Environmental variables in the JS were measured at eight stations during four sampling periods: June, September, and November 2017, and March 2018. In addition, surface seawater samples were collected specifically for the analysis of planktonic protists, with three of the eight stations chosen for this purpose ( Figure 1A ). These stations were located in three distinct regions: the shallow-water (S) region adjacent to the Korean peninsula, the middle region of the strait where a seasonal thermohaline front (F) is present, and the deep-water (D) region near Jeju Island. However, due to adverse weather conditions, we were unable to operate the research vessel (R/V) between December and February as originally planned.

At each site, approximately one ton of surface seawater was collected using a submerged pump with a flow rate of 70 liters per minute. The collected seawater was subjected to a filtration process for the harvesting of the planktonic protists. Initially, the seawater was pre-filtered through a 200 μm mesh to eliminate large organisms and debris. Then, it was further filtered through a 1 μm Nylon conical-shaped net, which retain the planktonic protists while allowing excess water to pass through. The filtrate fractions were transferred to 50 mL conical tubes and centrifuged at 2,500×g for 30 min. This centrifugation step effectively removes any residual seawater and resulted in a pellet of concentrated plankton. The pellet was immediately frozen and stored in liquid nitrogen until genomic DNA (gDNA) extraction, in order to preserve the DNA quality as much as possible.

Temperature and salinity profiles were recorded using a CTD probe (Sea-Bird Electronics, WA, USA). Seawater samples from various depths were collected using Niskin bottles attached to a CTD-Rosset sampler. Nutrients analysis, including dissolved inorganic nitrogen (DIN), dissolved inorganic phosphorus (DIP), and dissolved inorganic silicate (DISi), was conducted using water samples obtained from multiple stations at depths ranging from 0 to 125 m. To assess pH levels, water samples were taken from the surface, middle, and bottom. Total chlorophyll a (TChl a) concentrations were specifically determined from water samples collected at depths of 0, 10, and 30 m. To assess the nutrient levels in seawater, a 100 mL sample was filtered through a 0.45 μm PTFE syringe filter (Advantec, Japan) and stored at -20°C. The sample was later measured using an auto-analyzer (Quattro, Seal Analytical, Norderstedt, Germany). The pH of the seawater samples was measured at 25°C using spectrophotometry with the m-cresol purple indicator, following established procedures described by Clayton and Byrne (1993). In addition, the TChl a concentration was determined using a chlorophyll a fluorometer (ECO-AFL/FL, Sea-Bird Electronics, WA, USA) attached to a CTD probe. The fluorescence measurements were converted to TChl a concentration using a regression method based on the relationship between fluorescence and TChl a, which was established using high-performance liquid chromatography (HPLC) as the reference method.

The small organelle-enriched metagenome pioneer version (SoEM-pv) method, as validated by Jin et al. (2023), was used to identify marine plankton via environmental DNA (eDNA). DNA extraction involved homogenizing the pellet sample in a buffer of 250 mM sucrose, 30 mM Tris-HCl, 10 mM EDTA, and pH 7.5, using an IKA-T10 homogenizer (IKA, Staufen, Germany). Small organelles were concentrated from the homogenate through a modified differential centrifugation method (Jo et al., 2019) and the DNA was extracted using the Qiagen DNeasy Blood/Tissue Kit (Qiagen, Hilden, Germany) following the manufacturer’s instructions.

For library construction, the TruSeq Nano DNA kit (Illumina, CA, USA) was used, fragmenting approximately 200 ng of DNA to a 550 bp insertion size with a LE220 Focused-ultrasonicator (Covaris, MA, USA). The DNA fragments underwent end repair, the addition of a single “A” base at the 3’ end, and ligation of the Illumina adapters. The libraries were purified and template size distribution analysis was performed using the Agilent Technologies 2100 Bioanalyzer and a DNA 1000 Chip (Agilent Technologies, CA, USA). Quantification used qPCR following the protocol guide for Illumina sequencing platforms. Paired-end sequencing (2 × 301 bp) was carried out on the Illumina Miseq platform by Macrogen Inc (Seoul, Korea). The raw sequencing data were deposited in the NCBI SRA database under the accession numbers from SRR24790320 to SRR24790330, and under the NCBI Bioproject accession number PRJNA974865.

