Grid sampling was conducted in the East China Sea (ECS) at a total of 30 stations in July, 2019 ( Figure 1 ). Surface water (4-6 m at 30 stations) and DCM water (15-85 m at 20 stations) were sampled. Seawater was collected using Go-Flo bottles mounted on a CTD rosette (Sea-Bird Electronics, USA). Ten liters of seawater was prefiltered using a 200 μm nylon mesh, and then processed with size-fractions (20-200 μm, 3-20 μm and 0.2-3 μm). For DNA collection, 20 μm pore size was collected on 90 mm-diameter nylon membranes (Millipore) while 3 μm- and 0.2 μm- pore sizes were collected on 142 mm-diameter polycarbonate membranes (Millipore). The filters were immediately preserved in liquid nitrogen and transferred and stored at -80°C until DNA extraction. In this study, chlorophyte composition was studied exclusively in pico-sized fractions (0.2-3 μm) due to the greater number of reads after subsampling.

Temperature and salinity were measured using a CTD profiler. The concentrations of ammonium, nitrite, nitrate, phosphate and silicate were measured according to standard methods used in a previous study (Gong et al., 2000). To measure chlorophyll a concentration, 1 L of seawater was filtered through 47-mm- diameter GF/F filters and then stored at -20°C until analysis. The chlorophyll on the filters was extracted with 90% acetone and determined with a fluorometer (Gong et al., 2000). Flow cytometry samples were preserved with paraformaldehyde (0.2% final concentration), and stored at -80°C. PPEs were quantified with chlorophyll red fluorescence using flow cytometry at an excitation wavelength of 488 nm (CytoFLEX, Beckman Coulter).

PCR conditions and methods of DNA extraction are described in Lin et al. (2021). Briefly, DNA was extracted using DNeasy PowerWater Kit (Qiagen) following the manufacturer’s instructions. The V4 region of the 18S rRNA gene was amplified with the eukaryotic universal primer set TAReuk454FWD1 (5’-[illumina adaptor]- CCAGCASCYGCGGTAATTCC-3’) and TAReukREV3 (5’- [illumina adaptor]- ACTTTCGTTCTTGATYRA-3’) (Stoeck et al., 2010). Amplicon sequencing was carried out on an Illumina MiSeq platform (pair-end 2 × 300 bp). The primers in raw paired-end reads were trimmed with Cutadapt (Martin, 2011), and the reads were analyzed using DADA2 pipeline (Callahan et al., 2016). The taxonomic annotation was carried out in accordance with PR2 version 4.12.0 database, resulting in 8 taxonomic levels: Kingdom, Supergroup, Division, Class, Order, Family, Genus and Species. DADA 2 infers exact amplicon sequence variants (ASVs) instead of operational taxonomic units (OTUs) to obtain an overall taxonomic composition for all samples ( Supplemental File 1 ).

Chlorophyte ASVs were selected from the eukaryotic ASV table produced by DADA2 under Division level of “Chlorophyta” ( Supplemental File 2 ). Subsequently, 400 chlorophyte ASVs were selected randomly and subsampled 100 times to avoid sequencing depth variations (Yeh et al., 2015). Four out of 50 samples with <400 chlorophyte reads were excluded from further analysis. We conducted multivariate and statistical analysis on the subsampled data. R package phyloseq was used to calculate the Chao1 and Shannon diversity indices. Chlorophyte composition was then clustered based on Bray-Curtis dissimilarity, used the Ward.D2 clustering method. Canonical correlation analysis (CCA) was then carried out to investigate the relationships between environmental variables and the 20 most abundant chlorophyte ASVs. To explore the relationships between niche distribution and environmental variables in each ASV, we performed a generalized additive model (GAM) to evaluate the 10 most abundant chlorophyte ASVs using a mgvc package. Prior to GAM modelling, ASV reads were natural log transformed. All statistical analyses were done with R 3.6.1.

To study the phylogenetic relationships in Chloropicophyceae and Mamiellophyceae, the 15 most abundant ASVs in each class were used to analyze the V4 region of 18S rRNA gene ( Figure S1 , S2 ). One ASV, O. tauri, was placed in Ostreococcus clade E based on phylogenetic analysis ( Figure S1 ). The taxonomic annotation of this ASV was then updated throughout the manuscript. Raw amplicon data with three size fractions have been deposited in the NCBI Sequence Read Archive under the BioProject number: PRJNA738614. Samples of size 0.2-3 μm (library_ID contains F002) from surface and DCM waters (library_ID contains -S & -D) were analyzed in this BioProject.

