Toxicity and Growth Assessments of Three Thermophilic Benthic Dinoflagellates (Ostreopsis cf. ovata, Prorocentrum lima and Coolia monotis) Developing in the Southern Mediterranean Basin

Harmful benthic dinoflagellates, usually developing in tropical areas, are expanding to temperate ecosystems facing water warming. Reports on harmful benthic species are particularly scarce in the Southern Mediterranean Sea. For the first time, three thermophilic benthic dinoflagellates (Ostreopsis cf. ovata, Prorocentrum lima and Coolia monotis) were isolated from Bizerte Bay (Tunisia, Mediterranean) and monoclonal cultures established. The ribotyping confirmed the morphological identification of the three species. Maximum growth rates were 0.59 ± 0.08 d−1 for O. cf. ovata, 0.35 ± 0.01 d−1 for C. monotis and 0.33 ± 0.04 d−1 for P. lima. Toxin analyses revealed the presence of ovatoxin-a and ovatoxin-b in O. cf. ovata cells. Okadaic acid and dinophysistoxin-1 were detected in P. lima cultures. For C. monotis, a chromatographic peak at 5.6 min with a mass m/z = 1061.768 was observed, but did not correspond to a mono-sulfated analogue of the yessotoxin. A comparison of the toxicity and growth characteristics of these dinoflagellates, distributed worldwide, is proposed.


Introduction
Harmful algal blooms (HABs) seem to have become more frequent, intense and widespread [1]. These events occur preferentially in coastal waters and sheltered areas throughout the world, such as harbors, small bays and coastal lagoons [2]. This phenomenon has been attributed either

Ostreopsis cf. ovata
Vegetative cells were ovoid to oblong, pointed toward the ventral area in apical view, with many golden chloroplasts. A sub-spherical nucleus was located at the posterior end of the cell ( Figure 1). For cells harvested in the exponential growth phase, mean length and width were 50.38 ± 4.36 µm and 36.80 ± 3.33 µm, respectively. Vegetative cell sizes of the O. cf. ovata OOBZT14 strain matched those reported for the Mediterranean populations and were close to those found by other authors [26,74,78,83]. During the stationary growth phase, cells with more rounded and irregular shapes were observed in O. cf. ovata cultures with a mean length and width of 55.18 ± 5.25 µm and 39.58 ± 3.61 µm, respectively (Table 1). Our results are in agreement with those of Accoroni et al. (2012) [84], Vanucci et al. (2012) [85] and Pezzolezi et al. (2014) [86], who reported an increase in cell size and biovolume at the stationary and the decline phases of the cultures. The appearance of these large anomalous cells can be interpreted as a response to unfavorable conditions (nutrient depletion) or as a precursor of pellicle cysts at the end of the growth phase [83]. 39.58 ± 3.61 μm, respectively (Table 1). Our results are in agreement with those of Accoroni et al. (2012) [84], Vanucci et al. (2012) [85] and Pezzolezi et al. (2014) [86], who reported an increase in cell size and biovolume at the stationary and the decline phases of the cultures. The appearance of these large anomalous cells can be interpreted as a response to unfavorable conditions (nutrient depletion) or as a precursor of pellicle cysts at the end of the growth phase [83].  Table 1. Morphometric characteristics of Ostreopsis cf. ovata (OOBZT14), Prorocentrum lima (PMBZT14) and Coolia monotis (CMBZT14) strains: mean, minimum, maximum values (µm) and standard deviation of the length and width of the cells harvested in both exponential and stationary growth phases (n = 30).

Prorocentrum lima
Cells were oval to oblong in valve view, and both valves were concave in lateral view. The periflagellar area was V-shaped and located on the right valve. The ring-shaped pyrenoid was situated in the center of the cell, and the nucleus occupied the dorsal part ( Figure 2). Cell length varied from 42. 98-48.80 µm and width 34.941-37.95 µm. The variation in cell shape, expressed by the length/width ratio, ranged from 1. 23-1.31. No size differences (p > 0.05) were observed between cells in exponential and stationary growth phases (Table 1). Cell sizes of P. lima from Tunisian waters fit well with the description of   [87] and Aissaoui et al. (2014) [70].

