Aphotic N2 Fixation in the Eastern Tropical South Pacific Ocean

We examined rates of N2 fixation from the surface to 2000 m depth in the Eastern Tropical South Pacific (ETSP) during El Niño (2010) and La Niña (2011). Replicated vertical profiles performed under oxygen-free conditions show that N2 fixation takes place both in euphotic and aphotic waters, with rates reaching 155 to 509 µmol N m−2 d−1 in 2010 and 24±14 to 118±87 µmol N m−2 d−1 in 2011. In the aphotic layers, volumetric N2 fixation rates were relatively low (<1.00 nmol N L−1 d−1), but when integrated over the whole aphotic layer, they accounted for 87–90% of total rates (euphotic+aphotic) for the two cruises. Phylogenetic studies performed in microcosms experiments confirm the presence of diazotrophs in the deep waters of the Oxygen Minimum Zone (OMZ), which were comprised of non-cyanobacterial diazotrophs affiliated with nifH clusters 1K (predominantly comprised of α-proteobacteria), 1G (predominantly comprised of γ-proteobacteria), and 3 (sulfate reducing genera of the δ-proteobacteria and Clostridium spp., Vibrio spp.). Organic and inorganic nutrient addition bioassays revealed that amino acids significantly stimulated N2 fixation in the core of the OMZ at all stations tested and as did simple carbohydrates at stations located nearest the coast of Peru/Chile. The episodic supply of these substrates from upper layers are hypothesized to explain the observed variability of N2 fixation in the ETSP.


Introduction
The efficiency of oceanic carbon (C) sequestration depends upon many factors, among which is the availability of nutrients to support phytoplankton growth in the illuminated surface ocean. In particular, large amounts of nitrogen (N) are required, as it is an essential component of proteins, nucleic acids and other cellular constituents. Dissolved N in the form of nitrate (NO 3 2 ) or ammonium (NH 4 + ) is directly usable for growth, but concentrations of fixed N are low (,1 mmol L 21 ) and often growth-limiting in most of the open ocean euphotic zone [1]. Dinitrogen (N 2 ) gas dissolved in seawater, on the other hand, is very abundant in the euphotic zone (ca. 450 mmol L 21 ) and could constitute a nearly inexhaustible N source for the marine biota. However, only certain prokaryotic 'N 2 -fixers' (or diazotrophs) are able to use this N source since they can break the triple bond between the two N atoms of the N 2 molecule, and convert it into a usable form (i.e. NH 3 ) for assimilation.
The focus of much recent marine N 2 fixation research has been on the NO 3 2 -poor environments of the surface tropical ocean, where it may sustain up to 50% of 'new' primary production [2,3]. The filamentous cyanobacterium Trichodesmium spp., which is widespread in the tropical ocean and has a macroscopic growth form [4], may fix from 60 [5] to 80 Tg of N per year [6]. Until the last decade, this organism was the focus of the bulk of research as it is conspicuous and easily collected [4]. However, since then, studies of the abundance and diversity of the nifH gene required for N 2 fixation have elucidated the importance of unicellular picoand nano-planktonic cyanobacteria [7,8], extending the geographical extent of diazotrophy beyond tropical waters [9], and potentially narrowing the gap between direct measurements and geochemically-based global marine N fixation rates [5]. These molecular tools have also revealed the presence of putative noncyanobacterial diazotrophs (possessing and potentially expressing the nifH gene) in diverse aquatic environments [10], including surface seawater, hydrothermal vents and lakes [11] and references therein). In marine waters, these diazotrophs seem to be almost ubiquitous [12], but few studies e.g. [13,14] have focused on these non-cyanobacterial diazotrophs, and our knowledge of their distribution in the ocean and their biogeochemical importance for the marine N budget is still very limited.
The N budget for the global ocean is poorly constrained, with some suggestions that sinks (denitrification and anammox) exceed sources (N 2 fixation) [15]. The high energy and iron (Fe) requirements [16,17] of the N 2 fixation reaction have implied that this process occurs mainly in the large oligotrophic areas of the ocean that are depleted in fixed N, and where fixing N 2 gives an ecological advantage. This may be particularly the case in areas which receive high Fe-rich Saharan dust such as the North Atlantic [18], or which are under the influence of terrigenous and submarine Fe sources, such as the North Pacific near Hawaii [19,20] or the South West Pacific [20,21,22,23]. However, recent studies [24,25] have hypothesized that N 2 fixation might also be associated with denitrified surface waters over oxygen minimum zones (OMZs), which have measureable NO 3 2 , but are depleted in N relative to phosphorus (P). This hypothesis has been recently confirmed in the coastal surface waters of the Peruvian-Chilean upwelling [26,27] as well as throughout the eastern tropical South Pacific Ocean (ETSP) [28], where depth-integrated rates over the upper water column were comparable to those found in subtropical gyres. nifH sequences recovered from these areas within the upper 200 m of the ocean were mostly noncyanobacterial and clustered with known heterotrophic sequences [26]. This led us to explore N 2 fixation in the aphotic zone of the ETSP.
Previous studies conducted in surface waters of the ETSP indicated that N 2 fixation was highly variable in space and time, with depth-integrated rates varying from 10-to 30-fold between cruises performed at the same locations [26,28]. Although the activities of heterotrophic diazotrophs might potentially be contributing to this high temporal variability, very few studies have examined the regulation of N 2 fixation by heterotrophic bacteria in marine waters. Organic C availability has been hypothesized to control marine heterotrophic N 2 fixation [29] as a consequence of the high energy requirements of the reaction, but, to our knowledge, the effect of organic molecules on heterotrophic N 2 fixation has never been studied in OMZs.
In this study, we investigated N 2 fixation along a transect across the ETSP in 2010 and 2011 through temperature, oxygen and nutrient gradients. We quantified N 2 fixation throughout the 0 to 2000 m depth range in order to evaluate its potential biogeochemical impact on the marine N budget, and we conducted aphotic nutrient addition bioassays in the core of the aphotic OMZ in order to investigate which nutrients might control N 2 fixation in this environment. We also phylogenetically characterized the diazotrophs community composition in the core of the OMZ and how it responded to some of the nutrient amendments.

