This study was conducted in the subtropical Kuroshio region of the Western North Pacific Ocean during the KH-22–5 research cruise (spring 2022) on the R/V Hakuho-Maru (Figure 1A) and during the KS-22–15 research cruise (autumn 2022) on the R/V Shinsei-Maru (Figure 1B). Samplings were carried out at 12 stations on the NA (Stations NA1–NA6) and T (Stations T1–T6) transect lines during spring and at 5 stations (Stations 3, 5–8, and E1) during autumn (Supplementary Table 1). Sampling stations were established in the East China Sea, on the Kuroshio Current, and in the open ocean across the Kuroshio Current for both research cruises, but surveys were not possible on the Kuroshio Current during autumn due to unfavorable weather conditions. The position of the Kuroshio Current during the study period was obtained from the Japan Coast Guard (Quick Bulletin of Ocean Conditions, https://www1.kaiho.mlit.go.jp).

Seawater was taken from four depths (10, 50, 100, and 150 m) for eDNA analysis and from 11 (10, 20, 30, 40, 50, 60, 80, 100, 125, 150, and 200 m during spring) or seven (10, 30, 50, 100, 150, 200 m and chlorophyll maximum during autumn) depths (< 200 m) for chlorophyll (chl.) a analysis using Niskin bottles (General Oceanics) equipped on a CTD-carousel system (SBE 911 plus CTD, Sea-Bird Scientific). For eDNA analysis, seawater samples (65 samples in total) were collected in 7-L plastic bags (Rontainer 5RH, Sekisui Seikei) after they were rinsed three times. Seawater (mean ± standard deviation: 6.85 ± 0.76 L) was immediately filtered through a 0.22-µm pore size cartridge filter (Sterivex SVGP01050, Merck) using a digital peristaltic pump (Masterflex 07528-10, Cole-Parmer) at a rotation speed of 60 rpm. The plastic bags and other materials (e.g. silicon tubes) were all sterilized before use to prevent contamination, and blank samples were also collected by filtering ultrapure water instead of seawater. After adding 2 mL of stabilization solution (RNAlater AM7024, Thermo Fisher Scientific) with 5-mL disposable syringes to avoid DNA degradation, the cartridge filters were stored at –20 °C until further analysis. The details of this filtration process were described by Yu et al. (2022).

For chl. a analysis, seawater samples were collected in 125-mL plastic bottles (Nalgene 2004-0004, Thermo Fisher Scientific) and immediately filtered through glass fiber filters (Whatman GF/F, GE Healthcare Life Sciences). After chl. a pigment on the filters was extracted with 5 mL of N,N-dimethylformamide at –20 °C for 24 h under the dark condition (Suzuki and Ishimaru, 1990), chl. a concentration was measured using a fluorometer (10 AU, Turner Designs) (Welschmeyer, 1994).

DNA was extracted from the cartridge filters and eluted in 50 µL of elution buffer using DNeasy PowerWater Sterivex Kit (14600-50-NF, Qiagen) after the stabilization solution was removed. The concentration of DNA extracted from blank samples was under the detection limit. Following an early-pooling protocol of Ushio et al. (2022), the 16S (V4–V5 region) and 18S (V4 region) rRNA genes of each sample were amplified in four replicates using 515F-Y/926R (Quince et al., 2011; Parada et al., 2016) and E572F/E1009R (Comeau et al., 2011) universal primer sets (with the sequences of the sample-specific tags and sequencing primers) for prokaryotes and microbial eukaryotes, respectively. The compositions of PCR mixtures and the thermal cycle conditions are shown in Supplementary Table 2, and PCR negative controls (H₂O instead of template DNA) were also prepared. The 1st PCR product of each sample was purified with ExoSAP-IT Express (75001, Thermo Fisher Scientific) and pooled at equimolar concentration. After purification with AMPure XP Reagent (A63881, Beckman Coulter), the 2nd PCR was performed to append adapter sequences. The 2nd PCR product was purified again with AMPure XP Reagent, and the target-sized DNA was excised using E-Gel SizeSelect II Agarose Gel (G661012, Thermo Fisher Scientific). The amplicon libraries were sequenced (2 × 250 bp PE) using a NovaSeq 6000 system (illumina).

