香京julia种子在线播放

The field study was conducted during the R/V Hakuho-Maru KH-20–9 cruise from 10 September to 5 October 2020. The sampling sites (34 stations) covered the Kuroshio region from the ECS off Okinawa to the south of Honshu, Japan ( Figure 1 ). All six transects were designed to cross the Kuroshio axis or the Kuroshio meander. In order to set the sampling stations, sea surface height during the cruise was acquired from E.U. Copernicus Marine Service Information ( Figure 1 ; Global Ocean Gridded SSALTO/DUACS Sea Surface Height L4 product, https://resources.marine.copernicus.eu/products).

At each station, Conductivity-Temperature-Depth (CTD) system (Sea-Bird Electronics 911plus CTD) was deployed to collect hydrographic data. Water samples were collected with Niskin-X bottles (General Oceanics, USA) equipped on a CTD-carousel system from 10 m, 30 m, 40 m, 50 m, 60 m, 80 m, 100 m, 125 m, 150 m, 200 m, 300 m, 400 m, 600 m, 800 m, and 1,000 m or at 50 m above seafloor if the depth is shallower than 1,000 m. Underwater light penetration was measured at daytime sampling stations once per day by deploying the radiometer PRR-800 with the radiometer PRR-810 on deck (Biospherical Instruments Inc.) to determine the depth of euphotic zone (EZ; 1% irradiance level relative to sea surface). To examine nitrate (NO₃ ⁻), nitrite (NO₂ ⁻), ammonium (NH₄ ⁺), and phosphate (PO₄ ³⁺) concentrations, water collected in Niskin-X bottles was transferred to 12-mL polycarbonate tubes in duplicate and stored in gastight aluminum bags at −30°C until measuring. Water samples for measuring chlorophyll-a (Chl-a) concentration were collected down to 200 m. For metabarcoding analysis, water samples (~10 L) were collected at 10 m, 50 m, 100 m, and 150 m of all stations except Sta. OK2 and OK4 and filtered through a Sterivex filter (0.45 μm, Millipore SVHV010RS, Merck, Tokyo, Japan). Once the filtration was completed, 2 mL of RNAlater was added into each Sterivex cartridge and stored in a deep freezer at −70°C until DNA extraction.

Ocean Data View 5.5.2 was used to create the map of the studying area, temperature-salinity (TS) diagrams. The vertical profiles of nutrients and Chl-a were created using ggplot2 3.3.5 package in R (Wickham, 2016). If a nutrient concentration was below its detection limit, zero was assigned. The potential density anomaly was calculated in the Ocean Data View 5.5.2. The mixed layer depth (MLD) was defined as where the potential density anomaly is 0.125 kg m⁻³ greater than the surface.

Right before DNA extraction, each Sterivex filter was thawed on ice, and RNAlater inside was removed by aspiration. DNA was extracted following the method in Wong et al. (2020) using a Qiagen DNeasy Blood and Tissue DNA extraction kit. The eluate was stored at −25°C. DNA of four dinoflagellate cultures isolated during the expedition and growing in 1/10 x Daigo’s IMK medium was extracted following the same procedure to create a positive control.

Intended to cover as much dinoflagellate species as possible, universal primer sets TAReuk454FWD1 and TAReukREV3 (Stoeck et al., 2010) and 1389F and 1510R (Amaral-Zettler et al., 2009) were used to amplify the hypervariable V4 and V9 regions of eukaryotic 18S rDNA of extracted DNA samples, respectively ( Supplementary Table 4 ). A two-step PCR was performed using 15 μL of PCR systems with 10 ng to 30 ng of DNA, 0.45 μL of forward and reverse primers (10 μM), Toyobo KOD -Plus- ver. 2 enzyme and reagents from Toyobo KOD -Plus- ver. 2 kit (Toyobo, Japan). The first step amplified the desired regions with the Illumina Paired-End adapters attached, started with an initial activation at 94°C for 2 min, followed by 30 cycles of 10 s at 98°C, 30 s at 56°C, and 60 s at 68°C, and ended at 68°C for 7 min. The second step added indexes allowing sample multiplexing in one Illumina MiSeq run, starting at 94°C for 2 min, followed by eight cycles of 10 s at 98°C, 30 s at 59°C, and 60 s at 68°C, and then final elongation at 68°C for 7 min. Negative controls without the addition of DNA and one positive control (only V4 region amplified) were made. Products of each PCR step were checked on electrophoresis gels.

