Enhanced nutrient supply promotes mutualistic interactions between cyanobacteria and bacteria in oligotrophic ocean

Abstract

Cyanobacteria can form complex interactions with heterotrophic microorganisms, but this relationship is susceptible to nutrient concentrations. Disentangling the cyanobacteria–bacteria interactions in relation to nutrient supply is essential to understanding their roles in geochemical cycles under global change. We hypothesize that enhanced nutrient supply in oligotrophic oceans can promote interactions among cyanobacteria and bacteria. Therefore, we investigated the planktonic bacteria and their interactions with cyanobacteria in relation to elevated nutrients caused by enhanced upwelling around a shallow and a deep seamount in the tropical western Pacific Ocean. We found obviously higher complexity of network occurred with significantly more cyanobacteria in the deep chlorophyll maximum layer of the shallow seamount when compared with that of the deep seamount. Cyanobacteria can shape bacterial interaction and community evenness in response to relatively high nutrient concentrations. The effects of the nutrients on cyanobacteria-related networks were further estimated based on the Tara Oceans data. Statistical analyses further showed a facilitative effect of nitrate concentrations on cyanobacteria–bacteria mutualistic interactions in the global oligotrophic ocean. By analysing the Tara Ocean macrogenomic data, we detected functional genes related to cyanobacteria–bacteria interactions in all samples, indicating the existence of a mutualistic relationship. Our results reveal cyanobacteria–bacteria interaction in response to nutrient elevation in oligotrophic ocean and highlight the potentially negative effects of global change on the bacterial community from the view of the bio-interaction.

1. Introduction

Globally upper-ocean thermal stratification has been enhanced as a consequence of global warming [1]. The strengthened stratification may prevent vertical nutrient supply to the euphotic zone [2], and thus directly affect the oceanic primary production. Marine planktonic photosynthetic organisms are responsible for approximately 50% of the Earth's primary production [3]. Within phytoplankton, cyanobacteria are one of the most ubiquitous groups, contributing about 25% of marine primary production [4]. Marine cyanobacteria and heterotrophic bacteria (hereafter bacteria) are actively involved in complex ecological interactions, which can directly or indirectly influence the biota and biogeochemical cycles in the environment [5,6]. Disentangling the interactions between cyanobacteria and bacteria in relation to nutrient supply is essential to understanding the roles of bacteria in marine geochemical cycles under global change.

A mutualistic relationship can develop between cyanobacteria and bacteria in the co-culture system, allowing their survival and growth in an oligotrophic environment without external nutrient supply [7]. However, this relationship is susceptible to the availability of nutrients. In a recent study, although the diversity of cyanobacteria and bacteria increased at the beginning of nutrient input, their mutualistic relationship was disrupted with the consuming of exogenous nutrients and their abundance and diversity decreased [8]. This mutualistic relationship was gradually recovered, and the diversity of bacteria increased after the exogenous nutrients were exhausted. Despite these observations in the co-culture system, the relationship between cyanobacteria and bacteria in relation to the nutrients in the ocean, which is not static with spatiotemporal changes in environmental factors, remains unclear.

Seamounts, which are seabed uplifts below the sea surface in the deep ocean and can enhance the vertical transport of nutrients [9,10], may serve as an appropriate research area to study the phytoplankton–bacteria interaction in relation to the nutrients. Depending on the distance from the seamount summit to the sea surface, seamounts can be categorized into shallow (less than 200 m), intermediate (200–400 m) and deep seamounts (greater than 400 m) [11]. Owing to the blocking effect of seamounts on ocean currents, unique hydrological environmental characteristics, such as Taylor column, internal waves, and upwelling, are formed around seamounts [12,13]. As a widely existing and prominent geological structure in the ocean [14,15], seamounts have a profound impact on the surrounding hydrological and chemical environments [16,17]. The unique hydrological environment of the seamount area has a significant impact on the distribution of nutrients, resulting in various distribution patterns of nutrients in the shallow seamount, deep seamount and non-seamount area [18]. The special physical and chemical environment of the seamount area further forms a unique biological community in the seamount area [17,19], which is manifested as higher plankton diversity and biomass [20]. These characteristics make seamounts excellent research areas to study the relationships among diversity, bio-interactions and nutrients. We hypothesize that elevated nutrient concentrations caused by enhanced upwelling in the seamount region support more cyanobacteria, which further promote the formation and strengthening of a mutualistic relationship with bacteria, to respond to the changes in nutrients in oligotrophic ocean. To test this hypothesis, we investigated the bacterial community and interaction at a shallow seamount named M4 in the Caroline Ridge and the deep seamount Kocebu Guyot, two seamounts with different characteristics of nutrient distribution in the oligotrophic waters of the western Pacific Ocean [21]. We constructed a bacterial interaction network for each water layer along the gradient of water depth in both the shallow and deep seamounts, particularly focusing on water layers shallower than 300 m, where the influences of the two seamounts are quite different. The bacterial communities and interactions in relation to nutrients were studied and compared. Furthermore, the interactions between cyanobacteria and bacteria in relation to the nutrient concentration on a global scale were studied using Tara Oceans data. An equation between the nutrient concentration and the cyanobacteria-related network was generated using regression analyses. The study aims to reveal the cyanobacteria–bacteria relationship in response to nutrient elevation in oligotrophic ocean waters and to highlight the potential effects of global change on the bacterial diversity and function from the view of the bio-interaction.

