Efficient CO2 fixation by surface Prochlorococcus in the Atlantic Ocean

Nearly half of the Earth's surface is covered by the ocean populated by the most abundant photosynthetic organisms on the planet—Prochlorococcus cyanobacteria. However, in the oligotrophic open ocean, the majority of their cells in the top half of the photic layer have levels of photosynthetic pigmentation barely detectable by flow cytometry, suggesting low efficiency of CO2 fixation compared with other phytoplankton living in the same waters. To test the latter assumption, CO2 fixation rates of flow cytometrically sorted 14C-labelled phytoplankton cells were directly compared in surface waters of the open Atlantic Ocean (30°S to 30°N). CO2 fixation rates of Prochlorococcus are at least 1.5–2.0 times higher than CO2 fixation rates of the smallest plastidic protists and Synechococcus cyanobacteria when normalised to photosynthetic pigmentation assessed using cellular red autofluorescence. Therefore, our data indicate that in oligotrophic oceanic surface waters, pigment minimisation allows Prochlorococcus cells to harvest plentiful sunlight more effectively than other phytoplankton.


Introduction
Oceanic phytoplankton constitute only a minute fraction of the photosynthetic biomass on Earth (Falkowski, 2012), but they contribute almost half of the yearly global net primary production Field et al., 1998;Behrenfeld et al., 2001). Prochlorococcus (Pro) cyanobacteria are the most abundant phytoplankton in the ocean, inhabiting nutrient-depleted environments from the equator up to 40-501 of latitude (Campbell and Vaulot, 1993;Zubkov et al., 2000). However, in surface waters the cellular pigment concentration of Pro is very low (B0.1 fg divinyl chlorophyll per cell), and consequently their carbon to chlorophyll ratios are very high (93.3-122 mg C per mg chlorophyll-a) in comparison with Synechococcus cyanobacteria and small picoeukaryotes (38-58 mg C per mg chlorophyll-a) (Veldhuis and Kraay, 2004;Bouman et al., 2006).
From their discovery onwards, Pro cells were almost exclusively enumerated by flow cytometry using their red autofluorescence and small size for identification (Chisholm et al., 1988;Partensky et al., 1999). Cellular red autofluorescence and chlorophyll pigment content are directly linked (Moore et al., 1995;Cavender-Bares et al., 1999;Dusenberry et al., 2001), although the relation may not always be linear and can be affected by pigment 'packaging' or variations in accessory pigments predominantly for larger, more heterogeneous cells (Sosik et al., 1989). Low red autofluorescence of Pro cells hampered their flow cytometric enumeration Dusenberry and Frankel, 1994) and technical improvements such as tightening of the beam spot, broadening of laser excitation, the introduction of quartz flow cells with integrated lenses as well as the reduction of the sheath pressure were used to detect low pigmented Pro cells Cavender-Bares et al., 1998). However, in the upper 40-80 m of stratified oligotrophic oceanic waters, it has remained a challenge to enumerate Pro cells unambiguously using their red autofluorescence (Chisholm et al., 1988;Zubkov et al., 1998;Partensky et al., 1999;Ting et al., 2002;Malmstrom et al., 2010). In such surface waters, the flow cytometrically visible Pro population is typically shaped in the form of a 'rising sun' emerging out of the background noise on a scatter plot of red autofluorescence (4650 nm) vs 901 side light scatter (Charles et al., 2005 and Supplementary Figure S1). In some studies, an extrapolation based on the visible part of the Pro population was applied to correct for the missing part (see, for example, Partensky et al., 1996). Nucleic acid staining of bacterioplankton was proposed as an alternative solution to the Pro cell detection problem (Zubkov et al., 2000). Later molecular identification confirmed that a distinct population observed following nucleic acid staining consists mainly of Pro cells (Zubkov et al., 2007;Mary et al., 2008;Gomez-Pereira et al., 2013). Direct comparison of the two approaches for Pro enumeration, that is, using red autofluorescence or cellular DNA-content/light scatter for identification, showed that in surface waters, up to a half of the Pro cells are unaccounted when red autofluorescence is used as the sole identifier (Zubkov et al., 2000).
