Karlodinium veneficum strain CCMP2778 was originally isolated from coastal area off Longboat Key near Sarasota, Florida USA and provided by the Provasoli-Guillard National Center for Marine Algae and Microbiota (NCMA) in Boothbay Harbor, Maine, USA. In our laboratory, the culture was maintained in L1 medium (NCMA recipe) amended seawater (salinity, 28 PSU), which was filtered through 0.22-μm membranes and autoclaved. Cultures were grown at 20°C under a 14 h: 10 h light dark cycle with a photon flux of 100 ± 10 μmol photons m^-2 s^-1. To obtain the cultures under contrasting P conditions, the cultures were first grown in L1 medium until it reached the exponential growth stage, and were then inoculated into L1 and L1-P (same as L1 except that no phosphate was added) medium under the same light environment, both conditions were treated in triplicate. For the cultures under different light conditions, algal cells in the exponential growth stage were inoculated into new L1 medium to be cultured at 20°C under different light conditions (50, 300, and 600 μmol photons m^-2 s^-1) with previous diurnal cycle, each treated in triplicate. The cultures were acclimated to these light intensities for six generations before the measurements were made. The experimental cultures described above were grown in a volume of 300 mL in 500-mL flasks. Cell counts were carried out using a Sedgwick–Rafter counting chamber (Phycotech, St. Joseph, MI, USA). The concentration of dissolved inorganic phosphate (DIP) in the medium was measured using the molybdenum blue method (Timothy et al., 1984). About 1 × 10⁶ cells were collected from each culture at the selected time by centrifugation (5000 g, 10 min) and resuspended in 1 mL Trizol reagent (MRC, Cincinnati, OH, USA) and stored at –80°C for subsequent RNA extraction.

Chlorophyll fluorescence of K. veneficum cells was measured using a FIRe fluorometer system (Satlantic, Halifax, NS, Canada). The algal cell concentration was diluted to approximately 10, 000 cells per mL during the NPQ measurements to avoid self-shading effect. The high luminosity blue light (maximum emission 455 nm, 60 nm bandwidth) in FIRe was used to excite chlorophyll fluorescence. K. veneficum cells were sampled at the 8th hour of the light cycle and were then dark adapted for 30 min at 20°C before measurement. NPQ was calculated as (Fm-Fm’)/Fm’ and the maximum quantum efficiency of PSII photochemistry Fv/Fm = (Fm–F₀)/Fm (Maxwell and Johnson, 2000; Baker, 2008), where F₀ is the minimal fluorescence obtained in the presence of the measuring light; Fm is the maximum fluorescence of dark-adapted algal cells measured during a very short and strong single turnover flash (STF); and Fm’ is the maximum fluorescence measured after the cultures were exposed to a continual actinic light (PAR, photosynthetically active radiation, wavelength range from 400∼700 nm) using the actinic light source (ALS) through the manual PAR acquisition or PAR stepping acquisition of FIRe. NPQ induction and relaxation kinetics under contrasting light and P conditions were observed through the manual PAR acquisition, algal cells after dark-adaption were exposed to actinic light for 10 min and then the actinic light was turned off for another 10 min, Fm’ were measured at the end of each minute during this process and NPQ was calculated accordingly. The algal cells cultured under 100 μmol photons m^-2 s^-1 were sampled for NPQ estimation under contrasting actinic light intensities. To measure the induction and relaxation of NPQ under contrasting P conditions, the algal cells were sampled at the 10th day upon inoculation to +P and –P conditions. The maximal NPQ capacity of the algal cells under +P and –P condition was also measured using the PAR stepping acquisition, in which the PAR intensity increased by 50 every 30 s from 0 to 700 μmol photons m^-2 s^-1.

