0.27370, C = 0.14067, G = 0.22437, T = 0.36125, assuming a gamma distribution shape (G = 0.850). For LSU D1-D3, the final alignment comprised 65 sequences with 745 base pairs. The best-fit model was also a TrN, with base frequencies of A = 0.29774, C = 0.15203, G = 0.23614, T = 0.31410, assuming gamma distribution shape (G = 0.708).
Phylogenetic analyses were performed with two reconstruction methods: maximum likelihood (ML) and Bayesian inference (BI). Considering that both ML and BI analyses gave the same tree topology and relationships among clades, only the majority-rule consensus tree of the ML analysis is shown herein. Six distinct clades were found in the phylogeny inferred from ITS sequences (C. malayensis, C. monotis, C. santacroce, C. palmyrensis, C. canariensis and C. tropicalis), and seven clades from LSU D1-D3 (same clades as those inferred from ITS plus Coolia guanchica) ( Figures 1 and 2). The sequences from Brazil fit within four (C. malayensis, C. santacroce, C. palmyrensis and C. tropicalis) out of the seven described species containing molecular data.

Morphology and Geographical Distribution
Cells of all Coolia species evaluated in the present study were nearly spherical, with slightly different degrees of anteroposterior compression among species. Living cells contained many golden-brown chloroplasts. The thecal plate formulae followed a Po, 3', 7", 5"', 2"" pattern (according to [1]).

Toxicity
Cells Toxicity was evaluated by feeding adults of Artemia salina with increasing biomass of Coolia spp. and proportionally decreasing biomass of a nontoxic algal species (Tetraselmis sp.) In control treatments, all A. salina individuals fed only with Tetraselmis sp. survived after 96 h of experiment, with no signs of impaired swimming or any other visual alteration.
Among individuals fed with increasing biomass of C. palmyrensis (strain LM112) or C. santacroce (LM113), no significant toxic effect was recorded (Figures 8 and 9). On the contrary, exposure to C. malayensis (LM036) and C. tropicalis (LM141) was lethal to A. salina, with mortality rates directly related to biomass and exposure time ( Figure 8C,D). C. malayensis killed up to 57% of A. salina individuals after 96 h of experiment, with no differences in mortality rates among those exposed to biomasses of toxic cells equivalent to 1500 ng C ml −1 or higher ( Figure 8C). C. tropicalis, in turn, caused significantly increasing mortality rates with increasing biomass of toxic cells and exposure time, killing > 90% of the individuals after 72-96 h of exposure to the highest biomass tested (16,000 ng C ml −1 ) ( Figure 8D). After 24 h, mortality rates caused by C. tropicalis were consistently higher than those provoked by equivalent biomass of C. malayensis at the same exposure time (Figures 8 and 9). Maximum biomass (max) tested was 19,300 ng C mL −1 for C. malayensis, 16,000 ng C mL −1 for C. tropicalis, 15,000 ng C mL −1 for C. santacroce, and 11,500 ng C mL −1 for C. palmyrensis. Half of the maximum biomass was also tested (max/2), as well as three fixed biomasses: 3000, 1500 and 750 ng C mL −1 .  Fourteen strains of Coolia spp. were analyzed in negative (ESI − ) and positive (ESI + ) full scan ionization mode on a high-resolution Q-Tof 6550 instrument. Raw data were processed following the Find by Formula (FbF) algorithm in the Agilent MassHunter Qualitative Analysis software, using a Personal Compound Database and Library (PCDL) created by Phycotoxins Laboratory (IFREMER, France). The in-house database used for Coolia spp. screening was composed of 81 compounds, including cooliatoxins, yessotoxins, ciguatoxins, maitotoxins, gambieric acids, gambierones, gambierol and gambieroxide (see Supplementary Material, Table S1). Screening with the PCDL allowed identification of compounds based on their formulae and thus detection of the compound itself or isomers.
In ESI + mode, two compounds were tentatively identified in C. tropicalis (strain LM141) as 44-methyl gambierone (MTX-3) at 6.0 and 6.6 min retention time (RT), with a score of > 90% and a mass error of < 5 ppm ( Figure 10A). The analysis in ESI − mode confirmed the presence of the same two compounds at 6.0 and 6.6 min RT in C. tropicalis, identified as 44-methylgambierone with a score of > 90% and a mass error of < 10 ppm (Supplementary Material, Figure S1). These compounds were additionally investigated in an extract of Gambierdiscus autrales, used here as reference, as no analytical standard is available. The peak at RT = 6.6 min assigned to 44-methylgambierone was also present in G. australes, but no other peak was detected. The identity of peak at RT = 6.0 min in C. tropicalis extract was further determined as a new isomer of 44-methyl gambierone, as detailed below in Section 2.4.2. The assigned positive and negative HRMS ion species for these two compounds are listed in Table 2.
No other known compounds were successfully identified (i.e., with a score > 90% and a mass error < 10 ppm) in any of the other thirteen Coolia strains, neither in positive nor in negative mode. Table 2. HRMS ion species corresponding to the accurate mono-isotopic m/z for 44-methyl gambierone and its isomer. Mass differences (∆ppm) were compared between measured and exact theoretical mass.

