Global gaseous nitrogen (N2) fixation rates may be underestimated and data is lacking from many regions without conspicuous diazotrophic cyanobacteria, such as cold oceans. We estimated N2 fixation rates at diverse sites in the Canadian Arctic, including the mouth of the Mackenzie River, the offshore Beaufort Sea, Lancaster Sound, Baffin Bay and a river influenced fjord. We also identified potential diazotrophic communities using a targeted survey of the nifH gene. Nitrogen fixation rates ranged from 0.02 nmol N L−1 d−1 in Baffin Bay to 4.45 nmol N L−1 d−1 in the Mackenzie River plume. Sequences recovered from the nifH gene survey belonged mainly to Cluster III, a group of nifH sequences associated with diverse microorganisms, with some α- andγ-proteobacterianifH genes at most sites. Cyanobacteria nifH genes with best matches to Nostocales, which are common in Arctic freshwaters, were recovered from the marine Beaufort Sea. The geographic pattern of N2 fixation rates and nifHgene identities suggest that the Mackenzie River is the source of a diazotrophic community that contributes new nitrogen to the nitrogen-depleted surface waters of the Beaufort Sea. This first record of N2 fixation at high latitudes refines our understanding of the global nitrogen budget.
 Global estimates of nitrogen fluxes indicate that the oceanic nitrogen cycle is not in balance [Codispoti et al., 2001]. Over long-time scales, the main source of biologically available nitrogen to the ocean is gaseous nitrogen (N2) fixation by diazotrophic (N2-fixing) cyanobacteria. Conversely, the major loss of this fixed nitrogen from the ocean to the atmosphere is mediated by other bacteria and occurs via denitrification, which ultimately releases N2 after the dissimilatory reduction of nitrate (NO3−), or via the anaerobic oxidation of ammonium (NH4+) [Ward et al., 2007]. Estimates of denitrification, in both the water column and in seafloor sediments, are globally greater than N2 fixation rates implying that the ocean is losing nitrogen to the atmosphere [Brandes and Devol, 2002; Mahaffey et al., 2005; Codispoti, 2007; Gruber and Galloway, 2008].
 A substantial portion of global denitrification occurs in the sediments of the wide shallow shelves that dominate the western Arctic seascape [Devol et al., 1997; Chang and Devol, 2009]. The ensuing loss of nitrogen to the atmosphere is thought to lower the N:P ratio of Pacific origin waters that transit through the Beaufort Sea and eventually enter the North Atlantic Ocean [Yamamoto-Kawai et al., 2006]. The low molar N:P ratio, which is below nine at the onset of the growing season in the southeast Beaufort Sea [Tremblay et al., 2008], could favor diazotrophy in the western Arctic Ocean, in the same manner as further south, in the Atlantic Ocean [Karl et al., 2002; Voss et al., 2004; Yamamoto-Kawai et al., 2006]. Nitrogen input from N2 fixation could also explain the continued drawdown of dissolved inorganic carbon and soluble reactive phosphorus following the depletion of NO3− in the surface mixed layer of the Beaufort Sea [Tremblay et al., 2008].
 To our knowledge, N2 fixation has never been reported from the Arctic Ocean and only once north of 30°N in the coastal temperate Atlantic Ocean [Mulholland et al., 2012]. Any N2 fixation in the Arctic Ocean would alter the current paradigm of the nitrogen balance between Pacific and Atlantic oceans and explain some of the discrepancy between input and removal estimates of fixed nitrogen. Nitrogen fixation could also potentially influence regional new production, which is the portion of primary production supported by allochthonous nitrogen and balanced by vertical carbon export under steady state conditions [Eppley and Peterson, 1979]. We addressed this data gap by combining rate measurements and nifH gene surveys to assess the potential for N2 fixation in the Arctic Ocean.
2. Materials and Methods
 This study was carried out aboard the Canadian icebreaker CCGS Amundsen in the coastal (Mackenzie Shelf) and offshore Beaufort Sea (19 July to 3 August 2008 and 31 July to 13 August 2009) and in Lancaster Sound and Baffin Bay (7–25 September 2008) (Figure 1). Stations that were too shallow for the ship in the Mackenzie River plume and Kugmallit Bay were accessed by helicopter or barge and the surface samples were collected directly into a clean carboy. At all other stations, water was collected with a rosette sampler equipped with twenty-four 12-L Niskin type bottles (OceanTest Equipment Inc.), a conductivity-temperature-depth (CTD) profiler (SBE-911, Sea-Bird Inc.), a fluorometer (Seapoint Sensors Inc.) and a sensor measuring photosynthetically available radiation (PAR; QCP2300, Biospherical Instruments Inc.).
