Seawater samples were collected using Go-Flo bottles mounted on a conductivity, temperature, and depth (CTD) rosette (Sea-Bird 91 Electronics, Bellevue, WA, USA). A total of 30 stations with different depths were sampled in the ECS during a summer cruise in July 2019 ( Figure 1 ). Hydrographic data were measured using a CTD profiler. The nutrient variables, including dissolved inorganic nitrate, phosphate, silicate, and chlorophyll a (Chl a) concentration, were measured according to the standard methods developed by Gong et al. (2003). The precision for the determination of nitrate, phosphate, silicate, and Chl a were ± 0.3 µM, ± 0.5 µM, ± 0.01 µM, and ±0.02 mg/m³, respectively.

For DNA samples, 10 L of three-layered seawater (surface, deep chlorophyll maximum (DCM), and bottom) were pre-filtered through a 200-μm nylon mesh, followed by filtering through a 20-μm nylon filter (90 mm diameter, 100 Millipore, Billerica, MA, USA), and then subsequently filtered through 3-μm and 0.2-μm pore size polycarbonate membranes (142 mm diameter, Millipore, USA) using a peristaltic pump. The filters were immediately preserved in liquid nitrogen and stored at −80°C until DNA extraction.

DNA was extracted using the DNeasy PowerWater Kit (Qiagen, Valencia, CA, USA). The V4 region of rDNA was amplified using TAReuk454FWD1 (5′-CCA GCA SCY GCG GTA ATT CC-3′) as the forward primer and the modified TAReukREV3 (5′-ACT TTC GTT CTT GAT YRA-3′) as the reverse primer (Stoeck et al., 2010). PCR conditions were as described in Lin et al. (2022). Amplicons were sequenced on an Illumina MiSeq platform, generating 300-bp paired-end reads. Data have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) database under the project number PRJNA738614.

The sequence primers were removed using cutadapt (Martin, 2011) and then underwent quality filtering and denoising using the standard pipeline of the DADA2 package (Callahan et al., 2016) under R software. Reads were filtered with the following parameters: truncLen and minLen = c (240, 180), truncQ = 2, and maxEE = c (2, 2). Taxonomic assignments of amplicon sequence variants (ASVs) were conducted using the Protist Ribosomal Reference Database (PR2) version 4.12.0 (Guillou et al., 2013). Planktonic ciliate amplicons were selected corresponding to the orders Tintinnida, Choreotrichida, and Strombidiida, among the division Ciliophora in PR2. Ciliate communities were subsampled with 250 reads for 100 times.

All statistical analyses were conducted using software R (Team, 2013). Alpha diversity indices (Shannon index H′ and richness) and the Bray–Curtis dissimilarities were calculated using the Vegan package (Oksanen et al., 2019). The non-metric multidimensional scaling (NMDS) plot was performed to visualize the patterns and identify potential environmental drivers for the ciliate community composition based on the log-transformed ASV abundances and scaled environmental variables. Venn diagram was made using the ggVennDiagram package (Gao et al., 2021). Ciliate communities were grouped by hierarchical cluster analysis using Ward’s minimum variance method and square-rooting Bray–Curtis distances with cluster package (Maechler et al., 2022). To understand the relationship between the environment and each cluster, differences in environmental variables among clusters were examined using Tukey’s honestly significant difference (HSD) post-hoc test, followed by an analysis of variance (ANOVA). Graphics were created using the ggplot2 package (Wickham, 2016). The plots of sampling stations, temperature–salinity (T-S) diagram, and vertical profiles of environment variables were generated in the Ocean Data View (Schlitzer, 2022). Small-sized planktonic ciliates (<20 µm) indicated sequences obtained from the nano-sized fraction. The pico-sized reads were counted for the presence of ASVs to reveal the distribution.

During this cruise, seawater temperature and salinity ranged from 17.8°C to 29.7°C and 29.8 to 34.8, respectively. Four water masses observed in the study area followed the definition by Gong et al. (1996): TCWW, Yellow Sea Mixed Water (YSMW), CDW, and Kuroshio water (KW) ( Figure 1 ). The YSMW influenced the bottom of the northern transect in the ECS, while the coastal surface area was influenced by the low-salinity CDW (S < 31; Supplementary Figure S1B ). The KW intruded in northeastern Taiwan, involving a topographic upwelling at St. 1 ( Supplementary Figure S1 ). The TCWW dominated the shelf area of most of the sampling sites ( Figure 1C ).