Sequence data were demultiplexed and processed (e.g. primer trimming, quality filtering, error correction, and merging) by the dada2 plugin (Callahan et al., 2016) of QIIME 2 (Bolyen et al., 2019), which generates the table of amplicon sequence variants (ASVs). Because some ASVs were detected in negative controls, the numbers of the reads of the ASVs in negative controls were subtracted from those in seawater samples. Then, the numbers of the reads of each sample were rarefied to the minimum number of reads among all samples (10,000 for both prokaryotic and eukaryotic samples) through random sampling to prevent the effect of sequence depth. The coverages of the rarefied samples were almost consistent for prokaryotic (99.96 ± 0.04%) and eukaryotic (99.97 ± 0.02%) communities. After rarefaction, rare ASVs (relative abundance < 0.1%) and ASVs present in only one sample were removed for robust statistical analysis. Taxonomy was assigned to each ASV with the greatest possible resolution at a confidence threshold of 70% using the SILVA 138 reference database (Yilmaz et al., 2014). ASVs classified as the non-target organisms (e.g. chloroplasts for prokaryotic samples and prokaryotes for eukaryotic samples) were removed.

The diversity of prokaryotic and eukaryotic communities was assessed by ASV richness and Shannon index. Cluster (hierarchical) and non-metric multidimensional scaling (nMDS) analyses were conducted based on the Bray-Curtis similarity matrix among samples calculated from microbial relative abundance data at the order level to classify samples into several groups. On the basis of the number of groups formed as well as the complexity of the community structures of prokaryotes and microbial eukaryotes, similarity level used for grouping was determined. The nMDS analysis was also performed using similarities among microbial taxa in order to compare their distribution patterns. Moreover, to evaluate the effect of water temperature and salinity (explanatory variables) on the variation in microbial community structure (response matrix), the distance-based linear model (DistLM) analysis was conducted using the similarity matrix among samples. In addition to water temperature and salinity, chl. a concentration was also used as an explanatory variable, assuming that the relationships between chl. a concentration and prokaryotic or eukaryotic communities partly reflect the potential interactions between phytoplankton and heterotrophic prokaryotes or eukaryotes. Before the analysis, the explanatory variables were square-root-transformed, normalized, and checked for multicollinearity. The best model was determined based on stepwise selection and the adjusted R2 criterion, and significance was assessed by 10,000 permutations. The relationships between the explanatory variables and microbial community structure were visualized by the distance-based redundancy analysis (dbRDA) ordination and vector overlay. These analyses were performed in the statistical software PRIMER ver. 7 with a package PERMANOVA+ (Anderson et al., 2008; Clarke and Gorley, 2015).

On the NA Line, water temperature (17.1–17.3 °C) and salinity (34.58) were low and vertically uniform at Station NA1 located in the East China Sea (Figure 2A). Salinity showed high values (34.68–34.78) in the surface layer (< 150 m) of Stations NA4–NA6 in the open ocean. On the Kuroshio Current, high water temperature (21.0–22.1 °C and 21.6 °C) and low salinity (34.57–34.58 and 34.58) were observed at shallower depths (< 100 m and < 40 m at Stations NA2 and NA3, respectively). At deeper depths of the both stations, water temperature decreased gradually, and salinity increased irregularly. Generally, chl. a concentration was high near the surface and decreased with depth. In the surface layer, high chl. a concentration (0.14–0.75 µg L^–1) was observed at Stations NA1 and NA3, and low values (0.09–0.55 µg L^–1) were observed at Stations NA4–NA6. At Station NA2, chl. a concentration showed high values (0.26–0.61 µg L^–1) at shallower depths but decreased sharply at deeper depths.