During the study period, temperature in the ECS increased north to south, varying from 23.6°C to 29.7°C in the surface layer and from 20.3°C to 28.3°C in the DCM layer ( Figure 1 , S3 , S4 ). The maximum temperature was observed in oligotrophic Kuroshio Water (St. K). Salinity increased from northwest to southeast, ranging from 29.8 to 34.4 at the surface and from 31.3 to 34.8 in the DCM layer ( Figure 1 , S3 , S4 ). Using a definition of water types (Gong et al., 1996), most sampling sites were influenced by the oligotrophic Taiwan Warm Current (TWC, temperature >23°C and salinity >31) ( Figure 1E ). In addition, southern offshore waters (Sts. K and 1) were influenced by Kuroshio Water (KW). A topography-induced upwelling of colder saline water caused by the westward subsurface intrusion of Kuroshio water was observed at St. 1 ( Figures 1C, D ). The northern ECS (St. 21) was influenced by eutrophic Yellow Sea Mixing Water (YSMW) with high phosphate concentrations ( Figure S3 ). The freshest water appeared in the central coastal region of the ECS (Sts. 30, 18 and 6; Figure 1 ), defined as Changjiang Diluted Water (CDW) (Gong et al., 1996), which was located further south of the Changjiang river mouth during the sampling period. The average PPE abundance was 1.1 x 10³ cells ml^-1 in the ECS with a maximum of 4,180 cells ml^-1 on the central coast at St. 6 ( Figure 1 ).

Following DADA2 pipeline, a total of 2,933,769 eukaryotic reads and 8,249 ASVs were obtained from 30 surface water samples and 20 DCM samples. After removing Opisthokonta reads, the major supergroups comprised Alveolata, Stramenopiles, Archaeplastida, Rhizaria etc. Rhizaria showed a higher amplicon contribution in the DCM than in the surface layer (12% vs 4%, Figure S5 ). We simplified PPEs into Orchrophyta, Chlorophyta, Haptophyta and Cryptophyta. Chlorophyta contributed a higher percentage of total PPE reads in DCM than in the surface layer (51% vs 40%, Figure S5 ), and were important components among pico-size PPEs. Overall, the amplicon of chlorophyta accounted for ~6% of total eukaryotic reads and 3% of total ASVs (196,467 reads and 189 ASVs prior to subsampling). The samples with low chlorophyte reads (<400 reads) were located in the northern ECS, where the temperature-salinity conditions were similar to those in Yellow Sea Mixing Water ( Figure 1E ). The following samples were excluded from further analysis: St. 20 (surface & DCM), St. 21 (DCM), St. 22 (surface).

After subsampling, Mamiellophyceae (42%) and Chloropicophyceae (41%) contributed similar read numbers in the surface water layer. Mamiellophyceae contributed slightly higher reads in the DCM layer at the class level (67% in DCM layer, Figures 2A, B ). At the genus level, Chloropicon (32%) and Chloroparvula (10%) were more important in surface water whereas Bathycoccus (28%) and Ostreococcus (22%) were more important in DCM water ( Figures 2C, D ). Micromonas reads were similar in both layers. With respect to genetic diversity, among Mamiellophyceae, Micromonas comprised 11 ASVs while Bathycoccus and Ostreococcus comprised 3 ASVs. Chloropicophyceae showed greater diversity than Mamiellophyceae on 18S V4 region, with Chloropicon and Chloroparvula comprising 16 and 22 ASVs, respectively. Apart from Mamiellophyceae and Chloropicophyceae, other chlorophyte classes included Palmophyllophyceae, prasinophyte clades V and IX, Pyramimonadales, and Trebouxiophyceae ( Figure 2 ). The most abundant chlorophyte ASVs in this study, after subsampling, included Bathycoccus prasinos (ASV_11), Chloropicon roscoffensis (ASV_8), Micromonas commoda clade A1(ASV_10) and Ostreococcus clade B (ASV_13) ( Figure 3 ).