We investigated our annotated dinoflagellate-specific spliced leader (DinoSL)-based K. veneficum cDNA database (Lin et al., unpublished, as briefly reported in Cui et al., 2016) for genes potentially related to NPQ. The sequences acquired were further confirmed by blastp against NCBI GenBank database. To obtain the full-length cDNA of these genes, we extracted RNA from K. veneficum cells as previously reported (Lin et al., 2010). Specific primers (Supplementary Table S1) were designed for both 3′- and 5′- RACE based on the partial sequences identified from the above-mentioned transcriptome dataset. The 21-bp highly conserved DinoSL was used as the 5′ forward primer for the 5′-RACE (Zhang et al., 2007; Lin et al., 2010). The amplicons were cloned into T-vectors and sequenced through Sanger sequencing.

Specific primers (Supplementary Table S1) were designed for RT-qPCR to examine the differential expression of the photoprotection genes identified in this work under different P and light conditions. Calmodulin (calcium-modulated protein; KM275627) was used as the reference gene because of its relative stable expression previously reported in some other dinoflagellates (Rosic et al., 2011; Shi et al., 2013). For standard curves, a purified PCR product for each gene was prepared in ten-fold dilution series (10³–10⁷ copies per μL). RT-qPCR was performed using Bio-Rad iQ SYBR Green Supermix Kit (Bio-Rad Laboratories, Hercules, CA, USA) with all the reactions set up in triplicate for each gene. Relative transcript levels of these genes were calculated in two ways to facilitate comparison: normalized to the amount of total RNA equivalent to the amount of cDNA used in each reaction, and to the expression levels of the reference gene calmodulin.

We carried out iTRAQ (isobaric tags for relative and absolute quantitation) analysis to identify the differentially expressed proteins in K. veneficum collected from +P (L1) and -P (L1-P) conditions. Cultures under contrasting P conditions were obtained as described above and grown in a volume of 1 L in 2-L flasks. After inoculation into different conditions, the +P cells were sampled at the 3rd day, and the –P cells were sampled at the 9th day, each in duplicate. About 3 × 10⁷ algal cells were harvested from each culture and protein was extracted as previously reported (Wang et al., 2013). Total protein was quantified through Bradford protein assay and 100 μg from each sample was used for iTRAQ labeling. Samples were labeled with the iTRAQ tags and SCX-fractionated with a LC-20AB HPLC pump system (Shimadzu, Kyoto, Japan). Liquid chromatography electrospray ionization tandem mass spectrometry (LC-ESI-MS/MS) analysis was then performed based on a TripleTOF 5600 System (AB SCIEX, Concord, ON, USA), followed by protein identification through Mascot search engine (Matrix Science, London, UK; version 2.3.02) against the above-mentioned DinoSL-based K. veneficum cDNA database. To reduce the probability of false peptide identification, only peptides at the 95% confidence interval by a Mascot probability analysis greater than “identity” were accepted, and each confident protein identification was represented by at least one unique peptide.

The abundance of a protein was quantified only when it was represented by at least two unique peptides in our proteomic data. The quantitative protein ratios were weighted and normalized by the median ratio in Mascot. The criteria as fold changes >1.2 and p-values < 0.05 was adopted to depict significantly differentially expressed proteins. Functional annotations of the proteins were conducted using Blast2GO program against NCBI non-redundant (nr) protein database and Uniprot database¹. The KEGG database² and the Clusters of Orthologous Groups (COG) database³ were used to classify and group these identified proteins.

Analysis of variance (ANOVA) was carried out using PASW Statistics 18 software package to evaluate the statistical significance of the differences between contrasting light and P conditions. Data shown in the figures are means with standard deviation calculated from different replicates.

Karlodinium veneficum cells after dark-adaption were exposed to different light intensities of actinic light using FIRe to observe the induction and relaxation of NPQ (Figure 1A). Under actinic light of 50 μmol photons m^-2 s^-1, the NPQ was induced briefly at the first three minutes and then the NPQ returned to zero, indicating that under this light intensity the algal cells did not need NPQ to dissipate the excess light. The increase of NPQ in the beginning was due to the sudden shift of dark-adapted cells to the light. When the dark-adapted algal cells were exposed to actinic light of 200, 300, and 600 μmol photons m^-2 s^-1, NPQ was induced quickly and significantly. After light was turned off at the 10th minute, NPQ relaxed quickly but not completely in several minutes. During the first two to three minutes of the light phase, an abrupt increase in NPQ was observed from the dark to light transition. Subsequently the NPQ showed slightly different fluctuations under the three different light conditions. Under 200 and 300 μmol photons m^-2 s^-1, the NPQ value decreased to a minimum and then increased to a relatively steady state before a sharp decline occurred at the 10th minute. Under 600 μmol photons m^-2 s^-1, the NPQ decreased gradually until it reached a relatively steady value.

