Temporal Variability of Virioplankton during a Gymnodinium catenatum Algal Bloom

Viruses are key biogeochemical engines in the regulation of the dynamics of phytoplankton. However, there has been little research on viral communities in relation to algal blooms. Using the virMine tool, we analyzed viral information from metagenomic data of field dinoflagellate (Gymnodinium catenatum) blooms at different stages. Species identification indicated that phages were the main species. Unifrac analysis showed clear temporal patterns in virioplankton dynamics. The viral community was dominated by Siphoviridae, Podoviridae, and Myoviridae throughout the whole bloom cycle. However, some changes were observed at different phases of the bloom; the relatively abundant Siphoviridae and Myoviridae dominated at pre-bloom and peak bloom stages, while at the post-bloom stage, the members of Phycodnaviridae and Microviridae were more abundant. Temperature and nutrients were the main contributors to the dynamic structure of the viral community. Some obvious correlations were found between dominant viral species and host biomass. Functional analysis indicated some functional genes had dramatic response in algal-associated viral assemblages, especially the CAZyme encoding genes. This work expands the existing knowledge of algal-associated viruses by characterizing viral composition and function across a complete algal bloom cycle. Our data provide supporting evidence that viruses participate in dinoflagellate bloom dynamics under natural conditions.


Introduction
Phytoplankton are important primary producers in marine ecosystems and play a critical role in element cycling, food web production, and broader ecosystem functions [1][2][3]. Over the last several decades, eutrophication, temperature increases due to global climate change, and modifications of the food web have been associated with an increased frequency of phytoplankton blooms with toxic or other deleterious properties. In recent years, worldwide increases have been reported in the occurrence, geographic expansion, and persistence of harmful algal blooms (HABs) [4][5][6], with severe impacts on human health and ecosystems, mainly through the release of toxins and the production of dissolved organic matter [7]. During the early part of the twenty-first century, urbanization has expanded, particularly in coastal and developing countries, making it essential to conduct intensive multidisciplinary research on HABs [5].
In recent years, increases in the frequency and global distribution of HABs has prompted researchers to investigate their triggers and formation mechanisms because it is crucial to understand each sample was preserved in 1% glutaraldehyde to quantify total bacterial abundance. In addition, 50 mL of each sample was preserved with 1% Lugol's iodine for the identification and quantification of phytoplankton. Algae were observed and counted under an optical microscope (×100 magnification).
In addition to the evaluation of each sample, physicochemical parameters of the seawater (temperature, salinity, pH value, and Chl α level) were recorded in vivo using a YSI 6920 water-quality sensor (Xylem Inc., Yellow Springs, OH, USA). Other environmental factors, including levels of ammonium nitrogen (NH 4 + ), nitrate nitrogen (NO 3 -), nitrite nitrogen (NO 2 -), and phosphate phosphorus (PO 4 3-), were measured according to the methods of Murphy [42] and Greenberg et al. [43].

Collection of Environmental DNA
The seawater samples were initially passed through 10 µm filter to remove particles or cells. For enrichment of marine viruses, we used FeCl 3 -mediated flocculation, as described previously [44]. In brief, we added 6 mL of 1% aqueous FeCl 3 into 60 mL of seawater and mixed it thoroughly. The sample was then incubated at room temperature for at least 1 h. The incubated seawater was then passed through a polycarbonate membrane filter (0.22 µm, Millipore, Billerica, MA, USA) and stored at 4 • C until analysis. The filtered membranes, including viral particles and microbial cells, were collected for DNA extraction. Total DNA was extracted using the MoBio PowerWater Viral DNA/RNA Isolation Kit (MoBio laboratories, Carlsbad, CA, USA) according to the manufacturer's instructions. The viral DNA was amplified by Qiagen REPLI-g Single Mini Kit (Qiagen, Hilden, Germany) using Phi29 DNA polymerase. The viral clone library was constructed using the whole-genome multiple displacement amplification (MDA) method. The sequencing library for next-generation sequencing (NGS) was prepared using NEBNext ® UltraTM DNA Library Prep Kit (New England Biolabs, Ipswich, MA, USA). Adapter-ligated fragments were then amplified using PCR and purified by gel electrophoresis. The DNA library was pair-end sequenced using Illumina HiSeq 2000 by MAGIGENE Biotech. Co. Ltd. (Guangzhou, China).

