During a research cruise in the period of 6–12 March 2018, the station B6 (117.59°N, 21.89°E) was investigated. The study station (B6) was well stratified and located to the south of Taiwan Strait in the South China Sea (Figure 1). The region is influenced by the Zhejiang-Fujian Coastal Current, the South China Sea Warm Current and the intruding Kuroshio. At the station, CTD SBE911 with the rosette bottle samplers was deployed to read the vertical profile of environmental data including salinity, temperature, pH, concentrations of dissolved oxygen (DO), and in vivo fluorescence chlorophyll. The depth of the DCM layer was determined and sampled according to the vertical distribution of the fluorescence (Figure 2A). We collected water samples from the surface (1 m) and DCM layer (50 m) water and conducted incubation experiments on board of the ship.

The water samples of the surface layer (1 m) and the DCM layer (50 m) at station B6 were pre-filtered through 200 μm nylon mesh to remove large zooplankton and debris, then dispensed into 45 acid-cleaned 2 L plastic bottles (totally 5 groups, 9 bottles per group). The experiment was designed in 5 treatments (groups) shown as Figure 3. For surface water, treatment (1) is Control: water sample + natural light; (2) water sample + nutrients + natural light. For the DCM layer water, treatment (3) is Control: water sample + 10% light; (4) DCM layer water + natural light; (5) DCM layer water + nutrients + natural light.

The added nutrients were from F/2 medium nutrient stocks and the final concentrations of NO₃⁻, PO 4   3 − and SiO 4   2 − were ~40 uM, ~2 uM and ~10 uM, respectively. All bottles were placed in water tanks with surface seawater flowing to ensure that the incubation temperature was consistent with the surface water temperature of the ocean with a light-dark cycle of 12 h: 12 h. In the group (3), the tank were covered with several layer of neutral density screens in order to reduce the ambient light by 90%, simulating the weak light of DCM layer. Experiments started at 7:00 in the morning, and ended at 3 days later. The samples during the incubation were taken at T=0, 24, 48, and 72 hours. During each sampling time, we consume three bottles (replicates) of each groups for analysis of nutrients, counts of cell abundance, analysis of DNA for microbial community, and continue to incubate the rest bottles until the end (Figure 3). Concentrations of ammonium (NH₄⁺), nitrite (NO₂⁻), and nitrate (NO₃⁻), dissolved active phosphate (PO₄³⁻), silicate ( SiO 4   2 − ) were determined with an autoanalyzer (Seal, Germany).

Samples for cytometric analyses were pre-filtered through 200 μm meshes, and fixed with formaldehyde (2% final concentration) in 2 ml tubes, and quickly frozen in liquid nitrogen until flow cytometry (FCM) analysis in the laboratory. Abundances of bacteria (Cyanobacteria, heterotrophic bacteria) and small (pico and nano-plankton) photosynthetic eukaryotes were determined on a BD FACSCanto II flow cytometer (BD Biosciences, San Jose, CA, USA) equipped with 488 nm laser excitation and standard filter setup as previously described in (Marie et al., 2000). First, we differentiated PPE (photosynthetic pico-eukaryotes) and PNE (photosynthetic nano-eukaryotes) based on orange fluorescence (FL2) versus red fluorescence (FL3) and side scatter versus FL3 signatures. Cyanobacteria (Prochlorococcus and Synechococcus) cell counts were determined from unstained samples, based on natural fluorescence from phycoerythrin (orange) and chlorophyll (red; Casotti et al., 2003). Heterotrophic bacteria were enumerated as in (Balestra et al., 2011) after staining with SYBR Green I (Lonza, final concentration 10⁻³ of stock solution) for 15 min in the dark at room temperature prior to analysis. Heterotrophic bacteria were identified in plots of red fluorescence versus green fluorescence and side scatter versus green fluorescence. The 1 μm fluorescent polystyrene calibration beads (Polysciences Co., USA) were added to all samples as an internal standard. Data were collected, saved, and analyzed with the Cell QUEST software.

Molecular analyses of microbial community (bacteria and eukaryotes) were performed for cultivation seawater samples (three replicates pooled together for one sample). Samples of 2 L were filtered through a 0.2 μm pore-sized membrane (Millipore, Billerica, MA, USA) under low light and vacuum pressure (< 100 mm Hg). All filter membranes were frozen in liquid nitrogen onboard until laboratory analysis. The DNA was extracted using a FastDNA Spin Kit (MP Biomedical, USA) according to the manufacturer’s instructions. The concentrations of purified DNA were measured using a spectrophotometer Nanodrop 2000c (ThermoFisher, USA).