High irradiance and resulting bleaching of photosynthetic pigments alongside potential nutrientlimitation in surface waters of oceanic subtropical gyres are common explanations for almost colourless Pro cells in surface ocean waters. Because in deeper (480 m) nutrient-replete but less illuminated parts of the water column, Pro pigmentation is much more intense (Partensky et al., 1999), the CO 2 uptake studies have focussed on Pro living in deeper waters (Chisholm et al., 1988;Li, 1994).
We put to the test two alternative explanations that account for the extremely low photopigmentation of Pro cells in the nutrient-depleted surface waters: (1) pigment levels are decreased as a result of the combined effects of high irradiance and nutrient limitation that diminishes capacity of the cells to cope with this stress, or (2) constitutively low cellular levels of photosynthetically active pigmentation are adequately physiologically balanced for these environmental conditions. In the former case, Pro red autofluorescence-normalised CO 2 fixation should be low compared with other phytoplankton, whereas red autofluorescence-normalised CO 2 fixation of Pro cells will be comparable with, or higher than, red autofluorescence-normalised CO 2 fixation by other phytoplankton cells if the latter were true.
Here, we present direct experimental evidence that red autofluorescence-normalised CO 2 fixation of surface Pro is high compared with the smallest eukaryotic phytoplankton and Synechococcus cyanobacteria. These results demonstrate that the dim red autofluorescence of surface Pro does not prevent them attaining high CO 2 fixation rates across the Atlantic Ocean.

Sampling
Pre-dawn seawater samples were collected from 20 m depth in 20 l Niskin (Miami, FL, USA) bottles attached to a standard conductivity-temperaturedepth profiler on the 20th cruise of the Atlantic Meridional Transect programme aboard the UK Royal Research Ship James Cook in October-November 2010 (Supplementary Figure S2). Seawater content of the entire Niskin bottle was decanted into an acid-rinsed polycarbonate carboy. To prevent exposure of photosynthetic cells to artificial light on board, the carboy was covered completely with two layers of dark plastic. Samples were processed immediately after collection. The sampling depth was chosen because it reflects the surface mixed layer, and the influence of ship movement and contaminants at that depth are minimal. At selected stations (indicated in Supplementary Figure S2) additional samples were taken from the bottom of the thermocline in order to compare CO 2 fixation rates of deeper vs surface phytoplankton communities.
Abundance measurements and definition of regional boundaries Concentrations of the Synechococcus (Syn) and Pro cyanobacteria were determined in unstained, fixed (1% paraformaldehyde, final concentration; Sigma-Aldrich, Hamburg, Germany) samples according to Olson et al. (1993) using a FACSort flow cytometer (Becton-Dickinson, Oxford, UK). Pro cells were counted in both unstained (Pro unst ) and stained fixed samples (Pro st ) on the basis of their red autofluorescence and 901 side light scatter, and their nucleic acid content (green fluorescence) and 901 side light scatter, respectively. Subsamples for Pro st counting were taken from a 20 l carboy and fixed with 1% paraformaldehyde for 1 h in the dark at room temperature and stained with SYBR Green I dye (Sigma-Aldrich) (Marie et al., 1997). Cellular abundances of small (o2 mm) and large (2-5 mm) plastidic eukaryotes (Plast-S and Plast-L, respectively) were determined in parallel from the same sample. Before flow cytometric analyses, a mixture of 0.5 and 1.0 mm multi-fluorescent beads (Polysciences, Eppelheim, Germany) at a calibrated concentration (Zubkov and Burkill, 2006) was added to both stained and unstained samples. The beads were used as an internal standard for calculating absolute cell concentrations and for normalising cellular red autofluorescence.
Catalysed reporter deposition fluorescence in situ hybridisations (CARD-FISH) on flow cytometrically sorted cells In order to confirm that the distinct, high-nucleic acid bacterial population observed by flow cytometry (Supplementary Figure S1) consists mainly of Pro cells, CARD-FISH hybridisations using the Pro-specific probe PRO405 (West et al., 2001) were carried out on sorted cells at selected stations covering each province (NG, EQ and SG). The in silico specificity of the probe was re-evaluated by running TestProbe (part of the Silva online software packages, www.arb-silva.de; Quast et al., 2013) against the Silva SSU r117 reference database. In addition, the Eubacteria-targeted probe mix Eub338I-III (Amann et al., 1990;Daims et al., 1999) was used to determine overall hybridisation efficiency. The details of contamination-free flow sorting of target cells and CARD-FISH are described in Gomez-Pereira et al. (2013). We analysed seven stations in the NG, three stations in the EQ and four stations in the SG.