 Samples for N2 fixation assays and molecular analyses in 2008 were collected from the surface or ≈5 m when collected with the rosette sampler. In 2009, N2 fixation assays were conducted on samples collected from ≈5 m and the subsurface chlorophyll maximum (SCM, between 30 and 57 m) to investigate diazotrophic potential deeper in the water column. The seawater was passed through a 200 μm Nitex mesh to eliminate large particles and zooplankton and then collected into 20 L acid-washed carboys that had been rinsed 3 times with sample water prior to filling.
2.2. Nitrogen Fixation Assays
 The 15N2 tracer method was selected since it is more sensitive than the acetylene (C2H2) reduction method and is preferred for oligotrophic systems with low diazotrophic activity [Montoya et al., 1996]. In 2008, water for the N2fixation assays was immediately dispensed into 250 mL clear glass bottles, which were filled to overflowing to avoid air bubbles. A gas-tight syringe (SGE Analytical Science) was used to inject 1 mL of15N2 (≥98 atom% 15N; Cambridge Isotope Laboratories) through twin-valve bottle caps, leading to an enrichment of ca. 16–20 atom%. Immediately after15N2 addition, T0 bottles were filtered under gentle vacuum (<10 in. Hg) to estimate the initial 15N/14N ratio of particulate organic nitrogen (PON). Triplicate or duplicate bottles, depending on water availability, were incubated over 24 h in an on-deck incubator covered with neutral-density screening simulating ca. 50% surface irradiance. Water was continuously pumped from the surface through the incubators to maintain in situ temperature. Incubations were terminated by gentle filtration onto 3μm silver filters for one set of bottles and onto 0.2 μm silver filters for another set. This smaller filter pore size was to retain the smallest cells.
 Although the high sensitivity of mass spectrometric analysis makes it possible to use low 15N enrichment [Montoya et al., 1996], the 15N/14N ratio of incubated samples fell within the range of T0 values at one of the offshore station in 2008. For this reason the incubation volume was increased to 500 ml in 2009 to improve precision. The quantity of 15N2added was the same as in 2008, leading to an enrichment of 8–9%. Since the effect of temperature was also assessed in 2009, the 24-h incubations were carried out in controlled laboratory incubators. One incubator was kept at 2.0°C and the other at 7.0°C. The in situ values at the surface and the SCM averaged 2.6°C and −1.1°C, respectively. To avoid potential light limitation of the community, the samples were incubated at 268μmol quanta m−2 s−1, which was within range of daily mean values at 5-m depth (mean of 154 ± 147μmol quanta m−2 s−1) and much greater than at the SCM (mean of 4.0 ± 4.7 μmol quanta m−2 s−1). PON was collected onto 24-mm diameter, pre-combusted 0.7μm GF/F filters in order to limit filtration times. On several occasions the filtrate was collected onto 0.2 μm filters to test whether 15N-labeled PON passed through the GF/F and to facilitate comparison with 2008 data.
 All filters were folded, desiccated at 60°C for 48 h and stored in 2 mL cryovials. The quantity and isotopic enrichment of the PON on filters was estimated using an elemental analyzer (ECS 4010, Costech Analytical Technologies Inc.) coupled with an isotope ratio mass spectrometer (Delta V Advantage, Thermo Electron Corporation). Sample runs started following two series of 10 injections of working reference gases, N2(5.0 grade, ultra-high-purity; Praxair) and CO2(4.5 LaserStar™, ultra-high-purity; Praxair), with a standard deviation <0.06‰ forδ15Ν. Several international reference standards of L-glutamic acid (USGS40 and USGS41 from the International Atomic Energy Agency) [Qi et al., 2003] were placed at the beginning and end of each analytical run and working standards (acetanilide; Costech Analytical Technologies Inc.) were interspersed between samples to correct for drift. Nitrogen fixation rates were estimated as per Montoya et al. . Initial N2 concentrations were calculated using constants proposed by Hamme and Emerson  and assuming equilibrium with the overlying atmosphere. The N2 fixation rates reported here should be considered as conservative, since the direct addition of a 15N2 gas bubble instead of previously equilibrated 15N2-enriched seawater may underestimate N2 fixation by ca. 20% during 24 h incubations [Mohr et al., 2010].
