Samples were collected from five stations located in the UB aboard R/V Tamgu 3, in August (summer) and December (winter) 2019 ( Figure 1B ). At each station during the two seasons, temperature, salinity, and dissolved oxygen (DO) were determined using an attached sensor system (Sea-Bird SBE 911 conductivity–temperature–depth) on a rosette sampler (Sea-Bird Electronics, USA). The CTD was calibrated by the manufacturer before we proceed with the measurements at stations and the accuracies of temperature, conductivity, oxygen, and pressure sensors were ± 0.001°C, ± 0.0003 S m ^-1, ± 2 % and ± 0.015% of the full-scale range, respectively. To determine the nutrients [nitrite ( N O 2 − ), nitrate ( N O 3 − ), phosphate ( P O 4 3 − , denoted as PO₄), and silicate (SiO₂)], seawater was collected in 15 ml pre-acid rinsed plastic bottles and analyzed onboard using a continuous flow auto-analyzer (QuAAtro 39, Seal Analytical, UK). The analytical precision of the nutrients was better than 1%. The nutrient measurements were performed using same method as the used by Hydes et al. (2010) in the GO-SHIP protocol and reference materials (Kanso RMNS) was used to track the performance of the system and the consistency of nutrient data. The nitrite + nitrate observed in this study is defined as dissolved inorganic nitrogen (DIN). To determine chlorophyll-a (Chl-a), 500 ml of seawater was filtered through 25 mm Whatman GF/F filters and frozen at -80°C. In the laboratory, pigments were extracted using 90% acetone (Parsons et al., 1984) and the Chl-a concentrations were measured using a Trilogy fluorometer with an analytical precision of ± 0.05 µg L^-1 (CHL NA, Model #7200-002; Turner designs, USA). For Chl-a measurement, EPA 445.0 was used as the standard operating procedure for fluorometric determination. At all stations (5 stations × 2 seasons = 10 samples), 2L of surface seawater was filtered through 0.22 µm cellulose ester membrane filters (Millipore, Ireland) to capture bacterial cells. Filtered samples were immediately frozen and stored at −80°C until DNA extraction.

Total DNA was extracted from the obtained samples following a standard protocol using a DNA isolation kit (Qiagen, Germany) and the quality of the extracted DNA was cross-checked using NanoDrop (Thermo Scientific, USA). After quality control (QC), the 16s rRNA sequencing library was constructed according to the 16S metagenomic sequencing library preparation protocol (Illumina, San Diego, CA, USA) targeting the V3 and V4 hypervariable regions of the 16S rRNA genes. PCR was performed using the 16S amplicon PCR forward primer: 5′- TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGTCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG-3′ and reverse primer: 5′- GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′. The extracted DNA samples were amplified and analyzed using Illumina MiSeq (MiSeq Control Software v2.4.1.3) and Real-Time Analysis (RTA v1.18.54.0). Subsequently, 16S tagging of DNA fragmentation (100–300 bp) was performed using the quantitative kit for sample DNA. Primers were prepared for all DNA fragments, amplified using an adapter, and subjected to library QC using a bioanalyzer, followed by clustering. Sequencing data were processed using QIIME1.9.1 to assemble paired-end reads into tags according to their overlapping relationships. In the pre-processing step, the primer was removed and then demultiplexing and quality filtering (Phred ≥20) were applied. USEARCH7 was used to perform denoising and chimera detection/filtering to group operational taxonomic units (OTUs). The Silva132 and National Center for Biotechnology Information (NCBI) databases were used to determine OTUs with 97% similarity using UCLUST and the close-reference analysis. The OTU table was normalized by dividing each OTU by its 16S copy number abundance. After filtering the generated OTU table using the Biological Observation Matrix format, the resulting sequences were clustered into OTUs based on a similarity threshold of ≥97% using Python Nearest Alignment Space Termination (PyNAST) (Caporaso et al., 2010). Using RDP classifiers, we performed comparative OTU assignments with the database regarding phylum, class, order, family, genus, and species. Permutational multivariate analysis of variance (PERMANOVA) was used to test for significant differences among the groups. The relative abundance of phyla, classes, orders, families, genera, and species was estimated using the STAMP program. Multivariate analysis was performed using principal component analysis, whereas univariate analysis was performed by ANOVA. Raw sequence data from this study were deposited in the NCBI with Sequence Read Archive (accession number: PRJNA778576).

