SML (0–1 mm depth) and SSW (2.5–5.0 m depth) waters were collected from 15 sites of the ECS (D3, D5, F1, F3, F5, P2, P3, P4, W1, W3, T1) and the YS (H8, H9, H11, H3) aboard the R/V Dongfang Hong 2 in June 2018 (Figure 1; Supplementary Table 1). SML water samples were collected by using the Garrett metal screen (MS; Agogué et al., 2004; Chen et al., 2016), whereas SSW samples were collected with a Sealogger CTD (SBE25, Electronic Inc., United States) rosette water sampler (~4 m below the surface). Samples containing 1 L seawater were filtered serially through 3 and 0.22 μm polycarbonate membranes (Millipore Corporation, Billerica, MA, United States), respectively. The particle-associated (PA) bacteria in the seawater were collected through 3 μm polycarbonate membranes, and the free-living (FL) bacteria in the seawater were collected through 0.22 μm polycarbonate membranes. For the quantification of Synechococcus (SYN), Prochlorococcus, picoeukaryotes (PEUK) and heterotrophic bacteria, 2 ml of water samples from each sample were placed into sterile tubes and immediately fixed with paraformaldehyde (final concentration 4%, v/v) for 30 min in the dark at room temperature. Liquid nitrogen was immediately used to freeze both membranes and 2 ml water samples, which were then stored at −20°C on board and transferred to −80°C in the laboratory.

Hydrological parameters (temperature, salinity and depth) were obtained by CTD equipped on the water sampler in situ. The concentrations of Chlorophyll a (Chl a) were measured as described previously (Zhang et al., 2014). Using GF/F filter (Whatman) with a pore size of 0.7 μm immediately filtered seawater samples after collection on board, which were then soaked in 90% (v/v) acetone in the dark for 24 h to extract Chl a. The F4500 fluorescence spectrophotometer was used to determine the concentration of Chl a in the extracts (Hitachi, Japan). DO was measured by Winkler method (Carpenter, 1965). Samples for nutrients (PO₄³⁻, NO₂⁻, NO₃⁻, SiO₃²⁻, and NH₄⁺) were filtered with 0.45 μm cellulose acetate membranes and were analyzed by an Auto-Analyzer (AA3, Seal Analytical Ltd., United Kingdom; Liu et al., 2015). The abundances of Synechococcus, Prochlorococcus, picoeukaryotes, and heterotrophic bacteria were measured by flow cytometer (BD FACSJazz, United States) in the laboratory (Zhao et al., 2017).

DMS and DMSP concentrations in seawater samples were measured in situ as described previously (Yang et al., 2011). DMS in seawater samples was captured by a cryogenic purge-and-trap pretreatment system and the concentration of DMS was measured by flame photometric detection with gas chromatography (Agilent GC-7890B; Tan et al., 2017). To avoid the influence of microorganism cell rupture caused by filtration pressure on DMSP concentrations, DMSP was captured by gravity filtration method (Tan et al., 2017). Total DMSP (DMSP_t) refers to the DMSP without filtration; Particulate DMSP (DMSP_p) refers to the DMSP that captured on the 0.45 μm filter membrane; Dissolved DMSP (DMSP_d) refers to the DMSP in the filtrate through the 0.45 μm filtration membrane.

Total DNA were extracted from 3 and 0.22 μm membranes of the SML and SSW samples using the Phenol-chloroform method (Sun et al., 2020). The extracted DNA was redissolved in 10 mM Tris–HCl (pH 8.0) and stored at −80°C. Subsequently, it was used for bacterial 16S rRNA high-throughput sequencing and qPCR to quantify functional genes.

The abundances of total bacterial 16S rRNA genes, DMSP producing genes dsyB, mmtN, catabolic genes dmdA (C/2, D/1 subclade) and dddP in seawater samples were quantified by qPCR. All primer sequences and annealing temperatures were listed in Table 1. PCR reactions and melting curves were performed as described by Sun et al. (2020). qPCR standard curve was made using pUCm-T vector (Biotech, China) that contained a single copy of the corresponding gene. Plasmids were extracted with Mini Plasmid Kit (Takara, Tokyo, Japan), linearized with restriction endonuclease XhoI, purified with TIANgel Mini Purification Kit (TIANGEN Biotech, Beijing) and quantified with Nanodrop-1,000 spectrophotometer. A standard curve was then generated using 10-fold serially diluted linearized plasmids, all of which showed linear correlations of R2 = 0.99. The qPCR gene amplification efficiency ranged from 95 to 105% (93% to 98% for dsyB and mmtN). Three techniques replicates were set for each sample. Double-distilled water was used as a template for negative control. The qPCR of all samples was performed on the StepOne ™ Real-time PCR System (Applied Biosystems), and the obtained data were analyzed by the StepOne software (version 2.2). The abundance of each gene in the sample was calculated according to the copy number obtained by qPCR and the volume of filtered water sample.