Methods
Our research was carried out during two cruises in the ETSP, aboard the R/V Atlantis in February and March 2010, and the R/V Melville in March and April 2011. Experiments were performed along a transect that began in northern Chile and ran west along 20uS, from the nutrient-rich waters at 82uW to the more oligotrophic and low-NO 3 2 waters at 100uW, and returned along 10uS (Fig. 1). No specific permissions were required for these locations/activities as both cruises took place in international waters. This study did not involve endangered or protected species. The coastal waters of this region of the ETSP are characterized by a permanent wind-driven upwelling of cool nutrient-replete water (Fig. 1), which supports high primary productivity and a persistent subsurface OMZ, where O 2 concentrations are low enough to induce the anaerobic processes of the N cycle, such as denitrification and anammox [30,31,32]. These O 2 -deficient waters are carried by Eckman transport westward beyond the limit of our transect. The ETSP is subjected to the inter-annual climactic variability of the El Niñ o-Southern

Hydrographic and nutrient measurements
Hydrographic and nutrient measurements were performed at 6 stations in 2010 and 7 stations in 2011 (Fig. 1). Vertical profiles of temperature, chlorophyll a, fluorescence and dissolved oxygen were obtained using a Seabird 911 plus CTD equipped with a model 43 oxygen sensor and a Wetlabs ECO-AFL/FL chlorophyll fluorometer. Oxygen values were calibrated by micro-Winkler [33]. Seawater samples were collected at selected depths using a rosette equipped with 24