2. Material and methods

(a) . Study area and sampling

The M4 Seamount is located on the Caroline Ridge in the oligotrophic tropical western Pacific Ocean. The M4 Seamount is a typical shallow seamount with a peak less than 100 m below the sea surface. In August 2017, a comprehensive scientific expedition onboard the R/V Kexue was carried out in the Caroline M4 Seamount. Two sections with five stations across the seamount summit were established for water sampling. Station C0 was located at the summit of the seamount. The maximum water depths at stations C0, C4, C9, C15 and C20 were 103, 1506, 4501, 2 763 and 2738 m, respectively. Samples were collected at the following water depths: surface, deep chlorophyll maximum (DCM), 200, 1000, 2000, 3000 m and bottom. A total of 33 samples were obtained (electronic supplementary material, table S1). In May 2019, another comprehensive scientific expedition onboard the R/V Kexue was carried out at the Caroline M4 Seamount. Two sections with nine stations (D2, D3, D4, F6, F7, N, E, W, S) across the seamount summit were established for water sampling. Station D3 was located at the summit of the seamount. Samples were collected at the following water depths: surface, DCM and 200 m. A total of 26 samples were obtained (electronic supplementary material, table S1).

Published data of the deep seamount Kocebu Guyot were also included in the analyses (electronic supplementary material, table S1) [19]. Kocebu Guyot belongs to the Magellan seamount chain in the oligotrophic tropical western Pacific Ocean. Kocebu Guyot is 4300 m high and is a typical flat-topped deep seamount, as its peak is 1150 m below the sea surface. Station A5 was located at the seamount summit. The maximum depth of stations A2, A8, B1 and B8 were 3900, 3500, 3200 and 5003 m respectively. A total of 38 samples were included in the analyses (electronic supplementary material, table S1).

Data from the Tara Oceans survey were downloaded and reanalysed to study the interactions between cyanobacteria and bacteria in relation to nutrient concentration on a global scale. An operational taxonomic unit (OTU) table based on the Tara Oceans shotgun metagenome sequence 16S miTAGs, metadata and the table of functional gene abundance from the Tara Oceans macro-genomic data was obtained through http://ocean-microbiome.embl.de/companion.html [22] (electronic supplementary material, table S2). The OTU table of functional gene abundance was obtained through the following steps. The raw metagenomic sequencing data were assembled and gene prediction was performed. A reference gene set was constructed using the predicted gene coding sequences. Functional annotation of the reference gene set was conducted using the eggNOG and KEGG databases. OTUs annotated as bacterial were filtered out and normalized by subsample at a depth of 10 000.

The station map is visualized through Ocean Data View and shown in figure 1 [23].

Figure 1.

Sampling stations (top) in the sea areas of the Caroline M4 Seamount (left) and Kocebu Guyot (right), and for the Tara Oceans data (lower map).

A total of 20 l seawater was collected from each water layer at each station with a conductivity–temperature–depth (CTD) rosette system equipped with Niskin bottles. The seawater samples were prefiltered using 200 µm bolting-silk and then filtered by a 0.22 µm mixed cellulose esters (MCE) membrane. The filters were then immediately transferred to cryovials and stored in a −80°C refrigerator after adding RNA later RNA Stabilization Reagent (Qiagen, Germany).

Water temperature (T) and salinity (S) were measured in situ by probe. After fixation with manganese sulfate and alkaline potassium iodide solution, dissolved oxygen (DO) was also measured in situ with a relative standard deviation of ≤2% by the Winkler iodimetry method [24]. Chlorophyll a was extracted with 90% acetone at 4°C for 24 h and quantified using a Turner Designs Trilogy Fluorometer (Turner Designs, USA) [25]. Nutrient concentrations in the GF/F filtrate were determined with a continuous flow analyser (QuAAtro, Seal Analytical Limited, UK) according to the Joint Global Ocean Flux Study (JGOFS) spectrophotometric method [26].

