Transfer of diazotroph-derived nitrogen towards non-diazotrophic planktonic communities: a comparative study between Trichodesmium erythraeum, Crocosphaera watsonii and Cyanothece sp.

Abstract. Biological dinitrogen (N2) fixation is the major source of new nitrogen (N) for the open ocean, and thus promotes marine productivity, in particular in the vast N-depleted regions of the surface ocean. Yet, the fate of the diazotroph-derived N (DDN) in marine ecosystems is poorly understood, and its transfer to auto- and heterotrophic surrounding plankton communities is rarely measured due to technical limitations. Moreover, the different diazotrophs involved in N2 fixation (Trichodesmium spp. vs. UCYN) exhibit distinct patterns of N2 fixation and inhabit different ecological niches, thus having potentially different fates in the marine food webs that remain to be explored. Here we used nanometer scale secondary ion mass spectrometry (nanoSIMS) coupled with 15N2 isotopic labelling and flow cytometry cell sorting to examine the DDN transfer to specific groups of natural phytoplankton and bacteria during artificially induced diazotroph blooms in New Caledonia (southwestern Pacific). The fate of the DDN was compared according to the three diazotrophs: the filamentous and colony-forming Trichodesmium erythraeum (IMS101), and the unicellular strains Crocosphaera watsonii WH8501 and Cyanothece ATCC51142. After 48 h, 7–17 % of the N2 fixed during the experiment was transferred to the dissolved pool and 6–12 % was transferred to non-diazotrophic plankton. The transfer was twice as high in the T. erythraeum bloom than in the C. watsonii and Cyanothece blooms, which shows that filamentous diazotrophs blooms are more efficient at promoting non-diazotrophic production in N-depleted areas. The amount of DDN released in the dissolved pool did not appear to be a good indicator of the DDN transfer efficiency towards the non-diazotrophic plankton. In contrast, the 15N-enrichment of the extracellular ammonium (NH4+) pool was a good indicator of the DDN transfer efficiency: it was significantly higher in the T. erythraeum than in unicellular diazotroph blooms, leading to a DDN transfer twice as efficient. This suggests that NH4+ was the main pathway of the DDN transfer from diazotrophs to non-diazotrophs. The three simulated diazotroph blooms led to significant increases in non-diazotrophic plankton biomass. This increase in biomass was first associated with heterotrophic bacteria followed by phytoplankton, indicating that heterotrophs took the most advantage of the DDN in this oligotrophic ecosystem.

Trichodesmium spp. and the diatoms-diazotrophs associations (DDAs) were first thought to be the main contributors to oceanic N 2 -fixation (Capone et al., 1997;LaRoche and Breitbarth, 2005;Mague et al., 1974). However, the use of molecular tools has demonstrated that the diversity of diazotrophs was greater than previously thought, highlighting in particular the role of pico-and nano-planktonic unicellular cyanobacteria, termed UCYN (Needoba et al., 2007;Zehr et al., 1998Zehr et al., , 2001. The latter are now considered to be of a major importance in the global N 2 fixation budget due to their broad distribution and high abundance in several oceanic basins (Luo et al., 2012;Moisander et al., 2010;Needoba et al., 2007). These observations are confirmed by the high contribution of N 2 fixation rates reported in the < 10 µm size fraction (Bonnet et al., 2009;Dore et al., 2002;Montoya et al., 2004).
While studies dealing with the diversity and the biogeographical distribution of diazotrophs in the ocean are on the increase, little is known regarding the fate of the fixed N 2 by the diazotrophs (hereafter called diazotroph-derived N, DDN) in the ocean. It remains unclear whether the DDN is preferentially directly exported out of the photic zone, recycled by the microbial loop, or transferred into larger organisms, subsequently enhancing indirect particle export. Some studies report low δ 15 N signatures on zooplankton, evidencing the transfer of DDN towards higher trophic levels (Montoya et al., 2002). This transfer can be directly through the ingestion of diazotrophs (O'Neil et al., 1996;Wannicke et al., 2013), or indirect, i.e. mediated through the release of dissolved N by diazotrophs (Capone et al., 1994;Glibert and Bronk, 1994;Mulholland et al., 2004), which is taken up by heterotrophic and autotrophic plankton (Bonnet et al., 2016c), and is subsequently consumed by the zooplankton (e.g. O'Neil et al., 1996). Other studies performed in the tropical North Atlantic and Pacific Oceans report low δ 15 N signatures on particles from sediment traps, suggesting that at least part of the DDN is ultimately exported out of the photic zone (Bourbonnais et al., 2009;Karl et al., 2002;Knapp et al., 2005). However, the export efficiency appears to depend on the diazotrophs involved in N 2 fixation in surface waters: while it has been demonstrated that DDAs directly contribute to particle export Subramaniam et al., 2008;Yeung et al., 2012), Trichodesmium spp. is rarely found in sediment traps (Walsby, 1992) mainly due to its positive buoyancy, regulated by the production of carbohydrates (Romans et al., 1994). Data on the export efficiency of UCYN are scarce. During the VAHINE mesocom experiment designed to track the fate of DDN in the surface oligotrophic ocean, Berthelot et al. (2015b) showed that the production sustained by UCYN (mainly related to group C) resulted in a higher rate of particle export compared to the production sustained by DDAs. In this same special issue, Bonnet et al. (2016a) confirmed that UCYN-C significantly contribute to POC export (up to 22.4 ± 5.5 % at the height of the UCYN-C bloom).
However, most of the particle export associated with UCYN-C was probably mainly indirect through recycling processes and DDN transfer to surrounding planktonic communities (Bonnet et al., 2016a). However, such transfer of DDN to the surrounding planktonic communities and its potential impact on export production is poorly understood and rarely quantified.
The transfer of DDN to surrounding plankton is mediated through the dissolved pool as diazotrophs release a significant fraction of the fixed N (10-50 %) under the form of ammonium (NH + 4 ) and dissolved organic N (DON; Benavides et al., 2013;Glibert and Bronk, 1994;Konno et al., 2010;Mulholland and Bernhardt, 2005;Mulholland et al., 2004). This release of DDN by diazotrophs has been linked to exogenous processes such as viral lysis (Hewson et al., 2004;Ohki, 1999), copepods sloppy feeding (O'Neil et al., 1996) or programmed cell death (Berman-Frank et al., 2004). Significant N release was also reported in axenic cultures, suggesting that it is also an endogenous process (Mulholland et al., 2004). Once released, fixed N compounds are potentially transferred to non-diazotrophic plankton communities, as suggested by massive developments of diatoms (Devassy et al., 1979;Dore et al., 2008;Lee Chen et al., 2011) and dinoflagellates (Lenes and Heil, 2010;Mulholland et al., 2006) during or following blooms of Trichodesmium spp. 15 Nenrichment measured in size-fractioned pico-plankton after 15 N 2 incubations also supports the idea of a DDN transfer within the planktonic community; Bryceson and Fay, 1981;Garcia et al., 2007). However, this method probably overestimates the DDN transfer as it is not possible to discriminate between DDN that has been transferred to pico-plankton and N 2 fixation by pico-plankton itself. Bonnet et al. (2016c) recently measured the actual transfer of DDN from several Trichodesmium spp. blooms to different groups of autotrophic and heterotrophic plankton using single cell mass spectrometry analyses (nanoSIMS) coupled with cell sorting by flow cytometry after 15 N 2 labelling, and showed that the DDN was predominantly transferred to diatoms and bacteria, and DDN was mainly converted to diatom biomass. This study was performed during naturally occurring Trichodesmium spp. blooms, but comparative studies on the transfer efficiency of DDN from different diazotrophs are lacking. Trichodesmium spp. and UCYN exhibit distinct patterns of N 2 fixation (the first fix during the day, while the second fix is during the night, e.g. Bergman et al., 2013;Dron et al., 2012) and inhabit different ecological niches (Luo et al., 2012), thus having potentially different fates in the marine food webs, that remains to be explored.
Here, we compared N 2 fixation rates, the quantity and the quality of DDN released in the dissolved pool and the transfer of DDN towards non-diazotrophic plankton from three distinct diazotrophic groups: Trichodesmium erythraeum, Crocosphaera watsonii and Cyanothece sp. For this purpose, we simulated blooms of these three diazotroph phylotypes by inoculating freshly sampled seawater containing  (Bonnet et al., 2016b). The present experiment performed in microcosms was designed to complement the mesocosm experiment and compare the fate of DDN originating from distinct groups of diazotrophs.

