Determination of total N2 fixation rates in the ocean taking into account both the particulate and filtrate fractions

Using theN2 tracer method and high-sensitivity δ15N analytical systems, we determined N 2 fixation rates for ocean samples by dividing them into particulate ( >0.7 μm) and filtrate ( <0.7 μm) fractions. While N2 fixation in the filtrate fraction had been ignored in previous studies, we found a significant N2 fixation rates in the filtrate fraction in our study. The areal N2 fixation rates in the western North Pacific Ocean estimated from the particulate fraction varied from <1 to 160 μmol N m−2 d−1, and those rates estimated from the filtrate fraction ranged from<0.5 to 54 μmol N m−2 d−1. Thus, N2 fixation in the filtrate fraction accounts for on average 50% (ranging from<10% to 84%) of the total N2 fixation rates. If these results are confirmed generally in the ocean, the new total N 2 fixation flux, which includes fixation in the filtrate fraction, possibly doubles the original estimates; therefore, the revised influx may reduce the imbalance in the global oceanic fixed nitrogen budget.


Introduction
Over the last three decades, the global budget of oceanic fixed nitrogen (NO − 3 , NO − 2 , NH + 4 , particulate organic nitrogen (PON), and dissolved organic nitrogen (DON)) has been extensively studied as one of the representative nutrients that regulate primary production in the ocean (e.g. Wada et al., 1975;Codispoti and Christensen, 1985;Gruber and Sarmiento, 1997;Brandes and Devol, 2002). The Correspondence to: U. Konno (utakonno@mail.sci.hokudai.ac.jp) total fixed nitrogen is predominantly controlled by the total influx of fixed nitrogen through N 2 fixation and by the total efflux of fixed nitrogen through denitrification (Codispoti et al., 2001;Brandes and Devol, 2002;Deutsch et al., 2004). However, the estimated values of both fluxes obtained in different studies are highly variable (e.g., Wada et al., 1975;Liu, 1979;Codispoti and Christensen, 1985;Gruber and Sarmiento, 1997;Codispoti et al., 2001;Brandes and Devol, 2002;Capone and Knapp 2007;Naqvi et al., 2008). Earlier estimates have revealed the efflux of 95 ± 20 Tg N yr −1 through sedimentary denitrification (Gruber and Sarmiento, 1997) and the influx of 110 ± 40 Tg N yr −1 through oceanic N 2 fixation (Codispoti et al., 2001). These values almost balance the total fixed nitrogen budget. However, recent studies on denitrification in sediments and water columns have revealed a considerably large efflux of 275-450 Tg N yr −1 (Codispoti et al., 2001;Brandes and Devol, 2002) that substantially exceeds the influx. Therefore, a larger influx by N 2 fixation can be expected to balance the oceanic fixed nitrogen budget (e.g. Altabet, 2007;Brandes et al., 2007;Codispoti et al., 2007).
The two most commonly used incubation methods to estimate the N 2 fixation rate are the 15 N 2 tracer method and the acetylene (C 2 H 2 ) reduction method. The former method measures the 15 N uptake rate of diazotrophs by determining the temporal variation in the nitrogen isotope compositions (δ 15 N = ( 15 N/ 14 N) sample /( 15 N/ 14 N) AirN2 − 1) of diazotrophs incubated under 15 N 2 (Montoya et al., 1996). The GF/F filter (pore size = 0.7 µm) has been traditionally used to gather diazotrophs as particulate organic matter for the 15 N 2 tracer method, because glass fiber filters are characterized by lower N blanks than other filters (in this paper, unless otherwise 2370 U. Konno: Determination of total N 2 fixation rates in the ocean noted, we have used the GF/F filter and defined the border between particulate and filtrate fractions to be 0.7 µm) (e.g. Montoya et al., 1996;Zehr et al., 2001;Mulholland et al., 2006;Shiozaki et al., 2009). Glibert and Bronk (1994) found that roughly half of recently fixed N could be released into filtrate fraction (<0.2 µm) during the incubation of Trichodesmium thiebautii and T. erythreum collected in the Caribbean Sea. Besides, several recent studies have indicated the possibility of diazotrophy in picoplanktonic organisms (Bird et al., 2005;Langlois et al., 2005;Church et al., 2008) that can pass through the GF/F filter. As a result, the estimated N 2 fixation rate determined only for the particulate fraction on the GF/F filter (e.g. Montoya et al., 1996;Zehr et al., 2001;Mulholland et al., 2006;Shiozaki et al., 2009) could be underestimated if a considerable amount of N is released into the filtrate fraction during incubation experiments (Bronk and Glibert, 1991;Glibert and Bronk, 1994) or active N 2 fixation in the filtrate fraction. Therefore, in addition to the particulate, one should determine the initial and final values of the concentration and δ 15 N for the filtrate in each incubation bottle in order to estimate the total N 2 fixation rates using the 15 N 2 tracer method. However, it is difficult to determine the δ 15 N value of filtrate in natural samples using the conventional elemental analyzer isotope ratio mass spectrometry (EA-IRMS) techniques (Mulholland et al., 2004;Meador et al., 2007).