The surface salinity data for August and surface temperature data for December used in this study were obtained from NASA Ocean Color (https://oceancolor.gsfc.nasa.gov/). Multiple statistical and similarity analyses (ANOSIM; a non-parametric permutation procedure that tests whether differences in dissimilarity between groups exceed differences within groups) were conducted with bray-curtis method using the R program (v.3.6.2, https://www.R-project.org) with gplots, ggplot2, Heatplus, and vegan packages. Non-metric multidimensional scaling (NMDS) analysis was performed using the unweighted Unifrac method to indicate surface bacterial dissimilarity. Shannon diversity indices and Chao1 species abundance indices were used to estimate the alpha diversity of bacterial communities. Distance-based redundancy analysis (db-RDA) was performed to understand the relationships between the bacterial communities and the surrounding physicochemical parameters. Spearman’s rank-order correlation coefficients were used for non-normally distributed data (i.e., between the relative abundance of bacterial taxa and physical and biogeochemical factors) to define the degree of dependence among the variables, giving a value between -1 and +1 (1 represents a total positive correlation, 0 represents no correlation, and -1 represents a negative correlation) using R software.

The environmental parameters obtained during the investigation showed variable results at each station ( Figure 1 and Table 1 ). In particular, the summer (August) sea surface salinity data showed lower salinity CDW (<31 PSU) widely distributed from Changjiang River mouth to the ECS ( Figure 1C ). A large discharge of CDW extends to near 28°N, and in turn, the salinity of CDW increases (32–33 PSU) because of the mixing of warm and saline TWW (>20°C and <33.8 PSU) via the Tsushima Current (Park, 1978; Yang et al., 1991; Cho et al., 1997; Chang and Isobe, 2003; Moon et al., 2009). TWW and CDW flowed into the UB through the western channel of the KS ( Figures 1B, C ), the main path of fresh CDW resulting from strong halocline stratification (Senjyu et al., 2006; Kim et al., 2014). Consequently, the conducting temperature-salinity diagram explains the distribution of the three different water masses during summer ( Figure 1D ). The surface temperature and salinity observed at St-3 were 25.01°C and 32.74 PSU, respectively, with relatively high PO₄ (0.02 µM) and SiO₂ (5.63 µM) concentrations ( Table 1 ). The vertical salinity profile also showed that CDW distinctly affected St-3 at pressures of 0–8 dbr ( Figure 1E ). The mean surface temperature and salinity at Sts-1, 2 and 4 were 25.8°C (range: 24.9–26.46°C) and 33.45 PSU (33.54–33.38 PSU), which were higher than those at St-3, while mean nutrient concentrations were relatively lower (DIN, 0.39 µM; PO₄, 0.01 µM; and SiO₂, 3.40 µM) ( Table 1 and Figure 1D ), indicating the influences of TWW. Relatively low temperature (21.55°C) and high DO (8.21 mg L⁻¹) and DIN concentration (0.86 µM) were observed at St-5 ( Table 1 ), located at the highest latitude among other stations with North Korea Surface Water (NKSW), originating from the northern ES and is characterized by low temperature and salinity near 40°N in the UB (Kim and Lee, 2004). Based on these observations, it is evident that the physical mixing of three different water masses (TWW, CDW, and NKSW) influences the UB region during summer. We categorized these observed stations into three groups: TWW (Sts-1, 2, and 4), CDW (St-3), and NKSW group (St-5).