Comparative Fragmentation Between 44-Methyl Gambierone and the New 44-Methyl Gambierone Isomer
As the FbF workflow procedure is not sufficient to unequivocally identify a compound, HRMS/MS spectral acquisition in both ESI − and ESI + mode were necessary to confirm that the new compound was in fact an isomer of 44-methyl gambierone.
In ESI − , fragmentation of the molecular anion [M-H] − at m/z 1037.4785 showed that both molecules shared the same product ions (Table 3, Figure S2). The fragments at m/z 899.3741 [C43H63O18S]-and 96.9601 [HOSO3], which exhibited very small mass error (∆ppm < 3) in our analysis, corresponded to the two fragments also described in previous studies [5,35].
The compound with RT at 6.0 min was further confirmed as an isomer of 44-methyl gambierone, by comparing the positive HRMS/MS with that of 44-methyl gambierone itself (Table 3, Figure 10B). Indeed, the fragmentation pathways showed the formation of protonated fragment ions with a small mass error (∆ppm < 5 ppm) at m/z 959.5330, 941.5216, 923.5133, 905.5012 and 887.4945, corresponding to the sulfite loss followed by successive water losses. Furthermore, similar fragmentation patterns were observed from m/z 303 to 95 for both compounds, and the gambierone specific fragment ion was detected at m/z 109.0645, in the isomer as well (∆ppm = −2.7).  Fragments common to both compounds were marked in red in the mass spectrum and reported in Table 3.

Quantification of Gambierone Toxins with LC-LRMS/MS (System B) in C. tropicalis
The extract of Coolia tropicalis (strain LM141) was further analyzed on a low-resolution API4000 Qtrap instrument, in order to confirm the previous results and to quantify the gambierone toxins ( Figure 11). A targeted MRM method was performed in negative ionization mode for the screening of MTXs, gambierones and gambieric acids, as described in Section 5.5.2. Retention times (RTs) were slightly shorter in system B compared to system A, due to differences in dead volumes of the UHPLC systems. The strain of C. tropicalis contained 73 and 20 pg MTX-1 eq. cell −1 of 44-methylgambierone and 44-methylgambierone isomer, respectively. Therefore, under the culture conditions described in Section 5.1, this C. tropicalis strain contained 3.6-fold greater intracellular amounts of 44-methyl gambierone than those of its isomer.

