Metabolomic Profiles of Dinophysis acuminata and Dinophysis acuta Using Non-Targeted High-Resolution Mass Spectrometry: Effect of Nutritional Status and Prey

Photosynthetic species of the genus Dinophysis are obligate mixotrophs with temporary plastids (kleptoplastids) that are acquired from the ciliate Mesodinium rubrum, which feeds on cryptophytes of the Teleaulax-Plagioselmis-Geminigera clade. A metabolomic study of the three-species food chain Dinophysis-Mesodinium-Teleaulax was carried out using mass spectrometric analysis of extracts of batch-cultured cells of each level of that food chain. The main goal was to compare the metabolomic expression of Galician strains of Dinophysis acuminata and D. acuta that were subjected to different feeding regimes (well-fed and prey-limited) and feeding on two Mesodinium (Spanish and Danish) strains. Both Dinophysis species were able to grow while feeding on both Mesodinium strains, although differences in growth rates were observed. Toxin and metabolomic profiles of the two Dinophysis species were significantly different, and also varied between different feeding regimes and different prey organisms. Furthermore, significantly different metabolomes were expressed by a strain of D. acuminata that was feeding on different strains of the ciliate Mesodinium rubrum. Both species-specific metabolites and those common to D. acuminata and D. acuta were tentatively identified by screening of METLIN and Marine Natural Products Dictionary databases. This first metabolomic study applied to Dinophysis acuminata and D.acuta in culture establishes a basis for the chemical inventory of these species.


Introduction
Dinophysis species produce diarrhetic shellfish poisoning (DSP) toxins and pectenotoxins (PTXs) and pose a major concern to public health and the aquaculture industry in Western Europe [1,2]. Their impact is particularly serious in the Galician Rías Baixas (NW Spain), site of intensive mussel (Mytilus galloprovincialis) production (up to 3 × 10 5 metric tons per year; [3]). In this region, recurrent spring-summer proliferations of D. acuminata, followed in some years by those of D. acuta, cause lengthy harvesting bans whenever toxins in shellfish exceed regulatory levels [4]. Even though Dinophysis toxins (DTXs) emerged as a major risk since the 1970s, it was not until the recent cultivation of D. acuminata in the laboratory [5] that it became possible to undertake physiological studies of the
During the second phase of Experiment 2, ciliates were exhausted by day 7 in all the triplicate flasks, except those of D. acuta fed Danish M. rubrum ( Figure 2C,D). Results on specific growth rates and yields showed maximal values for D. acuminata fed Danish M. rubrum (µ0-8 = 0.25 day −1 ). Nevertheless, D. acuminata fed Spanish M. rubrum (µ0-10 = 0.08 day −1 ) and D. acuta fed Danish M. rubrum (µ4-8 = 0.08 day −1 ) had a very poor growth and D. acuta fed Spanish M. rubrum had hardly any growth (µ6-10 = 0.07 day −1 ). Thus, D. acuta did not seem to feed well on Danish ciliate prey and culture densities declined from day 10 onwards. The opposite was observed in D. acuminata, which reached a density of ~2200 cells mL −1 by the end of Experiment 2 ( Figure 2C).
During the second phase of Experiment 2, ciliates were exhausted by day 7 in all the triplicate flasks, except those of D. acuta fed Danish M. rubrum ( Figure 2C,D). Results on specific growth rates and yields showed maximal values for D. acuminata fed Danish M. rubrum (µ 0-8 = 0.25 day −1 ). Nevertheless, D. acuminata fed Spanish M. rubrum (µ 0-10 = 0.08 day −1 ) and D. acuta fed Danish M. rubrum (µ 4-8 = 0.08 day −1 ) had a very poor growth and D. acuta fed Spanish M. rubrum had hardly any growth (µ 6-10 = 0.07 day −1 ). Thus, D. acuta did not seem to feed well on Danish ciliate prey and culture densities declined from day 10 onwards. The opposite was observed in D. acuminata, which reached a density of~2200 cells mL −1 by the end of Experiment 2 ( Figure 2C).
The biomass of D. acuta and Danish M. rubrum harvested for toxins and other analyses on the second phase of Experiment 2 represented 81.6 ± 4.0% and 18.4 ± 4.0% of the total pellet biomass, respectively. In contrast, during the first phase of Experiment 2 these percentages were 93.5 ± 1.4% and 6.5 ± 1.4% for D. acuta and Danish M. rubrum, respectively, D. acuta being well fed in both situations (Table 1).

Mar. Drugs 2018, 16, x FOR PEER REVIEW 4 of 24
The biomass of D. acuta and Danish M. rubrum harvested for toxins and other analyses on the second phase of Experiment 2 represented 81.6 ± 4.0% and 18.4 ± 4.0% of the total pellet biomass, respectively. In contrast, during the first phase of Experiment 2 these percentages were 93.5 ± 1.4% and 6.5 ± 1.4% for D. acuta and Danish M. rubrum, respectively, D. acuta being well fed in both situations (Table 1).