In summer, metagenome samples collected from five stations and a total of 490,087 operational taxonomic units (OTUs) were identified, belonging to seven phyla, 10 classes, 23 orders, and 37 family levels in the surface bacterial communities ( Supplementary Table S1A–D ). To determine the influence of physical mixing on bacterial composition, we performed alpha (Shannon diversity and Chao1 species richness indices) and beta diversity (clustering dendrogram, NMDS, dbRDA, and heatmap) analyses ( Figures 2 and 3 ). Clustering analysis computed by Bray–Curtis dissimilarity based on OTUs revealed that stations affected by the same water mass were clustered first: Sts-1, 2, and 4; St-3; and St-5, in sequence (r = 0.12, P < 0.5) ( Figure 2A ). The CDW group was clustered with the TWW group before the NKSW group ( Figure 2A ) as the CDW was transported from the ECS to UB by the Tsushima current system ( Figure 1A ). The NMDS plot represents the dissimilarity of the surface bacterial community with distances in the configuration (Shepard, 1962; Kruskal, 1964; Holland, 2008), showing remarkable dissimilarity of St-3 ( Figure 2B ) during summer. This was further confirmed by the Shannon diversity and Chao1 species abundance indices with lower values (4.84 and 422.15, respectively) at St-3 ( Figures 2C, D ). Moreover, the Shannon and Chao1 indices across Sts-1, 2, and 4 were consistently high (mean value: 5.37 and 497.64, respectively). Commonly, at higher latitudes, lower values of Shannon and Chao1 were discovered within the TWW group, whereas St-5 showed a contradictory pattern with relatively high Shannon value but low Chao1 values of 5.38 and 433, respectively ( Figures 2C, D ). The Chao1 value of St. 4 which belongs to the TWW group, was low in the TWW group; however, it was relatively high compared to that of Sts-3 and 5. As St. 4 is close to St. 5, the St. 4 Chao1 value seems to stand out within the same group.

Visual inspection of a heatmap revealed differences in the relative abundance at the family level (top 37 genera, relative abundance >0.01%) of surface bacterial communities from five stations during summer ( Figure 3A ). Even within the same TWW group, some bacteria showed differences in relative abundance by station. The relative abundances of Spongiibacteraceae, Halomonadaceae, Cryomorphaceae, Pseudomonadaceae, Neisseriaceae, Pseudoalteromonadaceae, Hyphomonadaceae, and Erythrobacteraceae were high at St. 1 whereas those of Bacillaceae, Williamsiaceae, Microbacteriaceae, Enterobacteriaceae, Planctomycetaceae, Geodermatophilaceae, and Moraxellaceae were high at St. 2. In addition, Alteromonadaceae, Flammeovirgaceae, Rhizobiaceae, Mycobacteriaceae and Hyphomicrobiaceae showed high relative abundances at St. 4. However, Vibrionaceae, Alcanivoracaceae, Idiomarinaceae, Pelagibacteraceae, and Verrucomicrobiaceae were abundant in Sts-1 and 2. In St-3, the relative abundance of Micrococcaceae (belonging to the class Actinobacteria) and Rhodospirillaceae (Alphaproteobacteria) were higher, whereas that of Flavobacteriaceae (Flavobacteriia) and Halieaceae, were high at St-5, even if it was a low value. As shown in Figures 2C, D , which showed high diversity and species abundance, high relative abundance species were more diverse in the TWW group (Sts- 1, 2 and 4) than in other stations. The db-RDA analysis was conducted to identify the environmental factors shaping the bacterial community composition at the family level ( Figure 3B ; with relative abundance >0.01% in at least one of the samples). Sts-1 and 2 (affected by warm and saline TWW) were closely linked to temperature and salinity. Additionally, owing to the influence of high temperature and saline TWW and its geographical features closely located to St. 5, St. 4 appears to be associated with salinity and DIN ( Figure 3B ). St-3 (affected by CDW with relatively high PO₄ and SiO₂ concentration) and St-5 (affected by NKSW with high DO and DIN concentration) were associated with PO₄ and DIN, respectively (St. 3 is not closely matched with salinity because the dbRDA plot has shown that high values are connected). These results suggest that the bacterial structure and composition at each station were closely related to the physicochemical characteristics (P < 0.05) that were influenced by different water masses.