Metagenomic Sequencing
Among the 11 time-point samples obtained during the HAB event, we selected 8 for metagenomic analyses: YT1, YT3 (pre-bloom), YT4-YT6 (during bloom), and YT8, YT10-YT11 (post-bloom). DNA extraction was performed using the aforementioned methods. Purified genomic DNA was sheared to 150-200 bp using the Covaris S220 (Woburn, MA, USA) acoustic system. Metagenomic libraries were prepared using the NuGen Encore multiplex systems I and IB (NuGen, Carlsbad, CA, USA). The manufacturer's instructions were followed for end repair, adapter ligation, amplification, and barcoding, with unique DNA barcodes flanking the adapters for each sample. The metagenomes were sequenced using the next-generation sequencing platform Illumina HiSeq 2000 (MAGIGENE Biotech. Co. Ltd., Guangzhou, China).

Data Analysis
For environmental parameters, SPSS version 13.0 was used to perform statistical analysis. Then, one-way analysis of variance tests was conducted to evaluate the differences between different HAB stages, and adjusted p-values (0.05 or 0.01) were determined. For viral composition, the cluster and relationships between viral community composition and environmental parameters were investigated using correlation matrices (such as Pearson correlation) and Euclidean distances under R software (www.r-project.org). Inter-species associations between virus and target species (G. catenatum) were examined using a variance ratio (VR) test [45]. To visualize the results, a network was created to identify inter-species associations and correlations among species [46].
Raw sequencing reads underwent quality trimming using Trimmomatic to remove adaptor contaminants and low-quality reads [47]. Read counts were normalized according to Yang et al. [48], dividing the raw-read count by the total number of reads in the sample and by the (average) genome size. The abundance of each gene within a sample was calculated as follows: where a i represents relative abundance of (i) gene; L i represents the length of (i) gene sequence; x i represents the number of (i) gene-mapping reads; b i represents the copy number of (i) gene; and j represents the number of all of reads. Uncontaminated high-quality reads (clean reads) were used for subsequent analysis. Clean reads were then independently assembled to scaffolds for each sampling site using SPAdes [49]. The resulting scaffolds were merged, and redundancy was removed by CD-hits at 90% identity. MaxBin 2.0 (http://downloads.jbei.org/data/microbial communities/MaxBin/MaxBin.html, Version 2.2.4) was used to perform scaffold binning by adopting the following conditions: (1) assembled scaffolds had lengths >1000 bp; (2) tetranucleotide frequencies and coverage level of scaffolds were calculated; (3) the parameters were substituted into the expectation maximization (EM) algorithm to calculate the probability that all scaffolds belong to each bin. Subsequently, the contigs were uploaded to virMine to perform automated analysis and annotation. Open reading frames (ORFs) were predicted from the scaffolds. To investigate the functional diversity of HABs viruses, clusters of orthologous group (COG) annotation of viromes was determined by Blastp comparisons of predicted ORFs with the eggNOG database (http://eggnog.embl.de/) with an E-value of 1 × 10 −5 . ORFs affiliated to COG function classes of carbohydrate transport and metabolism were further clustered to generate unique carbohydrate metabolic ORFs with CD-hits. Subsequently, CAZymes from these viral ORFs were identified on the dbCAN web server based on CAZyme family-specific HMMs [50]. ORFs related to carbohydrate metabolism and CAZymes were compared with NCBI nr and Pfam databases to determine the most accurate annotation and similarity for each ORF. For map analysis, scaffolds containing CAZymes were retrieved, and ORFs were identified with Meta-Gene. The ORFs were then compared with NCBI for functional and taxonomic annotation based on protein sequences.

Data Availability
The raw data in this study have been deposited in the NCBI database under accession number SRP158562. In addition, the analyzed data of viral scaffolds can be found in Supplementary File 1.

Algal Bloom Characteristics
During this algal bloom, the target species was G. catenatum, and cell concentrations ranged from (0.51 ± 0.042) × 10 2 to (5.4 ± 0.61) × 10 3 cells/mL, with the highest biomass appearing at the peak stage (p < 0.01). Five phases were documented over the course of the bloom's cycle: pre-stage (YT1 and YT2), exponential growth stage (YT3 and YT4), peak-stage (YT5 and YT6), decline-stage (YT7-YT9), and terminal-stage (YT10 and YT11). During the HAB sampling period, temperature, salinity, and pH ranges were 27.5-28.7 • C, 27.8%-34.6%, and 7.36-8.52, respectively. There was distinct temporal heterogeneity in the main nutrient concentrations (NO 3 -, NO 2 -, NH 4 + , and PO 4 3-) throughout the HAB, and the lowest values appeared at the bloom's peak. Detailed information is given in our previous work [51]. The changes in these parameters reflect the basic ecological characteristics of the algal bloom, indicating the typical dynamic characteristics of dinoflagellates [52].