Quantitative Toxin Analysis Using Low Resolution Mass Spectrometry
Dinophysis acuminata and D. acuta showed different toxin profiles. In both experiments, D. acuta produced OA, DTX2, and PTX2, whereas D. acuminata produced only OA (Table 2). Supernatants were not analyzed in this study and PTX2 derivatives could not be studied due to toxin degradation during the hydrolysis procedure [46]. Toxin quantification, pre-and post-hydrolysis, was used to indicate the potential presence of esters of the OA-group toxins in the samples from the two experiments.
During mid-exponential phase in Experiment 1, hydrolyzed samples of D. acuminata had 1.6 times more OA per cell than the non-hydrolyzed ones. Dinophysis acuminata contained 2.4 times more free OA (22 pg cell −1 ) and three-fold more total OA (35 pg cell −1 ) than D. acuta, the latter, however, contained a significant cell quota of PTX2. Results of a one-way analysis of variance (ANOVA) when comparing both triplicates were statistically significant (F(9.96) > 7.71; p < 0.05). In the case of D. acuta, total OA was 33% higher than free OA, whereas the amounts of DTX2 were nearly the same.
Estimated biovolumes of the different species were 65 µm 3 for Teleaulax amphioxeia, 2169 µm 3 and 2900 µm 3 for the Spanish and Danish Mesodinium, respectively, and 16,752 µm 3 and 54,333 µm 3 for D. acuminata and D. acuta respectively. Thus, the Danish Mesodinium was approximately 33% larger than the Spanish one, and D. acuta approximately three times larger than D. acuminata.
During the first phase of Experiment 2, D. acuta contained more OA per cell than D. acuminata. During the second phase of the experiment, cells from prey-limited cultures of D. acuta had more OA and DTX2 per cell than the well-fed cells and the same pattern was found for OA in D. acuminata cultures.  13; p < 0.05) for PTX2. Total OA content per cell in prey-limited D. acuminata cultures (close to stationary phase) were on average 3.6 (Danish Mesodinium) and 3.5 times (Spanish Mesodinium) higher than well-fed cells (closer to exponential growth phase). In fact, ANOVA results were only significant for OA in these two cases, being (F (55.53)>10.13; p < 0.05) and (F (59.59)>7.71; p < 0.05) for the Spanish and Danish M. rubrum feeding, respectively. In the case of total and free DTX2 in D. acuta, again results were significant only with D. acuta fed Spanish M. rubrum (F (17.13)>10.13; p < 0.05). When comparing hydrolyzed and non-hydrolyzed samples, ANOVA analysis was significant only for D. acuta well-fed with Spanish M. rubrum. Table 2. Toxin cell contents per biovolume and per cell (average ± standard deviation, n = 3) of D. acuminata and D. acuta cells from the baseline experiment (Experiment 1, mid-exponential growth phase) and from cultures under different nutritional status and prey (Experiment 2). Abbreviations: Sp = Species, PO = Prey origin, GP = Growth Phase, NS = Nutritional Status, ME = mid-exponential, ES = Spain, DK = Denmark.

Esters of OA-Group Toxins
The quantitative analysis using low resolution mass spectrometry suggested a large difference between total and free OA in D. acuta well-fed with Spanish Mesodinium. Indeed, spectra that were acquired from targeted high-resolution tandem mass spectrometry in both positive and negative ionization modes strongly suggested the presence of an OA ester, which had not been previously described in D. acuta [47].

Data Treatment for Non-Targeted Liquid Chromatography High Resolution Mass Spectrometry
Three variables were considered in the data analysis: (i) "species" (D. acuta, D. acuminata, M. rubrum and T. amphioxeia); (ii) "prey origin" (Danish and Spanish M. rubrum strain) and (iii) "nutritional status" (well-fed and prey-limited). Variable "species" was used in Experiment 1 to separate the response of different organisms in the principal component analysis (PCA): The X-axis explained the highest variability, i.e., 45.66% in ESI + mode (data not shown) and 57.08% in ESI − mode ( Figure 3). Esters of OA-Group Toxins The quantitative analysis using low resolution mass spectrometry suggested a large difference between total and free OA in D. acuta well-fed with Spanish Mesodinium. Indeed, spectra that were acquired from targeted high-resolution tandem mass spectrometry in both positive and negative ionization modes strongly suggested the presence of an OA ester, which had not been previously described in D. acuta [47].

Data Treatment for Non-Targeted Liquid Chromatography High Resolution Mass Spectrometry
Three variables were considered in the data analysis: (i) "species" (D. acuta, D. acuminata, M. rubrum and T. amphioxeia); (ii) "prey origin" (Danish and Spanish M. rubrum strain) and (iii) "nutritional status" (well-fed and prey-limited). Variable "species" was used in Experiment 1 to separate the response of different organisms in the principal component analysis (PCA): The X-axis explained the highest variability, i.e., 45.66% in ESI + mode (data not shown) and 57.08% in ESI − mode ( Figure 3). A similar response was observed using the same variable in Experiment 2 ( Figure 4). Species separation was clearly visible, with replicates of Mesodinium, Teleaulax, and the two species of Dinophysis forming separate clusters in principal component analysis. Intraspecific differences in each of the two Dinophysis species were larger than in Experiment 1, most likely due to the increased complexity of Experiment 2 since this one included treatments with different prey origins and nutritional status.
When Teleaulax and Mesodinium were excluded from the dataset, D. acuminata and D. acuta were clearly separated by the variable "species" (Figure 5A), also showing the highest number of species-specific compounds (p < 0.001; see Tables S3A-C and S4A-C for compounds specific to both Dinophysis species ranked by increasing p-values). This example also highlights the meaning of each principal component, as component 1 ( Figure 5A) corresponds to the separation according to species, whereas component 2 corresponds to grouping according to prey origin. As component 2 explains less than half of the variability as compared to component 1, the variable "species" can be considered most important, followed by prey origin. A similar response was observed using the same variable in Experiment 2 ( Figure 4). Species separation was clearly visible, with replicates of Mesodinium, Teleaulax, and the two species of Dinophysis forming separate clusters in principal component analysis. Intraspecific differences in each of the two Dinophysis species were larger than in Experiment 1, most likely due to the increased complexity of Experiment 2 since this one included treatments with different prey origins and nutritional status.
When Teleaulax and Mesodinium were excluded from the dataset, D. acuminata and D. acuta were clearly separated by the variable "species" (Figure 5A), also showing the highest number of species-specific compounds (p < 0.001; see Tables S3A-C and S4A-C for compounds specific to both Dinophysis species ranked by increasing p-values). This example also highlights the meaning of each principal component, as component 1 ( Figure 5A) corresponds to the separation according to species, whereas component 2 corresponds to grouping according to prey origin. As component 2 explains less than half of the variability as compared to component 1, the variable "species" can be considered most important, followed by prey origin.  In the case of D. acuminata, many compounds (107) were specific to one of the two nutritional conditions ( Figure 6). Still, taking the previous observations into account, only metabolites appearing under "species" and "prey origin" variables were considered in the following sections.
While the above analyses did not investigate the nature of the features that were detected in the non-targeted HRMS analysis, the following sections give an overview of the tentative identifications after comparisons with the databases METLIN and/or Marine Natural Products Dictionary.   In the case of D. acuminata, many compounds (107) were specific to one of the two nutritional conditions ( Figure 6). Still, taking the previous observations into account, only metabolites appearing under "species" and "prey origin" variables were considered in the following sections.
While the above analyses did not investigate the nature of the features that were detected in the non-targeted HRMS analysis, the following sections give an overview of the tentative identifications after comparisons with the databases METLIN and/or Marine Natural Products Dictionary.  In the case of D. acuminata, many compounds (107) were specific to one of the two nutritional conditions ( Figure 6). Still, taking the previous observations into account, only metabolites appearing under "species" and "prey origin" variables were considered in the following sections.
While the above analyses did not investigate the nature of the features that were detected in the non-targeted HRMS analysis, the following sections give an overview of the tentative identifications after comparisons with the databases METLIN and/or Marine Natural Products Dictionary. of Mesodinium used as prey are represented as triangles and rectangles for Denmark and Spain, respectively; (B) Principal component analysis of features detected in positive ionization mode in Experiment 2 (grouping by species and prey origin). First, second and third principal components (PC1, PC2, and PC3, the three components representing the largest fraction of the overall variability) are displayed on the X, Y, and Z-axis Variability explained by each component: PC1 (X-axis) 37.7%, PC2 (Y-axis) 26.7%, PC3 (Z-axis) 9.9% (total 74.3%). Nota bene: D. acuta samples in top half of the graph and D. acuminata samples in the bottom half of the graph, i.e., separation by species driven by the second component (Y-axis). "Species" and "prey origin" are represented in the figure as for A; (C) Venn diagram explaining overlap of features specific to species, prey-origin, and interactions of these parameters (see also Supplementary Table S4A-C).