On the T Line, water temperature was low (19.4–20.7 °C), and salinity was high (34.64–34.76) in the surface layer (< 150 m) of Stations T1–T3 in the open ocean (Figure 2B). In contrast, on the Kuroshio Current (at Stations T4–T5), high water temperature (20.4–21.7 °C) and low salinity (34.57–34.67) were observed at shallower depths (< 140 m). Similar to Stations NA2–NA3, water temperature and salinity at Stations T4–T5 decreased and increased, respectively, at deeper depths. At Station T6 located on the Kuroshio Current as well, water temperature showed a high value (21.4 °C) at shallower depths (< 20 m) but decreased sharply at deeper depths whereas salinity remained vertically uniform and low (34.57–34.61) in the surface layer. Chl. a concentration was generally high near the surface except at Station T6 where high values (0.72–1.11 µg L^–1) were observed even at deeper depths. Chl. a concentration in the surface layer showed low values (0.06–0.61 µg L^–1) at Stations T1–T3 and high values (0.06–0.86 µg L^–1) at Stations T4–T5. Chl. a concentration was considerably low (0.06 µg L^–1) at deeper depths of Station T5.

During autumn, the water column was strongly stratified due to heavy precipitation (e.g. 16.0–25.0 mm at Stations 5–8) during the study period (Figure 2C). Low water temperature (24.6–25.2 °C) and salinity (33.94–34.07) were observed at shallower depths (< 60 m) of Stations 5–8 in the East China Sea. On the other hand, water temperature (26.8–27.1 °C) and salinity (34.24–34.38) showed high values at shallower depths (< 70 m) of Stations 3 and E1 in the open ocean. At deeper depths, water temperature decreased sharply at the five stations. A gradual increase in salinity was observed at deeper depths of Stations 3 and E1 while salinity increased irregularly below shallower depths at Stations 5–8. Chl. a concentration at Stations 5–8 was high (0.45–0.62 µg L^–1) at shallower depths and decreased sharply at deeper depths. At Stations 3 and E1, the maximum of chl. a concentration (0.22 µg L^–1) was observed at 90 m and 50 m, respectively.

To characterize the microbial communities of different seasons, locations, and depths, samples were classified into several groups according to the similarity of prokaryotic or eukaryotic community structure at the order level among samples. For prokaryotes, cluster analysis divided 65 samples into six groups (a–f) with the similarity of 69% (Supplementary Figure 1A), and similarity distances among samples and groups were shown by nMDS analysis (Figure 3A). Similarly, the similarity of 59% was used for microbial eukaryotes, which formed nine groups (A–I) (Figure 3B, Supplementary Figure 1B).

Prokaryotic samples were primarily divided by seasons (Table 1A). Samples of spring and autumn were classified into Groups a–c and d–f, respectively, except a sample at 150 m depth of Station NA2. During spring, samples at Station NA1 (Group b) and samples at 100–150 m depths (Group c) formed different groups. Samples at 10–50 m (Group d) and 100–150 m (Groups e–f) depths were grouped differently during autumn. Flavobacteriales were the most dominant (17.0–37.5%) during spring (Groups a–c), and Synechococcales (17.4%), SAR11 Clade (18.2%), and Nitrosopumilales (16.2%) were also abundant in Groups a, b, and c, respectively (Figure 4A, Supplementary Figure 2A). During autumn, Groups d (44.5%) and f (16.2%) were dominated by Synechococcales, and Nitrosopumilales dominated (17.8%) Group e.

Most of eukaryotic samples were classified into Groups A (28 samples) and C (6 samples) regardless of seasons, locations, and depths. The other samples were divided by depths (Table 1B). Groups B, G, and I were composed of samples at shallower depths (0–50 m) whereas samples at deeper depths (100–150 m) formed Groups D–F and H. Notably, samples at 150 m depth of Stations NA2 and T5 and at deeper depths of Stations 5–8 were grouped together as Group H. Dinophyceae and Syndiniales were dominant in Groups A (22.9% and 13.1%, respectively) and C (17.0% and 12.2%, respectively) with more abundant Siphonophorae (26.8%) in Group C (Figure 4B, Supplementary Figure 2B). In Groups B, G, and I, heterotrophs such as Copelata (35.4%), Phyllodocida (42.4%), and Doliolida (63.8%) dominated eukaryotic communities, respectively. Groups D, E, and F, were dominated by Peridiniales (29.8%), Syndiniales (26.6%), and Fungi (33.9%), respectively, and Spumellaria were the most dominant (57.3%) in Group H.