The PCR products were pooled by adding 100 ng of PCR products of each sample and controls. Because the concentration of the V4 negative control was too low, only the culture mix was pooled. The pooled samples were sent to the commercial sequencing company FASMAC (Atsugi, Japan) and run separately on the Illumina MiSeq platform. The pooled V4 sample was run on one lane of a pair-end 2× 300-bp sequencing run, and the pooled V9 sample was run on a lane of a pair-end 2× 250-bp sequencing run.

The concentrations of NO₃ ⁻, NO₂ ⁻ NH₄ ⁺ and PO₄ ³⁺ were measured using absorption spectrophotometry (Strickland and Parsons, 1972) adapted for an auto-analyzer (QuAAtro, BL-Tec). The detection limits for NO₃ ⁻, NO₂ ⁻, NH₄ ⁺, and PO₄ ³⁺ were 0.05 μM, 0.07 μM, 0.05 μM, and 0.01 μM, respectively. Chl-a samples were filtered on board through Whatman GF/F filters under low vacuum pressure (<0.02 MPa). The filters were soaked in 5 mL of N,N-dimethylformamide and stored at −20°C in dark for 24 h for extracting pigment (Suzuki and Ishimaru, 1990). Chl-a concentration was then measured with a Turner Designs fluorometer on board following a non-acidification method (Welschmeyer, 1994).

The raw sequence data (refer to the Data availability statement section) were processed to generate absolute sequence variant (ASV) table of amplified eukaryotic V4 and V9 regions using the DADA2 1.20.0 package in R Studio based on the standard procedure (Callahan et al., 2016). The quality profiles of reads were checked using FASTQC prior to trimming (Andrews, 2010). Taxonomic assignation of generated ASVs was performed using the default setting of function “assignTaxonomy” in DADA2 against the DADA2 trained Protist Ribosomal Reference (PR²) database ver. 4.14.0 (Guillou et al., 2013) that includes a curated 18S rDNA reference database of dinoflagellate (Mordret et al., 2018). The output data comprising of ASV tables and taxonomic annotation were merged with environmental data of sampling stations and directly input to Phyloseq 1.36.0 package in R Studio (McMurdie and Holmes, 2013). The ASV tables were first cleaned by excluding these detected in the negative control, which could be contamination during the experiment. Then, only ASVs belonged to the phylum Dinoflagellata were extracted to form new ASV tables. The new tables were again cleaned by excluding ASVs annotated as symbiotic and parasitic dinoflagellates ( Supplementary Table 1 ). Read depth was standardized to median read depth of samples in Phyloseq, and data artifacts (ASV with a total read count < 0.1% of the sample sum) were excluded from the standardized ASV tables. The ASV will be referred as metabarcodes hereinafter.

For analyzing the relation between species and environmental variables, read counts of metabarcodes assigned to the same taxon at the species level were merged to create a new species read count table using tax_glom command in Phyloseq. The trophic strategy of species detected in both V4 and V9 metabarcodes were assigned on the basis of published literatures and databases (refer to Supplementary Table 2 ). For simplification, MTD is defined as plastid-containing species with observed predation behavior or food vacuoles in this study.

Relative abundance of metabarcodes was calculated as a Phyloseq object using “transform_sample_counts” function or directly in R Studio. To visualize the relative abundances, the values were log10-transformed for plotting heatmaps of species detected by both V4 and V9. Where the relative abundance was zero, 0.01 was assigned. Metabarcode α-diversity indices and Pielou’s evenness index were calculated using Vegan 2.5–7 package in RStudio (Oksanen et al., 2020). Analysis of variance (ANOVA) test was performed followed by Tukey multiple comparisons of means to understand the difference in metabarcode diversity between groups.

Prior to the multivariate analysis, standardized read counts of dinoflagellate metabarcodes were Hellinger-transformed in Vegan. Then, Bray-Curtis–based non-metric multidimensional scaling (NMDS) was calculated using Vegan to examine regional dissimilarity, and detrended correspondence analysis (DCA) was used to explore the correlation between the diversity of dinoflagellate communities and environmental variables in Vegan, as a unimodal fit was expected. Once significant variables identified by p-value, the relation between these variables and the distribution of species annotated in both V4 and V9 was analyzed by canonical correlation analysis (CCA) in Vegan. From species read count table, a species occurrence data was constructed by assigning 1 (present) or 0 (absent) to each species annotated. To assess niche of each species in the study area, we applied an Outlying Mean Index (OMI) analysis using the Subniche 1.5 package in R, which is based on the occurrence data and PCA results of environmental condition (Dolé Dec et al., 2000; Karasiewicz et al., 2017). The OMI is the index of marginality that measures the distance between the average habitat conditions of one species and the average conditions of the whole sampling region. Here, OMI would be used to relate the distribution to the sampling station categories. To know the preferred habitats of each species, the coordinates on the first two factorial axes were plotted with the polygons of environmental spaces based on PCA analysis. The above analysis was visualized using ggplot2 3.3.5 package in R (Wickham, 2016).