Taxonomy and Phylogeny of Coolia Species
Coolia is usually considered a harmful genus of benthic dinoflagellates [1], even though marked species-specific differences in toxicity are observed, as reported herein and discussed later in Section 3.3. Therefore, accurate taxonomy is desirable before any toxicological discussion concerning Coolia spp. For instance, prior to the description of C. malayensis, and without performing genetic analysis, the occurrence of toxic (CAWD77) and non-toxic (CAWD39 and CCMP304) strains of C. monotis was reported in New Zealand [31]. However, by retrieving genetic sequences of the same strains from GenBank, it is possible to find out that two out of three strains belonged, in fact, to C. malayensis, and that the actual C. monotis strain (CCMP304) was non-toxic. In the present study, before screening our strains for the presence of toxic compounds and toxic effects, we thus assessed Coolia spp. taxonomy based on morphology and genetic data, with special focus on the two most toxic species, i.e., C. malayensis and C. tropicalis.
From all eight species of Coolia described so far, only one (C. areolata) has no molecular data available to date [15]. Genetic separation among the species is, in general, well resolved. Despite several studies suggesting that C. malayensis and C. monotis might be the same species [1,37], currently the separation into two different species is more widely accepted [14,19]. In the present study, the phylogenetic analysis on Brazilian strains of Coolia spp. revealed similar topology to previous investigations, clearly separating all species based on both ITS (ITS-1, 5.8S rDNA, ITS-2) and LSU (D1-D3) sequences. However, these two DNA fragments resulted in distinct relative distances separating both C. palmyrensis and C. canariensis from the other species. Thus, we strongly recommend that phylogenetic analyses of Coolia include at least two markers in order to confirm species identification.
The species group encompassing C. tropicalis is characterized by a larger cell size compared to the C. monotis group (present study, Table 1; [15,16,18,19]). Both species groups are also distinguishable phylogenetically, based on the LSU D1-D3 alignment dataset (see Figure 2). Furthermore, C. tropicalis is easily separated from the other species within its own group (i.e., those possessing a rectangular 7" plate), based on the smooth cell surface on both epitheca and hypotheca (present study, Figure 7A-B; [15,23,28]).
Conversely, species within the C. monotis group are difficult to separate from each other based solely on morphological features. Plate arrangements are quite similar, size ranges overlap, and thecal plates lack any kind of perforation patterns in these smaller-celled species. For C. malayensis, a larger third post-cingular plate (3"') is the most noticeable feature distinguishing it from other species from C. monotis group. In fact, we did observe this feature in most of the cells examined in this study. However, its use as diagnostic of C. malayensis may be questionable, as plate 3"' can be, in some cases, similar in size to plate 4"' (present study, [27,38]). As an additional complicating issue, measuring a large plate like 3"' can be inaccurate due to the pronounced cell curvature, making it inappropriate to base cell identification on morphometrics in this case. Alternatively, each of the species from C. monotis group could be differentiated from C. malayensis, based on a particular feature. C. palmyrensis, for instance, may be easily separated from other species based on the density and number of pores on the cell surface [19]. In the present study, we confirmed that C. palmyrensis had a lower pore density than its sister species C. malayensis and C. santacroce (Table 1). However, as previously reported [19], some variability exists, and the ranges of both pore numbers and density can overlap between species. In this sense, C. malayensis may sometimes possess low pore density, thus mimicking C. palmyrensis, and vice versa (Table 1). Such inconstant thecae features can be attributed to natural variability inside the population, deformation due to culture conditions or to the cell-cycle phase [38]. In conclusion, as already indicated for C. monotis and C. santacroce [20], species identification within this group of species (C. monotis group) should always be supported by genetic data, as otherwise, it can lead to misidentification.

