Niche partitioning by photosynthetic plankton as a driver of CO2-fixation across the oligotrophic South Pacific Subtropical Ocean

Oligotrophic ocean gyre ecosystems may be expanding due to rising global temperatures [1–5]. Models predicting carbon flow through these changing ecosystems require accurate descriptions of phytoplankton communities and their metabolic activities [6]. We therefore measured distributions and activities of cyanobacteria and small photosynthetic eukaryotes throughout the euphotic zone on a zonal transect through the South Pacific Ocean, focusing on the ultraoligotrophic waters of the South Pacific Gyre (SPG). Bulk rates of CO2 fixation were low (0.1 µmol C l−1 d−1) but pervasive throughout both the surface mixed-layer (upper 150 m), as well as the deep chlorophyll a maximum of the core SPG. Chloroplast 16S rRNA metabarcoding, and single-cell 13CO2 uptake experiments demonstrated niche differentiation among the small eukaryotes and picocyanobacteria. Prochlorococcus abundances, activity, and growth were more closely associated with the rims of the gyre. Small, fast-growing, photosynthetic eukaryotes, likely related to the Pelagophyceae, characterized the deep chlorophyll a maximum. In contrast, a slower growing population of photosynthetic eukaryotes, likely comprised of Dictyochophyceae and Chrysophyceae, dominated the mixed layer that contributed 65–88% of the areal CO2 fixation within the core SPG. Small photosynthetic eukaryotes may thus play an underappreciated role in CO2 fixation in the surface mixed-layer waters of ultraoligotrophic ecosystems.


Sampling, hydrography and underwater light field
Sampling was carried out during the RV Sonne "UltraPac" cruise (SO245) from Antofagasta, Chile (17 December2015 A HyperPro II profiling system (SeaBird Scientific, former Satlantic) was used to collect hyperspectral underwater light field data in free-falling mode. The system consisted of one hyperspectral irradiance and one hyperspectral radiance sensor. A second hyperspectral irradiance sensor was mounted on the research vessel for matching above-water irradiance reference measurements. Profiles were conducted at each station depending on sea and weather conditions, with deployments at 50 m away from the ship to avoid ship shadow. All sensors were pre-calibrated by the manufacturer and validated prior to the cruise with a reference lamp.
Data were recorded with the SatView software (V 2.9.5_7), pre-processed from raw to Level 3a using the ProSoft Processing software (V 7.7.19_2) and binned in 1 m depth intervals. A dark correction was made automatically based on shutter measurements by the instrument. Postprocessing was made in accordance with (6,7). The photosynthetic active radiation (PAR) was integrated from 400 to 700 nm, considering the onboard reference measurements. The spectral types of 1% downwelling irradiance (Ed) were determined to compare penetration depth and wavelength bands for each station.

Nutrients, Chlorophyll a and Dissolved Organic Carbon and Nitrogen
Dissolved inorganic phosphorus (PO4 3-), nitrite (NO2 -) and nitrate (NO3 -) (and silicate (Si)) concentrations were measured with the QuAAtro39 autoanalyzer (Seal Analytical) according to the methods of Strickland and Parsons (8,9). Seawater low-nutrient standards (OSIL) were used as a secondary standard to test primary standard calibrations. Low with a pore size of 0.7 µm and stored in 10 mL polypropylene tubes at -20 °C in the dark until further processing. Chl a extraction and measurements were performed on board after (10,11).