Family level bacteria with high relative abundance at one or two stations in each three group (i.e., TWW [Sts-1,2, and 4], CDW [St-3], and NKSW [St-5]) were identified (20 genera). Consequently, they showing considerable differences between the three groups with a correlation value of ≥ 0.5 with the physical and biogeochemical factors of influencing water masses ( Figure 4 ). Verrucomicrobiaceae (belonging to class Verrucomicrobiae), Alcanivoracaceae, Pseudomonadaceae, Spongiibacteraceae (Gammaproteobacteria), and Cryomorphaceae (Flavobacteriia) were highly abundant at Sts-1, 2, and 4 [TWW], influenced by high temperature, salinity, and poor-nutrients depicting a strong positive correlation with temperature (r = 0.97, 1.0, 0.97, 0.89, and 0.89, respectively), and salinity (r = 0.2, 0.3, 0.41, 0.67, and 0.67, respectively) and a negative correlation with DIN concentration (r = -0.97). In addition, Nocardiaceae, Williamsiaceae (belonging to class Actinobacteria), and Vibrionaceae, Spongiibacteraceae (Gammaproteobacteria) were highly dominant in Sts-1, 2, and 4, while Pseudoalteromonadaceae (Gammaproteobacteria), Hyphomonadaceae, Erythrobacteraceae (Alphaproteobacteria), and Cryomorphaceae (Flavobacteriia) at the family level showed positive and negative correlations with temperature and DIN. At St-3 [CDW], the relative abundance of Micrococcaceae, belonging to class Actinobacteria which is mainly distributed in the Changjiang River (Chung et al., 2015), and Rhodospirillaceae (Alphaproteobacteria) were higher and showed a positive relationship with PO₄ (r = 0.58 and 0.4, respectively) and SiO₂ concentrations (r = 0.2 and 0.98, respectively) ( Figure 4 ). Flavobacteriaceae (belonging to class Bacteroidetes) reflects specific substrate conditions related to the phytoplankton community (Rink et al., 2011) and showed high relative abundance at St-5 [NKSW] ( Figure 3A ), with a positive correlation with DIN (r = 0.9; P < 0.001), DO and Chl-a concentration (r = 0.7 and 0.21). Based on these results, environmental conditions with specific water mass (TWW, CDW, and NKSW) characteristics ( Figure 4 ) are related to a specific group of bacterial relative abundances ( Figure 3 ) during summer.

Spatial patterns in the marine bacterial structure and community composition have been reported across oceanic water masses with distinct hydrographic properties (Baldwin et al., 2005; Giebel et al., 2009). However, no studies have shown the influence of physical process on bacterial composition in ES, where complex physical mixing of water masses could cause variation of bacterial community composition. In the Kuroshio and Oyashio water masses from the western North Pacific Ocean, Kataoka et al. (2009) reported that different hydrographic conditions, that is temperature, salinity, and nutrient concentrations influence the elevation of bacterial composition, and variation in their phylotype structure in each water mass. Yokokawa et al. (2010) revealed the influence of water masses’ on the spatial variations in bacterial composition abundance, similarity, and activity in the eastern Mediterranean Sea, reflecting their responses to the surrounding physicochemical characteristics. Along with this physiological evidence, recent metagenome studies from the Western Subarctic Pacific Ocean (Li et al., 2018) and the Western Pacific Ocean (Zhang et al., 2018) have shown that surrounding physicochemical parameters could elevate the fundamental metabolic activity of bacterial communities and their composition. Similar to these observations, water masses were classified in this study according to their physicochemical properties, which can elevate the surrounding bacterial communities and their metabolic activities during summer. This assumption was further confirmed by alpha diversity of the bacterial community (Shannon and Chao1 indices) and a positive correlation with the features of water masses associated with physicochemical parameters ( Figure 2 ).