The hypervariable regions V4 of 16S and V4 of 18S rRNA genes, respectively, were amplified for further sequencing. Amplification of V4 region of 16S rRNA genes was performed using 515f and 806r primer pairs (Walters et al., 2016) and V4 region of 18S rRNA genes using the 528F and 706R primer pairs (Elwood et al., 1985). The 20 μL reaction solution contained 1 unit Taq polymerase (Life Technologies, Carlsbad, CA, USA), 1× reaction TransGen buffer, 0.2 mM dNTPs, 1.5 mM MgCl₂ and 0.2 μM of each primer. The PCR program was as follows: an initial denaturation step at 95°C for 5 min; 30 cycles of 94°C for 30 s, 47°C for 45 s and 72°C for 1 min; and a final extension at 72°C for 5 min. The PCR products from triplicate reactions for each sample were pooled together and sequenced on the MiSeq System (Illumina, San Diego, CA, USA) in a commercial company (Novogene, China).

Data were processed using the Quantitative Insights into Microbial Ecology (Qiime) pipeline for 16S rRNA and 18S rRNA data sets (http://qiime.org/tutorials/index.html; Caporaso et al., 2010). In brief, the sequences with a quality score below 25 and 200 bp in length were trimmed and then assigned to water samples based on unique 10 bp barcodes. For the 16S rRNA gene, after a first assignment against Silva v119 (Quast et al., 2013), chloroplast and mitochondria sequences were removed. In the case of 18S rRNA gene, Metazoa sequences were removed after a first assignment against the PR² database (Guillou et al., 2013). OTUs at 97% similarity were generated using default settings and clustering was performed with furthest neighbor algorithm. OTUs formed by 10 or less sequences in the case of 16S rRNA gene and 5 or less sequences in the case of 18S rRNA gene, were removed from further analysis. Singletons (OTUs containing a single read across all samples) were excluded prior to further analysis. Rarefaction curves and diversity indexes were obtained with Mothur.

For calculating alpha diversity estimators (i.e., OTU richness, Shannon, Simpson, and Chao1) of bacteria and eukaryotes, the reads were rarefied by randomly re-sampling 50000 reads per sample for 10 times, which minimized bias associated with sequencing depths and allowed for comparison of diversities for all samples. The bacterial and eukaryotic variations during the incubations based on Bray-Curtis similarities were visualized using NMDS ordination implemented within PRIMER package V5.0 (Primer-E, UK).

The vertical profiles of physical and chemical variables at station B6 were shown in Figure 2 and Supplementary Table S1. The water column was highly stratified with the strong thermocline between 10 m and 50 m (Figure 2). From the surface to the depth of 180 m, temperature dropped from 29.05°C to 15.81°C (Figure 2C), and salinity increased slightly from 34.02 PSU to 34.56 PSU (Figure 2B). The DCM layer occurred around 50 m, and chlorophyll was 0.13 μg L⁻¹, being much higher than the surface layer (0.06 μg L⁻¹, Figure 2A). In the DCM layer, the concentrations of DIN, PO₄³⁻ and SiO 4   2 − were 6.73, 0.62 and 4.90 μM (Figure 2).

On-deck incubation experiments showed the different responses of surface and DCM microbial communities to light and nutrients. For the surface water incubation experiments, in the control, initial NO³⁻ and PO₄³⁻ were low and remained near depletion levels during the incubation, but SiO₄²⁻ decreased to 3.80 ± 1.11 µM in 24 h, and increased to 6.85 ± 2.20 µM at the end of the experiment (Supplementary Figure S1A). In contrast, in the nutrient addition treatment, ambient NO³⁻ and PO₄³⁻ were drawn down from 41.4 ± 3.15 to 30.67 ± 3.20 µM and from 1.67 ± 0.20 to 0.98 ± 0.24 µM, respectively, while initial SiO 4 2 − remained relatively stable (Supplementary Figure S1C).

For the DCM water incubation experiments, in the control (10% light addition group), ambient NO³⁻ and PO 4   3 − decreased from 5.30 ± 1.06 to 3.50 ± 0.68 µM, and from 0.63 ± 0.15 to 0.51 ± 0.14 µM, as SiO 4   2 − increased little during the incubation (Supplementary Figure S1E). More rapid uptake of ambient nutrients (NO³⁻, SiO₄²⁻ and PO₄³⁻) took place as the nutrients decreased dramatically after 72 h incubation in the nutrients + light addition and the light addition groups (Supplementary Figures S1B, D).

The cell abundance of different microbial taxa changed dramatically in different incubation treatments (Figure 4). For the surface water incubation experiments, in the control (the light addition group), the density of bacterial cells was decreasing as fluorescence PPE and PNE increased little (Figures 4A, B). In contrast, in nutrients + light addition group, the microbial cells increased rapidly after 72 h incubation, especially for the heterotrophic bacteria and PPE (from 0.65 ± 0.16 to 1.72 ± 0.3 × 10⁴ cells ml⁻¹, and from 0.69 ± 0.07 to 3.40 ± 0.67 × 10³ cells ml⁻¹, respectively, Figures 4C, D).