For prokaryotic communities, both richness and Shannon index showed high values in Groups c (medians: 453 and 7.15, respectively), e (862 and 8.48), and f (585 and 7.87) (Figures 5A, C) which consist of samples at deeper depths (100–150 m) (Table 1A). The richness of eukaryotic communities was high (505) in Group A and low (159) in Group F (Figure 5B). High Shannon index of eukaryotic communities was observed in Groups A (7.35) and E (7.33) while the index was low in Groups F (4.53) and I (4.08) (Figure 5D).

The DistLM analysis showed that water temperature, salinity, and chl. a concentration explained 64.6% of the variance in the community structure of prokaryotes at the order level (Table 2). Among the parameters, the contribution of chl. a concentration was the largest (35.0%), followed by water temperature (20.0%) and salinity (9.57%). On the other hand, for microbial eukaryotes, only 35.4% of the variance was accounted for by the three parameters. Chl. a concentration contributed most (24.7%) to the variation in eukaryotic community structure, but water temperature (8.59%) and salinity (4.72%) each made a minor contribution. The axes 1 and 2 of the dbRDA accounted for 39.3% and 23.4% of the variance in prokaryotic community structure, respectively (Figure 6A). For microbial eukaryotes, 26.8% and 6.04% of the variance in community structure was explained by the dbRDA axes 1 and 2, respectively (Figure 6B). The axis 1 positively and highly correlated with chl. a concentration whereas water temperature (negative) and salinity (positive) showed stronger correlations with the axis 2 for both prokaryotic and eukaryotic communities (Supplementary Table 3).

To compare the distribution patterns of microbial communities across multiple trophic levels, the similarities among both prokaryotic and eukaryotic taxa at the order level were examined. Prokaryotes, dinoflagellates (Dinophyceae, Syndiniales, Peridiniales, and Gymnodiniales), Spirotrichea, and Copelata showed relatively similar distribution patterns and formed the same cluster at the similarity of 29% (Figure 7, Supplementary Figure 3). In contrast, distinct distribution patterns were observed for Siphonophorae, Spumellaria, Doliolida, Phyllodocida, and Fungi. The distribution patterns of these heterotrophic eukaryotes were also independent from each other.

This study showed the key characteristics of physical and biological environments in the subtropical Kuroshio region. For example, water temperature and salinity were low during each season with high chl. a concentration at shallower depths of Stations NA1 and 5–8 (Figure 2), which is typical of shelf water in the East China Sea affected by abundant river discharge (Lie and Cho, 2002). In the surface layer of Stations NA3 and T4 located at the edge of the Kuroshio Current, chl. a concentration showed high values during spring. Similarly, high phytoplankton abundance was found near the surface of the Kuroshio front due to the upward flux of nutrients caused by the enhanced turbulence (Kaneko et al., 2013). At deeper depths of Station T6, high chl. a concentration as well as low water temperature and salinity were observed probably because of the discharge of coastal water, as reported by Endo and Suzuki (2019).

The Kuroshio Current, which flows along the continental slope of the East China Sea, causes complex hydrodynamic processes in the subtropical Kuroshio region. Previous studies reported that the shoreward upwelling of offshore intermediate water occurred under the Kuroshio surface water due to steep topography (Chen et al., 1995; Ito et al., 1995). Thus, the observed changes in physical conditions and chl. a concentration (i.e. the decrease in water temperature and the irregular increase in salinity) during spring at deeper depths of Stations NA2 and T5 located on the Kuroshio Current (Figure 2) probably resulted from the intrusion of intermediate water from the open ocean. Furthermore, similar changes in these parameters were also observed at deeper depths of Stations 5–8 during autumn. This may have been because the shoreward upwelling reached the East China Sea and showed strong seasonality with maximum during autumn (Guo et al., 2006; Isobe, 2008; Wu et al., 2014). Interestingly, the different water masses during spring and autumn were reflected in the distinct structure of microbial communities at 150 m depth of Stations NA2 and T5 and at 100–150 m depths of Stations 5–8 consisting of taxa abundantly detected in intermediate waters (Table 1, Figure 4). These results indicate that high microbial diversity in the subtropical Kuroshio region is supported by the dynamic interplay of multiple water masses.