Similar to the present study area (semi-closed marginal sea and occurrence of seasonal physical mixing) multiple investigations have been conducted in the Baltic Sea (Herlemann et al., 2011; Koskinen et al., 2011), demonstrating that the salinity gradient influenced the bacterial composition. However, these studies did not show a clear pattern of richness or diversity contrary to the observations in the present study, which proved a clear distinction between the three water masses (TWW, CDW, and NKSW). Specifically, a bacterial family of 20 genera, among 37 genera showed evident relative abundance differences by physically mixing the three water masses ( Supplementary Table S2 ). For example, 17 genera were dominant at Sts-1, 2, and 4 (influenced by TWW), strongly correlated with temperature and salinity, and negatively correlated with DIN concentration ( Figure 4 ). In addition, two genera, Micrococcaceae and Rhodospirillaceae showed remarkably high relative abundances in St-3 (influenced by CDW) with a high correlation to PO₄ and SiO₂ concentrations. The highly abundant Flavobacteriaceae at St-5 (affected by NKSW) strongly correlated with the DIN, DO, and Chl-a concentrations. These results suggest that our observation of hydrographic characteristics was formed by three different physical mixing processes that influenced surface bacterial communities and composition during summer in the UB.

Sea surface temperature data obtained from satellite observations and environmental data collected from field observations during winter (December) showed distinct physicochemical parameters ( Figure 5 and Table 1 ). This variation could be reflected in the influence of the clockwise UWE ( Figure 5A ) during this investigation, which is identifiable as a 2°C lower cold-core (16°C) wrapped by the warm filament (18°C). The surface temperature data further depict the EKWC (northward branch of warm and saline Tsushima current that flows along the Korean coast) curved between 37 and 38°N and entrained high Chl-a concentration in coastal waters that formed the UWE ring, supplying highly productive coastal waters to the center of the UB (Hyun et al., 2009; Kim et al., 2012). With the occurrence of UWE in UB during winter, the physicochemical characteristics at St-1 (located at the UWE ring; Figure 5A ) showed a lower temperature (13.22°C) but higher Chl-a concentration (1.05 µg L⁻¹) compared to those outside of the eddy ring at St-2 temperature (14.62°C) and Chl-a concentration (0.67 µg L⁻¹) ( Figure 5B and Table 1 ). St-3 (located between Ulleung Island and Dokdo) environmental parameters ( Figure 5A ) showed a fluctuating environment, which could be the reason for the strong wind (Yamada et al., 2004; Baek et al., 2018) characterized by the highest DIN (6.60 µmol L⁻¹), PO₄ (0.34 µmol L⁻¹), and SiO₂ (7.02 µmol L⁻¹) concentrations ( Table 1 ). Temperature and salinity in Sts-4 and 5 were similar at 14.21–14.22°C and 33.86–33.89 PSU, respectively, with the lowest Chl-a concentration (0.38–0.54 µg L^-1), indicating that these stations were influenced with surface water mass ( Figure 5B , Table 1 ). The temperature-salinity diagram ( Figure 5B ) confirmed that the occurrence of UWE resulted in differences in physicochemical data from five stations and was categorized similarly within three groups: the UWE ring group (Sts-1 and 2, varied between inside and outside of the UWE ring), Ulleung Island and Dokdo (UID) station (St-3, affected by the geographical features between UID), and surface water group (Sts-4 and 5, affected by wintertime surface water).

In total, 254,712 OTUs were identified in winter belonging to 10 genera in the phylum, 11 genera in the class, 23 genera in the order, and 33 genera at the family level ( Supplementary Table S3A–D ). Cluster analysis of the five samples with Bray–Curtis dissimilarity showed ( Figure 6A ) that bacterial communities were divided into three clusters with UWE ring group (Sts-1 and 2), surface water group (Sts-4 and 5), and UID station (St. 3). The NMDS plot, describing the dissimilarity of the surface bacterial community ( Figure 6B ), showed that Sts-1 and 2 were geographically distant from the other three stations, indicating that the bacterial community dissimilarity at Sts-1 and 2 were higher. The Shannon diversity indices of the five stations ( Figure 6C ) displayed that the lowest Shannon index with a value of 3.92 was observed at St-1, located at the UWE ring. A study reported that these low trophic level organisms (heterotrophic bacteria) dominate in high-nutrient UWE rings formed with coastal upwelling water (Hyun et al., 2009). In addition, Shannon diversity indices at Sts-1 and 2 ( Figure 6C ) showed a reverse pattern with the Chao1 richness index ( Figure 6D ) as the Shannon diversity index in St-1 was lower than that at St-2 (3.92 and 4.04, respectively), and the Chao1 species abundance index at St-1 was higher than that at St-2 (327.5 and 308.16, respectively). However, lower Shannon and Chao1 values were observed at St-4 (3.99 and 311.96) than at St-5 (4.24 and 367), unlike Sts-1 and 2 ( Figures 6C, D ). The highest Shannon and relatively high Chao1 index values (4.24 and 344.88) were observed at St-3, characterized by the highest DIN, PO₄, and SiO₂ concentrations ( Table 1 ), indicating that this result is associated with nutrient concentration, which significantly affects bacterial niche structure (Cullen, 1991), shaping bacterial communities (Dai et al., 2022) with changes in community diversity and an increase in species abundance as nutrients increased (Gilbert et al., 2012).

