Biological N2O Fixation in the Eastern South Pacific Ocean and Marine Cyanobacterial Cultures

Despite the importance of nitrous oxide (N2O) in the global radiative balance and atmospheric ozone chemistry, its sources and sinks within the Earth’s system are still poorly understood. In the ocean, N2O is produced by microbiological processes such as nitrification and partial denitrification, which account for about a third of global emissions. Conversely, complete denitrification (the dissimilative reduction of N2O to N2) under suboxic/anoxic conditions is the only known pathway accountable for N2O consumption in the ocean. In this work, it is demonstrated that the biological assimilation of N2O could be a significant pathway capable of directly transforming this gas into particulate organic nitrogen (PON). N2O is shown to be biologically fixed within the subtropical and tropical waters of the eastern South Pacific Ocean, under a wide range of oceanographic conditions and at rates ranging from 2 pmol N L−1 d− to 14.8 nmol N L−1 d−1 (mean ± SE of 0.522±1.06 nmol N L−1 d−1, n = 93). Additional assays revealed that cultured cyanobacterial strains of Trichodesmium (H-9 and IMS 101), and Crocosphaera (W-8501) have the capacity to directly fix N2O under laboratory conditions; suggesting that marine photoautotrophic diazotrophs could be using N2O as a substrate. This metabolic capacity however was absent in Synechococcus (RCC 1029). The findings presented here indicate that assimilative N2O fixation takes place under extreme environmental conditions (i.e., light, nutrient, oxygen) where both autotrophic (including cyanobacteria) and heterotrophic microbes appear to be involved. This process could provide a globally significant sink for atmospheric N2O which in turn affects the oceanic N2O inventory and may also represent a yet unexplored global oceanic source of fixed N.


Introduction
Nitrous oxide (N 2 O) is an important greenhouse gas that contributes to stratospheric ozone depletion. The transfer of this gas within the Earth System is intimately linked to physical and biological processes occurring in the ocean, atmosphere and soils. In the ocean, N 2 O saturation levels depend upon oceanographic variables such as temperature and salinity while processes producing N 2 O are mainly being controlled by organic matter and dissolved O 2 [1]. A detailed understanding of its biological production and consumption is necessary to predict the effects of long term changes in N 2 O on the Earth's climate [2].
According to current knowledge, nitrification performed by bacteria [3] and archaea [4], via aerobic NH 4 + oxidation or nitrifier denitrification (the pathway in which NH 4 + is oxidized to NO 2 2 followed by the reduction of NO 2 2 to NO, N 2 O, and N 2 [5]) is the main process responsible for most of the N 2 O production under oxic or even microaerophilic conditions; whereas partial denitrification, via the anaerobic reduction of NO 2 2 to N 2 O, can produce this gas under suboxic conditions [3,6].
Contrary to its production, N 2 O can only be consumed by photolysis in the stratosphere [7] and by canonical denitrification via dissimilative reduction of N 2 O to N 2 , a pathway that only exists under anoxic condition [3,6,8]. However, early studies suggest that N 2 O, and even NO 2 2 , could act as substrates for the enzymatic nitrogenase complex (NifH) [9,10]. Indeed, it is also known that given the properties of the NifH, certain diazotrophs are able to reduce not only N 2 , but also other multi-bonded substrates, such as acetylene, azide, cyanide, methyl isocyanide and even N 2 O [11,12]. Regarding N 2 O, this molecule could bind to metals belonging to the NifH complex through the N-bound nitro form and the N-O bond form [13]. However, there is no direct evidence so far that indicates how or where the N-O bond containing molecule N 2 O binds to NifH, or that indicates the chemical mechanism of its reduction.