Prokaryotes showed seasonally and vertically different community structures because prokaryotic communities were mainly grouped by seasons and depths (Table 1A, Figure 4A). During spring, Flavobacteriales dominated prokaryotic communities (Groups a–c). The dominance of Flavobacteriales is often associated with phytoplankton blooms (Buchan et al., 2014) as they efficiently utilize the high-molecular-weight organic compounds derived from phytoplankton (Teeling et al., 2012; Fernández-Gómez et al., 2013; Williams et al., 2013). Indeed, a significant positive correlation was observed between chl. a concentration and the relative abundance of Flavobacteriales throughout samples (r = 0.489, n = 65, p < 0.001) (Supplementary Figure 4A). In addition, Flavobacteria are known to have proteorhodopsins (PRs) functioning as the light-driven H+ pumps, and H+ gradient across membranes is used to produce energy via ATP synthases (Béjà et al., 2000; Yoshizawa et al., 2012). Gómez-Consarnau et al. (2007) proved the active photo-heterotrophic growth of the PR-containing Flavobacteria under light exposure, which is considered to support their prosperity in oligotrophic, subtropical environments. SAR11 clade, one of the most dominant bacteria in marine surface waters (Morris et al., 2002; Giovannoni, 2017), was also abundantly detected during spring (Group b). Similar to Flavobacteria, SAR11 clade members contain PRs, allowing them to sustain growth under the light and carbon-limited conditions (Giovannoni et al., 2005; Steindler et al., 2011).

During autumn when the water column was more stratified in the surface layer, different prokaryotic communities were observed between shallower and deeper depths, as reported by Treusch et al. (2009). Prokaryotic communities at 10–50 m depths were overwhelmingly dominated by Synechococcales during autumn (Group d) (Table 1A, Figure 4A), which are dominant primary producers in warm, oligotrophic environments (Agawin et al., 2000; Hirata et al., 2011) since smaller cells have higher surface-to-volume ratio and are more beneficial in nutrient uptake (Pasciak and Gavis, 1974; Kiørboe, 1993; Marañón, 2015). Laboratory experiments showed that the optimal growth temperature of marine Synechococcus was mostly between 21 °C and 28 °C (Moore et al., 1995; Mackey et al., 2013). The water temperature of samples classified into Group d was within this range (24.6–27.1 °C) (Supplementary Figure 5A), and water temperature was significantly positively correlated with the relative abundance of Synechococcales throughout samples (r = 0.804, n = 65, p < 0.001) (Supplementary Figure 4A). On the other hand, Nitrosopumilales, ammonia-oxidizing archaea (Qin et al., 2016), were abundant at 100–150 m depths during autumn and spring as well (Groups e and c, respectively) probably because ammonia oxidation is photo-inhibited in surface waters (Horak et al., 2018). These results correspond to the vertically stratified community structure of prokaryotes with a sharp transition around 50–100 m depths (DeLong et al., 2006; Mende et al., 2017). The present study indicates that prokaryotic communities are structured seasonally by physical and biological environments and vertically by water column stratification and light environments.

Eukaryotic communities were mostly dominated by dinoflagellates (Dinophyceae, Syndiniales, and Peridiniales) (Groups A and C–E, 50 out of all 65 samples) (Table 1B, Figure 4B). Dinoflagellates are globally distributed in the epipelagic layer (Le Bescot et al., 2016) and known to dominate the subtropical region of the Western North Pacific Ocean (Zhang et al., 2018; Wu et al., 2020). Dinoflagellates are highly diverse in trophic strategies (Cohen et al., 2021). Although about half of marine dinoflagellate species are autotrophic or mixotrophic, the other half are heterotrophic (Gómez, 2012). Particularly, Syndiniales are parasitic and infect a wide range of hosts such as other dinoflagellates, ciliates, and radiolarians (Guillou et al., 2008; Anderson and Harvey, 2020; Käse et al., 2021). Syndiniales often dominate eukaryotic communities in various oceanic regions, which has been observed by recent 18S rDNA metabarcoding surveys (Clarke et al., 2019; Rizos et al., 2023; Anderson et al., 2024), and can regulate host populations (Chambouvet et al., 2008; Montagnes et al., 2008). The dominance and diverse trophic strategies of dinoflagellates suggest their important roles in the marine microbial food web.