It has been well demonstrated that the Kuroshio took bimodal paths off southern coast of Honshu, Japan (Kawabe, 1995). The Kuroshio flows easterly along the continental shelf approximated to 33°N, but the path left the coast and directed to the south approximated to 30°N and then returned to the north, forming a large meander path (LMP) ( Figure 1A ). When the expedition started, the LMP was observed off Honshu around 137°E, and the southern edge reached to 29°30̬′N ( Figure 1A ; Supplementary Video 1 ). At where the path met the M transect (Sta. M1 to M8), it started extending southward from 14 September and finally separated from the Kuroshio axis around 28 September. As a result, a cyclonic eddy was formed ( Figure 1B ). The sampling along the M transect started from 30 September after the formation of the cyclonic eddy. We set Sta. M4 at the center of the eddy.

Water property in the study area exhibited subtropic water features of high temperature and salinity, having surface water temperature above 26°C and salinity over 34 ( Figure 2 ). Based on the indicator temperature of the Kuroshio axis at 200 m (Kawabe, 2003), which is 17°C in the ECS, 16°C in the east of the Kyushu, and 15°C in the south of the Honshu, most stations were positioned on the Kuroshio axis or the south of the axis. The exceptions were some stations on the continental shelf and inside the cyclonic eddy (Sta. M3 and M4) at which low salinity (<34.0) was observed. It was apparent that the water in the eddy was transported from coast, as the water property of Sta. M3 and M4 resembled it of Coastal stations, being colder and less saline comparing to other stations along the transect M ( Figure 2 ). From the water properties and formation of the cyclonic eddy, we grouped sampling stations into three categories: Coastal stations (Sta. OK1, OK1b, TA1, and TC1) characterized by cooler and less saline water, Eddy stations inside the eddy (Sta. M3 and M4) with similar water property to Coastal stations, and KC stations on the axis and south with warm and high salinity water ( Table 1 ). The mixed layer depth determined by a difference of 0.125 σθ respect to the sea surface was shallow at Coastal stations (21.2 ± 14.8 m) compared with one in the KC station (46.3 ± 14.3 m) ( Table 2 ). At Eddy stations, the bottom of mixed layer shoaled to 40 m at Sta. M3 and 28 m at Sta. M4, showing an uplift of subsurface water in the cyclonic eddy. The depth of euphotic zone (EZ) was from 32 m to 107 m ( Table 3 ). The shallowest was observed at coastal Sta. OK1 (32 m), followed by Eddy station M4 (56 m) and Coastal station TA1 (69 m). All the EZ of KC stations examined (range from 76 m to 107 m; mean ± SD, 92.1 ± 10.0 m) were deeper than those in the Coastal and Eddy stations.

Nutrients such as NO₃ ⁻, NO₂ ⁻ and PO₄ ³⁺ were depleted in EZ and increasing with depth ( Figure 3 ; Supplementary Data Sheet 1 ). Concentration of NH₄ ⁺ was below the detection limit at most stations throughout. Exceptions were stations at shallower shelf regions and inside the cyclonic eddy, where higher nutrient concentrations were observed in the EZ. The depth of Chl-a maximum was generally observed between 50 m and 100 m, except Coastal stations where it was above 50 m.