Along with these similarities and diversities ( Figure 6 ), the relative abundance of the top 33 family level bacteria is shown in the heatmap ( Figure 7A ), and the relative abundance differences in each group were different. Specifically, the relative abundance of Enterococcaceae, Prochloraceae, Flavobacteriaceae, and Porticoccaceae was higher at St. 1, whereas Pseudoalteromonadaceae was abundant at St-2 ( Figure 7A ). However, Verrucomicrobiaceae showed a high relative abundance at Sts-1 and 2. In addition, Staphylococcaceae, Propionibacteriaceae, and Streptococcaceae were highly abundant at St. 4, whereas Planctomycetaceae, Vibrionaceae, Mycobacteriaceae, Rhodobiaceae, and Alcanivoracaceae were highly abundant at St-5 (Shewanellaceae was abundant in Sts-4 and 5, surface water group). However, there was a physicochemical difference between Sts-1 and 2 due to the occurrence of the UWE ring ( Figure 5B ), clustering ( Figure 6A ) and heatmap ( Figure 7A ) that showed that community composition similarity was high, and bacteria with high relative abundance in common at Sts-1 and 2 are similar. This is considered to be due to other biological factors (i.e., interactions with other microorganisms and phytoplankton community) not covered in the present study potentially influencing the bacterial community structure of the UWE ring group. However, eight bacteria, Hyphomonadaceae, Sphingomonadaceae, Balneolaceae, Lactobacillaceae, Alteromonadaceae, Bifidobacteriaceae, Methylobacteriaceae, and Nocardiaceae showed high relative abundance at St-3, consistent with the result of the high Shannon diversity index ( Figure 6C ). The db-RDA analysis revealed the physicochemical factors that played a determinant role in shaping surface bacterial community composition at the family level (relative abundance >0.01% in at least one of the samples; Figure 7B ). In addition, Sts-1 and 2 were closely linked to Chl-a and salinity St-3 was related to DIN concentration, and Sts-4 and 5 were intertwined with temperature ( Figure 7B ). These results revealed significant differences in relative abundance (at the family level) in each of the three groups of bacterial communities, which were greatly influenced by environmental features developed by the occurrence of UWE.

The specific influences of environmental characteristics on the bacteria at the family level (20 genera) showed evident differences in the relative abundance of the three groups (UWE, UID, and surface water) ( Figure 8 ). Flavobacteriaceae (member of the order Flavobacteriales), observed to be dominant at Sts-1 and 2, was particularly highly abundant at St-1, located in a high Chl-a concentration UWE ring with a positive correlation with DO and DIN concentrations (r = 0.3 and 0.1). This genus was previously reported to be associated with phytoplankton and responds rapidly to nutrient-rich eddy circulation (Fuhrman et al., 2015). Similarly, a high abundance of Flavobacteriales in euphotic eddy-influenced samples has been observed in the northeastern South China Sea (McGillicuddy et al., 2007). Moreover, Prochloraceae and Verrucomicrobiaceae abundant at Sts-1 and 2 showed high correlations with salinity (r = 0.60) and Chl-a concentration (r = 0.67). In particular, Prochloraceae belonging to the genus Prochlorococcus are significantly influenced by warm or cold eddies (Jing and Liu, 2012; Wang et al., 2016). However, Sphingomonadaceae, Nocardiaceae, Lactobacillaceae, Hyphomonadaceae, Bifidobacteriaceae, Balneolaceae, and Methylobacteriaceae were positively correlated with high DIN, PO₄, and SiO₂ concentrations, which were abundant at St-3. Vibrionaceae, Erythrobacteraceae, Staphylococcaceae, Streptococcaceae, and Mycobacteriaceae were abundant at Sts-4 and 5 (affected by low-temperature surface water) and showed a positive correlation between temperature and relative abundance.