On the other hand, the C 2 H 2 reduction method is used to measure the C 2 H 4 production rate through the reduction of C 2 H 2 by nitrogenase (Capone, 1993). However, this method requires the use of a conversion factor to convert the observed C 2 H 2 reduction rates to N 2 fixation rates. While the theoretical reduction ratio of C 2 H 2 :N 2 is 3:1 (mol:mol) (Montoya et al., 1996;Postgate, 1998), little evidence has been found to support the reliability of this ratio under natural conditions (Mulholland et al., 2006;Tsunogai et al., 2008). Nitrogenase-dependent H 2 evolution, which is inhibited by C 2 H 2 , results in deviations from this theoretical stoichiometry (Robson and Postgate, 1980;Mulholland et al., 2006Mulholland et al., , 2007. Therefore, the conversion factor is generally determined using the same field measurements by comparing the N 2 fixation rate calculated using both the C 2 H 2 reduction method and the conventional 15 N 2 tracer method. Although Capone and Montoya (2001) recommended a conversion factor of 4, Mulholland et al. (2006) showed that the factor was variable, with values ranging from 3.7 to 15.7 even in experiments conducted over a span of a few days. Furthermore, it is impossible to accurately estimate N 2 fixation rates using the conversion factors estimated by the conventional 15 N 2 tracer method based only on the 15 N 2 uptake rate by diazotrophs collected on GF/F, as already presented. Therefore, it is impossible to estimate the quantitative values of the total N 2 fixation rates by the C 2 H 2 reduction method.
The only way to solve the above-mentioned problem is to determine the δ 15 N values not only for the particulate but also for the filtrate during 15 N 2 tracer incubation. Recent developments in high-sensitivity δ 15 N analysis of organic nitrogen have now enabled us to determine the δ 15 N values of filtrate as well (Tsunogai et al., 2008). The primary objective of this study is to quantify the total N 2 fixation rates (N 2 fixation in both the particulate and the filtrate fractions) in the field to estimate the difference from those determined only for the particulate fraction. Our estimate would contribute to reassess the global oceanic fixed nitrogen budget. Further, in order to clarify the mechanisms of the N 2 fixation signal in the filtrate fraction, we quantify nifH gene abundances in the same field study to investigate what types of diazotrophs are dominant when significant N 2 fixation signals are detected in the filtrate fraction. If small sized (<GF/F filter) diazotrophs are dominant, direct N 2 fixation by these diazotrophs in the filtrate fraction could be the main reason, whereas recently fixed N release into filtrate fraction could be the main reason if large sized (>GF/F filter) diazotrophs are dominant. This is the first report on the estimation of the total N 2 fixation rates in the ocean, including the filtrate fraction.  Table 1. Water samples were collected from depths between 5 and 150 m using a CTD-Carousel multi-sampling system (911plus; Sea-Bird Electronics Inc.). Seawater samples were sub-sampled into 250-mL Pyrex bottles (KH06-2, KH07-2, and KH08-2) or 500-mL polycarbonate bottles (KH08-2) with septum caps without headspace. Then, 1.0 mL of 15 N 2 (99 atom %; Shoko Co. Ltd., Tokyo, Japan) was injected into each incubation bottle using a gas-tight syringe. The bottles were gently shaken and then incubated in thermostatic baths on a deck covered with screens to simulate the in situ temperature and light intensity for periods ranging from 12 to 72 h. Although the duration of incubation was variable, the incubation was mainly carried out during diurnal periods (24, 48, or 72 h) to avoid the bias caused by the day-night cycle on the N 2 fixation rate. However, two samples were incubated for three different periods of 12, 36, and 60 h during the KH06-2 expedition and these may have been biased by the day-night cycle. Therefore, we corrected the bias by adopting the relative variations between 36 and 12 h (corresponding to 24 h incubation) or those between 60 and 12 h (corresponding to 48 h incubation).  