Vertical profiles of N 2 fixation
Rates of N 2 fixation were measured using the 15 N 2 tracer method [35]. Water samples were dispensed into acid-leached 4.5-L polycarbonate bottles. During the 2010 cruise, this work was exploratory and unreplicated (except for nutrient addition bioassays, see below) measurements were made at 12 to 14 depths between the surface and 2000 m at stations 1, 9 and 11. During the 2011 cruise, samples were collected at stations 1, 5, 6, 7, 9 and 11 in triplicates at 12 depths between the surface and 2000 m, with a specific focus on O 2 gradients. Depths were chosen in order to sample the oxycline, at least 3 depths within the core of the OMZ, as well as an additional 3 within the second increasing oxygen concentrations below the OMZ. Most of these depths were located in the aphotic zone.
On both cruises, specific care was taken to avoid O 2 contamination and to perform incubations under strict oxygenfree conditions as described in [36]. Briefly, before each profile, the 36 4.5-L bottles were filled with deionized water, then the deionized water was flushed with argon and finally filled with seawater via tubing into the bottom of the argon-filled bottles to minimize gas exchange. Bottles were then closed with septa and spiked with 3 mL 15 N 2 (99 atom % EURISO-TOP) via a gas-tight syringe. Each bottle was shaken 30 times to fragment the 15 N 2 bubble and facilitate its dissolution. Recent work has suggested that with this method, there may be incomplete equilibration of the added 15 N 2 gas bubble with the seawater sample, resulting in a dissolved 15 N 2 concentration in the sample that is lower than the equilibrium value assumed in the calculation of 15 N 2 fixation rates [37]. This may lead to a potential underestimate of N 2 fixation rates [38,39]. Therefore, the values given in the present study should be considered as minimum estimates (discussed below). Bottles were then incubated either in on-deck incubators at irradiances specific from the sampling depth using blue screening and cooled with circulating surface seawater (photic samples), or in dark rooms at 12uC or 5uC depending of the sampling depth. After incubation, the triplicate bottles from each depth were filtered onto precombusted (4 h at 450uC) 25-mm GF/F filters. Filters were stored at 220uC until the end of the cruise, then dried for 24 h at 60uC and stored dry until mass spectrometric analysis. During the 2011 cruise, an extra 4.5-L bottle was collected at each depth of the profile, spiked with 15 N 2 and immediately filtered in order to determine the initial background d 15 N in the particulate organic N (PON) for calculations of N 2 fixation rates. During the 2010 cruise, the value of d 15 N in air (0.00366) was used as a reference value for these calculations, which may introduce a potential bias, except at Station 1 where 15 N atom % of the PON at depth was available.
Nutrient addition bioassays in the core of the OMZ Nutrient addition bioassays of N 2 fixation were performed at one single depth in the core of the OMZ (based on O 2 -CTD profiles) at 3 stations (Stations 5, 7 and 11, between 140-and 450m depth) during the 2010 cruise and at 6 stations (Stations 1, 5, 6, 7, 9 and 11, between 320-and 475-m depth-) during the 2011 cruise. All experiments were performed in triplicate and under strict oxygen-free conditions (using the argon flushing method described above) to avoid inhibition of N 2 fixation by oxygen. Immediately after collection, bottles were capped with septa and amended with nutrients via syringes. During the 2010 cruise, at each of the 3 stations, triplicate bottles were left as unamended controls, and a second set of bottles was amended with glucose to obtain a final concentration of 10 mmol L 21 . During the 2011 cruise, triplicate bottles were left as unamended controls, and a second set of triplicate bottles was amended with a mixture of three simple carbohydrate substrates (39% glucose, 29% acetate and 32% pyruvate, final total concentration of 1 mmol carbohydrate L-1) to test the effect of a source of dissolved organic C (DOC) on N 2 fixation. A third set was amended with a mixture of three amino acids as a source of both DOC and dissolved organic N (DON) (20% leucine, 23% glutamic acid and 56% alanine) to reach a final concentration of 1 mmol amino acids L 21 . The proportion of each carbohydrate and amino-acid has been chosen in order to add the same quantity of organic C in the two treatments (4 mmol L 21 ). A fourth set was amended with ATP (source of dissolved organic P, DOP) to reach a final concentration of 1 nmol L 21 , and a fifth set was amended with 8 mmol L 21 of NO 3 2 to test its potential inhibitory effect on heterotrophic N 2 fixation. Bottles were then incubated in a dark cold room at 12uC for 24 h in order to leave enough time to induce any potential nutrient stimulation. After 24 h, all bottles were spiked with 15 N 2 as described above, and incubation was continued under the same conditions for an additional 24 h. At the end of each incubation, the three treatments and control replicates were filtered as described above in order to measure N 2 fixation rates, and amplification of the nifH gene (2010 only). Samples were also collected from bottles sacrificed at time zero in order to quantify background NO x and PO 4 32 concentrations at every station. NO x concentrations were also measured just after the NO 3 2 additions in order to confirm the added concentrations at the beginning of the incubations (data not shown).

Mass spectrometric analyses
The isotopic enrichment of particulate N after the incubation of seawater with 15 N 2 was measured by continuous flow isotope ratio mass spectrometry of pelletized filters (Europa Integra-CN), calibrated every 10 samples using reference material (International Atomic Energy Agency [AIEA], Analytical Quality Control Services). The linearity of 15 N atom % as a function of increasing sample PON mass was verified as detailed in [40] on both natural and 15 N enriched material. This step is critical in ultraoligotrophic environments or deep waters, where suspended PON concentrations are low. 15 N atom % was linear (Fisher test, p,0.01) between 0.20 and 39 mmol N, which is within the range of PON measured in all of our samples (0.27 to 4.91 mmole N depending on the station and depth).
Detection and quantification limits for particulate N were calculated daily, as 3 times and 10 times the standard deviation of 15 N analysis of blanks, respectively. Detection limits ranged from 0.10 to 0.17 mmole N, and quantification limits ranged from 0.13 to 0.26 mmole N, depending on the station. The 15 N isotope enrichment of a sample was calculated using the 15 N atom % excess over the 15 N atom % in samples taken from the same station at time zero, which was determined on bottles filtered immediately after adding 15 N 2 . We considered the results to be significant when 15 N excess enrichments were greater than 3 times the standard deviation obtained with ten AIEA references ( 15 N atom % .0.0005). The quantification limit of N 2 fixation in this study was 0.01 nmol L 21 d 21 . If only one of the 3 replicate measurements was quantifiable, the average of the 3 replicates was forced equal to zero, in order to provide minimum estimates of N 2 fixation.
In order to determine areal rates, N 2 fixation measurements were trapezoidally depth-integrated from the summed products of the average of two adjacent rate measurements (including those equal to zero) with the depth interval between them. The standard deviation on the triplicates (2011 cruise) was also used for a trapezoidally depth-integration in order to obtain the standard deviation on integrated rates.