The total bacterial 16S rRNA gene was amplified by Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China) using primers 515modF and 806modR (Walters et al., 2016). The PCR amplification system (20 μl) was conducted as follows: 4 μL of 5 × Fast Pfu buffer, 2 μL of 2.5 mM dNTPs, 0.8 μL of 5 μM forward and reverse primers, 1 U of TransStart Fastpfu DNA polymerase, 10 ng of template DNA, 0.2 μL of BSA (bovine serum albumin), and add double-distilled water to 20 μL. The PCR cycling condition was conducted as follows: a. pre-denaturation at 95°C for 3 min, b. 29 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 30 s, extension at 72°C for 45 s, c. extension at 72°C for 10 min, and 10°C until halted. The PCR amplification product was purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, United States), and the DNA was quantified using QuantiFluor™-ST (Promega, United States). The purified amplicons were merged in equimolar and paired-end sequenced on the Illumina MiSeq platform (Illumina, San Diego, United States) according to the standard protocols of Majorbio Bio-Pharm Technology Co. Ltd. (Shanghai, China). After subsampling each sample to an equal sequencing depth according to the minimum number (65,552) of sample sequences, OTUs were clustered using Usearch7.0 method of the QIIME1.9.1 with 97% similarity cutoff. The taxonomic position of each OTU representative 16S rRNA gene sequence was analyzed by Silva 128 16S rRNA database¹ using confidence threshold of 70%. The absolute abundance of potential DMSP biosynthetic and catabolic genera were estimated by their relative abundance determined by 16S rRNA gene amplicon sequencing and the total bacteria abundance quantified by qPCR analysis of 16S rRNA gene. This is a semi-quantitative approach as it is solely based on the presence of these genes in isolates/genomes belonging to similar genera reported in previous publications.

Mothur² was used to calculate the Alpha diversity indices such as Shannon, Chao1 and Good’s coverage to measure the species richness and diversity of the community (Sun et al., 2016; Bullock et al., 2017). For beta diversity, non-metric multidimensional scaling analysis (NMDS) and hierarchical clustering trees were performed with ANOSIM based on Bray-Curtis distance matrices using the “vegan” package in R software (version 4.1.1). The difference of bacterial community structure between the SML and SSW was analyzed by Wilcoxon rank-sum test. The differences in bacterial diversity and richness between SML and SSW were analyzed by Wilcoxon signed-rank test. The relationship between environmental factors and bacterial community structure was evaluated by distance-based redundancy analysis (db-RDA) with 999 Monte Carlo permutation tests using the Canoco software (version 5.0, Microcomputer Power). The correlations between environmental factors and functional gene abundance were conducted using Spearman correlation test. The difference of environmental factors and functional gene abundance between SML and SSW was conducted by Wilcoxon signed-rank test. Differences in environmental factors and gene abundance between seasons and between regions were conducted using the Mann–Whitney tests. All statistical analyses were performed on SPSS version 25.0 (SPSS, Chicago, IL, United States) and the significance threshold for all tests was set with p < 0.05 and p < 0.01. The map of sampling sites was created using Ocean Data View (ODV, v5.1.7) and figures were drawn by Origin 2021 software³ or GraphPad Prism 6.01.

Raw reads from the summer 16S rRNA gene amplicon sequencing have been deposited in the NCBI BioProject database under the accession number PRJNA648032. Raw reads from the spring 16S rRNA gene amplicon sequencing were deposited into the NCBI Sequence Read Archive (SRA) database with accession number SRP174872 under the BioProject PRJNA511511 (Sun et al., 2020).

The environmental parameters of all samples were listed in Supplementary Table 1. In the eastern China marginal seas (ECS and YS), DOC was significantly higher in the SML than in the SSW (p < 0.05, Wilcoxon signed-rank tests, Supplementary Tables 2–4). DMSP_d and DMSP_t concentrations were significantly higher in the SML (162.66 ± 324.28 nM and 403.09 ± 647.46 nM) than in the SSW samples (7.51 ± 3.94 nM and 91.49 ± 55.22 nM; ~21.67-fold and 4.41-fold, respectively, p = 0.001 and 0.019, Wilcoxon signed-rank tests, Supplementary Table 3). DMS, DMSP_p and Chl a concentrations showed no significant difference between the SML and SSW samples (Supplementary Tables 2–4).