Cultures maintenance
Three unialgal cultures of diazotrophs abundant in the southwestern Pacific (e.g. Bonnet et al., 2015;Turk-Kubo et al., 2015) were used in this study to simulate blooms of the filamentous colony forming Trichodesmium erythraeum IMS101, and the UCYN strains Crocosphaera watsonii WH8501 and Cyanothece ATTC51142. They were grown in batch cultures under close to lagoon conditions, and maintained in exponentially growing phase under 120 photons m −2 irradiance on a 12 : 12 light : dark cycle at 27 • C. The culture medium was composed of 0.2 µm filtered and sterilized seawater collected in the New Caledonian lagoon (166.44 • E, 22.48 • S), at the study site where the DDN transfer experiment described below was performed. The collected seawater was characterized by lowF nitrate + nitrite (NO x ) concentrations (< 0.1 µmol L −1 ). It was amended with phosphate (PO 3− 4 ) and micronutrients according to the N-deplete YBCII medium recipe (Chen et al., 1996), except for PO 3− 4 concentration, which was reduced to 10 instead of 50 µmol L −1 in the original medium. Cultures were acclimated to this medium for at least 10 generations before the experiment started. They were not axenic but manipulations under laminar flow hood and sterilization of the lab materials were performed in order to limit bacterial contamination. Before inoculation into natural seawater, and in order to control the biomass of diazotrophs added, cultures were monitored microscopically every 1-2 days on a Malassez counting cell for UCYN and on a 10 µm polycar-bonate filter for T. erythraeum, using an epifluorescence microscope (Zeiss Axioplan, Jana, Germany) fitted with a green (510-560 nm) excitation filter.

DDN transfer experiment
Seawater containing the natural planktonic community was collected at the experimental study site on 2 February 2014 at 2 m depth, using an air-compressed Teflon pump (AstiPure ™ ) connected to a polyethylene tubing. At the time of the sampling, the seawater temperature was 25.4 • C.  and NO x concentrations were < 0.2 µmol L −1 . Seawater was transferred into 15 HCl-washed 4.5 L polycarbonate bottles equipped with septum caps and quickly brought back to the laboratory. Bottles were divided into five sets of three replicates. The first set was immediately amended with Trichodesmium erythraeum (hereafter referred to as "T. erythraeum treatment"), the second with the UCYN Crocosphaera watsonii (hereafter referred to as "C. watsonii treatment"), the third one with the UCYN Cyanothece spp. (hereafter referred to as "Cyanothece treatment"), the fourth set was left unamended and served as a control (hereafter referred to as "Control treatment"), and the last set was immediately processed as described below to characterize the initial conditions (T0). To simulate blooms of the different diazotrophs, we added 5.10 3 trichomes L −1 for T. erythraeum treatment and 1.10 6 cells L −1 for the UCYN treatments, to be representative of the diazotroph blooms observed in the southwestern Pacific region Moisander et al., 2010;Rodier and Le Borgne, 2008;Shiozaki et al., 2014). Care was made to introduce a similar biomass of diazotrophs in each treatments in order to be able to compare the different treatments. The initial cultures were sufficiently concentrated in cells in such a way that the volume of culture added represented less than 1 % of the 4.5 L bottles volume, so nutrient concentrations, especially PO 3− 4 concentrations were not significantly influenced by these additions, which represented < 0.05 µmol L −1 of added PO 3− 4 . Immediately after the diazotrophs inoculation, all 4.5 L bottles were amended with NaH 13 CO 3 (EURISOTOP, 99 atom % 13 C, 5 g in 60 mL of deionized water) to obtain a ∼ 10 atom % 13 C-enrichment (1 mL in each 4.5 L bottles) and 15 N 2 (98.9 atom % 15 N, Cambridge isotopes) enriched seawater, according to the protocol developed by Mohr et al. (2010) and fully described in Berthelot et al. (2015a). Briefly, 15 N 2 enriched seawater was prepared by circulating 0.2 µm filtered seawater collected at the same site as described above through a degassing membrane (Membrana, Minimodule ® , flow rate 450 mL min −1 ) connected to a vacuum pump (< 850 mbar) for at least 1 h. The degassed seawater was transferred to a 2 L gas tight Tedlar ® bag and amended with 1 mL of 15 N 2 per 100 mL of seawater. The 15 N 2 bubble was vigorously shaken for 5 to 10 min until its complete dissolution. The incubation bottles were then amended with 5 % vol : vol enriched seawater and closed without headspace with septum caps. The final 15 N-enrichment of the N 2 pool in the incubation bottles was measured using a Membrane Inlet Mass Spectrometer (Kana et al., 1994) and was found to be 3.5 ± 0.2 atom % (n = 9). The potential contamination by 15 NO x and 15 NH 3 of the 15 N 2 bottles, recently highlighted by Dabundo et al. (2014), was tested on one of our 15 N 2 Cambridge Isotope batch. According to the model described in Dabundo et al. (2014), it appeared that the low level of contamination measured (1.4×10 −8 mol of 15 NO 3 mol −1 of 15 N 2 and 1.1×10 −8 mol NH + 4 mol −1 of 15 N 2 ) would only contribute to ∼ 0.05 % of the DD 15 N measured in our study and was thus neglected.
Except for the T0 set of bottles, all bottles were incubated for 48 h under in situ-simulated conditions in on-deck incubators at ∼ 26.5 • C with continuous water flowing irradiances corresponding to the sampling depth using neutral screening. Bottles were gently mixed three times per day during the experiment to insure homogeneity. After incubation, the four sets of bottles (the three diazotrophs-amended treatments and the control treatment) were recovered and subsampled to analyze the following parameters: heterotrophic bacteria and phytoplankton abundances, N 2 fixation rates, DDN release, organic and inorganic nutrients concentrations and cellular 15 N-and 13 C-enrichment on diazotrophs and non-diazotrophic plankton groups (see below for detailed protocols). Unless otherwise stated, samples were taken individually in each bottle of each set, so each parameter was measured in triplicate in every treatment.