Immediately after incubation, the particulate in each incubated water sample was collected on a pre-combusted (450 • C for 4 h) Whatman GF/F filter (pore size = 0.7 µm) by gentle vacuum filtration. The pressure difference was strictly controlled to be <100 mm Hg to avoid the leakage of small particles from the filters. In addition to the samples for incubation, seawater samples without 15 N 2 addition were sub-sampled into 500-mL polycarbonate bottles and filtered within an hour after the sampling through the pre-combusted GF/F filter for natural organic nitrogen analysis in a manner identical to that for the samples for incubation.

Sample collection and incubation experiments
The <0.7 µm filtrate was collected in a light-resistant polyethylene bottle (100 ml) and frozen until analysis. The particulate collected on the filter was further washed with filtered (using a pre-combusted GF/F filter) clean seawater, placed in a plastic case, frozen instantaneously, and stored in a deep freezer (−80 • C) until analysis. For quantitative polymerase chain reaction (qPCR) assays targeting partial nifH fragments, seawater (1 L) was filtered onto 25-mm Supor filters (pore size: 0.2 µm, Pall Corporation) under gentle vacuum (<100 mm Hg). The obtained filters were frozen in a deep freezer (−80 • C) until analysis.

Geochemical data analysis
The concentrations and δ 15 N values of organic nitrogen in both the particulate and the filtrate, including those incubated under 15 N 2 addition, were analyzed using the method developed by Tsunogai et al. (2008). This method involves oxidation/reduction methods such as the oxidation of organic nitrogen to nitrate using persulfate (Knapp et al., 2005;Tsunogai et al., 2008Tsunogai et al., , 2010, reduction of nitrate to nitrite using spongy cadmium, and further reduction of nitrite to nitrous oxide using sodium azide. The total recovery rate of N was more than 90% for the samples. The blank level was <10 nmol N for the particulate (corresponding to 0.02 µmol N L −1 when the filtrate volume was 500 mL) and <1.0 µmol N L −1 for the filtrate. All the data presented herein had already been corrected for blank contributions. The standard deviation of the sample measurements was less than 0.3‰ for samples containing more than 50 nmolN and less than 0.5‰ for those containing more than 20 nmolN. For the filtrate samples, not only organic nitrogen but also inorganic fixed nitrogen (nitrate, nitrite and ammonium) were included in the determined concentrations and δ 15 N values. Because the concentrations of the fixed inorganic nitrogen, which were quantified by using an AutoAnalyzer (AACS II; Bran + Luebbe), were always low (mostly below detection levels) at the studied sites (Table 1), we neglected these contributions and interpreted that the determined values represented those of organic nitrogen. Even though the contributions of inorganic nitrogen were significant for the samples for 15 N incubation, they did not affect the final results of the N 2 fixation rates. The concentrations of particulate and filtrate nitrogen ranged from 0.11 to 0.60 µmol N L −1 and from 4.0 to 7.0 µmol N L −1 , respectively (Table 2). These values are the typical concentrations of organic nitrogen in oligotrophic oceans (e.g. Minagawa et al., 2001;Mino et al., 2002;Meador et al., 2007). 15 N enrichment was clearly observed over time in most of the filtrate and particulate samples incubated under 15 N 2 addition (Fig. 1). The enrichment in δ 15 N values ranged from 2372 U. Konno: Determination of total N 2 fixation rates in the ocean  <2 to 1090‰ in particulate fixed nitrogen and from <0.5 to 10.7‰ in filtrate fixed nitrogen. This result indicates that a part of the recently fixed nitrogen was transferred not only into particulate fraction but also into filtrate pools or that active N 2 fixation occurred in the filtrate fraction during the incubation experiment. The N 2 fixation rate was calculated for each incubation bottle using the results for both the concentration and the δ 15 N values of the filtrate/particulate