Prorocentrum lima
Cells were oval to oblong in valve view, and both valves were concave in lateral view. The periflagellar area was V-shaped and located on the right valve. The ring-shaped pyrenoid was situated in the center of the cell, and the nucleus occupied the dorsal part ( Figure 2). Cell length varied from 42. 98-48.80 μm and width 34.941-37.95 μm. The variation in cell shape, expressed by the length/width ratio, ranged from 1. 23-1.31. No size differences (p > 0.05) were observed between cells in exponential and stationary growth phases (Table 1). Cell sizes of P. lima from Tunisian waters fit well with the description of   [87] and Aissaoui et al. (2014) [70].

Molecular Analysis and Phylogeny
Sequences of 930, 856 and 887 base pairs of the partial large subunit (LSU) rDNA (D1-D3) have been obtained from O. cf. ovata (OOBZT14), P. lima (PLBZT14) and C. monotis (CMBZT14) strains, respectively. They were deposited in GenBank with Accession Numbers KX845008 (OOBZT14), KX845009 (PLBZT14) and KX845010 (CMBZT14). These sequences were similar to a batch of sequences from France, Italy and Greece for O. cf. ovata, from Spain, Italy and Australia for P. lima and from Greece, Italy and Netherlands for C. monotis; all available in GenBank and identified as O. cf. ovata, P. lima and C. monotis. The identity of the three strains OOBZT14, PLBZT14 and CMBZT14, determined on the basis of morphological examination of the cells, was then confirmed. The phylogeny inferred from LSU rDNA showed that all of these sequences clustered in a highly supported clade, which indicated that the LSU sequences of the strains from Bizerte Bay are identical to strains mainly found in the Mediterranean Sea and Atlantic Ocean. Results revealed that the OOBZT14 strain was close to strains found in France, Italy and Greece. For P. lima, the PLBZT14 strain was identical to strains from Italy and from the Atlantic Spanish coast. The CMBZT14 strain grouped with other stains from Greece and (e,f) hypothecal view (g,h) U-shaped nucleus located in the dorsal region of the cell. Scale bars, 10 µm. Po: pore plate; S: sulcus; C: cingulum.

Molecular Analysis and Phylogeny
Sequences of 930, 856 and 887 base pairs of the partial large subunit (LSU) rDNA (D1-D3) have been obtained from O. cf. ovata (OOBZT14), P. lima (PLBZT14) and C. monotis (CMBZT14) strains, respectively. They were deposited in GenBank with Accession Numbers KX845008 (OOBZT14), KX845009 (PLBZT14) and KX845010 (CMBZT14). These sequences were similar to a batch of sequences from France, Italy and Greece for O. cf. ovata, from Spain, Italy and Australia for P. lima and from Greece, Italy and Netherlands for C. monotis; all available in GenBank and identified as O. cf. ovata, P. lima and C. monotis. The identity of the three strains OOBZT14, PLBZT14 and CMBZT14, determined on the basis of morphological examination of the cells, was then confirmed. The phylogeny inferred from LSU rDNA showed that all of these sequences clustered in a highly supported clade, which indicated that the LSU sequences of the strains from Bizerte Bay are identical to strains mainly found in the Mediterranean Sea and Atlantic Ocean. Results revealed that the OOBZT14 strain was close to strains found in France, Italy and Greece. For P. lima, the PLBZT14 strain was identical to strains from Italy and from the Atlantic Spanish coast. The CMBZT14 strain grouped with other stains from Greece and Netherlands. Hence, the three benthic strains from Bizerte Bay belonged to the Mediterranean/Atlantic clades (Figure 4a Netherlands. Hence, the three benthic strains from Bizerte Bay belonged to the Mediterranean/Atlantic clades (Figure 4a-c). (a)
In our study, the OOBZT14 growth curve showed a lag phase from Days 0-2, and an exponential phase characterized by three steps: an initial exponential growth from Days 2-8, then a slower growth from Days 8-14, followed by a resumption of growth from Days 14-24. Cells reached the stationary growth phase after 24 days of culture ( Figure 5a). Scalco et al. (2012) [83] reported also an exponential growth phase of 20 days, but cultures reaching the stationary phase earlier, after 10 or 13 days, were noted by Guerrini et al. (2010) [22] and Brissard et al. (2014) [19], respectively.
In our study, the OOBZT14 growth curve showed a lag phase from Days 0-2, and an exponential phase characterized by three steps: an initial exponential growth from Days 2-8, then a slower growth from Days 8-14, followed by a resumption of growth from Days 14-24. Cells reached the stationary growth phase after 24 days of culture ( Figure 5a). Scalco et al. (2012) [83] reported also an exponential growth phase of 20 days, but cultures reaching the stationary phase earlier, after 10 or 13 days, were noted by Guerrini et al. (2010) [22] and Brissard et al. (2014) [19], respectively.

