Differentiating Two Closely Related Alexandrium Species Using Comparative Quantitative Proteomics

Alexandrium minutum and Alexandrium tamutum are two closely related harmful algal bloom (HAB)-causing species with different toxicity. Using isobaric tags for relative and absolute quantitation (iTRAQ)-based quantitative proteomics and two-dimensional differential gel electrophoresis (2D-DIGE), a comprehensive characterization of the proteomes of A. minutum and A. tamutum was performed to identify the cellular and molecular underpinnings for the dissimilarity between these two species. A total of 1436 proteins and 420 protein spots were identified using iTRAQ-based proteomics and 2D-DIGE, respectively. Both methods revealed little difference (10–12%) between the proteomes of A. minutum and A. tamutum, highlighting that these organisms follow similar cellular and biological processes at the exponential stage. Toxin biosynthetic enzymes were present in both organisms. However, the gonyautoxin-producing A. minutum showed higher levels of osmotic growth proteins, Zn-dependent alcohol dehydrogenase and type-I polyketide synthase compared to the non-toxic A. tamutum. Further, A. tamutum had increased S-adenosylmethionine transferase that may potentially have a negative feedback mechanism to toxin biosynthesis. The complementary proteomics approach provided insights into the biochemistry of these two closely related HAB-causing organisms. The identified proteins are potential biomarkers for organismal toxicity and could be explored for environmental monitoring.


Introduction
The genus Alexandrium is one of the major harmful algal bloom (HAB)-causing genera in terms of global diversity, magnitude and consequences of blooms [1][2][3]. The proliferation of this dinoflagellate in the marine environment causes deleterious effects to the economy and human health. Most Alexandrium species produce Paralytic Shellfish Toxins (PSTs) such as saxitoxin and analogues which can cause paralysis and eventual death [4].
Alexandrium minutum Halim 1960 and Alexandrium tamutum Montresor 2004 are closely related species. Morphologically, both A. minutum and A. tamutum are oval to elliptical in shape, with the latter slightly pentagonal in some cases. In the cyst stage, A. minutum is hemispherical in shape whereas A. tamutum is elliptical (Figure 1).
A. tamutum and A. minutum have comparable cell length between 19-34 µm and 15.5-29 µm, respectively [5,6]. The main diagnostic character between A. minutum and A. tamutum is that the latter possesses a relatively wide and large sixth precingular plate (6") [7,8]. This is in contrast with A. minutum which has a narrower and smaller sixth precingular plate [6]. The extremely similar size of A. tamutum and A. minutum often leads to difficulty in cell counting and morphological differentiation. Negative control for A. minutum (C) and A. tamutum (D) after the preparative stages (pigment removal using methanol). Immunostaining of A. tamutum (E,F) and A. minutum (G,H). Scale bar: 10 µm at 630× magnification. (I) Quantitation of fluorescence intensity for A. tamutum and A. minutum and negative controls. Immunostaining with primary antibody for A. tamutum and A. minutum. A significant difference in fluorescence was observed (** p < 0.01), with A. minutum having the highest fluorescence signal when bound to the A. minutum-specific primary antibody. Data is presented as mean ± SD (n = 10) and analyzed using ANOVA with post Tukey's HSD Test (** p < 0.01).
Genomics, transcriptomics and proteomics have been utilized for differentiating species, elucidating molecular mechanisms and understanding cellular dynamics among HAB-causing organisms [20]. However, the large genome size and permanently condensed chromosomes of dinoflagellates has impeded the use of genomics for molecular analysis [21]. Furthermore, transcriptomics has the limitation of providing little information on protein activity and fitness for functional and ecological analysis since not all transcripts are translated into functional proteins [22]. Owing to these, proteomics presents an alternative approach to reveal great insights on the various cellular processes among dinoflagellates [23,24]. Dinoflagellate proteomics have been employed to study cell growth and regulation [25], response to environmental stresses [26,27], toxin biosynthesis [2,28], species identification and evolution [29], cell wall and cell surface protein identification [30,31] and symbiotic relationship among these organisms [32]. HAB proteomics has led to the discovery of biomarkers of toxicity [33] and development of antibody-based species identification [34] for HAB species identification and monitoring.
In this work, we differentiated A. minutum and A. tamutum based on their toxin, immunostaining and proteome profiles. Proteomic studies of these two organisms using gelbased (two-dimensional difference gel electrophoresis or 2D-DIGE) and gel-free approaches (isobaric tags for relative and absolute quantitation or iTRAQ) were complementarily used and provided insights on the biochemical machineries of these related organisms.