Species Distribution and Diversity in Brazil
Until quite recently, the distribution of benthic dinoflagellates was poorly documented in Brazilian waters, with the genus Coolia, represented at that time by C. malayensis, found exclusively along the southeastern coast [21]. Recent studies, however, have documented a larger number of species (C. malayensis, C. tropicalis and C. canariensis), occurring over a wider geographical distribution, including the south [12], southeast [26,28] and northeast [27,28] sectors of the Brazilian coast. In the present study, we reported for the first time the species C. santacroce and C. palmyrensis occurring in South Atlantic waters, and confirmed the high genetic diversity of this genus in Brazil, as suggested by Nascimento et al. [28] Similar species composition has been found in the Canary Islands [15,23], suggesting a high degree of connectivity over the Atlantic Ocean.
The most widely distributed species in Brazil is C. malayensis, present along the entire coast (present study, [12,[26][27][28]). The other toxic species reported herein, C. tropicalis, is more restricted to the northeastern warm waters (present study) and the tropical oceanic island of Trindade [28]. The other three species, C. canariensis [28], C. santacroce and C. palmyrensis (present study), were found exclusively in offshore sites: Trindade Island for C. canariensis, Abrolhos Archipelago for C. santacroce and C. palmyrensis, and an offshore diving site in Pernambuco State for C. palmyrensis. Similar patterns of species distribution have been observed in Australia [22] and in the Iberian Peninsula [24], with C. malayensis more broadly distributed, C. tropicalis restricted to warmer waters, and C. palmyrensis present in offshore sites. Coolia santacroce, in turn, had only been recorded, so far, in the Caribbean Sea (GenBank sequences). Thus, a general pattern of Coolia spp. distribution seems to arise from this and previous studies: C. malayensis is globally distributed (see [14]); C. monotis is probably broadly distributed in coastal waters of Europe and the East Atlantic [24]; while distributions of C. tropicalis, C. canariensis, C. santacroce and C. palmyrensis are more restricted, mainly to warmer waters and/or to offshore sites with lower hydrodynamics (present study, [22]).