Coolia monotis
For C. monotis, the maximum cell density was 27,057 cell•mL −1 after 24 days of culture. The maximum growth rate was 0.35 ± 0.01 d −1 . To our knowledge, studies characterizing the growth potential of Coolia monotis are very limited. Faust (1992) [97] reported a doubling time of 3-4 days during the logarithmic phase of growth for cultures grown at 23 °C. Morton et al. (1992) [98] reported growth rates, ranging from 0.2-0.6 d −1 for strains grown under different temperatures, salinities and light intensities.
CMBZT14 growth patterns were similar to those of O. cf. ovata, with a lag phase from Days 0-2 and an exponential phase characterized by three steps: initial exponential growth from Days 2-12, slower growth from Days 12-16 and a resumption of growth from Days 16-24. The beginning of the stationary phase was also observed after 24 days of culture ( Figure 5c).

Coolia monotis
For C. monotis, the maximum cell density was 27,057 cell·mL −1 after 24 days of culture. The maximum growth rate was 0.35 ± 0.01 d −1 . To our knowledge, studies characterizing the growth potential of Coolia monotis are very limited. Faust (1992) [97] reported a doubling time of 3-4 days during the logarithmic phase of growth for cultures grown at 23 • C. Morton et al. (1992) [98] reported growth rates, ranging from 0.2-0.6 d −1 for strains grown under different temperatures, salinities and light intensities.
CMBZT14 growth patterns were similar to those of O. cf. ovata, with a lag phase from Days 0-2 and an exponential phase characterized by three steps: initial exponential growth from Days 2-12, slower growth from Days 12-16 and a resumption of growth from Days 16-24. The beginning of the stationary phase was also observed after 24 days of culture ( Figure 5c).
Our results showed that the O. cf. ovata growth rate (0.59 ± 0.08 d −1 ) is clearly higher than those of Coolia monotis (0.35 ± 0.01 d −1 ) and P. lima (0.33 ± 0.04 d −1 ), which suggests that O. cf. ovata has an ecological advantage and can predominate in coastal waters. However, maximum cell densities were inversely proportional to growth rates, with the lowest density recorded for O. cf. ovata (13,095 cell·mL −1 ) and the highest for P. lima (32,019 cell·mL −1 ).
Growth characteristics of these dinoflagellates originating from various ecosystems are summarized in Tables A1-A3. Data are from laboratory experiments of cultured strains growing in different environmental conditions (irradiance, temperature and salinity) corresponding to local conditions. Results from the literature, gathered in these tables and shown in Figure 6 Our results showed that the O. cf. ovata growth rate (0.59 ± 0.08 d −1 ) is clearly higher than those of Coolia monotis (0.35 ± 0.01 d −1 ) and P. lima (0.33 ± 0.04 d −1 ), which suggests that O. cf. ovata has an ecological advantage and can predominate in coastal waters. However, maximum cell densities were inversely proportional to growth rates, with the lowest density recorded for O. cf. ovata ( 13,095 cell•mL −1 ) and the highest for P. lima (32,019 cell•mL −1 ).
In our study, no clear pattern was observed for toxin content in relation with growth phase. OOBZT14 cells were harvested at the early (Day 12) and late (Day 20) exponential growth phase. Ovatoxin-a and -b levels decreased slightly after 20 days ( Figure 8a). Many reports indicated that O. cf. ovata increases toxin production during the progression of growth from the exponential to stationary phase [9,22,23,89,100]. Nevertheless, toxin content can vary considerably during each growth phase. Scalco et al. (2012) [83] noted that the cellular toxin content was markedly lower during the post-exponential growth phase than during the exponential phase for cells cultivated at 22 • C and under a 15L:9D illumination cycle. Moreover, based on hemolytic bioassays, Granéli et al. (2011) [9] found that the hemolytic activity on Day 14, was higher than that on Day 20 for O. cf. ovata cultures growing at 23 • C.  [19]. In our study, no clear pattern was observed for toxin content in relation with growth phase. OOBZT14 cells were harvested at the early (Day 12) and late (Day 20) exponential growth phase. Ovatoxin-a and -b levels decreased slightly after 20 days ( Figure 8a). Many reports indicated that O. cf. ovata increases toxin production during the progression of growth from the exponential to stationary phase [9,22,23,89,100]. Nevertheless, toxin content can vary considerably during each growth phase. Scalco et al. (2012) [83] noted that the cellular toxin content was markedly lower during the post-exponential growth phase than during the exponential phase for cells cultivated at 22 °C and under a 15L:9D illumination cycle. Moreover, based on hemolytic bioassays, Granéli et al. (2011) [9] found that the hemolytic activity on Day 14, was higher than that on Day 20 for O. cf. ovata cultures growing at 23 °C.  For O. cf. ovata, recorded data did not show significant differences in toxin profiles between Mediterranean strains (Table A1). Ovatoxin-a is the predominant toxin except for the Italian strain CBA2-122, which produces higher amounts of ovatoxin-f [27]. In Atlantic waters, Nascimento et al. (2012) [89] found higher levels of OVTX-b than OVTX-a in the Brazilian strains LCA-B7 and LCA-E7. Ovatoxin profiles seem to be strain specific, with isolates that cannot produce some ovatoxins. Our OOBZT14 produced only OVTX-a and -b, and the Italian strain CBA29-2012 was found not to produce OVTX-b and -c [82]. Recently, the presence of a new ovatoxin analog, named ovatoxin-h, was reported for the French strain IFR-OST-03V [65]. Concerning the mascarenotoxins (McTX-a and McTX-c), their presence was only observed for the strain D483 originating from the Gulf of Naples [28,83]. For the putative palytoxin (pPLTX), it is usually found at low levels and is not systematically present in all strains from Mediterranean and Atlantic waters (Figure 9a). pg PLTX (palytoxin) equivalent•cell −1 produced by Ostreopsis cf. ovata; (b) okadaic acid (OA) and dinophysistoxin-1 (DTX-1) in pg•cell −1 produced by Prorocentrum lima.
For O. cf. ovata, recorded data did not show significant differences in toxin profiles between Mediterranean strains (Table A1). Ovatoxin-a is the predominant toxin except for the Italian strain CBA2-122, which produces higher amounts of ovatoxin-f [27]. In Atlantic waters, Nascimento et al. (2012) [89] found higher levels of OVTX-b than OVTX-a in the Brazilian strains LCA-B7 and LCA-E7. Ovatoxin profiles seem to be strain specific, with isolates that cannot produce some ovatoxins. Our OOBZT14 produced only OVTX-a and -b, and the Italian strain CBA29-2012 was found not to produce OVTX-b and -c [82]. Recently, the presence of a new ovatoxin analog, named ovatoxin-h, was reported for the French strain IFR-OST-03V [65]. Concerning the mascarenotoxins (McTX-a and McTX-c), their presence was only observed for the strain D483 originating from the Gulf of Naples [28,83]. For the putative palytoxin (pPLTX), it is usually found at low levels and is not systematically present in all strains from Mediterranean and Atlantic waters (Figure 9a).   Concerning toxin production, only two Mediterranean strains (IFR-OST-03V isolated in France and IRTA-SMM-12-62 isolated in Spain) showed high toxic levels (250-300 pg·cell −1 ) [19,99]. All of the other isolates displayed lower toxin contents (3.51-57.5 pg·cell −1 ). Both O. cf. ovata strains, developing in Atlantic waters, showed high toxin levels (200 and 468 pg·cell −1 ) [89] (Figures 9a and 10a). Few data are available for Ostreopsis ovata toxicity in Pacific and Indian waters. These strains seem to be non-toxic or to display low toxicity [101][102][103][104]. Rhodes [106] resulted in no palytoxin-like activity in an O. cf. ovata Indian isolate. Further investigations on Indo-Pacific strains are needed. Nevertheless, O. cf. ovata toxicity seems to be closely related to their genetics, with toxic strains belonging to the Mediterranean-Atlantic clades and less toxic isolates belonging to the Indo-Pacific clade; even if many other environmental factors are suspected to be involved in toxin production.