Figure 6.
Venn diagram indicating the distribution of compounds specific to or common within groups of Dinophysis acuminata depending on "prey origin" and "nutritional status" from Experiment 2 (ESI − ).

Common to All Species
Dinophysis cells that were harvested from Experiment 1 had been previously washed to eliminate the remains of cryptophytes and M. rubrum from the cultures, but this step was not repeated for Experiment 2. For this reason, only compounds appearing in both experiments are shown in Table S1A,B.
Within the compounds common to all species, the one with the highest score was a molecule (Compound ID: S1A-3) which showed the maximal relative abundance in T. amphioxeia, decreasing progressively from M. rubrum to Dinophysis. When considering the normalized abundance of this compound in both experiments, its transfer through the food chain from T. amphioxeia to Dinophysis is plausible. This compound was previously found in the marine dinoflagellate Heterocapsa circularisquama and at a concentration ≥0.5 µg mL −1 it showed cytolytic activity toward oyster heart cells [48]. Another compound common to all species in both experiments was a galacto-glycerolipid-like compound (Compound ID: S1A-2 and S1B-8), in particular, a mono-galactosyl-diacyl-glycerol (MGDG 18:5), which had large quantities of polyunsaturated fatty acids. Mono-and di-galactosyl-diacylglycerols (MGDGs/DGDGs) are present in cyanobacteria and chloroplasts, and are used to preserve the Photosystem II (PSII) components. In higher plants, changes in the mono-and di-galactosyl-diacylglycerol ratios (MGDG/DGDG) are related to environmental changes that are affecting the structure of thylakoid membranes [49]. In fact, MGDGs with different structures (MGDG 20:5/16:3) appeared in Experiment 2 (Compounds ID: S1B-6, S1B-10, and S1B-11). All three compounds had the highest content in T. amphioxeia, and this relative content decreased from Mesodinium to Dinophysis (particularly in the case of D. acuta). Different structures of these lipids (MGDG and DGDG) have been suggested to be responses to changes in temperature [50]. Such responses were observed in different strains of Pyrocystis lunula, P. noctiluca, and P. fusiformis, in which DGDG (20:5/18:4) and DGDG (20:5/18:5) were the dominant galactolipids . Venn diagram indicating the distribution of compounds specific to or common within groups of Dinophysis acuminata depending on "prey origin" and "nutritional status" from Experiment 2 (ESI − ).

Common to All Species
Dinophysis cells that were harvested from Experiment 1 had been previously washed to eliminate the remains of cryptophytes and M. rubrum from the cultures, but this step was not repeated for Experiment 2. For this reason, only compounds appearing in both experiments are shown in Table S1A,B.
Within the compounds common to all species, the one with the highest score was a molecule (Compound ID: S1A-3) which showed the maximal relative abundance in T. amphioxeia, decreasing progressively from M. rubrum to Dinophysis. When considering the normalized abundance of this compound in both experiments, its transfer through the food chain from T. amphioxeia to Dinophysis is plausible. This compound was previously found in the marine dinoflagellate Heterocapsa circularisquama and at a concentration ≥0.5 µg mL −1 it showed cytolytic activity toward oyster heart cells [48]. Another compound common to all species in both experiments was a galacto-glycerolipid-like compound (Compound ID: S1A-2 and S1B-8), in particular, a mono-galactosyl-diacyl-glycerol (MGDG 18:5), which had large quantities of polyunsaturated fatty acids. Mono-and di-galactosyl-diacylglycerols (MGDGs/DGDGs) are present in cyanobacteria and chloroplasts, and are used to preserve the Photosystem II (PSII) components. In higher plants, changes in the mono-and di-galactosyl-diacylglycerol ratios (MGDG/DGDG) are related to environmental changes that are affecting the structure of thylakoid membranes [49]. In fact, MGDGs with different structures (MGDG 20:5/16:3) appeared in Experiment 2 (Compounds ID: S1B-6, S1B-10, and S1B-11). All three compounds had the highest content in T. amphioxeia, and this relative content decreased from Mesodinium to Dinophysis (particularly in the case of D. acuta). Different structures of these lipids (MGDG and DGDG) have been suggested to be responses to changes in temperature [50]. Such responses were observed in different strains of Pyrocystis lunula, P. noctiluca, and P. fusiformis, in which DGDG (20:5/18:4) and DGDG (20:5/18:5) were the dominant galactolipids in all the strains at temperatures of 25 • C and 15 • C, respectively, whereas MGMG (20:5/18:5) was predominant at 15 • C.
In our study, the presence of glycerolipids was generally reduced under prey-limited conditions, but these lipids did not seem to be specific to either D. acuminata or D. acuta.
Compound ID: S1B-4, which was a minor compound in both species of Dinophysis fed Danish M. rubrum (also present in all the species), was identified as a phosphatidyl glycerol (PG)-like molecule (through the METLIN database albeit with a low score of 53.04%). This phospholipid is an important component of thylakoid membranes [51], and its absence produces photosynthesis photo-inhibition [52]. After the description of the two previous compounds, it is worth mentioning that from algae to higher plants, P-limitation leads to an increased galactoglycerolipid/phosphoglycerolipid ratio [53,54]. However, the same compound was also identified as the methylester of melanodocin, an isomer of acanthifolicin (Marine Natural Products Dictionary, score of 99.3%). Acanthifolicin is a sulphide of okadaic acid, and this compound had not been reported from dinoflagellates but had been initially isolated from the sponge Halichondria melanodocia. However, the presence of this compound in all species, including Teleaulax amphioxeia, supports the hypothesis of a (PG)-like molecule. In any case, further studies are necessary to confirm the compound identity and gain new insights into its role.