Metagenomic Data
We got 242.5 gigabases of raw reads (mean of 31,051,745 per sample) from the eight metagenomic samples. Details of the data can be found in a previous work [51]. After primary assembly, a total of 15,472,449 scaffolds (>200 bp) were obtained in the test samples. We selected the scaffolds (>1000 bp) that were predicted as viral and queried them via megablast against the NCBI nr/nt database online. A total of 428,502 scaffolds were compared to the nt database, resulting in the identification of 25,263 viral scaffolds, which were taxonomically assigned as viral sequences.

Taxonomic Composition
The most abundant viral communities were double-stranded DNA (dsDNA) viruses (73.08%), with a relatively lower proportion (<1%) of ssDNA and RNA virus communities. Most of the dsDNA groups of viruses were bacteriophages, which reflects the high abundance and diversity of bacterial hosts during the algal bloom. In addition, the most dominant dsDNA viruses were from the order Caudovirales. Previous viral metagenome studies on marine environments have also found a high abundance of Caudovirales [53,54]. Our analysis indicated that the relative abundance of Caudovirales increased from the pre-bloom stage (YT1 and YT3) to the growth and peak stages (YT4 to YT6); it reached its highest values at YT5 (50.6%) and subsequently decreased during the post-bloom stage (YT8, YT10, and YT11) ( Figure 1). This trend suggests that the relative abundance of Caudovirales has distinct alterations in different algal bloom stages. In addition, Caudovirales has been reported to infect a wide range of bacterial hosts, especially species in the Proteobacteria and Bacteroidetes phyla, which are dominant taxa in blooms [51]. Bacterioplankton are important microbial components of food-web fluxes [39], and they have been associated with natural plankton populations [55]. Therefore, it is possible that Caudovirales potentially or indirectly affects the structure of the plankton community during blooms.
bloom, indicating the typical dynamic characteristics of dinoflagellates [52].

Metagenomic Data
We got 242.5 gigabases of raw reads (mean of 31,051,745 per sample) from the eight metagenomic samples. Details of the data can be found in a previous work [51]. After primary assembly, a total of 15,472,449 scaffolds (>200 bp) were obtained in the test samples. We selected the scaffolds (>1000 bp) that were predicted as viral and queried them via megablast against the NCBI nr/nt database online. A total of 428,502 scaffolds were compared to the nt database, resulting in the identification of 25,263 viral scaffolds, which were taxonomically assigned as viral sequences.