We obtained a total of 15,980 ASVs represented by ~11,465,721 DNA reads. Alveolata contributed largely to total reads, particularly in nano-sized fraction (73.9%, data not shown). A total of 314 ASVs (21,336 reads) remained as planktonic ciliates across three size fractions after subsampling (19, 36, and 29 samples from micro-sized, nano-sized, and pico-sized fractions, respectively). A complete list of ciliate taxa from this dataset is reported in Supplementary Table 1 . Overall, nano- and pico-sized fractions share more similarities than micro-sized where the Bray–Curtis distances were closer in the NMDS plot ( Supplementary Figure S2D ). The alpha diversity and species richness were lower in the micro-sized fraction when compared to both the nano- and pico-sized fractions ( Supplementary Figure S2A ). Tintinnida accounted for more than 50% of the ciliate relative abundance in the micro-sized fraction, in which Salpingella sp., Amphorellopsis acuta, and Stenosemella pacifica were found to be the major contributors ( Supplementary Figure S3 ). Strombidiida dominated in two small-sized fractions, which were Strombidiida_G_XX_sp., Strombidiidae_H_X_sp., and Strombidiidae_G_X_sp. in the nano-sized fraction, and Strombidiida_G_XX_sp., Spirotontonia_sp., and Tontonnidae_B_X_sp. in pico-sized fractions (43.97% and 52.93%, respectively; Supplementary Figures S2B , S3 ). Leegaardiella sp. was the most abundant taxon across three size fractions in the ECS.

Hierarchical cluster analysis based on the Bray–Curtis dissimilarities was performed on nano-sized ciliate communities. These clusters were subsequently organized into two groups, defined by a distance threshold of 1.7, resulting in one group containing three clusters and another with two clusters ( Figure 2 ). In nano-sized samples, both cluster analysis and NMDS revealed a strong correlation with depth ( Figures 2 , 3 ). The first group, which included Clusters I, II, and IV, was influenced by the deeper water column. Cluster IV, which comprised St. 20 and St. 21 across all depths in the water mass of TCWW and YSMW, was dominated by Choreotrichida. The taxonomic compositions in Cluster I were mostly from deeper water samples in TCWW, with only one exception in YSMW. Although Clusters II and IV shared similar taxonomic compositions according to the analysis of Bray–Curtis dissimilarity, the former was mainly influenced by TCWW and KW ( Figure 2 ), whereas the latter was mainly influenced by YSMW. The second group containing Clusters III and V was primarily composed of samples collected from shallow depths, particularly the surface layer. These shallow water samples were mostly from TCWW and positively correlated to temperature but significantly different in salinity ( Supplementary Figure S4 ; Table 1 ). Clusters III and V, both from shallow water, were higher in temperature and oxygen concentration. The biological parameters indicated the potential prey (Synechococcus, photosynthetic picoeukaryotes, and bacteria), and only Synechococcus was higher in Cluster III ( Supplementary Figure S4 ).

The top four taxa in nano-sized fraction, Leegaardiella sp., Strombidiida_G_XX_sp., Strombidiidae_G_X_sp., and Strombidiidae_H_X_sp., were widespread and across all three layers ( Figures 4 , 5 ). Lynnella semiglobulosa, Pelagostrobilidium minutum, Rimostrombidium venilia, Strombidium capitatum, Strombidium caudispina, and Strombidium triquetrum were only found in the surface layer (data not shown).

Multiple vertical distribution patterns were observed based on different annotated species. The presence of the mixotrophic family Tontonnidae, Pseudotontonia simplicidens, Pseudotontonia_sp., Spirotontonia grandis, Spirotontonia_sp., Spirotontonia turbinata, Tontonnidae_A_X_sp., and Tontonnidae_B_X_sp., decreased in the deeper layer. There was an opposite pattern with the Tontonnidae, with Tintinnidae_X_sp. and Parastrombidinopsis shimi in low abundance/absent in the surface layer ( Figure 5 ). Strobilidium caudatum and Strombidium_M_sp. were found across three layers but on the north transect ( Figure 4 ).