The NPQ induction and relaxation of K. veneficum cells under +P and –P conditions were studied through a continual exposure to 700 μmol photons m^-2 s^-1 for 10 min followed by dark treatment for 10 min (Figure 1B). The results showed that NPQ under P-depleted condition was induced more quickly and the values were generally higher compared to that under P-replete condition. Upon switch to dark, NPQ relaxed quickly while it still maintained at a higher level for the P-depleted cells compared to the P-replete cells.

We also conducted a 10-day experiment to further estimate the NPQ capacity and the Fv/Fm under the two P conditions. Algal growth rate under +P condition was higher than that under the –P condition (Figure 2A). The DIP depletion and the cessation of the population growth under the –P condition indicated that the cultures were experiencing P deprivation from the fourth day of the experiment (Figures 2A,B). Fluorescence measurement showed that the Fv/Fm decreased over time under the -P condition while that in the +P cultures kept at relatively stable and higher levels (Figure 2C), indicating that P deprivation led to a lower photochemical efficiency. In accordance, the NPQ capacity of the P-depleted cells was significantly higher than that of P-replete cells (Figure 2D).

Genes encoding five LHCX proteins (Genebank No. KX524133 to KX524137) and a homolog of Phototropin-2 (PHOT2, KX524138), which is involved in chloroplast avoidance movement in plants (Kasahara et al., 2002), were identified in the cDNA library of K. veneficum (Table 1). With specific primers designed from the partial sequences obtained, our RACE yielded cDNAs containing the complete open reading frame (ORF), with existence of DinoSL at the 5′ UTR of these gene transcripts confirming their dinoflagellate origin. RT-qPCR was conducted to study their transcript abundances. The results showed that the transcriptional regulation of the lhcx and the phot2 genes was quite limited in response to different P conditions (Figure 3). Transcript abundances of lhcx1, 3, 4 and phot2 did not show significant differences between +P and –P conditions. However, lhcx2 showed a significant up-regulation under –P condition (Figure 3B). Moreover, lhcx5 was remarkably down-regulated under the P-depleted condition compared to P-replete condition (Figure 3E).

Transcript abundances of LHCX proteins and PHOT2 in K. veneficum cells cultured under different light conditions were also studied through RT-qPCR. The growth rate of cultures under 50, 300, and 600 μmol photons m^-2 s^-1 was 0.239, 0.2, and 0.17, respectively, indicating that under the three light intensities employed in this study, the higher the light intensity used, the lower growth rate the cultures achieved (Figure 4A). The transcript abundance of lhcx1 was very high, even higher than calmodulin, the reference gene used in this study; however, the transcript level was similar under the three light conditions (Figure 4B). Similarly, lhcx3, 4 and 5 did not show a significant differential expression under the three light conditions (Figures 4C,D). In contrast, lhcx2 exhibited changes in transcript abundance, higher under 600 μmol photons m^-2 s^-1 than under 50 and 300 μmol photons m^-2 s^-1 with the latter two conditions producing no significant difference. Phot2 also showed a transcriptional regulation in response to different light conditions but the pattern was different from lhcx2. The transcript abundance of phot2 was higher at the light intensities of 300 and 600 μmol photons m^-2 s^-1 than at 50 μmol photons m^-2 s^-1, albeit at a small magnitude, but the former two light conditions did not elicit significant difference.

