3.1. Origin and Fate of Hydrothermally Discharged Ammonium
 During high-T hydrothermal circulation, Mg2+is almost completely removed from the aqueous solution through various water-rock reactions [Edmond et al., 1979; Seyfried, 1987]. At lower temperatures, mixing of hydrothermal fluids with Mg2+-rich and cold seawater elevates the [Mg2+] again. Hence, a linear (inverse) relationship between the temperature of the HV fluid and its Mg2+-content exists (r2 = 0.90, Figure 2a). [Mg2+] can thus be used as an indicator of mixing between zero-Mg2+pure high-T hydrothermal fluids and Mg2+-rich (52.9 mmol/kg) low-T crustal seawater.
Figure 2. Magnesium concentrations versus average fluid temperatures at Endeavour Segment (star = Sasquatch, down triangle = High Rise, square = Main endeavor Field, diamond = Mothra), Cobb Segment (triangle) and Axial Volcano (circle). Magnesium is almost completely removed from the hydrothermal vent fluids at higher temperatures and only slowly removed at lower temperatures. The colors represent the sampling years (yellow = 2004, purple = 2006, green = 2007, blue = 2008 and red = 2009). Standard deviation for sub-samples (n = 2 to 5) collected during the same dive and location is indicated.
Download figure to PowerPoint
 Ammonium is the dominant DIN species under the highly reducing conditions encountered in high-T fluids (∼200°C to 350°C, or <10 mmol/kg Mg2+), with average concentrations decreasing from North to South along the Juan de Fuca Ridge: 1078 ± 4.7 μmol/L (n = 2) at Sasquatch, 863 ± 129 μmol/L (n = 11) at High Rise, 410 ± 53 μmol/L (n = 23) at MEF, and 396 ± 20 μmol/L (n = 5) at Mothra at the Endeavour Segment, 44 μmol/L (n = 1) at Cobb Segment, and 14 ± 3 μmol/L (n = 22) at Axial Volcano (Figures 3b–3d). NH4+ concentrations behave mostly conservatively with respect to Mg2+, indicating the dilution of high-NH4+hydrothermal fluids with zero-NH4+, oxygenated deep-seawater.
Figure 3. (a) Nitrate (all sites) and ammonium concentrations at (b) Endeavour Segment, (c) Cobb Segment and (d) Axial Volcano versus [Mg2+] (note the different scales of the yaxis). In low-T vents, nitrate concentrations that fall below the mixing line between zero-nitrate, ammonium-rich, pure hot hydrothermal vents and nitrate-rich crustal seawater are indicative of microbial nitrate consumption. Ammonium concentration above ∼0μmol/L in low-T waters at Axial Volcano, occurring mainly at the site Marker 113 and its surrounding areas (black stars), is a sign of microbial ammonium production. Symbol and color scheme as inFigure 2legend. The seawater end-member is represented by a turquoise hexagon. Regression lines in Figure 3b: Sasquatch = solid line, High Rise = long dash, Main Endeavor Field = medium dash and Mothra = short dash). Standard deviations for sub-samples (n = 2 to 5) are indicated by the error bars.
Download figure to PowerPoint
 The δ15N of ammonium in high-T fluids of the Endeavour Segment did not differ among years, but differs significantly among sites, with average values (all years combined) of 4.3 ± 0.7‰ (n = 3) at Sasquatch, 4.2 ± 0.5‰ (n = 11) at High Rise, 3.6 ± 0.4‰ (n = 19) at MEF and 3.1 ± 0.3‰ (n = 5) at Mothra (Figure 4a) (Kruskal-Wallis, p-value = 0.009). Since we were primarily interested in variations among vent fields, we grouped the values at the four Endeavour Segment sites for the subsequent discussion. It should be noted that the low average ammoniumδ15N in high-T vent fluids of the Endeavour Segment (3.7 ± 0.6‰ (n = 37)) does not closely match the only existing report on the N isotopic composition of NH4+ in hydrothermal fluids by Lilley et al. , who measured a δ15N value of 12.4‰, but is consistent with a δ15N of 2.1‰ for an extinct sulfide chimney from Dante (at MEF) collected during the 2009 cruise (unpublished data).
Figure 4. δ15N–NH4+ versus [Mg2+] and ln [NH4+] at (a, b) Endeavor and Cobb Segments and (c, d) Axial Volcano. The average δ15N-NH4+ in high temperature fluids is indicated by the horizontal lines (dashed line for Cobb Segment). Symbol and color scheme as in Figure 2legend. In Figures 4b and 4d, high-T end-members are encircled. The highest15N isotope enrichment was observed at Axial Volcano: Marker 113 (black stars), The Spot (circle), and CASM (down-triangle).
Download figure to PowerPoint
 The average Endeavour Segment ammonium δ15N was significantly different from the δ15N of nitrate in Deep-Pacific seawater at ∼2100 m depth (5.6 ± 0.2‰ (n = 5), p-value = 0.0003). The averageδ15N of ammonium in high-T fluids of the Cobb Segment was 4.1 ± 0.1‰ (n = 3), i.e., not significantly different from theδ15N of ammonium at Endeavour Segment (Figure 4a). In contrast, the average δ15N of ammonium in high-T fluids at Axial Volcano was 6.7 ± 1.0‰ (n = 16) (Figure 4c), significantly greater than the δ15N observed at the Endeavor (p-value = 8 × 10−9) and Cobb (p-value = 0.007) segments, but not significantly different from theδ15N of deep-seawater nitrate in the ambient water column (6.4 ± 0.2‰, n = 4) (Figure 4c).