CO2 fixation experiments
Rates of CO2 fixation, were determined at six depths over the upper water column (< 300 m) at stations, 1 (only two depths), 2, 4, 6, 8, 10, 12, 14, and 15. CO2 fixation rates were determined through a 24-hours-incubation of collected seawater with the addition of dissolved inorganic 13 C (DI 13 C). The 13 C method applied in this study is comparable in approach, and for oligotrophic regions, comparable in sensitivity to the classical 14 C method (13). Analogous to the 14 CO2 tracer method, which after more than a half-century of use and refinement is the basis for calibrating algorithms used to estimate surface productivity from satellite data (14), the 13 CO2 method yields rates lying between gross primary productivity (GPP) and net primary productivity (NPP) (14-16) and references therein). The difference between GPP and NPP depends on the rate of oxidation of fixed C back to CO2 via photorespiration and mitochondrial respiration by the CO2-fixing phototrophs during the course of the experiment. As net primary productivity is the measure of CO2 uptake that we wish to constrain, incubation times are set to 24 hours in order to allow phototrophic mitochondrial respiration to run its full course. As we are also interested in obtaining near instantaneous CO2 fixation rates, incubation times are as short as possible, thus limited to one 24-hour incubation cycle.
Incubations were performed in on-deck incubators that were kept at surface water temperature via seawater flow-through and adjusted for three light levels in order to bracket the light conditions experienced by phytoplanktion within the water column. High-light, photoinhibition effects can lead to an underestimation of CO2 fixation rate. Conversely, underillumination may lead to an overestimation, as isotope labeling methods 14 (17). For open ocean regions, they found that incubation derived CO2 fixation rates values accurately represented in situ rates for a freely mixed euphotic zone (17). Decades of evaluation have shown that when performed with care 14 C and 13 C incubation methods provide robust estimates of CO2 fixation into biomass in the euphotic zone, i.e., net primary productivity.
Incubations were performed in triplicates with the addition of DI 13 C (NaH 13 CO3, ≥98% Sigma-Aldrich) to ~5% DI 13 C. The natural abundance of 13  Additionally, caffeine standards were measured every 6 samples, in order to evaluate precision and drift, as well as providing standards for absolute C and N content.
CO2 fixation rates were calculated from the incorporation of DI 13 C into biomass using the following equation (18).
where % is the atom % 13 C in particulate organic carbon (POC); % is the atom percent of DI 13 C in the total DIC pool; and % % is the natural abundance of particulate C (atom %) in the control incubations that were simultaneously incubated with the other bottles with no added DI 13 C. At every IRMS run a pre-combusted GF/F filter was measured as a blank and samples were blank-corrected. Samples with POC/PON concentrations that were estimated to be too low to be precisely quantified by the EA-IRMS were spiked with 89.1 nmol of caffeine Accurate determinations of Prochlorococcus in high light intensity, oligotrophic environments such as the SPG are known to be highly problematic (19). Prochlorococcus populations with dim chlorophyll fluorescence were identified by gating on green fluorescence (FL1) and chl a fluorescence (FL3) after staining with SYBR Green (Molecular Probes S7585) ( Figures S4-S8). Prochlorococcus cell numbers were subsequently cross-checked with direct cell counts and onboard contemporaneous DNA sequence reads (2) ( Figure S12). Heterotrophic prokaryotes were determined in subsamples that had been stained with SYBR Green I Nucleic Acid Stain (Biosciences) for at least 15 minutes in the dark at room temperature.

Single-cell sample imaging
Catalyzed reporter deposition-fluorescent in situ hybridization (CARD-FISH) was used to identify Prochlorococcus cells according to (20). Filters were punched out into 5 mm circles The data was processed using Look@NanoSIMS software (22). For every measurement run the recorded secondary ion images of at least 10 of the measured planes were drift corrected and accumulated. Region of interests (ROIs) around cell structures were defined using the RGB combined image of 12 C 14 Nand 32 S -. For each ROI the 13 C/ 12 C ratios were calculated. The variability of all background values (average ± 3xSD; 1.13 atom %) of the dataset was used to estimate the detection limit for significant isotope enrichments. Poisson error across all planes was <5%. The natural abundance (NA= 1.1 atom %) of 13 C atom % was used from the EA-IRMS measurements (average from the respective samples measured with the nanoSIMS).

Single cell calculations
The cell size, as determined with Look@NanoSIMS software (22), was used to calculate biovolume (BV). For size determination of cells, ROIs were drawn after the first two planes were accumulated, as we have noticed that during the measurement that cell size is sometimes reduced as the beam removes parts of the cell.
Prochlorococcus and photosynthetic small eukaryotes were considered as a prolate spheroid and thus the following equation was used for BV calculations: where b is the width and a is the length of the cell.
The carbon (C) per cell was calculated based on the BV with the formula by (23)  Growth rate were estimated based on the incorporations of DI 13 C into biomass assuming exponential growth and an even distribution of the isotopes in the biomass during cell division: where % is the atom % excess over background in the DIC pool; % − is the atom % excess over background measured in the photosynthetic small eukaryote and Prochlorococcus cells; and is incubation time in days.

Contributions of small photosynthetic eukaryotes and Prochlorococcus to the bulk CO2
fixation were calculated with the median single-cell CO2 fixation rates of the respected group of organisms, station and depth multiplied by the abundance. For the rates measured at 20 m depth abundance the respective data taken from the surface to 40 m were used and for the chl a max abundance data within the 0.5 µg chl a l -1 isoline were used.  (36). We removed non-chloroplast sequences of the 16S rRNA metagenomic data, as well as sequences from the class of Embryophyceae (land plants), which were assumed to be contaminants. To retrieve only photosynthetic 18S rRNA sequence we filtered based on the results of the chloroplast 16S rRNA metagenomic approach.

petB Gene Distributions
In order to assess the distribution of Prochlorococcus ecotypes, we mapped the metagenomic reads from the metagenomes described above to a custom database of the petB gene, a high-resolution taxonomic marker for Prochlorococcus ecotypes (3). We also extracted reads from additional shotgun metagenomes (44)