(b) . DNA extraction, amplification and sequencing

Environmental DNA was extracted from the filters using the AllPrep DNA/RNA Mini Kit (Qiagen, Germany) according to the manufacturer's instructions. For 2017 data, two universal primers for bacteria, U341F (5′CCTACGGGRSGCAGCAG3′) [27] and R685 (5′ATCTACGCATTTCACCGCTAC3′) [28], were used to prepare 16S rDNA amplicons. To reduce PCR bias, three PCR replicates were set for each sample before sequencing. Libraries were prepared and sequenced by the Illumina MiSeq platform (Illumina, San Diego, CA, USA) using the 2 × 250 bp paired-end protocol. For 2019 data, two universal primers for bacteria, PrimerF (5′GTGCCAGCMGCCGCGG3′) and PrimerR (5′CCGTCAATTCMTTTRAGTTT3′), were used to prepare 16S rDNA amplicons. Synthetic sequences of known copy number of spike-in standards wereadded to the DNA samples before PCR amplification. These sequences were used as internal standards and helped in absolute quantification of the sample. Next, Accu16S™ absolute quantitative sequencing was performed by the Genesky Company (Shanghai, China) (electronic supplementary material, table S3). To reduce PCR bias, three PCR replicates were set for each sample before sequencing. Libraries were prepared and sequenced by the Illumina NovaSeq platform (Illumina, San Diego, CA, USA) using the 2 × 250 bp paired-end protocol.

(c) . Amplicon sequence analysis and statistical analysis

Raw sequencing data were processed according to the Easy Amplicon pipeline [29]. Primers and low-quality reads were removed using VSEARCH (v. 2.15.2). Then sequences were denoised to obtain an amplicon sequence variants (ASVs) table using USEARCH (v. 10.0.240). Taxonomic annotation was implemented using VSEARCH (v. 2.15.2) with the reference database RDP (v. 16). Before further analysis, the ASV table was normalized by subsampling to ensure that all samples were compared at the same level. Diversity indices were calculated by means of the Easy Amplicon pipeline and visualized using R package ‘ggplot2’.

(i) . Community assembly processes analysis

By analysing the phylogenetic and taxonomic composition of these communities, a null model framework was used to assess the relative importance of stochasticity (random fraction) and determinism (non-random fraction) in shaping microbial communities. Phylogenetic trees were first reconstructed through the Easy Amplicon pipeline. Next, the β-Nearest Taxon Index (β_NTI) was calculated by Stegen's method [30,31]. A β_NTI valueof more than 2 indicates that community assembly is driven by determinism, whereas otherwise it is driven by stochasticity. The Raup–Crick matrix (β_RC) was used to determine the role of dispersal during community assembly [32]. A β_RC value of more than 0.95 indicates that community assembly is driven by probabilistic dispersal, whereas otherwise it is driven by undominated factors.

(d) . Source tracking analysis

Source tracking analysis was performed through the R package FEAST [33]. In short, for each water layer, the samples of this layer were analysed as the sink and all other samples as the source. The results were visualized by using R package ‘ggplot2’.

(e) . Network construction and analysis

Network analysis is an important method to describe and visualize microbial interactions. Only ASVs with a relative abundance of more than 0.1% were retained in the network analysis. All networks were constructed by using the Molecular Ecological Network Analysis Pipeline (MENAP) (http://ieg2.ou.edu/MENA) [34,35] and Integrated Network Analysis Pipeline (iNAP) (http://mem.rcees.ac.cn:8081) [36]. In short, based on the results of cluster analysis, the samples from each seamount were divided into five groups: surface layer, DCM layer, 200–300 m layer, 500–1000 m layer, and ≥2000 m layer. Then, for each group, ASVs present in more than half of the samples were screened for subsequent analysis. Next, a method based on random matrix theory (RMT) was used to automatically determine the appropriate threshold. Finally, the bacterial co-occurrence network was constructed by the Pearson method with 0.95 as the threshold. Absolute quantitative sequencing data were divided into three groups (surface, DCM, 200 m) and networks were constructed using the same method.

Networks based on the Tara Oceans data were constructed in the same method. Considering community similarities, environment similarities and geographical distances together, surface seawater samples from the Tara Oceans survey were screened and divided into seven groups, each containing five stations: IO_SRF1 (TARA_041, TARA_042, TARA_045, TARA_048, TARA_052), IO_SRF2 (TARA_056, TARA_057, TARA_062, TARA_064, TARA_065), MS_SRF (TARA_023, TARA_025, TARA_030, TARA_007, TARA_009), NPO_SRF (TARA_109, TARA_132, TARA_137, TARA_138, TARA_140), SAO_SRF (TARA_068, TARA_070, TARA_072, TARA_076, TARA_078), SPO_SRF1 (TARA_111, TARA_112, TARA_096, TARA_098, TARA_099), SPO_SRF2 (TARA_122, TARA_123, TARA_124, TARA_12, TARA_128). A co-occurrence network was constructed with a threshold of 0.95 for each group.

To investigate the role of nutrients in the network, a co-occurrence network containing ASVs and environmental factors was constructed. The ASVs with a relative abundance of more than 0.1%, the nutrients, and the physical parameters in the M4 Seamount surface layer, DCM layer and 300 m layer were used to construct the network. In order to better calculate the correlation between ASVs and environmental factors, the data were normalized by logarithmic transformation and the Spearman method was used to calculate the correlation. A co-occurrence network containing bacteria and environmental factors was constructed with 0.82 as the threshold.

