Construction and ecological exploration of TOGSMMs.

a. Graphical abstract of genome-scale metabolic model reconstruction applied on globally distributed metagenomes (n=127) and metatranscriptomes (n=94) (left). For this, known physics and chemistry of metabolic reactions are captured into metabolic networks (schematic right, where genes are depicted in purple, enzymes in pink, and metabolic reactions in blue that transform metabolites in white). b. The number of reactions for each metabolic model was reconstructed per sample from its metagenome, and the matching metatranscriptome was plotted against temperature to assess potential bias due to the latitudinal gradient. In all cases, metagenome-based model are larger (contains more reactions) than their corresponding metatranscriptome-based model, which was interpreted as the per-sample genetic potential (metagenome) versus genetic expression (metatranscriptome) sampled functional repertoires, respectively. Color coding represents previous ecological zones defined by microbial and virus community data27,26 (TT = Temperate and Tropical, ANT = Antarctic, ARC = Arctic), and the degree of color represents sampling depth as indicated in the figure. c and d. Ordination (UMAP, or Uniform Manifold Approximation and Projection) of per-sample presence/absence reaction predictions show that genome-scale metabolic models generated from metagenomes (c) and metatranscriptomes (d) structure along known ecological zones. Colors are the same as b.

Computing metabolic synergy across the global ocean and comparing other parameters.

a. Graphical abstract to capture TOGSMM workflows. For each TOGSMM, the metabolic solution space reflects all mechanistic solutions for any environmental conditions (i.e., the fundamental metabolic niche of the community30, which we suggest here represents a ‘metabolic phenotype’ for each sample). Because of the complexity of the mechanistic solutions, we summarize them by a flux sampling procedure that estimates each metabolic reaction’s flux distribution. To estimate the reaction’s strength for controlling the metabolic phenotype, pairwise correlations (all reactions vs all other reactions) were computed for each reaction within a network and summarized as a Reaction Cumulative Correlation (RCC) as done elsewhere30. For a given network, the RCCs’ distribution determines its metabolic synergy, offering a quantitative score for comparing metabolic phenotypes. b. Correlogram of metabolic synergy and environmental variables to assess statistically significant correlations (Spearman correlation).

Metabolic synergy across the global ocean.

a. Distribution of Tara samples based on metabolic reaction synergy from autonomous (blue) to interdependent (red) systems. The node size is proportional to the per-sample prokaryotic organism diversity. Directed edges represent Lagrangian connections between samples realized in less than two months to describe the potential acclimation of water masses. The darker the arrow’s color is, the stronger the Lagrangian connection is. In other words, the darker the arrow is, the higher the connection probability between the water masses is. b. Projection on the metabolic synergy across the global ocean visualized from an equator- (left) and pole- (right) centric perspectives. The data points were interpolated at each station (halo color on each map), and actual data values are the colored dots. c. Relationships between the metabolic synergy of the microbial ecosystem against the normalized impact of phosphate (purple), nitrogen (blue), and carbon (red) reactions and the potential prokaryotic ecosystem biomass production (orange).

RCC and impact of reactions targeted by virus-encoded Auxiliary Metabolic Genes (AMGs).

a. Representative example from sample 125_SRF of the RCC of all (blue) and randomly sampled (gray, see Materials and Methods) metabolic reactions as compared to the reactions targeted by virus-encoded AMGs (red). All samples showed a similar, statistically significant high RCC AMG spike (see Supplementary Fig. 14). b. Metabolic pathways associated with high RCC AMG-targeted reactions across ecological zones. The RCC of reactions targeted by AMGs, present in all networks and statistically belonging to the high RCC spike, were averaged per KEGG pathway (i.e., lines) and ecological zone (i.e., columns). RCCs are then displayed by opacity. Note that a single reaction may be shared among several KEGG pathways. c. Comparison between the summed RCC of reactions targeted by AMGs and the number of reactions per sample. Each dot represents a sample. Its color code indicates the temperature associated, and its shape characterizes the system: a circle for polar systems and a cross for temperate systems. Grey dots represent the summed RCC of randomly defined viromes. d. Contribution of AMGs to exometabolites and endometabolites through increased (i.e., red) or decreased (i.e., blue) metabolite production. Here we consider metabolites that are only exometabolites (n=26), or only endometabolites (n=28)41.