We conducted sequence analyses by modifying the bioinformatics analysis pipeline of SoEM-pv (Jo et al., 2019). Raw sequences were preprocessed with Trimmomatic v.0.39 to exclude adapter sequences and low-quality reads below the Phred score average of 33 (Bolger et al., 2014). The paired-end reads were merged using FLASH v1.2.11, with a minimum overlap of 20 bp (Magoč and Salzberg, 2011).

Operational taxonomic units (OTUs) were identified from merged sequences using a BLAST search against the non-redundant nucleotide database (accessed November 11, 2021, from https://www.ncbi.nlm.nih.gov/nucleotide/) with an E-value under 1e-10 and a query sequence length of at least 400 bp. Confident OTUs were selected based on a minimum of five read counts. The taxonomic identification was obtained by assigning all taxonomic ranks from kingdom to species using the NCBI taxonomy database. Super-kingdoms other than eukaryotes were discarded as part of this process (Jin et al., 2023).

Taxonomic classification and inhabitance (i.e., terrestrial, fresh, brackish, and marine) of each OTU were identified using the scientific name and the World Register of Marine Species (WoRMS) database with the worrms package v.0.4.2 (Chamberlain, 2020). To focus on protists, the supergroups of SAR, Excavata, and Amoebozoa were intentionally selected (Adl et al., 2012). Non-marine OTUs that were assigned by WoRMS, were removed. Functional traits of 252 OTUs were assigned to two size groups (nano-size 2–20 μm and micro-size 20–200 μm) and six trophic modes (obligate photoautotroph (OPA), heterotroph (HET), constitutive mixotroph (CM), generalist non-constitutive mixotroph (GNCM), endosymbiotic specialist (eSNCM) and plastidic specialist non-constitutive mixotroph (pSNCM)) (Flynn et al., 2019). The assignments were made using Ramond et al. (2019)’s functional traits database of marine protists. Such functional traits can give insights into the potential roles and ecological functions of the planktonic protist community (PPC) in the JS.

We determined that our sampling had sufficient sequence depth based on the rarefaction curves of each sample, but we noted significant variation in sequence depth ( Figure S1 ). To avoid false negative error, we did not rarefy (McMurdie and Holmes, 2014), but instead normalized individual OTU read counts by relative abundance for downstream analysis.

We calculated alpha diversity indices, such as species richness and the Shannon-Wiener index of diversity, using the vegan package v.2.6.4 (Oksanen et al., 2022). Faith’s phylogenetic diversity (PD) and the net relatedness index (NRI) were computed using the picante package v.1.8.2 (Kembel et al., 2010). The phylogenetic tree was created with the phylostratr package v.0.2.1 (Arendsee et al., 2019), and all branch lengths were set to “1” using the “compute.brlen” function in the ape package v.5.7.1 (Paradis and Schliep, 2019).

To identify statistical differences among seasons and stations, we performed an analysis of variance (ANOVA) or a Kruskal-Wallis H tests, with post hoc multiple comparisons based on the Bonferroni test. Non-metric multidimensional scaling (NMDS) and analysis of similarity (ANOSIM) were used to assess differences in PPC using the Jaccard distance to estimate community dissimilarities. Principal component analysis (PCA) and ANOVA were employed to examine variations in environmental conditions across stations and seasons. We utilized several R packages, including phyloseq, dplyr, lawstat, and ggplot2, in R v.4.2.2 for data analysis and visualization (McMurdie and Holmes, 2013; Wickham et al., 2023; R Core Team, 2022).