 Assuming no secondary alterations due to microbial or thermic processes, the δ15N of ammonium in high-T fluids should reflect that of the original N source. During high-T subsurface hydrothermal circulation, under reducing conditions, nitrate from the deep sea can be abiotically converted to ammonium by transition metal oxides and sulfides as it infiltrates the subsurface through cracks and fissures [e.g.,Brandes et al., 2008]. By mass balance, if fixed N is neither consumed nor produced during high-T reactions, discharging hydrothermal waters should contain NH4+at concentrations (and N isotopic composition) that are equivalent to the nitrate of inflowing crustal waters with a deep-sea origin (i.e., ∼40μmol/L), at least at the un-sedimented (i.e., no apparent organic N source) hydrothermal systems of the Juan de Fuca Ridge. In this context, the elevated ammonium concentrations observed at the un-sedimented Endeavour Segment appear anomalous. These have been argued to originate from the decomposition of subsurface organic material in sediments buried at an earlier stage of the ridge formation. Pleistocene turbidite flows from deposits located about ∼40 km north of the Juan de Fuca Ridge might be the source of these sediments [Lilley et al., 1993]. If OM from these deposits was indeed the source for the ammonium at Endeavor, their 15N content must have been low compared to the mean δ15N-NO3−of deep-seawater.
 Ammonium can, in theory, also be produced by abiotic reduction of N2by Fe-minerals in the subsurface high-T hydrothermal fluids, producing newly fixed NH4+ with a low δ15N [Brandes et al., 1998a; Schoonen and Xu, 2001; Dörr et al., 2003]. It is widely accepted that the δ15N of newly fixed N is close to the δ15N of the dissolved N2, which can either originate from a magmatic source (with a δ15N value of −5 ± 2‰ [Javoy and Pineau, 1991; Marty and Humbert, 1997]) or from circulating deep-seawater (with aδ15N value of ∼0‰). As pointed out by Lilley et al. , it is unlikely that abiotic N2 fixation would occur to such extent only at the Endeavor and Cobb Segments. Moreover, it can be predicted that the average δ15N of hydrothermal ammonium would be even lower than 3.7‰, at least if high-temperature N2 fixation was the main ammonium source.
 At Axial Volcano, the similarity of the δ15N of ammonium in high-T fluids and deep-seawater nitrate (6.7‰ and 6.4‰, respectively) suggests that seawater nitrate that penetrates from the water column into the anaerobic subsurface through cracks and fissures represents the substrate for complete abiotic reduction to ammonium. However, the ammonium-N concentration in high-T fluids of Axial Volcano was lower than the nitrate-N concentration in the Deep-Pacific by ∼26μmol/L. This difference, indicating the net loss of fixed N in the high-T fluids, is likely caused by high-T ammonium ion substitution in secondary minerals during fluid interactions with basaltic rocks. For instance,Hall  and Busigny et al.  observed that N (mainly occurring as ammonium ions substituting for K+ and Na+/Ca2+ in minerals) indeed gets enriched in rocks during basalts alteration. Details of the actual N removal mechanism aside, the accordance of the δ15N of inflowing nitrate and of high-T end-member ammonium implies that the N scavenging in hot hydrothermal fluids must occur without significant N isotope fractionation. In the same line,Holloway et al.  found that the δ15N of ammonium in hydrothermal waters of the Yellowstone National Park with a pH < 5 (the pH range for the high-T fluids in this study was ∼3.5 to 5) remained more or less unaffected by water rock interactions.
3.2. Biological Uptake and Isotopic Fractionation of Ammonium N in Low-T Fluids
 Ammonium in low-T fluids (<∼50°C) of the subsurface biosphere of hydrothermal vents can be produced either by organic matter remineralization (of both sedimentary and in situ produced organic N, e.g., by biotic N2 fixation [Mehta et al., 2003, 2005; Mehta and Baross, 2006]), and/or dissimilative nitrate reduction to ammonium (DNRA), which has been demonstrated to occur at temperatures up to 70°C [Vetriani et al., 2004; Voordeckers et al., 2005; Perez-Rodriguez et al., 2010]. On the other hand, ammonium can be consumed by biological processes in low-T fluids, displaying distinct N-isotope effects, such as ammonium assimilation, aerobic microbial ammonium oxidation [Lam et al., 2004, 2008], or anammox [Byrne et al., 2009].
 In most cases where ammonium concentration gradients appeared to behave conservatively with respect to Mg2+, the δ15N was invariant, confirming simple mixing of high-NH4+hydrothermal waters with zero-NH4+seawater in low-T hydrothermal fluids (Figures 3b–3d and Figure 4). However, at some sites of the Endeavour Segment and Axial Volcano, net ammonium production or consumption are evidenced by concentrations above or below those expected from conservative mixing in the low-T hydrothermal fluids (Figures 3b and 3d). Net ammonium production, either via partial DNRA, organic matter remineralization or the N2 fixation/remineralization cycle, corresponded with a decrease of δ15N-NH4+ with increasing [NH4+] in low-T fluids (i.e., Hulk at MEF and Hermosa, Vixen, Village and Escargo at AV,Figures 4b and 4d).
 Anomalously high ammonium concentrations were observed at Marker 113 diffuse vent fluids during all sampling campaigns (Figure 3d). Yet the δ15N-NH4+was either similar to the high-T end-member value or even greater (up to ∼10‰,Figures 4c and 4d), which appears inconsistent with ammonium production by N2 fixation or N isotope fractionation during partial DNRA (both processes would act to produce low δ15N-NH4+). Remineralization of high δ15N organic material to ammonium can also not account for the elevated δ15N-NH4+ values. Reported values of δ15N of hydrothermal vent fauna are generally low (∼−10 to +4‰ [Rau, 1981, and reference therein]), as was the δ15N of particulate material collected at Axial Volcano sites (4.6 ± 0.2‰ at Marker 33 and Gollum, 2009 cruise, unpublished data; 4.3 ± 1.2‰ [Levesque et al., 2005]). Therefore, we conclude that the elevated ammonium δ15N values reflect N-isotope fractionation during partial ammonium consumption by bacterial assimilation or nitrification occurring in tandem with ammonium production by processes mentioned earlier. That is, while net ammonium production is evidenced by non-conservative behavior of the ammonium concentration, theδ15N indicates that ammonium consumption occurs concurrently. The ammonium 15N enrichment was also observed at other sites, where NH4+ was clearly consumed relative to conservative mixing (up to ∼5.5‰ at CASM, Axial Volcano, Figures 4c and 4d).