fixed nitrogen using the following equation (referred from Montoya et al., 1996): where ρ denotes the N 2 fixation rate (µmol N L −1 d −1 ), [N] av denotes the average fixed nitrogen concentration in filtrate/particulate during experiments (µmol N L −1 ), T denotes the incubation time, A f denotes the final abundance ratio of 15 N (100 × 15 N/( 15 N + 14 N)) in filtrate/particulate fixed nitrogen, A 0 denotes the initial abundance ratio of 15 N in filtrate/particulate fixed nitrogen, and A N2 denotes the initial abundance ratio of 15 N in N 2 in the incubation bottle. As for the values of the initial concentration and δ 15 N of organic nitrogen, we used those determined for the natural organic nitrogen samples without adding 15 N 2 . If the increased δ 15 N values estimated from the 24 h incubation experiments were less than 2‰ for the particulate and 0.5‰ for the filtrate, we classified the N 2 fixation rates as less than the detection limit.
The vertical distributions of the N 2 fixation rates estimated from the particulate fraction during the KH06-2 expedition are shown in Fig. 2. The profiles indicate that the N 2 fixation rates at the water surface were high and that these rates linearly decreased to nearly zero at depths of ca. 100 m. Therefore, we calculated the areal N 2 fixation rates by integrating the N 2 fixation rates on a volume from the surface to 100 m depths, assuming linear attenuation toward zero with depths of up to 100 m including the stations where the estimation was limited to the surface.

Quantification of nifH gene abundance
DNA extraction was performed according to the method proposed by Short and Zehr (2005) with slight modifications. In brief, 600 µl of XS buffer (1% potassium ethyl xanthogenate; 100 mM Tris-HCl, pH 7.4; 1 mM EDTA, pH 8.0; 1% sodium dodecylsulfate; 800 mM ammonium acetate) and ca. 0.2 g of 0.1-mm glass beads were added to the vials containing the filters. The vials were placed in a bead beater (BioSpec Products) and agitated three times at 4800 rpm for 50 s. The samples were incubated at 70 • C for 60 min. After incubation, the XS buffer was transferred to a 1.5-ml microtube, vortexed for 10 s, and placed on ice for 30 min. Cell debris was removed by centrifugation at 15 000 g at 5 • C for 15 min; the supernatants were then decanted into another 1.5-ml microtube with the same amount of isopropanol. The samples were incubated at room temperature for 10 min, and the precipitated DNA was pelleted by centrifugation at 15 000 g for 15 min at 5 • C. Isopropanol was decanted, and the DNA pellets were washed with 70% ethanol, vacuum dried, and resuspended in 100 µl of 10 mM Tris-HCl (pH 8.5). The obtained samples were stored at −20 • C until further analysis.
Quantitative PCR (qPCR) assays targeting partial nifH fragments were carried out with a Thermal Cycler Dice Real Time System (TP800; TaKaRa) using primers and TaqMan probes designed by Church et al. (2005); they determined five nifH phylotypes including the cyanobacteria Crocosphaera spp. (termed Group B), an uncultivated phylotype termed Group A that was presumed to be a unicellular cyanobacterium, Trichodesmium spp., heterocystous cyanobacteria, and g-proteobacteria in the North Pacific Ocean. For each set of primers and probes set, standard curves were derived using duplicate or triplicate serial dilutions of linearized pUC18 plasmids (TaKaRa) containing the positive control insert. The number of molecules of a plasmid was estimated from the amount of DNA according to the equation derived by Short and Zehr (2005). The PCR amplification mixture solution (25 µl) contained 12.5 µl of Premix Ex Taq (Perfect Real Time, TaKaRa), 0.05 µl of each primer (final conc.: 0.2 µM), 0.1 µl of the probe (final conc.: 0.4 µM), 11.3 µl of sterile Milli-Q water, and 1 µl of DNA template. In each qPCR run, environmental DNA and no template controls (i.e. sterile Milli-Q water) were also prepared in duplicate or triplicate. The thermal cycling reactions were carried out as follows: 95 • C for 10 s, and 50 cycles of 95 • C for 5 s followed by 60 • C for 30 s.