The result of DADA2 pipeline produced a total of 189 chlorophyte ASVs, of which 156 ASVs were exclusively from the surface layer, 74 ASVs were exclusively from the DCM, and 41 ASVs were found in both layers. Prior to further analysis, subsampling was conducted to correct for variations in read numbers between samples. The Chao1 and Shannon indices were significantly higher in the surface layer than in the DCM (P < 0.01, Figure 4A ). For surface water, the maximum Shannon index occurred in KW water (St. K). Both Chao1 and Shannon were at a minimum in the central ECS, in coastal water defined as CDW (St. 6) ( Figure 4 ). Where the amplicon of Picochlorum sp. (ASV_31) among Trebouxiophyceae overwhelmingly dominated chlorophyte reads (~90%), it occurred alongside maximum chlorophyll a concentration and PPE abundance (chlorophyll a: 1.48 mg m^-3 and PPEs: 4,180 cells ml^-1, Figures 1 , S3 ; Supplemental File 3 ). Other eukaryotic reads at St. 6 were dominated by Syndiniales in dinoflagellate ( Supplemental File 1 ). Picochlorum might therefore be major contributors to chlorophyll a concentration and PPE abundance in our study.

To examine chlorophyte composition, we created a Bray-Curtis dissimilarity matrix and grouped samples based on hierarchical clustering (Ward.D2). The composition varied along the coastal-offshore gradient, and also with vertical depth in offshore samples. Based on geographic and vertical characteristics, four clusters were identified and named as follows: surface water (n=17), DCM water (n=15), coastal water (n=11) and Kuroshio Water (n=3) ( Figure 5 ).

There has been extensive discussion on the relative advantages of choosing operational taxonomic units (OTU) and ASVs from DADA2 (Xu et al., 2022). Generally, OTU is a consensus sequence which can represent different sequences from multiple species through clustering, typically using 97-99% similarity. In contrast, an ASV represents an exact sequence by running an error model to extract true sequences. The ASV method has been broadly accepted for chlorophytes due to their relatively low genetic diversity (Tragin and Vaulot, 2019). The operational taxonomic units (OTU) method is unable to discriminate the distinct Ostreococcus clades, even with a 99% identity threshold (Tragin and Vaulot, 2019). Thus, ASV is better than OTU at representing the taxonomic groups for chlorophytes. The distribution patterns of certain chlorophyte groups have already been demonstrated, for example that Mamiellophyceae is dominant in coastal water whereas Chloropicophyceae and prasinophytes clade IX are dominant in oceanic waters as well as in tropical regions (Tragin et al., 2016; Dos Santos et al., 2017a; Charvet et al., 2021). This study conducted grid sampling at surface and DCM layers on the continental shelf region to provide a high-resolution examination of chlorophyte community variation. Previous studies have not discussed the distribution of chlorophytes with respect to nutrient concentration. Our results show that the DCM layer was dominated by Mamiellophyceae, indicating that nutrient levels play an important role in determining their composition. In surface water, the dominance of both Mamiellophyceae and Chloropicophyceae might result from the mixing of coastal and Kuroshio waters on the continental shelf, with chlorophyte composition thus reflecting the hydrographic dynamics ( Figure 6 ). Some ASVs in surface water resulted from the mixing of coastal and offshore water, e.g., C. roscoffensis (ASV_8), C. sieburthii (ASV_42) and Picochlorum. On the other hand, some ASVs appeared exclusively in the transitional mixing water and occupied a narrow ecological niche, e.g., some ASVs in Dolichomastigaceae-B sp, Chloroparvula clades B2 and B3 and prasinophyte clade IX ( Figure 5 ). Collectively, chlorophyte composition reflects the hydrographic conditions and can thus be used as a model for studying the dynamics of water masses in the ECS shelf system.

Since the first Chloropicophyceae genome of Chloropicon primus has been published, the differentiation of ecological niches between Mamiellophyceae and Chloropicophyceae might be indicated by genomic features (Lemieux et al., 2019). Using genomic prediction, it has been hypothesized that C. primus might be a bacterivore since it possesses WASH complex and DNaes II genes which are associated with phagocytic capacity (Bock et al., 2021). Although C. primus was not an important species in the ECS, there is still little genomic information about other species in Chloropicophyceae. Additionally, visible vacuoles were observed in Chloropicon during ultrastructure inspection (Dos Santos et al., 2017b), but did not appear in either Bathycoccus or Ostreococcus (Limardo et al., 2017). These differences in genomic and physiological features might lead to Mamiellophyceae and Chloropicophyceae occupying different ecological niches.