Plankton abundance determination
Samples for micro-phytoplankton were collected from the 4.5 L incubation bottles in 250 mL glass bottles and fixed with lugol (0.5 % final concentration). Diatoms, dinoflagellates and micro-zooplankton (ciliates) were identified and enumerated to the lowest possible taxonomic level from a 100 mL subsample following the Utermohl methodology (Hasle, 1978), using a Nikon Eclipse TE2000-E inverted microscope equipped with phase-contrast and a long distance condenser.
Pico-, nano-phytoplankton and bacterial abundances were determined using flow cytometry. For this purpose, samples were collected in 1.8 mL cryotubes, fixed with paraformaldehyde (final concentration 2 %), left at ambient temperature for 15 min in the dark, flash frozen in liquid N 2 and stored at −80 • C. Analyses were carried out at the PRECYM flow cytometry platform (https://precym.mio.univ-amu.fr/) using standard flow cytometry protocols (Marie et al., 1999) to enumerate phytoplankton and heterotrophe bacteria, using a FACSCalibur analyzer (BD Biosciences, San Jose, CA). Samples were thawed at room temperature and just before analyses, were added to each sample: 2 µm beads (Fluoresbrite YG, Polysciences), used as internal control (and to discriminate picoplankton < 2 µm < nanoplankton populations), and Trucount beads (BD Biosciences), used to de-termine the volume analyzed. An estimation of the flow rate was calculated by weighing three tubes of samples before and after a 3 mn run of the cytometer. The cell concentration was determined from both Trucount beads and flow rate measurements. For picoplankton cells, the red fluorescence (670LP, related to chlorophyll a content) was used as trigger signal and cells were characterized by three other optical signals: forward scatter (FSC, related to cell size), side scatter (SSC, related to cell structure), and the orange fluorescence (580/30 nm, related to phycoerythrin content). Phytoplankton communities were clustered as Synechococcus spp. cell like (hereafter called Synechococcus), Prochlorococcus spp. cell like (hereafter called Prochlorococcus) and picoand nano-eukaryotes (< 20 µm, hereafter called small eukaryotes). In addition, in the UCYN treatments, C. watsonii and Cyanothece clusters were determined. The resolution of these clusters was realized by comparing the UCYN treatments cytograms with the control one. The proportion of diazotrophic cells in these clusters (i.e. the proportion of the new counts in the UCYN treatments compared to the control treatment) was > 98 and > 90 % for C. watsonii and Cyanothece, respectively. For heterotrophic bacteria (hereafter called "bacteria") samples were stained with SYBR Green II (Molecular Probes, final conc. 0.05 % [v/v], for 15 min at room temperature in the dark), in order to stain nucleic acids; then cells were characterized by two main optical signals: side scatter (SSC, related to cell size and structure) and green fluorescence (530/40, related to SYBR Green fluorescence). For the calculation of heterotrophic prokaryotes abundances, phytoplankton cells, Prochlorococcus and Synechococcus particularly, were gated out on the basis of their chlorophyll a content (red fluorescence; Sieracki et al., 1995). All data were collected in log scale and stored in list mode using the CellQuest software (BD Biosciences). Data analysis was performed a posteriori using SUMMIT v4.3 software (Dako).
The abundance of T. erythraeum added to the natural planktonic assemblage was monitored microscopically: 300 mL from the 4.5 L bottles were filtered on a 10 µm polycarbonate filter in each triplicate bottle. The cells were fixed with paraformaldehyde (2 % final concentration) for at least 1 h at 4 • C and stored at −20 • C until counting using an epifluorescence microscope (Zeiss Axioplan, Jana, Germany) fitted with a green (510-560 nm) excitation filter.

N 2 fixation rates determination
For net N 2 fixation, 2 L from each 4.5 L bottle were filtered onto precombusted (450 • C, 4 h) GF/F filters. Filters were stored at −20 • C and dried at 60 • C for 24 h before analysis. The particulate organic N (PON) content and PON 15 N isotopic enrichment of each filter were measured by continuousflow isotope ratio mass spectrometry coupled to an elemental analyser (EA-IRMS) using an Integra-CN mass spectrometer. The analytical precision associated with the mass deter-Biogeosciences, 13, 4005-4021, 2016 www.biogeosciences.net/13/4005/2016/ mination averaged 2.8 % for PON. The analytical precision associated with 15 N was ±0.0010 atom % 15 N for a measured mass of 0.7 µmol N. The particulate inorganic N contribution was not taken into account. N 2 fixation rates were calculated according to Montoya et al. (1996). We considered the results to be significant when 15 N excess enrichment was higher than three times the standard deviation obtained with time zero samples (n = 3).