Toxicity and Toxin Production
Toxicity within the genus Coolia (as Coolia sp.) was first reported in the early 1980s, based on hemolytic activity via in vitro assays, although no toxicity to mice and fish was registered for the same methanol extract [29]. Since then, this genus of benthic dinoflagellates has been considered potentially toxic [1]. However, later studies did not detect any toxic activity in several strains/species, using different cell models and organisms. For instance, ethanol extracts of cell pellets of C. monotis (strain CCMP304) or C. malayensis (CAWD39) were not toxic to mice following intraperitoneal injections in mouse bioassays (MBA) [31]. Additionally, Penna et al. [33] tested methanol extracts of C. monotis (strain CM2V) and C. malayensis (strain CCMP1345), and did not find any hemolytic activity to human erythrocytes. Similarly, methanol extracts of other C. monotis strains (CCMP2582 and CCMP304) exhibited no cytotoxicity to Rhabdomyosarcoma cells derived from bone marrow [19]. Moreover, Artemia franciscana nauplii were not affected upon exposure to living cells of C. monotis (strains Dn23EHU, DN24EHU) or C. canariensis (strains Dn28EHU, Dn29EHU) [34], and no toxic effects were observed in A. salina nauplii exposed to filtered medium of C. guanchica cultures [15]. In the present study, likewise, no toxic effects were observed in adults of A. salina exposed to increasing abundances of living C. santacroce or C. palmyrensis cells. In contrast, cells of C. malayensis and C. tropicalis were lethal to the micro-crustaceans at equivalent biomass ranges (see Figure 8), confirming the marked species-specific variability in Coolia toxicity.
After the early study by Nakajima et al. [29], Holmes et al. [30] evaluated the toxicity of C. tropicalis (as C. monotis) using MBA. The authors reported mouse mortality caused by the butanol-soluble fraction of the extract, but no toxicity from either hexane-or water-soluble fractions. Later on, acetone and ethanol extracts from two different C. malayensis strains (CAWD77 and CAWD151) were also lethal to mice via MBA [31,39]. Moreover, C. malayensis (methanol extracts) exhibited the strongest cytotoxic effects to Rhabdomyosarcoma cells when compared to C. santacroce (intermediate toxicity) and C. palmyrensis (low toxicity) [19]. Finally, methanol extracts from Brazilian strains of both C. malayensis (UFBA044) and C. tropicalis (UFBA055) showed hemolytic activity to sheep erythrocytes [27]. These are the same species here reported as lethal to A. salina upon short-term (24-96 h) exposure to living cells. In our experiments, a northeastern Brazilian strain of C. tropicalis (LM141) was relatively more toxic than a C. malayensis strain isolated from the southeastern coast (LM036).
Taking the results from this and previous toxicity assessments together, we believe that at least C. tropicalis and C. malayensis should be considered toxic species. Toxicity of other Coolia species may vary geographically and should be more carefully evaluated, perhaps using a combination of different assays. For example, C. santacroce and C. palmyrensis strains from the Caribbean or the Pacific Ocean (Palmyra Atoll) were reported as cytotoxic by Karafas et al. [19], but strains from the present study were not lethal to A. salina. Nevertheless, considering only genetically sequenced strains, C. monotis, C. canariensis and C. guanchica have not shown, up to now, any sign of toxic activity (see discussion above), while the toxicity of C. areolata has not been examined [25].
The compounds responsible for the toxic activity in C. malayensis and C. tropicalis are still controversial. In 1995, an analogue of yessotoxin (YTX), then named cooliatoxin, was described in C. tropicalis (as C. monotis) using low-resolution LC-MS/MS [30]. The exact molecular structure of that compound, however, was not elucidated. Later on, related, yet unique, compounds with fewer oxygen atoms than cooliatoxin or YTX were detected in C. malayensis from Okinawa (Japan), and described as disulphated polyether analogues of YTX based on high-resolution LC-MS/MS [32]. However, YTX analogues (including cooliatoxin) have never been detected again in other Coolia spp. cultures (present study, [23]). Besides YTX analogues, our strains of C. malayensis, C. palmyrensis and C. santacroce also lacked any other toxic compound produced by another genus of benthic dinoflagellates, Gambierdiscus, including maitotoxins, gambierones, gambieroxide, and gambieric acids. A strain of C. tropicalis, however, contained relatively high intracellular levels of 44-methyl gambierone (previously referred to as MTX-3 [35]) and a novel isomer of the same compound. Spectral data presented by Holmes et al. [30] suggest that 44-methyl gambierone was not present in that extract (absence of 1037.5 in the negative ionspray mass spectrum), suggesting either misidentification of species, intra-specific variability of toxin production or divergence of C. tropicalis between the Pacific and Atlantic Oceans.
According to Boente-Juncal et al. [5], 44-methyl gambierone exhibits similar biological activities to gambierone and CTX3, leading to the decreased viability of undifferentiated neuroblastoma cells and modified expression of excitatory neurotransmitter receptor subunits. This compound can be produced by diverse Gambierdiscus species, mainly by G. australes, G. belizeanus and G. polynesiensis, which produce large amounts (reviewed in Longo et al. 2019). Intra-cellular contents of 44-methyl gambierone ranged from 5.8 to 74.1 pg MTX1 eq. cell −1 in G. polynesiensis (Longo et al. 2019). In the smaller C. tropicalis cells (G. polynsiensis is twice the size of C. tropicalis, see [40] and Table 1 above), we measured 73 and 20 pg MTX1 eq. cell −1 of 44-methylgambierone and 44-methylgambierone isomer. Such surprisingly high toxin levels are especially relevant considering that 44-methyl gambierone may be implicated in the neurological manifestations related to ciguatera poisoning (CP) in humans [5]. Thus, the role of C. tropicalis as another causative agent of CP deserves to be considered in further investigations.

Conclusions
Coolia is a potentially toxic marine dinoflagellate genus, with many taxonomical and toxicological issues yet to be evaluated and resolved. The smaller-celled Coolia species, including C. malayensis and similar species, cannot be clearly distinguished from each other based only on morphological features. Thus, in studies of any strain from the C. monotis species group (C. monotis, C. malayensis, C. santacroce and C. palmyrensis), the use of molecular data is mandatory. C. malayensis has proved to be the most broadly distributed species of the genus, found in both temperate and tropical waters of the Atlantic and Pacific Oceans, while other species occur in more restricted areas. This study increased from three to five (out of eight) the number of Coolia species reported in Brazilian waters so far, highlighting Brazil as an area of biodiversity for this genus.
Assessment of Coolia toxicity can be rather controversial due to the distinct assays/techniques used, but also due to species-specific differences in the capacity of producing toxic compounds. In the present study, C. malayensis and C. tropicalis cells were toxic to adult Artemia individuals in feeding experiments, while C. santacroce and C. palmyrensis were not. Using both low-and high-resolution LC-MS/MS, we detected considerable amounts of 44-methyl-gambierone (MTX3)-previously limited to Gambierdiscus spp.-and a new 44-methyl gambierone isomer in C. tropicalis. According to previous studies, this compound exhibits a powerful cytotoxic effect, which might explain the toxicity in bioassays involving this species.