Prorocentrum lima
Okadaic acid (OA) and dinophysistoxin-1 (DTX-1) were detected in PLBZT14 cells, with the OA being the most predominant toxic compound (Figure 7b). Toxin production was greater for cells harvested on Day 12 (OA = 28.33 pg·cell −1 , DTX-1 = 7.4 pg·cell −1 ) than after 60 days of culture (OA = 7.13 pg·cell −1 , DTX-1 = 2.23 pg·cell −1 ) (Figure 8b). PLBZT14 strain seems to be highly toxigenic. Our results are close to those reported by Lee et al. (1989) [31] for Spanish isolates from Vigo (OA = 5-24.5, DTX-1 = 6-14.3 pg·cell −1 ) and a Japanese isolate from Okinawa (OA = 26, DTX-1 = 13 pg·cell −1 ) (Table A2) Concerning the toxin production during growth, maximum toxin concentrations for P. lima cells during the stationary phase were described by many authors [39,90,108,109]. At the end of our experiment (Day 60), P. lima cells densities are still increasing, and toxin content was lower than after 12 days of culture. Some studies have shown that the toxin production did not increase exponentially during cell growth. In order to draw conclusions, we must determine the toxin production kinetics of OOBZT14 and PLBZT14 throughout the entire growth cycle, by harvesting cells at different growth phases and at different times of a same phase.
For Prorocentrum lima, recorded data showed that okadaic acid and dinophysistoxin-1 are the most common compounds in Mediterranean, Atlantic and Pacific strains (Table A2). OA levels are usually higher than those of DTXs (Figure 9b). However, some variability in toxin profiles and production can be observed. Pan et al. (1999) [94] reported the dominance of dinophysistoxin-4 in a Canadian isolate, and Morton and Tindall (1995) [34] found higher methyl-okadaic acid levels in Australian clones from Heron Island. Low DTX-4 and DTX-2 levels were detected for strains from the United Kingdom [40] and from Spain [39], respectively. Prorocentin or 4-hydroxyprorocentrolide and 14-O-acetyl-4-hydroxyprorocentrolide were only reported for Pacific strains PL021117001 and PL01 from Taiwan [42,110]. Some OA esters, such as OA-D6, OA-D8 and OA-D9, were found for the strain IO66-01 from Portugal [96]. OA-D10a and OA-D10b were also reported for strains from Southern China [111]. No clear pattern emerges for toxin production as a function of the geographical distribution (Figure 9b). Results from the literature summarized in Table A2 showed for P. lima an important variability in the toxin content (0.39-14.3 pg·cell −1 for total DTXs and 1.9-41 pg·cell −1 for OA), whatever the geographic location considered (Figure 10b).

Coolia monotis
For C. monotis, Holmes et al. (1995) [52] reported that an Australian isolate contained a monosulfated polyether toxin, named cooliatoxin. A chromatographic peak at 5.6 min was observed in CMBZT14 extracts (Figure 7c). This peak corresponded to a mass close to that of the cooliatoxin. In order to compare the spectra of yessotoxin (YTX) masses and the hypothetical cooliatoxin, an MS/MS fragmentation was performed on samples, as well as on a standard of YTX (Figure 7c). YTX standard was detected at 6.2 min with a mass m/z = 1141.4669 and an error of 4.2 ppm compared to the calculated mass. The mass spectrum allowed finding the characteristic fragment ions of the molecule at m/z = 1061.5111, 924.4142, 855.3802 and 713.3157. The detected peak at 5.6 min had a mass m/z = 1061.768 with an important error of 239 ppm compared to the calculated mass of a mono-sulfated form of YTX. The MS/MS spectrum did not confirm a similar structure to the YTX. In conclusion, for CMBZT14 strain, the major detected peak at 5.6 min with a mass m/z = 1061.768 close to that of the cooliatoxin (m/z = 1061.5), did not correspond to a mono-sulfated analogue of the yessotoxin. The structure and toxicity of the unknown compound detected in the CMBZT14 strain needs to be investigated. Fraga et al. (2008) [112] analyzed the toxic content of several C. monotis strains (CM2V, CM6V, VGO782, RIKZ4, CCMP1345 and VGO858). A peak at m/z = 1067 was detected, but the ion was rejected as a YTX analog after complementary mass fragmentation. Observations on other Atlantic and Mediterranean strains confirmed a lack of toxicity in C. monotis [20,58,64,113]. Toxic C.  [114]. These strains were recently re-identified as Coolia tropicalis and Coolia malayensis based on the application of molecular techniques [57,115]. Hemolytic activity of C. monotis was reported by Nakajima et al. (1981) [53] for Japanese isolates, and Pagliara and Caroppo (2012) [54] showed that the cell lysate of C. cfr. monotis from Italy had low hemolytic activity and inhibited sea urchin embryo development. However, these studies did not mention genetic data. Further investigations are needed in order to affirm or deny C. monotis toxicity. Isolation and characterization of natural compounds from C. monotis deserve also more interest, knowing that a ceramide with a novel branched-chain and an unprecedented dioxocyclononane named cooliatin were already identified [116,117].