Total and cell-specific CO 2 fixation Before each experiment, 60 ml Pyrex glass bottles (Fisher Scientific, Loughborough, UK) were acidsoaked overnight (10% HCl) and rinsed twice with 30 ml sample sea water. After washing, 60 ml of seawater sample was added to each bottle and spiked with trace metal-clean 14 C radiolabelled sodium bicarbonate (34.66 mM NaH 14 CO 3 ; DHI, Hørsholm, Denmark). Samples were then incubated at ambient temperatures (regulated by a refrigerated water bath (Grant Instruments, Shepreth, UK)) in a 6 l water tank illuminated by a warm white light-emitting diode array (Photon Systems Instruments, Drasov, Czech Republic) adjusted to a constant output of 500 mmol photons m À 2 s À 1 . The chosen light intensity equals half the irradiance reaching the water surface at noon-time in the equatorial region (Jitts et al., 1976), because on average at 20 m depth the light intensity is reduced by 33-55%. Moreover, no photoinhibition occurs at this light intensity (Morel et al., 1996). In contrast to incubations at ambient light, the constant light output made it possible to compare CO 2 fixation rates at different stations.
Two different concentrations of NaH 14 CO 3 were used to determine total CO 2 fixation during a time series and to measure CO 2 fixation of flow cytometrically sorted phytoplankton populations, respectively. Time series were carried out to ensure linear uptake of label and to guarantee that the small volumes of sorted cells are representative of the whole community (Supplementary Figure S3). Subsamples of 1.6 ml were taken at 0 and 10 h for flow cytometric analyses to ascertain that community composition remained unchanged for the duration of the experiment (Supplementary Figure S4).
For time series CO 2 fixation measurements, 3.7 kBq ml À 1 NaH 14 CO 3 was added to 60 ml seawater sample. In total six Pyrex glass bottles were prepared. Five bottles were incubated for 2, 4, 6, 8 and 10 h in the light, and the remaining bottle was incubated for 10 h in the dark. At each discrete time point, the whole sample was fixed by adding 1% paraformaldehyde (final concentration) and incubated for 1 h at room temperature. Subsequently, the complete sample was filtered onto a 0.2 mm polycarbonate filter (Nuclepore, Whatman, Little Chalfont, UK), washed three times with ultra-clean water (MQ system, Millipore, Whatman, Walford, UK) and placed in a scintillation vial. Before addition of 5 ml scintillation cocktail (Goldstar, Meridian, Epsom, UK), 1 ml of 10% HCl was added, the vial gently swirled and incubated for 10-30 min to fume out non-incorporated NaH 14 CO 3 . Dark CO 2 fixation rates were o3% of paired CO 2 fixation rates in the light in all experiments (Supplementary Figure S3).
To determine group-specific CO 2 fixation rates, higher NaH 14 CO 3 concentrations had to be used because of the small size of the organisms. To 60 ml seawater sample in a Pyrex glass bottle, 246 kBq ml À 1 NaH 14 CO 3 was added, the sample incubated for 10 h and then fixed with 1% paraformaldehyde (final concentration). Three 1.6 ml subsamples were taken directly to determine total CO 2 fixation and to sort Pro st . In order to sort adequate cell numbers of cyanobacteria (Pro unst and Syn), 20 ml of the sample was concentrated on a 0.6 mm polycarbonate filter (Nuclepore, Whatman) mounted in a filtration unit (Swinnex, Millipore) using a syringe pump (KD Scientific, Holliston, MA, USA) at a flow rate of 2.5 ml min À 1 . This pore size was selected as it was shown in an earlier publication (Zubkov et al., 1998) that the Pro population with visible red autofluorescence had a cell diameter of 0.63 ± 0.03 mm. Moreover, similar 901 side light scatter values of Pro unst before and after concentration suggest that there is no selective enrichment of larger cells because of the concentration procedure (Student's t-test, P ¼ 0.871, Supplementary Figure S5). The remaining sample was concentrated on a 0.8 mm polycarbonate filter (Nuclepore, Whatman) the same way to enrich eukaryotic phytoplankton. Apart from the 0.6 mm concentrated fraction, all samples were stained with SYBR Green I (Marie et al., 1997), stored at 4 1C and sorted flow cytometrically within 10 h.