The eastern South Pacific (ESP) region hosts the most extreme range of biogeochemical conditions in the global ocean, from the very eutrophic, nitrate-rich and oxygen-poor waters of the Peruvian and Chilean coastal upwelling (CU ,71u-76uW) to oxygenated and severely nutrient-limited waters of the central subtropical Pacific gyre (STG ,110uW) [14]. Dissolved O 2 vertical distribution denotes the well-known oxygen minimum zone (OMZ),which has relatively shallow suboxic-anoxic waters (0$ O 2 #11 mmol L 21 ) in subsurface waters off Peru and northern Chile [15] and also the oxygenated waters towards the subtropical gyre and south of ,37uS. N 2 O concentrations in seawater and its concomitant exchange across air-sea interface also reflect the previously mentioned biogeochemical gradients in the ESP. Thus, N 2 O content in the ESP can be as high as 400% saturation with strong effluxes within coastal upwelling and sub-saturated or slightly supersaturated N 2 O levels (,80-105%) with near zero airsea fluxes within the subtropical gyre (STG) [16,17]. A number of studies reports sub-saturated N 2 O values in surface waters throughout the world's major oceans (Table S1 and Text S1); these low values have been assumed to be analytical artifacts [18] given the absence of any known chemical or biological mechanism able to remove N 2 O from surface waters. In suboxic subsurface waters, in contrast, such as those found in the OMZ of the ESP (at depths from 50 to 400 m), sub-saturated N 2 O concentrations (as low as 40-60%) have been solely attributed to canonical denitrification, which is so far the only known process able to use N 2 O as an electron acceptor instead of dissolved O 2 . Canonical denitrification along with anaerobic NH 4 + oxidation (anammox), both processes particularly active in the ESP [19,20], are the main sinks for fixed N budgets.
There has been long-standing uncertainty as to whether or not the ocean is losing N faster than it is being incorporated via marine N 2 fixation [8,21]. Important advances in the understanding of regulation, rates and the microorganisms involved in N 2 fixation have been made in recent years [22][23][24]. These studies indicate that NifH is present in a diverse range of microbial groups which include the well-studied diazotrophic cyanobacteria, diatomdiazotroph assemblages and gammaproteobacteria [23,25,26]. In fact, diazotrophs display high levels of metabolic diversity, much greater than previously thought, among which can be found not only well-known photoautotrophs but also chemo-and heterotrophic diazotrophs [27]. Recent results showing the expression of nitrogenase gene (nifH), suggest that heterotrophic bacteria dominate the diazotrophic community in the oligotrophic waters of the western South Pacific [28] and even in the OMZ of the Arabian Sea [29].
The possible occurrence of biological N 2 O fixation or rather the well-known process which removes N 2 O was investigated using an enriched, labeled substrate ( 15 N 2 O). This process was explored using field samples and cultures of marine cyanobacteria such as Trichodesmium and Crocosphaera. These bacteria, particularly Trichodesmium, are present in most tropical and subtropical gyres, being the dominant diazotrophic species and having global significance in relation to the introduction of new N into the ocean [30,31].

Results and Discussion
N 2 O fixation was explored in field experiments carried out during several cruises covering a wide range of geographical locations (13u-36.5uS; 72u-110uW) and depths (from the surface down to 400 m). The study areas have extreme biogeochemical conditions reflected in surface chlorophyll-a (Chl-a; Fig. 1A), dissolved fixed N (mainly NO 3 2 ) and O 2 . The latter variable varies from the oxygenated waters in the subtropical gyre and south of ,37uS to suboxic-anoxic waters (0$ O 2 #11 mmol L 21 ) off Peru and northern Chile. Oxygen deficient waters are clearly observed (Fig. 1B), delimiting an OMZ that has become one of the shallowest and most intense in the world ocean [19]. Oceanographic conditions of the sampled stations are summarized in Table 1. N 2 O fixation was observed in 92% of sampled depths, with rates ranging between 2 pmol N L 21 d 21 and 14.8 nmol N L 21 d 21 (mean 6 SE of 0.52261.06 nmol N L 21 d 21 , n = 93). These field samples, which came from different trophic (Fig. 1A) and O 2 regimes (Fig. 1B), were incubated on board under a light gradient (65% to 4% of surface irradiance) and under dark conditions, and with temperature and dissolved O 2 levels maintained close to those found under in situ conditions (Table S2). N 2 O fixation was blocked in control experiments treated with HgCl 2 (total experiments n = 15), confirming the biological nature of this process, hereafter referred to as assimilative N 2 O fixation.