Conclusions
O. cf. ovata (OOBZT14) and P. lima (PLBZT14) strains from Tunisian waters are toxic, with ovatoxin-a and okadaic acid being the most abundant compounds, respectively. Tunisian marine ecosystems, as in the whole Mediterranean, are facing water warming. This could promote the development of these thermophilic toxic species. O. cf. ovata, characterized by a high growth rate in comparison with other benthic species, can out-compete the co-existing microalgae. Blooms of O. cf. ovata could threaten human health through the emission of noxious aerosols. P. lima, characterized by high cell densities and low dispersion capacities, can form toxic hot spots in localized areas and lead to catastrophic effects in the proximity of shellfish farming areas. Measures to protect human health and economic activities must be taken. Monitoring programs have to determine the risk of impacts from toxic benthic microalgae and need to include regular analyses for the related toxins. For C. monotis (CMBZT14), further investigations are required to elucidate the chemical structure of the detected compounds and to clarify their toxicological properties by performing mouse bioassays, hemolytic tests and cytotoxicity experiments.
Temperature, salinity and irradiance are the most important environmental factors influencing the growth and cell toxin content of dinoflagellate species [11,[118][119][120][121][122][123]. The culture medium and origin of the water used for cultivation could also affect these biological parameters, highlighting specific requirements regarding certain trace elements [124]. The genetic and related physiological plasticity of the strains could also explain the variability of the responses of the dinoflagellates to specified environmental factors [118]. The comparison of the growth and toxin content of O. cf. ovata and P. lima developing in large marine ecosystems, including Mediterranean, Atlantic and Indo-Pacific waters, performed on the basis of data available in the literature, suggests a huge intraspecific variability and that toxin production and growth could be driven by both the intrinsic and the prevailing environmental factors.

Sampling Site
Ostreopsis cf. ovata, Prorocentrum lima and Coolia monotis were collected from Bizerte Bay, North of Tunisia (37 • 16 7 N 9 • 52 58 E), in April and July 2014 ( Figure 11). Bizerte Bay is situated in a harbor area and is connected to a semi-enclosed lagoon, the Bizerte Lagoon. Several oyster and mussel farms are implemented in this coastal lagoon, which represents one of the major aquaculture areas in Tunisia. Sporadic HABs events were recorded in Bizerte lagoon in association with PSP (Alexandrium spp.), DSP (Dinophysis sp., P. lima, P. mexicanum and P. minimum) and ASP (Pseudo-nitzschia spp.) episodes [125]. P. lima represents a significant part of the seawater microphytoplanktonic community of the Bizerte Lagoon, reaching concentrations higher than 10 4 cell·L −1 [69]. No published data are available for the Bizerte Bay. Nonetheless, this bay can shelter toxic dinoflagellates, which can increase their range via the channel and proliferate in the lagoon.

Isolation and Culture Conditions
The three benthic dinoflagellates were isolated from the macrophyte Cymodocea nodosa. Fresh leaves of this magnoliophyte were hand-collected in Bizerte Bay (0.5-1-m depths), placed in plastic jars containing seawater and transported to the laboratory. Twenty grams of leaves were placed into a jar containing 250 mL of seawater (previously filtered through 180 μm) and vigorously shaken to allow the dislodgement of epiphytic microalgal cells. The sample was then concentrated on a 20-μm mesh sieve and observed under an inverted photonic microscope. Cells were harvested in April 2014 for

Isolation and Culture Conditions
The three benthic dinoflagellates were isolated from the macrophyte Cymodocea nodosa. Fresh leaves of this magnoliophyte were hand-collected in Bizerte Bay (0.5-1-m depths), placed in plastic jars containing seawater and transported to the laboratory. Twenty grams of leaves were placed into a jar containing 250 mL of seawater (previously filtered through 180 µm) and vigorously shaken to allow the dislodgement of epiphytic microalgal cells. The sample was then concentrated on a 20-µm mesh sieve and observed under an inverted photonic microscope. Cells were harvested in April 2014 for P. lima and C. monotis and July 2014 for O. cf. ovata. The three strains were isolated by picking a large number of single cells using the capillary pipette method. Non-axenic monoclonal cultures were grown in enriched natural sea water medium (ENSW) [126], at stable conditions of salinity 36, temperature 25 • C and irradiance 80 µmol photons·m −2 ·s −1 in a 12:12 light:dark cycle. The strains were named OOBZT14, PLBZT14 and CMBZT14 corresponding to O. cf. ovata, P. lima and C. monotis, respectively.