Toxin Analysis
PSTs were quantified using high performance liquid chromatography (HPLC) coupled to fluorescence detection of the oxidation products. From the toxin analysis, A. minutum contained 14.05-19.50 fmol STX equiv/cell, with GTX1,4 and GTX2,3 as the primary detectable toxins. On the other hand, no toxin peaks were detected for A. tamutum (Supplementary Figure S1).

Immunological Analysis
An immunostaining differentiation between A. minutum and A. tamutum using an antibody raised against A. minutum cell lysates was conducted to assess the biochemical difference between these organisms. This monoclonal antibody, developed by Carrera et al. (2010), was raised in BALB/c mice, with the mouse sera extracted at 10-14 days post-immunization [35]. The monoclonal antibody showed significant and selective staining of A. minutum as compared to other Alexandrium species such as A. tamarense and A. andersonii [35].
Average fluorescence intensity for A. tamutum was significantly lower compared to A. minutum (Figure 1), based on One-Way ANOVA analysis and Tukey's HSD Test (p < 0.01) comparison. Autofluorescence was not observed for either species.

Gel-Free Proteomics Approach
Gel-free comparative proteomics analysis was done by iTRAQ labelling of proteins from A. tamutum and A. minutum followed by tandem mass spectrometry (MS/MS) analysis. Protein identification was done using MASCOT NCBInr (non-redundant) query which had 90,971,994 sequences and 33,504,913,701 residues available at the time of analysis. The search result gave a total of 154,069 queries with a total of 1436 proteins identified. The majority (93%) of the identified proteins have available protein annotation in UniProt. Out of the 1436 proteins identified, 1266 (88%) of the identified proteins based on UniProt were expressed by both A. minutum and A. tamutum. Common proteins in these two Alexandrium species gave 99% UniProt annotation. In terms of gene ontology of the common proteins, 46% are involved in molecular function, 22% are cellular component and 32% are involved in biological processes. The enzyme classes are oxidoreductases (33%), transferases (26%), hydrolases (16%), lyases (10%), isomerases (4%) and ligases (11%).
From the 1436 proteins identified in A. tamumtum and A. minutum, 96 proteins showed varying abundance between the two species based on the iTRAQ labeling approach (Supplementary Table S1).
A significant ratio (A. tamutum/A. minutum intensity ratio) of 1.3 and 0.77 at 95% confidence interval was used as criteria for high and low protein abundance, respectively.
An interspecies comparison of protein abundance showed 56 proteins were more abundant in A. minutum than A. tamutum. In contrast, 40 proteins had higher amounts in A. tamutum. The more abundant proteins in A. minutum were oxidoreductases and hydrolases. On the other hand, abundant proteins in A. tamutum were proteins involved in DNA-directed DNA polymerase reactions and acid anhydride catalysis such as 3-hydroxybutyryl-CoA dehydratases. These enzymes are known to be involved in carbohydrate metabolism.
To further improve the protein identification coverage, we conducted a restricted database search using the available in-house mRNA sequence information for the dinoflagellate Pyrodinium bahamense (NCBI accession: PRJNA261863, ID: 261863). Protein abundance levels using this reference transcriptome were determined following the significant ratio as mentioned previously.
Utilizing a P. bahamense transcriptome sequence data as reference, we identified an additional 19 proteins (Supplementary Table S2) with differential abundance in the two Alexandrium species.
The abundant proteins in A. minutum include osmotic growth proteins, type 1 polyketide synthase and Zn-dependent alcohol dehydrogenase, serine/arginine splicing factor 6, mitochondrial ubiquinol cytochrome c oxidoreductase core beta subunit-like protein 5, and contig06626_2 and contig00658_4 which both have no annotations and no conserved domains (Table 1). More abundant proteins for A. tamutum include ATP synthase subunit, 3-hydroxyacyl-CoA dehydrogenase, photosystem II protein D2, light-harvesting protein, eukaryotic translation initiation factor, methionine synthase, S-adenosyl-methionine trans-ferases, heat shock protein 90, plastid C1 class II fructose bisphosphate aldolase, malate dehydrogenase, and the remaining numbers included predicted, uncharacterized and hypothetical proteins (Table 1).