DDN released to the dissolved pool
300 mL of the filtrate obtained during N 2 fixation filtrations was recovered and stored in 500 mL SCHOTT glass flasks, poisoned with HgCl 2 (final concentration 10 µg L −1 ) and stored at 4 • C for further measurement of the 15 Nenrichment of the dissolved pool. This was achieved using the two-step diffusion method extensively described in Berthelot et al. (2015a) and derived from Slawyk and Raimbault (1995). This method enables the differentiation of the NH + 4 and DON pools and measures their respective 15 Nenrichment. It should be noted that in the DON recovery step, NO x were also recovered. However, NO x concentrations were very low during our experiments (< 0.2 µmol L −1 ) with respect to DON concentrations (∼ 4.5 µmol L −1 ). Furthermore, they were unlikely to be released by diazotrophs, thus unlikely 15 N-enriched. Nitrification, that converts NH + 4 to NO − 3 at rates rising 5-10 nmolL −1 d −1 in N-depleted surface waters (e.g. Yool et al., 2007) may have contributed to the underestimation of the transfer of DD 15 N in the NH + 4 pool and to an overestimation of the DD 15 N in the DON pool. Nevertheless, in surface water, nitrification fluxes are found to be several orders of magnitude lower than NH + 4 regeneration (Raimbault and Garcia, 2008) and were thus neglected in the interpretation of the results. Net DDN release rates were calculated according to Berthelot et al. (2015a).

Organic and inorganic nutrient analyses
Samples for NH + 4 concentrations determination were collected in duplicate in 40 mL SHOTT flasks and NH + 4 concentrations were measured according to Holmes et al. (1999) using a trilogy fluorometer (Turner Design, detection limit = 3 nmol L −1 ). Samples for inorganic nutrients were collected in triplicate in 20 mL acid washed scintillation vials, poisoned with HgCl 2 (10 µg L −1 final concentration) and stored in the dark at 4 • C until analysis. NO x and PO 3− 4 concentrations were determined by standard colorimetric procedures (Aminot and Kérouel, 2007) on a segmented flow auto-analyzer. The quantification limit was 0.05 µmol L −1 . Samples for determination of DON concentrations were collected in 40 mL SHOTT flasks after filtration onto combusted GF/F filters (450 • C, 4 h) and stored at −20 • C until analysis. Concentrations were measured by wet oxidation according to Pujo-Pay and Raimbault (1994).

Cell sorting and sampling for nanoSIMS analyses
For flow cytometry cell sorting and subsequent analysis using nanoSIMS, samples were collected as follows to preconcentrate cells and facilitate cell sorting: for each treatment, 300 mL of each triplicate from the 4.5 L bottle were pooled and filtered onto 0.2 µm pore size 47 mm polycarbonate filters. Filters were quickly placed in a 5 mL cryotube ® filled with 0.2 µm filtered seawater with PFA (2 % final concentration), for at least 1 h at room temperature in the dark. The cryovials were vortexed, for at least 10 s, in order to detach the cells from the filter and were stored at −80 • C until analysis. Cell sorting was performed on a Becton Dickinson Influx ™ Mariner (BD Biosciences, Franklin Lakes, NJ) high speed cell sorter of the Regional Flow Cytometry Platform for Microbiology (PRECYM), hosted by the Mediterranean Institute of Oceanography, as described in Bonnet et al. (2016c). Planktonic groups were separated using the same clusters as for the phytoplankton abundance determination as described above. After sorting, the cells were recovered in Eppendorf tubes and immediately filtered onto a 0.2 µm pore size 25 mm filter. Particular care was taken to drop the cells on the surface as small as possible (∼ 5 mm in diameter) to ensure the highest cell density possible to facilitate further nanoSIMS analyses. In the UCYN treatments, additional "diazotroph" sort gates were defined. The gates were delimited around the new populations that appeared in the UCYN treatments, compared to the control.
Large phytoplanktonic cells (T. erythraeum and diatoms) were visible and easily recognized on the CCD camera of the nanoSIMS and thus did not require any cell sorting step. Thus, to recover these cells, 300 mL of each triplicate 4.5 L bottle were pooled together and filtered on 10 µm pore size 25 mm polycarbonate filters. The cells were fixed with PFA (2 % final concentration) for at least 1 h at ambient temperature. The filters were then stored at −20 • C until nanoSIMS analyses.

NanoSIMS analyses and data processing
NanoSIMS analyses were performed using a NanoSIMS N50 at the French National Ion MicroProbe Facility according to Musat et al. (2008) and Bonnet et al. (2016c). Briefly, a ∼ 1.3 pA Cesium (16 KeV) primary beam focused onto ∼ 100 nm spot diameter was scanned across a 256 × 256 or 512 × 512 pixel raster (depending on the image size) with a counting time of 1 ms per pixel. Samples were presputtered prior to analyses with a current of ∼10 pA for at least 2 min to achieve sputtering equilibrium and insure the analysis to be performed inside the cells by removing cell surface. Negative secondary ions ( 12 C − , 13 C − , 12 C 14 N − , 12 C 15 N − and 28 Si − ) were collected by electron multiplier detectors, and secondary electrons were also imaged simultaneously. A total of 10-50 serial quantitative secondary ion images were generated, that were combined to create the fiwww.biogeosciences.net/13/4005/2016/ Biogeosciences, 13, 4005-4021, 2016 nal image. Mass resolving power was ∼ 8000 in order to resolve isobaric interferences. From 20 to 100 planes were generated for each cells analyzed. NanoSIMS runs are timeintensive and not designed for routine analysis, but at least 20 cells from each community were analysed to assess the variability in isotopic composition under the same conditions. Thus, for diatoms only the three dominant species present in our experiment and previously counted microscopically were analysed. Data were processed using the LIMAGE and Look@NanoSIMS (Polerecky et al., 2012) software. Briefly, all scans were corrected for any drift of the beam and sample stage during acquisition. Isotope ratio images were created by adding the secondary ion counts for each recorded secondary ion for each pixel over all recorded planes and dividing the total counts by the total counts of a selected reference mass. Individual cells were easily identified in nanoSIMS 12 C, 14 N and 28 Si images that were used to define regions of interest (ROIs) around individual cells. For each ROI, the 15 N-and 13 C enrichment were calculated. In total, almost 1000 ROIs were used for this study.