Statistical analysis
Controls and experimental nutrient treatments were compared using a 2-tailed non parametric Mann-Whitney mean comparison test (n = 3, a = 0.05, unpaired samples).

Phylogenetic characterization of diazotrophs
In order to characterize the potential diazotrophs present in the core of the OMZ that responded to the addition of glucose, nucleic acid samples were collected from triplicate bioassays during the 2010 cruise for amplification of the nifH gene. At T0 and at the termination of the experiment, bottles were immediately filtered as described in [41] onto 25-mm, 0.2-mm Supor filters (GE Osmotics, Minnetonka, MN), and immediately flash frozen in liquid N 2 . All filters were stored at 280uC thereafter.
DNA samples were extracted using the Qiagen All Prep kit (Valencia, CA), according to manufacturer's guidelines, with modifications to include freeze-thaw and bead-beating steps to disrupt the cells [42]. The wash steps of this protocol were automated using a QIAcube (Qiagen). DNA extracts were stored at 220uC until use.
Nested PCR amplification targeting a fragment of the nifH gene was carried out using degenerate primers nifH1-4 [43,44] using the reaction and thermocycling conditions described in [42]. Amplicons were purified using a QIAquick Gel Extraction Kit (Qiagen) and cloned using the TOPO TA Cloning Kit for Sequencing (Invitrogen, Carlsbad, CA) according to the manufacturer's guidelines. Purified recombinant plasmids containing partial nifH sequences were recovered from clones using the Montage Plasmid Miniprep96 Kit (Millipore, Billerica, MA) and sequenced using Sanger technology at the UC Berkeley DNA Sequencing Center. All DNA extractions and as PCR preparations were performed in a PCR-amplicon free facility at UCSC described in [42].
Sequencher 5.1 sequence analysis software (Gene Codes Corporation, Ann Arbor, MI) was used to remove vector contamination and low-quality reads from raw sequences. All resulting partial nifH sequences were imported into a curated nifH database (http://pmc.ucsc.edu/,wwwzehr/research/database/), translated into amino acid sequences, aligned to the existing hidden Markov model alignment using the Quick Align function, and nucleic acids were realigned to the aligned amino acids in the ARB software environment. Sequences generated from the nutrient addition bioassays were clustered at 97% nucleotide similarity using CD-HIT-EST [45]. Nucleic acid trees used the Jukes-Candor correction for branch length. Trees generated in ARB were exported into iTOL for the display of associated metadata. All partial nifH sequences recovered were submitted to Genbank under Accession numbers KF515738 -KF515848.