The functional genes associated with cyanobacteria–heterotrophic bacteria interactions were extracted from the Tara Oceans macro-genomic data according to the catalogue provided by Nair et al. [8]. These genes are associated with phosphorus cycling, nitrogen cycling, vitamin B12 synthesis, bacterial secretory systems, quorum sensing (QS), quorum quenching (QQ), chemotaxis and biofilm formation.

(f) . Correlation analysis

The correlation coefficients between nutrients and ASVs contained in the bacterial co-occurrence network in the DCM layer of the M4 Seamount were calculated via the R package ‘Hmisc’. The correlation coefficients within the 0–300 m water layer were obtained and the Mantel test between the nutrients and the key taxa were performed by using the R package ‘linkET’. The correlation between the network index and environmental factors of the Tara Oceans co-occurrence network was calculated and visualized by means of the R package ‘linkET’.

3. Results

(a) . Species composition and source tracking of bacteria in the seamounts

In the M4 Seamount area, cyanobacteria were the most abundant in both the surface and DCM layers, followed by Gammaproteobacteria, Alphaproteobacteria and Flavobacteriia (figure 2a). Gammaproteobacteria and Alphaproteobacteria had the highest relative abundance in deeper water layers. In contrast, Gammaproteobacteria was the dominant group in both the surface and DCM layers of Kocebu Guyot, while the relative abundance of cyanobacteria was much lower than at the M4 Seamount in both layers (electronic supplementary material, figure S1).

Figure 2.

(a) Relative abundance and diversity of major assemblages in nine water layers of M4 Seamount. (b) Sources of each water layer at the M4 Seamount and Kocebu Guyot (Koc). DCM, deep chlorophyll maximum layer.

The source tracking analysis indicated that the components in the surface, 200 and 300 m water layers were respectively sourced more from the adjacent deep layer at the M4 Seamount than were those at Kocebu Guyot (figure 2b). The difference in the source between the two seamounts decreased in the water layers below 500 m.

(b) . Bacterial co-occurrence network in the seamounts

Network complexity represented by the number of edges in both seamounts showed the same trend along the gradient of water depth. The edge number of the network increased first, peaked in the DCM layer, and then decreased with the increase of water depth (figure 3a). The network complexity in the DCM layer of the M4 Seamount was much higher than that in other layers, although bacterial species richness was not the highest and the bacterial community in the DCM layer showed similar proportions to those in the surface layer (figure 3b).

Figure 3.

(a) Bacterial co-occurrence network from the surface layer, deep chlorophyll maximum (DCM) layer, 200–300 m layer, 500–1000 m layer, and ≥2000 m layer at the M4 Seamount with cutoff of 0.95. The degree of each node in the network is mapped to the size. The red edges indicate the positive correlations and the blue ones indicate the negative correlations. (b) Bacterial average richness in each sample of the surface, DCM, 200–300 m, 500–1 000 m, 2000 m, and deeper water layers at the M4 Seamount and Kocebu Guyot. (c) Degree (number of edges connected to a node, reflects the importance of the node in the network) and relative abundance (average read) of each group in the M4 Seamount and the Kocebu Guyot networks.

Positive correlations were dominant in all the networks. With increasing water depth, the proportion of positive correlations gradually increased: about 65.9% in the surfacer layer, 72.5% in the DCM layer, 77.9% in the 200–300 m layer, 87.7% in the 500–1000 m layer and 100% in the water layers below 2000 m at the M4 Seamount. The proportion of positive correlations of the Kocebu seamount networks showed a slightly decreasing and then increasing trend: about 73.5% in the surface layer, 70.6% in the DCM layer, 76.2% in the 200–300 m layer, 93.1% in the 500–1000 m layer.

By comparing the bacterial co-occurrence networks at the two seamounts, the network complexity was similar in both the surface layer and the 200–300 m layer, but distinctly different in the DCM layer (640 edges at the M4 Seamount versus 108 edges at Kocebu Guyot), though the number of nodes is similar (table 1). In terms of the taxonomic composition, the proportion of the total edge number of Gammaproteobacteria was higher in the euphotic zone of both seamounts, while Cyanobacteria (26.1%) in the DCM network of the M4 Seamount replaced Gammaproteobacteria (21.3%) to occupy the most edge (figure 3c). The co-occurrence network was also constructed using absolute quantitative data from the M4 Seamount in 2019. Similar values and trends were obtained with the relative quantitative data: the DCM layer had the highest network complexity (106 nodes and 844 edges), followed by the surface layer (91 nodes and 317 edges), and the 200 m layer had the lowest network complexity (54 nodes and 60 edges) (table 1).

Table 1.