The iTRAQ proteomic analysis identified 4, 922 proteins for K. veneficum grown under +P and –P conditions (Supplementary Figure S1); 82 of them were found to be up-regulated under the –P condition (Supplementary Table S2), with 43 being classified into different COG functional categories, mainly carbohydrate metabolism, energy production and amino acid metabolism (Figure 5). Among these up-regulated gene categories there were two inorganic pyrophosphatases, which catalyze the hydrolysis of pyrophosphate into phosphate. Also identified were a glycerol-3-phosphate dehydrogenase and a putative sterol carrier protein, which might be involved in lipid metabolism. Three proteins were related to inorganic ion transport and sulfur metabolism. One photosystem II protein and two mitochondrial tricarboxylate transporters were also identified among the up-regulated proteins.

Totally 136 proteins were down-regulated under the –P condition and 41 of them were classified into 11 COG functional categories (Figure 5 and Supplementary Table S3). Most of these proteins are related to nucleic acid and protein synthesis (e.g., involved in nucleotide transport and metabolism, DNA replication, transcription, translation, and posttranslational modification, Figure 5). Three proteins in signal transduction pathways, including two Ca²⁺-binding proteins, were identified. A predicted histidines or aspartates domain phosphohydrolase was also down-regulated under the –P condition (Supplementary Table S3). From the 93 down-regulated proteins that were not grouped into COG functional categories, a thylakoid luminal protein, a polymerase, a peptidase, a deoxyribonuclease II, a cold shock protein, two RNA-binding proteins, two cathepsins and four enzymes related to amino acid metabolism were identified (Supplementary Table S3). Most of the other proteins were annotated as putative uncharacterized proteins and predicted proteins. Fifty-four other down-regulated proteins under P stress had no matches in the database, potentially novel proteins responding to P stress.

Seventy-two light harvesting protein complexes (LHCs) were identified and five of them were annotated as stress-related chlorophyll a–b binding proteins LI818 (Savard et al., 1996; Richard et al., 2000; Peers et al., 2009), denoted as LHCX proteins in this study (Table 1). Totally 27 of the LHCs were up-regulated under the –P condition (Supplementary Table S2), these included three LHCX proteins, LHCX1, LHCX2, and LHCX4 (Figure 6). LHCX5 was excluded from the comparative proteomic analysis because its abundance was too low. PHOT2 was identified in the K. veneficum proteome and its abundance was also found to be higher under the –P condition (Figure 6). Besides, two VDE proteins (KX524139 and KX524140) and one ZEP protein (KX524141) were also identified in the proteome and cDNA library of K. veneficum (Table 1).

Phytoplankton live in surface water and often face excess light, and thus have developed many strategies to protect them from photo-oxidative damage caused by excess light energy (Niyogi, 1999; Erickson et al., 2015). While recent study about NPQ in dinoflagellates mainly focus on the symbiotic species of corals, NPQ mechanism in HAB-forming dinoflagellate species, which play an ecologically important role in the marine ecosystem, has rarely been explored (Goss and Lepetit, 2015). From the induction kinetics of NPQ observed in the present study, it is evident that transfer from darkness to light, even low light, can lead to a brief induction of NPQ in K. veneficum cells. Within the range used in this study (50–600 μmol photons m^-2 s^-1), NPQ increased with the intensity of ambient light. Reduced growth rate of the K. veneficum cultures was found under high light condition in this study (Figure 4A), indicating that excess light caused photoinhibition to this species. The induction of NPQ under high light and its relaxation in the dark shows that NPQ is an important mechanism for K. veneficum to cope with absorbed excess light energy. Our results showed that the induction and relaxation of the NPQ in K. veneficum was very rapid, indicating that qE, which enables the organism to cope with the frequently and rapidly changing light field in the coastal marine ecosystem, is a major constituent of NPQ in this species, as is the case for most plants and algae (Müller et al., 2001; Erickson et al., 2015; Goss and Lepetit, 2015) Moreover, from our K. veneficum proteomic dataset, we detected the LHCX proteins, which have been confirmed to play a vital part in the qE of many microalgal groups (Peers et al., 2009; Bailleul et al., 2010; Goss and Lepetit, 2015). NPQ did not relax completely in ten minutes under dark environment, especially in the P-deprived condition, suggesting that other components of NPQ such as qT and qI, which needs longer time to relax, were also induced, as has been described in other algae (Wykoff and Grossman, 1998; Zhu and Green, 2010). It has also been proposed that diatoxanthin could also contribute to the sustained part of NPQ (Lavaud and Lepetit, 2013).

