This study set out to test six hypotheses under the assumption that 16S rRNA sequences can offer significant insight into microbial ecological functions in oceanographic studies. The metabolic reconstructions showed strong latitudinal gradients in the microbial metabolic pathways which coincide geographically with the four oceanographic provinces and are in agreement with the current mechanistic understanding of microbial biogeography in the South Pacific Ocean 7, 17 (confirming H1, Fig. 1B). More specifically, the trends in microbial functions mirrored the latitudinal variation in physico-chemical and biological measurements (which included temperature; nutrient bioavailability; diagnostic pigments, such as fucoxanthin, peridinin, chl-b, and zeaxanthin; and the isotopic fractionation of particulate organic nitrogen). This shows that our 16S rRNA sequencing data can complement previously published datasets on biomass and rate measurement for microbial communities in the South Pacific Ocean and, thus, improve our understanding of ecological changes observed across this basin. We were also able to find evidence which supported H2-H6, as described in more detail in the following sections.
Primary productivity shapes ecological functions of the bacterial community (H2)
Our second hypothesis, based upon the fact that primary productivity (PP) is stimulated by increases in nutrient concentrations at frontal zones, proposed that the types of microbial pathways (and overall functional community structure) in frontal systems should be related to biomass production. Raes, van de Kamp 17 showed that, within the GO-SHIP P15S transect, the SO and the SPSG were areas of low PP, whereas the STF and the PED had relatively high PP (Fig. 1 B, 3 A). The authors also observed an important trend along this transect: a switch from net autotrophy (i.e. high C-fixation) in the STF to heterotrophy (i.e. high nitrification and degradation of organic matter) in the SO 17. In the oligotrophic SPSG we observed high NH4+ assimilation rates, an increase in the δ15N-PON, and a higher abundance of picoplankton, which suggests an active microbial loop in this region. All pathways associated with CO2-fixation, energy metabolism, and nucleotide biosynthesis showed similar latitudinal trends, which aligned with variations in PP (Fig. 3 A, B, C). In our model predictions, concentrations of nanoplankton and chl-b alone explained 58% of the latitudinal changes in PP, but also explained ~50% of the latitudinal trends for CO2-fixation and energy pathways (Fig. 1 B and 3, A, B and C). Besides supporting H2, these results also support previous findings that PP (energy production) is a main driver for archaeal and bacterial richness across frontal boundaries in the South Pacific Ocean 7.
Temperature-regulated growth and nutrient limitation (H3 and H4)
Our third hypothesis was formulated upon the concept that micro-organisms follow a second order growth curve with temperature 19, 20. We recorded an increase in the percentage of the bacterial community which displayed cell structure and cell wall biosynthesis pathways as temperature -a proxy for latitude- increased (Fig. 4 E). However, we also found that NH4+ and the concentration of photosynthetic prokaryotes, together with temperature, explained ~80% of the variance in cell structure and cell wall biosynthesis pathways in our models (Supplementary Table 3). This is in agreement with previously published data for the tropical (i.e., high temperatures), oligotrophic SPSG, which was characterized by relatively higher NH4+ uptake rates and picoplankton concentrations as well as organic matter with higher δ15N 17; indications of a food web that is dominated by high turnover of organic material (greater cell wall biosynthesis). It seems, thus, that the presence and relative abundance of cell growth-related pathways is not only directly related to temperature, but also to picoplankton abundance.
Our fourth hypothesis postulated that the inferred metabolic predictions would result in latitudinal trends which reflect microbial strategies in coping with trace metal and macro nutrient limitations. As expected, we observed bimodal latitudinal trends for the biosynthesis of secondary metabolites, co-factors (which were identified as key pathways), and vitamins, which we related to trace metal and co-nutrient limitations. In the more productive regions (STF and PED), expression of the above-mentioned pathways was lower relative to the least productive regions. Iron supply to the SO is limited, causing the region to have low chlorophyll production despite its high nutrient concentrations22. Similarly, inorganic macronutrient concentrations in the oligotrophic SPSG are at or below the detection limit in surface waters 23. The above-mentioned biosynthesis pathways thus represent strategies used by the microbial community to cope with (essential) micro and macro nutrient limitation.