 The net ammonium consumption N isotope effect can be estimated from the correlation between the natural logarithm of the ammonium concentration and the δ15N of the residual ammonium (closed-system Rayleigh Model [Mariotti et al., 1981]). However, open-system aspects and spatial/temporal variability in both the ammonium concentration andδ15N are likely to prevent any clear Rayleigh-type ammonium N isotope dynamics (Figures 3 and 4), and, even more importantly, in situ regeneration of NH4+ and the N isotope effects associated with this regeneration will bias estimates for εuptake for natural assemblages of bacteria in the diffuse HV fluids. Furthermore, the N isotope effects during ammonium oxidation (ε = +14 to +38‰ [Delwiche and Steyn, 1970; Mariotti et al., 1981; Yoshida, 1988; Casciotti et al., 2003]) and ammonium assimilation (ε = +14 to +27‰ [Hoch et al., 1992; Waser et al., 1998]) by microorganisms and algae in aquatic systems are highly variable and likely influenced by environmental conditions (e.g., substrate concentration and uptake rate), making it difficult to tell the two processes apart based solely on the degree of N-isotope enrichment. A plot ofδ15N-NH4+ versus the ln [NH4+] (Figures 4b and 4d) does not indicate obvious Rayleigh-type N-isotope dynamics. A significant relationship betweenδ15N-NH4+ and ln [NH4+] was only observed at Cobb Segment in 2007 (ε = 1.2‰, r2 = 0.7) and Axial Volcano in 2006 (ε = 2.6‰, r2= 0.3), although sample sizes for these data sets were limited. The computed ammonium N isotope effects can be taken as community N-isotope effects for net ammonium removal, and were much lower than both the N isotope effect expected for aerobic (and anaerobic) ammonium oxidation, as well as for ammonium assimilation at elevated NH4+ concentrations [Hoch et al., 1992]. Clearly, production of low-δ15N NH4+ occurs through gross ammonium regeneration.
 While the δ15N-NH4+ data alone do not allow us to constrain the actual ammonium removal pathway (uptake versus nitrification) the nitrate N and O isotopes in combination with elevated δ15N-NH4+ at Marker 113 suggest that partial nitrification of ammonium to nitrate must occur to some extent. The nitrate N and O anomalies that appear to be indicative of nitrate regeneration are discussed in detail below (see section 3.4).
3.3. Nitrate Consumption and Associated N and O Isotope Effects in Hydrothermal Vent Fluids
 Nitrate δ15N and δ18O values close to the N and O isotopic composition of ambient seawater nitrate (∼6‰ and 2‰, respectively) (Figures 5a–5d) suggest that nitrate in both high and low-T (<∼50°C) HV fluids mainly originates from mixing with nitrate-replete crustal seawater (∼40μmol/L in the deep Northeast Pacific Ocean). At several low-T fluids sites, however, elevated nitrateδ15N and δ18O (i.e., high Mg2+) (Figure 5), concomitant with decreased [NO3−] (Figure 3a), were observed. This clearly indicates a N and O isotope fractionating nitrate-consuming process in the low-T subsurface biosphere prior to venting. The15N and 18O enrichment in the HV fluid nitrate was greatest at Axial Volcano (up to ∼3‰ for δ15N and ∼11‰ for δ18O, Figures 5c and 5d) compared to the Endeavor and Cobb Segments (between ∼1.4‰ and ∼2‰ for δ15N and ∼3‰ and ∼5‰ for δ18O, respectively, Figures 5a and 5b). Moreover, we consistently observed higher relative isotope enrichments for 18O versus 15N (Figure 5). Nitrate δ15N and δ18O values in diffuse vent fluids sampled at the same locations fluctuated between years, with no clear temporal trend. In general, the 15N and 18O isotope enrichment was greater in 2006 and 2009, and lowest in 2008. In the subsequent discussion, we treat each data set separately.
Figure 5. Nitrate δ15N and δ18O versus [Mg2+] at (a, b) Endeavor and Cobb Segments and (c, d) Axial Volcano. Same symbol and color scheme as in Figure 2. The seawater (SW) δ15N end-member is indicated by the solid black lines. Error bars represent the standard deviation for sub-samples (n = 2 to 5) collected at the same time and location. Sites where the highest heavy-isotope enrichments were observed at Axial Volcano are indicated by crosses (Bag City), black stars (Marker 113) and black squares (Cloud). Black arrows indicate isotopic fractionation during nitrate consumption in diffuse fluids (seeFigure 3a).
Download figure to PowerPoint
 The development of a nitrate 15N and 18O isotope enrichment in diffuse fluids (due to faster consumption of the lighter isotopologues during biological reactions) depends on (1) the isotope effect associated with the biologic reaction and (2) the degree to which the N-removal process has advanced. The latter is, in turn, a direct function of the nitrate removal rate and the fluid flow velocity (i.e., nitrate supply). Indeed, the greatest nitrate15N and 18O enrichments were generally observed at the same vent sites in all sampling years (e.g., Bag City, Marker 113 and Cloud at Axial Volcano, Figures 5c and 5d), suggesting some microbiological, chemical or physical characteristic unique to these sites that make them conducive to nitrate removal and the associated enrichment of 15N and 18O in the residual nitrate. However, no significant relationship was observed between temperature, and the apparent nitrate 15N and 18O isotope enrichment. In fact, highest nitrate δ15N and δ18O values at the respective sites and sampling events encompass a broad range of temperatures (at the low-T end of the spectrum): e.g., ranging from ∼6°C at Cloud (Axial Volcano) to 71°C at Milli-Q (Endeavour Segment).