Compounds Exclusive to the Ciliate Mesodinium
No compounds that were specific to the M. rubrum strains were found. This result may well be related to some residual contamination from T. amphioxeia in ciliate cultures, but also to the fact that Mesodinium keeps the full set of cryptophyte organelles (nucleus, mitochondria, chloroplast) and the cryptophyte transcriptional machinery active after sequestration [26,55].

Compounds Exclusive to Teleaulax amphioxeia
Two compounds that were specific to the cryptophyte T. amphioxeia (Table S2, compounds ID: S2-1 and S2-2) were found. One of them (Compound ID: S2-2) matched the group of phospholipids, which form part of eukaryotic cell membranes with the suggested function of maintaining their fluidity [56].

Compounds Exclusive to One of the Two Species of Dinophysis
A glycerol-like compound (Compound ID: S7A-1) was identified only in D. acuminata from Experiment 2 (Table S7A). This compound is required for the synthesis of triacylglycerols, phospholipids, and glycolipids. Amongst glycolipids, glycosphingolipids act as cell adhesion mediators and signal transduction modulators. When they are associated with other proteins (glycosynapse), they act as functional groups through which glycosylation-dependent cell adhesion coupled with signal transduction takes place [57]. These glycosphingolipids, together with other polysaccharides, are involved in membrane formation and were found to be overexpressed during bloom conditions in field populations of Levanderina fissa [58]. Stigmastane-like compounds (Compounds ID: S10-3 and S10-4) were only present in well-fed cells of D. acuminata (fed Danish M. rubrum; Table S10). The cellular content of this kind of sterols was found to increase in Gymnodinium sp. cultures when they reached the stationary phase [59]. Maximum levels of other sterol lipids, e.g., compounds ID: S11B-1 and S11B-4, were found in starved cells of D. acuminata (Danish M. rubrum) (Table S11B).
Protochlorophyllide (Compound ID: S11B-9), which is an ester-like molecule that is involved in the biochemical route for the synthesis of chlorophyll compounds [60], was only found in prey-limited cells of D. acuminata initially fed Danish M. rubrum (Table S11B), together with a phaeophytin a-like compound (Compound ID: S11B-5; chlorophyll a degradation product).

Compounds First Found in Other Marine/Terrestrial Organisms
Many of the compounds that were found in either or both Dinophysis species were tentatively identified as polyketides. Polyketides are a structurally diverse group that is constituted by a carbon chain synthesized by a family of enzymes called polyketide synthases (PKSs). Genes coding for PKSswere first cloned and sequenced in bacteria and fungi, and later identified and associated with the dinoflagellate Karenia brevis by Snyder et al. [61,62].
Compounds Exclusive to One of the Two Species of Dinophysis

Dinophysis acuminata
Several compounds previously isolated from sponges were tentatively identified in D. acuminata (Table S7B, Compound ID: S7B-1), but were not related with a particular "prey-origin" nor "nutritional-status" condition. Other compounds that were first isolated also in sponges and skates (genus Raja) were identified in well-fed cells of this species (Compound ID: S10-5), regardless of the Mesodinium strain that was used as prey (Table S10). Different alkaloid-like compounds previously found in starfish and sponges were found in food-limited D. acuminata cells (Table S11A,B, compound ID: S11B-11).

Dinophysis acuta
Brevetoxins and OA-related toxins may be classed as polyketides [65]. While OA was present in both D. acuta and D. acuminata, a peak of an isomer appeared just a few seconds after OA in D. acuta only, which corresponded to dinophysistoxin-2 (DTX2). Dinophysistoxin 1 (DTX1) and DTX2 often co-occur with OA, but the two DTXs rarely co-occur. This is probably due to the fact that DTX2 has a stereochemistry opposite to that of DTX1 at their C-35 [66], which suggests a significant difference in the biosynthetic pathway.
In agreement with results from targeted analysis by low resolution mass spectrometry (Table 2), PTX2 was also specific to D. acuta in the non-targeted analysis (Table S6B, compound ID: S6B-5). The abundance of PTX2 was slightly higher in prey-limited than in well-fed cells, in particular, for the cells that were fed Spanish M. rubrum.
A compound (Compound ID: S9A-4) that was tentatively identified as Prorocentin was the most abundant of all the compounds specific to prey-limited D. acuta fed Danish M. rubrum. Prorocentin, which was first isolated from the toxic dinoflagellate Prorocentrum lima [67], is a polyketide suggested to share a biosynthethic pathway with OA.
A compound originally described in sharks (Compound ID: S8-1) was found in well-fed (Danish M. rubrum) cells of D. acuta (Table S8). However, when food was limiting, a putative anti-feedant compound (Compound ID: S9A-1) that was initially isolated from the cyanobacteria Lyngbya majuscula [70], was identified together with other compounds firstly identified in sponges, molluscs, and bacteria (Table S9A,B).

Growth Curves
Growth rates that were obtained in Experiment 2 for both Dinophysis species demonstrated the significant effect of prey source (Spanish or Danish M. rubrum strain). Interestingly, D. acuminata showed the highest growth rates under both nutritional conditions when it was initially fed Danish M. rubrum. Only 0.03% of genetic dissimilarity was observed when comparing both Mesodinium strains, but the biovolume of the Danish strain was 33% larger than that of the Spanish one. In this regard, it is well known that prey size and morphological features affect the selectivity of predators towards specific prey organisms. This phenomenon has previously been demonstrated for the heterotrophic dinoflagellates Oxyrrhis marina, Protoperidinium pallidum, and P. steinii: when they were fed algae with different sizes, their growth was mainly supported by the larger-sized specimens [71,72].