Hydrographic and nutrient profiles
During both cruises, oceanographic conditions were consistent with active wind-driven upwelling off the coast of Northern Chile and Peru (Fig. 1), associated with a vertically and horizontally extensive OMZ (Fig. 2C and 3B (Fig. 5). The overall range of rates measured over the cruise was from the detection limit to 0.80 nmol L 21 d 21 . The highest rates were measured in O 2 deficient waters at the oxyclines or in the core of the OMZ (Fig. 5 Table 1). The average integrated rate over the cruise was 317 mmol N m 22 d 21 . Integrated N 2 fixation rates in the aphotic zone accounted for 73 to 99% of the rates measured over the entire water column depending on the station. When considering all the stations, the average areal rate in the aphotic zone was 87% of the total rate over the entire water column (Table 1).
During the 2011 cruise, N 2 fixation rates were significantly greater than zero in 140 of the 216 measurements made (Fig. 3F,  Fig. 6). The overall range of rates measured was from detection limit to 0.2660.12 nmol L 21 d 21 . In the northern transect (Fig. 3F, Fig. 6), the highest rates of N 2 fixation over the vertical profiles were measured in the oxycline as in 2010, and mean rates (n = 3) reached 0.1560.13 nmol L 21 d 21 at Station 7 and 0.1960.28 nmol L 21 d 21 at Station 9 at the oxycline. At station 11, the highest rates were found in surface waters (0.2260.19 nmol L 21 d 21 ) but rates at the oxycline were also measurable (0.0660.03 nmol L 21 d 21 ). Below the OMZ (ca. 400-2000 m), rates were also measurable and ranged from 0.0060.01 to 0.2160.13 nmol L 21 d 21 , the highest rates being measured at station 9 at 1000 m depth. In the southern transect (Fig. 4F, Fig. 6  90% of total rates measured over the entire water column (Table 1) Fig. 7). At the 2 most oceanic Stations 5 and 7, glucose amendments did not result in any significant increase of N 2 fixation (p.0.05). At Station 11 near the Peruvian coast, glucose amendments resulted in a significant (p,0.05) increase in N 2 fixation rates by a factor of 3.2, to reach 0.566 0.04 nmol N L 21 d 21 (Fig. 7).
During the 2011 cruise, nutrient concentrations in the core of the OMZ (320 to 475 m) where experiments were performed ranged between 33.12 and 38.72 mmol L 21 for NO x and from 2.33 to 3.03 mmol L 21 for PO 4 . Mean N 2 fixation rates in the control bottles ranged from 0.0060.01 at Station 11 to 0.0760.01 and 0.0760.04 nmol N L 21 d 21 at Stations 9 and 5, respectively (n = 3; Fig. 8). At Stations 1 and 9, N 2 fixation rates were significantly (p,0.05) stimulated by simple carbohydrate additions

Phylogenetic characterization of diazotrophs in 2010 glucose addition bioassays
A full phylogenetic characterization of diazotrophs in the upper 200 m of the ETSP water column was performed during the same cruises and is detailed in a companion paper [42]. In this study we report the complementary phylogenetic characterization of samples from the core of the OMZ (Fig. 8). Partial nifH sequences recovered during deep glucose addition bioassays during 2010 at Stations 5, 7 and 11, indicated that diazotrophs were present in the deep waters of the OMZ. The diazotrophic community was comprised of non-cyanobacterial diazotrophs affiliated with nifH clusters 1K (predominantly comprised of a-proteobacteria), 1G (predominantly comprised of c-proteobacteria), and 3 (sulfate reducing genera of the d-proteobacteria as well Clostridium spp., Vibrio spp, etc.) (Fig. 9). Clear differences exist between OMZ diazotrophic community composition at each station. The Station 5 community was dominated by nifH cluster 1K sequences, many of which are closely related to a phylotype (94-97% nucleic acid similarity) originally reported at Hydrostation S (North Atlantic) from a depth of 1000 m (BT5167A10 (DQ481253) [47]), although a few putative c-proteobacteral (1G) sequences were also recovered that affiliated with cETSP3, a cluster recovered from the ETSP [42]. Although the lowest number of total sequences was recovered from Station 7, they were mainly affiliated with cluster 1G, along with a few 1K sequences. In contrast, clone libraries from Station 11 were dominated by cluster 3 sequences, along with a few 1G sequences, but no 1K sequences (Fig. 9).
Despite being prevalent in clone libraries, both aETSP1 and cIII-ETSP groups were not detected at abundances great enough to quantify using TaqmanH qPCR assays during the course of these experiments (see Figuer S1 in File S1). Because the abundances of these targets did not increase as a result of nutrient amendments, it is difficult to speculate whether any of them were responsible for the increased N 2 fixation rates we measured after glucose addition.