Regionally, DOC in the ECS (130.34 ± 30.65 μmol C/L) was lower than that of the YS (195.75 ± 26.97 μmol C/L), while Chl a of the ECS (2.01 ± 1.21 μg/L) was higher than that of the YS (0.50 ± 0.18 μg/L) in the SML (Supplementary Table 1). The DMS and DMSP concentrations reached the maximum at the SML site F3 of the ESC among all samples (26.67 nM and 2123.95 nM, respectively, Supplementary Figure 1). DMS and DMSP_t concentrations of the SML in the ECS were higher (~1.61-fold and ~2.47-fold) than in the YS (Supplementary Figure 1; Supplementary Table 1), while no difference was observed for DMS and DMSP_t concentrations in the SSW of the ECS and YS (p > 0.05, Supplementary Figure 1; Supplementary Table 5).

In both spring and summer, DMSP_d and DMSP_t concentrations were higher in SML than in SSW, and DMS and Chl a were not significantly different between these two water layers (Supplementary Figure 1; Supplementary Tables 1, 3, 4; Sun et al., 2020). Additionally, the concentrations of DMS, DMSP_d, DMSP_p and DMSP_t in the summer SML samples were higher than those in spring (Supplementary Table 5). Chl a in the SML of the ECS in summer samples was also higher (~2.00-fold) than that in spring (Sun et al., 2020).

The total abundance of the bacterial 16S rRNA gene (the sum of FL and PA bacteria, Supplementary Figure 2B) quantified by qPCR was consistent with the changes of heterotrophic bacteria number (Supplementary Figure 2C) shown by flow cytometer in SML and SSW of the eastern China marginal seas. The average counting of heterotrophic bacteria in the SML (1.72 ± 3.02 × 10⁹ cells L⁻¹) was higher than in the SSW (7.57 ± 6.30 × 10⁸ cells L⁻¹). Total abundance of the bacterial 16S rRNA gene in SML (2.91 ± 3.28 × 10⁹ copies L⁻¹) was also significantly higher than in SSW (6.36 ± 4.95 × 10⁸ copies L⁻¹，p < 0.01, Wilcoxon signed-rank tests, Supplementary Figure 2B; Supplementary Table 2), and was more abundant in FL fraction than in the PA fraction (~8.28-fold for SML and ~ 7.27-fold for SSW, respectively, Supplementary Figure 2A).

Regionally, the average counting of heterotrophic bacteria in the ECS SML (2.10 ± 3.53 × 10⁹ cells L⁻¹) was higher than those of the YS SML (7.56 ± 5.14 × 10⁸ cells L⁻¹), so was that for the SSW samples (9.03 ± 7.16 × 10⁸ cells L⁻¹ in the ECS and 3.58 ± 1.83× 10⁸ cells L⁻¹ in the YS). Total abundance of the bacterial 16S rRNA gene in the ECS (3.30 ± 3.71 × 10⁹ copies L⁻¹ in the SML and 7.90 ± 4.93 × 10⁸ copies L⁻¹ in the SSW) was also higher than those of the YS SML (1.81 ± 1.52× 10⁹ copies L⁻¹ in the SML and 2.11 ± 0.88 × 10⁸ copies L⁻¹ in the SSW; Supplementary Figure 2B; Supplementary Tables 3, 4). The abundance of both FL and PA bacteria was significantly higher in the SML of the ECS than in the SSW (p < 0.05, Wilcoxon signed-rank tests, Supplementary Table 3), but there was no significant difference between the two water layers in the YS (p > 0.05, Wilcoxon signed-rank tests, Supplementary Table 4). Seasonally, the total abundance of bacterial 16S rRNA gene in summer SML and SSW samples was ~3.00-fold and ~ 5.42-fold higher than those in spring (1.10 ± 0.57 × 10⁹ copies L⁻¹ for the SML and 1.46 ± 0.91 × 10⁸ copies L⁻¹ for the SSW, Sun et al., 2020), respectively.

For eukaryotes in the eastern China marginal seas, picoeukaryotes and Synechococcus were more abundant in the SSW (7.74 ± 17.65 × 10⁵ and 1.31 ± 1.84 × 10⁷ cells L⁻¹) than in the SML (1.17 ± 2.55 × 10⁵ and 1.06 ± 1.46 × 10⁷ cells L⁻¹, Supplementary Table 1). Synechococcus (SYN) in the ECS SML (1.71 ± 1.89 × 10⁷ cells L⁻¹) was more abundant than that in the YS SML (1.72 ± 1.03 × 10⁶ cells L⁻¹; p < 0.05, Mann–Whitney tests, Supplementary Table 4).