A total of 15,491 V4 and 17,963 V9 metabarcodes were generated by DADA2 from a total of 7,311,304 and 9,300,530 reads. After matching against the PR² database and filtering out data artifacts, 7,269 V4 and 5,110 V9 metabarcodes were annotated under the phylum Dinoflagellata. Both regions detected a high genetic diversity of parasitic Syndiniales, which accounted 62% V4 and 32% V9 dinoflagellate metabarcodes. The most abundant orders of Syndiniales were Dino-Group-I and Group-II ( Supplementary Figure 1 ). After filtering out metabarcodes of Syndiniales and other symbiotic or parasitic taxa ( Supplementary Table 1 ), 2,702 V4 and 3,402 V9 metabarcodes remained. Once the read depth was standardized, metabarcodes of less than 0.1% relative abundance was removed, and 1,571 V4 and 2,800 V9 metabarcodes remained. An average of 122.41 ± 49.57 metabarcodes per sample present in the V4 dataset and 170.98 ± 56.01 metabarcodes per sample in the V9 dataset. Among these, 610 V4 and 274 V9 metabarcodes were annotated to 47 and 41 genera under eight orders, respectively. The taxa ended with one or more X such as the order Dinophyceae_X and Dinophyta_XX contained unknown Dinoflagellate sequences detected by contributors of PR² database, mostly from environmental samples. The most frequently detected order was Gymnodiniales, followed by Peridiniales ( Figure 4 ), and the most frequently detected genera were Gyrodinium, Karlodinium, Gymnodinium, Lepidodinium, and Tripos. Aligning results from V4 and V9 regions, a total of 30 dinoflagellate species were detected ( Figure 5 ; Supplementary Table 2 ). Most species occurred at the KC stations, and Gyrodinium fusiforme and Karlodinium veneficum prevailed at all stations. The average relative abundance of unknown V9 metabarcodes was 48.53 ± 12.97%, larger than it observed in the V4 region (25.57 ± 5.10%).

Because many V4 and V9 dinoflagellate metabarcodes could not be annotated to lower taxonomic levels, the genetic diversity that was assessed by Shannon diversity and Pielou’s evenness indices was calculated on the basis of the number of metabarcodes and standardized read counts of metabarcodes. While a general high genetic diversity and evenness were observed, the mean value of Shannon index of V4 metabarcodes (4.04) was significantly lower than one of V9 (4.66), as well as the means of Pielou’s evenness index (0.856 for V4 and 0.915 for V9 metabarcodes). The differences among depths and categories were illustrated on Figure 6 . Whereas Shannon diversity of V4 metabarcodes decreased with depth, that of V9 metabarcodes remained unchanged due to increasing evenness. Communities at Coastal and Eddy stations at 150 m were generally less diverse and even compared with KC stations. The ANOVA test verified the significant differences in genetic diversities among depths and station categories (p < 0.001, Table 4 ). The Tukey test showed that the differences were attributed to communities from surface (10 m and 50 m) and subsurface (100 m and 150 m), as well as from KC and Coastal stations ( Supplementary Table 3 ).

Metabarcode compositions of dinoflagellates between samples were compared by NMDS analysis ( Figure 7 ). Surface communities from 10 m to 50 m were highly similar with each other especially at the KC stations, and the exceptions were Coastal stations like Sta. OK1 and TA1. Subsurface communities at 100 m and 150 m were distanced from the surface communities and from each other, and their compositions varied noticeably more with depth, especially between station categories. These agreed the results of ANOVA and Tukey tests ( Table 4 ; Supplementary Table 3 ). Although facing alike environmental conditions, the subsurface communities at Coastal stations OK1, OK1b, and TA1 were similar between the Eddy stations, and the communities from subsurface layers (100 m and 150 m) of Sta. TC1 resembled to communities at KC stations.

To explore the dinoflagellate community structure and the influence of environmental factors, DCA analysis was conducted on the basis of metabarcodes and their read counts at each station ( Figure 8 ). Major differences in communities could be related to temperature and sampling depth, which was exhibited along the axis of DCA1, separating communities of different sampling depth. Surface communities of 10 m and 50 m again formed a tighter group comparing with subsurface communities of 100 m and 150 m, excepting those at Coastal stations. Coastal stations demonstrated their distinct community structures from KC stations along the axis of DCA2, which could be correlated to the higher nutrient concentrations at these stations. Among all environmental variables, NH₄ ⁺ was not significant by both V4 and V9 and NO₂ ⁻ was the least significant factor in DCA analysis of V4 datasets.

To further explore the relationship between dinoflagellate species and environmental factors, metabarcodes of dinoflagellate species detected by both V4 and V9 regions were extracted along with the numbers of read counts and the trophic strategy of each species ( Supplementary Data Sheet 1 ). Each species annotated present at an average of 13.23 ± 12.04 and 14.33 ± 12.50 stations based on V4 and V9 metabarcodes. Out of the 30 dinoflagellate species, nine are autotrophy (ATD), eight are heterotrophy (HTD), and 13 are mixotrophy (MTD).