Energy storage and degradation (H5 and H6)
Our fifth hypothesis suggested that pathways associated with energy storage, such as lipid and carbohydrate synthesis, would be most active in the SO and in the STF. Lipids and carbohydrates are structurally essential molecules and important energy sources 29, such that, when nutrients are abundant, microorganisms allocate C to lipid biosynthesis 30. The SO and STF are highly seasonal environments where the strong differences in light availability between the winter and summer profoundly impact lipid trophodynamics, such that energy stored during the light season becomes critical for survival in the darker months 24, 31. Our study was conducted during the beginning of winter in the SO and the STF, when the remainder of the biomass produced during the light season is consumed and, most likely, allocated to energy storage in the form of lipids. This could explain the high relative abundance of lipid biosynthesis pathways in the bacterial community at the higher latitude (the SO and STF; Fig. 3 K, L). North of the STF in the tropical region the seasons are not as distinguishable, and the conditions are oligotrophic (Fig. 1 B and C). This means that organic matter is likely to be rapidly recycled and taken up for cell growth (e.g. high NH4+ uptake in the SPSG see Raes, van de Kamp 17), which may explain the rapidly declining lipid and carbohydrate biosynthesis pathways observed north of the STF (Fig. 3 K, L). Another possible explanation for the declining trend in lipid biosynthesis pathways might be related to the bioavailability of PO43- 32. It has been shown that, in PO43--deficient environments (such as the SPSG; Fig. 1C), heterotrophic bacteria and photosynthetic prokaryotes (picocyanobacterial) are able to engage in lipid remodelling (substituting phospholipids with alternative, non-phosphorus lipids, such as sulfolipids or glycolipids28), a strategy which increases their survival at an evolutionary scale in oligotrophic areas of the ocean 28, 33. As this lipid remodelling is expressed at a community level, the shift in trends of the metabolic pathways might give insight regarding how bacterial communities cope with PO43--limitation in the South Pacific Ocean.
Our last hypothesis suggested that we would detect more degradation-type pathways during the onset of winter in the SO, given that it has been previously shown that the SO is a region of high nutrient recycling rates and breakdown of organic matter in winter (e.g., measurements of high nitrification rates 17 ). Relative to the productive STF we measured lower inorganic C assimilation rates in the Southern Ocean, during the onset of the darker winter months (Fig. 1 B). Our result also supports those of Manganelli, Malfatti 34, who concluded that bacteria and archaea are the most important producers of organic particles via heterotrophic production (organic degradation) when light availability is reduced at higher latitudes during the winter. Overall, these results independently confirm that C-based degradation pathways are indeed key functions in the Southern Ocean during winter.
Other Important pathways
The SO is also a hotspot for sulphur cycling, in particular the production of dimethylsulfoniopropionate (DMSP; as shown by the high presence of Phaeocystis in Raes, Bodrossy 7 and Sow, Trull 35)) and the climate cooling dimethylsulfide gas (DMS; Berresheim 36 and Sheehan and Petrou 37). Recent work by Landa, Burns 38, who surveyed 1.4 million bacterial genome equivalents from the Tara Ocean’s dataset found that 1 in 5 of those genomes have the capacity to use DMSP. Members of the SAR11 and Planktomarina genera are known DMSP degraders ; they were dominant bacteria in our data set in the SO 7, explaining the higher relative abundance and presence of sulfur metabolising genes in the SO (Supplementary Fig. 4 A). Sulphur pathways declined significantly north of the STF but were still detectable up to the equator (Supplementary Fig. 4 A). These findings are in agreement with the study from Landa, Burns 38, which claims that a large range of marine bacteria have the capacity to use dissolved organic sulfur metabolites, and that the sulphur metabolites play an important part in the global pelagic ocean.