In total, 2,871,910 reads were obtained with an average sequence length of 273 bp via the 16S rRNA gene amplicon sequencing. After quality control and subsampling, a total of 3,219 OTUs were assigned at the 97% sequence similarity threshold level. The good’s coverage values (99.58%–99.76%, Supplementary Table 6) indicated that the sequencing results can cover most of the bacterial community in the samples. The NMDS analysis and hierarchical clustering trees based on Bray-Curtis distances showed a clear separation of communities by sampling site. Basically, all the samples were partitioned into four geographic clusters (stress = 0.141), i.e., ECS_SML, ECS_SSW, YS_SML, and YS_SSW (Figure 2). The Shannon and Chao 1 indices were used as indicators of the bacterial community diversity and richness in SML and SSW samples, respectively (Supplementary Table 6). Differences of the bacterial community of SML and SSW in the whole eastern China marginal seas, and differences of the bacterial community of SML and SSW between ECS and YS and between summer and spring were analyzed (Supplementary Figures 3–6). In general, the Shannon diversity index was significantly higher in the SSW than in the SML (p < 0.05, Wilcoxon signed-rank test, Supplementary Figure 4A), while the Chao 1 index was not significant different between the SML and SSW samples (p > 0.05, Wilcoxon signed-rank test, Supplementary Figure 4B). Regionally, the Chao 1 and Shannon diversity indices in the ECS SML were significantly higher than that of YS SML (p < 0.05, Wilcoxon signed-rank test, Supplementary Figures 5A,B). The bacterial diversity in the YS SSW was higher than that of SML (p < 0.01, Wilcoxon signed-rank test, Supplementary Figure 5A). Both the Chao 1 and Shannon diversity indices showed no significant differences between SML and SSW samples of the ECS in summer, which were significantly higher than those in spring (p < 0.001, Wilcoxon signed-rank test, Supplementary Figures 5C,D).

Microbial community of SML and SSW samples from the eastern China marginal seas was analyzed to identify potential DMSP producers and consumers via 16S rRNA gene amplicon sequencing. From the eastern Chinese marginal sea, it possesses more Gammaproteobacteria in the SML, while Cyanobacteria, Bacteroidetes Incertae Sedis, Thermoplasmata, Verrucomicrobiae, and Clostridia were significantly more abundant in the SSW (p < 0.05; Supplementary Figure 3A). The relative abundance of Alphaproteobacteria was not significantly different between the two layers (Figure 3). At the genus level, the abundance of Pseudoalteromonas, Erythrobacter, Psychrobacter, Vibrio, Halomonas, Pseudomonas were higher in the SML (p < 0.05; Supplementary Figure 3B). Meanwhile, the structure of bacterial communities in the ECS and YS was significantly distinct (Figure 3).

Bacterial communities in SML and SSW of the ECS were both dominated by Alphaproteobacteria, followed by Gammaproteobacteria, Flavobacteria, Cyanobacteria and Actinobacteria (Figure 3; Supplementary Figure 6A). Synechococcus and Erythrobacter were the most abundant genera in both SML and SSW samples, followed by unclassified Surface 1, NS5 marine group, Candidatus Actinomarina, Ruegeria, Paracoccus, norank SAR86 clade, Alteromonas and Sulfitobacter. Among them, Erythrobacter and Psychrobacter were more abundant in the SML (p < 0.001, Wilcoxon rank-sum test), while norank SAR86 clade was more abundant in the SSW (p < 0.05, Wilcoxon rank-sum test, Supplementary Figure 6B). Unlike in summer, the bacterial communities of SML and SSW in spring were dominated by Gammaproteobacteria, followed by Alphaproteobacteria, Actinobacteria and Flavobacteria (Sun et al., 2020).

Bacterial community structure in the YS was dominated by Gammaproteobacteria (43.63%–72.80% in SML and 11.54%–24.52% in SSW, respectively), which was significantly higher in SML than that of SSW (p < 0.05, Wilcoxon rank-sum test, Figure 3; Supplementary Figure 6C). Pseudoalteromonas (14.03%–52.00% in SML and 0.57%–11.05% in SSW, respectively) was the most abundant genera in the YS. Additionally, Alphaproteobacteria, Flavobacteria and Cyanobacteria were also with higher proportions in SML than in SSW samples. At the genus level, the relative abundances of Acinetobacter, Alteromonas, Formosa, Halomonas, Pseudomonas and Vibrio were higher in SML than those in SSW (p < 0.05, Wilcoxon rank-sum test, Supplementary Figure 6D).