Gel-Based Proteomics Approach
A 2D-DIGE experiment was utilized as a complementary method to determine potential species-specific proteins (Supplementary Figure S2). A total of 420 protein spots were identified for A. minutum and A. tamutum. Out of these, 41 protein spots (10%) were observed to have >2 intensity ratio, indicative of differing abundance. A total of ten protein spots were prioritized for further analysis based on the signal intensity ratio of >2 and protein spot size (~2 mm diameter size) as criteria for selection ( Figure 3).  Figure S2). Proteins were separated in the first dimension using an immobilized pH gradient isoelectric focusing pI 3-10. Proteins were further separated in the second dimension based on molecular weight by SDS-PAGE. 2D gels were subjected to silver staining. Ten manually excisable protein spots were identified by tandem mass spectrometry (MS/MS). Protein spots with increased amounts in A. tamutum and A. minutum are labelled with red and blue circles, respectively. This was set to ensure successful protein identification after potential protein losses that may arise from extraction and handling since other protein spots were too small for excision. These spots were excised and further analyzed using MS/MS analysis for identification.
In the gel-based approach, high abundant proteins for A. minutum include ribulose biphosphate carboxylase (Rubisco), histidine kinases and hypothetical proteins (Table 3). On the other hand, hypothetical proteins, uncharacterized proteins, tRNA modification GTPase and thiamine monophosphate kinase (Table 3) were present in high amounts in A. tamutum.