In order to determine the influence of nutrients on CO 2 fixation, at three stations (NG and SG) a parallel incubation was carried out where 2.6 ml of nutrientenriched sea water from 300 m depth was added to 60 ml of seawater sample from 20 m to simulate a mixing event (Supplementary Figure S2). The experiment was run in parallel to our standard 20 m incubations for 10 h under the same light regime. The sample was processed as described above for the standard incubations. The nutrient addition corresponds to an B20-fold increase in nitrite/nitrate concentration (0.03 and 16.83 mmol l À 1 average ambient concentration at 20 and 300 m, respectively; Harris and Woodward, 2014) along the whole transect. At four stations (Supplementary Figure S2, EQ and SG), samples from the bottom of the thermocline were incubated in parallel at the same light intensity as surface samples to compare CO 2 fixation rates of phytoplankton groups living at the two depths.
Flow cytometric sorting Different phytoplankton populations were sorted according to light scattering properties (901 or side light scatter), relative concentration of SYBR Green I stain per particle (green fluorescence; FL1, 530 ± 30 nm), phycoerythrin content (orange fluorescence; FL2, 580 ± 30 nm) and chlorophyll content (red fluorescence; FL3, 4650 nm) using a FACSort instrument (Becton-Dickinson). Because of their low pigmentation in surface waters, we used two approaches to sort Pro cells. A distinct bacterial population, verified to be mainly Pro by CARD-FISH (486% , Table 1), was sorted from unconcentrated, SYBR Green I stained samples (Pro st ) according to side scatter and green fluorescence properties. In addition, Pro was sorted according to red autofluorescence from 0.6 mm concentrated unstained samples (Pro unst ), as earlier studies indicated a cell diameter of 40.6 mm (Chisholm et al., 1988;Vaulot et al., 1990;Zubkov et al., 2000). From the same sample, Syn cells were sorted according to their phycoerythrin content. Plast-S and Plast-L populations were sorted from 0.8 mm concentrated, stained samples using side scatter, SYBR Green I stain and red autofluorescence as defining parameters. For each population, 4-6 replicates of different cell numbers were sorted. Bacterial and eukaryotic cells were collected on 0.2 and 0.8 mm polycarbonate filters, respectively, and treated following the same procedure as for total CO 2 measurements (see above) before counting. Radioassaying of samples was carried out using an ultra-low-level liquid scintillation counter (1220 Quantulus, Wallac, Waltham, MA, USA).

Cell biomass estimation of Pro and other phytoplankton
Cell diameters of Pro st and Pro unst surface populations were determined on Atlantic Meridional Transect (AMT)-4 at 11 stations spanning NG, EQ and SG regions using a size fractionation method (Zubkov et al., 2000). Briefly, cell concentrations were measured in unfiltered samples and filtrates after filtering samples through polycarbonate filters (Nuclepore) of different pore sizes. The filter pore size versus the percentage of cells in the corresponding filtrates relative to cell concentration in the unfiltered sample were plotted to estimate the pore size that would retain 50% of cells. That pore size was interpreted as a mean cell diameter. Average cell diameters of Pro st (0.52 ± 0.03 mm, n ¼ 30) and Pro unst (0.6±0.05 mm, n ¼ 35) were significantly different (t-test, Po0.001). Mean cell biovolumes were calculated assuming a spherical cell shape. For Pro and Syn cells, conversion factors of 184 and 211 fg C mm À 3 (Heldal et al., 2003) and cell diameters of 0.52 ± 0.03 and 0.95 ± 0.31 mm (Zubkov et al., 2000) were applied. Details of all conversion factors used in this study can be found in Table 2. A conversion factor of 200 fg C mm À 3 (Waterbury et al., 1986) was used to calculate biomass-specific CO 2 fixation rates assuming spherical cell shape and average cell diameters of 2.0 ± 0.1 and 3.1 ± 0.3 mm for Plast-S and Plast-L cells, respectively (Hartmann et al., 2012). These cell diameters were established for all studied regions on two consecutive AMT cruises, including AMT-20 where the here presented CO 2 fixation rates were measured.