 Both assimilative (nitrate uptake by vent organisms) and dissimilative (denitrification) nitrate reduction are likely candidate processes to remove nitrate from HV, as both are known to fractionate nitrate isotopes in other environments [e.g., Brandes et al., 1998b; Voss et al., 2001; Lehmann et al., 2003; Sigman et al., 2003]. Biogeochemical evidence for respiratory denitrification by autotrophic bacteria has been reported previously for HV systems. For example, it has been shown that bacterial symbionts within the vestimentiferan Riftia pachyptila at the Genesis vent site (East Pacific Rise) can facultatively use nitrate as an electron acceptor for the oxidation of hydrogen sulphide, with either nitrite or N2 gas being the metabolic N product [Hentschel and Felbeck, 1993]. Similarly, sulphur oxidizing bacteria (e.g., Beggiatoasp.), found at Guaymas Basin hydrothermal vents, can accumulate nitrate at concentrations that are at least 3,000-fold higher than ambient concentrations in their vacuoles, which they subsequently also use for the oxidation of hydrogen sulphide [McHatton et al., 1996]. In diffuse fluids at Axial Volcano Butterfield et al. observed comparatively high concentrations of nitrite and nitrous oxide (20 to 600 nmol/L), typical intermediates and byproducts of denitrification in intermittently anoxic aquatic environments. While these previous studies have highlighted that active denitrification by microbes is likely to occur in hydrothermal fluids, nitrate assimilation for bacterial growth is another important process to be considered. The observed heavy-isotope enrichments were mostly restricted to the lower-T fluids (<50°C), where microbial cell densities are by far the highest (i.e., up to ∼10 times more than in seawater [Butterfield et al., 2004]). While ammonium is generally the form of fixed N that is preferred during N assimilation, by both photosynthetic organisms and bacteria [Dortch, 1990; Dugdale et al., 2007], Lee and Childress showed that S-oxidizing bacteria that live in symbiosis with the HV tubewormRiftia pachyptila exclusively assimilate nitrate, even under ammonium–replete conditions.
 The degree of community N and O isotope fractionation can potentially help us elucidate the pathway of nitrate removal in the HV diffuse fluids. Denitrification in the environment generally occurs with a significant nitrate isotope effect (ε) for both N and O isotopes of ∼20 to 30‰ [Cline and Kaplan, 1975, and references therein]. In contrast, nitrate assimilation seems to be associated with a significantly lower N-isotope effect of ∼5‰ in laboratory cultures and the natural environment [Altabet, 2001; Granger et al., 2004]. Here we attempt to estimate, for the first time, the community N and O isotope effects for nitrate removal in HV fluids, with the goal of assessing nitrate removal pathways. Analogous to the approach used to estimate the ammonium consumption N isotope effect above, the nitrate removal N and O isotope effects can be approximated using a closed system (Rayleigh) model, where a closed nitrate pool is consumed with a constant isotope effect as described by the equation [Mariotti et al., 1981]
where the initial nitrate concentration is calculated from the [Mg2+] content of the fluid, and assuming a strict linear mixing relationship between pure hydrothermal vent fluids and seawater (Figure 3a). As explained above, HV systems do not really behave as closed systems, so that the Rayleigh approach is likely to underestimate the community N isotope fractionation [Lehmann et al., 2003, 2007, 2009]. Alternatively, in an open steady state model, we assume that new seawater with a fixed δ15N for nitrate is constantly being mixed into the hydrothermal conduits, balancing the loss of nitrate by denitrification and/or N uptake so as to yield a steady state. The associated community N isotope effect is then calculated using the following equation [Altabet, 2001; Sigman et al., 2003]:
(Figure 6). The results from both approaches are shown in Table 1. The highest isotope effects were obtained using the open system model at Axial Volcano in 2006 (1.9‰ for 15εk and 8.6‰ for 18εk). The lowest isotope effects were observed using a closed-system model at the Endeavour Segment in 2008 (0.4‰ for both15εk and 18εk). Overall, the differences between the respective models were not very large. Strictly speaking, both models (closed-system and steady state) may not be representative of the real situation, as during hydrothermal circulation mixing is not necessarily continuous but rather episodic. However, modeling efforts bySigman et al.  demonstrated that for different mixing regimes between open steady state and closed systems (i.e., if there is sporadic mixing), the calculated isotope effect should fall between the two extremes.
Figure 6. δ15N and δ18O versus f(open system “steady-state” model), wheref is the fraction of nitrate consumed ([NO3−]initial − [NO3−]measured); [NO3−]initial is considered to be equal to [Mg2+]measured/[Mg2+] sw × [NO3−]sw, which is the [NO3−] for the corresponding [Mg2+] on the mixing line, i.e., ∼40 μM, in HV fluids near the seawater end-member (Figure 3a), at (a, b) Endeavor and Cobb Segments and (c, d) Axial Volcano. The isotope effects are estimated based on the slopes of the linear regression lines (dashed line for Cobb Segment). Symbol and color scheme as in Figure 2. Linear regressions for each year are shown. The colors of the lines correspond to the colors of the symbols (i.e., sampling years). See Table 1 for N isotope effect estimations and the r2 of the linear regressions.