2.3. Molecular Analysis
 In a clean laboratory on board the ship, 2 to 6 L of seawater were sequentially filtered through a 52 μm mesh, a 47-mm diameter 3μm pore size Nucleopore polycarbonate membrane filter and a 0.2 μm pore size Sterivex®units (Durapore, Millipore) using a peristaltic pump (Masterflex, Cole-Parmer). Buffer (50 mmol L−1Tris-HCl, 0.75 mol L−1 sucrose, 40 mmol L−1 EDTA, pH 8.3), was added to the Sterivex® units and samples were stored at −80°C.
 The microbial biomass collected on the 0.2 μm filter (0.2–3 μm fraction) was used for all subsequent analysis. Although filamentous Nostocales and the diazotroph Crocosphaera may be larger than 3μm, these organisms have never been reported from the marine Arctic. Thus, molecular analyses were only performed using 0.2–3 μm fraction since a priori the expected diazotrophic community would be either picocyanobacteria or other bacteria [Gradinger and Lenz, 1995; Waleron et al., 2007]. DNA was extracted using a salt-extraction protocol [Aljanabi and Martinez, 1997] slightly modified as explained in Harding et al. . The potential diazotrophic community was characterized by way of clone libraries constructed using specific primers targeting the nifH gene that encodes for dinitrogenase reductase within the operon responsible for N2 fixation [Zehr et al., 2003a]. Five stations were selected from regions representing a range of N2 fixation rates and environmental conditions (stations indicated by a star in Figure 1).
 Nested polymerase chain reaction (PCR) with two sets of degenerate primers was used to amplify the nifH gene [Zehr and Turner, 2001]. Reaction mixtures without addition of template DNA were used for negative controls for all PCR reactions, these were always negative. PCR amplifications were performed with an iCycler thermal cycler (BioRad). Reaction mixtures were composed of 0.2 to 1 μL of DNA template, Taq polymerase (0.05 units μL−1 final concentration), reaction buffer, deoxynucleotide triphosphates (dNTP, 0.2 mmol L−1final concentration) (Feldan-Bio), bovine serum albumin (BSA, BioLab), the forwardnifH3 (5′-ATRTTRTTNGCNGCRTA-3′) and reversenifH4 (5′-TTYTAYGGNAARGGNGG-3′) primers (2μmol L−1 final concentration) [Zehr and Turner, 2001]. Autoclaved and UV treated milli-Q water was added for a final reaction volume of 20μL. The amplification protocol consisted of an initial denaturing step at 94°C for 4 min, then 30 cycles of 1 min each at 94°C, 55°C, and 72°C, followed by a final extension at 72°C for 7 min. The second PCR entailed the same concentration of reagents as the initial PCR reaction, using 3 μL of the PCR-product from the first PCR and the forwardnifH1 (5′-TGYGAYCCNAARGCNGA-3′) and reversenifH2 (5′-ADNGCCATCATYTCNCC-3′) primers at a final concentration of 2μmol L−1 [Zehr and Turner, 2001]. The same amplification protocol was repeated but annealing temperature was increased to 57°C. PCR-products the target size of 359 base pairs (bp) were verified on a 2.5% agarose gel as the high-density gel facilitated band cutting. The 359 bp fragments were excised and purified with the QIAQuick Gel extraction kit (Qiagen) following the manufacturer's protocol. PCR-products were then cloned (StrataClone PCR cloning kit, Agilent Technologies) and 60–90 positive clones for each library were selected and verified following PCR amplification using the plasmid T7 primers. Amplicons the correct size were sequenced with an Applied Biosystems 3730xl system by Service de séquençage et génotypage du Centre Hospitalier de l'Université Laval (CHUL). Sequences from this study have been deposited in Genbank with accession numbers HQ130009-HQ130029.
 Sequence reads were checked and edited using Chromas software version 2.3 (Technelysium Pty Ltd.). The nucleotide sequences were checked against the National Center for Biotechnology Information (NCBI) nr database using BLASTn [Altschul et al., 1990]. The phylogeny of our sequences was inferred by comparison with other sequences that had the highest similarity to the recovered sequences and sequences from other environmental nifH studies [Zehr et al., 1998; Falcón et al., 2004; Church et al., 2005, 2008; Langlois et al., 2005]. The sequences were aligned using ClustalX 2.0.12 [Thompson et al., 1997]. The software PHYLIP 3.69 [Felsenstein, 1989] was used to construct maximum likelihood bootstrapped (1000 replicates) phylogenetic trees and operational taxonomic units (OTUs) were defined at a level of ≥97% nucleotide similarity using BioEdit.