Discussion
The dinoflagellate Alexandrium sp. has been implicated in harmful algal bloom through the production of paralytic shellfish toxins [36,37] and/or fish kills due to bioactive reagents such as reactive oxygen species and polyunsaturated fatty acids [38]. In this study, two Alexandrium species from Manila Bay, Bataan, Philippines were characterized. A. minutum produced GTX1,4, and GTX2,3 as primary toxins, with similar toxin profiles to A. minutum from southern Taiwan [39]. A. tamutum did not give any detectable toxins, in agreement with the observed toxin profiles for other A. tamutum isolates from the Mediterranean Sea, Italy and Malaysia [9,10]. Reported toxic Alexandrium species in the Philippines now include A. minutum aside from earlier reported Alexandrium cf. pacificum and Alexandrium affine [2].
Immunostaining significantly differentiated the cells of the two species collected at the exponential phase. The monoclonal antibody used in this experiment specifically binds to an antigen in A. minutum. The exact antigen is, however, unidentified since the antibody was raised using whole cell lysate of A. minutum [35].
An exhaustive characterization of these two closely related Alexandrium species was done using gel-based and gel-free proteomics approaches. A total of 420 protein spots were identified from the gel-based approach, almost three-fold less than the number of proteins identified using the gel-free approach (1436). The low number from 2D-DIGE may be attributed to overlapping and sometimes too small protein spots. The gel-free approach presented a challenge with the de novo peptide sequencing of proteins with very low protein identification scores, often due to the quality of the MS/MS data.
Despite the significant difference in detected proteins in 2D-DIGE and iTRAQ, both approaches showed little variance between the A. minutum and A. tamutum proteomes. Using the iTRAQ technique, 12% of the identified proteins showed different abundance in both organisms, while 10% of proteins spots were observed to vary using 2D-DIGE.
Our iTRAQ-based proteomic approach yielded comparable results with the gel-free proteomic profiling of Alexandrium catenella. Comparison of the protein profiles between the toxic A. catenella (ACHK-T) and the non-toxic mutant (ACHK-NT) strain showed 185 differentially expressed proteins from the 3,488 proteins identified (5.3%), with 1.2 and 0.83 significant ratio values as criteria [40]. Moreover, in comparison to a gel-based study conducted by Chan, et al. (2005) [23] for toxic and nontoxic cultured A. minutum from Taiwan, proteins were found to vary significantly at pIs ranging from 4.8 to 5.3 with molecular masses between 17.5 to 21.5 kDA using a pI 4-7 strip gradient. We observed differentially expressed proteins at pI of 4-7 using a pI 3-10 gradient and molecular mass of 17-56 kDa. In both studies, only few differences were detected between the toxic and the nontoxic species, in agreement with our observation.
Having the majority of proteins expressed with the same abundance in both organisms suggests that A. minutum and A. tamutum are most likely to follow similar cellular and biological processes at the exponential stage. This result provides molecular and cellular evidence to the initial study of Figueroa, et al. (2007) [41] which showed that the two species have major common life-cycle patterns and tend to have the same response to modifications in the external nutritional levels. We mapped the proteins with similar abundance levels in the two organisms to biological pathways such as biosynthesis of secondary metabolites, metabolism through carbon fixation, nitrogen metabolism, carbohydrate and lipid metabolism. These are the primary biological pathways involved in the response of the organism to differing nutritional levels in the environment.
Furthermore, proteins identified in both organisms are involved in amino acid biosynthesis, purine metabolism, carbohydrate metabolism, amino acid degradation, co-factor biosynthesis, pyrimidine metabolism, lipid metabolism and one-carbon metabolism such as folate metabolism. As cells are actively dividing at the exponential phase, much of the energy is allocated to the biosynthesis of amino acids. Amino acids are essential building blocks for a variety of biological pathways through cofactor activation as cellular component of new cells, as nutrients for the organism and as energy source [42]. In addition, pyrimidine and purine metabolism are involved in biosynthesis of DNA across all life forms [43]. Processes involving carbohydrates are essential for cellular compartment of the organism and for energy source. Both organisms have thecal plates or "armor" that consist primarily of cellulose [44]. Lipid metabolism among dinoflagellates is essential in adapting to the changing environmental conditions through the alteration of the lipid composition [45]. Folate-mediated one-carbon metabolism is typical among actively dividing eukaryotes as this process supplies one-carbon units to biological pathways such as nucleic acid biosynthesis, mitochondrial and chloroplast biosynthesis, methyl group biogenesis, vitamin metabolism and amino acid metabolism [46].
Proteins with differing abundance in A. minutum and A. tamutum are involved in vitamin and toxin biosynthesis and osmotic response. Furthermore, our results agree with other proteomic studies which identified differing abundance of proteins related to toxin production such as rubisco [47], alcohol dehydrogenase and S-adenosylmethionine transferase [48].
The gel-based approach showed proteins involved in photosynthesis such as rubisco and histidine kinases being more abundant in A. minutum. Rubisco, which among dinoflagellates is encoded by the nuclear DNA instead of the chloroplast DNA, is a protein mainly involved in photosynthesis [49]. In a study of Jiang, et al. (2015) [50], they observed that proteins involved in photosynthesis were upregulated during toxin production. However, the cause-effect relationship has yet to be established. Histidine kinases are transferases that play a major role in signal transduction across cellular membranes. Histidine kinases were recently discovered to be important for alphaproteobacteria and algae interaction via quorum sensing [51].
Abundant proteins for A. tamutum include thiamine monophosphate kinase, a key enzyme for vitamin B1 (thiamine) synthesis and converts thiamine monophosphate into thiamine pyrophosphate (coenzyme B1). Vitamin B1 is often derived from de novo synthesis or from the unphosphorylated thiamine (vitamin B1). Most harmful algal bloom species are known to be vitamin B1 and B2 auxotrophs or organisms that are dependent from available nutrients in the environment [52]. The prevalence of photosynthetic microalgae dependent on vitamin auxotrophy are as follows: 22% require vitamin B1 (thiamine), approximately 5% require biotin (vitamin B7), and more than 50% require vitamin B12 (cobalamin) [53,54]. Tang, et al. (2010) [52] showed that A. minutum CCMP113 from Vigo, Spain does not require thiamine from its environment for its proliferation. In this study, we report the presence of thiamine monophosphate kinase in A. tamutum. Interestingly, Cruz-Lopez and Maske (2016) [55] showed that in some marine dinoflagellates, vitamin B1 and vitamin B12 are provided by the associated bacteria. This indicates the potential role of endosymbionts in A. tamutum vitamin B1 production. The role of differential quantities of vitamin B1 and vitamin B12 has been implicated in toxin production and regulation. Since the cultures used in this study are xenic, utilization of axenic cultures can be explored to assess whether these dinoflagellates are prototrophic or auxotrophic in relation to the mentioned vitamins.
In terms of toxin-biosynthesis, the present study was able to identify proteins involved in toxin production in A. tamutum and A. minutum. Further, this proteomic result corroborates with the recent study of Vigniani, et al. (2020) in which transcripts related to saxitoxin-biosynthesis were detected in the non-toxic clone of A. tamutum isolated in Italy [56]. However, most of these detected proteins did not significantly differ in abundance (Table 2). This parallels the results of Zhang, et al. (2015) [40] for the toxic and nontoxic strains of A. catenella where only a small proportion of proteins involved in toxin synthesis were observed to vary.
Among the abundant proteins in A. minutum, type I polyketide synthase and alcohol dehydrogenases are directly involved in PST biosynthesis. Alcohol dehydrogenases are involved in the reduction of the terminal aldehyde in the eighth step of the predicted saxitoxin biosynthesis [57], while type I-polyketide synthase is involved in the initiation of toxin biosynthesis. In addition, osmotic growth protein, which catalyzes the reduction of fumarate to succinate and essential for anaerobic growth, was also observed to be more abundant in A. minutum. The osmotic growth protein enables dinoflagellates to grow despite the osmotic pressure from the environment [58,59]. Errera and Campbell (2011) [60] showed that osmotic stress triggered brevetoxin production in Karenia brevis. It was observed that the rapidly changing salinity which simulate a shift from oceanic conditions to a decreased salinity and mimics coastal conditions triggered a 14-fold increase in brevetoxin cell quota while the growth rate remain unchanged. However, this claim that osmotic pressure leads to toxin production has been challenged by other researchers [61]. The observed upregulation of histidine kinase and type-I polyketide synthase in this study parallels the results of Wang et al. (2012), where these proteins showed higher abundance in the toxic A. catenella compared to the non-toxic clone [28]. The decreased levels of these proteins in non-toxic Alexandrium species may limit the available biosynthetic precursors of PSTs and affect the rate of toxin biosynthesis. Detecting histidine kinase and type Ipolyketide synthase as differentially expressed proteins in multiple toxic and non-toxic Alexandrium species is suggestive of the potential utility of these proteins to serve as biomarkers for toxin production. However, in the follow up study of Zhang et al. (2015), there was no significant difference in the expression of toxin-related enzymes and instead, differential toxicity was related to carbon and energy utilization [40].
Noticeably, methionine synthase and S-adenosylmethionine (SAM) transferase, proteins involved in the methionine cycle [62], have increased levels in A. tamutum. Methionine synthase is responsible for the production of methionine from homocysteine. ATP when reacted with methionine in the presence of S-adenosylmethionine synthase forms SAM. SAM when methylated via S-adenosylmethionine (SAM) transferases becomes S-adenosylhomocysteine (SAH). SAH is reduced to homocysteine and other forms through the action of S-adenosylhomocysteine hydrolases. Despite the high abundance of SAM transferase in A. tamutum, there was no significant difference in the abundance of S-adenosylhomocysteine hydrolases, suggesting potential increase in SAH levels. Excessive SAH acts as an inhibitor of methylation reactions because it has a high affinity for most methyltransferases. This feedback mechanism has so far been reported in humans and plants but not in dinoflagellates [63,64]. This initial finding may suggest that the mechanism for SAH production might be a preferred reaction. It is possible that a negative feedback inhibition as a result of higher abundance of the SAM transferases may lead to inhibition of methylation ( Figure 4).
Methylation is a key reaction in the initial steps of saxitoxin production. Potential feedback mechanism to toxin biosynthesis has also been proposed for the non-toxic A. catenella strain [28], albeit with different set of proteins. In addition, the importance of SAM for toxin biosynthesis has also been highlighted in previous studies [40,48]. The toxic A. catenella strain showed increased levels of enzymes related to SAM biosynthesis [40,48] compared to the non-toxic strain. Our findings further support the potential contribution of feedback mechanisms to toxin production and warrant further investigation. Additional experiments using these proteins as target are suggested to probe the effect of these enzymes on toxin production.