Discussion
Active N 2 fixation in deep and NO 3 -rich waters of the ETSP In this study, we measured during 2 consecutive years N 2 fixation in surface waters affected by the OMZ, but reveal that N 2 fixation below the euphotic zone is more important: 87 and 90% of total areal N 2 fixation were measured in the aphotic zone in 2010 and 2011, respectively. In these aphotic layers, volumetric N 2 fixation rates were relatively low (,1.00 nmol N L 21 d 21 ), but when integrated over the whole aphotic layer, they ranged from 84 to 501 mmol N m 22 (Table 1). In 2011, rate measurements were replicated (triplicates) and calculations performed very carefully using a real T0 for every depth. These 2011 measurements are thus more reliable than those measured in 2010. These measurements in aphotic waters add new information compared previously published studies [26,28] in the area. The hypotheses explaining the persistence of N 2 fixation in these high NO x (ca. 40 mmol L 21 ) environments are largely developed in the companion paper [28]. First, fixed N loss processes occur in this region [31,48], creating a deficit of N relative to P, which is potentially favorable for N 2 fixation [24]. In particular, anammox removes NH 4 + , which has an immediate inhibitory effect on N 2 fixation [49]. Secondly, N 2 fixation is an anaerobic process [50] due to the irreversible inactivation of the nitrogenase enzyme by O 2 [51]. It is possible that the low O 2 concentrations in the OMZ and down to 2000 m contribute to the protection of nitrogenase [52], decrease the energy cost of maintaining intracellular  anaerobiosis [53], and thus facilitate N 2 fixation. Finally, redox conditions in the OMZ favor the equilibrium formation of the most bioavailable form of iron Fe 2+ [54], which could help to support the high Fe requirements of nitrogenase [16,17]. For these reasons, OMZs and deeper waters may represent favorable ecological niches for N 2 fixation, as shown in this study.

Potential impacts on N budgets in the ETSP
Aphotic N 2 fixation is currently ignored in oceanic N budgets based on biogeochemical rate measurements. However, this dataset indicates that rates in aphotic waters of the ETSP are of the same order of magnitude than those commonly measured in the tropical and sub-tropical NO 3 2 -depleted surface ocean (Table 2), where N 2 fixation has commonly been studied. The potential significance of the N 2 fixation rates measured in our study can be evaluated by comparing them with fixed N losses via denitrification and anammox measured in the same region. N losses in the ETSP have been estimated to range from 9 to 25 Tg N yr 21 (Table 3, [31,55,56,57]) in the upwelling area extending 175 km offshore, and 1860 km along the Peruvian-Chilean coast with an area extant of 3.26610 11 m 2 [31,56]. If we consider the same spatial extent for N 2 fixation, this process could potentially add 0.04 to 0.9 Tg N yr 21 (Table 3), counterbalancing 0.16 to 10% of the estimated N loss processes in this area (calculations have been performed only using numbers from the 2011). However, the anammox and denitrification measurements mentioned above [31] were performed under conditions of excess substrate availability and therefore represent maximum estimates of N loss rates. In contrast, the 15 N 2 bubble method used to quantify N 2 fixation [58] may underestimate rates [38,39].
Secondly, denitrification and anammox are restricted to subsurface suboxic or anoxic waters [59], whereas N 2 fixation is not. Further, denitrification and anammox appear to be restricted to the coastal upwelling system within ca. 175 km of the Peruvian-Chilean coast (the few data available at open ocean stations indicate that N loss processes were below detection limit during the 2010 cruise, Hamersley et al., (Pers. Com.)). N 2 fixation in the ETSP is active over a much greater spatial extent than N loss processes. If we consider the spatial extent of the N 2 fixation measurements in the ETSP covered by our cruises (2.23610 12 m 2 ), we estimate that N 2 fixation could potentially add 0.3 to 1 Tg N to the system in this area and therefore could compensate for up to 11% of the estimated N loss processes in the upwelling region of the ETSP (Table 3) (without taking into account methodological under-or overestimations). These estimates of N gains are the minimum ones calculated by taking into account only the 2011 cruise. If we take into account the 2010 cruise, N 2 fixation could potentially compensate up to 0.3 to 7 Tg N to the system (Table 3) (i.e. up to 78% of N losses). N 2 fixation in deep waters of the ETSP may be a significant source of N for the ETSP, and needs to be taken into account in future N budgets. Further coupled measurements between N gain and loss processes at the same stations/depths need to be performed to better constrain the magnitude of N gains in this region.