Vertical distributions of N 2 O fixation rates are shown in Fig. 2A, D, G, J along with the semi-conservative tracer N* [21] (Fig. 2B, E, H, K) and DN 2 O [32] (Fig. 2C, F, I, L). Off Peru (CUP) and northern Chile (CUNC), N 2 O fixation rates peaked at the surface and/or around the base of the oxycline (Fig. 2D, 2G) and were well correlated with fluorescence and particulate organic carbon and nitrogen (POC/PON) distributions (data not shown). Off central Chile (CUCC, Fig. 2J), N 2 O fixation rates were higher in the photic-oxic layer decreasing toward deeper waters (90 m depth). At the STG (BR-7 station, Fig. 2A), remarkably, vertical N 2 O fixation showed a maximum level at the base of the eufothic zone, where chlorophyll fluorescence and POC revealed a maximum level [33]. Additionally, experiments performed in the photic zone during the Big Rapa (BR-1 and BR-7 stations) showed active N 2 O fixation rates under dark as well as in situ light conditions (see Table S2), but without any significant differences. This supports the idea that microorganisms assimilating N 2 O are not only photoautotrophs and that part of microbes fixing N 2 O could be heterotrophs. Table 2 shows ranges of N 2 O fixation rates and their statistics, along with basic biogeochemical variables in predefined areas i.e., the STG and coastal upwelling (CU) centers. There was significant variation in N 2 O fixation rates among the study areas (Kruskal Wallis test, p,0.05), being higher at the CUNC. Further analyses involving a multiple linear regression (p,0.05) showed that 27% of the variation observed in N 2 O fixation rates was linked to dissolved O 2 concentrations, with higher rates observed at lower O 2 values; less variance percentage was ascribed to differences in Chl-a and fixed N pools among areas. In this sense, it has been suggested that dissolved O 2 controls surface diazotrophic activity which could be enhanced in close proximity to water column denitrification areas [34].
Assimilative N 2 O fixation rates in the photic layer samples, under different light intensities, averaged 0.49362.27 nmol N L 21 d 21 (n = 42). Despite the fact that certain maxima of N 2 O fixation rates are observed in the photic zone, these values were not significantly different (Mann-Whitney test, p = 0.03) from those measured in the aphotic layer, under the occasional influence of the OMZ, which averaged 0.10460.190 nmol N L 21 d 21 (n = 32).
In addition, all sampled areas in this study share a moderate to severe iron and NO 3 2 deficiency [35][36][37] which is reflected in the N* index. Indeed, in each predefined area N* profiles started with values close to zero in surface water and decreased with depth, reaching negative N* values as low as 230 (Fig. 2B, E, H and K). Thus, N* indicates a depletion or deficit of N compared to P, and could be a consequence of the lateral and upward transport of denitrified waters from the OMZ belonging to the ESP, causing a (lower than expected) deviation of the Redfield N:P ratio. If PO 4 23 is forced above the Redfield ratio, N fixing organisms may gain a selective advantage, which increases the inventory of NO 3 2 . Consequently, the impact of N 2 O fixation may lie in compensating N 2 O loss caused by canonical denitrification. Particularly, the N deficit relative to P is a common pattern observed at the STG [38] ( Fig. 2 B) and may be accentuated by weak vertical mixing and very scarce aeolian dust supply toward this region [14]. There, N* values close to zero or negative zero are usually observed within the gyre explaining the predominance of advected denitrification over N-fixation [21,38].
On the other hand, within the STG (St. BR-7) noticeable negative values of DN 2 O in oxygenated water were detected (Fig. 2C). However, these could not be possible by heterotrophic denitrification given the oxygenated condition of the entire water column. Therefore, the DN 2 O values suggest that the N 2 O deficit could be caused by assimilative N 2 O consumption. In contrast, in the upwelling of Peru, northern Chile DN 2 O values decrease from positive values at surface, peaking sometime at the oxyclines, to close to zero or lightly negative towards the OMZ's core (Fig. 2F, I); there N 2 O is consumed by canonical denitrification [39].
It is important to note that N 2 fixation rates were detected at the same stations and depths [40] as the detection of N 2 O fixation rates in our sampled CU areas (in 60% of the samples). In addition, in both CUP and CUNC study areas, dinitrogenase reductase genes (nifH) have been found and phylogenetic analysis revealed a diverse diazotrophic community [40] in which nifH sequences fell within three of the four known clusters for this gene [41,42]. Importantly however, no sequences associated with cyanobacteria were found during that study [40] which agrees with nifH transcripts retrieved in the western South Pacific gyre that showed active heterotrophic communities, mainly c-proteobacteria from cluster I [42], able to fix N 2 [28].