The environmental variables proven to be significant (p < 0.001) by the DCA analysis were chosen for the constrained CCA analysis ( Figures 9A, B ). The correlation illustrated by V4 and V9 metabarcodes approximately agreed with each other. Over half of the 30 species were more correlated to warmer water, and they were more likely to appear and being abundant at the surface oligotrophic habitats. The abundance of most MTD species was also related to the oligotrophic and warm euphotic surface habitats. For ATDs, they tended to appear at warmer surface, although this trend was not that obvious in community structures of V9 metabarcodes. Appearance of HTDs was more corresponding to higher nutrient levels and lower Chl a concentration. Rare species were situated relatively far from the environmental vectors on the CCA plots. For example, Polykrikos schwartzii (MTD), which was the farmost species on both CCA plots along the CCA2 axis, only present at the surface of Sta. OK1 and OK1b ( Figures 5 , 9 ).

The realized niche of each species was plotted against the realized environmental space by each category and sampling depth. On the plots ( Figure 10 ), the origin of the OMI ordination represents the average habitat condition, which resembles the environment of the KC stations at 100 m with moderate levels of all nutrients, temperature, and salinity in our data set. The space of Coastal stations was boarder due to great environmental variance among Coastal stations, such as the lower NO₃ ⁻ and PO₄ ³⁻ concentrations at Sta. OK1b and TV1 comparing to Sta. OK1 and TA1 above 150 m ( Figure 3 ). The area of Eddy stations was mostly within Coastal stations due to similarly low salinity and higher nutrient levels.

Most species could be found within 10 m and 50 m of KC and Eddy stations and approximated to the average habitat condition. Within KC stations, the realized niche of species having same trophic strategy was generally closer. When moving into the Coastal and Eddy habitats on OMI plots, the realized niche of species became more distanced. Most MTDs were detected within 10 m and 50 m and had adjacent niche. Comparing to MTDs, a few ATDs were detected below 100 m, and HTDs lived below 50 m in general. The exception was Polykrikos schwartzii (HTD), which only appeared at 10 m of Sta. OK1 and OK1b; thus, its realized niche was at the edge of Coastal stations ( Figures 5 , 10 ). Because the occurrence of some species was different in V4 and V9 metabarcodes, the realized niches of these species determined by V4 and V9 were different consequently. Tripos furca is a cosmopolitan species and was detected at almost every station in V4 metabarcodes, but it was only present at 10 m of Sta. OK1 and TC1 in V9 metabarcodes ( Figure 5 ).

A total of 15 HAB species was detected by V4 but only nine of them were detected by V9 ( Supplementary Data Sheet 2 ). Most HAB species annotated present at KC and Eddy stations. Gonyaulax polygramma, Karlodinium veneficum, and Takayama pulchella, all associated to fish killing, appeared at almost every station by both of the V4 and V9 ( Figure 5 ).

Distribution of most HAB species was correlated to warmer surface habitats at 10 m and 50 m, and only two HAB species correlated to subsurface ( Figure 10 ). One was Takayama pulchella having a maximum relative abundance approximate to nutriclines of most KC stations, which were generally below 50 m ( Figures 3 , 5 ). The other was Amphidinium gibbosum that also had a realized niche between 50 m and 100 m. Some infrequent species only appeared at Eddy or KC stations, except the benthic species Amphidinium gibbosum (Murray et al., 2004) which only occurred at 50 m of Sta. TC1 and 100 m of Sta. TV4 ( Figures 3 , 5 , 10 ).

The relative abundances of HAB assemblages were not significantly greater at Coastal or Eddy stations comparing with KC stations ( Table 5 ). A lower number of HAB species were detected at those stations. No HAB species’ realized niche was solely correlated to the coastal environment by OMI analysis, and only the presence of few species, Gonyaulax polygramma, Karlodinium veneficum, and Takayama pulchella, at Coastal stations was confirmed by both V4 and V9. Margalefidinium (= Cochlodinium) polykrikoides, one of the reoccurring red tide species along Japan coast (Iwataki et al., 2008; Furuya et al., 2018), was not detected at Coastal stations ( Figure 5 ).

From the year of 2017, the offshore LMP had been observed till the beginning of the cruise (Sugimoto et al., 2021) ( Figure 1A ). Our sampling stations originally intended to cover several transects of the Kuroshio axis along the LMP from the ECS off Okinawa to the south of Honshu, Japan and offshore region south of the axis, including regions onshore and offshore. While moving onshore, a cold eddy was formed and separated from the Kuroshio axis ( Figure 1B ). This change in path greatly influences local marine environment, biogeography, and productivity in the area (Nakata and Hidaka, 2003).