N 2 fixation rates in particulate fraction
The areal N 2 fixation rates estimated from the particulate fraction varied from <1 to 160 µmol N m −2 d −1 during the three expeditions undertaken in the western North Pacific (Table 2). Using the conventional 15 N 2 tracer method for the particulate fraction using GF/F filters, Shiozaki et al. (2009) estimated the areal N 2 fixation rates in the western North Pacific region to be 29-152 µmol N m −2 d −1 in early spring. The areal N 2 fixation rates estimated from the particulate fraction in the present study during the early summer expedition (KH06-2 cruise; 22-160 µmol N m −2 d −1 ) agreed well with those reported by Shiozaki et al. (2009). On the other hand, the areal N 2 fixation rates estimated during the late summer expeditions (KH07-2 and KH08-2 cruises; <20 µmol N m −2 d −1 ) were lower than those estimated during the early summer expeditions. The concentration of the sea surface chlorophyll-a was lower during the late summer expeditions (0.01 to 0.05 µg L −1 ) than during the early summer expeditions (0.07 to 0.21 µg L −1 ), indicating that the late summer expeditions coincided with the post-blooming season, when nutrients are limited. Thus, the observed difference in the areal N 2 fixation rates can be attributed to the seasonal variations in the N 2 fixation rates (Sañudo-Wilhelmy et al., 2001;Moutin et al., 2005). The lowest observed N 2 fixation rates at the two stations during the late summer expeditions (Stns.19 and 20, KH08-2) ( Table 2) can also be attributed to the lack of nutrients in the post-blooming season because the observed column-integrated quantities (from the surface to a depth of 100 m) for PO 3− 4 (<2 µmol PO 3− 4 m −2 at Stn. 19 and <1 µmol PO 3− 4 m −2 at Stn. 20) were the smallest for those stations during the late summer expeditions (Table 1).
However, the column-integrated quantities of PO 3− 4 were also small at Stns. 5, 6, and 7 (KH06-2) during the early summer expedition, where active N 2 fixation (>22 µmol N m −2 d −1 ) was observed in the particulate fraction. Moutin et al. (2005) also found few direct links between the P availability and accumulation of Trichodesmium U. Konno: Determination of total N 2 fixation rates in the ocean spp. during the early summer. Therefore, explanations other than the limited availability of nutrients may be required to explain active N 2 fixation during the early summer. Owing to the ability to store P, Trichodesmium spp. is active for a few months after PO 3− 4 deficiency (Thompson et al., 1994;Moutin et al., 2005). Therefore, during early summer, the N 2 fixers were using the stored P supplied during winter or spring, while they may exhaust the stored P during the late summer when we found the direct links between the Pavailability and the N 2 fixation rate. In conclusion, the estimated values of N 2 fixation rates for the particulate fraction were representative of those observed in the western North Pacific area during each season.
The δ 15 N values of the initial particulate also supported our conclusions ( Table 2). The lower δ 15 N values of particulate (+0‰ and +0.9‰; Stns. 1 and 5 during the KH06-2 expedition, +0.8‰ and +1.5‰; Stns. 16 and 22 during KH08-2 expedition) were found at the station where a higher N 2 fixation rate in the particulate fraction was observed. Because the PON derived from N 2 fixation had nearly 0‰ of δ 15 N derived from atmospheric N 2 (Minagawa and Wada, 1986;Carpenter et al., 1997;Montoya et al., 2002), the geographical variations in δ 15 N of particulate also support the significant N 2 fixation in the particulate fraction.