In order to check whether typical diazotrophic organisms have the same capacity to fix N 2 O as they have for fixing N 2 , three of the most studied strains of marine cyanobacteria [43] belonging to Trichodesmium (H-9 and IMS101) and Crocosphaera (W-8501) were cultured under laboratory conditions. It is important to note that they used NifH to transform N 2 into PON, but lacked the nosZ genes that encode the catalyzing enzymes of dissimilative N 2 O reduction to N 2 [44], thereby only leaving assimilative N 2 O fixation as the potential N 2 O consumption process. By contrast, Synechococcus (RCC 1029) was used as a negative control, given that this species does not possess either nifH or nosZ genes in its genome [45].
Several types of experiments were carried out with these cyanobacterial strains (Table S2). All incubated strains except strain RCC 1029 showed a significant excess of 15 N in PON (referred to as atom%) of 2 to 5 fold higher with respect to the natural abundance of 15 N in PON (,0.369 atom%) and its variation seemed to depend on the cell density of each cultured strain (quantified as particulate organic nitrogen ''PON''). Time course assays of N 2 O fixation with these strains at different cell densities (measured throughout PON) are illustrated in Figure 3. When observed, the incorporation of 15  This observed trend suggests that nitrogenase has an increasing affinity for N 2 O as N 2 O doses increased (Fig. 4A). It is important to remark that a slight enrichment in atom% (and even atom% excess with respect to natural isotopic abundance) was observed at minimal N 2 O doses (Fig. 4), whose final concentration was close to those environmental levels found in the STG (,6-10 nmol L 21 ). This suggests that even at very low N 2 O concentrations, N 2 O incorporation may occur. In surface waters associated with coastal upwelling areas, N 2 O concentrations can reach 100 nmol L 21 or more [46], and if we look at water under the influence of subsurface O 2 deficiency, N 2 O   concentrations as high as 400 nmol L 21 have been reported [46]. Assuming that an increase in atmospheric N 2 O is a likely trend under certain future intensification of eutrophication scenarios [47], and hypoxia and warming [48], this final added concentration of N 2 O in field experiments could be plausible for some marine environments. Given that some of the N 2 O enrichments used in our field experiments (e.g., the STG Fig. 2A, D, G, J) largely exceeded natural N 2 O levels in the surface ocean by 100 and 1000 times, rates of N 2 O fixation obtained for these sampled areas should be considered as a potential. A simple calculation of the N 2 O turnover time was carried out in surface waters of the STG and CU areas, taking into consideration surface N 2 O inventories and assimilative N 2 O fixation rates 100 and 1000 times smaller than those respectively measured, i.e. around 1 to 10 pmol L 21 d 21 ( Table 2). Estimates showed that the surface N 2 O inventory could be removed entirely between 2 and 12 years. Further kinetic studies and half-saturation constant determinations will be necessary in order to assess the ability of diazotrophic organisms to use N 2 O at its naturally occurring levels in the ocean.
One question that is worth addressing is why N-fixers would fix N 2 O instead of N 2 . When answering this question one has to bear in mind two points: 1) that both pathways are associated with nonselective NifH; and 2) that there might be benefits related to energy for using N 2 O instead of N 2 . The Gibbs free energy required during N 2 O assimilation is thermodynamically advantageous compared to that of N 2 because the dissociation energy for breaking the N-N bond in the case of N 2 O is only half that required for the N 2 molecule [49,50]. Thus, if available, N 2 O may appear to be a more energetically favorable substrate than N 2 .
It is however possible that this process does not directly occur. Therefore, dissimilative reduction of N 2 O to N 2 followed by the biological fixation of N 2 into PON may take place, in which case assimilative N 2 O fixation does not constitute a process in itself. This possibility was tested by looking at the isotopic composition of dissolved N 2 in seawater medium of the cultures (taken thought exetainer) following the addition of 15  This study provides further arguments in favor of the fact that assimilative N 2 O fixation should exist in marine waters to reduce uncertainties for the marine N cycle. One such argument is that the current isotopic and isotopomeric composition of accumulated atmospheric N 2 O cannot yet be precisely constrained by the N and O isotopic (and isotopomeric) compositions of N 2 O on the ocean surface [51]. A second argument is that it is well-known that there is an imbalance between the sources and sinks of the global N budget [6,19]. This should become less pronounced if a new N 2 O utilization pathway is included, in some way affecting the global N 2 O cycle [52].   . 5).