Morphology
Morphometric features were determined using a photonic microscope (Leica microsystems CMS GmbH, DM IL LED model, Wetzlar, Germany). Vegetative cells in exponential and stationary growth phases were fixed and cell dimensions determined at 400× magnification using Leica Application Suite software (LAS, Version 3.0, Leica Microsystems Ltd, Heerbrugg, Switzerland). For each strain, the length and width of up to 30 cells were measured. To determine the thecal plate morphology, cells were stained with calcofluor (0.5 mg·mL −1 , Sigma-Aldrich, St. Louis, MO, USA) [127] and observed under a Leica epifluorescent microscope (Leica microsystems CMS GmbH, DM2500 M model, Wetzlar, Germany). DAPI staining was also performed to observe nuclear DNA. The identification of the three benthic dinoflagellates was then confirmed by ribotyping.

DNA Extraction and PCR
Total genomic DNA was extracted from the pellets of the three strains (OOBZT14, PLBZT14 and CMBZT14) obtained by centrifuging cultures of 30 mL during 10 min at 3500× g and 4 • C. For the extraction, the classical phenol-chloroform method was used [128]. The cellular material was released by enzymatic lysis, using proteinase K digestion. The DNA was separated from protein by phenol:chloroforme:isoamyl alcohol (25:24:1) extraction, then extracted using chloroform:isoamyl alcohol (24:1). The separation of the aqueous and organic phases was performed by centrifugation. The aqueous phase contains the DNA, which was ultimately recovered in solid form, as a result of precipitation in ethyl alcohol. DNA was then resuspended on ultra-pure water. For PCR, the oligonucleotide primers and methods used were those described in Nézan et al. (2014) [129]. We focused on Internal Transcribed Spacers ITS regions and D1-D3 areas of the 28S rRNA of the strains, since these regions have been shown to be efficient to discriminate species.
All reagents were purchased from Sigma-Aldrich, St. Louis, MO, USA.

Phylogeny
For the phylogenetic analyses, the sequences of the Bizerte bay strains were aligned together with other related sequences in three independent datasets. For Ostreopsis cf. ovata, a matrix of 684 bp and 42 LSU rDNA sequences including the strain OOBZT14 and 38 other Ostreopsis sequences and 3 sequences of Coolia (as the outgroup) retrieved from GenBank was used. For Prorocentrum lima, a matrix of 977 bp and 36 sequences including PLBZT14 strain and 33 sequences of Prorocentrum and two sequences of Scrippsiella (as the outgroup) retrieved from GenBank was used. For Coolia monotis, a matrix of 581 bp and 28 LSU rDNA sequences including the strain CMBZT14 and 25 other Coolia sequences, and two sequences of Ostreopsis (as the outgroup) retrieved from GenBank was prepared. The matrixes of Ostreopsis and Coolia sequences were aligned using MAFFT software Version 7 [130], with selection of the Q-ins-i algorithm, which considers the secondary structure for the alignment, while the Prorocentrum sequences were aligned using MUSCLE software v. 3.7 [131]. The three alignments were refined by eye and analyzed by two methods of phylogenetic reconstruction: maximum likelihood (ML), using PhyML v.3.0 software [132] and Bayesian inference (BI) using MrBayes v. 3.1.2 [133]. The software jModeltest v 0.1.1 [134] was first used to select the most suitable model of substitutions. The general-time reversible model (GTR + I + G) was chosen as indicated by the hierarchical likelihood ratio tests (hLRTs), Akaike Information Criterion 1 (AIC1), Akaike Information Criterion 2 (AIC2) and Bayesian information criterion (BIC) tests implemented in jModeltest. Bootstrap values (support for branches) of trees were obtained after 1000 iterations in ML. For Bayesian inference, four Markov chains were run simultaneously for 2 × 10 6 generations with sampling every 100 generations. On the 2 × 10 4 trees obtained, the first 2000 were discarded (burn-in), and a consensus tree was built from the remaining trees. The posterior probabilities corresponding to the frequency, with which a node is present in preserved trees, were calculated using a coupled Monte Carlo Metropolis approach-Markov Chain (MCMC).