M. commoda clade A1, M. bravo clade B2 and Micromonas clade B5 appeared more frequently in the ECS ( Figure 3 ), while Micromonas clade B3 and M. polaris were absent. A previous study indicated that Micromonas clade B3 and M. polaris were important in polar regions (Tragin and Vaulot, 2019). M. commoda clade A2 seems to be adapted to cold water at high latitudes (Tragin and Vaulot, 2019), but rarely appeared in the ECS, being present only in the northern Yellow Sea Mixing Water (St. 23). It was absent in a survey of the subtropical coastal water of Taiwan (Lin et al., 2017). M. commoda clade A1 was common in our study, and was detected in most of the samples. GAM prediction indicated that M. commoda clade A1 increased with temperature >25°C ( Supplemental File 4 ). It has been suggested that Micromonas clade B5 is a warm-adapted clade as it is found in the warm waters of the South China Sea and off the Philippines and Taiwan (Wu et al., 2014; Lin et al., 2017; dela Peña et al., 2021). Our results support this, and in our study Micromonas clade B5 peaked at the edge of coastal water where salinity was 32-34 and silicate concentration was ~ 2 μM. (St. 5; Supplemental File 4 ). It occurred a little more frequently in the DCM layer than in surface samples, although the total number of reads was the same in the two layers ( Figure 3 ). Overall, Micromonas clade B5 preferred nutrient-rich water.

Ostreococcus clade B was abundant in the ECS during the summer ( Figure 3 ) and was an important component of chlorophyte composition in DCM water, appearing alongside Bathycoccus prasinos ( Figure 5 ). It has been described as an oceanic ecotype, associated with low-light and deeper environments (Rodríguez et al., 2005; Demir-Hilton et al., 2011). Ostreococcus clade E, which has two bases different from clade B in 18S V4 region, was established more recently (Tragin and Vaulot, 2019). In our study, Ostreococcus clade E had low reads and peaked in more coastal waters ( Figure 5 ; Supplemental File 2 ), implying a preference for eutrophic warm waters. On the other hand, O. lucimarinus belonged to a coastal ecotype (Simmons et al., 2016), and its related reads were detected only at St. 31 ( Supplemental File 2 ). Although its sequences were found to be abundant in the Kuroshio front located off southeastern Japan (Clayton et al., 2017), it was almost absent in the ECS during the summer. Bathycoccus prasinos reads were negatively correlated with temperature in the ECS ( Supplemental File 4 ), but there were no clear ecotypes based on 18S V4 amplicon data. Mamiella gilva comprised 6 ASVs here and the dominant ASV showed a preference for coastal surface water and was absent in the DCM layer ( Figures 3 and Supplemental File 4 ). Mantoniella clade B was detected only in two offshore samples, represented by one ASV ( Supplemental File 2 ). Overall, the majority of Mamiellophyceae ASVs seem to be more important in eutrophic waters.

In marine environments, Chloropicophyceae comprises two main genera, Chloropicon and Chloroparvula, both of which were extensively represented in the ECS samples ( Figures 3 and Supplemental File 4 ). Notably, the phylogeny in the Chloropicon species indicated by using V4 region of the 18S rRNA was inconsistent with the taxonomic annotation with DADA2 ( Figure S2 ). Therefore, we have pointed out the ASV number among Chloropicon. So far, little is known about their distribution, particularly in tropical-subtropical regions. However, Dos Santos et al. (2017a) have reported that global surveys indicate Chloropicophyceae has a preference for oligotrophic and warm water. In our study, all clades in Chloropicon and Chloroparvula were detected, apart from Chloropicon maureeniae. C. roscoffensis (ASV_8) was the most abundant ASV ( Figure 3 ) having a wide distribution over the continental shelf. In the ECS, C. roscoffensis reads correlated negatively with silicate concentration ( Supplemental File 4 ). However, it was abundant in the Changjiang estuary at low nutrient levels. The next most abundant ASV among Chloropicophyceae was C. sieburthii (ASV_42), which was slightly more abundant in the DCM layer than in the surface layer ( Figure 3 ), and was the most important ASV in KW ( Figure 5 ). Different C. roscoffensis ASVs exhibited different distribution patterns. ASV_8 occurred in a wide range of salinity, whereas ASV_84 appeared only where salinity was >32. Chloroparvula clades B2 and B3 exhibited similar distribution patterns and were found mainly in the central offshore area of the ECS. Both clade B2 and clade B3 were absent in the DCM layer ( Figure 3 ), and generally peaked in low chlorophyll a water ( Supplemental File 4 ). A previous study indicated that Chloropicon and Chloroparvula were important in offshore water (Dos Santos et al., 2017a). Our results were consistent with that study as the majority of Chloropicophyceae ASVs were negatively correlated with silicate concentration and inhabited more oligotrophic surface water than other chlorophyte ASVs ( Supplemental File 4 ).