Conclusions
An exhaustive comparative quantitative proteomics enabled the characterization and differentiation of two closely related Alexandrium species. Biochemical machineries such as toxin biosynthesis-related proteins were expressed by both organisms. The difference in abundance levels may possibly lead to changes in cellular dynamics and consequently, affects PST biosynthesis. A possible feedback mechanism is proposed. The favored reaction towards S-adenosylhomocysteine may lead to inhibition of methylation which is an important step in toxin production. Decreased levels of toxin biosynthetic enzymes and other proteins that create feedback mechanisms affecting the availability of PST biosynthetic enzymes and precursors may affect the toxin biosynthesis capability of Alexandrium species. Overall, this study has provided insights on the commonalities and divergence between two Alexandrium species and provided insights into the critical cellular machinery of these organisms.

Cultivation and Cell Collection
Subcultures of A. minutum Halim (AminBAT11713) and A. tamutum Montresor (Alex2LBol041313) were obtained from the Red Tide Laboratory, Marine Science Institute, University of the Philippines. The initial cultures of the two species were collected from Manila Bay, Bataan, Philippines on 17 January 2013 (A. minutum) and 13 April 2013 (A. tamutum). Monoclonal cultures were then subcultured in four 4 L flasks with F/2 culture medium under the conditions described previously by Azanza-Corrales and Hall, 1993 [65] and Subong, et al. 2017 [2], with the following modifications: temperature of 24 • C (±2), light intensity of 200 ± 50 µEm −2 s −1 following a 12 h:12 h light: dark cycle. Cell counts were taken every 4-5 days to monitor growth. Starting cell density was normalized to~200 cells/mL. Two biological replicates at the exponential phase were performed for each species.