Toxin Quota and Target Analysis
Toxin content per cell of all toxins was higher in Dinophysis cultures during the prey-limited phase (Experiment 2), except for PTX2 in D. acuta fed Danish ciliate, which was probably related to the poor physiological state of that culture ( Table 2). Earlier experiments with D. acuminata [73] with D. fortii [74] showed that prey limitation reduced its growth and increased toxin content per cell due to an imbalance between cell division and toxin production rates. The different results that were observed for D. acuta are in accordance with the observations on its feeding behavior: D. acuta did not feed well on the Danish ciliate, a fact that would also explain the lack of differences between well-fed and prey limited phases ( Table 2, Experiment 2). In contrast, D. acuta reached the highest content of OA per cell when feeding on Spanish Mesodinium (Table 2, Experiment 2).
Toxin content per cell observed in both Dinophysis species during Experiments 1 and 2 (D. acuta: 70 pg PTX2 and 75 pg OA cell −1 ; D. acuminata: 32 pg OA cell −1 in hydrolyzed samples) was significantly higher than any ever-recorded in picked cells of D. acuminata and D. acuta strains from the Galician Rías. Earlier studies on Dinophysis strains from this region showed maximum values of 9.4, 6.6 and 6.1 pg cell −1 of OA, DTX2, and PTX2, respectively [75], and a simpler profile with only PTX2 (32.3 pg PTX2 cell −1 ) in D. acuta from another year [76].
Toxin analysis that was carried out during the baseline experiment (Experiment 1) showed the same toxins as in Experiment 2, but their proportions were somewhat different and the cell-toxin quota was lower in Experiment 1. Thus, PTX2 and total DTX2 toxin content in D. acuta was only 22 and 4 pg cell −1 , respectively, in Experiment 1, while the largest amount of these toxins were 70 and 30 pg cell −1 , respectively, during parts 1 and 2 of Experiment 2. During Experiment 1, OA in D. acuminata reached the highest content per cell (~21.5 pg OA cell −1 ), similar to those that were acquired during Experiment 2. Contrarily, inExperiment 2, it was D. acuta that reached the highest content of OA per cell (~76 pg OA cell −1 ), while during Experiment 1 OA toxin content per cell was found in a minor quantity (~9 pg OA cell −1 ). The time lag between Experiments 1 and 2 was approximately three months, so it is unlikely that toxin loss occurred in D. acuminata cultures, as has been observed with other dinoflagellate cultures. As an example, the losses of toxin content in a strain of Alexandrium lusitanicum (=A. minutum) have been related to slower growth in the laboratory [77].

Semi-Quantitative Non-Targeted Analysis: Comparison of Larger Parts of Metabolomes
A multivariate analysis was used to estimate differences between phenotypes, at the chemical level, in relation to three variables: "species", "prey origin", and "nutritional status". The variable "species" led to the highest separation within samples, followed by "prey origin". The "nutritional status" did not provide clear-cut results in the case of D. acuta, neither during the growth experiments nor in the PCA analysis. In contrast, in the case of D. acuminata, distinct results were obtained depending on the "nutritional status". Statistical analysis that was carried out here gave enough separation between the organisms, however future studies should include partial least square-discriminant (PLS-DA) analysis and Variable Importance in Projection VIP score plot, in addition to PCA to underscore such separation.
Only a few lipid-derivatives could be tentatively identified in every species in the present study and were likely transferred through the three-species food chain (Teleaulax-Mesodinium-Dinophysis). Metabolomic studies aim to unveil a comprehensive view of the metabolome, i.e., the set of metabolites a biological entity may produce, although datasets in this field can be convoluted [78]. This part of the study would fit in the field of partial ecometabolomic studies (PEM), a term coined by Sardans et al. [79], which encompasses ecophysiological metabolomic studies focused on the identification of metabolites that are produced in response to specific biotic/abiotic effects on organisms. Formally speaking about metabolomics, intra-and extracellular metabolites should be analyzed [80]. The main objective of this work was to identify intracellular compounds (whether by formulae or by name) that were (i) common to all the species that were used in this study (T. amphioxeia, M. rubrum, Dinophysis species); (ii) specific to some of them; or, (iii) species-specific. It was not intended to be a comparison between both batches of samples (Experiments 1 and 2) as these types of analysis are "snapshots" of the cell at the time of being harvested, and they were, indeed, harvested at different times. Due to the diversity of the metabolome, it is not possible to identify all metabolite types [33]. This study represents a first application of these next-generation technologies to Dinophysis, and should be useful, among others, to identify molecular markers that are involved in predator-prey recognition and/or Dinophysis identification.
Regarding biological results, maximum yields were obtained in the D. acuminata strain fed Danish M. rubrum. Contrary effects were seen when the population of D. acuta, feeding on the same prey, declined during the second part of Experiment 2. Therefore, important questions regarding Experiment 2 were: could such compounds be involved in predator-prey recognition? If so, do these compounds belong to the Danish M. rubrum or are specific for D. acuminata or D. acuta?
Different alkaloids and lipids were mainly identified in prey-limited cells of D. acuminata that were initially fed Danish M. rubrum. In the case of D. acuta, the putative characterization of a compound with anti-feeding properties (Compound ID: S9A-1) could be associated with the shift to early stationary and senescent phases in this species. In fact, the release of secondary metabolites by dinoflagellates, and their effects have gained much attention during the last years. The term "allelopathy" can be understood as an adaptation [81] related to the secretion of secondary metabolites and other biochemicals [82]. Particularly, the lipopeptide Malyngamide A was shown to have anti-feeding properties that deterred the uptake by two species of coral red fishes of artificial food (mixture of water, agar, and powdered green alga) containing the lipopeptide [83]. Many identified features were first isolated in higher marine organisms implying roles in various ecological relationships. Okadaic acid was first isolated from the sponges Halichondria okadai and H. melanodocia [36], so it could well be that compounds found in other marine organisms in fact were derived from smaller-sized cells. Toxins from certain strains of the dinoflagellate Karlodinium veneficum displayed anti-feeding properties against the copepod Acantia atonsa [84]. In this context, Denticulatin A, which has been tentatively identified in both Dinophysis species (Compound ID: S4A-38), can be mentioned as a possible ichtyotoxic agent due to its polypropionate character.
Regarding the issue of how Dinophysis can recognize the ciliate M. rubrum as a prey, different glycolipids (e.g., Glycerol 1, 2-dialkanoates in D. acuminata) and other compounds identified only tentatively by their formulae were found. Raho et al. [13] used different lectin markers to detect the presence of cell-surface binding carbohydrates in Galician strains of D. acuminata and D. acuta. Such lectin markers showed positive results in both Dinophysis as well as in M. rubrum cells, suggesting that lectins could play a key role in predator-prey recognition. Wood-Charlson et al. [85] showed that the cell surface of Symbiodinium contained glycan ligands (α-Mannose/α-Glucose and α-Galactose residues), which bound with two different lectins (ConA and Jac) that were found in Fungia scutaria coral larvae, meaning that some recognition mechanism is playing an important role in the coral/dinoflagellate symbiosis. Several recent studies pointed out the excretion by Dinophysis of a mucus trap to catch its prey [86][87][88]. Mafra et al. [86] demonstrated that the "toxic substance" in the mucus was not a lipophilic toxin. These authors suggested that excreted toxic substances, together with the mucus, were part of Dinophysis feeding strategies; however, prey capture by Dinophysis using a mucus trap was not observed in the present study. Finally, we repeat that all the identifications in this study must be considered as tentative, since no reference compounds are available for many natural products, and full scan high resolution mass spectrometry only allows for an initial comparison with existing databases, but without complete structural confirmation. Further studies need to be carried out to follow up on these results and confirm the actual identity of the metabolites.

