One unit of enzyme activity is defined as the amount of NR that is capable of catalyzing the conversion of 1 μmol nitrate to nitrite per minute at optimal pH and temperature. Assimilatory nitrate reductase analyses were carried out by reconstituting the freeze-dried Arabidopsis, yeast, and maize enzymes in 25 mM potassium phosphate, 0.1 mM Na₂EDTA, pH 7.5, at 22–25°C, to a final concentration of 1 unit/mL. For the Aspergillus enzyme, a 100 mM potassium phosphate buffer, 0.1 mM Na₂EDTA, pH 7.5 was used. The NR K_m for NO⁻₃ in Arabidopsis and maize ranges from 15 to 40 μM (Su et al., 1997; Campbell, 1999), 199 μM for Aspergillus (Gilliam et al., 1993), and 470 μM for the dissimilatory NR in E. coli (Adams and Mortenson, 1982). There is no information for the specific yeast NR used in our study, but NRs from other genera of yeast have a K_m ranging from 110 (Hipkin et al., 1986) to 740 μM (Morozkina et al., 2005).

The enzyme reaction solution contained 1 unit (1 mL) of reconstituted NR, 165 mL of additional phosphate buffer (pH 7.5), and 70 μmol of NADH or NADPH for the plant and fungal enzymes, respectively. Additionally, the Aspergillus solution contained 4 μmol of FAD. After mixing, 10 mL of solution was added to each of 16 sterile 15 mL centrifuge tubes. One mL of KNO₃ solution (0.2 M KNO₃) was then added to each centrifuge tube, and the tubes were placed into a water bath at 30°C (25°C for Aspergillus) for 30 min. The use of 0.2 mmol of KNO₃ in the reaction tubes was calculated to provide a surplus of NO⁻₃ for the enzymatic reaction, while minimizing the possibility that residual NO⁻₃ might saturate the mass spectrometer detectors. Therefore, the enzyme reactions were not necessarily operating under NO⁻₃ saturating conditions. The KNO₃ was enriched with different levels of ¹⁵N from natural abundance to >99 atom % ¹⁵N to examine how substrate ¹⁵N enrichment influences the estimation of NR ¹⁵N discrimination using an IRMS. At the end of the incubation, samples were placed into 20 mL scintillation vials. The reaction was halted by adding 0.25 mL of 1 M HCl and swirling the vials (decreasing pH to below 4). After 1 min, 40 μL of 10 M KOH were added to the solution, raising its pH to 10–12 to prevent oxidation of NO⁻₂ to NO⁻₃. Concentrations of NO⁻₂ and NO⁻₃ were determined using the Griess reaction (Miranda et al., 2001).

The reaction catalyzed by dissimilatory E. coli NR requires reducing conditions. The enzyme was reconstituted in 2 mL of cold degassed deionized water and added to 55 mL of degassed buffer (0.125 M MOPS, pH 7.0) containing 30 mg benzyl viologen (BV; 1,1′-Dibenzyl-4,4′-bipyridinium dichloride, Sigma) as an electron donor. Ten milliliter of buffer/enzyme solution were added to 15 mL centrifuge tubes containing 0.2 mmoles of KNO₃ of various ¹⁵N enrichments and 0.25 mL of a sodium dithionite/sodium bicarbonate solution (500 mg of each added to 50 mL degassed DI H₂O) to maintain reducing conditions. If the blue color of the reduced BV dye started to fade, additional dithionite solution was added to the vial. The vials were incubated in a water bath at 30°C for 90 min. To stop the reaction, vials were shaken vigorously to oxidize the BV. At the end of the incubation, 75 μl of 10 N NaOH was added to each tube to minimize conversion of NO⁻₂ to NO⁻₃.

For all forms of the NR enzyme, we analyzed the ¹⁵N isotope composition of the headspace N₂O produced through bacterial denitrification of NO⁻₃ and NO⁻₂. Bacterial cultures and solutions were prepared as in Böhlke et al. (2007). The bacterial cultures used in this study were Stenotrophomonas nitritireducens (ATCC# BAA-12) and Pseudomonas chlororaphis (ATCC# 43928), both of which were obtained from the American Type Culture Collection (www.atcc.org). Both species lack the enzymatic ability to reduce N₂O to N₂. S. nitritireducens is limited to reducing NO⁻₂, while P. chlororaphis is capable of reducing both NO⁻₃ and NO⁻₂ to N₂O. This specificity allowed a sequential reduction of our experimental samples to estimate the ¹⁵N concentrations of the NO⁻₃ and NO⁻₂ independently.