Download figure to PowerPoint
Table 1. Estimated Nitrate N and O Isotope Effects According to a “Rayleigh” Closed System and an Open “Steady-State” System Model at Endeavour Segment, Cobb Segment and Axial Volcano on the Juan de Fuca Ridge From 2006 to 2009a
|15εk (‰)||18εk (‰)||15εk (‰)||18εk (‰)||15εk (‰)||18εk (‰)||15εk (‰)||18εk (‰)||15εk (‰)||18εk (‰)|
|Endeavour Segment|| || || || || || || || || || |
|Closed system||1.3 (0.49*)||2.6 (0.57*)||0.41 (0.57*)||0.41 (0.31*)||1.0 (0.97*)||2.3 (0.99*)||1.7 (0.71*)||4.3 (0.59*)||1.1 ± 0.6||2.4 ± 1.6|
|Open system||2.0 (0.56*)||3.9 (0.64*)||0.90 (0.64*)||0.96 (0.41*)||1.7 (0.94*)||3.7 (0.98*)||1.8 (0.71*)||4.7 (0.59*)||1.6 ± 0.5||3.3 ± 1.6|
|Cobb Segment|| || || || || || || || || || |
|Closed system||na||na||na||na||1.4 (0.85*)||3.5 (0.88*)||na||na||1.4||3.5|
|Open system||na||na||na||na||2.1 (0.76*)||5.3 (0.80*)||na||na||2.1||5.3|
|Axial Volcano|| || || || || || || || || || |
|Closed system||0.69 (0.41*)||3.2 (0.93*)||1.9 (0.57*)||3.0 (0.58*)||0.28 (0.11)||1.6 (0.63*)||1.2 (0.84*)||5.1 (0.89*)||1.0 ± 0.7||3.2 ± 1.4|
|Open system||1.5 (0.55*)||5.9 (0.89*)||2.9 (0.48*)||4.6 (0.51*)||0.90 (0.20*)||3.8 (0.60*)||1.9 (0.63*)||8.6 (0.75*)||1.8 ± 0.9||5.7 ± 2.1|
 The herein derived nitrate isotope effects are significantly lower than the ∼20–30‰ isotope effect expected for canonical denitrification, and generally closer to the N and O isotope effect of 5‰ expected for nitrate assimilation only. Assuming that denitrification occurs with a N and O isotope fractionation at levels similar to those reported for ocean oxygen minimum zones, the nitrate isotope data suggest that denitrification can account only for a minor fraction of the total nitrate removal in the hydrothermal fluids. It is also possible, however, that the N and O isotope fractionation of HV denitrification is suppressed by diffusion limitation [Lehmann et al., 2007]. Cell growth in the sub-seafloor primarily occurs on fixed surfaces (e.g., microbial mats [Moyer et al., 1995]). Analogous to the N-isotope-effect suppression due to substrate limitation reported for denitrification in benthic marine environments [Brandes and Devol, 1997; Lehmann et al., 2004, 2007], the diffusive nitrate supply to the actual sites of denitrification in such mats may be limiting so that nitrate is completely consumed. As a consequence, the nitrate N and O isotope effects would be significantly reduced, possibly with an apparent isotope effect of less than 2‰ [Brandes and Devol, 2002]. Finally, analogous to our above considerations regarding net N isotope effects of ammonium consumption, NO3−depletion is likely to be the net result of co-occurring nitrate consumption and production. Hence, here-reported N and O isotope effects represent community isotope effects that may partly be biased by the regeneration of nitrate and the isotope effects associated with the regeneration processes [e.g.,Lehmann et al., 2004]. At least for the nitrate N isotope ratios, the low apparent isotope effects may demonstrate the occurrence of gross nitrate production, by either nitrification or N2 fixation (and the subsequent remineralization/nitrification of newly fixed organic N to nitrate). Both processes would increase nitrate concentrations while decreasing nitrate δ15N values, and thus erase, or at least mask, any N isotope signals resulting from denitrification or nitrate assimilation. Figure 6 and Table 1 show that the community nitrate isotope effects are larger for 18O than for 15N, especially at Axial Volcano. A ratio of 18O versus 15N isotope enrichment >1 (between ∼1 to 4 (Figures 7a and 7c)) is atypical for stand-alone denitrification and/or nitrate assimilation in marine environments [Granger et al., 2004, 2008]. Given previous work [Lehmann et al., 2004; Sigman et al., 2005; Knapp et al., 2008; Bourbonnais et al., 2009], such a decoupling of the 15N versus 18O nitrate isotope enrichment is expected if quantitative N regeneration occurs simultaneously to net N consumption in the vent fluids. In the next section, we will discuss the observed δ15N-δ18O relationship in the context of possible N regeneration pathways within the HV fluids that can lead to the observed nitrate N-to-O isotope anomalies.
Figure 7. δ18O-NO3− versus δ15N-NO3− and Δ (15,18) versus Mg2+ concentration at (a, b) Endeavor and Cobb Segments and (c, d) Axial Volcano. Symbol and color scheme as in Figure 2 legend. The ratio of nitrate 18O versus 15N enrichment are estimated from slopes of the linear regressions (dashed-line for Cobb Segment in Figure 7a).
Download figure to PowerPoint
3.4. Nitrate N-to-O Isotope Anomalies: Possible Causes and Constraints on N Regeneration
 Analogous to the approach proposed by Sigman et al. , we quantify the deviation of nitrate isotope values from a 1:1 N versus O isotope fractionation relationship expected for pure denitrification/assimilation, using Δ(15,18):
where δ15Nm = ∼6‰ and δ18Om = ∼2‰ are the mean δ15N and δ18O of deep-seawater, respectively, and18ε/15ε is the ratio of N versus O isotope enrichment during denitrification, i.e., 1. An equal enrichment for N and O isotopes (Δ(15;18) = 0‰) can be assumed as a baseline characteristic for pure denitrification [Sigman et al., 2005; Granger et al., 2008] and NO3− assimilation [Granger et al., 2004; Lehmann et al., 2005]. From Figures 7b and 7d, it can be derived that the Δ(15,18) is negative in Mg2+-rich, low-T hydrothermal fluids, where denitrification (and possibly nitrate assimilation) results in the net consumption of nitrate. The Δ(15,18) was significantly higher (up to ∼−8.5‰) at Axial Volcano compared to the Endeavor and Cobb Segments (up to ∼3‰). The nitrate isotope anomalies also varied between years, (with generally higher anomalies in 2006, 2007, and 2009) although no general temporal pattern could be discerned.