Data analyses
Cell-specific CO 2 fixation rates were determined from average per cell values of each of the sorted replicates and converted to fg C cell À 1 h À 1 according to (Parsons et al. (1984). Statistical analyses were carried out using SigmaPlot (London, UK). In case of normal distribution and equal variance, t-tests were carried out for comparison. If the data were nonnormally distributed or the equal variance test failed, Mann-Whitney rank-sum tests were used.

Results
Determination of Pro cell abundance and CO 2 fixation rates Because of the low red autofluorescence of Pro cells in surface waters and the resulting unreliability of detection, we focussed on a distinct population within the bacterioplankton based on flow cytometrically determined cellular nucleic acid content and 901 light scatter, called Pro st (Supplementary Figure S1; Zubkov et al., 2000). The taxonomic identity of cells within this population was verified by flow sorting followed by Card-FISH using a Pro-targeted probe (Pro405, West et al.2001) at 14 stations along the AMT (Table 1). The majority of flow sorted cells (86-94%) hybridised with the Pro-specific probe, corroborating the results of previous molecular studies of the same distinct population (Table 1; Zubkov et al., 2007;Mary et al., 2008;Gomez-Pereira et al., 2013). As a control, Pro cells were in parallel enumerated on the basis of red autofluorescence in unstained samples (Pro unst , Supplementary Figure S1) and compared with Pro st cell numbers revealing a significant Table 1 Identification of a flow cytometrically sorted, distinct population with high-nucleic acid content as Prochlorococcus cyanobacteria (Pro) in different regions of the Atlantic Ocean using Pro-targeted CARD-FISH (Pro405)
There could be two reasons why the Pro unst cells systematically fixed 50% more 14 CO 2 than the Pro st cells: 1. The presence of non-photosynthetic bacteria among the sorted Pro st cells could lower the average cellular 14 C content (because the total measured 14 CO 2 fixation was divided by the total number of sorted cells). The cells among the sorted Pro st populations not hybridising with the Pro-targeted probe could either be Pro cells with ribosomal contents below detection level of FISH (for example, Pro cells were compromised or dead) or they could be by-sorted non-Pro cells displaying similar DNA fluorescence and side light scatter properties. A small proportion of sorted cells (5-6%) could not be detected by FISH with the universal bacterial probe (Table 1), and this lends support to the former explanation. However, even in the extreme (assumed) case if all probe-negative cells were by-sorted non-Pro cells, one could explain a discrepancy of only 6-14% between Pro st and Pro unst rather than the measured 50% difference. 2. On average, larger Pro unst cells that contain proportionally higher chlorophyll amounts resulting in detectable red autofluorescence could fix more 14 CO 2 than smaller Pro st cells.
Indeed, cell diameter estimates carried out on an earlier AMT cruise (AMT-4) using size fractionation revealed that Pro st cells were systematically smaller than Pro unst cells across the Atlantic Ocean (Table 2; Zubkov et al., 2000). These observations suggest that only larger or dividing Pro cells with higher red autofluorescence were visible above the threshold of the red fluorescence photomultiplier detector. Indeed, the abundance of Pro unst cells was significantly lower than that of Pro st cells (on average 58 ± 18%, Figure 1a). Consequently, the Pro unst cells are not representative of the entire Pro population in surface waters. Sorting of Pro unst would therefore lead to overestimation of the actual cellular CO 2 fixation rates by Pro, and we used hereafter Pro st cells for more realistic measurement of CO 2 fixation by Pro cells.
The Pro cell-specific CO 2 fixation rates in the NG and SG were similar, but their rates were more than doubled in the EQ (t-test, P ¼ 0.002; Supplementary Figure S6 and Supplementary Table S1). The CO 2 fixation rates of Plast-L cells followed the same pattern (t-test, Po0.001), showing increased rates in the EQ (Supplementary Figure S6 and Supplementary  Table S1). In contrast, Plast-S as well as Syn cells showed similar rates in the NG, SG and EQ, but significantly lower CO 2 fixation rates in the ST (t-test, P ¼ 0.042 and P ¼ 0.032, respectively, Supplementary Figure S6 and Supplementary Table S1).