The denitrifying bacteria were streaked on tryptic soy agar (TSA) plates composed of TSA (40 g/L) and tryptic soy broth (TSB; 10 g/L) for S. nitritireducens, and TSA (40 g/L) and KNO₃ (1 g/L) for P. chlororaphis. The TSA plates for the S. nitritireducens cultures were prepared with 10 mM Na₂WO₄ to minimize any potential NO⁻₃ reduction (Nielsen et al., 2004). Both cultures were grown aerobically at 22–24°C. After 5–7 days, bacterial colonies were streaked onto a second plate. If re-plated more than 2 or 3 times, higher background levels of N₂O were observed for both bacteria. Therefore, all N₂O analyses were made using colonies from the second plate; additional analyses required starting the process over from the original frozen aliquot.

Individual S. nitritireducens colonies were transferred 4–5 days after streaking to sterile 50 mL centrifuge tubes with 35 mL of a growth medium containing TSB solution (15.0 g/L), (NH₄)₂SO₄ (0.25 g/L), and KH₂PO₄ (2.45 g/L). Tubes were capped leaving an air headspace and agitated on a reciprocal shaker for 1–2 days at 22°C. The tubes were opened daily to allow for aerobic growth. The contents of the centrifuge tubes were combined and an antifoaming agent was added, 2–3 drops of Antifoam B (Sigma) per 105 mL solution. Four milliliter of the combined solution were pipetted into each 20 mL vial, which was capped with a septum and crimped, and then placed on a manifold that purged the vials with N₂ gas for 4 h to remove any residual oxygen, CO₂, and N₂O. The contents of these vials were used for ¹⁵N analysis of NO⁻₂, as well as for providing substrate for the NO⁻₃ reducing P. chlororaphis cultures.

Pseudomonas chlororaphis cultures were used to inoculate a starter tube containing 1.5 mL of TSB solution (30 g/L), KNO₃ (1 g/L), and (NH₄)₂SO₄ (0.066 g/L), which was placed on a reciprocal shaker for 1 day. TSB solution (400 mL) was added to a 500 mL media bottle, which was inoculated with the contents of the starter tube and sealed tightly. The media bottle was placed on the reciprocal shaker, and the bacteria were allowed to grow for 6–10 days under anaerobic conditions. At the end of the 6–10 days, the solution was centrifuged (10 min at 20°C and 6500 rpm). The supernatant was decanted into a large beaker and the pellets remaining in the centrifuge tubes were resuspended in 60 mL of the supernatant solution. An antifoam agent—one drop of Antifoam B (Sigma) per 60 mL—was added and the concentrated bacterial solution was mixed thoroughly. Two milliliter of this solution were pipetted into each 20 mL headspace vial, capped with a septum and crimped, and then purged for 4 h on the manifold system with N₂ gas. The manifold system was carefully sterilized between sample sets.

Solutions from the enzyme reaction were added to the purged S. nitritireducens vials to determine the ¹⁵N enrichment of N₂O derived from NO⁻₂. The amount of solution added to the vial was determined by the concentration of the NO⁻₂ in the solution. A target value of 0.025 μmol of NO⁻₂ were added to each vial. This corresponded to volumes ranging from 0.25 to 1.5 mL depending on the enzyme. The vials were then placed on a reciprocal shaker for 24 h. At the end of 24 h, a sample was removed from the vial using a gas-tight syringe and injected into a vial containing P. chlororaphis. Again, a volume of solution equivalent to 0.025 μmol of NO⁻₃ was added to the vials (volumes ranged from 25–200 μL). Amounts of N added to all samples were matched to ensure that the samples remained in the linear range of the detectors in the IRMS. The P. chlororaphis vials were placed on the reciprocal shaker.