Other dominant eukaryotic communities were largely different depending on depths. Copelata (Group B) and Doliolida (Group I) dominated eukaryotic communities locally but overwhelmingly at shallower depths (Table 1B, Figure 4B) where high chl. a concentration (0.57–1.05 µg L^–1) was observed (Supplementary Figure 5B). These tunicates feed on biological particles including the pico-sized and nano-sized plankton through mucous filter-feeding (Deibel, 1998; Bone et al., 2003), which cannot be ingested efficiently by crustacean zooplankton (copepods). Appendicularian houses and doliolids are then grazed by copepods (Takahashi et al., 2013; Nishibe et al., 2015) and fish larvae and juveniles (Watanabe and Saito, 1998; Watanabe and Kawaguchi, 2003). Thus, Copelata and Doliolida probably proliferated in their favorable food environments and linked small primary producers and the early life stages of fish efficiently, supporting high fishery production in the subtropical Kuroshio region as the tunicate food chain (Okazaki et al., 2019). Nevertheless, although tunicates were identified around the region in plankton net samples (Kobari et al., 2018) and the gut contents of fish larvae and juveniles (Okazaki et al., 2019), the detailed information on their local distributions has been limited. The present study successfully captured the spatially heterogeneous distribution of eukaryotic communities in marine environments by applying eDNA metabarcoding analysis at horizontally and vertically high resolutions.

At deeper depths, Spumellaria were most dominant (Group H) (Table 1B, Figure 4B). A previous study reported that the class Polycystinea (include the order Spumellaria) was abundant exclusively in the deep layer (around 500 m depth) of the Eastern North Pacific Ocean (Schnetzer et al., 2011). This is consistent with the results of water column structure, indicating that the water mass of samples classified into Group H was derived from offshore intermediate water, as discussed above. Moreover, fungi were dominant at 100–150 m depths of Station T6 (Group F), which was probably affected by coastal water. Because fungal abundance decreased with increasing salinity in the Delaware River estuary (Burgaud et al., 2013) and increased with riverine input at a coastal time-series site of the Western English Channel (Taylor and Cunliffe, 2016), salinity may be important for the prosperity of fungi in the ocean. These results indicate that the intrusion of water masses brings microbial communities and increases biodiversity in an oceanic region.

The diversity of both prokaryotic and eukaryotic communities varied vertically. High diversity (both richness and Shannon index) of prokaryotic communities was observed in groups composed of samples at deeper depths (Table 1A, Figures 5A, C). Thus, when the diversity of all samples is compared by depth, prokaryotic diversity was significantly different between depths and higher at deeper depths (Supplementary Table 4, Supplementary Figures 6A, C). This is consistent with the vertical variation in epipelagic waters reported in previous studies and was probably due to adaptation to a wider range of ecological niches at deeper depths such as the particle-associated micro-environments (Sunagawa et al., 2015; Mende et al., 2017). In contrast, the obvious trend of eukaryotic diversity was not observed by comparison among groups (Figures 5B, D). In terms of vertical variation, the diversity of eukaryotic communities tended to be high at shallower depths, decrease at 100 m depth, and then increase at 150 m depth although significant difference was not found between depths (Supplementary Table 4, Supplementary Figures 6B, D). Schnetzer et al. (2011) investigated eukaryotic community structure at multiple depths (< 880 m) of the Eastern North Pacific and found high diversity at the surface (1.5 m) and middle depth (150 m). They pointed out that high diversity was supported by diverse autotrophic species at the surface and various niches with sharp environmental gradients including water temperature and particle concentration at middle depth. The vertical variations in diversity were different between prokaryotic and eukaryotic communities, which suggests their adaptation to different environmental conditions in the ocean.