These surface values at the STG coincide with those reported for the subtropical North and South Pacific gyre [54][55][56], and support the idea that the isotopic N 2 O composition of surface waters cannot simply be the result of mixing between the atmospheric and subsurface marine N 2 O pools. Therefore, some surface water biological process should operate to maintain the isotopic compositions of d 15 N, d 18 O and SP as observed. In view of the prevailing levels of oxygenation in these waters, nitrification would be expected to be the process of greatest importance. In cultures of ammonia-oxidizing bacteria, regardless of the specific mechanism of N 2 O production, this gas is produced as a byproduct with much more depleted isotopic compositions for N and O than those observed in the ocean surface. For example, nitrifying bacterial cultures exhibit average signals of d 15 N bulk from 254.9% to 26.6%, d 18 O = 40% and a SP ,33.5% via NH 2 OH decomposition, but an SP of about 20.8% via nitrifier denitrification [57,58]. With the recent finding that marine Archaea can produce N 2 O via ammonium oxidation, [4,59], resulting in a heavier isotopic composition than that observed in ammonia oxidizing bacteria (d 15     Here, some type of biological process such as assimilative N 2 O reduction into PON that enriches the dissolved N 2 O pool in d 15 N bulk and d 18 O is able to resolve the observed discrepancies. This process could also explain the estimated negative values of DN 2 O (Fig. 1C) in the STG. However, the negative values of N* and positive values of DN 2 O as observed within the CU do not exclude the possibility that assimilative N 2 O fixation is taking place but simply indicate that other N 2 O producing processes (nitrification or partial denitrification), and/or in situ total denitrification along with the advection of denitrified waters are occurring faster than assimilative N 2 O fixation, camouflaging both expected N* and DN 2 O signals.
Secondly, if N 2 O fixation is significantly taking place in the world's oceans, it should have an important effect on the global ocean DN 2 O inventory which is illustrated in Table S3 and Text S3. As part of assimilative N 2 O fixation appears to be carried by phototrophic diazotrophs, (even if one cannot exclude other micro-organisms), it should be directly linked to diazotroph biomass or abundance. Thus, significant contributions of N 2 O fixation should be observed in these regions where diazotrophs are of quantitative importance. Indeed, emerging patterns of marine N fixation suggest that the Pacific Ocean, in particular the South Pacific (although poorly studied), has the lowest abundance and diazotroph activity compared with other regions such as the North Atlantic, the Indo-Pacific oceanic region and even the Baltic Sea [22]. Further research on N 2 O fixation in other oceanic basins is recommended and several insights appear to indicate that assimilative N 2 O fixation could be higher in regions where diazotrophic microorganisms were more abundant or more active The gathered evidence draws attention to the importance of this pathway for regional and even global N 2 O removal, preventing part of its potential efflux toward the atmosphere. Thus, N 2 O represents a form of yet unreported fixed N that could change our vision and understanding of the oceanic N cycle.

Methods
This study covers coastal upwelling centers off Peru (CUP  Table 1 summarizes the location, water depth and other oceanographic variables and parameters measured at the sampling stations. During all cruises, vertical profiles of temperature, salinity, dissolved O 2 , fluorescence and PAR (Photosynthetically Active Radiation) were obtained using a Conductivity Temperature Depth CTD-O 2 probe (Sea-Bird Electronics Inc., USA). The O 2 sensors from the upcast CTD-O were calibrated with discrete samples obtained by Winkler titration (see below). In the case of the KN182-9 cruise, O 2 sensors were calibrated pre-and postcruise at Woods Hole Oceanographic Institution (USA). During the Galathea-3 cruise, an ultrasensitive sensor STOX was tested in situ [60]. Water column light irradiance, averaged over the visible spectrum (400-700 nm), was measured using a LI-COR (LI-190) quantum sensor, and fluorescence was measured using a WetStar sensor.