Once the subsample had been removed from the vials containing S. nitritireducens, and after 24 h incubation for the P. chlororaphis vials, 0.2 mL of 10 M NaOH were added to both sets of vials killing the bacteria and removing any residual headspace CO₂. All vials were then taken to the UC Davis Stable Isotope Facility where they were analyzed for [N₂O] and ¹⁵N (ThermoFinnigan GasBench + PreCon trace gas concentration system interfaced to a ThermoScientific Delta V Plus IRMS, Bremen, Germany). To control for potential contamination, blanks (including bacterial solution alone and with DDI water), natural abundance NO⁻₂ and NO⁻₃ samples, and appropriately ¹⁵N-enriched NO⁻₂ and NO⁻₃ standards, were included in each sample run. The blanks also allowed us to correct for any N₂O derived from residual NO⁻₃ remaining in the P. chlororaphis solution and instrumental leaks. Samples of the enzyme reaction solution were also tested for inorganic N contamination and found not to contain detectable amounts. No remaining inorganic N was found in the spent nutrient solution after the bacterial incubations. Thus, conversion of the NO⁻₃ and NO⁻₂ to N₂O was complete and quantitative.

We defined ⁴⁵R and ⁴⁶R as the ratios of measured ion peak areas [⁴⁵R = (m/z 45)/(m/z 44), and ⁴⁶R = (m/z 46)/(m/z 44)]. ¹⁵R or the ratio of ¹⁵N/¹⁴N of the N₂O headspace was determined as:

where ¹⁷R (¹⁷O/¹⁶O) is fixed at the isotopic oxygen signature of ambient N₂O (¹⁷R = 0.00038855; Kaiser et al., 2003). ¹⁵R may equivalently be determined using ⁴⁶R as described in Stevens and Laughlin (1998):

where ¹⁸R (¹⁸O/¹⁶O) is fixed at the isotopic oxygen signature of N₂O (δ¹⁸O = 44.62 and ¹⁸R = 0.0020947; Kaiser et al., 2003).

Isotope analyses are often expressed in terms of either δ¹⁵N or atom % ¹⁵N. δ¹⁵N is expressed as the per mil difference between the ¹⁵N/¹⁴N ratio of a sample relative to the primary reference material for nitrogen, atmospheric N₂(“Air,” 0‰):

For all analyses, initial measurements are made relative to a working reference gas of the same species (i.e., N₂O, N₂), co-measured during analysis. Subsequently, all measurements were corrected using working standards previously calibrated to the following standard reference materials IAEA-N1, IAEA-N2, IAEA-N3, USGS-40, and USGS-41.

There are, however, no internationally accepted reference standards for N₂O at high ¹⁵N enrichments and, because of this, we compared the ¹⁵N/¹⁴N ratio of the sample N₂O to the ¹⁵N/¹⁴N ratio of the substrate NO⁻₃ as determined by the quantitative combustion of substrate NO⁻₃ to N₂ (Stickrod and Marshall, 2000; See Supplemental Table 1) via a PDZ Europa ANCA-GSL Elemental Analyzer interfaced to a PDZ Europa 20-20 IRMS (Sercon Ltd., Cheshire, UK) according to:

This approach minimizes the impact of absolute errors in calibration because only differences between the substrate and product are considered, and the δ notation is very sensitive to large absolute differences between the sample and reference ratios (Fry, 2006). Also, in order to examine the performance of the bacterial reduction method at high levels of ¹⁵N enrichment, we calculated the atom % ¹⁵N from the atomic ratios, ¹⁵R:

Alternatively, the atom % ¹⁵N could be determined from the molecular ratios (e.g., ⁴⁵R and ⁴⁶R) as outlined in Stevens and Laughlin (1998).

The kinetic isotope effect (KIE) for a chemical reaction is the relative reactivity of molecules labeled with different isotopes (Farquhar et al., 1989). Data for competitive KIE experiments is generally taken at different values of fractional reaction of the initial substrate and fitted to Equation (6a), where f is fractional conversion of the initial substrate, R₀ is the isotopic ratio of the initial substrate (equal to the isotopic ratio of the product at 100% conversion), and R_f is the isotopic ratio of the product at fractional reaction f. Here, because all reactions were halted at very low conversions of the initial NO⁻₃ (0.5% maximal conversion), the KIE is well approximated by the molar ¹⁵N/¹⁴N ratio of the initial nitrate substrate to the nitrite product (Equation (6b); Bigeleisen and Wolfsberg, 1958):

Therefore, the KIE is related to the measured isotopic discrimination (Δ) by

Alternatively, it can be derived by the following equation (Evans, 2001):

where δ¹⁵N₀ and δ¹⁵N_f are the ¹⁵N content of the initial substrate and the product at fractional reaction f, respectively. Enzyme discrimination data will be presented using the Δ notation for the remainder of this paper.