Comparison of cellular, red autofluorescencenormalised and biomass-specific CO 2 fixation of Pro, Syn and small eukaryotic phytoplankton Although cellular CO 2 fixation rates positively correlated with cell sizes (R 2 ¼ 0.83, Supplementary Figure S6), no such relationship was found between red autofluorescence-normalised or biomass-specific CO 2 uptake (see Table 2 for details on biomass in this  Supplementary Figures S1b and d) were used for normalisation. Consequently, the derived normalised values for Pro should be treated as conservative lower estimations. Red autofluorescence-normalised CO 2 fixation shows that Pro despite an order of magnitude lower red autofluorescence than Syn (Table 2) can fix up to 4 times more CO 2 than other small phytoplankton (t-test, Pp0.005, Figures 2a and b) whereas CO 2 fixation rates of Syn and plastidic eukaryotes are comparable. Biomass-specific CO 2 fixation rates of plastidic eukaryotes were significantly lower than those of cyanobacteria (Mann-Whitney, Po0.001) whereas Syn showed on average 60% higher biomassspecific CO 2 fixation than Pro (t-test, P ¼ 0.04; Figures 2c and d). Because of the combined effect of comparatively high CO 2 uptake rates and high abundance, Pro led microbial CO 2 fixation across the low-latitude Atlantic Ocean (Mann-Whitney, Pp0.038; Figures 2e and f).
To assess whether cellular CO 2 fixation and red autofluorescence of surface Pro could be influenced by a lack of nutrients, additional experiments with added nutrients were performed. Nutrient addition to surface samples in the form of deep water (300 m depth) had no significant effect on CO 2 fixation rates of either Pro or Syn cells (t-test, P40.5, Figure 3) and did not influence red autofluorescence of Pro (t-test, P ¼ 0.12). Moreover, comparisons of CO 2 fixation rates of Pro populations from surface waters and deeper water (bottom of the thermocline) revealed no significant differences in CO 2 fixation rates when exposed to the same light conditions, despite stronger red fluorescence of deeper Pro cells (Figure 4).

Discussion
Direct determination of group-specific CO 2 fixation rates using 14 C-tracer is technically challenging and has been attempted only in three other studies so far (Chisholm et al., 1988;Li, 1994;Jardillier et al., 2010) that were spatially restricted to small areas of the Atlantic Ocean. This new data set provides for the first time insight into CO 2 fixation rates of four distinct phytoplankton groups across the Atlantic Ocean. Cellular CO 2 fixation rates measured in the equatorial region are comparable to those measured in the North East Atlantic (Li, 1994;Jardillier et al., 2010;Supplementary Table S3). Our slightly lower values can be most likely attributed to the differences in light regime between the studies (artificial vs ambient light). Inferred from estimates of diel synchronised cell division in the photic layer (Vaulot et al., 1995) and measurements of CO 2 fixation by Pro inhabiting the deeper waters (Chisholm et al., 1988), Pro could contribute more than a half to the total CO 2 fixation in the lowlatitude Ocean. These estimates concur with our results from surface waters (Figures 2e and f).
A combination of high sunlight irradiance (up to 3000 mmol photons m À 2 s À 1 ; Jitts et al., 1976), slow rates of vertical mixing, low inorganic nutrient availability and potential preferential grazing pressure by mixotrophic protists  creates a harsh habitat for Pro cyanobacteria in surface waters of the low-latitude Atlantic Ocean. How do Pro cells remain numerous, effective CO 2 fixers in these waters (Figure 2) with virtually undetectable red cellular autofluorescence (indicative of extremely low photosynthetic pigmentation)?