N 2 fixation rates in the filtrate fraction
N 2 fixation signal was also found in the filtrate fraction. This signal could be attributed to both the active N 2 fixation into filtrate fraction (direct path) and/or the recent fixed nitrogen release into filtrate fraction (indirect path). The areal N 2 fixation rates estimated in this study for the filtrate fraction (ranging from <0.5 to 54 µmol N m −2 d −1 ) accounted for on average 50% (ranging from <10 to 84%) of the total N 2 fixation rates (Table 2). Glibert and Bronk (1994) also estimated that the rates of N release by Trichodesmium spp. into the filtrate (DON) fraction (<0.2 µm) could account for 50% on an average of the N 2 fixation rates in the particulate (PON) fraction (>0.7 µm). Furthermore, Mulholland et al. (2006) estimated the rates of release of fixed nitrogen into the filtrate (DON + DIN; <0.7 µm) fraction in the ocean on the basis of the discrepancies in the N 2 fixation rates estimated by the 15 N 2 tracer method and the C 2 H 2 reduction method for the same samples; they found that the filtrate (DON + DIN; <0.7 µm) fraction comprised 52% (ranging from 9.1% to 81%) of the total N 2 fixation by using a theoretical conversion factor of 3 for the C 2 H 2 reduction method. Both the average N 2 fixation rate for the filtrate fraction within the total N 2 fixation rate and the range of variation estimated in this study corresponded well with the estimated rate and variation in past studies. The N 2 fixation rates in the filtrate fraction estimated in this study may be highly reliable for estimating those in the ocean.
The discrepancies in the estimates for the N 2 fixation rates between the 15 N 2 tracer method and the C 2 H 2 reduc-tion method have been noted in the western North Pacific region as well. Using the C 2 H 2 reduction method, Kitajima et al. (2009) found high N 2 fixation rates of 0.5-12 nmol N L −1 d −1 in the western North Pacific region; this range is more than twice as high as that estimated for the particulate fraction in this study (0.4-3.2 nmol N L −1 d −1 during the KH06-2 expedition) based on the 15 N 2 tracer method. It is difficult to attribute the differences in the estimations to the seasonal variations because both experiments were performed during the same early summer season (May to June). Because the C 2 H 2 reduction method resulted in higher N 2 fixation rates in comparison with the 15 N 2 tracer method for the particulate fraction when the N 2 fixation rates in the filtrate fraction was significant, the systematic difference between the estimates obtained by the two different methods for the same region implies that the N 2 fixation rates for the filtrate fraction were almost as significant as those for the particulate fraction.
The δ 15 N values of the initial filtrate also support our conclusions. Within the whole δ 15 N values of the initial filtrate in surface water (ranging from +5.5‰ to +8.8‰, Table 2), which agree with those reported in a previous study in the Central Pacific region (Meador et al., 2007), the lowest δ 15 N value of the filtrate (+5.5‰; Stn. 16 during the KH08-2 expedition) was found at the station where the highest N 2 fixation rate in the filtrate fraction was observed. Because the fixation of atmospheric N 2 (δ 15 N = 0‰) produces organic nitrogen with δ 15 N values similar to atmospheric N 2 (Bourbonnais et al., 2009), the geographical variations in δ 15 N of the filtrate also support the significant N 2 fixation rates in the filtrate fraction. In conclusion, the estimated significant N 2 fixation rates in the filtrate fraction are representative of those in the western North Pacific region.

Mechanisms of N 2 fixation signal in the filtrate fraction
As already presented, the significant N 2 fixation rates observed in the filtrate fraction can be explained by the following two mechanisms: (1) N 2 fixation in the filtrate fraction by small unicellular plankton and (2) secondary release of recently fixed nitrogen into the filtrate fraction from the particulate fraction during incubation on board, which includes viral cell lysis (Hewson et al., 2004), grazing (O'Neil et al., 1996), cell death (Berman-Frank et al., 2004), or direct release of N-compounds (Glibert and Bronk, 1994). The large N 2 fixation rates in the filtrate fraction were found at high latitudes (Fig. 3). In several previous studies, it has been observed that γ -proteobacteria are more abundant in waters characterized as both cooler in temperature and richer in nutrients than waters where usual cyanobacterial N 2 fixers are dominant (Bird et al., 2005;Langlois et al., 2005;Church et al., 2008). In particular, Church et al. (2008) found that the nifH gene is actively expressed in γ -proteobacterial phylotypes at stations far north up to 44 • N in the north  The white and gray regions denote N 2 fixation in the particulate and filtrate fractions, respectively. For the stations where the N 2 fixation rates in the filtrate fraction were below the detection limit, the maximum values are shown in light gray.