The bacterial reduction method was compared with a chemical reduction method to examine potential bacterial discrimination against ¹⁵N during the reduction of NO⁻₂ to N₂O. For the chemical reduction, we used sulfamic acid to reduce NO⁻₂ to N₂O (Granger and Sigman, 2009). As a result of this reaction, the solution NO⁻₂ supplies one of the N atoms in the N₂O molecule, while sulfamic acid supplies the other (Granger and Sigman, 2009). No remaining NO⁻₃ was observed in the samples after the reduction.

The experiments were formulated as a randomized complete block design. Blocks were represented by the enzyme batch to control for inter-batch variation in enzyme activity and purity. Depending on the particular enzyme, a single batch of enzyme was used to produce from one to three repetitions of each ¹⁵NO₃ concentration to NO⁻₂. A linear mixed effects model (PROC MIXED, SAS 9.1, Cary, NC, USA) was used to analyze the data.

The results show that ¹⁵N discrimination in the NR-catalyzed reaction is similar across eukaryotic NRs, particularly among non-recombinant NRs. All NRs had ¹⁵N discrimination falling between 22 and 32‰ (corresponding to KIEs of 1.023 and 1.033, respectively; Table 2). This is higher than the 15‰ typically cited in the literature (e.g., Robinson, 2001; Werner and Schmidt, 2002). The relatively high ¹⁵N discrimination observed here may partially derive from differences in how the measurements were made. Other work has generally focused on in vivo discrimination (e.g., Mariotti et al., 1982) or in vitro experiments using isolated organelles or cytosolic fractions (e.g., Ledgard et al., 1985). In vitro experiments with purified enzymes typically produce greater isotopic discrimination than experiments with cellular organelles, multiple enzymes, or whole organisms because additional factors diminish ¹⁵N discrimination as the catalytic system becomes more complex (Werner and Schmidt, 2002). We performed in vitro assays using purified NRs and electron donors, providing unlimited substrate, and operating under optimal conditions to prevent subsequent conversion of the product NO⁻₂. Olleros-Izard (1983: cited in Schmidt and Medina, 1991) used a similar approach with purified corn NR and observed a discrimination of 30‰, which agrees with our values for non-recombinant enzymes. Karsh et al. (2012) using purified enzymes and the bacterial reduction method found that the assimilatory NR from A. niger discriminated against ¹⁵NO⁻₃ by 26.6 ± 0.2‰ (Karsh et al., 2012), again a value similar to ours.

The four assimilatory eukaryotic NRs (Arabidopsis, yeast, maize, and A. niger) are structurally similar, although the two recombinant enzymes (Arabidopsis and yeast) lack a portion of the complete native enzyme. For example, the recombinant Arabidopsis NR, which is natively homotetrameric, is missing in three of its four subunits the Mo-MPT cofactor, and thus the NO⁻₃ reducing site (Skipper et al., 2001). Such differences in structure may account for the lower ¹⁵N discrimination observed in the recombinant NRs than in the maize and A. niger enzymes (Table 1). Some of the observed difference between the Arabidopsis and maize NRs may be due to differences between monocot and eudicot NRs; however, there is no published evidence for this possibility. Differences in glycosylation between native enzymes and recombinant enzymes expressed in P. pastoris also might affect enzymatic activity and possibly isotope discrimination (e.g., Henriksson et al., 2003).

Despite the structural disparity (Moreno-Vivián et al., 1999) between the eukaryotic and prokaryotic NRs, the E. coli NR did not grossly differ in NR ¹⁵N discrimination from the other NRs tested.

The electron donor did not strongly affect ¹⁵N discrimination. The plant-derived NRs (Arabidopsis and maize) used NADH as an electron donor, while the fungal NRs used NADPH. The similar level of discrimination that we found despite the difference donors supports the hypothesis that substantial structural differences among the eukaryotic enzymes are largely localized to the FAD binding domain (Campbell, 1999; Karsh et al., 2012). The dissimilatory E. coli NR used BV as an artificial electron donor, which provides electrons to the Mo-MGD cofactor directly, in contrast to the biological donor quinol, which provides electrons through the FeS or cytochrome c cofactors. No studies have measured the effects of the biological electron donor on ¹⁵N discrimination in nitrate reduction in prokaryotes, and it is not clear whether the use of a different electron donor would affect ¹⁵N discrimination. Nonetheless, use of an artificial electron donor (e.g., methyl viologen) rather than NADH had little to no influence on eukaryotic NR ¹⁵N discrimination (Karsh et al., 2012).