Biosynthesis of photosynthetic pigments like chlorophyll requires certain inorganic nutrients, for example, nitrogen and iron, that are depleted in the subtropical gyres, and constrained nutrient bioavailability can lead to reduced cellular pigmentation (Riemann et al., 1989;Staehr et al., 2002). However, both plastidic protists and Syn cyanobacteria can easily be detected in the same waters by their photosynthetic pigmentation, suggesting that the required nutrients are still bioavailable to those cells. Furthermore, CO 2 fixation rates of Pro remained unchanged in our deep water addition experiments to simulate mixing events, suggesting that Pro are not nutrient limited, at least with regard to CO 2 fixation (Figure 3). These findings are in accordance with a study in the equatorial Pacific Ocean (Vaulot et al., 1995), where close to maximal (that is, nutrient unlimited) growth rates of Pro were estimated. In addition, Pro cells populating the deeper parts of the mixed layer, where nutrients are still scarce but irradiance is less intense, exhibit stronger red autofluorescence (Zubkov et al., 1998). Figure 2 Detailed (a, c, e) and average (b, d, f) red autofluorescence-normalised, biomass-specific and population-specific CO 2 fixation of Pro in comparison with Syn cyanobacteria and smaller and larger plastidic eukaryotes (B2 mm, Plast-S and B3.1 mm, Plast-L) in the Atlantic Ocean (EQ, equatorial region; NG, Northern Gyre; ST, Southern temperate waters; SG, Southern Gyre and). Units on the y-axes are the same for (a and b, c and d and e and f). Student's t-test confirmed significant differences in biomass-specific CO 2 fixation between all phytoplankton groups (P ¼ 0.001-0.047, see Result section for details).
Perhaps Pro cells produce only low amounts of photosynthetic pigmentation in order to achieve efficient CO 2 fixation with minimal effort using light energy for photosynthesis as well as for photoheterotrophy, that is, redirecting a part of the collected light energy for actively importing organic molecules (Casey et al., 2009;Zubkov, 2009). This would allow them to compete with Syn cells that spend considerable energy on the production of photoprotective pigments (Raven, 1991;Aráoz and Häder, 1999). Molecular studies of cultured Syn and Pro showed that the Pro response to photo-damage is modelled to minimise energy demand, for example, during high irradiance periods of the day the main metabolic processes are downregulated (Mella-Flores et al., 2012). Lower biomass-specific CO 2 fixation rates of Pro in comparison with Syn (Figures 2c and d) are, perhaps, a price worth paying to avoid constant repair of photo-damaged reaction centres in highly irradiated surface waters. That might be a reason (additional to inorganic nutrient limitation (Tarran et al., 1999;Vaulot et al., 1996)) for low concentrations of Syn cells in oligotrophic waters despite their higher biomass-specific CO 2 fixation rates. On the other hand, the difference in biomass-specific CO 2 fixation rates between Pro and Syn is comparatively small given the higher pigment content, up to 10 times at 1000 mmol m À 2 s À 1 light in cultures (Moore et al., 1995) and according to red autofluorescence (Figures 2a and b) of the latter.
High red autofluorescence-normalised CO 2 fixation by Pro is likely related to the high geometrical absorption cross-section owing to their small cell size (Morel et al., 1993;Bailey et al., 2005) and package effect that states that the light-harvesting to the effect of increased pigmentation is reduced because of a parallel decrease in the absorption cross-section (Dubinsky et al., 1986;Berner et al., 1989). Light harvesting efficiency of Pro is further increased by the unique pigment, chlorophyll a 2 (Chisholm et al., 1988;Goericke and Repeta, 1992) with absorption maximum that coincides with the wavelength of higher energy blue light. The cumulative outcome of these numerous adaptations has enabled Pro cells with minimal photosynthetic pigmentation to become highly efficient CO 2 fixers (Figure 2).

Conclusion
The results presented here demonstrate that Pro are highly efficient CO 2 fixers in surface waters of the Atlantic Ocean and their red autofluorescencenormalised CO 2 fixation rates are higher than those of Syn and small plastidic protists. These findings indicate that Pro should be specially taken into account when photosynthetic pigmentation data are used for deducing biological CO 2 fixation in the oligotrophic open ocean. Figure 4 Average biomass-normalised CO 2 fixation of Pro cells in surface waters (that is, 20 m depth, white) and deeper water layers (grey) in the Equatorial region (EQ) and the Southern gyre (SG). Numbers on x-axis indicate depth of the deep water sample.

Figure 3
Average per cell CO 2 fixation of surface Pro (empty) and Syn (pattern) without (white) and with addition of nutrients (grey, that is, addition of sea water from 300 m depth). NG, Northern Gyre; SG, Southern Gyre.