To understand the influence of environmental factors on microbial communities, the predictor variables of environmental factors in SML and SSW of the ECS and YS were analyzed by db-RDA based on Bray-Curtis distances (Figure 4). Latitude, DOC, DMS, DMSP_d and DMSP_t were significantly correlated with the distribution of bacterial communities in SML samples (explained 45.57% by first axis and 7.34% by second axis, Figure 4A). Environmental factors (latitude, DOC, DO, Chl a, temperature, PEUK, PO₄³⁻ and NO₂⁻) were the main contributors affecting the bacterial community structure of SSW in the ECS and YS (explained 32.60% by first axis and 14.40% by second axis, Figure 4B). Different from the main influencing factors (longitude and DMS) in spring (Sun et al., 2020), the bacterial community of the ECS in summer was mainly affected by latitude, DOC, temperature, DMSP, PEUK and Chl a.

As the dominant DMSP biosynthesis gene in the eastern Chinese marginal sea, the total abundance of dsyB and mmtN in the SML were ~7.81 and 2.93 folds higher than in the SSW (p < 0.01, Wilcoxon signed-rank tests, Figure 5A; Supplementary Figure 7A; Supplementary Table 2). dsyB was generally more abundant (~33.79-fold) than mmtN (Supplementary Figures 7A,B) in all SML and SSW samples. dsyB also showed higher abundance in the FL fractions (1.03 ± 1.51 × 10⁶ copies L⁻¹) than in the PA fractions (1.01 ± 0.73 × 10⁵ copies L⁻¹) in the SML (Figure 5A), whereas the mmtN did not differ between these two lifestyles (Figure 5B). The abundance of dsyB showed an increasing trend from north to south, and was negatively correlated with latitude (p < 0.01, Spearman correlation tests, Supplementary Table 7). Moreover, we found that dsyB was negatively correlated with DOC in the SML, and significantly correlated with DO and temperature in the SSW (p < 0.05, Wilcoxon signed-rank tests, Supplementary Tables 7, 8). mmtN (especially the FL fraction) was correlated with longitude, pH, and DMSP in the SSW, while it was not found to be influenced by environmental factors in the SML (p < 0.01, Spearman correlation tests, Supplementary Tables 7, 8).

Geographically, dsyB gene abundance in the ECS was higher than that of the YS (~21.87-fold in SML, ~29.31-fold in SSW, respectively), but no significant difference was found for mmtN (Supplementary Figures 7A,B; Supplementary Table 5). Compared with the ECS samples in spring, dsyB and mmtN of the SML and SSW samples were significantly more abundant in summer (p < 0.01, Mann–Whitney tests, Supplementary Table 5), and the abundance of DMSP biosynthesis genes were consistently higher in the SML than in the SSW among both summer and spring samples.

For DMSP catabolic genes in the eastern Chinese marginal sea, dddP and dmdA (C/2 and D/1) were more abundant in the SML (3.86 ± 4.05 × 10⁷ copies L⁻¹ and 4.99 ± 7.69 × 10⁸ copies L⁻¹) than those in the SSW samples (6.16 ± 8.75 × 10⁶ copies L⁻¹ and 9.28 ± 9.33 × 10⁷ copies L⁻¹; ~6.27-fold and 5.38-fold, p < 0.01, Wilcoxon signed-rank tests, Figures 6A,B; Supplementary Figures 7C,D; Supplementary Table 2). Both dddP and dmdA genes showed higher abundance in the FL fractions than in the PA fractions (Figures 6A,B), and dmdA was more abundant (~12.92-fold for the SML and ~15.06-fold for the SSW) than dddP, indicating that dmdA-mediated demethylation is the main pathway of DMSP catabolism in the eastern Chinese marginal sea. Additionally, dmdA D/1 subclade was far more abundant than C/2 subclade both in the SML and SSW (Figure 6B; Supplementary Figure 7D). The dmdA D/1 subclades (especially the FL fraction) was negatively correlated with longitude, DMS, and DMSP in the SML (p < 0.05, Spearman correlation tests, Supplementary Tables 7, 8).