Cell-specific biomass and DDN transfer calculations
The biomass of the added diazotrophs was measured at T0 by filtering an aliquot of each culture on a precombusted GF/F filter for PON determination as described above. The total biomass was divided by the number of cells determined microscopically to obtain the cell-specific biomass. For diatoms, the biovolume of the three most abundant diatom taxa (Chaetoceros spp., Bacteriastrum spp. and Thalassionema nitzschioides) was estimated by measuring their cross, apical and transapical sections in order to calculate their biovolume according to Sun and Liu (2003). At least 50 measurements were performed for each diatom taxon. Biovolume was then converted to N cellular content according to Smayda et al. (1978) and using a C : N ratio of 6.6 : 1 (Redfield, 1934). These three taxa represented ∼ 75 % of the total diatom abundance in this experiment. The remaining 25 % was mainly composed of smaller diatoms (e.g. Pseudo-Nitzschia spp., Cylindrotheca spp. and Leptocylindrus spp.) that probably weakly contributed to the total diatom biomass.
For Synechococcus, the C content reported in Buitenhuis et al. (2012) was used (255 fg C cell −1 ) and converted into N content according to the Redfield ratio of 6.6 : 1 leading to a value of 3.2 ± 0.9 fmol N cell −1 . For bacteria, the average N content of 0.15 ± 0.08 fmol N cell −1 (Fukuda et al., 1998) was assumed. For the small eukaryotes, the cellular N content of 9.2 ± 2.9 fmol cell −1 was used as reported in Gregori et al. (2001). The cellular N content of each group multiplied by their abundances allowed the calculation of the biomasses associated with each plankton group.
The DD 15 N cell-specific N 2 fixation and transfer (in nmol L −1 48 h −1 ) that depict the amount of 15 N 2 transferred from diazotrophs towards the non-diazotrophic plankton was calculated for each plankton group analysed as follows: where R cell is the mean 15 N-enrichment of individual cells (in atom %) after 48 h of incubation, R N 2 is the 15 Nenrichment of the 15 N 2 in the dissolved pool (in atom %), N con is the cellular N content (in nmol N cell −1 ) and A is the plankton group specific abundance (in cell L −1 ).

Statistical analyses
The effect of the diazotrophs treatments on the biomass associated with non-diazotrophs was tested using an Tukey HSD (honest significant difference) test. The differences in the 15 N-enrichment of cells between the different treatments and the natural abundance were tested using an unpaired nonparametric Mann-Whitney test, as the dispersion of values did not follow a normal distribution pattern. The statistical significance threshold was 5 % (p < 0.05). All the uncertainties associated with the parameters measured were taken into account and propagated over the different computations made.

Plankton abundance and biomass
At the start of the experiment (T0), (i.e. ambient waters in which the DDN transfer experiment was performed), diatoms dominated the micro-phytoplanktonic community (89 % of the total abundance), mainly driven by the contribution of Chaetoceros spp. (6130 cells L −1 ), Thalassionema spp.
(5345 cells L −1 ) and Bacteriastrum spp. (2391 cells L −1 ), which together represented ∼ 75 % of the total diatom community (Table S1 in the Supplement). Dinoflagellates were an order of magnitude less abundant than diatoms and were mainly composed of Gymnodinium spp. and Gyrodinium spp. Few Trichodesmium spp. filaments were observed in the natural assemblage at abundances lower than 40 trichomes L −1 . Ciliate abundance was 430 cells L −1 including 40 to 100 tintinnids cells L −1 . The initial abundance of Synechococcus, Prochlorococcus, small eukaryotes and bacteria determined by flow cytometry was 5.4 ± 1.1×10 4 , 2.2 ± 0.4×10 4 , 1.4 ± 0.1 × 10 3 and 5.9 ± 1.5 × 10 5 cells mL −1 , respectively (Table S1). Converted to biomass, Synechococcus dominated to phytoplanktonic biomass at T0 (120 ± 40 nmol N L −1 ), followed by bacteria (90 ± 40 nmol N L −1 ) and diatoms (40 ± 14 nmol N L −1 ). The biomass associated with small eukaryotes and Prochlorococcus together represented less than 10 nmol L −1 (i.e. 3 % of the total biomass). The dinoflagellate and ciliate biomass values were 1-2 orders of magnitude lower than the diatom biomass, respectively, and were thus not considered in detail in this study.  Figure 1. Relative increase of biomass associated with nondiazotrophic plankton groups considered in this study in the three diazotrophs-amended treatments relative to the control (%) after 48 h of incubation. Errors bar represent the standard deviations on triplicate incubations of both diazotrophs-amended treatments and control treatment. * Depict significant increase in biomass (unpaired Tukey HSD test, at 95 % levels of confidence).
In the control treatment after 48 h of incubation, the abundance of total diatoms and dinoflagellates increased by a factor of 2.3 and 1.9, respectively, while the abundances of bacteria remained stable and Synechococcus and Prochlorococcus abundances decreased by a factor of 1.4 and 1.3 respectively (Table S1). In the diazotrophs-amended treatments, the abundance of added diazotrophs decreased slightly in the T. erythraeum treatment (from 5 × 10 3 to 3.9 ± 0.5 × 10 3 trichomes L −1 ) and remained stable around 1 × 10 6 cells L −1 in the UCYN treatments (Table S1).
After 48 h of incubation, the biomass associated with nondiazotrophs increased in all the diazotrophs-amended treatments compared to the control (Fig. 1). The highest increase was observed in the T. erythraeum treatment (62 ± 39 %), mainly driven by a bacterial biomass increase of 90 ± 6 % and to a lesser extent by a Synechococcus (47 ± 22 %) and diatom (37 ± 17 %) biomass increase (Figs. 1 and 2). In the C. watsonii and Cyanothece treatments, the increase of biomass associated with non-diazotrophic plankton was 39 ± 39 and 35 ± 46 %, respectively. It was mainly driven by bacterial (58 ± 12 %), Synechococcus (23 ± 10 %) and diatom (30 ± 16 %) biomass increase in the C. watsonii treatment, and by bacterial biomass increase only (116 ± 16 %) in the Cyanothece treatment. The effect of diazotrophs on the biomass of small eukaryotes was less noticeable.
In all the treatments, the sum of the N biomass associated with every group of plankton was in good agreement with the actual PON concentrations measured by EA-IRMS after 48 h, indicating that the cellular N contents used in this study (described in Sect. 2) are realistic (Fig. 2).