Discrete water samples, for chemical analyses and experiments, from the surface (,2 m), down to 400 m were collected using Niskin bottles (12 L) attached to a rosette sampler. Core parameters, including dissolved O 2 and N 2 O, nutrients (NH 4 + , NO 3 2 , NO 2 2 , PO 4 32 ), Chl-a, particulate organic carbon and nitrogen (POC and PON), and their natural C and N isotopic compositions, were determined at all stations. N*, a quasiconservative tracer defined as a linear combination of NO 3 2 and PO 4 32 , was estimated from nutrient concentrations throughout the water column [21]. Apparent N 2 O production (DN 2 O) was obtained from the difference between the N 2 O saturation at equilibrium with the atmosphere and its concentration measured in seawater [61].
In terms of the employed analytical methods, dissolved O 2 was analyzed in triplicate by automatic Winkler titration. The samples for N 2 O analyses were transferred directly into 20-mL glass vials (triplicates), preserved with 50 mL of saturated HgCl 2 and sealed with butyl rubber and aluminum cap stoppers. N 2 O was determined by Helium equilibration in the vial, followed by quantification with a Varian 3380 Gas Chromatograph (GC) using an electron capture detector maintained at 350uC, for more details see [46]. A calibration curve was made with 5 points (He, 0.1 ppm, air, 0.5 ppm, and 1 ppm) and the detector linearly responded to this concentration range. The analytical error for the N 2 O analysis was less than 3%.
Nutrient samples were collected with a 60 mL plastic syringe and filtered through a glass fiber filter (pore size 0.7 mm) into highdensity polypropylene scintillation vials. Samples were stored at -20uC until laboratory analysis, except during the KN182-9 and Galathea-3 cruises when dissolved NO 2 2 and PO 4 23 concentrations were immediately determined on board [62]. For determination of Chl-a, a fluorometry method was used for filtered seawater through a 45 mm Whatman GF/F filter [63]. Analyses of particulate organic C and N (POC and PON) and their natural 13 C and 15 N isotopic compositions were carried out after filtering 1 L of seawater through pre-combusted 0.7 mm glass fiber filters (22 mm Whatman GF-F) and stored at -20uC until analysis. Filters were dried at 60uC for 12 h before determining their isotopic composition via continuous-flow isotope ratio mass spectrometry (IRMS; Finnigan Delta Plus). Reproducibility for 13 C and 15 N was greater than 0.11% and 0.02%, respectively, based on the acetanilide standard used as reference material. Isotope ratios were expressed as per mil deviations from the isotopic composition of Vienna PDB and air, for 13 C and 15 N, respectively [64]. Significant differences were checked between the enrichment as 15 N atom% and its atom% excess of PON in the experiments with respect to the natural background or natural isotopic composition of PON taken at each sampled station and depth.
Additionally, oceanographic/meteorological and biogeochemical variables/parameters shown in Table 1 were measured and estimated. The mixed layer depth (Z m ) was obtained from vertical density profiles measured every 1 dbar using the CTD sensor, and the depth of the euphotic zone (irradiation at 1% of its surface value) was estimated from the attenuation coefficient of downwelling irradiance averaged over the visible spectrum (400-700 nm) measured by a LI-COR sensor. In the few instances where light profiling was not possible (night sampling), light profiles were estimated using surface irradiation (assumed 4% surface reflection) and the vertical attenuation coefficient of PAR (K) from the previous day.
In order to detect differences in N 2 O fixation rates among the sampled areas (i.e., STG, CUP, CUNC and CUCC), non parametric Kruskal Wallis test was carried out using statistical language R [65]. Additionally, a multiple linear regression model was performed to assess the variables that determine the variance of N 2 O fixation rates, with a prior logarithmic transformation of this dependent variable [65]. Categorical variables associated with the different study areas were also included. A step-wise selection was used to test the significance of each variable in the model. The best model was obtained by determining its heteroscedasticity and p-value (p,0.05). Models were then contrasted using the Akaike Information Criteria (AIC). The comparison among light (65,30,4 Table S2).