Taxonomic Composition
The most abundant viral communities were double-stranded DNA (dsDNA) viruses (73.08%), with a relatively lower proportion (<1%) of ssDNA and RNA virus communities. Most of the dsDNA groups of viruses were bacteriophages, which reflects the high abundance and diversity of bacterial hosts during the algal bloom. In addition, the most dominant dsDNA viruses were from the order Caudovirales. Previous viral metagenome studies on marine environments have also found a high abundance of Caudovirales [53,54]. Our analysis indicated that the relative abundance of Caudovirales increased from the pre-bloom stage (YT1 and YT3) to the growth and peak stages (YT4 to YT6); it reached its highest values at YT5 (50.6%) and subsequently decreased during the post-bloom stage (YT8, YT10, and YT11) (Figure 1). This trend suggests that the relative abundance of Caudovirales has distinct alterations in different algal bloom stages. In addition, Caudovirales has been reported to infect a wide range of bacterial hosts, especially species in the Proteobacteria and Bacteroidetes phyla, which are dominant taxa in blooms [51]. Bacterioplankton are important microbial components of food-web fluxes [39], and they have been associated with natural plankton populations [55]. Therefore, it is possible that Caudovirales potentially or indirectly affects the structure of the plankton community during blooms. At the family level, the overall viral community was primarily represented by Siphoviridae, Podoviridae, Myoviridae, Phycodnaviridae, Mimiviridae, and Microviridae ( Figure 2). As the bloom developed, significant changes were observed in the viral community. The relative abundance of Podoviridae showed a declining trend, from 32.8% ± 0.6% in the pre-bloom stage to 15.1% ± 7.6% in the peak stage, followed by a mild increase during the post-bloom stage (21.4% ± 5.6%). The patterns observed among the Myoviridae and Siphoviridae were somewhat different, with a gradual increase over time and a maximum proportional abundance appearing at the peak stage, with approximately 21.9% ± 11.1% and 50.5% ± 24.6% of the total OTU tags, respectively. Similarly, the abundance of the At the family level, the overall viral community was primarily represented by Siphoviridae, Podoviridae, Myoviridae, Phycodnaviridae, Mimiviridae, and Microviridae ( Figure 2). As the bloom developed, significant changes were observed in the viral community. The relative abundance of Podoviridae showed a declining trend, from 32.8% ± 0.6% in the pre-bloom stage to 15.1% ± 7.6% in the peak stage, followed by a mild increase during the post-bloom stage (21.4% ± 5.6%). The patterns observed among the Myoviridae and Siphoviridae were somewhat different, with a gradual increase over time and a maximum proportional abundance appearing at the peak stage, with approximately 21.9% ± 11.1% and 50.5% ± 24.6% of the total OTU tags, respectively. Similarly, the abundance of the other members (such as Phycodnaviridae and Microviridae) showed a dramatic increase during the G. catenatum bloom and were enriched (with abundances of more than 10%) at the end of bloom. Siphoviridae, Podoviridae, and Myoviridae were the major members, which is not surprising because these species have been found to be the most abundant groups of viruses in the aquatic environment [54,56,57]. Previous metagenome investigations of environmental samples have shown ubiquitous dsDNA eukaryotic viruses (such as Iridoviridae, Phycodnaviridae, Mimiviridae, Poxviridae, and Polydnaviridae) in the DNA viromes. In our study, only Phycodnaviridae and Mimiviridae were detected, and they accounted for a relatively low proportion of the whole. The main reason for this was likely our choice of sampling sites: Gong et al. [54] used samples from deep sea areas, whereas our samples were collected near shore in a phycosphere environment. The relatively higher abundance of Phycodnaviridae and Mimiviridae in this work indicates that they were the main species associated with G. catenatum and perhaps participated in the decline of algal bloom [58][59][60][61]. In addition, we identified an extremely small fraction (0%-3.0%) of Microviridae, as the sole ssDNA viruses. This low level of representation may reflect differences in the abundances of these viruses across bloom stages. However, many new ssDNA viral genomes have been assembled from environmental samples from Saanich Inlet, the Strait of Georgia, and the Gulf of Mexico [62]. This suggests that there exist a great number of ssDNA viruses in the ocean that are as yet unrecognized [63]. In future work, researchers should improve the methods of amplification and increase sequencing depths to mine information on ssDNA viruses in algal bloom environments. catenatum bloom and were enriched (with abundances of more than 10%) at the end of bloom. Siphoviridae, Podoviridae, and Myoviridae were the major members, which is not surprising because these species have been found to be the most abundant groups of viruses in the aquatic environment [54,56,57]. Previous metagenome investigations of environmental samples have shown ubiquitous dsDNA eukaryotic viruses (such as Iridoviridae, Phycodnaviridae, Mimiviridae, Poxviridae, and Polydnaviridae) in the DNA viromes. In our study, only Phycodnaviridae and Mimiviridae were detected, and they accounted for a relatively low proportion of the whole. The main reason for this was likely our choice of sampling sites: Gong et al. [54] used samples from deep sea areas, whereas our samples were collected near shore in a phycosphere environment. The relatively higher abundance of Phycodnaviridae and Mimiviridae in this work indicates that they were the main species associated with G. catenatum and perhaps participated in the decline of algal bloom [58][59][60][61]. In addition, we identified an extremely small fraction (0%-3.0%) of Microviridae, as the sole ssDNA viruses. This low level of representation may reflect differences in the abundances of these viruses across bloom stages. However, many new ssDNA viral genomes have been assembled from environmental samples from Saanich Inlet, the Strait of Georgia, and the Gulf of Mexico [62]. This suggests that there exist a great number of ssDNA viruses in the ocean that are as yet unrecognized [63]. In future work, researchers should improve the methods of amplification and increase sequencing depths to mine information on ssDNA viruses in algal bloom environments.  [56] found that Pelagibacter phage HTVC010P, Ostreococcus lucimarinus OIV5/OIV1, and Roseobacter phage SIO1 were the most common species in Goseong Bay. Gong et al. [54] suggested that Cellulophaga phages, Pseudomonas phages, and Vibrio phages were responsible for the high percentage of surface water viromes in Prydz Bay. In Chile Bay, an integrative omics study demonstrated that the phage community associated with bacterioplankton was dominated by a new Pseudoalteromonas virus (PpCBA) during a diatomdominated phytoplankton bloom [64]. As the main driver, bacteria play multiple ecological roles in phytoplankton blooms, killing hosts, degrading biopolymers, and recycling algae-derived organic matter. The high abundance of phages in the phycospheric environment may act as a potential factor to affect bacterial composition and eventually contributes to the initiation and termination of algal blooms.  [56] found that Pelagibacter phage HTVC010P, Ostreococcus lucimarinus OIV5/OIV1, and Roseobacter phage SIO1 were the most common species in Goseong Bay. Gong et al. [54] suggested that Cellulophaga phages, Pseudomonas phages, and Vibrio phages were responsible for the high percentage of surface water viromes in Prydz Bay. In Chile Bay, an integrative omics study demonstrated that the phage community associated with bacterioplankton was dominated by a new Pseudoalteromonas virus (PpCBA) during a diatom-dominated phytoplankton bloom [64]. As the main driver, bacteria play multiple ecological roles in phytoplankton blooms, killing hosts, degrading biopolymers, and recycling algae-derived organic matter. The high abundance of phages in the phycospheric environment may act as a potential factor to affect bacterial composition and eventually contributes to the initiation and termination of algal blooms. Figure 4 shows the associations between viral composition and environmental parameters. Among the physical factors, the dominant member was temperature, which indicates that temperature is a major abiotic force that shapes the dynamics of viral community dynamics. Among the chemical factors, multiple inorganic nutrients ( [65], and demonstrate that the nutrient-mediated succession of the virioplankton community contributes to bottom-up viral composition. Teeling et al. [66] found that material resources trigger succession of microbial populations; based on this, we speculated that community succession (like virus) in algal blooms is also regulated by substrate availability. However, more ecological data will be needed to confirm this in future work. Although abiotic factors generally control viral communities, biotic correlations may also play a role. In addition, in order to investigate the contribution of viral interactions to the host, we constructed a network based on correlations between the predominant taxa and the target species, G. catenatum (Figure 5). At the genus level, there were forty-four correlations (twenty-seven negative and fourteen positive) between the viral phenotypes and target algae (G. catenatum). G. catenatum was strongly positively correlated with Pelagibacter phage HTVC008M, Flavobacterium phage FL1, and Bathycoccus sp. RCC1105 viruses in the network (p < 0.01), and negatively correlated with Synechococcus phage S-EIV1, uncultured Mediterranean phage uvDeep-CGR1-KM17-C101, and CGR2-KM23-C896, among others (p < 0.01). This multiplicity of positive inter-species associations may have provided an opportunity for species to adapt to the surrounding environment and contribute to the duration of algal bloom [67]. By contrast, the negative correlations reflect the appearance of kill or lyse algal cells. Lysis activity related to bloom-forming species, such as Emiliania huxleyi, Phaeocystis pouchetii, and Phaeocystis globose, has been well studied [38,64]. During such blooms, viruses exhibit a strong regulation activity and contribute to HAB collapse in a boom-and-bust pattern [68][69][70]. Similarly, we observed a decline in G. catenatum at the post-HAB stage, which can be explained with the "killing the winner" hypothesis [71], wherein viral lysis of the most abundant algae releases new nutrients and opens a niche that other algal species then compete for.   Figure 4 shows the associations between viral composition and environmental parameters. Among the physical factors, the dominant member was temperature, which indicates that temperature is a major abiotic force that shapes the dynamics of viral community dynamics. Among the chemical factors, multiple inorganic nutrients (NO2 − , NO3 − , NH4 + , and PO4 3− ) co-contributed to the variability of viral communities. Specifically, Roseobacter phage SIO1, Flavobacterium phage FL-1, and Verrucomicrobia phage P8625 were positively correlated with NO2 − and NO3 − and were negatively correlated with PO4 3− ; Prochlorococcus phage Syn1 and Pseudoalteromonas phage BS5 were positively correlated with PO4 3− and NO2 − to some extent, but NH4 + had a negative influence; and in uncultured can be explained with the "killing the winner" hypothesis [71], wherein viral lysis of the most abundant algae releases new nutrients and opens a niche that other algal species then compete for.   can be explained with the "killing the winner" hypothesis [71], wherein viral lysis of the most abundant algae releases new nutrients and opens a niche that other algal species then compete for.