Eddies exert a major control on the phytoplankton productivity by nutrient flux (Huang et al., 2010; He et al., 2016; He et al., 2017; Liu et al., 2017), and on the associated bacterial community structure, diversity, and function (Harke et al., 2021). Eddy offshore shifts from large to small phytoplankton cells and higher to lower autotrophic biomass, has a complex and variable system regarding its hydrological structure and distribution of biological properties and has the highest organic matter concentration (Arístegui et al., 2004). Bacterial communities were also observed to be influenced by cyclonic and anticyclonic eddy circulation. For example, earlier studies noted that bacterial responses to mesoscale eddies (approximately 200km diameter) elevated their biomass and production inside a cold-core eddy compared to the surrounding waters (Thyssen et al., 2005; Zhang et al., 2009; Baltar et al., 2010). More specifically, a recent study by Devresse et al. (2021) on oligotrophic oceans observed that eddies influence the elevation of bacterial abundances and their metabolic activities, that is respiration rate, biomass production, and growth efficiency. Similar to these studies, present study results showed that the difference in bacterial composition at the UWE influenced (Sts-1 and 2) and not influenced stations. In the Sargasso Sea, prokaryotic heterotrophic production increases at the perimeter of the cyclonic eddy relative to the eddy center (Ewart et al., 2008). Likewise, considerable spatial heterogeneity in bacterial community composition was apparent inside and outside the cold-core cyclonic eddies of the subtropical marginal South China Sea (Zhang et al., 2011). Moreover, Li et al. (2014) demonstrated that bacterial communities derived from anticyclonic eddy sites displayed abnormal compositions compared to those of adjacent sites in the Northern South China Sea.

The five stations during winter were categorized based on their environmental features: the UWE ring group (Sts-1 and 2), distinguished from the Chl-a concentration that classifies the effects of the UWE ring with relatively high salinity values; the UID station (St-3), which showed high DIN, PO₄, and SiO₂ concentrations; and surface water group (Sts-4 and 5), influenced by wintertime surface water. These physicochemical features played a decisive role in the similarity, diversity, and distribution of the surface bacterial composition at each station related to the relative abundance of bacteria ( Supplementary Table S4 ). Interestingly, we observed a clear distinction in the bacterial communities at the family-level Pelagibacteraceae, Halieaceae, Verrucomicrobiaceae, Porticoccaceae (belonging to Proteobacteria), Planctomycetaceae (belonging to Planctomycetes), Enterococcaceae (belonging to Firmicutes), Flavobacteriaceae (belonging to Bacteroidetes), and Prochloraceae (belonging to Cyanobacteria), which were more predominant in Sts-1 and 2. In contrast, Sphingomonadaceae, Hyphomonadaceae, Methylobacteriaceae (belonging to Alphaproteobacteria), Lactobacillaceae (belonging to Firmicutes), Bifidobacteriaceae, Nocardiaceae (belonging to Actinobacteria), and Balneolaceae (belonging to Balneolaeota) were predominantly found in St-3. Meanwhile, the strongest indicator species for Sts-4 and 5 were Vibrionaceae, Erythrobacteriaceae (belonging to Proteobacteria), Mycobacteriaceae (belonging to Actinobacteria), Staphylococcaceae, and Streptococcaceae (belonging to Firmicutes). Collectively, our data indicate that the physical circulation of the UWE during winter was influences the physicochemical conditions of the surrounding water and consequently alters the surface bacterial composition in the UB, ES.