N 2 fixation rates and DDN release
Net N 2 fixation rates determined by EA-IRMS in the control treatment were 1.5 ± 0.1 nmol L −1 48 h −1 (Fig. 3). This N 2 fixation was attributed to the diazotrophs already present in the natural assemblage (probably Trichodesmium spp. that were found at low abundances in the control, data not shown). In the diazotroph-amended treatments, net N 2 fixation rates were 10 to 40 times higher than in the control, indicating the that diazotroph blooms artificially induced worked well: ∼ 60 nmol L −1 48 h −1 in the T. erythraeum and C. watsonii treatments and 16 nmol L −1 48 h −1 in the Cyanothece treatment (Fig. 3). The DDN released to the dissolved pool by diazotrophs represented 16.1 ± 6.7 % of the total N 2 fixation (where total N 2 fixation is defined as the sum of N 2 fixed recovered in the PON, DON and NH + 4 pools) in the T. erythraeum treatment, 13.8 ± 1.9 % in the C. watsonii treatment, 30.5 ± 10.4 % in the Cyanothece treatment and 66.0 ± 21.9 % in the control treatment. In all cases, most of the 15 N released in the dissolved pool after 48 h of incubation was under the form of DON, which represented 77 to 81 % of the total N release in the diazotrophs-amended treatments without any differences between the treatments. The NH + 0.239 atom % (n = 18) and 2.501 ± 0.300 atom % (n = 46), respectively (Fig. 5). T. erythraeum 15 N-enrichment averaged 1.147 ± 0.233 atom % (n = 68). The 13 C-enrichment was similar for T. erythraeum (3.316 ± 0.634 atom %) and C. watsonii (3.124 ± 0.670 atom %) and higher for Cyanothece (4.612 ± 0.837 atom %). The correlation between 13 Cenrichment and 15 N-enrichment was significant for T. erythraeum (r 2 = 0.50, p < 0.001, n = 68), weaker but still significant for C. watsonii (r 2 = 0.39, p = 0.005, n = 18), and not significant for Cyanothece (r 2 = 0.01, p = 0.500, n = 46).
The amount of DD 15 N transferred to non-diazotrophs corrected from N 2 fixation detected in the control treatment was higher in the T. erythraeum treatment (9.5 ± 4.9 nmol N L −1 ) compared to the C. watsonii and Cyanothece treatments, where it was 4.1 ± 2.3 and 1.2 ± 0.9 nmol N L −1 , respectively. It represented 11.7 ± 4.4 % of total N 2 fixation in the T. erythraeum treatment and was significantly higher than in the C. watsonii (5.8 ± 2.7 %) and Cyanothece treatments (4.9 ± 2.4 %) ( Table 1).

Discussion
The fate of DDN in the marine food web has been poorly studied, mainly due to technical limitations. Using 15 N and 13 C labeling coupled with cell sorting by flow cytometry and nanoSIMS analyses at the single cell level, we were able to trace the transfer of DD 15 N from the diazotrophs to the dissolved pool and to the non-diazotrophic plankton, and compare the DD 15 N transfer efficiency as a function of the diazotroph groups dominating the community.

Cell-specific photosynthesis and N 2 fixation
Cell-specific N 2 fixation rates measured using nanoSIMS are in the range of previous N 2 fixation rates measured in cultures using conventional N 2 fixation methods for the same strains of T. erythraeum, C. watsonii and Cyanothece (Berthelot et al., 2015a). This confirms the ability of nanoSIMS to accurately measure N 2 fixation rates, as previously shown in former studies (Finzi-Hart et al., 2009;Foster et al., 2013;Ploug et al., 2010). The high N 2 fixation rates induced by the inoculation of diazotrophs in the natural planktonic community (7-30 nmol N L −1 d −1 ) are representative of those reported in the southwestern Pacific region under blooming conditions (Berthelot et al., 2015b;Bonnet et al., 2015;Garcia et al., 2007). Thus, the artificial diazotroph blooms induced for the purpose of this study provided realistic conditions to study the DDN transfer to non-diazotrophic plankton.
The significant correlation between 13 C-and 15 Nenrichments in T. erythraeum cells analyzed after 48 h of incubation argue that both PP and N 2 fixation occur simultaneously within the cells (Fig. 5). This appears in opposition   with the idea of the cells specialization in N 2 fixation (called diazocytes) where high respiration rates and degradation of glycogen and gas vacuoles reduce the O 2 concentration enabling the expression of nif genes allowing daytime N 2 fixation (Bergman and Carpenter, 1991;Berman-Frank et al., 2001;Sandh et al., 2012). However, it has to be noticed that after 48 h of incubation with the tracers, it is highly probable that both 15 N and 13 C have been exchanged between cells, leading to a homogenization of the cells isotopic enrichments.
More surprisingly, the coupling between 13 C-and 15 Nenrichments for individual UCYN cells after 48 h of incubation is weaker than for T. erythraeum cells, in particular for Cyanothece. This appears counter-intuitive as UCYN are supposed to perform both N 2 fixation and photosynthesis within the same cell. This uncoupling suggests that UCYN cells might be at least partially specialized in photosynthesis or N 2 fixation, similarly to Trichodesmium spp. These results confirm the patterns already observed for C. watsonii (Foster et al., 2013). In addition, the weaker correlation between 13 C-and 15 N-enrichments in UCYN cells compared to T. erythraeum also suggests weaker extracellular fixed N and C exchanges between cells. These differences might be the result of the greater spatial proximity of Trichodesmium spp. cells within colonies and filaments compared to free living UCYN cells in the water column. According to this vision, the high production of extracellular polymeric substances observed in different C. watsonii strains (Sohm et al., 2011;Webb et al., 2009) might be a strategy to agglomerate the free living UCYN together to form colonies (Bonnet et al., 2016a;Foster et al., 2013), ensuring a spatial proximity and thus facilitating the exchange of metabolites between cells.