Functional Prediction of Algal Bloom Sample Viromes
Given many viral species are uncultured in the laboratory, the reference genome database for viral metagenomic analysis is still far from complete, and an extraordinary amount of uncharacterized viral "dark matter" continues to exist in seawater. For this reason, there are many sequences that remain without an assigned function [72][73][74]. Most recognizable functions were related to conventional viral functions, such as replication, recombination and repair, intracellular trafficking, secretion, vesicular transport, biogenesis of cell wall/envelope/membrane, DNA/RNA transport and metabolism, as well as transcription (Figure 6a), which are critical for viral reproduction and survival. According to the annotation results from the CAZy database, the ORFs of 43 auxiliary CAZyme genes are mostly affiliated with glycoside hydrolase (GH) and glycosyl transferase (GT) (Figure 6b), implying the importance of these functional genes for the degradation of marine organic carbons. Actually, complex carbohydrates, such as starch, xylan, cellulose, chitin, alginate, mannan, and pectin, are major components of algae cell walls and microbial intercellular spaces, and they are highly difficult to degrade [75]. Notably, in viruses, multiple genes encode CAZymes, such as hydrolysis enzymes and auxiliary enzymes, which are necessary for the degradation of various extracellular polymeric substance (EPSs) [76], indicating viral abilities in the biolysis of complex polysaccharides during bloom events. In this study, we found most viral carbohydrate metabolic genes belonged to CAZymes with glycoside hydrolase activities, which allow viruses to primarily participate in the decomposition of organic carbon during algal bloom events. To better elucidate the distribution of CAZymes at different bloom stages, the relative abundance of GH and GT was investigated (Figure 6c). The relative abundance of GH significantly increased (from YT01 to YT04), reaching peak abundance (>75%) at YT04, and then gradually increased from the later during-bloom stages to the post-stages. The trend of the relative abundance of GT was opposite to that of algal bloom formation and collapse. It seems that this big variation in the relative abundances of GH and GT was associated with bloom dynamics. In order to identify the potential contributions to carbon degradation, the relative contributions of viral species to GH and GT genes were performed. Figure 6d shows that genes encoding components of the GH enzyme were mostly affiliated with the uncultured Mediterranean phage, Celeribacter phage, Flavobacterium phage, and Roseobacter phage; whereas genes in the GT proteins mostly belonged to uncultured Mediterranean phage, Verrucomicrobia phage P8625, Vibrio virus nt1, and uncultured marine virus (Figure 6d). In particular, the uncultured Mediterranean phages co-contributed to GH and GT genes, suggesting that these viruses are main contributors to organic carbon decomposition [77]. Viral "auxiliary metabolic genes" (AMGs) for carbon metabolisms have been extensively investigated in marine environments; most of them are involved in central carbon metabolism to facilitate viral replication [78]. In this work, we found various viruses associated with GH/GT genes, and we speculated that virus-encoded GHs/GTs can act as AMGs and increase host accessibility through producing more viruses [79]. Recent work proposed that virus-encoded GHs potentially augment the breakdown of complex carbohydrates to increase energy production and boost host metabolism during viral infection [80,81]. These results hint that viruses promote host infection and induce HAB decline.