Field samples were incubated on board, under temperaturecontrolled conditions, using an in situ range of light intensities, as well as dark conditions. For this purpose, seawater was dispensed from Niskin bottles using a gas-tight Tygon tube, to avoid any oxygenation, into 1.5-2 L double-laminated aluminum-polyethylene or transparent TedlarH bags. The volume and weight of the filled bags were controlled at the beginning and end of the incubation process, and real volumes were used in rate calculations. As an additional precaution, a permeability test was performed on bags prior to the experiments. The bags were filled with pure helium and monitored for 24 h using gas chromatography. They showed no atmospheric gas (N 2 O or CH 4 ) intrusions, thus ensuring hermetically sealed conditions. Atmospheric O 2 could not be tested in the bags via chromatography (given the sensitivity of the chromatographic method), however tests were performed using atmospheric N 2 O as a tracer in vacuum emptied bags. As N 2 O intrusions were not detected inside the bag during the days following the sampling, O 2 intrusions during incubations in similar bags were considered unlikely. Each bag had a hose/ valve with a septum through which 15 N 2 O tracer and different treatments, i.e., 15 N 2 O, HgCl 2 (see Table S2) were injected using gastight syringes. Tracer addition was carried out at a final concentration of 10,20 mmol L 21 (1 or 2 mL of tracer gas into 1.5-2 L incubation volume). The relative tracer ( 15 N 2 O) concentration with respect to natural 14 N 2 O background varied between 100 and 1000 fold (depending on seawater N 2 O levels), taking into consideration solubility and partition coefficients, as well as the ratio between gas and liquid phases of N 2 O in the bag [61].
Most of the incubations were performed using six deck incubators maintained at sea surface temperature and with light intensities ranging between 65% and 4% of incident light (Lee FiltersH). Samples from below the base of the euphotic layer were incubated in the dark in a thermo-regulated bath (Johnson ControlH, KN182-9 cruise) or a temperature controlled incubator at temperatures close to in situ. During the Big Rapa cruise, surface duplicate samples were simultaneously incubated under in situ light and dark conditions. The incubations lasted 24 h and they were terminated by gentle filtration onto pre-combusted 0.7 mm glass fiber filters (Whatman GF/F filters) using a vacuum (,100 mm Hg) or a peristaltic pump. Filters were stored at 220uC until laboratory analysis. Off central Chile, samples were incubated at two times (t = 12 and t = 24) with artificial light in a temperaturecontrolled room.
The diazotrophic strains were cultivated at 24uC under artificial light according to a daily cycle of 12 h light/12 h dark, in YBCII artificial seawater medium with no nitrogen sources [67], while Synechococcus was cultivated with the same cycle in PCR-S11 medium [68] at 18uC. Several batches of these strains were then transferred to GC bottles (50 or 125 mL), and 50 or 100 mL of 15 N 2 O (depending on of the vials) was injected into each bottle through septa. Different doses were inoculated, with final concentrations going from 10 to 400 nmol L 21 (expected levels in the study area). Cell density was variable depending on the type of cultivated strain and date when experiments were undertaken. PON level (mg L 21 ) measured during each time incubation was used as a biomass index. No variations of PON (Fig. 3) were recorded during time course experiments, indicating no net growth during incubations.
Prior to each experiment and tracer inoculation, the health and purity of cultures were checked using flow cytometry and microscopy. The IMS101, H-9, WH-8501, RCC-1029 strains were not axenic cultures and, therefore, bacterial cell density was periodically checked in the cultures with a FACSCalibur flow cytometer equipped with an ion-argon laser of 488 nm of 15 mW (Becton Dickinson). The picoplanktonic bacterial abundance was estimated from samples previously stained with SYBR-Green I (10,000 x; Molecular Probes) following Marie et al, 2000 [69]. Cell Quest Pro and Cytow software were used for data acquisition and analysis. In the case of WH-8501, abundances varied from 5610 3 to , 4610 4 cell mL 21 , but the abundance of WH-8501strain was an order of greater magnitude.
Finally, after addition of 15 N 2 O to selected experiments with H-9 and WH-8501 strains, liquid samples were collected in exetainers (LABCO) with particular care to avoid air contamination, and the isotopic composition of N 2 was measured in order to check if one-step reaction transforming N 2 O directly into NH 4 + proceeded without intermediary N 2 . The isotope ratios of N 2 were measured in gas mixtures (headspace) using a Thermo Finnigan GasBench+PreCon trace gas concentration system interfaced to a Thermo Scientific Delta V Plus isotope-ratio mass spectrometer (U Davis, USA; http://stableisotopefacility.ucdavis.edu/n2.html).