Toxin Analysis
Two 50 mL-aliquots of A. mimutum and A. tamutum cultures were collected at the exponential phase using vacuum filtration with a 0.2 µM nylon filter. These were subjected to cell counting and toxin analysis. A. tamutum and A. minutum cells were extracted with 0.01 N HCl to obtain the paralytic shellfish toxins (PSTs) extract, as described by Oshima et al. (1995) [66]. PST concentration was determined using the pre-oxidation HPLC-fluorescence detection (HPLC-FD) method described by Lawrence and Ménard (1991) [67,68].
Periodate analysis was done for saxitoxin (STX), neo-saxitoxin (neoSTX) and gonyautoxin 1, 4 (GTX1,4) detection. In brief, a 50 µL aliquot of the PST extract was reacted with 0.03 M H 5 IO 6 , 0.3 M Na 3 PO 4 and 0.3 M NH 4 HCO 2 (1:1:1) for 3 min prior to the addition of concentrated CH 3 COOH. The resulting product was analyzed by HPLC-FD. STX, dcSTX and GTX2,3 content was analyzed after peroxide oxidation of the PST extract. The same volume of PST extract was reacted with 1.0 M NaOH and 10% H 2 O 2 (10:1), and concentrated glacial CH 3 COOH was added after 2 min. The resulting peroxide oxidation product was analyzed by HPLC-FD. HPLC-FD was done using analytical C18 (Inertsil ODS-3V, 250 × 4.6 mm, 5 µm) with 1.0 M NH 4 HCO 2 (pH 6.0) (Solvent A) and 5 % acetonitrile (Solvent B) as mobile phase. The oxidation products were eluted using the following mobile phase gradient: 1% Solvent B for 2 min; 5% Solvent B for 6 min; 6% Solvent B for 5 min; 10% Solvent B for 10 min at a flow rate of 1.2 mL/min. Oxidation products were monitored at 330 nm (excitation) and 400 nm (emission). The oxidation products of PSTs gave the following retention times (± 0.  [69,70]. Toxicity conversion factors were based on the method of Oshima (1995) [66]. Paralytic Shellfish Toxin (PST) concentrations were quantified by integrating the peak areas and calculated using the following equations:

Immunostaining Using Confocal Laser Scanning Microscopy
Immunofluorescence of cultured whole algal cells was performed using the procedure described by Lin and Carpenter (1996) and Carrera, et al. (2010) [35,71] with some modifications. Cells at exponential phase were harvested, collected by centrifugation (500g for 5 min, 37 • C) in a refrigerated centrifuge and fixed overnight using 4% paraformaldehyde in 0.1 M PBS. Samples were then rinsed with 0.1 M PBS (3×) and were treated with methanol at −20 • C for 24-72 h to remove the pigments. A final concentration of 5 ×10 4 cells in PBS/1% BSA for each aliquot was subjected for antibody labeling. Labeling was performed using 50 µL of hybridoma supernatant containing monoclonal antibody (with concentration of~5-25 mg/mL) for 1 h at room temperature with occasional shaking. The cells were washed with PBS/1% BSA and subsequently incubated with 50 µL of secondary antibody, fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG (H + L) (1:200 in PBS/1% BSA). The samples were incubated in the dark for 1-2 h at room temperature and subsequently washed with PBS/1% BSA. The samples were resuspended in 200 µL of PBS/1% BSA. Finally, the cells were imaged using a 630× magnification of a confocal laser-scanning microscope (Zeiss LSM 710, Carl Zeiss AG, Oberkochen, Germany). Two technical replicates were done per biological replicate. A negative control was also performed, with no antibody labeling to assess potential pigment fluorescence. For the single cell analysis of immunofluorescence, 5-10 cells for each sample per biological replicate were imaged for each experiment. Images were then processed for green intensity quantitation using ImageJ (Bankhead, 2014) [72] bundled with Java 1.80.0 software (US National Institutes of Health, Bethesda, MD, USA). The antibody was a generous gift from Prof. Africa Gonzales-Fernandez of Area de Inmunologia, Universidad de Vigo, Spain.

Protein Extraction for iTRAQ-Method
A total of 3 × 10 6 cells for each sample was used for protein extraction. Samples were lysed (burst of 5 s) using a sonicator probe for 5 min and cell lysis was confirmed by visual inspection using a microscope. The lysis buffer contained 6 M urea, 3 M thiourea, 20 mM TEAB, 5 mM TECP, 0.1% SDS, and 0.1% protease inhibitor mix. Cell lysates were centrifuged at 5000g, 4 • C for 5 min. The supernatants were recovered and precipitated using 20% TCA/acetone precipitation to yield the protein pellets. The protein pellets were washed with acetone (3×) to remove residual TCA. Protein pellets were resuspended in 9 M urea and 0.5% SDS and protein concentrations were measured using Bradford assay. In total, two biological replicates were performed.