Growth Characteristics
For each strain, cell concentrations were followed in triplicates, every two days, during 34 days for O. cf. ovata and C. monotis and during 60 days for P. lima. After homogenization, 3 mL of each culture were taken out axenically from the 250-mL flasks, always at the same time of the day. Fixed cells were counted on a Sedgewick-Rafter counting slide, under an inverted photonic microscope. In accordance with Guillard [135], the maximum growth rate (µ m ; expressed in day −1 ) was calculated from the slope of a linear regression over the entire exponential phase of growth by the least square fit of a straight line to the data after logarithmic transformation; µ m = Ln (N 1 ) − Ln (N 0 )/T 1 − T 0 in units of day −1 , where N 1 and N 0 were the cell density at time T 1 and T 0 , respectively, during the linear portion of the exponential growth phase.

Toxin Analysis
Twenty eight milliliters of the corresponding culture were harvested at Days 12 and 20 for O. cf. ovata and C. monotis and at Days 12 and 60 for P. lima. Cells were centrifuged at 3500× g during 10 min and the supernatant carefully removed. The pellets were stored at −20 • C until toxin analysis.

Sample Preparation
Culture pellets were dissolved with 1 mL methanol 100% for P. lima and C. monotis and with 1 mL methanol 90% for Ostreopsis cf. ovata. Mixtures were ground with glass beads (0.25 g) in a mixer mill (Retsch MM400, Germany) for 30 min. After centrifugation at 5000× g during 10 min, the supernatants were collected and filtered through 0.2 µm before injection to the liquid chromatograph and mass spectrometry in tandem (LC-MS/MS).

Instrumentation: LC-MS/MS Systems
The analyses were carried out using two LC-MS/MS systems: (A) triple quadripole (QqQ); (B) high resolution quadrupole time of flight (Q-TOF). For the liquid chromatography conditions, a C 18 Kinetex column (Phenomenex, Torrance, CA, USA) was employed with a linear gradient using water as Eluent A and 95% acetonitrile/water as Eluent B, both eluents containing 2 mM ammonium formate and 50 mM formic acid.
System A is composed of an LC system (UFLC Nexera, SHIMADZU, Tokyo, Japan) coupled to a hybrid triple quadrupole/ion-trap mass spectrometer (API4000Qtrap, SCIEX, Redwood City, CA, USA) equipped with a Turbospray ® interface (SCIEX, Redwood City, CA, USA). The instrument control, data processing and analysis were conducted using Analyst software. Mass spectrometry detection was performed in both negative and positive mode using multiple reaction monitoring (MRM) and scanning a minimum of two transitions for each toxin. System B is composed by a UHPLC system (1290 Infinity II, Agilent Technologies, Santa Clara, CA, USA) coupled to a 6550 ifunnel Q-TOF (Agilent Technologies, Santa Clara, CA, USA) equipped with a Dual Jet Stream™ (Agilent Technologies, Santa Clara, CA, USA) -ESI source. The instrument was operated in full scan and targeted MS/MS mode. The experiments were acquired in negative or positive, depending on the compound ionization.
The mass spectra were acquired over the m/z 100-1700 range with an acquisition rate of 2 spectra/s. The targeted MS/MS mode was applied over the m/z 50-1700 range with an MS scan rate at 10 spectra/s and an MS/MS scan rate at 3 spectra/s. Three fixed collision energies (20, 40 and 60 eV) were applied to the precursor ions to obtain an overview of the fragmentation pathways. The instrument control, data processing and analysis were conducted using Mass Hunter software (Agilent technologies, Santa Clara, CA, USA).
Author Contributions: Mohamed Laabir and Ons Kéfi-Daly Yahia designed the field survey and laboratory experiments; Hela Ben-Gharbia isolated the species, performed the morphology and the growth experiment; Zouher Amzil and Manoella Sibat performed the toxin analyses; Nicolas Chomérat performed the molecular analyses and phylogeny; Eric Abadie and Estelle Masseret contributed to correcting and improving the manuscript; Habiba Zmerli Triki contributed with materials/analysis tools; Habiba Nouri contributed to maps making; Hela Ben-Gharbia and Mohamed Laabir wrote the paper.

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

Abbreviation
The following abbreviations are used in this manuscript: Doubling time µmax Maximum growth rate S-PES Seawater with Provasoli's ES supplement S-ES-1 Seawater with ES-1supplements Appendix A. Table A1. Summary for Ostreopsis ovata established laboratory cultures from various marine ecosystems. Culture conditions, growth rates and toxicity are specified when available.