Conclusions
This study surveyed viral communities in a natural dinoflagellate G. catenatum bloom and found that the viruses exhibited remarkable heterogeneity in their responses to algal bloom events. Viral community structures were affected by multiple factors, including physiochemical parameters, substance availability, and interactions. A multitude of viral interactions (positive and negative) are likely involved in determining a given algal bloom fate at a specific time and location. Functional predictions indicated that the viruses exhibited glycoside hydrolase activities, facilitating the understanding of the primary roles of viruses in the degradation of organic carbon. Detailed comparisons of the metabolic characters of phytoplankton and viral communities should be performed to enable a better understanding of the ecological mechanisms of HABs from a virological perspective.
Supplementary Materials: The following are available online at www.mdpi.com/xxx/s1.

Conclusions
This study surveyed viral communities in a natural dinoflagellate G. catenatum bloom and found that the viruses exhibited remarkable heterogeneity in their responses to algal bloom events. Viral community structures were affected by multiple factors, including physiochemical parameters, substance availability, and interactions. A multitude of viral interactions (positive and negative) are likely involved in determining a given algal bloom fate at a specific time and location. Functional predictions indicated that the viruses exhibited glycoside hydrolase activities, facilitating the understanding of the primary roles of viruses in the degradation of organic carbon. Detailed comparisons of the metabolic characters of phytoplankton and viral communities should be performed to enable a better understanding of the ecological mechanisms of HABs from a virological perspective.