Cultures
Dinophysis acuminata (strain VGO1349) was isolated from Ría de Vigo in July 2016 and Dinophysis acuta (VGO1065) from Ría de Pontevedra in October 2010. A strain of the ciliate Mesodinium rubrum (AND-A0711, Acc. Number KP142651) and the cryptophyte Teleaulax amphioxeia (AND-A0710; Acc. Number KP142646) were isolated from Huelva, Southwest Spain, in 2007. Cultures of M. rubrum, fed the cryptophyte T. amphioxeia, were periodically given to Dinophysis as prey. An additional strain of M. rubrum (MrDK-2009, Acc. Number MG018339), which was isolated from Helsingør Harbor (Denmark) during summer 2009, was used in this study. Images of these organisms are shown in Figure 7.
together with the mucus, were part of Dinophysis feeding strategies; however, prey capture by Dinophysis using a mucus trap was not observed in the present study. Finally, we repeat that all the identifications in this study must be considered as tentative, since no reference compounds are available for many natural products, and full scan high resolution mass spectrometry only allows for an initial comparison with existing databases, but without complete structural confirmation. Further studies need to be carried out to follow up on these results and confirm the actual identity of the metabolites.

Cultures
Dinophysis acuminata (strain VGO1349) was isolated from Ría de Vigo in July 2016 and

DNA Extraction, Polymerase Chain Reaction (PCR), 18S Sequencing and Phylogenetic Analysis of Mesodinium
A culture of M. rubrum (Mr-DK2009) fed the cryptophyte Teleaulax acuta was starved for two weeks until no prey could be detected. Single cells were picked and transferred to 0.2 mL PCR tubes containing 100 µL of milli-Q water and 10% (w/v) Chelex 100 (Sigma-Aldrich #C7901, St. Louis, MO, USA). For DNA extraction, the PCR tubes were vortexed for 5 s, spun down in a microcentrifuge for 10 s, and subsequently incubated at 95 °C for 20 min [89]. After incubation, the tubes were centrifuged again for 10 s and stored at 4 °C until use for PCR reactions.
PCR products were tested on a 2% agarose gel and purified with a QIAquick PCR purification kit (Qiagen #28106). Purified samples were sent to Macrogen (Macrogen Europe, Amsterdam, The

DNA Extraction, Polymerase Chain Reaction (PCR), 18S Sequencing and Phylogenetic Analysis of Mesodinium
A culture of M. rubrum (Mr-DK2009) fed the cryptophyte Teleaulax acuta was starved for two weeks until no prey could be detected. Single cells were picked and transferred to 0.2 mL PCR tubes containing 100 µL of milli-Q water and 10% (w/v) Chelex 100 (Sigma-Aldrich #C7901, St. Louis, MO, USA). For DNA extraction, the PCR tubes were vortexed for 5 s, spun down in a microcentrifuge for 10 s, and subsequently incubated at 95 • C for 20 min [89]. After incubation, the tubes were centrifuged again for 10 s and stored at 4 • C until use for PCR reactions.
PCR products were tested on a 2% agarose gel and purified with a QIAquick PCR purification kit (Qiagen #28106). Purified samples were sent to Macrogen (Macrogen Europe, Amsterdam, The Netherlands) for Sanger sequencing in both directions. Sequence analysis (trimming, assembly, BLAST) was done with Geneious version 10.1.3. For the phylogenetic analysis, additional sequences of the SSU rRNA gene were downloaded from GenBank and were aligned using T-Coffee, as implemented by intcoffe.crg.cat. The alignment included 2780 characters and was uploaded to the South of France bioinformatics platform for PhyML 3.0 analysis with Smart Model Selection (best model was GTR + G + I), using the Akaike Information Criterion and performing 1000 bootstrap replicates [90,91]. Bayesian Inference was performed with MrBayes 3.2.6 using a GTR + I + Γ model, as implemented in Geneious ® 10.2.2 (Biomatters Ltd., Auckland, New Zealand) [92]. The following settings were used: four simultaneous Markov chain Monte Carlo (MCMC) run for 1,000,000 generations, sampling every 100 generations. The first 25% of trees were discarded as burn in. Finally, a neighbor-joining tree was build, using the Jukes-Cantor genetic distance model and 10,000 bootstrap replicates, as implemented in Geneious ® 10.2.2.

Biovolume
Measurements of the organisms that were studied (cryptophytes, Mesodinium, and Dinophysis) were carried out to estimate their biovolumes. These estimates were used to normalize data related to the identified features (for "feature" meaning explanation see below at Section 4.7.1). Imaging of 30 specimens of each Dinophysis species was carried out with an Axiocam HRC digital camera (Zeiss, Oberkochen, Germany) and measurements that were taken under a light microscope (Leica DMR, Germany) at 630 X magnification. Dinophysis acuta geometric shape was considered equivalent to a cone plus a truncated cone and D. acuminata to a flattened ellipsoid. Biovolume estimates were based on the equations used in Olenina et al. [98]. Biovolumes of the two M. rubrum strains and T. amphioxeia were estimated with a particle counter (Multisizer 3 Coulter counter, Beckman, Roissy Charles De Gaulle, France) using triplicates of 10 mL for each species.