The systematic and large deviation from expected values of the KIEs at high ¹⁵N enrichment (particularly >50 atom %; Figure 2) cannot derive from substrate dependent changes in the rate limiting nature of N-O bond cleavage because the observed KIEs are greater than their theoretical limits (Tcherkez and Farquhar, 2006). Our measurements at high ¹⁵N enrichment were affected by artifacts. Similar artifacts may have affected the ¹⁵N pulse labeling experiment in our previous work (Bloom et al., 2010). The following discusses some of these potential artifacts.

First, although the sequential bacterial reduction method to assess ¹⁵N discrimination at natural abundance ¹⁵N worked very well, we observed increasingly large standard errors and anomalously low ¹⁵N enrichment in the N₂O derived from NO⁻₃ reduction by P. chlororaphis as ¹⁵N NO⁻₃ enrichment increased (Supplemental Table 2). At enrichments of 99 atom %, the N₂O derived from the NO⁻₂ was substantially more enriched in ¹⁵N than the N₂O derived from NO⁻₃ (97.6 vs. 58.3 atom % ¹⁵N; Supplemental Table 2). If we used a different NO⁻₃ reducing bacterium, Pseudomonas aureofaciens (Sigman et al., 2001), the observed differences were even larger (97.6 vs. 7.9 atom %; Supplemental Table 2). By contrast, S. nitritireducens produced N₂O from NO⁻₂ with ¹⁵N enrichment comparable to that produced via chemical reduction (Figure 1). These results suggest that the sequential bacterial reduction method should be limited to ¹⁵N enrichments of less than 10 atom %. The reason for the variation among the bacterial strains is not clear, although ages of the bacterial strains or possible bacterial contamination may play a role. NO⁻₃ concentrations in the spent bacterial reduction solutions contained no measurable NO⁻₃ after reduction, so the reaction was apparently quantitative (data not shown).

Second, instrument performance and the mathematical models currently available for the calculation of ¹⁵N in N₂O at high levels of ¹⁵N enrichment are likely responsible for some of the observed variation. Using an IRMS rather than using an elemental analyzer coupled with a mass spectrometer contributed to the observed error, particularly at ¹⁵N enrichments greater than 50%, however, the use of the IRMS was necessary to measure NR discrimination at natural abundance ¹⁵N concentrations. High enrichments of ¹⁵N lead to problems with IRMS collector saturation, and we could not measure the 44 peak necessary for the calculation of the mass ratios.

The results from highly ¹⁵N enriched substrates presented here illustrate a more general methodological issue regarding stable N isotopes: one should avoid the use of highly ¹⁵N enriched substrates in discrimination studies, particularly if small experimental changes in ¹⁵N concentration are expected. High ¹⁵N enrichments exacerbate methodological difficulties in sample preparation, measurement, and instrument calibration. This ultimately increases the final experimental error and unnecessarily confounds experimental outcomes.

Indeed, the problems in the use of highly enriched ¹⁵N encountered here may affect measurements at larger scales. For example, it is common in agricultural field studies to use enriched ¹⁵N as a tracer for the cycling of N among different soil pools. Experiments conducted using enriched ¹⁵N tracers consistently estimate lower N cycling rates than those conducted using natural abundance or non-isotopic methods (e.g., Hauck and Bremner, 1976; Androsoff et al., 1995; Ladha et al., 2005). Other factors beyond the measurement issues described above certainly play a role in the observed differences among these approaches, but artifacts linked to the use of high ¹⁵N enrichments such as those observed here may contribute to these differences.

In conclusion, despite the use of different (i.e., natural and artificial) electron donors, NR ¹⁵N discrimination at natural abundance levels of ¹⁵N varied between 22 to 32‰ among four assimilatory eukaryotic NRs as well as a prokaryotic dissimilatory NR. These discriminations are similar to those observed or reported in other recent studies (Tcherkez and Farquhar, 2006; Karsh et al., 2012). Difficulties encountered using heavily ¹⁵N-enriched samples demonstrate the undesirability of protocols requiring such labeling for the measurement of enzyme isotope discrimination.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.