Geographically, the abundance of dmdA (D/1 and C/2) in both SML and SSW of the ECS was higher (6.19 ± 8.74 × 10⁸ copies L⁻¹ for SML and 9.64 ± 10.02 × 10⁷ copies L⁻¹ for SSW, respectively) than that of the YS (1.69 ± 1.17 × 10⁸ copies L⁻¹ for SML and 8.29 ± 8.36 × 10⁷ copies L⁻¹ for SSW, respectively, Figure 6B; Supplementary Figure 7D). Similarly, the abundance of dddP in SML and SSW of the ECS was also both higher (4.29 ± 4.61 × 10⁷ copies L⁻¹ for SML and 8.04 ± 9.60 × 10⁶ copies L⁻¹ for SSW, respectively) than that of the YS (2.70 ± 1.83 × 10⁷ copies L⁻¹ for SML and 9.93 ± 9.82 × 10⁵ copies L⁻¹ for SSW, respectively, Figure 6A; Supplementary Figure 7C). DMSP metabolic genes (dddP and dmdA, especially in the FL fraction) in the SML of the ECS rather than the YS were negatively correlated with longitude (p < 0.05, Spearman correlation tests, Supplementary Tables 9–12). Compared with the spring ECS samples, the abundance of dddP and dmdA genes in summer SML was higher (~8.73-fold and ~38.56-fold, p < 0.05, Mann–Whitney tests, Supplementary Table 5).

16S rRNA gene abundance (qPCR) was used to evaluate the absolute abundance of DMSP producing and catabolic genera in SML and SSW samples of the eastern Chinese marginal sea (Dickson et al., 1980; Sunda et al., 2002; Curson et al., 2017; Williams et al., 2019). The DMSP biosynthetic genera Ruegeria (with dsyB and mmtN), Alteromonas (with mmtN) and Croceicoccus (with mmtN) were the most abundant in the eastern Chinese marginal sea, and their relative abundances in the SML were much higher (3.67%, 2.88%, and 0.19%) than that in the SSW (3.11%, 2.07%, and 0.14%, Figures 5C,D). However, the distribution of DMSP biosynthetic genera was significantly different in the ECS and YS. For the ECS samples, the DMSP biosynthetic genera that may contain dsyB included Thalassobius, Oceanicola, Hoeflea and Albimonas, as well as Ruegeria, Roseovarius and Labrenzia which may contain both dsyB and mmtN (Williams et al., 2019) were higher (~1.28–8.51 folds) in SML than SSW samples (p < 0.05, Wilcoxon rank-sum test, Figures 5C,D). Thalassobaculum containing dsyB (Williams et al., 2019) only presented in near shore SML sample (D3 and W1), but were not found in SSW. Alteromonas, Croceicoccus and Novosphingobium, containing mmtN (Williams et al., 2019), were ~4.46–7.32 fold higher in SML than SSW samples (p < 0.05, Wilcoxon rank-sum test, Figure 5D). The genus Marinobacter, which can produce DMSP but the synthesis gene is unknown, was also more abundant (~4.82-fold for ECS, p < 0.05, Wilcoxon rank-sum test) in SML than SSW samples. As for the samples from YS, Ruegeria, Croceicoccus, Marinobacter, and Alteromonas was also more abundant in SML than that of SSW, and Thalassobaculum, Novosphingobium, and Streptomyces only presented in SML samples. In contrast, Thalassospira and Nisaea only presented in SSW samples (Figures 5C,D). Most of the DMSP-synthesis bacteria (except that Nisaea, only presented in YS) were more abundant in the ECS than those in the YS (Figure 5). In contrast to the spring samples from ECS, the DMSP-synthesis bacteria were more diverse and more abundant in summer, and Oceanicola, Thalassobius, and Thalassobaculum only appeared in summer samples. The abundance of Ruegeria was higher in summer than in spring samples from the ECS (Figures 5C,D; Sun et al., 2020).

The relative abundance of DMSP consumers, e.g., Roseobacter clade bacteria (Sulfitobacter, Paracoccus, Rugeria, Phaeobacter, Pseudophaeobacter), SAR11 clade, SAR116 clade, were higher (range from 1.02 to 14.77 folds) in SML than in SSW in the eastern Chinese marginal sea (Figures 6C,D). The distribution of DMSP-catabolic bacteria differed between the ECS and the YS. Loktanella, Thalassobius, Roseovarius and Labrenzia were more abundant (~1.83–8.51-fold) in SML of the ECS than in SSW, which were inverse in the YS. In contrast, we found higher abundance of Rubellimicrobium in the YS SML (containing dddP genes, ~4.14-fold higher than in the SSW). The abundance of most DMSP-demethylation and cleavage genera in the ECS were much higher than that in the YS, such as Roseobacter clade bacteria, SAR11 clade and Pseudophaeobacter (Figures 6C,D). Shimia containing dddP and dmdA only presented in the ECS sample, but were not found in the YS. In both summer and spring ECS, some representative DMSP-degrading bacteria such as Sulfitobacter, Paracoccus, Rugeria, Labrenzia, Shimia, SAR11 and SAR116 were far more abundant in SML than in SSW samples (Sun et al., 2020), However, DMSP-degrading bacteria were more abundant during the summer than in the spring, and Phaeobacter, Thalassobius, and Pseudophaeobacter only appeared in summer but not in spring (Figures 6C,D; Sun et al., 2020).