 We argue that the Δ(15,18) anomaly observed at some sites is due to the gross production of nitrate coupled to microbial nitrate consumption (assimilation and/or denitrification). However, we currently have few quantitative constraints on the different N fluxes involved or the physical and chemical characteristic of the subsurface biosphere (e.g., average fluid residence time, volume of hydrothermal conduits, and timing of mixing), so it is impossible to derive rates of processes from our isotope data. Therefore, we will only qualitatively discuss possible causes for the observed nitrate N and O isotope anomalies below. The exact causes for generally larger nitrate isotope anomalies at Axial Volcano compared to the Endeavor and Cobb Segments are unclear, as are the causes of any inter-annual variations. However, supplementary data on bacterial cell densities suggest a link between bacterial growth/activity and the observed expression of the N-to-O isotope anomalies. In fact, we observed a significant relationship (r2= 0.61, p-value = 0.0001) between bacterial cell counts in diffuse fluids (data from J. A. Baross and J. F. Holden laboratories) and the N-to-O nitrate anomalies for the years where cell counts were available (2008 and 2009) at Axial Volcano (Figure 8). Yet, at Endeavour Segment, where maximum NH4+concentrations were observed in the subsurface fluids, the nitrate N- versus O isotope anomaly was rather moderate and there was no correlation with bacterial cell density. Therefore causal links between bacterial biomass, “anomalous” nitrate isotope behavior and, possibly, nitrate regeneration through either ammonium or nitrite oxidation remain uncertain.
3.4.1. Partial Nitrification and N2 Fixation
 Ammonium oxidation in deep-sea hydrothermal vent plumes of the Main Endeavor Field, Juan de Fuca Ridge, has been reported byLam et al. [2004, 2008]. Byrne et al. also documented the diversity and activity of anammox bacteria in hydrothermal vent chimney samples on the Mid-Atlantic Ridge. Here, the partial re-oxidation of ammonium at the seawater end of the redox gradient in low-T, aerobic vent fluids would leave the residual NH4+ enriched in 15N, while the nitrate produced would be depleted in 15N [Delwiche and Steyn, 1970, and references therein]. While the δ15N of newly nitrified nitrate is directly dependent on the δ15N of the ammonium pool (with an offset that corresponds to the nitrification N-isotope effect), nitrification represents an absolute source with regard to the O isotopes.
 The general understanding at the moment is that two-thirds of the O atoms in new nitrate are being derived from ambient water, and that, because of important isotope effects during oxygen atom incorporation during ammonium and nitrite oxidation, theδ18O of nitrified nitrate should be 0.7–8.3‰ lower than the δ18O of the seawater (which was measured to be ∼0‰ in our 2004 hydrothermal fluid samples, unpublished results) [Buchwald and Casciotti, 2010; Casciotti et al., 2010]. On the other hand, 16O is preferentially extracted during nitrate reduction, causing the δ18O of the eliminated NO3− to be even lower than the δ18O of the nitrified nitrate. Hence, while nitrification tends to counteract the denitrification/assimilation–driven enrichment of both the δ18O and δ15N, this balancing effect is generally more pronounced for N than for O so that it can produce nitrate N-vs-O anomalies (more precisely, Δ(15,18) minima) [Sigman et al., 2005]. The Δ(15,18) effect is most marked if ammonium oxidation is not complete, either in an ammonium-replete setting, or when ammonium uptake and nitrification occur simultaneously (i.e., ammonium branching, with the N isotope effect associated with ammonium assimilation being lower than the N isotope effect associated with nitrification) [e.g.,Wankel et al., 2007].
 The N isotope balance predicts that in the HV fluids where partial nitrification generates low-δ15N nitrate, it simultaneously leaves 15N-enriched ammonium substrate behind (according to the nitrification N isotope effects of 14–38‰ [Delwiche and Steyn, 1970, and references therein]). The observed increase of the δ15N-NH4+in the low-T fluids at some sites of the Endeavor and Cobb Segments, as well as Axial Volcano (Figure 4), hence, confirms that oxidation of ammonium is occurring to some extent. We would expect the nitrification-driven Δ(15,18) minimum to be most pronounced at Endeavour Segment and, to a lesser extent, Cobb Segment, where the measured ammonium concentrations in diffuse HV fluids (up to ∼65μmol/L and ∼10 μmol/L, respectively) were often higher than the nitrate concentrations. Furthermore, the average δ15N of the hydrothermally discharged ammonium was also significantly lower at these sites compared to Axial Volcano, further enhancing the expression of the nitrate isotope anomaly. As discussed above, there were indeed systematic inter-site variations with regard to the Δ(15,18). However, curiously, the lowest Δ(15,18) was observed at Axial Volcano, where the influence of incomplete ammonium oxidation is likely to be less important and where, as a consequence, the Δ(15,18) should be less pronounced ([NH4+] in diffuse fluids was close to 0 μmol/L at most sites).