2.4. Nutrients Determination and Bacterial Counts
 Samples for nutrients were collected directly from the Niskin type bottles after filtration through GF/F filters (Whatman). Inorganic nutrients, NO3− and phosphate (PO43−), were analyzed using routine colorimetric methods adapted from Hansen and Koroleff  with an Autoanalyzer3 (Bran and Luebbe). The analytical detection limit was 0.03 μmol L−1 for NO3− and 0.05 μmol L−1 for PO43−.
 Bacterial abundance was estimated for surface samples collected in 2008 using standard procedures based on 4′, 6-diamidino-2-phenylindole (DAPI) staining [Porter and Feig, 1980]. After fixation of samples with 1% glutaraldehyde (Canemco Inc.), 15 mL was filtered by vacuum filtration onto 0.2 μm black polycarbonate filters (25 mm; Poretics) and incubated with DAPI (50 μg mL−1final concentration; Sigma-Aldrich) for 5 min. The filters were placed on slides with a drop of Zeiss non fluorescent immersion oil and stored at −20°C until examined [Porter and Feig, 1980], within 6 to 12 months. Cells were counted on non-overlapping fields using an epifluorescent Olympus IX71 microscope at 1000X magnification using ultraviolet excitation. Pico- and nano-cyanobacteria abundances were also determined for surface samples collected in 2008 using an Epics Altra flow cytometer (Beckman Coulter) after fixation of samples with 0.1% final concentration glutaraldehyde (Sigma-Aldrich). Details are given in the study ofTremblay et al. . Samples were flash frozen in liquid nitrogen and then stored at −80°C until analysis within six months.
3.1. General Conditions at the Sampling Stations
 In 2008, sea surface temperatures were warmer in the offshore Beaufort Sea (7.0°C) compared to Lancaster Sound (0.2°C) and Baffin Bay (2.1°C). Temperatures at the shallow stations of the Mackenzie River plume and Kugmallit Bay were above 15°C (Table 1). In 2009, Beaufort Sea surface temperatures were cooler, both in the coastal (average 4.0°C) and the offshore (average 1.2°C) zones (Table 2). In these two zones, temperatures were below 0°C at the SCM. The depth of the SCM corresponded to the bottom of the euphotic zone, defined here as 1% of surface irradiance for most sites (Table 2 and Figure 2). During both years, N:P ratios were greater on the Mackenzie Shelf and lower offshore; ratios were especially low on the eastern side of the Canadian Arctic. The 2008 bacterial concentrations were highest in the Mackenzie River plume and lower eastward toward Baffin Bay; Gibbs Fjord values were relatively higher than those of offshore Baffin Bay (Table 1). Within the Beaufort Sea region, cyanobacteria concentrations were extremely low with values of ca. 400 cells ml−1 in the Mackenzie River plume and less than 100 cell ml−1 elsewhere (Table 1).
Table 1. Environmental Variables, Bacterial Counts, and N2 Fixation Rate in the Surface Layer of the Different Regions Sampled During Summer 2008
Table 2. Environmental Variables and N2 Fixation Rate (Mean and Standard Error of Replicates) in the Surface Layer and at the SCM for the Estuarine and Marine Beaufort Sea During Summer 2009
Nitrate (μmol L−1)
Phosphate (μmol L−1)
N2 Fixation (nmol N L−1 d−1)
 Surface temperatures and NO3− concentrations decreased while surface salinity and PO43− concentrations increased from the Mackenzie River plume to the offshore Beaufort Sea in 2008 (Table 1). Moreover, CTD data from 2009 revealed that vertical PAR attenuation was much greater on the shallow shelf (stations with bottom depth <100 m) than offshore, presumably due to high concentrations of suspended material. The base of the euphotic zone was at ca. 26 m and ca. 63 m at coastal and offshore stations, respectively (Figure 2). Coastal Beaufort Sea stations also exhibited the relatively high N:P ratios and warm surface temperatures indicative of the Mackenzie River (Table 2). In the rest of the text these stations are labeled “estuarine” and all others are designed “marine” for 2009.
3.2. Nitrogen Fixation Rates
 In summer 2008, N2 fixation rates ranged from 4.45 nmol N L−1 d−1 in the Mackenzie River plume to 0.02 nmol N L−1 d−1 in Baffin Bay. Rates were highest at nearshore stations: the Mackenzie Shelf and Gibbs Fjord stations had higher N2 fixation rates than marine stations further offshore, especially compared to stations in the main Baffin Bay and in Lancaster Sound (Table 1). Despite differences in incubation protocols, N2 fixation rates in 2009 were similar to those in summer 2008 for the Beaufort Sea. River influenced and estuarine rates were nearly an order of magnitude greater than marine rates (Tables 1 and 2). To assess whether this difference was caused by dilution of diazotrophic river communities, results from the two sampling years were plotted against salinity. Nitrogen fixation rates decreased nonlinearly with increasing salinity dropping to low levels at salinities above 20 (Figure 3). In 2009, there was also an estuarine to marine gradient in N2 fixation rate for the SCM samples, whose rates were similar to or greater than those in the surface layer (Table 2).