Protein Labelling and Strong Cation Exchange Chromatography
For each biological replicate, one technical replicate was provided for iTRAQ analysis. Protein samples (50 µg) were reduced using DTT, alkylated with iodoacetamide and digested with trypsin. Tryptic digests were desalted using C18 zip-tip prior to iTRAQ 4-plex (AB Sciex, Foster City, CA, USA) labeling. A. minutum was labeled with 114 and 117 isobaric tags while A. tamutum with 115 and 116 isobaric tags. Labeled samples were combined in a 1:1:1:1 ratio (with BSA as internal control) and prefractionated offline using strong cation exchange column, polysulfoethyl A (200 × 2.1 mm; 5 µm) (PolyLC INC., Columbia, MD, USA) column, on an Agilent 1100-HPLC system (Agilent Technologies, Palo Alto, CA, USA). Solvent A was: 25% CH 3 CN; 0.1% HCOOH while Solvent B was 25% CH 3 CN; 0.1% HCOOH; 500 mM KCl. A linear gradient from 0-100% Solvent B in 45 min was performed with flow rate set at 200 µL/min. Peptides were detected using a UV detector set at λ = 214, 260 and 280 nm. A total of 21 fractions were collected using a fraction collector. Fractions were further analyzed using High-Performance Liquid Chromatography-Mass Spectrometry (HPLC-MS).

Mass Spectrometry Analysis, Protein Identification and Annotation
Collected fractions were dried and resuspended in loading buffer for reversed phase chromatography coupled to an LTQ-Orbitrap XL mass spectrometer. Mass spectrometry analysis was performed at Proteome Factory AG (Berlin, Germany). The LC component comprised of an Agilent ZORBAX SB-C18 (150 × 0.075 mm; 3.5 µm) column attached to an Agilent 1100 HPLC system (Agilent Technologies, Palo Alto, CA, USA). Fractions 1-7 and 20-21 were analyzed using short gradients (60 min each) while the remaining fractions were analyzed using a long gradient (240 min). Solvent A is 5% CH 3 CN + 0.1% HCOOH while the Solvent B is 100% CH 3 CN + 0.1% HCOOH. The gradient for the 60 min run consisted of 5% Solvent B for 5 min, 5-32% Solvent B in 50 min, 32%-90% Solvent B in 1.5 min, 90% Solvent B for 4 min. The gradient for the 240 min run includes 5% Solvent B for 5 min, 5%-20% Solvent B in 190 min, 20%-32% Solvent B in 30 min, 32%-90% Solvent B in 3 min, 90% Solvent B for 6 min. The MS method was performed following a data-dependent acquisition with MS-overview scan (MSI) between 350-16,000 m/z with 30,000 resolution. This was followed by MS/MS fragmentation of top five ions with collision-induced dissociation (CID) fragmentation at collision energy of 35 eV and fragmentation of the same precursor ions with higher-energy collisional dissociation (HCD) at normalized collision energy of 45 eV. The signal intensity threshold was set to 20,000, isolation width of 3 Da m/z and activation of 30 ms. Ions with unassigned charged states and singly charged were rejected for data dependent acquisition. Dynamic exclusion was activated as follows: 2 repeats in 30 s (top 5 most intense ions fragmented twice in CID and HCD) with exclusion list size 500, exclusion duration 240 s and exclusion mass width of 40 ppm. Tune file information include AGC target FTMS full scan of 2 × 10 5 , maximum injection time of 500 ms, AGC target MS n ion trap: 1 × 10 4 ; maximum injection time: 500 ms, AGC target MS n FTMS: 1 × 10 5 , maximum injection time of 500 ms. All scans were performed using only one micro scan.
Results were queried against MASCOT, v.2.5 using NCBInr (November 2016) to map the proteomes of A. minutum and A. tamutum. Subsequent focused database queries against in-house transcriptome sequencing data of Pyrodinium bahamense at the Marine Science Institute, University of the Philippines using MaxQuant v1.4 [73,74] was also performed to determine proteins with differing abundance in the two Alexandrium species. This was performed to increase the number of identifiable proteins. In-house genome sequence data is publicly available at NCBI accession: PRJNA261863, ID: 261863.
Protein hits which contained <1% false discovery hits at confidence interval (CI) of 95% are reported. Protein score threshold of 20 was utilized for MaxQuant searches [75] while a protein score of 10 was utilized for MASCOT [76].
A cut-off significant ratio (A. tamutum/A. minutum) of ≥1.3 (CI of 95%) and ≤0.77 was used to describe high or low abundance proteins in A. tamutum. Data normalization was performed using the average of the median intensities for the different labels.