The number of nodes and edges of each bacterial co-occurrence network for the two seamounts. DCM, deep chlorophyll maximum.

	surface			DCM			200–300 m
seamount	M4 (2017)	M4 (2019)	Kocebu	M4 (2017)	M4 (2019)	Kocebu	M4 (2017)	M4 (2019)	Kocebu
total nodes	68	91	77	108	106	102	50	54	57
total links	82	317	113	640	844	119	77	60	105

(c) . Relationship between nutrients and co-occurrence network

Obvious upwelling occurred in the M4 Seamount area, resulting in higher nutrient level and phytoplankton abundance than those in the surrounding non-seamount area [21]. Stronger upward nutrient transport was observed in the M4 Seamount area than at Kocebu Guyot (figure 4). The concentrations of the ${\mathrm{NO}}_3^{-}-\mathrm{N}$, ${\mathrm{SiO}}_3^{2-}-\mathrm{Si}$ and ${\mathrm{PO}}_4^{3-}-\mathrm{P}$ showed a similar variation trend along the depth gradient in the two seamounts. These nutrient concentrations were higher in water layers from 500 to 200 m at the shallow M4 Seamount when compared with those at the deep seamount Kocebu Guyot.

Figure 4.

Variation of average nutrient concentrations with depth in the M4 Seamount and Kocebu Guyot areas. The M4 Seamount exhibits nutrient uplift at shallower depths.

The analysis of community assembly processes indicates that determinism, namely ecological selection imposed by biotic and abiotic factors, accounted for 60.8% of the community in water depths shallower than 300 m at the M4 Seamount, and the probabilistic dispersal, which means that the species with high dispersal capacity make the community homogeneous, contributed 12.4% (figure 5a).

Figure 5.

(a) The proportion of determinism, probabilistic dispersal and undominated in community assembly at the M4 Seamount. (b) Bacterial co-occurrence network containing environmental factors from the water layer shallower than 300 m at the M4 Seamount, and the influence of the ${\mathrm{NO}}_3^{-}\text{-}\mathrm{N}$, ${\mathrm{NO}}_2^{-}\text{-}\mathrm{N}$, ${\mathrm{PO}}_4^{3-}\text{-}\mathrm{P}$, ${\mathrm{SiO}}_3^{2-}\text{-}\mathrm{Si}$ in the network. The degree of a point in the network is mapped to the size. DO, dissolved oxygen; Chla, chlorophyll a; T, temperature; S, salinity.

To further determine the role of nutrients in the determinism process, a co-occurrence network containing environmental factors and bacteria was constructed for water layers shallower than 300 m at the M4 Seamount and Kocebu Guyot (figure 5b; electronic supplementary material, figure S2). The results showed that the nutrients ${\mathrm{NO}}_3^{-}-\mathrm{N}$, ${\mathrm{NO}}_2^{-}-\mathrm{N}$, ${\mathrm{PO}}_4^{3-}-\mathrm{P}$ and ${\mathrm{SiO}}_3^{2-}-\mathrm{Si}$ had relatively high degrees (44, 41, 36 and 35, respectively) in the M4 Seamount network, while most of the nutrients had low degrees in the Kocebu Guyot network.

To determine the effects of nutrients on the community of the DCM layer at the M4 Seamount, Spearman correlation coefficients of all bacterial ASVs contained in the DCM network and environmental factors were calculated (figure 6a). Among 108 ASVs contained in the M4 Seamount DCM layer network, 45 ASVs were significantly correlated with nutrients, accounting for 41.7% of the total ASVs. These 45 ASVs mainly belong to the Cyanobacteria-Prochlorococcus, Alphaproteobacteria-SAR11 group, Gammaproteobacteria-Alteromonadaceae, and Flavobacteriia-Flavobacteriaceae. In addition, 56 ASVs were indirectly related to nutrients. The ASVs directly or indirectly related to nutrients accounted for 93.5% of the total ASVs in the network.

Figure 6.

(a) The proportion of directly and indirectly related ASVs of nutrients in the bacterial co-occurrence network for the M4 Seamount deep chlorophyll maximum (DCM) layer. (b) The correlation between environmental factors and high abundance groups for water layers shallower than 300 m at the M4 Seamount. T, temperature; S, salinity.

The Mantel test further showed that Cyanobacteria-Prochlorococcus and Alphaproteobacteria-SAR11 were significantly correlated with ${\mathrm{NO}}_3^{-}-\mathrm{N}$, ${\mathrm{PO}}_4^{3-}-\mathrm{P}$ and ${\mathrm{SiO}}_3^{2-}-\mathrm{Si}$ (Mantel's p < 0.05, figure 6b).

(d) . The role of cyanobacteria

The cyanobacteria were mainly distributed in the surface layer and DCM layer of the two seamounts, with higher diversity in the DCM layer than in the surface layer and higher relative abundance at the shallow M4 Seamount than at the deep Kocebu Guyot. Prochlorococcus was the main group of cyanobacteria in the two seamounts, accounting for 84.8% of the total relative abundance of cyanobacteria. The ASV-3, belonging to the high-light fitness II (HL II) Prochlorococcus, had a higher relative abundance in the DCM layer of the M4 Seamount compared with the Kocebu Guyot.