Research on microbial community composition and diversity using metabarcoding has been more prevalent globally but relatively scarce in the ECS (Bik et al., 2012; De Vargas et al., 2015; Gimmler et al., 2016; Malviya et al., 2016; Noan et al., 2016; Cordier et al., 2022). Metabarcoding has the potential to reveal large-scale patterns in plankton diversity and community structure that were largely invisible in morphological analyses due to large numbers of undescribed species and the difficulty in characterizing the species present in the plankton community (Massana et al., 2006; Ichinomiya et al., 2016; Morard et al., 2018; Tragin and Vaulot, 2019). Studies focusing on aloricate ciliate communities often face challenges in the resolution of morphological characteristics due to differences in fixation methodologies, which results in fewer studies based on morphology, as staining requires specific expertise (Wang et al., 2013; Sun et al., 2019; Huang et al., 2021; Kim et al., 2021). Identification of tintinnids could be easier and more complete for tintinnid communities by their external loricae compared to aloricate ones (Dolan et al., 2012; Dolan and Pierce, 2013; Zhang et al., 2017; Romano et al., 2021; Wang et al., 2021b; Li et al., 2023; Mieczan et al., 2023) However, ciliate community investigations can be more accessible without elaborate staining procedures and subjective observation under microscopy once the standardized pipeline of molecular sequencing and data filtering was followed (Bachy et al., 2013; Santoferrara et al., 2020; Burki et al., 2021). Metabarcoding provides rigor standards comparing dissimilarities among plenty of communities. Small-sized aloricate ciliates (<20 µm) were up to 50% in planktonic ciliate communities in many regions, including the ECS by morphological data (Yang et al., 2020; Wang et al., 2023a, b). Thus, this study can improve the understanding of planktonic ciliate community structure, particularly for small and naked species.

To date, many surveys by metabarcoding have shown ciliate sequences in the pico-sized fraction (Romari and Vaulot, 2004; Wu et al., 2017; Xu et al., 2017; Grattepanche et al., 2022; Flegontova et al., 2023). Canals et al. (2020) indicated that these sequences provided useful information about ciliate communities where over 94% of Ciliophora abundance from the Tara Ocean expedition in each size fraction corresponded to operational taxonomic units (OTUs) detected in the three fractions. There are hypotheses to explain these sequences in the pico-sized fraction: a) cell breakage during sequential filtration or b) membrane flexibility of ciliates to pass the small pores on filters (Cheung et al., 2008). It is possible for ciliates with a minute oral diameter (i.e., 2–5 µm) to pass through 3 µm, but molecular data are lacking (Yang et al., 2020). Comparing reads in nano- and pico-sized fraction from the same sampling sites, only Strombidiida_G_XX_sp. differed significantly in read abundance ( Supplementary Figure S5 , p < 0.05). Our results supported ciliate sequences in the pico-sized fraction from the cell breakages during filtration.

Leegaardiella sp., a cosmopolitan species reported by Agatha (2011), was dominant across three size fractions in this cruise ( Supplementary Figure S4 ). Additionally, with morphological data, Leegaardiella ovalis was identified in the same region (Yang et al., 2020). However, more sequences from the database are needed to establish a connection between the two species. In the micro-sized fraction, tintinnid contributed to half of the relative abundance, which decreased as size decreased, which was similar to previous studies in both coastal and pelagic samples ( Supplementary Figure S2B ; Grattepanche et al., 2016b; Canals et al., 2020). Tintinnid loricae are larger (50–400 μm in length) than aloricate oligotrich and choreotrich ciliates, facilitating the retention of cells on 20-μm filters. Most tintinnid ciliates were reported with morphological records except S. pacifica, Tintinnopsis platensis, and Ascampbelliella acuta (Xu et al., 2013; Li et al., 2016; Feng et al., 2018; Li et al., 2018; Sun et al., 2019; Yang et al., 2021; Wang et al., 2021b; Yu et al., 2022; Li et al., 2023). In both nano-sized and pico-sized fractions, dominant taxa were similar but different in relative abundance ( Supplementary Figure S3 ). P. minutum was absent in the micro-sized fraction, which corresponded to its morphological characteristics (Chen et al., 2017).

We focused more on ASVs, which were annotated with aloricate morphospecies ( Figure 3 ). Dadayiella ganymedes was found in offshore stations in the ECS and recorded on a large scale in the western Pacific region (Li et al., 2021; Feng et al., 2022; Li et al., 2023). Strombidinopsis acuminata, S. caudatum, P. shimi, Strombidium_M_sp., and Varistrombidium kielum were found only on the north transect across different layers, which might be related to the Changjiang River discharge-induced aloricate ciliate abundance in the summer (Chiang et al., 2003; Tsai et al., 2011; Yu et al., 2016). The mixotrophic family Tontonnidae was found vertically and horizontally in the ECS, whereas several species were frequently recorded in adjacent areas (Yang et al., 2020, 2021). However, a common species, Loboea strobila, was not detected during this cruise (Sun et al., 2019; Huang et al., 2021).