DMSP concentrations in SML were higher than that of SSW in both summer and spring, while no significant difference for DMS levels between SSW and SML was detected, irrespective of seasonality. The higher DMSP concentration could be related to the highly active DMSP biosynthesis in the SML. Although the DMSP cleavage pathway was also more active in the SML, the rapidly release of DMS from SML may lead to the observed similar DMS level in the SML and SSW. Seasonally, it is commonly reported the concentration of DMS or DMSP are higher in summer than in spring (Jian et al., 2019; Mao et al., 2021), which is consistent with our findings that the concentrations of DMS and total DMSP in SML of the ECS in summer were more abundant than those in spring (Supplementary Table 1; Sun et al., 2020). Additionally, the Chl a concentration of the ECS SML in summer was higher (~2.00-fold) than that in spring, reflecting the higher primary productivity and more active biological metabolism in summer, and this could be an explanation for the higher DMSP/DMS concentration in summer. However, there was no significant difference for Chl a level between SSW and SML (Supplementary Tables 3, 4), and no significant correlation was found between DMSP concentration and Chl a in both SML and in SSW. Combined with the study by Sun et al. (2020) in the SML of the ECS, it reinforces the idea that heterotrophic bacteria instead of eukaryotic algae, may have an important contribution to the higher DMSP concentration in SML.

The total bacterial abundance of the SML samples in the eastern Chinese marginal sea in summer was significantly higher (~4.58-fold) than that of SSW samples (Supplementary Figure 2B), which was consistent with the results of the ECS (~7.50-fold) in spring (Sun et al., 2020). This is also consistent with Sieburth et al. (1976) who found a ~10²–10⁴-fold bacterial enrichment in SML from the North Atlantic compared with SSW samples. The higher bacterial abundance in SML may be due to its higher concentrations of nutrients and organic matter, higher temperature, as well as the physically stable environment under the action of surface tension (Kuznetsova et al., 2004; Zhang et al., 2006; Galachyants et al., 2018). The abundance of bacteria in the ECS was higher (~2.02-fold) than that in the YS. The higher ECS bacteria abundance was most likely the result of the regional hydrography including Taiwan Warm Current and Changjiang plume flowing northward near the Yangtze Estuary, the joint influence of the Yellow Sea cold water mass and the Yellow Sea warm current, as well as the discharge of Yangtze Diluted Water (Lee and Chao, 2003; Mi et al., 2012; Yeh et al., 2015; Yu et al., 2021). Bacteria were more abundant (~3.28-fold) in summer than those in spring from the ECS (Sun et al., 2020), and the SML bacteria decreased with the increase of offshore distance in summer. This is consistent with Zhao (2014) who found bacterial abundance in summer were higher than those in spring in both the SML and SSW samples from the North Yellow Sea, and the SML bacteria were more abundant in the nearshore area in summer. This may be due to generally higher levels of available nutrients in the nearshore area (Maki, 2003) and higher primary productivity in summer.

The relative abundance of Gammaproteobacteria in the eastern Chinese marginal sea was higher in the SML compared to the SSW, while there was no significant difference in Alphaproteobacteria between the two layers (Figure 3; Supplementary Figure 3A). On the contrary, Sun et al. (2020) found that Alphaproteobacteria were more abundant in the SSW of the ECS in spring. Many genera, Pseudoalteromonas, Erythrobacter, Psychrobacter, Vibrio, Halomonas and Pseudomonas were more abundant in the SML samples (Supplementary Figure 3B). There were seasonal and regional differences in bacterial community composition between SML and SSW samples. Most genera, such as Erythrobacter, Ruegeria, Pseudoalteromonas, Alteromonas, Halomonas, Cobetia, Acinetobacter, Marinobacter and Vibrio were significantly higher in ECS SML samples compared with SSW samples both in summer and spring (Supplementary Figure 6B; Sun et al., 2020). However, the diversity and richness of bacterial community structure in SML and SSW samples of ECS in summer were significantly higher than those in spring (Supplementary Figures 5C,D). The relative abundances of Alphaproteobacteria, Flavobacteria, Cyanobacteria, Actinobacteria, Sphingobacteriia and Bacteroidia in the ECS SML were higher compared to the YS SML, whereas the relative abundances of Gammaproteobacteria and Bacilli were opposite (Figure 3), which may be due to the variation in levels of available nutrients, pH, DO, DOC, Chl a and temperature in different marine areas (Supplementary Table 5).