Experiment 1 (Baseline Study)
A preliminary experiment was carried out at IEO-Vigo to gain baseline information on the growth curves and metabolomes of the organisms. Triplicates of Dinophysis acuminata, D. acuta, M. rubrum, and T. amphioxeia where grown in 250 mL flasks filled with diluted (1:20) L1-Si medium [99] based on autoclaved seawater from Ría de Vigo with the salinity adjusted to 32 and a 12:12 h L:D cycle. Cultures were provided a light intensity of~250 µE m −2 s −1 and were kept in a temperature-controlled chamber at 15 • C. Mesodinium rubrum was fed T. amphioxeia at a 1:1 ratio (Teleaulax:Mesodinium) and Dinophysis acuminata and D. acuta were fed the Spanish M. rubrum at a 10:1 ratio (Mesodinium:Dinophysis). More prey was added to Mesodinium when no cryptophytes were observed under the light microscope. To carry out toxin and non-targeted analyses, cultures that were harvested at mid-exponential phase were centrifuged at 3000× g for 20 min at 4 • C, the supernatant removed and the pellets kept at −80 • C before being sent on dry ice to IFREMER-Nantes.

Experiment 2 (Growth Curves)
Experiment 2 was carried out at IFREMER-Nantes and was developed in two phases, the first one lasting 7 days and the second 15 days, under identical experimental conditions. For the first phase The two strains of M. rubrum were provided separately as prey for both Dinophysis species with a ratio 10:1 (M. rubrum:Dinophysis) on day 1. These were labeled as "well fed" treatments. The second phase (15 days) was carried out as previously described. The only difference was the longer duration of the second experiment (15 days), and that Dinophysis was fed only on day 0, and was therefore considered "prey-limited" at day 15 (starvation would have required being kept for weeks with no prey).
To estimate cell densities, samples of 10 mL (for Coulter counter) and 2 mL (sedimentation chambers) were taken every 2 days and were fixed with acidic Lugol's solution. Dinophysis cells were counted in sedimentation chambers under a Zeiss Invertoskop D microscope (Carl Zeiss AG, Oberkochen, Germany), Mesodinium with a Coulter counter, and T. amphioxeia, with a Nageotte hemocytometer counting chamber.
Specific growth rates (µ, day −1 ) of Dinophysis species were calculated as: where N o was the initial and N t the final density of Dinophysis after time t.

Toxin Analysis
All the toxin analyses were performed at the Phycotoxins Laboratory (IFREMER-Nantes).

Extraction and Hydrolysis
Samples from the first phase of Experiment 2 were collected during the exponential growth phase of each species, i.e., T. amphioxeia on day 3 and the M. rubrum controls and Dinophysis species on day 7, in Falcon tubes of 15 or 50 mL and centrifuged (Sigma 3-18K, Fisher Bioblock Scientific, France), as in Experiment 1. Pellets from Experiment 1 and 2 were extracted twice, resuspended in 0.5 mL MeOH, vortexed, and sonicated at 45 KHz for 15 min. Then, 200 mg of 200 µm glass beads were added and the mixture placed in a Mixer Mill MM 400 (Retsch GmbH, Haan, Germany) for 30 min to disrupt the cells, centrifuged at 12,000× g and 4 • C for 10 min and the supernatant (500 µL) filtered using 1.5 mL Eppendorf tubes with a 0.2 µm mesh included.For the second part of Experiment 2, extractions of each Dinophysis species triplicates that were fed each M. rubrum were carried out. Cultures of M. rubrum strains and the cryptophyte on day 12 were already senescent and discarded. For quantification, a basic hydrolysis of the Dinophysis and M. rubrum pellets from both experiments was carried out to convert esters of the main toxin (OA, DTX, and PTX) into their parent toxins (free toxins) [100]). For this purpose, 250 µL of methanol extracts of the samples were mixed with 30 µL of NaOH in hermetic-closing opaque-glass tubes and heated at 75 • C for 40 min. Samples were then left to cool down, their volumes verified and 30 µL of HCl added before being stirred, and then filtered through a 0.2 µm microfilter into 1.5 mL Eppendorf tubes, centrifuged (12,000× g, 10 min, 4 • C), and transferred to vials before injection.