 In 2008, >80% of the N2 fixation occurred in the 0.2–3 μm size fraction at the Mackenzie Shelf and Gibbs Fjord stations (Table 3). In the marine Beaufort Sea the 0.2–3 μm size fraction represented 37% of total N2 fixation. Rates were so low in Baffin Bay that we did not attempt to estimate difference between the size fractions. In 2009, sequential filtration onto GF/F filter (≈0.7 μm) and onto 0.2 μm showed that the >0.7 μm size fraction generally accounted for >90% of the total N2 fixation at both depths sampled for estuarine and marine stations of the Beaufort Sea (Table 4).
Table 3. Percentage of Surface N2 Fixation Occurring in Each Size Fraction Considered During 2008
Table 4. Percentage of N2 Fixation Occurring in Each Size Fraction at the Surface and the SCM During 2009 Assays
Estuarine Beaufort Sea
Marine Beaufort Sea
3.3. Temperature Effect
 In 2009, N2 fixation rates in the estuarine samples were higher at 7°C compared to 2°C. The slope of the linear regression of N2 fixation differed significantly from unity (t = 4.679, df = 11, p= 0.001, conformity T-test, Systat 11.0) (Figure 4). The corresponding Q10, which is the rate of change of a biological reaction following a 10°C increase in temperature, was estimated to be 1.44 using:
where 1.2 is the slope of the regression line in Figure 4 and T1 and T2 are the lower and higher temperatures used in the experiment, respectively. The effect of temperature on N2 fixation was not significant for the marine stations (t = 0.993, df = 10, p = 0.344).
3.4. Characterization of Diazotrophic Community
 Clone libraries were constructed from sites representing the wide range of different environments over the study region. Results from the different libraries varied and the lowest number of clones with the target sequence was from the Gibbs Fjord library and the most clones (25) were recovered from Baffin Bay (Table S1 in the auxiliary material). Following BLASTn, alignment and phylogenetic tree construction sequences clustered with cyanobacteria, α-,γ-proteobacteria andnifH Cluster III, a poorly defined group with representatives that include anaerobic archaea, δ-proteobacteria and sulphate reducers [Zehr et al., 2003a]. Proteobacteria and Cluster III sequences were compared with documented contaminants and with the soil bacterium VF003 [Zehr et al., 2003b; Goto et al., 2005], and a similarity matrix revealed no similarity >75% (Table S2).
 Most of the nifH sequences in this study belonged to Cluster III and were recovered from every library except the marine Beaufort Sea library. In the Mackenzie Shelf libraries, α-proteobacterianifH gene sequences were common while γ- andβ-proteobacterianifH genes were less frequently recovered. We found cyanobacteria nifH gene sequences from Kugmallit Bay and cyanobacteria were the only group recovered from the marine Beaufort Sea (Figure 5). One OTU was common to both regions (Table S1). The highest number of OTUs was from the Mackenzie Shelf, and only 2 OTUs were recovered from the marine Beaufort Sea (Table S1). Two nifH sequences retrieved in Gibbs Fjord fell into separate OTUs, one of which was also recovered from the Baffin Bay library and belonged to Cluster III.
 Phylogenetic analysis revealed that the non-cyanobacterial groups, especially Cluster III, were generally distant from other environmental sequences in the NCBI database (Figures 6, 7, and 8). Except for one Gibbs Fjord OTU related to γ-proteobacteria (Gibbs_9), which was only 92% similar to an environmental sequence from the Pacific Ocean (Figure 7), we did not find any sequences coming close to those from the Atlantic and the Pacific oceans, although several were similar to marine microbial mat bacteria. In contrast, many sequences had best matches to sequences previously reported from terrestrial or freshwater environments. Cluster III sequences clustered with both putative terrestrial and marine-/brackish-derived (Figure 8). All OTUs related to cyanobacteria in the estuarine and marine Beaufort Sea had their highest BLASTn matches to sequences reported from the order Nostocales (Nostoc spp., Figure 9).