Sampling and Cultures
Toxicity-Nineteen sampling campaigns were conducted from October 2016 to March 2018, at nine sampling sites along the Brazilian Coast (Figure 4), including coastal rocky shores and islands. Samples were collected and processed following the procedures described in Tester et al. [41] Seaweed samples were vigorously shaken to detach particles, and the seawater containing Coolia cells was used for microscopic observation and isolation of living cells. Cells of Coolia were isolated using a capillary pipette following successive washing in sterile, local filtered seawater. After initial growth through consecutive cell divisions, the volume of culture was successively doubled by transferring the old aliquot to a larger microplate well, containing an equivalent volume of sterile, 50% diluted f/2 media (f/4), without silica and~32 salinity. From 10 mL wells, cultures were transferred to 50 and then 250 mL Erlenmeyer flasks, where they were maintained at 26 • C under a 12:12 h light cycle (irradiance of 70 ± 20 µmol m −2 s −1 ). Fourteen strains were successfully established and used in the present study (Table 4). For toxin analysis, cultivated cells were harvested at two growth stages (exponential and stationary growth phase). Cells were concentrated by centrifugation (2332 g, 5 min), the supernatant was removed, and samples were stored at −20 • C. Prior to toxin analysis, frozen cell pellets were lyophilized.

DNA Amplification, Sequencing and Molecular Phylogeny
Cultured cells of Coolia spp. were harvested by centrifugation (2332 g, 5 min). The supernatant was removed and replaced by ethanol to preserve samples until DNA analysis. Before amplification, single cells from ethanol-preserved samples were isolated with a glass capillary and washed six times with deionized water. Single Coolia cells were placed in PCR tubes (at least two tubes for each sample), containing 1-3 µL of deionized water and stored at −20 • C before direct PCR amplifications.
Two consecutive PCR reactions (nested PCR) were performed to amplify the rDNA regions ITS1-5.8S-ITS2 (ITS) and LSU (D1-D3). For the first PCR reaction, 2.5 µL of each primer (ITSfw and D3B, Table 5), 12.5 µL of PCR Master Mix 2X (Promega, Madison ® , WI, USA) containing the Taq DNA polymerase, dNTPs, MgCl2 and reaction buffers, and 6.5 µL of nuclease-free water were added to each tube. The PCR were performed in a Biometra TOne thermocycler (Analytik Jena), as follows: one initial denaturation step at 95 • C for 2 min, then 35 cycles of 30 s at 95 • C, 1 min at 62 • C (melting temperature, "MT") and 1 min at 72 • C, and a final elongation step of 5 min 72 • C. For the second PCR reaction, 1 µL of the first product was added to a new tube containing 2.5 µL of each primer (ITSfw and 28S364r for ITS region; D1R and D3B for D1-D3; Table 3), 12.5 µL of GoTaq ® G2 Hot Start Green Master Mix (Promega ® , Madison, WI, USA) and 6.5 µL of nuclease-free water. The second PCR was performed as the first one, changing the MT to 50 • C for ITS, and 56 • C for D1-D3 region. DNA amplifications were controlled by electrophoresis on agarose gel. Positive samples were purified and sequenced, as described in Chomérat et al. [42] The alignment and phylogenetic analyses were performed as described in Chomérat et al. [42], with modifications as described below. Both ITS and D1-D3 rDNA region datasets were aligned using MAFFT algorithm, with selection of the q-ins-i strategy [43]. Poorly aligned positions were removed using Gblocks algorithm [44], and the most appropriate model of sequence evolution was selected using jModeltest2 v. 2.1.10 [45]. For both rDNA regions, TrN+G were the models used for Maximum Likelihood (ML) and Bayesian Inference (BI) analysis, with 2,000,000 generations performed in BI analysis for both alignments, and sampling every 100 generations. The posterior probabilities of each clade were calculated from the remaining 20,000 trees. For some samples, the primer Coo5.8f (Table 5) was used in the sequencing reaction to obtain clearer sequences from ITS2.