High Resolution Mass Spectral Analysis-System A: (QTOF 6550)
UPLC-HRMS analyses were carried out with a UHPLC system (1290 Infinity II, Agilent technologies, Santa Clara, CA, USA) coupled to a high resolution time-of-flight mass spectrometer (Q-Tof 6550 iFunnel, Agilent technologies, CA, USA), equipped with a Dual Jet Stream ® electrospray ionization (ESI) interface operating in both negative and positive mode in separate runs. Chromatographic separation was carried out on a reversed-phase C 18 Kinetex column (100 Å, 2.6 µm, 50 × 2.1 mm, Phenomenex, LePecq, France) at 40 • C using a mobile phase that was composed of water (A) and 95% acetonitrile/water (B) both containing 5 mM ammonium formate and 50 mM formic acid. The flow rate was set at 0.4 mL min −1 and the injection volume 3 µL. Separation was achieved using the following mobile phase gradient: from 10 to 50% B in 2 min, to 90% B over the next 3 min, held for 5 min before return to the initial condition (10% B) in 0.5 min, and a re-equilibration period (10% B) for 5.0 min.
Mass spectral detection was carried out in full scan and targeted MS/MS mode in negative (ESI − ) and positive (ESI + ) ion acquisition. The full scan acquisition operated at a mass resolution of 40,000 Full Width at Half Maximum (FWHM) over a mass-to-charge ratio (m/z) range from 100 to 1700 with a scan rate of 2 spectra s −1 . The targeted MS/MS mode was performed in a Collision Induced Dissociation (CID) cell using a mass resolving power of 40,000 FWHM over the scan range m/z from 50 to 1700 with a MS scan rate of 10 spectra s −1 and a MS/MS scan rate of 3 spectra s −1 . Three different collision energies (i.e., 30, 50, and 70 eV) were applied to the precursor ions to obtain adequate fragmentation.
Acquisition (see Table 4) was controlled by Mass Hunter software (Agilent Technologies, CA, USA). Raw data were processed using the Molecular Feature Extraction (MFE) algorithm of the Agilent Mass Hunter Qualitative Analysis software (version B.07.00, service pack 1).
Toxins were separated using the same chromatographic conditions, as described above (system A). The injection volume was set at 5 µL. For quantitation, the mass spectrometer was operated in multiple reactions monitoring (MRM) acquisition mode, scanning two transitions for each toxin. Positive and negative acquisition experiments were established using the following source settings: curtain gas set at 30 psi (ESI + ) and 20 psi (ESI − ), ion spray at 5500 V (ESI + ), and −4500 V (ESI − ), Turbogas temperature of 550 • C, gas 1 and 2 set, respectively, at 40 and 50 psi, and an entrance potential of 10 V. Pectenotoxins (PTX) were analyzed in positive mode and OA group toxins in negative ionization mode (Table 4). Data acquisition and quantification were carried out using Analyst software version 1.6.2 (SCIEX, Redwood City, CA, USA). ANOVA was carried out at the end of the experiments to evaluate the statistical significance of toxin content per cell in Experiment 2.
Certified calibration solutions were purchased from the National Research Council Canada (NRCC, Halifax, NS, Canada): PTX2, OA, DTX1, and -2. HPLC grade methanol, acetonitrile, and formic acid (98%) were obtained from Sigma Aldrich (Steinheim, Germany). Ammonium formate was acquired from Fluka (St. Louis, MO, USA). Milli-Q water was obtained in-house to 18 MΩ cm −1 quality, using a Milli-Q integral 3 system (Merck Millipore, Guyancourt, France). For HRMS, acetonitrile and high purity water were obtained from Fisher Scientific (Illkirch, France). Features are distinct ion clusters, belonging either to compounds or to provisionally identified metabolites, which possess an abundance, a mass, a retention time, and, if present in available databases, a score thatallows their qualitative identification and assignment of a sum formula. Features were extracted from raw data files by a molecular feature extraction (MFE) algorithm using a Mass Hunter Workstation Qualitative Analysis Software (Agilent Technologies, Santa Clara, CA, USA) to create a work-flow to subtract procedural blanks in batches of samples (ESI + and ESI − ionization modes in extracted and hydrolyzed samples). An abundance cut-off value was selected so that features with peak heights of 5000 or higher were transformed into compound exchange format (.cef) files with the Mass Hunter Reprocessor, and were subsequently analyzed with Mass Profiler Professional software (MPP, version B 13.1.1) to identify statistically meaningful differences of features between groups. Compound alignment parameters were: RT = 0.0% + 0.15 min and mass = 15.0 ppm + 2.0 mDa.
Resulting abundance data were normalized using external scalars of each triplicate sample group. After this normalization, organisms were treated as equal-sized organisms and size effects on the results were avoided (e.g., compounds abundance on the smallest organism-in this study, Teleaulax amphioxeia-would be too small to allow for a comparison with the largest organism, Dinophysis acuta in this case. External scalars were calculated, as follows: where B is the biovolume of species X normalized in relation to that of Teleaulax, A is the normalized biovolume of the total number of cells in the sample (N tot ), S is the external scalar that was used for abundance data extracted from LC-HRMS. Compounds satisfying a fold change cut-off of 2.0 and a 100% of frequency in each sample group triplicate were used to identify common and unique features within or between groups.
Sample quality assurance was tested by Principal Component Analysis (PCA). Significance testing was carried out using one-way and two-way ANOVA, an asymptotic p-value computation, a multiple testing correction of Benjamini-Hochberg, and a p-value cut-off of 0.001.
Similar and specific compounds were selected manually and verified with the Find Similar Entities tool displaying expression profiles of entities whose correlation coefficients to the target profiles were above the correlation cut-off range (0.5-0.85, 1.0) using an Euclidean similarity measure. P-values for "in both positive and negative ESI modes" were calculated with MPP, according to the variables "origin" (Spain and Denmark) and "feeding-status" (well-fed and prey-limited) for all the features in the samples. Samples list and grouping following the foregoing variables are shown in Table S12. Features were grouped into one variable depending on their distribution pattern. For features that were fulfilling the same conditions as above, only p-values of <0.001 were used.
Resulting compounds were identified with the Agilent Mass Hunter ID Browser B.07.00 against databases (METLIN and the Dictionary of Marine Natural Products) with a mass-match tolerance of 2.00 ppm, and formulae were generated with the Molecular Formula Generator (MFG). Compound identification default settings were as follows: mass and retention time (RT) score were 100.00; isotope abundance and spacing scores were 80.00. Selected expected data variations were: 1.0 mDa + 2.0 ppm for MS mass, MS isotope abundance of 7.5%, MS/MS mass of 5.0 mDa + 7.5 ppm, and a RT of 0.115 min. For positive ions, neutral water losses and H + , Na + ; K + , and NH 4 + adducts were selected as charge carriers, whereas for negative ions, neutral water losses, and H − , Cl − , Br − , HCOO − , CH3COO − , and CF3COO − were selected as charge carriers. Compounds were clustered once they were identified following (a) "physiological meaning": common and specific compounds of all and each organism, respectively, and (b) "compounds first found in other marine/terrestrial organisms" that were identified either in both or particularly in one of the Dinophysis species.

Conclusions
Metabolomic profiling allowed for us to differentiate Dinophysis species and treatments ("prey origin" and "nutritional status", especially in D. acuminata), and some of the compounds that were involved were tentatively identified. Non-targeted analyses serve as a powerful screening tool and led to the identification of a new diol ester in D. acuta. The variable "species" gave the highest separation between groups of D. acuminata and D. acuta, followed by the variable "prey origin" (Danish and Spanish M. rubrum strain). Future studies should help in building up a "Dinophysis-metabolites library" initiated by this study, to confirm or to reject some of the hypotheses that are formulated in this paper. This library would provide valuable information regarding (i) metabolic fingerprints and pathways; (ii) predator-prey-recognition mechanisms related to target metabolites; and, (iii) clarify biogeographical distributions as well as the biological origin of compounds often identified first in higher, filter-feeding marine organisms, such as sponges.