4.1. Distribution of N2 Fixation in the Canadian Arctic Ocean and Surrounding Seas
 Nitrogen fixation and sea surface temperature are correlated throughout the world's oceans [Karl et al., 2002; LaRoche and Breitbarth, 2005]. Nitrogen fixation estimates range from 0.15 nmol N L−1 d−1 [Falcón et al., 2004] to 3 nmol N L−1 d−1 [Voss et al., 2004] in the tropical Atlantic Ocean and from 0.38 nmol N L−1 d−1 [Zehr et al., 2001] to 1.8 nmol N L−1 d−1 [Montoya et al., 2004] in the tropical/subtropical Pacific Ocean. For the latter, rates decrease to 0.25 nmol N L−1 d−1 north of 30°N [Needoba et al., 2007]. Our marine averaged N2 fixation estimates (up to 0.14 nmol N L−1 d−1) are thus below the lower end of the range in the Atlantic and Pacific oceans. Similarly, when compared with other coastal rates, our estuarine rates (up to 4.45 nmol N L−1 d−1) appear low; N2 fixation rates of 50 nmol N L−1 d−1 have been measured in the coastal temperate Atlantic Ocean [Mulholland et al., 2012]. Yet the rates reported here appear disproportionately high given the very low temperature of the Arctic Ocean relative to the global mean, suggesting the diazotrophic community may have specific adaptation to cold. However, additional genetic information on the organisms involved is required to verify this hypothesis [Rodrigues and Tiedje, 2008].
 Within the study area, N2 fixation rates varied with temperature during summer 2008 and 2009. Rates were highest within the relatively warm water of the Beaufort Sea compared to the colder Lancaster Sound and main Baffin Bay, where N2 fixation rates were negligible. Within the Beaufort Sea, highest rates were in the warmer and fresher estuarine zone, where experiments in 2009 documented that increasing temperature could significantly increase N2 fixation rates. However, this positive effect was small, since the Q10 of 1.44 estimated for N2 fixation is at the lower end of values for different metabolic processes in planktonic assemblages [Robinson and Williams, 1993; Bianchi et al., 1997]. Thus, temperature does not seem to act as a major control of diazotrophic activity at the regional scale in the Arctic, especially given the relatively high N2 fixation rate and low temperature in Gibbs Fjord.
 The availability of macronutrients can impact the N2-fixing capacity and the distribution of diazotrophic communities [Karl et al., 2002]. It is often stated that a low N:P ratio favors N2 fixers [Karl et al., 2002; Voss et al., 2004; Yamamoto-Kawai et al., 2006] since low nitrogen availability precludes the utilization of excess phosphorus by other microbes. While this view is supported by the coincidence of relatively high rates of N2 fixation and very low N:P ratios in Gibbs Fjord, the highest fixation rates overall occurred on the Mackenzie Shelf, where N:P ratios were greatest. Consequently, other factors such as micronutrients and the supply of organic material may be important and explain the regional differences in N2 fixation rates estimated here and elsewhere [Moore et al., 2009]. For example, diazotrophs require iron and molybdenum [Karl et al., 2002]. Iron is abundant on the inner Mackenzie Shelf [Moore et al., 2004] and glacial meltwater, an important source of iron elsewhere [Smith et al., 2007], may be a source of iron in Gibbs Fjord. Molybdenum is generally more available in freshwater compared to marine waters and in water with high concentrations of dissolved organic carbon. In seawater, abundant sulphate competes with molybdenum since both can be taken up by the same molecular transporters [Howarth and Cole, 1985; Stal et al., 1999]. The diazotrophic communities of the estuarine Beaufort Sea and Gibbs Fjord could benefit from the freshwater inflows. Moreover, if diazotrophs on the Mackenzie Shelf were mostly heterotrophic microbes, their activity would be stimulated by the availability of higher concentrations of organic matter in the river [Garneau et al., 2006; O'Brien et al., 2006].
4.2. Diazotrophic Community
 Since the nifH gene cluster is large and represents a significant portion of the bacterial genome in N2 fixing cells, it has been suggested that only species with actual capacity for N2 fixation under favorable conditions would maintain the gene [Zehr et al., 2003a]. However, there are reports of high diversity of nifH genes from environments where N2 fixation is negligible [Moisander et al., 2007]. Therefore, in the absence of expression data, the presence of the nifH gene indicates only a potential for N2 fixation.