Consistent mixing was observed throughout all seasons at the shallow-water (S) station in the JS. In contrast, strong stratification was observed in September at the thermohaline front (F) and deep-water (D) stations displayed ( Figures 1B, C ). There were no significant differences in the environmental variables between spatial locations ( Table S1 ). However, in November, distinct environmental conditions were observed near the F station due to a thermohaline front ( Figures 1B–D ). However, the strait experienced significant changes in environmental conditions over time ( Figure S2 ). Surface water temperature peaked in September (26.33 ± 0.86°C) and dropped to its lowest in March (13.58 ± 1.14°C). Conversely, salinity showed an opposing trend, with values ranging from 30.49 to 34.41 ( Figures 1C ; S2 ). DIP and DISi were inversely correlated with temperature and displayed significant correlations (r = -0.705, -0.727, p < 0.05); however, DIN did not show a significant correlation (r = -0.589, p = 0.057). In contrast, pH had a significant positive correlation with temperature (r = 0.909, p < 0.001).

Preprocessing the Illumina sequencing of small organelle-enriched DNA molecules (as per section 2.3.) revealed a total of 252 protistan OTUs from the surface water of three stations in the JS. The taxonomic and phylogenetic diversity of the PPC did not significant vary with spatial locations or temporal changes. However, the indices of phylogenetic structure differed ( Figure 2 ). In November, NRI was higher (p = 0.052) compared with other months, and Faith’s PD was somewhat lower, indicating a more clustered phylogenetic structure and less phylogenetic diversity in the PPC. During November, the PPC demonstrated a more clustered phylogenetic structure (NRI > 0, Figure 2D ), implying that environmental filtering played a significant role in PPC structuring. This suggests that specific environmental factors selected certain protist species, leading to a less diverse community (see to Table 1 in Skouroliakou et al., 2022). The NRI values showed weaker phylogenetic clustering in June and March, indicating less environmental filtering and a more diverse PPC during these seasons.

NMDS ordination results indicated that PPCs in the JS clustered more strongly by season than by station ( Figure 3A ). ANOSIM also supported this finding, showing that seasonal variations significantly influenced the PPC structure (R = 0.987, p < 0.001) more than spatial variations (R = -0.133, p = 0.907). Essentially, differences in community composition were driven more by seasonal environmental changes than by variations in water masses or locations.

The number of unique species and species richness varied between seasonal protist communities, while spatial communities demonstrated similar values ( Figure 3B ). When comparing ten different phyla, the PPCs in November were primarily composed of Ochrophyta (mostly diatoms, at 91%), constituting 95% of the community ( Figures 4A, D ), while PPCs displayed a more balanced composition in March, with a variety of phyla. Relative abundance of Myzozoa (averaging 49% across three stations), including Alexandrium catenella (10%), A. pacificum (12%), and A. tamarense (16%), was considerably higher in March compared with other periods. Similarly, the relative abundance of Chlorophyta Bathycoccus prasinos (7.3%) and Cryptophyta Teleaulax amphioxeia (6.8%) was also higher. Although the taxonomic composition was generally similar across size fractions, the relative abundance of micro-sized protists increased by about 50% from June to November, then decreased in March ( Figures 4B, E ). In the JS, Bacillariophyceae (Ochrophyta), commonly known as diatoms, and Dinophyceae (Myzozoa) were the most abundant and diverse groups within the PPCs. Bacillariophyceae accounted for 59.4% of the sequence abundance and 48.4% of the OTU richness, while Dinophyceae represented 27.4% of the sequence abundance and 35.3% of the OTU richness.

The taxonomic composition of protists, classified by their trophic mode, demonstrated a trend similar to the relative abundance of Myzozoa ( Figure 4 ). Myzozoa, which are dinoflagellates capable of mixotrophy (a combination of phototrophy and phagotrophy), showed a positive correlation with the proportion of nano-sized protists in the community.

The NCM was applied to investigate the variability in the occurrence frequency of the entire PPCs in the JS. The model accounted for a large proportion (R^{2 =} 0.628) of this variability ( Figure 5A ), suggesting that stochastic processes predominantly drove the assembly of the PPC in this strait. The null model analysis results showed that the assembly of PPCs in the JS was greatly influenced by stochastic processes ( Figure 5B ). Within the neutral group, stochastic processes comprised undominated processes (50.9%), homogenizing dispersal (47.3%), and dispersal limitation (1.8%).