The hydrology of the ECS was intricate and influenced by factors such as the Changjiang River outflow, Taiwan Strait, and Kuroshio. In general, the clustering pattern in nano-sized communities was similar to the current circulation patterns in summer ( Figure 1B ; Yuan et al., 2008). Kuroshio Bottom Branch Current to the northeast of Taiwan (KBBCNT), which is bifurcated from Kuroshio subsurface water (KSSW), formed year-round upwelling off northeast Taiwan with its collision on the shelf (Liu et al., 1992; Wu et al., 2008; Kao et al., 2023). We observed upwelling in St. 1 and St. 1A by decreasing temperature ( Supplementary Figure S1A ; Chen et al., 2015). Then, KBBCNT flowed northwestward along the 60-m isobath, carrying high‐salinity and nutrient water and mixing with TCWW, toward the Changjiang River mouth (Yang et al., 2011). This circulation influenced Clusters I and II, in which Tintinnida and Strombidiida were dominant in both clusters ( Figure 2 ). In addition, the TCWW flowed northward in the upper layer mixed with Kuroshio surface water (KSW) and CDW (Ichikawa and Beardsley, 2002). It might reflect the distribution as Cluster III.

The Changjiang River discharge carried a large amount of terrestrial nutrients in summer, which moved northeastward, subsequently mixing with saline ambient waters (Chen et al., 2009). However, only St. 30 in Cluster V was classified under CDW ( Figure 1C ). Considering the location of St. 19 and St. 29 and the highest Chl a concentration measured at the surface in St. 19 (3.39 mg/m³), which were possibly influenced by CDW and TCWW, resulting in the lowest salinity among all clusters ( Supplementary Figure S4 ). The highest average Chl a value observed in Cluster V. Strombidiida_G_XX_sp. was dominant in Cluster V, indicating potential predation on phytoplankton (66%, Supplementary Table S3 ). Moreover, the two mixotrophic genera, Spirotontonia and Pseudotontonia, up to 12% in this cluster suggested the bottom-up environment. Cluster IV, which was from St. 20 and St. 21, was influenced by YSMW. Strombidinopsis sp. and S. caudatum were distinct and abundant species that implied different niches of YSMW, as in lower temperature and salinity (25% and 14%, Supplementary Table S3 ). However, different ASVs contributed to different clusters, which indicated the possibility of cryptic species of different niches (data not shown). The strong correlation between the ordination and depth (R² = 0.68) and particularly density (R² = 0.76), coupled with significant differences in temperature and salinity between clusters, implied that cluster dissimilarity was influenced by the currents ( Table 1 ).

The definitions of the water masses are not consistent in the ECS (Gong et al., 2000; Ichikawa and Beardsley, 2002; Yang et al., 2011; Quan et al., 2013; Zhou et al., 2018). Surface water in the ECS is mainly mixed by the CDW, KSW, and TCWW, while the deeper layer is influenced by KSSW, Shelf Mixing Water (SMW), and TCWW by the Optimum Multiparameter analysis (Zhou et al., 2018). Ciliate community composition and diversity reported differ markedly with depth (Countway et al., 2010; Grattepanche et al., 2016a, b; Santoferrara et al., 2023; Tucker et al., 2017; Zhao et al., 2017; Wang et al., 2021a), water masses (Sun et al., 2020; Yang et al., 2020; Gu et al., 2021), and geographical position (inshore/offshore) (Grattepanche et al., 2015; Santoferrara et al., 2016; Grattepanche et al., 2016b). While Yang et al. (2020) indicated that planktonic ciliates were potentially reliable indicators of water masses in the ECS, this was not universally consistent, possibly due to methodological differences in definitions of water masses or ciliate analysis approaches (Zhang et al., 2015). The NMDS plot revealed a correlation with multiple environmental variables, particularly density ( Figure 3 , Table 1 ). Additionally, the distribution of clusters was similar to current circulation patterns, suggesting that currents influence community composition. Here, we presented evidence that the composition of nano-sized ciliate communities varied with the circulation of currents.

In summary, detailed planktonic ciliate community composition was reported by metabarcoding in the ECS. The specified vertical and horizontal distributions of major small-sized taxa were first revealed. We supported hydrographic conditions shaping community structure by the similarities of cluster distribution. Building a more comprehensive database of reference sequences in future studies can provide more information for ciliate communities.