 N2 fixation in hydrothermal vents has been previously documented by Mehta et al. [2003, 2005] and Mehta and Baross  who found expressed nitrogenase genes (nifH) in anaerobic hydrothermal fluids at Axial Volcano and who were able to isolate a methanogenic archeaeon that can fix nitrogen at a temperature up to 92°C. N2 fixers are assumed to thrive in environments where fixed N forms that are energetically more favorable are scarce. Moreover, an anaerobic environment may be conducive to N2 fixation, as nitrogenase, the enzyme involved in N2-fixation, is inhibited by oxygen [e.g.,Berman-Frank et al., 2005]. N2 fixation may occur in microsites, e.g., microbial mats and particulate material, where O2and DIN concentrations may be low, providing an important source of bioavailable N for hydrothermal vent organisms in low-T fluids [e.g.,Proctor, 1997; Zehr et al., 2003]. N2 fixation produces organic material with a δ15N of ∼−2 to 0‰ [Carpenter et al., 1997, and references therein]. Upon ammonification of the N2-fixation-derived biomass and subsequent (incomplete) nitrification after mixing of the anaerobic vent fluids with oxic seawater, low-δ15N N can be transferred to the NO3− pool. Given the nitrate O isotope systematics described above, the δ18O, however, is rather insensitive toward this N2fixation as the incorporation of O atoms into the newly produced nitrate molecule does not discriminate between possible origins of the precursor N compounds. The above-described nitrification/denitrification mechanism to produce theδ15N-δ18O decoupling would, hence, be enhanced if newly produced nitrate is derived from the remineralization and nitrification of chemosynthetically fixed N2. And even in the case of complete nitrification of the ammonium, a negative Δ(15,18) can be expected [Bourbonnais et al., 2009].
3.4.2. Nitrate/Nitrite Redox Cycle
 Nitrate is not only regenerated by the oxidation of ammonium, it can also originate from the re-oxidation of nitrite. Under suboxic conditions in hydrothermal vent fluids, nitrate is reduced to nitrite. Along redox gradients or upon mixing of anaerobic and oxic waters, a large fraction of product nitrite may be re-oxidized to nitrate. We are only beginning to understand nitrite N and O isotope systematics in nature [Casciotti and McIlvin, 2007; Casciotti et al., 2010; Buchwald and Casciotti, 2010]. For example, factors that control observed offsets between the δ15N of nitrate and nitrite in the Eastern Tropical North Pacific OMZ (up to 30‰) are uncertain. However, previous work suggests that nitrite re-oxidation is an even more efficient mechanism for lowering the nitrate Δ(15,18) [Casciotti and McIlvin, 2007; Sigman et al., 2005], as it adds high δ18O to the nitrate pool. That is, by mass balance, nitrate reduction from a particular nitrate pool and subsequent (complete) nitrite re-oxidation should produce nitrate with a similarδ15N to that initially consumed. The δ18O of the re-oxidized nitrite, on the other hand, will likely be higher than theδ18O of the nitrate consumed, due to the “branching fractionation” (preferential extraction of 16O) during nitrate reduction (i.e., produced nitrite is enriched in 18O) and the incorporation of an O-atom with a relatively higherδ18O (compared to the δ18O of the O-atom lost during nitrate reduction) during re-oxidation of nitrite to nitrate. If portions of the nitrite are further reduced to gaseous forms of N, both the nitriteδ15N and the δ18O are increased in parallel. Therefore, while this redox cycle leaves the δ15N of nitrate essentially unchanged, it would act to increase the δ18O. If anything, we would expect nitrite oxidation to be the driver of N-to-O nitrate isotope anomalies particularly in higher-T diffuse fluids, where nitrate reduction to nitrite (favored by the more reducing conditions) and subsequent nitrite re-oxidation after mixing with oxygenated deep seawater, is most likely to occur. However, as mentioned earlier, the highest nitrate isotope anomalies were not observed in the highest T (generally < ∼50°C) diffuse fluids. In summary, several processes (partial nitrification, N2 fixation, and nitrite oxidation) can theoretically produce similar negative nitrate Δ(15,18) signatures [Sigman et al., 2005; Casciotti and McIlvin, 2007], just as observed at the Juan de Fuca Ridge vent sites. In the next section, we will present results from a simple isotope box model, which we used to assess the relative fluxes of the candidate processes that could explain the N-to-O nitrate isotope anomalies in diffuse vent fluids.