The OTUs were classified into three groups–above, below, and neutral–based on the NCM predictions. Most of the OTUs (87%) fell under the neutral prediction group, while a smaller portion was classified as non-neutral above or non-neutral below (4% and 8%, respectively). OTUs in the non-neutral above prediction group, which exhibited higher frequency and lower relative abundance than the neutral prediction group, comprised two Pseudo-nitszchia species, three Chaetoceros species, four Ochrophyta, and two Ciliophora ( Figure 5C ; Table S2 ). In contrast, OTUs in the non-neutral below prediction group, which demonstrated lower frequency and higher relative abundance, included five Alexandrium species, Protoceratium reticulatum, Prorocentrum triestinum, two other Myzozoa, six Ochrophyta, and four Chlorophyta and Haptophyta.

There were no differences in the size fraction of the OTUs between the neutral and non-neutral prediction groups ( Figure 5D ). However, trophic mode composition varied between the groups ( Figure 5E ). The above prediction group included a HET ciliate, Sterkiella histriomuscorum, and an eSNCM Ciliophora, in addition to mainly OPA ( Table S2 ). By contrast, the below prediction group was predominantly composed of CM dinoflagellates and a GNCM Cercozoa species, Ebria tripartita.

The correlation-based network of the PPC in the JS included 149 nodes representing distinct species and 209 edges indicating their interactions ( Figure 6A ; Table 1 ). The presence of 61 unique modules, many of which consisted of a single node, suggests the existence of distinct subgroups within the community with a certain level of functional differentiation. Approximately 72% of the interactions were positive (gray lines in Figure 6A ), indicating symbiotic relationships, while 28% were negative (blue lines), indicating the presence of competitive interactions within the community. An average clustering coefficient of 0.39 and a path length of 3.95 suggests a propensity for species to form clusters and establish efficient relationships within the network. A network diameter of 9 indicates a well-connected community. The average betweenness centrality score of 72.5 highlights the existence of specific key species that act as bridges, facilitating interactions between different parts of the network.

Within the primary network, there were 86 nodes and 202 links, primarily arranged into eight major modules, which accounted for 92% of the network. The modules and their proportions were as follows: Module I (12.8%), Module II (9.4%), Module III (10.1%), Module IV (6.0%), Module V (6.0%), Module VI (4.0%), Module VII (2.7%), and Module VIII (2.0%) ( Table S3 ). The transition pattern of the major protist community modules was evaluated using normalized relative abundance, and the correlation between modules and environmental variables was studied via a partial Mantel test ( Figures 6B, C ). The results showed that in the JS, the PPC transitioned in the following sequence: Modules III and IV, Module VIII, Modules I and V, Modules II and VII ( Figure 6B ). Modules III, IV, and VIII exhibited a positive correlation with temperature (r = 0.360‒0.566, p < 0.05), while Modules III and VIII demonstrated a strong positive correlation with salinity (r = 0.610–0.646, p < 0.005). Module III was primarily composed of nano-sized diatoms such as Skeletonema, Thalassiosira, and Chaetoceros. Module IV, comprising Pseudonitzschia, Alexandrium minutum, and Akashiwo sanguinea, showed a positive correlation with pH (r = 0.482, p = 0.002). Modules I and V displayed the highest normalized relative abundance in November, while Module I, which comprised micro-sized diatoms such as Coscinodiscus, Rhizosolenia, and Biddulphia, was significantly associated with DISi (r = 0.397, p = 0.024). In March, Modules II and VII were abundant and had a high proportion of dinoflagellates in OTU composition. The raphid pennate diatom Seminavis robusta played a role as a hub in Module II, and the toxic dinoflagellate Alexandrium fundyense was the connector linking the primary network and Module VII ( Figure 6C ; Table S3 ). These findings were not related to any environmental variables ( Figures 6A ; S3).