Effects of nutrients on N 2 fixation
The diazotrophic community of the ETSP characterized in this study, as well as in a companion study [42] is comprised of an assemblage of non-cyanobacterial diazotrophs, and little can be inferred about their metabolism from partial nifH sequences. However, we performed nutrient addition bioassays using molecules representing common labile components of the dissolved organic matter pool in marine waters (simple carbohydrates, amino-acids and ATP), which shed some light on nutrient control of N 2 fixation in the core of the OMZ. Our results indicated that simple carbohydrate additions significantly stimulated N 2 fixation at stations located nearest the coast during both cruises and at Station 9 during the 2011 cruise (Figs. 7, 8). In OMZs, organic C is largely supplied by vertical flux of planktonic production from shallower layers or by horizontal transport [60]. Thus this supply of organic C is not constant but rather episodic, which could explain why N 2 fixation appears so variable in space, in time, and between cruises and years, as reported in the present study and by [26]. This seems to be the case for N loss processes as well, since organic C supply has been correlated with regional denitrification [60] and anammox [31,61] rates in OMZs. Ward et al. [62] demonstrated that denitrification rates were significantly stimulated in the OMZ of the ETSP by organic C additions. To our knowledge our study is the first designed to study the response of diazotrophs to nutrient additions in the OMZ. In surface waters, significant stimulation of N 2 fixation rates by glucose additions have been reported during the same cruise at Station 9 [28]. A significant stimulation of bacterial production after glucose amendments in surface waters of the Chilean upwelling system have also been reported [63]. Finally, in surface waters of the southwest Pacific [64], reported a significant increase of nifH gene copies of unicellular diazotrophic cyanobacteria such as Group A (UCYN-A) and Crocosphaera after glucose and mannitol additions, hypothesizing that this capacity may allow conservation of energy by rapid uptake and recycling of sugars. However, it has to be noted that the large variability in the response to carbohydrates addition (high standard deviation at Station 11 for example) could be explained by the fact that it may be coincidental whether the taxa that benefit from the enrichment possess the nifH gene.
Because the organic C molecules tested here are also energyrich molecules easily entering catabolic pathways, one could interpret our results to be indicative of limitation either by energy or by assimilative C availability. However, in our experiments N 2 fixation was not stimulated by ATP additions at any station, indicating that C and not energy might have been the proximate limiting factor. In some oligotrophic P-limited environments, ATP is also a source of P for bacteria and uptake rates of ATP exceed those of glucose [65]; however, in OMZs, P is not limiting relative to N, which may further restrict the ability of ATP to stimulate N 2 fixation rates in our bioassays. In contrast, the addition of free amino acids stimulated N 2 fixation at all stations tested; this has also been shown in aphotic oxynenated waters of the Red Sea [66]. Amino acids are a source of both C and N, and it has been suggested that it is energetically advantageous for microbes to use preformed compounds such as amino acids rather than glucose as C sources [67]. In terrestrial legume-rhizobium symbioses, the diazotrophic bacteria assimilate amino acids such as glutamic acid provided by the host, which facilitate both dicarboxylate oxidation and ammonium assimilation into asparagine [68]. In Azospirillum sp., additions of glutamic acid also stimulated N 2 fixation activity [69] by providing a C and energy source to the diazotrophs, while N was still provided via N 2 fixation. It may be that similar nutrient assimilation dynamics are occurring in diazotrophs in the ETSP OMZ.
Ambient NO x concentrations were high (ca. 30-40 mmol L 21 ) at all stations where nutrient additions were performed, and NO 3 2 additions (8 mmol L 21 ) never resulted in N 2 fixation inhibition at any station. As the metabolic potential of diazotrophs present in  the OMZ have not yet been fully characterized, we do not know if they possess genes for reduction and assimilation of NO 3 2 or NO 2 2 . Detailed studies at the single cell level would be needed to characterize the metabolism of these organisms and understand why microbes fix N 2 in the presence of so much NO 3 2 . In addition to possible energy and C, N and P sources derived from molecules like amino acids, carbohydrates or ATP, electron sources and donors are also very important to know for characterizing the physiology of the diazotrophs present in the ETSP. Molecules like O 2 , NO 3 2 and less favorably SO 4 22 are common electron acceptor and they are used for different types of respirations like aerobic respiration, or anaerobic denitrification and sulfate reduction. These different respiratory pathways potentially supporting N 2 fixation are performed by organisms with different physiology which each have their own environmental sensitivities for fixing N 2 .