3.4.3. Estimates on the Relative Importance of the Possible Nitrate Regenerating Processes
 We attempt here to assess the role of ammonium oxidation, N2fixation or nitrite re-oxidation in the hydrothermal conduits of the Juan de Fuca Ridge, applying a simplified steady state one-box model (analogous to the one we used in previous work [Bourbonnais et al., 2009]) in order to calculate Δ(15,18) as a function of relative changes in potential N-regenerating and consuming processes (Figure 9). In this model, we included nitrate inputs through partial nitrification of hydrothermal ammonium, N2fixation and seawater mixing, and nitrate removal through bacterial uptake and/or denitrification. Finally the model includes the internal nitrite/nitrate cycle, where a portion of the nitrite from nitrate reduction is re-oxidized to nitrate. Theδ15N of ammonium-derived nitrate is calculated according toMariotti et al. :
where f is the fraction of reactant remaining, δ15N-NH4+initial is the δ15N of the initial reactant pool, and εnit is the kinetic isotope effect of ammonium oxidation to nitrate. We used an average N isotope effect of 26‰ during ammonium oxidation [Delwiche and Steyn, 1970, and references therein]. N2 fixation (i.e., the remineralization of newly fixed OM) adds nitrate with δ15N of −1‰ (with insignificant isotopic fractionation) [Carpenter et al., 1997, and references therein]. We assumed that part of the assimilated nitrate would be returned following organic matter ammonification and nitrification (recycled production term in Table 2). Therefore, net nitrate uptake is the gross nitrate uptake minus the recycled production. Independent of the original N source (hydrothermal ammonium versus OM from N2 fixation or bacterial uptake), a δ18O of −3.8‰ is assumed for nitrified nitrate (mean value taken from Buchwald and Casciotti ). With regard to the internal nitrite/nitrate cycling, complete nitrite oxidation returns nitrate with a δ15N equal to the original nitrate, and a δ18O of nitrate of 0‰, as also assumed in the study by Sigman et al. . Subsurface mixing with seawater adds nitrate with a δ15N of ∼6.0‰ and a δ18O of ∼2.0‰. Nitrate removal occurs either by assimilation or by denitrification. We used average isotope effects of 5‰ for nitrate assimilation [Altabet, 2001], 25‰ for denitrification [Cline and Kaplan, 1975, and references therein] and 1.5‰ [Brandes and Devol, 2002] for nitrate consumption occurring in bacterial mats. Figure 10 shows the model results for both Axial Volcano and Endeavour Segment. We simulated seven representative scenarios, in which we varied the relative importance of single N fluxes in order to gain information on their respective potentials for generating the observed Δ(15,18) minima in the diffuse vent fluids (see Table 2 and Figure 10 for more details on the parameters used for the different scenarios). In the first 3 scenarios, nitrate is produced only by the oxidation of hydrothermal ammonium and we varied the relative rates and extent of hydrothermal NH4+ oxidation relative to net nitrate uptake. In scenarios 4, 5, and 6, the oxidation of hydrothermal ammonium was suppressed and we varied the relative rate of nitrate production from N2fixation or nitrite re-oxidation relative to net nitrate uptake. Finally, in scenario 7, partial ammonium oxidation and nitrite re-oxidation were combined. Across all considered scenarios, we also varied the ratio of denitrification and N uptake, and differentiated between denitrification in the open HV conduits versus that by microbial mats and biofilms on conduit walls (seeTable 2for more detail). For all scenarios, decreasing the mixing with deep-seawater (while concomitantly increasing the input of nitrate from one of the three processes mentioned above), caused a decrease in the Δ(15,18).
Figure 9. Simplified steady state model used in section 3.4.3 (adapted from Sigman et al.  and Bourbonnais et al. ): (a) representation of the addition of nitrate from N2 fixation (with a δ15N of −1‰) or from hydrothermal-ammonium oxidation with aδ15N that corresponds to the integrated product of partial ammonium oxidation [Mariotti et al., 1981], and a δ18O of −3.8‰ (average value taken from Casciotti et al. ); (b) the input of nitrate from mixing with deep-seawater with aδ15N of ∼6‰ and a δ18O of ∼2‰; (c) the gross nitrate removal by nitrate assimilation and/or denitrification; (d) the remineralization of newly biosynthesized organic N to ammonium, coupled to nitrification, returning nitrate with a δ18O of −3.8‰ (and not changing the nitrate δ15N); (e) representation of the internal cycle of NO3− reduction and NO2−re-oxidation (see text for details).
Download figure to PowerPoint
Figure 10. Model results for Δ(15,18) for simulation scenarios presented in Table 2 for Axial Volcano (plain lines) and the Endeavour Segment (dashed lines): (a) partial ammonium oxidation, (b) N2fixation, and (c) nitrite re-oxidation (rates always relative to net nitrate uptake). Maximum Δ(15,18) anomalies observed at Axial Volcano (∼8.3‰) and Endeavour Segment (∼2.4‰) are indicated by horizontal black lines.
Download figure to PowerPoint
Table 2. Model Simulation Scenarios Described in Section 3.4.3a
| ||Fractional Crustal NH4+ Oxidation (Figure 10a)||N2 Fixation (Figure 10b)||NO2−Re-oxidation (Figure 10c)||NO2−Re-oxidation and Fractional NH4+ Oxidation (Figure 10c)|
|Scenario 1||Scenario 2||Scenario 3||Scenario 4||Scenario 5||Scenario 6||Scenario 7|
|Fraction NH4+ consumed (%)||25||50||75||na||na||na||75|
|Denitrification/gross nitrate uptake||0.25||0.5||0.25||0.5||0.25||0.4||0.4|
|% denitrification in microsites||0||50||25||25||50||0||0|
|Recycled production/NO3− assimilation||0.5||0.75||0.5||0.75||0.5||0.2–0.5||0.2–0.5|
 Our model simulation suggests that all 3 processes are capable of generating nitrate N and O isotope anomalies of a magnitude similar to that actually observed at the Endeavour Segment (minimum Δ(15,18) of −2.4‰). In our simplistic model, nitrite re-oxidation and partial ammonium oxidation can generate very low Δ(15,18) values (<−4.0‰), especially for a lower fraction of ammonium consumed, when N isotope fractionation is fully expressed, whereas N2 fixation has less potential to generate negative Δ(15,18) anomalies. Even at high rates of N2 fixation with no preformed NO3−, the Δ(15,18) does not exceed ∼−3‰. Therefore, at least at Axial Volcano, it seems very unlikely that the entire nitrate isotope anomaly can be attributed to the remineralization of fixed N. Figure 10c(Scenario 7) shows that only a combination of processes (for a given ratio of 0.5 for nitrite re-oxidation/net nitrate uptake and for a 75% fractional ammonium consumption can generate a Δ(15,18) of <−8‰, as at Axial Volcano (−8.3‰). While it seems clear that nitrate production is important in the investigated diffuse hydrothermal fluids, too many unknowns preclude a more quantitative assessment of actual N regeneration pathways. The observed nitrate N and O isotope data are consistent with all of the above candidate processes, but without independent data on actual rates, the Δ(15,18) data do not allow us to reliably predict the relative importance of nitrite re-oxidation and nitrification of hydrothermal ammonium versus nitrification of N2-fixation derived ammonium, particularly at those sites where nitrification is incomplete [Wankel et al., 2007].