The JS is characterized by dynamic environmental conditions influenced by different water masses, including the Korea Coastal Current (KCC), the JWC, and Changjiang Dilute Water (CDW). Our study revealed distinct oceanic phenomena occurring during different seasons, manifested as thermoclines and thermohaline fronts ( Figures 1B, C ). In June 2017, we observed the development of a seasonal thermocline near Jeju Island in the deep-water areas of the JS. A thermocline represents a layer within the water column where there is a rapid temperature change with increasing depth. Over time, this thermocline became deeper and wider, reaching its peak extent in September. Additionally, we noticed the presence of low salinity water extending above the thermocline throughout the JS. The formation of a thermocline leads to vertical stratification within the water column, resulting in distinct phytoplankton communities with varying nutrient. This vertical stratification has a significant influence on the structure of the PPCs (Mena et al., 2019; Venkataramana et al., 2021).

In November, we observed the occurrence of thermohaline fronts in the JS ( Figures 1B, C ). Thermohaline fronts serve as boundaries between different water masses based on variations in temperature and salinity. They indicate significant shifts in the temperature and salinity structures of the ocean during seasonal changes, impacting the distribution of marine life and nutrient availability (Franks, 1992). In the JS, the clockwise flow of the warmer and saltier JWC around Jeju Island, combined with the well-mixed coastal water off southwestern Korea, creates a thermohaline front near station F. This front potentially contributes to the observed temperature inversion (Kim et al., 2022). The formation of a thermohaline front along the 15–18°C isotherms in November corresponds to the horizontal distribution of surface temperature ( Figure 1B ). Thermohaline fronts are globally recognized for promoting ocean mixing through the horizontal intrusion of water masses. They also serve as biodiversity hotspots, fostering phytoplankton growth and attracting higher trophic level organisms, which support important fisheries with abundant zooplankton (Mcginty et al., 2014; Haberlin et al., 2019; Kim et al., 2020a).

In this study, we analyzed the alpha diversity indices of PPCs in the JS. Our analysis showed that seasonal changes in environmental conditions had a significant impact on Faith’s PD and NRI. Specifically, the NRI exhibited a substantial increase in November compared to other seasons ( Figure 2D ), which can be attributed to the presence of thermohaline fronts and the resulting distinctive environmental conditions. This finding suggests that community assembly during November was more deterministic, as indicated by the high NRI. This implies a higher likelihood of co-occurrence among closely related species than would be expected by chance. A positive NRI value signifies a clustered or converged phylogenetic structure within the community (Webb et al., 2002). The observed community assembly pattern may be influenced by environmental filtering or niche-based processes, commonly driven by recruitment and mortality dynamics affected by local environmental conditions. This suggests that environmental filtering plays a significant role in shaping the community structure (Letten et al., 2014; Chai et al., 2016).

Throughout the study period, no significant variations were observed in the alpha diversity indices across different spatial locations ( Figure 3 ). This indicates a relatively homogeneous spatial distribution and composition of PPC in the JS. It also suggests that factors influencing the presence and abundance of planktonic protist, such as nutrient availability, water temperature, salinity, and other ecological conditions, did not exhibit significant spatial variability during the study period. However, it is important to note that other aspects of PPC, such as functional traits or specific species compositions, may still exhibit differences, even in the absence of significant variations in alpha diversity indices (Roselli and Litchman, 2017; Anderson et al., 2021).

Overall, our study highlights the influence of seasonal changes in environmental conditions on the alpha diversity of PPCs in the JS. Furthermore, it emphasizes the role of environmental filtering and deterministic community assembly processes during specific seasons, as indicated by the observed patterns in NRI. These findings will contribute to our understanding of the ecological dynamics and assembly mechanisms governing PPCs in this strait.