Three station categories—Coastal, KC, and Eddy—were established as they were influenced by Kuroshio differently. All KC stations exhibited typical Kuroshio characteristics of warm and saline surface water. The water property of Coastal stations was influenced by the cold and less saline coastal water, especially above 200 m, and active mixing at Coastal stations was observed by irregular temperature and salinity. The unforeseen change of LMP brought great impact onshore, and some onshore stations such as Sta. CR1 and M8 were identified as KC stations, demonstrating that the influence of Kuroshio Current was more than its path displayed on the ocean surface topography ( Figure 1 ). The TS diagrams of Eddy stations were similar to that of Coastal stations and distinct from that of the KC stations ( Figure 2 ). At the center of the eddy (Sta. M4), the thermocline and halocline were observed at 29 m, suggesting that the original coastal water property and dinoflagellate community were kept inside the eddy.

The distinction among environmental characteristics of the three station categories was the nutrient concentration above 50 m ( Figure 3 ). At Kuroshio stations, the nutriclines were generally observed around 60 m to 100 m and nutrients were depleted above the nutricline, except these at the Tokara strait (e.g., Sta. TA2 and TC1) where strong turbulent mixing due to the shallow depth and sea mounts supplied nutrients to the surface (Nagai et al., 2021). At Coastal stations, nutriclines generally located at layers shallower than 50 m, and higher nutrient concentrations were observed at sea surface. Coastal stations were especially rich in NH₄ ⁺, whereas its concentration was below the detection limit (0.05 μM) at KC and Eddy stations. Similarly, the nutriclines were observed in layer shallower than 50 m in the eddy, and the NO₃ ⁻ and PO₄ ³⁺ concentrations at 50 m of Sta. M4 (3.91 and 0.30 μM, respectively) were similar to those at 100 m of the KC stations. As cyclonic eddy induces upwelling (Bibby et al., 2008; Uchiyama et al., 2017), this eddy may replenish nutrients from the uplifted thermocline. However, the uplift of thermocline and nutriclines did not reach to 10 m, and nutrients were depleted at 10 m in the eddy.

Dinoflagellate community compositions were localized horizontally and vertically in the study area regardless the station categories, and the dissimilarity was correlated greatly to sampling depth and nutrient levels ( Figures 7 , 8 ). Depth is the vital factor that corresponded to vertical temperature change and light gradient, separating communities of different sampling depth. The nutrient levels caused the variance of communities of different station categories, especially below 50 m. The formation of the tight group of surface KC and Eddy communities may be related to the well-mixed surface layers of the current, and the relatively stratified subsurface layers differentiated the subsurface communities. The vertical separation among surface and subsurface communities also appeared on the profiles of Shannon and Pielous’s indices, proven by the Turkey Test ( Supplementary Table 3 ). Similar vertical stratification of dinoflagellate and protist assemblages was observed in other ecosystems (Schnetzer et al., 2011; Cohen et al., 2021; Ollison et al., 2021; Carve et al., 2024). Carve et al. (2024) also mentioned that the wind-induced vertical mixing could lead to highly similar assemblages in the mixed surface layer. Irradiance could be another factor because, at 100 m of KC stations, it was being approximately 1% relative to surface, suggesting possible existence of autotrophy at this depth. However, the irradiance at 150 m of every station was too dim to support photosynthesis.

The levels of Chl-a and nutrients were mainly correlated to the separation of nutrient-rich Coastal stations from KC stations. At Eddy stations, the surface communities resembled the surface KC communities ( Figures 7 , 8 ) despite their similar water property to Coastal stations, suggesting potential influence of the water from Kuroshio axis. Below 100 m, the Eddy communities were similar to those at Coastal stations because the eddy trapped coastal water inside.

The dissimilarity among communities represents the variation in genetic diversity and evenness between communities. The diversity and evenness indices were generally higher at KC stations relative to Coastal and Eddy stations because more minor metabarcodes present at these stations ( Figure 6 ). Once metabarcodes were grouped by taxa, fewer taxa were observed at Coastal stations such as Sta. TA1. The ubiquitous species among the 30 species annotated demonstrated no preference to Coastal stations. A few rare species were specialized to local environment by their greater relative abundance at some Coastal stations. The most conspicuous example was Amphidinium gibbosum (ATD) whose only presence was relatively abundant at Sta. TC1 ( Figure 5 ). Others such as Abedinium dasypus and Protoperidinium tricingulatum, both HTDs that only present below 100 m in this study, were rare in the sampling area but slightly more abundant at Sta. TA1. It indicates a possible advantage that these species possessed in nutrient-rich environment. However, the physiology of rare species was hardly studies due to difficulties in sampling, identifying, and culturing, stressing the necessity of such knowledge in understanding their relation to environment.