DDN release to the dissolved pool
The DD 15 N released to the dissolved pool after 48 h accounted from 7 to 17 % of total N 2 fixation over the three diazotroph-amended treatments. These values are at the lower end of values (10-80 %) reported in Trichodesmium spp. blooms in the tropical Atlantic (Glibert and Bronk, 1994;Mulholland et al., 2006), southwestern Pacific (Bonnet et al., 2016c) or in mixed diazotroph assemblages of the North Pacific (Konno et al., 2010) and the Atlantic ocean Figure 5. 15 N-enrichment (atom %) measured in T. erythraeum (red), C. watsonii (green) and Cyanothece (blue) cells relative to the 13 C-enrichment. The coloured line are the linear regressions for T. erythraeum (red), C. watsonii (green) shown with their respective r-squared and p-values. Regression is not significant for Cyanothece and thus not shown on the plot. Box plots of 13 C-and 15 N-enrichments are shown, following the same colour code, on horizontal and vertical axes, respectively. (Benavides et al., 2013). In contrast, these values of N release are at least 1 order of magnitude higher than those reported in unialgal cultures (< 2 %) for the same strains as those studied here (Berthelot et al., 2015a). It is probable that, in culture, the cells are maintained in optimal growth conditions (exponential growth phase, appropriate light, temperature and nutrient conditions) and optimize the N use, either through a low excretion rate of DDN or through an efficient uptake of DDN . Conversely, in the field, the sampling does not necessarily occur during the exponential growth phase, and exogenous factors may affect the release of DDN, such as viral lysis (Hewson et al., 2004;Ohki, 1999) and sloppy feeding (O'Neil et al., 1996). In this study, the diazotrophs added to natural seawater were healthy but may have been affected by exogenous factors after inoculation, leading to a moderate proportion (7-17 %) of DDN released in the dissolved pool. These results indicate that the proportion of N 2 fixed released in the dissolved compartment both depends on the cell status and on exogenous factors more than the type of diazotrophs involved in N 2 fixation, as previously stated by Berthelot et al. (2015a).

DDN transfer efficiency and pathways
T. erythraeum transferred ∼ 12 % of DD 15 N towards the nondiazotrophic plankton. This is in good agreement with previous estimates by Bonnet et al. (2016c) using the same methodology, who report a DD 15 N transfer of 7 to 12 % in naturally occurring Trichodesmium spp. blooms. These results confirm that Trichodesmium spp. enhances the development of non-diazotrophic plankton, as already suggested by the frequent observations of co-occurrence or succession of Trichodesmium spp. and non-diazotrophs, particularly in N-depleted environments (Devassy et al., 1979;Dore et al., 2008;Lee Chen et al., 2011;Lenes and Heil, 2010).
The DD 15 N transfer was half as efficient (4-5 %) in the UCYN treatments compared to the T. erythraeum treatment (Table 1). This is in good agreement with the lower increase in plankton biomass associated with non diazotrophs in the UCYN treatments compared to the T. erythraeum treatment (Fig. 2). The ecology of UCYN is less characterized than that of Trichodesmium spp. and data on their co-occurrence with non-diazotrophic plankton in the ocean are scarce. In this issue, Bonnet et al. (2016a) used the single cell approach described here during a natural occurring bloom of UCYN-C (closely related to Cyanothece spp.) in the New Caledonian lagoon and measured a DD 15 N transfer of 21 ± 4 % of the total N 2 fixation, mainly towards pico-planktonic communities. The DD 15 N transfer reported in the latter study is ∼ 3 times higher than in the present study. This discrepancy may result from the physiological differences between the UCYN-C ecotypes involved in both studied, and/or from the DDN release from diazotrophic cells that is potentially higher is natural communities compared to cultured cells as discussed above. The bloom reported in Bonnet et al. (2016a) study co-occurred with a doubling of Synechococcus and pico-eukaryotes abundances, as well as an increase of diatoms (Leblanc et al., 2016) and PP (Berthelot et al., 2015b). Crocosphaera-like cells observed in association with the diatom Climacodium sp. (Carpenter and Janson, 2000) have also been shown to transfer the recently fixed N 2 towards the host diatom cell (Foster et al., 2011). All these data confirm that UCYN are able to provide DDN to non-diazotrophic plankton and thus promote marine productivity in N-depleted areas.
The transfer of DDN towards phytoplankton or bacteria requires the release of N in the dissolved pool. Surprisingly, the total amount of DD 15 N recovered in the dissolved pool was not a good indicator of the DD 15 N transfer efficiency: the highest release of DDN was measured in the Cyanothece treatment (16.6 ± 4.9 % of the total N 2 fixation) and led to the lowest DD 15 N transfer efficiency (4.9 ± 2.4 % of the total N 2 fixation). In the T. erythraeum and in C. watsonii treatments, the proportion of DD 15 N recovered in the dissolved pool was lower (10.3 ± 4.7 and 7.0 ± 3.0 % of the total N 2 fixation, respectively) but led to higher DD 15 N transfer efficiencies (11.7 ± 4.4 and 5.8 ± 2.7 % of the total N 2 fixation, respectively). This suggests that the N compounds released by Cyanothece were less available for the surrounding plankton communities than the compounds released by T. erythraeum and C. watsonii.
On the opposite, the 15 NH + 4 enrichment appeared to be a relevant indicator of the DDN transfer efficiency: it was twice as high in the T. erythraeum treatment compared to the UCYN treatments (Table 2), leading to a DD 15 N transfer efficiency twice as high in the T. erythraeum treatments (Table 1). This coupling between 15 NH + 4 enrichment and transfer efficiency suggests that NH + 4 is the major form of DD 15 N that is transferred to non-diazotrophic plankton, and that the DDN released under the form of DON is likely poorly available for the surrounding planktonic communities (Knapp et , 2009). This is in good agreement with the known higher bioavailability of NH + 4 for phytoplankton compared to DON (e.g. Bradley et al., 2010;Collos and Berges, 2002). However, some DON compounds such as urea or amino acids can also be a significant source of N for planktonic communities, e.g. heterotrophic bacteria and mixotrophic plankton (Antia, 1991;Bronk, 2007). Unfortunately, the methodology used here can not asserts the importance of DON compared to NH + 4 in the DDN transfer. It should be noted that the increase of plankton biomass associated with non-diazotrophs in the present study cannot only be explained by the DD 15 N provided by N 2 fixation within the time frame of the incubation (48 h). While the DD 15 N transferred to non-diazotrophic plankton biomass ranged between 1 and 10 nmol N L −1 in the diazotrophsamended treatments, the non-diazotrophic biomass increased from 90 to 160 nmol N L −1 in the diazotrophs-amended treatments. This suggests that production was also stimulated by DDN fixed prior the incubations, that was thus not 15 N labelled.