 Cyanobacteria nifH genes were not universally found over the region. The near absence of cyanobacteria from polar seas can be contrasted to their ubiquity in arctic freshwaters [Gradinger and Lenz, 1995; Vincent, 2000; Waleron et al., 2007] and it might be expected that cyanobacteria nifH genes would be most common at stations influenced by the Mackenzie River. Conversely, in the greater Beaufort Sea region all the cyanobacteria sequences except one from Kugmallit Bay were retrieved offshore. Interestingly, those sequences were closest to Nostocales nifH genes, which cluster apart from unicellular group A cyanobacteria that are reported from coastal marine waters between Cape Hatteras and Georges Bank, in the North Atlantic Ocean [Mulholland et al., 2012]. Concurrent studies at the same Beaufort Sea stations specifically aimed at investigating phytoplankton using microscopy failed to record cyanobacteria [Thaler and Lovejoy, 2012; M. Gosselin, personal communication, 2012] but PCR is a powerful technique where specific primers will amplify very rare DNA template. A few cyanobacteria cells were also detected using flow cytometry. A possible source of these cyanobacteria offshore may have been Richelia (Nostocales), which is an endosymbiont of the diatom Rhizosolenia. Rhizosolenia are common in Northern Baffin Bay [Lovejoy et al., 2002], and possibly occur in the Beaufort Sea. However, neither of the cyanobacterial nifH OTUs was particularly close to the publically available Richelia nifH sequences. The diversity of this group may be underestimated and we cannot rule out that Richelia or related endosymbionts may be a source of the sequences.
 The paucity of Nostocales in Arctic marine waters contrasts with their widespread presence in polar freshwater environments [Vincent, 2000; Jungblut et al., 2010] and cyanobacteria could be transported to the Beaufort Sea from terrestrial or freshwater environments by the Mackenzie River, whose plume has been tracked well offshore [Carmack and Macdonald, 2002; Garneau et al., 2006]. One of the cyanobacterial OTUs retrieved in both the Kugmallit Bay and marine Beaufort Sea clone libraries was >98% similar to Nostoc commune Vaucher (UTEX 584) [Wright et al., 2001], a terrestrial (freshwater) species. N2 fixation has been recorded in Arctic Nostocales dominated microbial mats [Vincent, 2000] and Waleron et al.  also recovered sequences belonging to Oscillatoriales with closest affinities to freshwater species in the marine Beaufort Sea, but not to those found in the Baltic Sea or saline lakes. Alternatively, windblown cyanobacteria could be transported offshore [Harding et al., 2011]. The lack of cyanobacteria sequences from the river plume at the time of sampling could also be an artifact due to poor coverage in the libraries [Zehr and Capone, 1996].
 Irrespective of the observation of nifH cyanobacteria sequences offshore, we did not recover cyanobacteria sequences from the estuarine Beaufort Sea and in Gibbs Fjord, where highest N2 fixation rates were recorded. Cluster III nifH genes are not common in brackish or marine surface water [Church et al., 2005; Langlois et al., 2005; Moisander et al., 2007]. However, Langlois et al. noted that Cluster III tended to be more abundant at the more northerly stations in the North Atlantic Ocean. Cluster III sequences retrieved here from the Arctic were not highly similar to sequences recovered elsewhere from either oceanic or terrestrial biomes. Within Cluster III, most sequences retrieved in the estuarine Beaufort Sea clustered more closely with terrestrial-derived sequences while sequences from Baffin Bay clustered with brackish and marine-derived sequences. Although less abundant, theα- andγ-proteobacteria that commonly dominate heterotrophic diazotrophic assemblages in oceans [Falcón et al., 2004; Langlois et al., 2008] were also present at most stations. Interestingly, α- andγ-proteobacteria phylogenies revealed that the estuarine Beaufort Sea sequences were closest to freshwater environmental sequences while the only marineγ-proteobacteria sequence was recovered in Gibbs Fjord, which is less influenced by freshwater runoff compared to the Beaufort Sea region.
4.3. Mackenzie River Influence on N2 Fixation Rates
 Overall, the majority of N2 fixation activity was detected in a region highly influenced by freshwater runoff, and was close to detection limits in most of the predominantly marine Baffin Bay. In the Beaufort region, the rapid decrease in diazotrophic activity away from sites influenced by the Mackenzie River, suggests that the river could be the source of active diazotrophs to the western coastal Canadian Arctic Ocean. The gene survey was consistent with this view but must be interpreted carefully. However, the exponential decrease of diazotrophic activity with increasing salinity away from the river cannot be explained strictly by the dilution of the river assemblage since such pattern should be linear. In the estuarine Beaufort Sea, where highest N2fixation rates were recorded, all OTUs except one were associated with non-cyanobacterial diazotrophs. The Mackenzie River is the largest single source of land-derived organic matter to the Arctic Ocean [O'Brien et al., 2006] and suspended organic material fuels high marine bacterial productivity in the coastal Beaufort Sea [Garneau et al., 2006]. Such a large input of organic matter could facilitate heterotrophic diazotrophy. Organic matter could also promote diazotrophy in the estuarine Beaufort Sea by its positive effect on the uptake of molybdenum [Howarth and Cole, 1985; Stal et al., 1999] and decreasing organic matter may contribute to the exponential decrease of N2 fixation rates away from the Mackenzie River plume.