Within communities, the environmental adaptation of the species detected was analyzed by means of OMI analysis. Despite the drawn polygon of Coastal stations appeared to contain other station categories, its broadness was due to the different environment in ECS and Tokara strait. Most species were found in the warm surface habitat of the KC and Eddy stations. A few species were shown to be specialized for deeper habitats such as Prorocentrum nanum (ATD) and Abedinium dasypus (HTD).

Most species annotated were photosynthetic ATD and MTD, and their correlation with shallower warm water was expected as a higher level of irradiance is required for photosynthesis. Most ATDs’ distribution correlated to temperature and had realized niches at 50 m. Alexandrium leei and Azadinium trinitatum are the examples. However, more MTDs were detected at 10 m than ATDs and appeared to be specialized for surface water, and the realized niche of MTDs was more aggregated compared to ATDs. Photoacclimation by adjusting Chl:C ratio is required to optimize photosynthesis and prevent photodamage, and nutrient limitation in surface water (above 10 m) could restrain the repairing ability of phytoplankton under strong irradiance (Litchman et al., 2002; Li et al., 2021). Phagotrophy could potentially help MTDs acquiring carbon without elevating Chl:C and adapting higher irradiance (Mitra et al., 2023). For MTDs possessing diverse prey types and varied surviving strategies regarding light and nutrient conditions (Hansen, 2011), it is possible that they occupied more adjacent realized niches in a habitat. Considering their abundance in the water column above 50 m, our results indicated a more complex and dynamic food web and support the hypothesis of MTDs enhancing trophic transfer efficiency (Flynn and Mitra, 2009; Ward and Follows, 2016; Stoecker et al., 2017). However, more in situ and laboratory experiments on their physiology and interaction with other protists are necessary to elucidate their roles in the food web.

Nonetheless, few rare ATDs were detected at deeper layers, such as the dinoflagellate Prorocentrum nanum, which had a realized niche at 100 m ( Figure 10 ). Because photosynthesis was almost impossible under 100 m, they could be either be cysts, an aggregate, or attach on macroalgae then sink below 100 m. It is disputable that they could possess unique physiology traits to survive in low light conditions unless they possess undiscovered mixotrophic ability, which helps them survive in the dark (Cohen et al., 2021).

The realized niche of HTDs varied largely among species. For example, of the two HTD species, Protoperidinium tricingulatum and Protoperidinium bipes, it was correlated to different environment variables. The first species often detected at 50 m and was more abundant at Coastal stations (Schneider et al., 2020). The other species was closely correlated to temperature because of its frequent appearance at 10 m and 50 m of oligotrophic KC and Eddy stations. However, when a species was rare and only detected at a few stations, its correlation with these environment variables is statistically unexplainable, hence locating far from the vectors.

The hypervariable V4 and V9 regions of the 18S rRNA are widely used DNA markers in marine protist research (Pawlowski et al., 2012; Burki et al., 2021). In this study, both regions exhibited their competence and arrived at similar statistical results based on community structures and environmental impacts. The V4 region showed more potent resolution power, possibly owing to its more completed dinoflagellate library and longer fragment size. The occurrence of most species was similar, as well as the tendency of their relative abundance in the sampling site; thus, similar pattern could be obtained. Karlodinium veneficum was such an example where coherent conclusion could be drawn despite subtle variations. However, huge variance was also observed in the relative abundance of some species. For example, G. gutrula was pervasive and also the second dominant species annotated by V9, but it appeared as a rare species among V4 metabarcodes. T. furca was the opposite, which being common based on V4 metabarcodes but rare in the V9 data set. Their correlations to environmental variables and realized niches based on V4 and V9 were distinct from each other.