Plankton groups benefiting from the DDN
Bacterial biomass increased from 60 to 120 % after the addition of three diazotrophs; it was the plankton group which responded the most to the diazotrophs inoculations, whatever the treatment considered ( Fig. 1 and Table S1). This is consistent with the high 15 N-enrichment of bacteria cells in T. erythraeum and C. watsonii treatments compared to the control treatment (Fig. 6). In contrast, the high bacterial biomass increase observed in the Cyanothece treatment contrasts with the relatively low 15 N-enrichment of bacterial individual cells measured in this treatment (Fig. 6). This suggests that, in the T. erythraeum and C. watsonii treatments, bacteria took advantage of the DD 15 N released during the incubation, while in Cyanothece treatment, bacteria may have mainly relied on DDN fixed prior to the beginning of the incubation. This is consistent with the higher ac-cumulation of DD 15 N in the DON pool in the Cyanothece treatment compared to the two other treatments, indicating that the DON compounds released by Cyanothece are likely less bio-available for the planktonic community compared to those released by T. erythraeum and C. watsonii.
The presence of bacteria in Trichodesmium spp. colonies has been widely studied (Hewson et al., 2009;Hmelo et al., 2012;Nausch, 1996;Paerl et al., 1989;Rochelle-Newall et al., 2014;Sheridan et al., 2002). Trichodesmium spp. harbours high heterotrophic bacterial activity (Nausch, 1996;Tseng et al., 2005) and abundance is found to be at least 2 orders of magnitude higher in Trichodesmium spp. colonies than in surrounding waters (Sheridan et al., 2002). Associations between bacteria and UCYN are less documented. However, similarly to Trichodesmium spp., tight relationships may occur between UCYN and bacteria, as suggested by the significant increases of bacterial abundances in the UCYN treatments, but further investigations would be needed to understand the nature of their interactions.
Phytoplankton (diatoms, Synechococcus and small eukaryotes) was also stimulated by the diazotroph blooms, although to a lower extend compared to bacteria in all treatments (Fig. 1). However, the increase in phytoplankton biomass in the T. erythraeum and C. watsonii treatments together with the significant 15 N-enrichments in diatoms and Synechococcus (Fig. 6) argue that phytoplankton also took advantage of the DDN. This confirms the ability of diazotroph to promote non-diazotrophic primary producers as suggested by previous studies (e.g. Bonnet et al., 2016c;Devassy et al., 1979;Lee Chen et al., 2011;Lenes and Heil, 2010). The enhancement of large phytoplanktonic cells such as diatoms by DDN observed within the timespan of this study reveals the tight relationship that may occurs between the new production fuelled by diazotrophy and particle export in oligotrophic areas.
In the present study, the plankton community composition remained relatively stable in comparison to the Bonnet et al. (2016c) study, in which the authors observed sys-tematic shifts from Trichodesmium spp. biomass towards diatom biomass. This difference is probably linked to the Trichodesmium cells status. In the Bonnet et al. (2016c) study, most of the analyzed Trichodesmium spp. cells were decaying, leading to the release of micromolar concentrations of NH + 4 accumulating in the dissolved pool, whereas the main form of DDN released in the present study was DON. This rapid increase in NH + 4 bio-availability was benefiting to diatoms, which are known to be highly competitive under high nutrient concentrations (Chavez and Smith, 1995;Kudela and Dugdale, 2000;Miller and Wheeler, 2012;Smetacek, 1998;Wilkerson et al., 2000). This synchronized destruction of the colonies has been shown to be possible within a few hours, mediated by programmed cell death (Berman-Frank et al., 2004). Furthermore, the dense bloom forming behaviour and maintenance of Trichodesmium spp. at the surface due to their positive buoyancy (Romans et al., 1994) may be an additional feature that helps the constitution of locally rich N layers promoting the diatom development. In the present study, the T. erythraeum colonies added to the incubation bottles were in exponential growing phase and the relative stability of their abundance throughout the experiment indicates no substantial cell breakages. The amount of DDN transfer in our study was thus mainly mediated by the release of NH + 4 and DON during active N 2 fixation rather than the N release due to cell breakage and was thus much more limited than in the Bonnet et al. (2016c) study, leading to attenuated changes in the planktonic community composition.

Conclusions and ecological implications
This study reveals the various short term fates of DDN in the ocean and highlight the complex interactions between diazotrophs and their environment. First, it shows that the DDN released by diazotrophs in the dissolved pool as NH + 4 is quickly transferred to non-diazotrophic plankton while the DDN released as DON is mostly accumulated in the dissolved pool. Second, the DDN transfer efficiency towards the non-diazotrophic plankton depends on the diazotrophs involved in N 2 fixation: it is twice as much for T. erythraeum compared to the DDN transfer associated with UCYN strains. This implies that T. erythraeum would be more efficient at promoting non-diazotrophic marine productivity in N-depleted areas than UCYN are. Finally, the results presented here suggest that diazotrophic activity first promotes heterotrophic plankton but also autotrophic plankton, albeit to a lower extent. Taken together, theses results show that the fates of DDN are diverse and would need further investigation, in particular in the vast open-ocean regions where primary productivity extensively depends on diazotrophy.
The Supplement related to this article is available online at doi:10.5194/bg-13-4005-2016-supplement.
Author contributions. S. Bonnet and H. Berthelot designed the experiments and S. Bonnet and H. Berthelot carried them out. All authors analyzed the samples. H. Berthelot prepared the manuscript, which was amended by S. Bonnet.