Morphological Observations
Prior to the scanning electron microscopy (SEM) observations, cultured Coolia cells were preserved with neutral and acidic lugol (1%). Small sample aliquots (2-5 mL) were placed on a piece of either a 5 µm Millipore filter or a 20 µm plankton net, rinsed with distilled water, and dehydrated in a series of increasing ethanol concentrations (30%, 50%, 70%, 90%, 95% and 100%), followed by critical point drying. Samples were finally mounted on a stub and sputter coated with gold palladium. Cells were observed using a JEOL ® JSM 6360-LV (Japan) microscope at 15 Kv. Species identification was based mainly on original and recent Coolia spp. descriptions [11,15,16,18].

Toxicity Experiments
The toxicity of selected Coolia spp. strains was evaluated through bioassays using adult individuals of the brine shrimp Artemia salina. After cyst hatching, A. salina larvae were kept in controlled tanks under constant aeration and fed non-toxic Tetraselmis suecica cells for 20-30 days. Then, adult specimens of A. salina were individually placed in wells of cell culture plates containing 5 mL of autoclaved sea water each. Before each test, plates containing A. salina were acclimated for 24-48h under the experimental conditions and fed with T. suecica in an amount equivalent to 150 ng C ind −1 h −1 . During the test, they were exposed to increasing cell densities (i.e., treatments) of Coolia spp. and a complementary amount of non-toxic T. suecica cells, in order to maintain a comparable food supply over all treatments. Cell density of each strain depended on their cell biovolumes, which were calculated from approximate geometrical shapes after measuring 50 cells of each strain [49]. Cell biovolume was then converted into carbon biomass, following conversion factors described in Menden-Deuer et al. [50] The in vivo toxicity assays aimed at providing quantities of Coolia spp. equivalent to 4.7, 9.4, 18.8, 37.5, 75 and 150 ng C ind −1 h −1 for 96 h. The maximum biomass possible (according to each culture cell densities at late exponential growth phase), as well as half of the maximum, were also used as additional treatments. Maximum Coolia spp. quantities tested were equivalent to 965, 800, 750, 575 ng C ind −1 h −1 for C. malayensis, C. tropicalis, C. santacroce, and C. palmyrensis, respectively.
Three cell culture plates were used for each experimental treatment, each containing twelve individuals of A. salina. From those, ten individuals were exposed to the toxic microalgae and the other two were exposed to the control condition, consisting of non-toxic T. suecica cells only. An extra plate containing twelve brine shrimps was used to increase the number of control individuals to 60, while 30 individuals were exposed to each treatment containing Coolia cells, adding up to 300 brine shrimps in each experiment. Survival of A. salina was evaluated after 1, 3, 12, 24, 48, 72 and 96 h of exposure. Individuals were considered dead if completely motionless at the bottom for 10 consecutive seconds.

Toxin Analysis
Prior to toxin analysis, cell pellets were sonicated in bath ultrasound (Transonic TI-H-15, Elma ® , Germany) at 45 kHz for 15 min with methanol/water (9:1, v/v). The mixture was centrifuged at 1200 g for 15 min. Supernatant was passed through a centrifuge NanoSep filter (0.2 µm Nylon, PALL ® , UK) and recovered into plastic vials with conical insert.
Filtered extracts from cell pellets were analyzed using two hybrid systems coupling ultra-high-performance liquid chromatography with tandem mass spectrometry (UHPLC-MS/MS), in either low-(LR) or high-resolution (HR).