The divergence in the read counts of metabarcodes depicts dinoflagellate community differently at certain stations. At the Eddy stations, their distant subsurface communities were positively correlated to nutrient concentrations on DCA plot based on V4 metabarcodes, but this correlation was lost on it built by V9 metabarcodes ( Figure 8 ). These differences could lead to separate interpretation if sole barcode region is used in a study and create uncertainty in comparison with other literature. The reason of it could be generalized down to the different resolution power and the genetic databases of the two. Based on the study by Ki (2012), the hypervariable V2–V4 has the best discriminative ability as the fragment size of V9 may be too short. In addition, the V4 regions detected more Prorocentrales and Syndiniales metabarcodes comparing to V9, and V9 detected more diversity in other subgroups like Peridiniales in our datasets. Similar observation also appeared in other studies (Stoeck et al., 2010; An et al., 2020; Choi and Park, 2020). It could be due to unintended preferential detection of primer sets or over-representation during amplification (Stoeck et al., 2010; An et al., 2020). Nevertheless, nearly half of the V4 and three quarters of the V9 metabarcodes could not be annotated to species level, and the relative abundance of species derived from V4 and V9 metabarcodes differed ( Figures 4 , 5 ). The present results based on existing genetic databases of dinoflagellate indicated the need of further efforts to improve the resolution to species level and complete dinoflagellate genetic database.

Although many HAB species were pervasive and found in all stations (e.g., Gonyaulax polygramma, Takayama pulchella, and Karlodinium veneficum), the number of taxa and the relative abundance of HAB assemblages were greater at KC stations and less at Coastal stations ( Supplementary File 2 ). These HAB species may have advantage in surviving oligotrophic environments, but non-HAB species also have the potential to surpass some cosmopolitan HAB species along coasts. Alexandrium leei, ATD that had caused HAB in coasts of Japan, is present infrequently at 10 m with low relative abundance; whereas Gyrodinium fusiforme, a HTD species feeding on dinoflagellate, diatom, haptophytes, and cryptophytes, was more abundant at Coastal stations (Naustvoll, 2000). This feeding strategy of G. fusiforme might help it to thrive at coastal environments where more preys live. At the same time, the high surface water temperature during this period may not favor Alexandrium leei because its optimal growth temperature was 20°C (Shikata et al., 2020). In our data, no HAB species solely presented at Coastal nor Eddy stations, except Amphidinium gibbosum that could either be brought up with bottom sediments or being endosymbionts of others (Murray et al., 2004).

Majority of the HAB species detected were absent at Coastal stations, such as Margalefidinium polykrikoides (MTD), indicating that Kuroshio could introduce HAB species to coastal regions and signifies the role of oceanic circulation that plays in the mechanism of HAB in addition to anthropogenic factors. M. polykrikoides has caused HAB first in East Asia, and its blooms spread globally after 1990. Studies have shown the outbreaks of this species were transported by currents from Korean coastal water to the southwest Sea of Japan and from the northern of Borneo, Malaysia to Palawan, Philippines, monitored using satellite image (Azanza et al., 2008; Onitsuka et al., 2010). Another study by Sildever et al. (2019) also monitored dispersal of this species along Hokkaido coasts by Soya warm current (a continuation of Tsushima Warm Current, a branch of Kuroshio). The molecular phylogenetic data also confirmed the differences in ribotypes in Asian coasts; American ribotype in Sabah, Malaysia, and Lampung, Indonesia; Philippine ribotype in the Philippines; and East Asian ribotype in Hong Kong, Japan, and Korea (Iwataki et al., 2007, 2008; Thoha et al., 2019). By performing LSU rDNA–based qPCR, a study conducted in Korean coastline and Tsushima Strait revealed that the Tsushima Warm Current could introduce the Philippine ribotypes into southern Korean water and potentially contributed the M. polykrikoides bloomings in 2009 in southern Korean coast (Park et al., 2018). Meanwhile, the study demonstrated the potential of the LSU rDNA–based metabarcoding analysis in identifying ribotypes of HAB dinoflagellates and elucidating the blooming mechanism. Another study conducted in the Jeju Warm Current, a branch of the Kuroshio Current in the northeastern ECS, also proposed the potential introduction of HAB species into the region by the Kuroshio Current (Kim et al., 2023). Some other HAB species mentioned in this study also present our study, such as Alexandrium affine, Margalefidinium polykrikoides, and Takayama pulchella (Kim et al., 2023). As Kawakami et al. (2023) suggest that the northward movement of Kuroshio greatly warms the surrounding region and may impact local marine biota, the introduction of HAB species along with warming water potentially increases risks of HAB along Japanese coast and prolonging it.

Additionally, the V9 regions may not be the best choice in detecting the bloom-forming dinoflagellate species. Only nine species were identified from V9 metabarcodes comparing to total of 15 species from V4 metabarcodes. The V9 primers may also have strong preference in Karlodinium veneficum, as its relative abundance among V9 metabarcodes was over 30% at all stations.