In this study, we observed that chlorophyte diversity varied with silicate concentrations in both surface and DCM layers. The results of CCA showed that salinity, temperature and silicate concentration had the greatest influence on the distribution of abundant chlorophyte ASVs ( Figure 7 ). Since chlorophytes seldom uptake silicate directly, their diversity and communities might be influenced by the diatom. The pico- and nano-sized diatoms have been found to be important in many marine systems recently (Leblanc et al., 2018; Charvet et al., 2021), including in this study ( Figure S5 ), and they might compete with the chlorophyte communities due to their similar size range. On the other hand, there may be other factors which covary with silicate concentration, e.g., turbidity. Turbidity and irradiance can significantly influence phytoplankton composition in estuarine systems since chlorophytes are photosynthetic organisms (Gameiro et al., 2011; Charvet et al., 2014). However, the sampling design of our study made it difficult to evaluate the impact of light. Nevertheless, we observed that certain chlorophyte ASVs showed a preference for specific water layers. Ostreococcus clade B, for example, was more significant in the DCM layer than at the surface based on read numbers ( Figure 3 ), and its cultures were generally isolated in deep water. A previous study demonstrated that the growth rate of different Ostreococcus strains varied with light intensity (Rodríguez et al., 2005; Six et al., 2008). Additionally, some ASVs were nearly absent in DCM water, for example, Chloroparvular clades B2 and B3, Mamiella gilva and prasinophyte clade IX ( Figure 3 ). On the other hand, an ASV in Pycnococcaceae (belonging to prasinophyte clade V, ASV_218), contributed moderate reads in both surface and DCM layers, and occurred frequently in the ECS summer, particular in KW ( Figures 3 , 5 ). According to BLAST search, this ASV is identical to a Pycnococcus provasolii sequence whose genome has been reported to encode unique dual photoreceptors of orange/far-red and blue to enable adaptation to a broad range of ocean depths (Makita et al., 2021). Overall, our data illustrated that chlorophyte communities seem to vary with terrestrial inputs of silicate concentration and salinity.

In the ECS, chlorophyte communities showed higher diversity in the surface layer than in the DCM layer. Chao1 and Shannon diversity indices were significantly correlated with nutrient concentrations, particular in surface water ( Figure 4B ). Temperature and salinity also play important roles in surface water ( Figure 4B ), with the Chao1 index being positively correlated with both temperature (r = 0.66, P < 0.001) and salinity (r= 0.65, P < 0.001) ( Figure 4B ). In the DCM layer, the diversity indices were related to few environmental factors, suggesting there might be other factors determining chlorophyte diversity, such as predators and irradiance levels (Charvet et al., 2014; Yang et al., 2018). Our data shows that the diversity indices were negatively correlated with silicate concentrations, indicating that terrestrial input to the ECS plays an important role in shaping chlorophyte diversity. A previous study indicated that sudden pulses of high nutrients can increase the growth of a single species, and thus decrease community diversity (Spatharis et al., 2007). In our study, the minimum chlorophyte diversity was observed in the coastal water (St. 6) of the central ECS dominated by Picochlorum (90% of total chlorophyte reads). We also observed high PPE abundance and chlorophyll a concentration. With global warming, Picochlorum might lead to harmful algal blooms since GAM indicates its reads correlate positively with temperature, ammonium and silicate concentrations ( Supplemental File 4 ). Foflonker et al. (2015) have indicated that Picochlorum can acclimate to a wide range of salinity and that excessive ammonium enhances their growth, and that these ecological features can be explained by genomic data. A recent study indicated that Picochlorum can be grazers on picoplankton (Pang et al., 2022). More than that, some taxa among chlorophytes have been reported as potential mixotrophs, e.g., Chloropicon, Dolichomastix, Micromonas, Nephroselmis, Pterosperma and Pyramimonas (McKie-Krisberg and Sanders, 2014; Bock et al., 2021). These pico-size chlorophytes are also predators on picoplankton, which might result in an increase in the efficiency of energy flows (Ward and Follows, 2016). On the other hand, being an important prey for protists and zooplankton (Orsi et al., 2018; Ma et al., 2021), their composition will influence the community diversity of predators (Yang et al., 2018). Therefore, as primary producers and consumers, chlorophyte community structure plays a key role in determining the efficiency of energy and nutrient flows in microbial food webs.