System A: HR-MS/MS
System A was composed of a UHPLC system (1290 Infinity II, Agilent Technologies, Santa Clara, CA, USA) coupled to a 6550 ifunnel Q-TOF (Agilent Technologies, CA, USA), equipped with a Dual Jet Stream ® ESI source. The high-resolution instrument was operated in both full scan and targeted MS/MS modes. Acquisition was carried out in positive and negative ionization modes, with optimized parameter sources. Temperature was set at 250 • C, drying gas flow at 16 L min −1 , nebulizer gas at 15 psi and sheath gas at 12 L min −1 and 400 • C. Capillary and nozzle voltages were set at 5000 V and 1000 V, respectively. Two reference masses m/z 121.0509 (purine) and m/z 922.0099 (hexakis phosphazine) were continuously monitored during the run.
The chromatographic conditions were similar to those described for system B in Section 5.5.2 below, except for the porosity of the Kinetex C18 column used (1.7 µm, instead of 2.6 µm). Mass spectra were acquired from 100 to 1700 m/z, with an acquisition rate of 2 spectra s −1 . The targeted MS/MS mode was applied over m/z 50-1700, with an MS scan rate of 10 spectra s −1 and an MS/MS scan rate of 3 spectra s −1 . Three fixed collision energies were applied by ionization mode (15, 30 and 45 eV in ESI + ; 30, 45 and 90 eV in ESI − ), to obtain an overview of the fragmentation pathways. Instrument control, data processing and analysis were conducted using Mass Hunter software v.8.0 (Agilent Technologies, CA, USA).
A linear gradient using water as eluent A and 95% acetonitrile as eluent B, both eluents containing 2 mM ammonium formate and 50 mM formic acid, was run through a Kinetex C18 column, 50 × 2.1 mm, 2.6 µm, 100 Å (Phenomenex, Torrance, CA, USA). The flow rate was 0.4 mL min −1 , the injection volume was 5 µL and the column temperature was 40 • C. The elution gradient was set as follows: 10% B to 95% B from 0 to 10 min, hold at 95% B for 2 min, decrease from 95% to 10% in 1 min and hold during 3 min to equilibrate. Mass spectrometric detection was performed in negative ionization mode using MRM scanning. The m/z transition used is listed in Table 6 for Gambierdiscus metabolites and in Table  S2 for YTXs toxins, for which certified standards were available. The optimized ESIparameters were set as follows: curtain gas at 25 psi, ion spray at −4500 V, turbo gas temperature at 500 • C, gas 1 and 2 at 50 psi, declustering potential at −210 V for Gambierdiscus metabolites and −120 V for YTXs toxins and an entrance potential at −10 V. Table 6. List of MRM transitions (m/z) used in ESI-to detect MTXs, gambierone toxins and gambieric acids on system B (API 4000QTrap).

Quantification of Gambierone Toxins in C. tropicalis
In order to quantify the MTX and gambierone toxins, a calibration curve of MTX1 was prepared from successive dilutions of a standard solution (Wako, Japan) in 50% MeOH, with concentrations ranging from 0.2 to 5.0 µg mL −1 . Due to the lack of analytical standards, each targeted compound was quantified from the MTX1 calibration curve prepared, assuming equivalent molar response.  Table S1. List of compounds extracted from the Personal Compound Database and Library (PCDL, Agilent Mass Hunter software), created by Phycotoxins Laboratory (IFREMER, France) and used for Coolia spp. Screening, Table S2: List of MRM transitions (m/z) used in ESI-to detect YTXs on system B (API 4000QTrap), Figure S1. (A) LC-HRMS chromatogram of C. tropicalis extract and high resolution full scan mass spectra acquired in negative mode on the apex of peaks at (A1) 6.0 min for 44-methyl gambierone isomer and (A2) at 6.6 min for 44-methyl gambierone, Figure S2

Conflicts of Interest:
The authors declare no conflicts of interest.