Fermentation pathways were included in the ten core ecosystem functions, which showed the highest relative abundances in the SO but were present across the entire transect (Supplementary Fig. 2). The anaerobic degradation of organic material (including fermentation) contributes significantly to the degradation processes in (marine) sediments 39. Because fermentation is favoured under anoxic environments, studies targeting the potential of this process in the photic zone are absent to the best of our knowledge. Anaerobic N-cycling processes such as denitrification and anammox have been shown to occur in anoxic and suboxic marine aggregates in oxygenated waters of the photic zone 40, 41, 42. These microhabitats offer niches for a diverse range of metabolic pathways 43, and the anoxic zones within marine snow particles could potentially harbour fermentative bacteria. We do note that the indication of fermentation pathways, however, could also be an artefact due to the presence of inactive sulphate-reducing bacteria and methanogenic archaea, which are capable of fermenting under favourable environmental conditions 39.
Considerations on the use of 16S rRNA sequencing for inferences on microbial functional ecology
We acknowledge that 16S rRNA metabarcoding is a broad-brush approach with a number of limitations for drawing conclusions about metabolic activity. Douglas, Maffei 15 clearly noted two main criticisms on functional estimates based on 16S rRNA amplicon-based hidden-state predictions. The first is that the predictions are obviously biased towards the available reference genomes, a limitation which will be partially addressed as the number of metagenome-assembled genomes (MAGs), and sequenced genomes in general, continues to increase. The second criticism is that the 16S rRNA-based predictions do not provide the necessary resolution to detect biogeographic pattern of ecotypes of interest, such as shown by Brown, Lauro 44 for the pelagibacter SAR11.
We should note two examples from our results that clearly illustrate these limitations of amplicon-based functional estimates. First, the metabolic pathways N2 fixation and nitrification, which have been shown to be important in the South Pacific Ocean17, were not statistically important, and therefore not present, in the PICRUSt2 MetaCyc outputs. This is likely because N2 fixation is not well resolved by 16S rRNA tag sequencing (e.g., Gaby and Buckley 45) and because bacteria involved in nitrification made up only 1% of the bacterial biomass (see supplementary Fig. 9 in Raes, van de Kamp 17). Neglecting these metabolic pathways that contribute to new (N2 fixation) and regenerated (nitrification) inputs of N will contribute to high uncertainties when estimating the f-ratio 46, 47, and thus, consequently lead to an underestimation when modelling global oceanic primary productivity. Secondly, our analyses do not provide the necessary resolution to detect biogeographic patterns of ecotypes of interest. The pelagibacter SAR 11, the cyanobacteria Prochlorococcus and the prymnesiophyte Phaeocystis, to name some examples, show ecological and evolutionary distinct ecotypes that are vertically and horizontally partitioned throughout the water column 35, 44, 48, 49. While these ecotypes appear functionally redundant in a broad, amplicon-based functional analysis, the fine-scale metabolic variations that have evolved among these ecotypes may have important bearing on the temporal and spatial structure of the community and productivity of the ecosystem.
The limitations noted above can, however, be addressed with additional existing tools in molecular ecology. For example, using this dataset, it was possible to combine amplicon-based analysis of the functional gene associated with N2 fixation (nifH) and direct rate measurements to reveal biogeographic patterns in the capacity and occurrence of these metabolic pathways17. Similarly for this dataset, focused taxonomic analysis of the 18S rRNA data revealed fine-scale ecotype distributions of the haptophyte Phaeocystis sp. 35
As with any endeavour in science, the tool a researcher chooses to use depends on the question being asked. Within the toolbox available to molecular ecologists, amplicon-based functional derivation 1) uses low cost, high throughput sample collection and standard analyses and 2) yields the information regarding ecosystem function that is required to construct biogeochemical and ecological models. Amplicon-based functional estimates yield functional information at the same breadth (i.e. basin-scale) and depth (i.e. a wide range of metabolic functions) as global ocean monitoring campaigns such as GO-SHIP and bioGEOTRACES. Mapping ecosystem function at this scale provides the opportunity to investigate the relationship between biodiversity and functional diversity, whether diversity drives productivity or the other way around, and whether biochemical resource limitation sets the ultimate control on productivity and ecosystem resilience. On a geological timescale, an analysis of 13-million-year-long nanoplankton abundance time series suggests that ecological functions, rather than species richness, are more important to community resilience and biochemical functions 50. Deriving functional profiles from 16S rRNA datasets obtained by oceanic sampling programs on a global scale may provide a better understanding of the components of a resilient marine ecosystem and of how that resilience is tested through existing and emerging environmental stressors.