Earlier pigment composition analysis revealed presence of violaxanthin but absence of antheraxanthin and zeaxanthin in K. veneficum CCMP2778, while the diadinoxanthin and diatoxanthin were very abundant in this strain (Bachvaroff et al., 2009), suggesting that Dd-Dt cycle is the main xanthophyll cycle in this species. This corresponds to previous findings that xanthophyll cycle in other groups of dinoflagellates is also made of the Dd-Dt cycle (Demers, 1991; Ambarsari et al., 1997). Violaxanthin in this species could be the precursors of diadinoxanthin, diatoxanthin and fucoxanthin, as previously reported in other algae containing Dd-Dt cycle (Lohr and Wilhelm, 1999; Goss and Jakob, 2010). The VAZ cycle is catalyzed by violaxanthin de-epoxidase (VDE) and zeaxanthin epoxidase (ZEP) while the Dd–Dt cycle depends on diadinoxanthin de-epoxidase (DDE) and diatoxanthin epoxidase (DEP) (Goss and Jakob, 2010; Goss and Lepetit, 2015). DDE and DEP were detected from our K. veneficum proteome but were annotated as VDE and ZEP (Table 1), as in the case of the diatoms P. tricornutum and T. pseudonana (Goss and Lepetit, 2015), because VDE versus DDE and ZEP versus DEP have high sequence identities (Coesel et al., 2008). VDE and DDE would differ in optimal activation pH, and ZEP and DEP differ in the regulation of enzyme activity (Jakob et al., 2001; Goss and Jakob, 2010). The exact structural and functional nature of the xanthophyll cycle in the dinoflagellates requires further studies.

From the RT-qPCR results, four of the five identified lhcx genes did not show transcriptional regulation under different light intensities, although the growth rate and the induced NPQ of the algal cells under the three light conditions were very different (Figure 4). Lhcx2 and phot2 were differentially expressed at the transcriptional level according to ANOVA; however, the fold change is rather limited. The limited transcriptional regulation of the photoprotection genes in K. veneficum cells was also found when they were cultured under different P conditions (Figure 3). Three lhcx genes (lhcx1, 3, and 4) as well as phot2 did not show transcriptional regulation between P-replete and P-depleted conditions. Interestingly, although both were lhcx genes, lhcx2 was up-regulated while lhcx5 was down-regulated under the P-depleted condition compared to the P-replete condition, indicating that different LHCX genes might respond differently to different environmental stresses. Similar results have also been observed in diatoms under different light and nutrient stresses, reflecting functional diversification of the LHCX gene family in the microalgal groups which enable them to adapt to different ecological niches in the ocean (Bailleul et al., 2010; Zhu et al., 2010; Taddei et al., 2016).

On the other hand, the iTRAQ analysis showed that the expression of LHCX1, 2, 4 and PHOT2 proteins were up-regulated under the P-depleted condition, indicating that the regulation of most of the LHCX proteins (except LHCX2) and PHOT2 rest at different levels. It seems that the regulation of the LHCX proteins and PHOT2 in K. veneficum lies mainly in the translational level. The nonsynchronous regulation of these proteins at the transcriptional and translational levels might be because the transcriptional regulation of genes in K. veneficum was quite limited, as has been observed in many other dinoflagellates (Lin, 2011).