A total of 295 edges were related to cyanobacteria in the DCM layer network of the M4 Seamount, and most of them were positively correlated (figure 7d). Among the cyanobacteria-related groups, Gammaproteobacteria contributed the highest proportion, followed by Flavobacteriia. In contrast, only 23 edges were related to cyanobacteria in the DCM layer network of Kocebu Guyot, with the proportion of positively correlated edges lower than that for the M4 Seamount (figure 7d).

Figure 7.

(a) Richness of heterotrophic bacteria (excluding cyanobacteria) in the surface and DCM layers of the M4 Seamount and Kocebu Guyot. (b) Pielou's evenness index of heterotrophic bacteria in the surface and DCM layers of the M4 Seamount and Kocebu Guyot. (c) Relative abundance of cyanobacteria in the surface and DCM layers of the M4 Seamount and Kocebu Guyot. (d) The proportion of cyanobacteria–cyanobacteria type edges (Cya-Cya) and cyanobacteria-other type edges (Cya-other) in the DCM layer network of the two seamounts, and the proportion of positively correlated edges (pp) and negatively correlated edges (np) in the two types. (e) Bacterial interaction network in the M4 Seamount and Kocebu Guyot surface layer and DCM layer before and after cyanobacteria removal.

To further reveal the importance of cyanobacteria in the network, the network was reconstructed after the removal of cyanobacteria. The surface network for the M4 Seamount, and both the surface and the DCM networks for Kocebu Guyot were not greatly affected. However, nearly half of the edge of the DCM network of the M4 Seamount was lost, with the decrease of the positive correlation proportion (figure 7e). After the removal of cyanobacteria, the community evenness of bacteria in the DCM layer of the M4 Seamount became significantly higher than that for Kocebu Guyot (figure 7a,b).

Moreover, in terms of cyanobacterial community structure, the DCM layer was dominated by many cyanobacterial ASVs, while the surface layer was dominated by only one cyanobacterial ASV in both seamounts (figure 7c).

To further reveal the role of cyanobacteria in the co-occurrence network, the absolute quantitative network of the DCM layer of the M4 Seamount was analysed. Similar to the results of the relative quantitative network, cyanobacteria-associated edges contributed a high percentage to all edges (265 edges, 31.4%). In addition, the vast majority (15 of 18) of cyanobacteria appearing in the network are Prochlorococcus. Among these cyanobacteria, the abundant ASV-4, ASV-8, ASV-15, ASV-16, ASV-42 and ASV-43 belong to Prochlorococcus MIT9313 (electronic supplementary material, table S4). In addition, we classified the taxonomic annotations of the ASVs associated with these cyanobacterial groups in the network and found that 51.8% of the ASVs belong to the SAR11 clade (electronic supplementary material, table S4).

At the community level, we found that domination by an individual ASV is not evident in the network, but, rather, multiple ASVs play important roles. Moreover, the importance of ASVs in the network (measured by the degree of ASV nodes, i.e. the number of ASVs interacting with the ASV) is not directly related to their relative or absolute abundance. We found that the communities with more evenly distributed ASV abundance had more complex network structures and more positive relationships when compared with the communities dominated by a single ASV in abundance.

(e) . Global bacterial co-occurrence network

Through the data of the Tara Oceans, a total of seven co-occurrence networks of surface plankton bacteria were created for the Indian Ocean, South Atlantic, North Pacific and South Pacific Oceans (figure 8a). Among them, the percentage of positive correlation in the SPO_SRF2 network was as high as 90.4%, while it was only 31.3% in the MS_SRF, with the lowest average nitrate concentration. By annotating the ASVs with correlations to cyanobacteria across all networks, we showed that the most widely interacting taxa with cyanobacteria belong to the SAR11 clade, which accounted for more than 30% of the total correlations in all groups, and more than 50% of the total in each of the four groups, IO_SRF1, IO_SRF2, SPO_SRF1 and MS_SRF (electronic supplementary material, table S5).

Figure 8.

(a) The bacterial co-occurrence networks in surface water of global oceans, and the proportion of cyanobacteria-related edges shared with those of other groups, as well as positive and negative correlations among cyanobacteria-related edges. (b) Correlation between nutrient levels and bacterial co-occurrence network indexes in surface water of global oceans. Asterisks indicate significance according to Spearman's correlation test: ***p < 0.001; **p < 0.01; *p < 0.05. (c) Regression analysis of mean nitrate (μmol l⁻¹, transformed by ln(x + 1)) with the proportion of positive cyanobacteria-related edges.