In the JS, Bacillariophyceae (Ochrophyta) and Dinophyceae (Myzozoa) were found to be the most abundant and diverse groups within PPCs. Several highly abundant OTUs identified in this study were also observed in previous investigations of protistan or microbial eukaryotic communities in the northern East China Sea (Faria et al., 2014; Kim et al., 2020b; Kim et al., 2022; Sun et al., 2022).

The composition of the PPCs varies throughout the year, primarily due to nutrient conditions. For instance, in June, areas characterized by high nutrient conditions, such as shallow-water and front zones, were mainly dominated by nano- and micro-sized diatoms ( Figures 4A, D ). However, in September, when the JS experiences strong stratification and nutrient depletion, the community was dominated by both diatoms and dinoflagellates. With their swimming ability and mixotrophic capabilities, dinoflagellates exhibit competitive advantages under these conditions, allowing them to persist in the euphotic zone and compensate for low nutrient concentrations (Flynn et al., 2019). In November, when nutrient concentrations were high (DIN: 5–10 µM, DIP: 0.6–0.8 µM), diatoms were found to dominate both in composition and relative abundance in the shallow and frontal waters, suggesting that they are strong competitors in environments with elevated nutrient concentrations (Burson et al., 2018).

The surface waters of the deep-water area (station D) in June and all stations in September, which were characterized by limited nutrient availability, account for a significant proportion of smaller nano-sized (2-20 μm) phytoplankton, referred to as nano-sized obligate photoautotrophs (OPAs) ( Figure 4 ). This preference for the smaller size is likely due to the higher surface area to volume ratio of smaller phytoplankton, which confers advantages in nutrient uptake in nutrient-limited environments (Harrison et al., 2015; Fernández-González et al., 2022). Additionally, interactions between predators and prey have been observed to influence the species composition of PPCs in such environments. It has been suggested that there is a stronger predator-prey coupling for smaller phytoplankton (Behrenfeld et al., 2021). Considering the seasonal changes observed in the species composition and relative abundance of the PPC in the JS, mixotrophic and heterotrophic protists showed contrasting quantitative changes compared with larger micro-sized (20-200 μm) PPC ( Figure 4 ). This dynamic could potentially favor the survival of smaller phytoplankton species over larger ones, as they may be more efficiently consumed by their predators, such as heterotrophs or mixotrophs ( Figures 4C, F ). The shift toward smaller phytoplankton can also impact the overall functioning of marine ecosystems and biogeochemical cycles, as areas dominated by smaller phytoplankton often exhibit lower biodiversity levels (Henson et al., 2021).

Dispersal limitation significantly influences plankton community assembly (Lear et al., 2014). Organisms tend to colonize habitats closer to their source, with community similarities decreasing as distance increases. Cell size is also an important factor in shaping community structure through dispersal limitation (de Vargas et al., 2015). The beta diversity index, which measures community similarity, reflects the influence of dispersal limitation. A higher index indicates less similarity and a greater influence of dispersal limitation, while a lower index suggests more similarity and less influence (Zhou and Ning, 2017).

The NCM is commonly used to assess the importance of stochastic processes, particularly neutral processes, in microbial community assembly (Chen et al., 2019; Ramond et al., 2021). The NCM has demonstrated that neutral processes play a dominant role in shaping community variations. In this study, the NCM explained a significant portion (R^{2 =} 0.62) of the variability in the occurrence frequency of the entire PPC, highlighting the importance of neutral processes in driving PPC variations. Similar findings regarding the high contribution of neutral processes have been observed in protistan or microbial eukaryotic communities in the rivers and coastal waters of the South China Sea (Chen et al., 2017; Zou et al., 2021; Xu et al., 2020). This is due to factors such as the high and random dispersal rate of microorganisms, which is caused by the absence of distance barriers and the rapid flow rates of rivers and ocean currents, preventing the formation of distance-decay patterns. These findings underscore the importance of neutral processes in shaping the assembly and variations of PPCs, particularly in environments where dispersal rates are high and distance barriers are limited. The prevalence of neutral processes suggests that chance events and random processes, including dispersal and ecological drift, have a stronger impact on community composition in the JS than deterministic factors.