4.4. Contribution of Arctic N2 Fixation to Global Nitrogen and Carbon Cycles
 A mean annual denitrification rate of 1 mmol N m−2 d−1 has been estimated for the western Arctic shelf sediment of the Bering, Chukchi and Beaufort seas [Devol et al., 1997]. We estimate that the upper limit of N2 fixation rates integrated down to the SCM in the marine Beaufort Sea, assuming uniform rates from the surface to the SCM, could reach 0.0065 mmol N m−2 d−1, which is over 2 orders of magnitude less than denitrification rates. Clearly the Arctic Ocean plays a significant role in global denitrification [Devol et al., 1997; Chang and Devol, 2009] and its contribution to the reverse flux appears rather small. However, our results indicate that N2 fixation in the Arctic Ocean and surrounding seas should not be dismissed out of hand, especially considering the lack of measurements adjacent to the other major rivers that enter the Arctic Ocean.
 The ecological importance of heterotrophic N2 fixation is well recognized in seagrass beds, where the release of newly fixed nitrogen by heterotrophic bacteria supports plant growth [Welsh, 2000]. A similar process could contribute to new phytoplankton production on Arctic shelves via direct release by diazotrophs or via the recycling of fixed nitrogen by bacteria and grazers. In the Beaufort Sea during early autumn 2002 and 2003, estimates of new primary production based on phytoplankton cell size ranged from 14 to 25 mg C m−2 d−1 [Brugel et al., 2009]. Applying a mean molar C:N ratio of 7.3 for this region [Tremblay et al., 2008], N2fixation-based new primary production would be equivalent to 0.6 mg C m−2 d−1 in the surface layer of the marine Beaufort Sea. It would thus represent up to 4.3% of new primary production in the Beaufort Sea. The contribution of N2 fixation to new primary production within the western Canadian Arctic Ocean, and consequently to the export of carbon toward the deep sea, is presently small but not negligible.
 This study demonstrates that substantial N2 fixation occurs on a coastal shelf of the High Arctic, an area thought to be unsuitable for diazotrophic organisms until now, and shows the potential for N2fixation in the Arctic Ocean and adjacent seas. This input of fixed N may compensate partly for sediment denitrification on shelves, thereby reducing the excess of phosphorus in Pacific-derived waters entering the Atlantic Ocean [Yamamoto-Kawai et al., 2006]. Even a small contribution of the Arctic Ocean to global N2 fixation could affect the oceanic N balance and the carbon cycle [Falkowski, 1997].
 The potential for N2 fixation in the Arctic Ocean has yet to be completely explored and its contribution to the nitrogen cycle requires further attention. The Russian Arctic seas are influenced by several major rivers and potentially contribute to N2 fixation in the eastern Arctic. Ongoing climate change causing increases in temperature [Intergovernmental Panel on Climate Change, 2007] and river discharge, notably of the Mackenzie River [Yamamoto-Kawai et al., 2009], may favor N2 fixation in the Arctic Ocean and could eventually lead to a revised global N budget. This study also provided baseline information by which to assess future alteration of the nitrogen cycle in the rapidly changing Arctic Ocean.
 This work was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) and the ArcticNet Network Center of Excellence (NCE) and is a contribution to the scientific program of Québec-Océan and three international polar year (IPY) research programs: Circumpolar Flaw Lead System Study (CFL), Canada's Three Oceans (C3O) and Malina. Marjolaine Blais received a graduate scholarship from NSERC and financial support from the Northern Scientific Training Program. We thank the captains and crews of the CCGS Amundsen for their invaluable support in the field. We are indebted to Véronique Lago and Dominique Boisvert for the collection and analysis of physical data, to Mariane Berrouard for field logistics, to Jessie Motard-Côté and Tommy Harding for help during laboratory analysis, and to Michel Gosselin for access to pico- and nano-cyanobacteria data. We thank Jon Zehr and Michel Gosselin for insightful discussions and comments on an earlier draft of the manuscript.