For the DCM water incubation experiments, in the control (the 10% light addition group), heterotrophic bacteria and Prochlorococcus maintained a higher cell density (1.08 ± 0.25 × 10⁴ cells ml⁻¹, 0.71 ± 0.08 × 10⁴ cells ml⁻¹) after 72h, while other microorganisms maintained a relatively low cell concentration (Figures 4I, J). Compared with the control, microbial cells in the nutrients + light addition group increased sharply, especially for PNE. The abundance of PNE increased to 4.61 ± 0.31 × 10³ cells ml⁻¹ after 72 h incubation (Figure 4H). In the light addition group, similar phenomenon was observed (from 0.04 ± 0.01 to 2.31 ± 0.12 × 10³ cells ml⁻¹, Figure 4D).

Overall, 16S rRNA (Bacteria) and 18S rRNA (eukaryotes) genes sequencing generated 1,685,903 and 1,730,235 sequences, respectively. During the 3 days incubation, the microbiome changed rapidly (Figure 5). Ecological indices showed that, in general, bacterial and eukaryotic communities are richer and more diverse in the initial than the end of the incubation experiments (Supplementary Tables S4, S5). The total number of OTUs, defined at a 97% similarity, ranged from 708 to 1,667 for bacteria and ranged from 430 to 1,463 for eukaryotes (Supplementary Tables S4, S5). In case of bacterial communities, Chao1 index was 921–1,532, Shannon index was 4.24–9.14 (Supplementary Table S4). For the surrounding eukaryote community, Chao1 index was 524–1,622 and Shannon index was 1.69–7.98 (Supplementary Table S5).

A total of 25 bacterial phyla were detected in the incubation experiments. The relative abundance varied greatly, ranging from 15.15 to 57.59% for Gammaproteobacteria, 10.15%–48.00% for Alphaproteobacteria, 1.18%–47.86% for Cyanobacteriia, 0.01%–24.45% for Thermoplasmata. Bacteroidia were less abundant (0.42%–7.10%), Acidimicrobiia (0.30%–6.43%), Bacilli (0.03–5.16%), Actinobacteria (0.02–2.51%) and others were relatively minor components of the communities (Figure 6A). In case of eukaryotes (Figure 6B), overall, the community of eukaryotes was dominated by Diatomea (4.89–95.85%) with Chlorophyta being 0.27–64.45%, Dinophyceae 0.30–34.36%, Protalveolata 0.29–31.56%, MASTs 0.31–11.37%.

Comparisons of the relative abundance of major microbial lineages in the incubation experiments provided information on how the community structurally shifted over a temporal scale of the incubation (Figure 6). For surface water incubations, in nutrients + light addition group, in the case of bacteria community, after 72 h incubation, the read proportions of Bacteroidia (5.79% vs. 1.81%), Alphaproteobacteria (31.49% vs. 20.15%) and Acidimicrobiia (6.43% vs. 3.94%) were significantly higher than the initial. However, the relative abundance of Cyanobacteriia slightly decreased (30.69% vs. 21.92%, Figure 6A; Supplementary Table S6). As for the eukaryotic community, in the nutrients + light addition group, the read proportions of Mamiellophyceae accounted for approximately 50.27%, which was about seven times higher than that in the initial (Supplementary Table S7). On the contrary, the proportions of Protalveolata and Dinophyceae were significantly decreased during the incubation experiments (Figure 6B; Supplementary Table S7).

For the DCM water incubation experiments, in case of bacteria, after 48h incubation, Cyanobacteria increased from 8.19% to 47.86% in 10% light addition group, from 8.19% to 44.08% in light addition group, from 8.19% to 34.93% in nutrients + light addition group, respectively, and the proportion of Thermoplasmata decreased dramatically (Figure 6A; Supplementary Table S6). In case of eukaryote community, diatoms (mostly the Mediophyceae and Bacillariophyceae) was significantly more abundant in the three groups, after 72h incubation, the relative abundance of diatom taxa accounted for the 95.85%, 51.97%, and 91.41% in the 10% light addition, light addition and nutrients + light addition groups, respectively (Figure 6B; Supplementary Table S7). More interestingly, the proportions of Dinophyceae and Protalveolata were significantly lower in the three groups after incubation (Figure 6B; Supplementary Table S7).