The dispersal rate of plankton can have varying effects on community structure. If the rate is high, it results in homogenization, while limited dispersal results in dispersal limitation (Vellend and Agrawal, 2010; Logares et al., 2018). The rate of immigration for PPCs in the JS was considerably lower (m = 0.014) than in other studies on microbial eukaryotic communities (m = 0.2–0.4) (Chen et al., 2017; Xu et al., 2020; Zhang et al., 2021), indicating that the PPC in the JS was highly limited by dispersal. This suggests that spatial dispersal limitation had minimal influence on the regional structure of PPC in the JS. Neutral processes, including ecological drift and homogenizing dispersal, primarily drive the variation within PPCs, overshadowing deterministic factors such as environmental conditions and species traits, as revealed by the NCM. Despite a low immigration rate, protists display a remarkable drive to disperse, facilitating a uniform distribution of species and diminishing distinct spatial patterns or distance-decay effects. These findings highlight the significant influence of stochastic processes and random events on community dynamics and protist distribution patterns.

Our results suggest that the assembly of PPCs in the JS is primarily driven by stochastic processes, particularly homogenizing dispersal and ecological drift, with minimal influence from dispersal limitation. Homogenizing dispersal, which accounts for 47% of the contribution, indicates that these plankton communities were widely dispersed and are distributed relatively evenly through the strait. This indicates a high level of similarity in community composition among different locations in the JS. In essence, there is a strong “mixing” effect, with organisms frequently moving and redistributing themselves throughout the ecosystem. Ecological drift, accounting for 51% of the contribution, suggests that random changes in species abundance over time also play a significant role in shaping these communities. The presence or absence of certain species in the community is largely influenced by chance events, such as random births and deaths or unpredictable environmental fluctuations. The very low contribution of dispersal limitation (1.82%) indicates that the movement of organisms across the strait is not significantly restricted. This could be attributed to the absence of physical barriers or the organisms’ strong dispersal abilities. Therefore, most species can reach almost all habitats across the strait, and community assemblages are not strongly influenced by the inability of species to disperse. Overall, these findings suggest that the PPCs in the JS are highly dynamic and are primarily shaped by random processes rather than deterministic environmental selection or dispersal constraints. It implies that these communities are highly resilient to changes in environmental conditions, as they are not tightly bound to specific environmental factors or locations. However, the strong influence of random processes also means that predicting the future development of these communities can be challenging.

In addition, a distinct group represented by yellow dots in Figure 5A has been characterized for several species within Chlorophyta and Myzozoa. This group exhibits a lower frequency but higher relative abundance compared to the prediction for neutrally occurring species, as shown by the NCM. Notably, species belonging to the Chloropicophyceae class are commonly found in oceanic waters, indicating their preference for such habitats (Lopes dos Santos et al., 2017). Another noteworthy genus is Bathycoccus, a cosmopolitan picophytoplankton genus classified in the Mamiellophyceae. Bathycoccus prasinos, the sole species within this genus, is a small non-motile green microalga characterized by intricate scales. It plays a vital role in the marine ecosystem by significantly contributing to primary oceanic production (Belevich et al., 2020). Interestingly, Bathycoccus prasinos has been identified as the most abundant species in the East China Sea during the Spring season (Xu et al., 2016). The below group comprises several species within the Myzozoa, specifically belonging to the Dinophyceae class, including notorious species such as A. catenella, A. pacificum, A. tamarense, Protoceratium reticulatum, and Prorocentrum triestinum, which are all known for causing harmful algal blooms (HABs) globally and in Korea specifically (Alves-de-Souza et al., 2014; Anderson et al., 2021; Min and Kim, 2023). HABs are typically episodic, occurring infrequently but in exceptionally high concentrations when they do happen. In the JS, these taxa exhibited a higher relative abundance in March compared with other periods. For example, the bloom of P. reticulatum in a Chilean fjord during the summers of 2016 and 2017 exemplified this occurrence pattern, with blooms observed at an average water temperature of 16°C (Alves-de-Souza et al., 2019).