The nitrate levels of each group in descending order were as follows: SPO_SRF2, SAO_SRF, IO_SRF2, NPO_SRF, SPO_SRF1, IO_SRF1, MS_SRF. The highest average nitrate concentration was observed in SPO_SRF2, about 4.08 µmol l⁻¹. Although nutrients did not show a significant correlation with the diversity of cyanobacteria and bacteria (electronic supplementary material, figures S3–S7), it was noted that the proportion of total positive edges, the proportion of cyanobacteria-related edges and the proportion of positive edges in cyanobacteria-related edges in the network were significantly positively correlated (p < 0.05) with total nitrate (figure 8b). At the same time, the proportion of the total negative edges in the network showed a significant negative correlation (p < 0.05) with the total nitrate (figure 8b). After regression analysis, the log-transformed (ln(x + 1)) nitrate concentration showed a linear relationship with the proportion of positive edges in cyanobacteria-related edges (figure 8c, R² = 0.73). The slope of the regression line increased after removing the two areas (SAO_SRF and SPO_SRF2) with relatively high nitrate concentrations (figure 8c).

The functional genes related to cyanobacteria–heterotrophic bacteria interactions are present in all the samples collected worldwide (electronic supplementary material, table S6). Although these genes did not exhibit significant statistical differences across different marine regions with various nutrient concentrations, some genes tend to have lower average relative abundances in regions with higher nutrient levels. For instance, the phosphate cycle genes showed the lowest average relative abundance in the group SPO_SRF2 with the highest nutrient concentration, while they had the highest average relative abundance in MS_SRF with the lowest nutrient concentration.

4. Discussion

In ecosystems, interactions such as predation, competition and symbiosis often occur between different species [37]. Owing to the complexity of species interactions and the difficulty in cultivation of organisms, it is challenging to verify these interactions experimentally. Therefore, bioinformaticians have developed a range of tools that can infer putative interactions from high-throughput sequencing. The correlation analysis and co-occurrence network based on the analysis of amplicon data have been commonly used to infer species interactions [35,38–41]. Despite some initial scepticism about the application of this method, the use of theories such as random matrix theory in co-occurrence network construction has gradually improved the reliability of this method [34]. Usually, significant spatial co-occurrence is considered an indicator of positive or mutualistic interactions, while co-exclusion is considered an indicator of negative interactions such as competition [42]. Therefore, in this study, we used correlation-based co-occurrence networks to infer putative ecological interactions between cyanobacteria and other bacteria.

As important primary producers, cyanobacteria play an important role in marine geochemical cycles [43–45]. Cyanobacteria tend to form complex interrelationships with other heterotrophic groups [6]. A recent study indicated that cyanobacteria could form a symbiotic relationship with heterotrophic bacteria in a cyanobacteria–bacteria co-culture system with natural seawater when the nutrients were depleted, and this relationship could be maintained for up to two years without exogenous nutrient input [7]. However, after the co-cultures of cyanobacteria and bacteria were transferred to seawater with high nutrient concentrations of 263.5 µmol l⁻¹ ${\mathrm{NO}}_3^{-}$ and 14.3 µmol l⁻¹ ${\mathrm{PO}}_4^{3-}$, their mutualistic relationship was disrupted [8]. This suggests that the input of additional nutrient may be detrimental to the maintenance of positive interactions between cyanobacteria and bacteria.

However, our study yielded an opposite conclusion, showing enhanced nutrient concentrations could promote the cooperative relationship between cyanobacteria and bacteria. This contradiction is likely largely due to the big difference in the nutrient level. The nitrate and phosphate concentrations used in the previous study were extremely high and uncommon in natural conditions. In contrast, the nitrate concentrations were only about 0.04–0.99 µmol l⁻¹ ${\mathrm{NO}}_3^{-}$ in the euphotic zone in the oligotrophic region of the Pacific Ocean. There might exist a nutrient threshold for regulating the positive and negative interactions. By analysing the data from the Tara Oceans survey on a global scale, we showed that nitrate had a facilitative effect on the positive relationship between cyanobacteria and bacteria at nitrate concentrations below 5.11 µmol l⁻¹. Under such circumstances, the positive relationship could be enhanced by relatively high nutrient concentrations in global oligotrophic ocean. Particularly, the increase in the nitrate concentration had a stronger contribution to the formation of cyanobacteria–bacteria interactions at lower nitrate concentrations. Further, cooperative relationships regulate the relative abundance of bacteria, improve the bacterial evenness and subsequently increase the community stability [46].