Physical conditions and chl. a concentration affected the community structures of prokaryotes and microbial eukaryotes differently. The DistLM analysis showed that chl. a concentration (35.0%) and physical conditions (29.6%) accounted for most (64.6%) of the spatial and seasonal variances in prokaryotic community structure (Table 2). Thus, it is indicated that the supply of resources such as organic matter and nutrients by phytoplankton and the physical environments of water masses mainly structure prokaryotic communities. This finding is consistent with the correlation of prokaryotic community structure with water temperature on a global scale (Sunagawa et al., 2015) and with phytoplankton community structure in the southern California region (Needham and Fuhrman, 2016). On the other hand, for microbial eukaryotes, chl. a concentration exerted a major (24.7%) influence on community structure, but physical conditions had a minor (10.7%) effect. Compared to prokaryotes, the effect of physical conditions on the community structure of eukaryotes was much smaller. A recent study also reported that water temperature and salinity less explained the variability of eukaryotic community structure in the Chinese marginal seas (Wang et al., 2024). In coastal waters of South Africa, protistan community structure showed a stronger relationship with chl. a concentration (Holman et al., 2021) as observed in the present study. Although the relationship between chl. a concentration (a proxy for phytoplankton abundance) and the community structure of eukaryotes including phytoplankton may not be surprising, the major contribution of chl. a concentration to structuring eukaryotic communities can be partly attributed to the importance of food environments for heterotrophs such as heterotrophic dinoflagellates and tunicates. However, a large proportion (64.6%) of the variance in eukaryotic community structure remained unexplained, indicating that other factors not considered in this study were more important. For instance, Lima-Mendez et al. (2015) examined the co-occurrence patterns of microbial taxa in the euphotic layer globally and found the importance of biological interactions, especially parasitism, in structuring plankton communities. Parasitic interactions were probably common in the present study since parasitoids (e.g. Syndiniales) were detected abundantly (Figure 4B, Supplementary Figure 2B). Moreover, predatory interactions (Zhu et al., 2023) and inorganic nutrients (Wu et al., 2020; Sogawa et al., 2022) may have been important. These results suggest that different factors shape the structures of prokaryotic and eukaryotic communities and thus cause the different spatio-temporal variations in their community structures.

The comparison of distribution patterns among both prokaryotic and eukaryotic taxa suggests biological interactions and ecological roles in marine environments. Prokaryotic growth is generally dependent on materials derived from phytoplankton (Cole et al., 1988; Ducklow and Carlson, 1992), and phytoplankton abundance and community structure are highly affected by the activity of symbiotic prokaryotes (Needham and Fuhrman, 2016). Ciliates (Jonsson, 1986; Rassoulzadegan et al., 1988) and appendicularians (Scheinberg et al., 2005; Troedsson et al., 2007) mainly feed on the pico-sized and nanosized plankton. Thus, relatively similar distribution patterns were observed for prokaryotes, dinoflagellates, Spirotrichea, and Copelata (Figure 7) probably due to their symbiotic and trophic interactions. Furthermore, these taxa were common and detected to some extent in most samples (Supplementary Figure 2), which implicates their roles in maintaining the stable structure and functions of marine ecosystems. In contrast, the other heterotrophic eukaryotes (Siphonophorae, Spumellaria, Doliolida, Phyllodocida, and Fungi) exhibited distinct and unique distribution patterns, showing the horizontal and vertical heterogeneity of their distributions in epipelagic waters. For example, doliolids form dense patches sporadically in subtropical waters with abundant food availability and impact local food web structure greatly through their high feeding rate (Takahashi et al., 2015; Frischer et al., 2021). The sporadic and massive blooms of Doliolida at shallower depths of the edge of the Kuroshio Current (Figure 4B, Supplementary Figure 2B) coincided with high chl. a concentration (0.76–1.05 µg L^–1) (Supplementary Figure 5B) and probably changed the structure of the microbial food web and energy transfer to higher trophic levels locally in their favorable environments due to active heterotrophy. The distinct distribution patterns of the heterotrophic eukaryotes may also have resulted from strong predatory relationships with other food sources and higher trophic levels not considered in this study. Warm-water species of Siphonophorae inhabiting the epipelagic layer, for instance, prey on copepods and fish (Mapstone, 2014; Damian-Serrano et al., 2021). These findings correspond to higher community stability of prokaryotes relative to eukaryotes under environmental disturbance (Wang et al., 2024) and more unique structure of eukaryotic communities compared to prokaryotes (Zhu et al., 2023). Different distribution patterns among prokaryotic and eukaryotic communities implicate the differences in their ecological roles in the ocean as well as their symbiotic and trophic interactions.