In this study, we investigated the relationship between eight distinct modules and various environmental variables, with a particular focus on monitoring changes in their normalized relative abundance over time. We found that Modules I, III, IV, and VIII had significant associations with specific environmental factors, indicating potential influences on their biological characteristics. However, it is important to note that correlation does not imply causality, and further research is needed to validate and obtain a more comprehensive understanding of these relationships.

The plankton community network analysis of all samples in this study revealed that Module I primarily consisted of diatoms and other silicate-utilizing organisms, including micro-sized diatoms such as Coscinodiscus, Rhizosolenia, and Biddulphia genera ( Table S3 ). The presence of Module I was found to positively correlate with DISi. This suggests that these genera can efficiently utilize dissolved inorganic silica to construct and maintain their cell walls (Martin‐Jézéquel et al, 2000).

Module III shows a clear seasonal pattern, with its normalized relative abundance peaking in June and reaching its lowest point in September. The partial Mantel test analysis revealed significant positive correlations between Module III and both temperature (r = 0.566, p = 0.002) and salinity (r = 0.610, p = 0.005). However, there were insignificant negative correlations with DIN, DIP, and DISi. These findings suggest that the species represented in Module III primarily comprised common nano-sized diatoms, including Skeletonema, Thalassiosira, and Chaetoceros, which have successfully adapted to thrive in warmer and saltier conditions. However, their activity and abundance may be curtailed during the summer months due to nutrient constraints in the JS.

Module IV, which displays a seasonal pattern similar to Module III, exhibits a significant positive correlation with both temperature and pH. This suggests that the organisms within this module may thrive under specific temperature conditions and pH levels. They could be adapted to warmer conditions and specific pH ranges that occur at certain times of the year, which explains the observed seasonal peak in June. The major species belonging to Module IV include Pseudonitzschia spp., Alexandrium minutum, and Akashiwo sanguinea, all types of plankton that play significant roles in marine ecosystems. These species are known for their ability to form blooms in coastal Korean waters, which can sometimes be harmful to other marine life because of the toxins they produce, such as domoic acid and neurotoxins, which can cause amnesic and paralytic shellfish poisoning in humans (Min and Kim, 2023).

The traits of Module VIII seem to be significantly affected by both seasonal transitions and certain environmental conditions, particularly temperature and salinity. The abundance of organisms in Module VIII follows a distinct annual pattern, peaking in September and gradually diminishing to zero by November. This pattern suggests a significant seasonal impact on the presence and prevalence of organisms within Module VIII, which may correspond to specific life cycle stages or the availability of resources. The organisms in this module might possess adaptations suited to environments with higher salinity and specific temperature ranges. However, because only four species within this module were identified, providing detailed information about their individual characteristics will be problematic.

Modules II, V, VI, and VII do not show significant correlations with the environmental variables, based on the results of partial Mantel tests, which suggests that these modules may operate independently from the tested variables and their variations. However, it is important to note that this interpretation is specific to the current context, and there may be other factors or complex interactions that could influence these modules.

This study employed SoEM sequencing methodology to investigate the seasonal and spatial dynamics of the PPC in the JS. The identification of 61 modules within the PPC network aligns with the concept of ecological niches, suggesting the presence of distinct subgroups or functional units. This modular structure indicates that different species within the PPC specialize in specific ecological roles or inhabit different habitats, contributing to the stability and resilience of the ecosystem (Forster et al., 2021). The modular organization of cellular networks, as observed in protein-interaction studies in yeast, suggests that specific biological responses may be controlled by such modules (Xue et al., 2018; Liu et al., 2019). Similarly, in the case of protists, clusters derived solely from network structure may define modules that fulfill similar biological roles (Rives and Galitski, 2003; Lewis et al., 2010). Thus, the modular structure observed in the PPC network indicates the presence of distinct ecological niches, which in turn contribute to ecosystem stability and resilience. This analysis enhances our understanding of species co-existence, resource utilization, and environmental adaptability.