Eastern Pacific region. Because bacterioplankton and/or picoplankton with sizes of ca. 0.2 to 2 µm can pass through the GF/F filter (pore size: 0.7 µm) and mix with the filtrate fraction, they can cause active N 2 fixation in the filtrate fraction.
However, the abundance of nifH gene copies determined by the quantified PCR method indicates that the dominant N 2 fixer for stations at high latitudes and exhibiting an active N 2 fixation signal in filtrate fraction is Trichodesmium spp. (Fig. 4). Although we could not directly compare the number of nifH gene copies with the N 2 fixation rates (Zehr et al., 2007), the large N 2 fixation rates observed in the filtrate fraction at high latitudes can be attributed to the active secondary release of N into the filtrate fraction from recently fixed nitrogen by Trichodesmium spp. Glibert and Bronk (1994) also found that the rates of release of N from Trichodesmium spp. into the filtrate fraction may account for on average 50% of the N 2 fixation rates in the particulate fraction. The release rate of N into the filtrate fraction corresponded well with our results for the natural samples. However, further studies are essential to confirm the precise mechanisms of active N 2 fixation and/or N 2 fixation signal in filtrate fraction.

Implication for total N 2 fixation flux in ocean
By using the N * parameter taking into account total N 2 fixation, the global oceanic N 2 fixation rates were estimated to be  . Abundance of nifH gene copies (bar charts: left axis) and the N 2 fixation rates (line charts: right axis) during the KH08-2 expedition. The dark gray, white, light gray, and black bars denote the results for Trichodesmium spp., heterocystous diazotroghs, nanoplanktonic diazotrophs, and picoplanktonic diazotrophs, respectively. The circles and squares on the line charts represent the total N 2 fixation rates (µmol N m −2 d −1 ), and the rates in the filtrate fraction, respectively. in the range of 70-150 Tg N yr −1 (e.g. Gruber and Sarmiento, 1997;Deutsch et al., 2007). However, Karl et al. (2002) noted that the N:P stoichiometry in the N * parameter method could not be accurately determined due to ignoring the dissolved organic matter (DON and DOP). On the other hand, the oceanic N 2 fixation rates were estimated to be in the range of 80-140 Tg N yr −1 on the basis of the 15 N 2 tracer method (Brandes et al., 2007); however, these estimates only accounted for the N 2 fixation rates in the particulate fraction. Using the present data, we can estimate the total N 2 fixation rates more accurately by correcting the past estimates. Using a roughly average N 2 fixation signal of 50% in the filtrate fraction that were underestimated in previous studies over oceans worldwide, the revised N 2 fixation inputs should be increased to 160-280 Tg N yr −1 . Codispoti et al. (2001) estimated the total influx and efflux of fixed nitrogen to be 287 and 482 Tg N yr −1 , respectively, for the current global fixed nitrogen budget in oceans; in this budget, the efflux exceeds the influx by ca. 200 Tg N yr −1 . The revised influx reduces the imbalance in the global fixed nitrogen budget. However, as observed during this study, the N 2 fixation rates in the filtrate fraction can be highly variable on different temporal and spatial scales. Further studies should be conducted to estimate the N 2 fixation rate in the filtrate fraction more accurately.
We found significant N 2 fixation rates in the filtrate fraction (<0.7 µm) from the western North Pacific region; N 2 fixation in this fraction had been ignored in previous studies. In our results, N 2 fixation rates in the filtrate fraction accounted for on average 50% (ranging from <10% to 84%) of the total N 2 fixation rates. The abundance of nifH gene copies determined by the quantified PCR method indicated that the large N 2 fixation rates observed in the filtrate fraction at high latitudes could be attributed to active secondary N release processes for filtrate fraction from recently fixed nitrogen by Trichodesmium. If these results are confirmed generally in the world's oceans, the new total N 2 fixation flux including the N 2 fixation signal in filtrate fraction possibly doubles the original estimates; therefore, the revised influx may reduce the imbalance in the global oceanic fixed nitrogen budget.