Phytoplankton have evolved many strategies to cope with P deprivation, including reducing the cellular demand of P and enhancing the ability to utilize other P sources such as dissolved organic phosphorus (Dyhrman et al., 2007; Van Mooy et al., 2009; Lin S. et al., 2016). Our results showed that although K. veneficum experienced growth inhibition under P deprivation, it could still maintain a stable population for an extended period of time (Figure 2). Accordingly, the comparative proteomic analysis reveals significant reconfiguration of the metabolic machinery in K. veneficum under P deprivation (Figure 7). Many proteins involved in the genetic information flow (e.g., DNA replication, transcription, translation and post-translation) were down-regulated to reduce the demand of P as protein synthesis is one of the major P sinks (Lin S. et al., 2016). Similarly, phosphonate metabolism was slowed down under P-depleted condition (Cui et al., 2016). However, K. veneficum cells maintained and even strengthened the functions related to energy production and processes demanding less P such as glycolysis pathway, tricarboxylic acid (TCA) cycle and lipid metabolism. Meanwhile, pyrophosphatases were up-regulated, which could hydrolyze pyrophosphate to release phosphate. Furthermore, the abundances of over a third of the LHCs were up-regulated under P-depleted condition. The metabolic machinery reconfiguration is consistent with the proposal that phytoplankton could increase the proportion of resource acquisition machinery such as P-poor proteins and pigments and decrease the production of growth machinery such as ribosomal RNA when resources are scarce (Klausmeier et al., 2004; Arrigo, 2005). Similar proteomic landscape changes such as the elevation of the ability to scavenge or economize P, increase of the LHC abundances, adjustment of the glycolysis pathway and down-regulation of protein synthesis under P deprivation have also been documented in the diatom T. pseudonana and the pelagophyte Aureococcus anophagefferens (Wurch et al., 2011; Dyhrman et al., 2012).

It is interesting to observe elevation in NPQ and up-regulation of NPQ-related and other photoprotective proteins in K. veneficum in response to P deprivation. The identification and the up-regulation of PHOT2 under P-depleted condition suggest that K. veneficum can potentially perform chloroplast avoidance movement to reduce the absorption of photons by the chloroplast. The higher NPQ measured in the P-depleted condition indicates that the algal cells could enhance their NPQ activity under P stress to dissipate the excess light stress and protect them from the potential photo-oxidative damage. This is the first documentation of this phenomenon in dinoflagellates, to the best of our knowledge. Our proteomic and transcriptional analyses discussed above, including the up-regulation of LHCX1, 2, and 4, have provided molecular evidence for the enhanced NPQ under P deprivation in K. veneficum.

The increase in the abundances of many light-harvesting proteins and the enhanced function of metabolic machineries related to energy production and conversion such as glycolysis, TCA cycle and pyrophosphate hydrolysis in the P-deprived cells indicate that the acquisition of light energy and the downstream energy flow were enhanced in the P-deprived cells (Figure 7). We suggest that the energy flow was accelerated in the P-deprived cells to increase the recycling rate of P-containing compounds such as ATP and NADPH to compensate for the very low external supply of P. Besides, ATPs generated from these metabolic processes are supposed to supply the energy needed for Pi acquisition, as the uptake of low concentration Pi and the utilization of DOPs from ambient environment by the P-deprived cells require energy (Petrou et al., 2008; Lin S. et al., 2016).

Despite the increased absorption of light energy, the algal cultures in our study exhibited a compromised photosynthetic efficiency when P was deprived, which is similar to the case of another dinoflagellate Amphidinium carterae (Li et al., 2016). The decreased photochemical efficiency would aggravate excess light stress (Wykoff and Grossman, 1998). Under this condition, elevated thermal dissipation through NPQ could protect the photosynthetic apparatus from photodamage and maintain the fluency of the energy flow. As photosynthesis is the basis of energy acquisition in algae, protection of the photosynthetic apparatus provided by the enhanced NPQ would be critical to the vulnerable algal cells suffering from P deprivation in a complex and changing light environment. Thus, NPQ plays an important role in the modulation of light energy and keeps the balance of the energy budget in the P-deprived algal cells. There could also be some interactions between light intensities and P-deprivation related to the regulation of NPQ capacity, which needs further work to be explored in the future.

Taken together, the results from this study suggest that NPQ functions are not only a protection from high light condition, but can also be an important adaptive mechanism for algal cells to cope with P deprivation. It gives flexibility to the P-deprived algal cells which need to acquire more energy with a lower photochemical efficiency to fuel the P acquisition processes and compensate for the P deprivation. This mechanism together with other photoprotective strategies could maintain the operation of photosynthesis and downstream functions related to energy flow and conversion and thus serve as an essential survival strategy for the dinoflagellate under P deprivation.