In our study, the major cyanobacterial group was Prochlorococcus. To adapt to oligotrophic environments, Prochlorococcus has evolved streamlined genomes, leading to a loss of some metabolic genes, such as genes encoding catalase [47]. In this case, Prochlorococcus cells rely on some heterotrophic bacteria for growth and environmental adaptation [48,49]. In turn, heterotrophic bacteria benefit from the organic carbon fixed by Prochlorococcus' photosynthesis and some essential organic compounds, such as enzymes and vitamin B12 [50]. Previous studies have identified a mutualistic relationship between the Prochlorococcus cyanobacteria and the heterotrophic bacterial SAR11 clade, and that between Prochlorococcus and Alteromonas as well as Marinobacter [51]. We found a similar mutualistic relationship from the co-occurrence network, in which the microorganisms analysed are mostly within the above categories. The widespread detection of such interactions provides support for our result. Moreover, previous studies have also proposed functional genes related to phosphorus cycling, nitrogen cycling, vitamin B12 synthesis, bacterial secretory systems, quorum sensing (QS), quorum quenching (QQ), chemotaxis and biofilm formation as indicators of mutualistic relationship between cyanobacteria and heterotrophic bacteria [8]. To verify the mutualistic relationship, we re-analysed the metagenomic data from the Tara Oceans survey, particularly those from the oligotrophic ocean (electronic supplementary material, table S6). The results showed that these functional genes were present in all the samples collected worldwide, indicating the existence of mutualistic relationship between the cyanobacteria and bacteria.

The tropical western Pacific region has the world's highest sea surface temperatures and is characterized by oligotrophic conditions [52,53]. Therefore, in this region, cyanobacteria and other phytoplankton often face strong nutrient limitations. Seamounts, which can promote the upward transport of nutrients, have been shown to induce a long-term increase in chlorophyll, especially at shallow seamounts [54]. Previous studies observed significant increases in phytoplankton, including cyanobacteria biomass in the euphotic zone of the M4 Seamount area [55]. Our study further revealed that the relative abundance of cyanobacteria at the shallow seamount was higher than that at the deep seamount. The higher relative abundance of cyanobacteria could be attributed to the greater nutrient supply caused by the upwelling around the shallow seamount [56]. This is confirmed by the fact that a high relative abundance of high-light fitness II Prochlorococcus, as a potential indicator for the enhanced vertical mixing [57], occupied the DCM layer of the M4 Seamount. However, the nutrientsupply caused by the upwelling does not change the oligotrophic state of the surface water, and nutrient limitation still persists. Therefore, the increased abundance of Prochlorococcus establishes and stabilizes more cyanobacteria–heterotrophic bacteria interactions based on metabolic products to meet their growing requirements and to fully use the increased primary productivity. At the same time, the nutrient supply in the seamount region is relatively stable and modest, which provides a stable environment for the formation of cooperative relationships compared with the drastic nutrient changes in the co-culture environment. Further, cooperative relationships regulate the relative abundance of bacteria, improve the bacterial evenness, and subsequently increase community stability [46].

Based on the findings of our study and the previous co-culture experiments, we propose that there might exist a threshold of nutrient level. Below the threshold more nutrient supply promotes the formation of cyanobacteria–bacteria mutualistic relationship, while such a relationship may be disrupted above the threshold. With an increase of additional nutrients, the nitrogen limitation disappears and there is no need for high-frequency exchange of metabolites between taxa to cope with nutrient deficiency. However, this hypothesis needs to be tested by further studies in the future.

Overall, enhanced nutrient supply in oligotrophic oceans promotes mutualistic interactions between cyanobacteria and heterotrophic bacteria, which perform important roles in marine geochemical cycles. However, global climate change has gradually caused surface seawater temperatures to increase [58]; such a process will intensify seawater stratification and impair the nutrient supply to the surface layer from the deep seawater [1]. In oligotrophic oceans, weakened nutrient supply might lead to a decrease in the relative abundance and diversity of cyanobacteria and a reduction in the interactions between cyanobacteria and heterotrophic bacteria, resulting in a potentially negative effect on marine primary productivity and biogeochemical cycles. In the nutrient-rich coastal region, human-related drastic changes in local nutrient supply may further disturb the collaboration between cyanobacteria and heterotrophic bacteria, impacting the stability and structure of the ecosystem. Our study highlights the potential effects of global change on bacterial diversity and function from the view of bio-interaction, and reveals cyanobacteria are vital in the nutrient cycle and in shaping bacterial diversity in oligotrophic oceans.

Ethics

This work did not require ethical approval from a human subject or animal welfare committee.

Data accessibility

The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive in the National Genomics Data Center, China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA: CRA008675), and are publicly accessible at https://ngdc.cncb.ac.cn/gsa.

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

W.L.: data curation, formal analysis, investigation, methodology, visualization, writing—original draft; F.Z.: conceptualization, data curation, formal analysis, supervision, validation, writing—original draft, writing—review and editing; X.L.: data curation; S.Z.: investigation; L.L.: formal analysis, writing—original draft; R.Z.: investigation; K.X.: conceptualization, funding acquisition, project administration, resources, supervision, writing—original draft, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Conflict of interest declaration

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

The research was supported by the National Natural Science Foundation of China (no. 41930533), the Strategic Priority Research Program of the Chinese Academy of Sciences (CAS) (no. XDB42000000), the Laoshan Laboratory (no. LSKJ202203102), and the Youth Innovation Promotion Association, CAS (no. 2022206).