Phylogenetic characterization of diazotrophs
Our characterization of the diazotrophic community in the core of the OMZ revealed the presence of potential N 2 -fixing heterotrophs based on the presence of the nifH gene. We did not detect the cyanobacterial diazotrophs commonly found in other regions of the open ocean; in contrast, most of the nifH genes amplified from the OMZ clustered with a-, cand d-proteobacteria. This result is consistent with the observations of Turk-Kubo [42] in the upper 200 m of the ETSP water column, where 96% of sequences were also affiliated with proteobacteria. Based on these results, and other studies conducted in the ETSP and the South Pacific Gyre [14,26,70], it is clear that the ETSP diazotrophic community is different from other well-studied tropical and subtropical oceans such as that of the North Pacific, North Atlantic and Indian Oceans. The cyanobacterial diazotrophic phylotypes commonly found at high abundances in these other ocean provinces appear to be either sporadically present at low abundances (i.e. Trichodesmium, UCYN-A), or undetected altogether (i.e. UCYN-B, diatom-diazotroph associations) in the ETSP.
The amplification of diverse non-cyanobacterial nifH-containing organisms from OMZ waters in the ETSP affiliated with nifH clusters 1K, 1G and 3, is consistent with the findings of other studies conducted in anaerobic waters [26,36,71] and in abyssopelagic waters [47]. However, the results from this study underscore the difficulty inherent in identifying the diazotrophic community responsible for N 2 fixation rates. It is important to note that nifH cluster 1K sequences have been reported as contaminants in many studies, including a study in the ETSP [42]. However, none of the sequences recovered here had greater than 90% amino acid similarity and 83% nucleic acid similarity to reported contaminants. Nevertheless, as a result of the use of highly degenerate primers and nested PCR cycles necessary to amplify this important but low-abundance gene target, contamination must always be considered as a source for heterotrophic diazotroph sequences, whether from PCR and DNA extraction reagents or from sampling or handling procedures, despite the screening of PCR and reagent blank controls as in this study.
Furthermore, although it is clear that a diverse assemblage of non-cyanobacterial nifH-containing organisms are present in the OMZ of the ETSP, the best methodologies currently available to characterize dominant members of the diazotrophic community (PCR amplification using degenerate nifH primers) often identify organisms present at extremely low levels when targeted using quantitative approaches (i.e. qPCR; Fig. S1 in File S1) [42,47,71]. This, in turn, makes it difficult to argue that these organisms are capable of fixing N 2 at cell-specific rates great enough to account for measured bulk rates. An analysis of the expected N 2 fixation rates based on abundances and plausible cell-specific N 2 fixation rates in the ETSP discussed in [42] indicate that these proteobacteria are unlikely to be responsible for all the measured bulk rates and therefore other N 2 -fixing organisms could be responsible for a part of N 2 fixation in this region but may remain uncharacterized. Identifying which organisms are actively transcribing nifH using techniques such as reverse transcription (RT)-qPCR might provide more insight into which diazotrophic taxa are actively fixing nitrogen. However, the challenge of identifying which organisms are important N 2 -fixers remains the same when designing qPCR primers from sequences derived from RT-PCR based clone libraries, and are further convoluted by potentially low transcript abundances per cell and/or the timing of sampling with respect to diel changes in nifH expression (even in the case of heterotrophs).

Conclusions
This study provides one of the first estimates of N 2 fixation rates in aphotic waters of the ETSP. It reveals that N 2 fixation in aphotic environments is the largest contributor to total areal N 2 fixation in ETSP. N 2 fixation in high [NO 3 2 ] environments remains an enigma as it requires an additional energetic cost relative to NO 3 2 or NH 4 + . Further physiological studies are needed to understand the physiological regulation of N 2 fixation, especially on newly discovered diazotrophic organisms. Contrary to N 2 fixation performed in euphotic layer which sustains new primary production [3], aphotic N 2 fixation may sustain organic matter remineralization. These new sources of N could potentially compensate for as much as 78% of the estimated N loss processes in ETSP, indicating that they need to be taken into account in marine N budgets. Phylogenetic studies confirm the presence of diazotrophs in the deep waters on the OMZ, which are distinct from cyanobacterial phylotypes commonly found in surface oligotrophic waters of the tropical ocean. Organic and inorganic nutrient addition bioassays reveal that amino acids and simple carbohydrates stimulate N 2 fixation in the core of the OMZ, and the episodic supply of these nutrients from upper layers may explain the large temporal and spatial variability of N 2 fixation in the ETSP. Research on marine heterotrophic N 2 fixation is at its beginning and significant progress needs to be made in the refinement of the methods to estimate planktonic N 2 fixation in OMZs ( 15 N 2 bubble method versus 15 N 2 -enriched seawater) from bulk measurements to single cells analysis. The 15 N-enriched seawater method should be coupled to oxygen-free and trace metal-clean procedures to provide more accurate estimates. Progress also needs to be made in the characterization of the community responsible for N 2 fixation in these deep waters, as well as the control of their population dynamics by the supply of organic matter. Estimates of global N 2 fixation based on field measurements [5,72] are presently lower than geochemicallybased (nutrient stoichiometry and isotopic ratio) estimates [73]. Taking into account deep N 2 fixation might help to resolve some of this discrepancy. However, progress also needs to be made in the quantification of N loss processes, as recent studies indicate that they may be less sensitive to oxygen than previously thought [59], further complicating the N budget in the ETSP. In future studies, N gain and loss measurements need to be coupled in space and time to further resolve the N budget in the ETSP.

Supporting Information
File S1 Supporting methods, Table S1, and Figure S1.