At station B6, microbial abundance and composition were significantly different in different water layers where there were gradients of vertical distribution of sunlight and nutrients in the water column. This agreed with the other observations (Liu et al., 2012; Casey et al., 2013; Giner et al., 2020). In our FCM data, the heterotrophic bacterial cells dominated the microbial cell assemblages in the surface and DCM waters. This is a common phenomenon that the heterotrophic bacteria dominate in oligotrophic seas in comparison to the phytoplankton groups (Cotner et al., 1992). In the upper layer of the oligotrophic sea, the growth of the heterotrophic bacteria is mainly supported by dissolved organic compounds produced by the photosynthetic taxa, and the remineralization of inorganic nutrients by the heterotrophic bacteria stimulates the growth of the photosynthetic groups (Linacre et al., 2015). The abundance of the Retaria groups increased in relative abundance with depth (Supplementary Table S7), as reported in earlier studies (e.g., Edgcomb et al., 2002; Giner et al., 2020). Previous studies found these groups likely have many rDNA operon copies (Massana, 2011), which might give an overestimate their abundance.

Compared to DCM water incubations, surface water incubations showed that the relative abundance of photosynthetic pico-eukaryotes (PPE) increased significantly. For instance, in the nutrients + light addition group, the read proportions of Mamiellophyceae (this taxa is typically PPE) accounted for approximately 50.27%, which was about seven times higher than that in the initial (7.59%; Figure 6B; Supplementary Table S7). The PPE cell abundance response patterns are coherent with those of eukaryote community composition (Figure 4F), which suggests that the Mamiellophyceae were the main responders to the surface water incubations after adding the nutrients. This typical pico-algae are globally distributed in aquatic environments, and can also increase under specific conditions in oceanic waters (Treusch et al., 2012) or Polar Regions (Li et al., 2009). The increasing in the surface water incubations of these taxa can be attributed to their enormous ecotypes, which may display specific photoacclimation strategies (Hanschen and Starkenburg, 2020).

At end of the DCM water incubations, cells of nano-sized phytoplankton became very high, indicating the occurrence of a bottled phytoplankton bloom (Figures 4D, H). It is interesting that the relative abundance of diatoms (mainly the Mediophyceae and Bacillariophyceae) was particularly high and the concentration of SiO 4 2 − decreased dramatically in the DCM water incubation experiments after 72 h (Figure 6B; Supplementary Figures S1B, D). Considering that diatoms are generally close or above to the nano size, we extrapolate the cells of diatoms were dominant in the incubation bottles. This result is not surprising, given that diatoms are often observed to be dominant during phytoplankton blooms after the typhoon events or other physical mixing, as shown in Table 1. In the SCS, Li et al. (2009) reported the dominance of diatoms in the “Kaemi” typhoon induced blooms in 2016. In the coast of Western Gulf of Mexico, diatoms cell abundance were observed to be dominant after the four super tropical cyclones (Anglès et al., 2015). According these, phytoplankton blooms likely occur after a physical mixing event such as typhoon or IW, due to seed phytoplankton and nutrients in the DCM layer being brought to the ocean surface to receive sufficient sunlight. Furthermore, previous researchers found that the surface layer has a higher zooplankton grazing rate than the DCM layer (Chen et al., 2009a), given that zooplankton may prefer to consume smaller phytoplankton (such as Mamiellophyceae) than bigger preys (such as larger diatoms; Gaul et al., 1999). The typical heterotrophic taxa such as Ciliophora and MASTs were found to be more abundant in the surface layer than in DCM in our study (Supplementary Table S7), which may imply stronger grazing rate at the surface than DCM layer during the incubations.

In our observations, Cyanobacteria (Prochlorococcus and Synechococcus) declined or increased slightly in cell abundance for all treatments (Figure 4). These results are not surprising, as in other studies abrupt decrease in Cyanobacteria abundance was observed both in mesocosm experiments and field observations (Cunliffe et al., 2009). Considering Cyanobacteria having a higher cell surface area/volume ratio, abnormally low nutritional requirements, cyanobacteria are better adapted to oligotrophic stratified conditions (Partensky et al., 1999). Moreover, the presence of small grazers (e.g., ciliates and MASTs) was suggested by past studies to be responsible for the decrease of this small photosynthetic group (Mackey et al., 2009).

Physical mixing events are widely recognized to control microbial community in aquatic systems (Johnson et al., 2010). The different community and their succession of microorganisms in surface and subsurface water after receiving nutrients and light indicates that the microorganisms in different water layers have distinct responses to the vertical mixing caused by physical events. According to the “dilution-recoupling” hypothesis, physical mixing also play a critical role in the initiation of the blooms, as mixing induced dilution of both phytoplankton and grazer biomass leads to fewer grazer-phytoplankton encounters (Behrenfeld, 2010; Johnson et al., 2010). Considering the carbon sequestration efficiency of the biological pump (BP) is shaped by the microbial community (Jiao et al., 2010). Those physical mixing events are important for the biological pump (BP) to perform carbon sequestration in the water column of oligotrophic oceans. This warrants more dedicated investigation.