Sample Extraction and Preparation for 2D-DIGE
A total of 3 × 10 6 cells were lysed using a sonicator probe with a buffer containing: 6 M urea, 2 M thiourea, 4% CHAPS, 5 mM MgOAc, 10 mM Tris pH 8.5 with protease inhibitor cocktail (1% v/v). Homogenization was performed on ice and was performed in bursts of 5 s. Samples were transferred to 15 mL tubes and proteins were precipitated using cold 10% TCA in acetone for 1 h. Protein pellets were washed with 80% 0.1 M NH 4 OAc dissolved in 80% CH 3 OH and 20% water (2×) and a final wash of 80% acetone. The air-dried pellets were dissolved in DIGE compatible buffer consisting of 6 M urea, 2 M thiourea, 4% CHAPS, 5 mM MgOAc, 10 mM Tris pH 8.5. Sample was cleaned using a 2D clean up kit according to manufacturer's protocol (GE Healthcare). The pellets were dissolved in previously mentioned DIGE compatible buffer. Protein content was done with Quick Start Bradford Assay from Bio-Rad according to the manufacturer's protocol.

2D-DIGE Labeling and 2D Runs
For 2D-DIGE experiment, 50 µg of A. minutum protein was labeled with 250 pmolar Cy3 (Chromis DGE Minimal Labelling Kit from Cyanagen, Bologna, Italy) while 50 µg of A. tamutum was labeled with 250 pmolar Cy5 (Chromis DGE Minimal Labelling Kit from Cyanagen, Bologna, Italy) according to manufacturer's protocol. Reciprocal labeling was also performed. Two technical replicates for each of the two biological replicates were done.
For second dimension separation, 12% SDS-PAGE gels (16 × 15 cm) were used. Gels were run using the following parameters: 20 mA/gel for 20 min followed by 40 mA/gel for a remaining time until the dye front leaves the gel.
Gels were scanned using a Typhoon 9400 fluorescent scanner using wavelengths adequate to Cy3 and Cy5 dye as per manufacturer specification. The resolution of the scan was 100 µm pixel size. Images were then loaded into DIA module (Differential in gel Analysis) of the DeCyder 5.2 software (GE Healthcare Life Sciences, Marlborough, MA, USA). Normalization was performed by the generation of a "master gel" from all the runs including reciprocal labeling runs. Distinct spots were identified following a Gaussian fitting (CI at 95%). Silver staining procedure that is compatible for MS analysis was used as a post-staining procedure. The staining reaction was developed in 30 s. Protein spots were then manually excised for further analysis.

Mass Spectrometry Analysis, Protein Identification and Annotation
Protein samples were trypsin digested and peptides were extracted. In brief, sufficient amount of trypsin (12.5 mg/mL trypsin, 25 mM ammonium bicarbonate) was added to the gel pieces and these were incubated at 37 • C. The digested peptides were extracted (3×) with CH 3 CN containing 1% TFA. The pooled extracts were dried by rotary evaporation and stored at −20 • C until further analysis. Mass spectrometry analysis was performed by 1st BASE (Selangor, Malaysia). Peptides were analyzed by electrospray ionization mass spectrometry with subsequent MS/MS fragmentation using a Shimadzu Prominence nano HPLC system (Shimadzu, Kyoto, Japan) coupled to a high resolution 5600 TripleTOF mass spectrometer (AB Sciex, Framingham, MA, USA). Tryptic peptides were loaded onto an Agilent Zorbax 300SB-C18 (150 × 4.6 mm, 3.5 µm; Agilent Technologies (Palo Alto, CA, USA) and separated with a linear gradient from 0-100% H 2 0/CH 3 CN/0.1% HCOOH (v/v) at a flow rate of 0.5 mL/min for 45 min. Protein identification was done using MASCOT, v2.5 sequence matching software (Matrix Science) with MSPnr100 database (December 2016). The following parameters were used: peptide tolerance: ±0.2 Da, MS/MS tol: tol: ±0.2 Da, peptide charge: 2+, 3+ and 4+, Mass: monoisotopic, Enzyme: trypsin, missed cleavage: 1, fixed modification: carbamidomethylation, variable modification: oxidation (methionine). A protein score of >60 at a CI of 95% was considered as a good hit and was utilized as one of the criteria for gel-based protein identification [2,24,80].