Correlated to the high DMSP concentrations, both the abundance of bacteria containing dsyB and mmtN (Figures 5A,B) and the abundance of DMSP-producing genera were higher in the SML samples (Figures 5C,D). Several DMSP biosynthesis genera, such as Thalassobaculum, were only found in the SML samples. At the gene level, dsyB and mmtN were more abundant in the SML in both regions and seasons, and were consistent with previous studies (Sun et al., 2020, 2021), the transamination pathway catalyzed by dsyB is the main DMSP production process in both SML and SSW. Furthermore, although dsyB and mmtN were more abundant in FL than in PA bacteria in spring ESC samples (Sun et al., 2020), no difference for mmtN was observed between the two lifestyles in summer samples (Figures 5A,B). However, considering the difference of abundant dsyB in FL and PA bacteria, FL bacteria could still be the main DMSP producers in the SML. Additionally, the major DMSP producing genera may be different between spring (Alteromonas) and summer (Ruegeria) of ESC (Sun et al., 2020).

Regionally, DMSP-production genes and the abundance and diversity of the corresponding DMSP-production bacteria were lower in the YS compared with the ECS (lower dsyB abundance, but no significant difference was found for mmtN). This may be due to the strong invasion of Kuroshio increases salinity in the summer ECS, phytoplankton and bacteria will produce more DMSP to balance intracellular osmotic pressure and thus the higher DMSP-synthesis bacteria and genes in the ECS (Sun et al., 2021).

The DMSP-catabolic genes (dddP and dmdA) and the corresponding bacteria (most of Roseobacter clade bacteria, SAR11 clade and SAR116 clade) were significantly more abundant in SML of the eastern Chinese marginal sea compared with SSW samples potentially indicating that bacterial DMSP catabolism more active in the SML. dmdA (C/2 and D/1 subclades) was higher than dddP in both summer and spring (Sun et al., 2020). This is consistent with Liu et al. (2018) who also found the genetic potential to cleave DMSP via the DddP DMSP lyase was far less prominent than that for DMSP demethylation in the ECS. It indicates that DMSP demethylation pathway may be more prominent than cleavage pathway. dmdA D/1 subclade in summer and dmdA C/2 subclade in spring had more potential to demethylate DMSP in the SML (Sun et al., 2020), indicating that the seasonal variation would affect the DMSP metabolism in gene subclades.

Consistent with the DMSP-synthesis gene (dsyB), DMSP-degradation genes (dddP and dmdA) in FL fraction were also more abundant than that in PA fraction. This is in agreement with Sun et al. (2020) who showed that the dmdA and dddP genes of FL bacteria was higher than PA bacteria in spring ECS. The dmdA D/1 subclades in the FL fraction was negatively correlated with DMSP in the SML of the eastern Chinese marginal sea (Supplementary Table 8). FL bacteria but not PA bacteria containing dmdA were significantly correlated with DMS and DMSP_t concentrations in ECS SML, and with DMSP_p concentration in ECS SSW (Supplementary Table 9). FL bacteria related to dddP and dmdA, but not PA bacteria, were correlated with DMSP_d concentration in both SML and SSW of the YS (Supplementary Table 12). Sun et al. (2020) also found the positive correlation between dddP and dmdA genes in FL bacteria and DMSP concentrations in SSW of the ECS in spring. These results indicate that FL bacteria may become the main DMSP consumers in the eastern Chinese marginal sea via the demethylation and cleavage pathways.

dmdA and dddP in SML and SSW samples from ESC was higher than that in the YS, which is consistent with the abundance of DMSP metabolic bacteria (Ruegeria, Sulfitobacter, Shimia and SAR11 clade; Figure 6). This result implied that the ECS may have a higher DMSP degradation potential compared with the YS. Furthermore, the significant negative correlation between the dddP and dmdA genes of FL bacteria and longitude in the ECS SML but not in the YS indicates that the DMSP-degradation capacity of the SML in the ECS may decrease with the increase of offshore distance. DMSP metabolic bacteria (Phaeobacter, Thalassobius and Pseudophaeobacter) and genes (dmdA and dddP) were higher in summer than spring in ECS (Sun et al., 2020). These results were also consistent with Kudo et al. (2018), who found dddP to be more abundant in summer compared to spring samples from Ofunato Bay, indicating that DMSP metabolic potential was more abundant in summer.