Advanced spectroscopic analysis and ¹⁵N-isotopic labelling study of nitrate and nitrite reduction to ammonia and nitrous oxide by E. coli

Abstract

Nitrate and nitrite reduction to ammonia and nitrous oxide by anaerobic E. coli batch cultures is investigated by advanced spectroscopic analytical techniques with ¹⁵N-isotopic labelling. Non-invasive, in situ analysis of the headspace is achieved using White cell FTIR and cavity-enhanced Raman (CERS) spectroscopies alongside liquid-phase Raman spectroscopy. For gas-phase analysis, White cell FTIR measures CO₂, ethanol and N₂O while CERS allows H₂, N₂ and O₂ monitoring. The 6 m pathlength White cell affords trace gas detection of N₂O with a noise equivalent detection limit of 60 nbar or 60 ppbv in 1 atm. Quantitative analysis is discussed for all four ¹⁴N/¹⁵N-isotopomers of N₂O. Monobasic and dibasic phosphates, acetate, formate, glucose and NO₃⁻ concentrations are obtained by liquid-phase Raman spectroscopy, with a noise equivalent detection limit of 0.6 mM for NO₃⁻ at 300 s integration time. Concentrations of the phosphate anions are used to calculate the pH in situ using a modified Henderson–Hasselbalch equation. NO₂⁻ concentrations are determined by sampling for colorimetric analysis and NH₄⁺ by basifying samples to release ¹⁴N/¹⁵N-isotopomers of NH₃ for measurement in a second FTIR White cell. The reductions of ¹⁵NO₃⁻, ¹⁵NO₂⁻, and mixed ¹⁵NO₃⁻ and ¹⁴NO₂⁻ by anaerobic E. coli batch cultures are discussed. In a major pathway, NO₃⁻ is reduced to NH₄⁺via NO₂⁻, with the bulk of NO₂⁻ reduction occurring after NO₃⁻ depletion. Using isotopically labelled ¹⁵NO₃⁻, ¹⁵NH₄⁺ production is distinguished from background ¹⁴NH₄⁺ in the growth medium. In a minor pathway, NO₂⁻ is reduced to N₂O via the toxic radical NO. With excellent detection sensitivities, N₂O serves as a monitor for trace NO₂⁻ reduction, even when cells are predominantly reducing NO₃⁻. The analysis of N₂O isotopomers reveals that for cultures supplemented with mixed ¹⁵NO₃⁻ and ¹⁴NO₂⁻ enzymatic activity to reduce ¹⁴NO₂⁻ occurs immediately, even before ¹⁵NO₃⁻ reduction begins. Optical density and pH measurements are discussed in the context of acetate, formate and CO₂ production. H₂ production is repressed by NO₃⁻; but in experiments with NO₂⁻ supplementation only, CERS detects H₂ produced by formate disproportionation after NO₂⁻ depletion.

---

1. Introduction

In the absence of oxygen (O₂), Escherichia coli (E. coli) can utilise alternative terminal electron acceptors for anaerobic growth, such as nitrate (NO₃⁻) and nitrite (NO₂⁻). The sequence of reductions from NO₃⁻ to NO₂⁻ to ammonia (NH₃, NH₄⁺ at physiological pH) is generally referred to as Dissimilatory Nitrate Reduction to Ammonia (DNRA).¹ The coupling of these reductions to the oxidation of organic substrates, such as formate, enables the generation of a proton gradient across the cytoplasmic membrane. DNRA is considerably more efficient for obtaining energy than the mixed acid fermentation pathways utilised when electron acceptors are unavailable. The expression of the respiratory NO₃⁻ and NO₂⁻ reductases is tightly controlled by FNR, an O₂ sensitive transcription factor, and NarXL/NarQP, both of which are two-component NO₃⁻/NO₂⁻ sensitive regulatory systems.^2,3

Although DNRA is the major NO₃⁻ reduction pathway in E. coli, the bacterium also generates minor amounts of the toxic radical nitric oxide (NO) from NO₂⁻ reduction. The low level of NO production by E. coli may be due to disproportionation of NO₂⁻ under acidic conditions or non-specific reduction by metalloproteins. The NADH-dependent cytoplasmic NO₂⁻ reductase (NirB),⁴ the membrane-bound periplasmic NO₂⁻ reductase (NrfA)⁵ and the major anaerobic NO₃⁻ reductase (NRA)^6,7 have all been proposed to be significant sources of NO formation as a by-product of their roles in the DNRA pathway. Aerobically, flavohemoglobin (Hmp) detoxifies NO by oxidation back to NO₃⁻; while anaerobically, NO is reduced further to nitrous oxide (N₂O) reportedly by Hmp,⁸ flavorubredoxin (NorV)⁹ and hybrid cluster protein (Hcp).¹⁰ N₂O is comparatively less toxic than NO and can rapidly diffuse out of the cell. E. coli is not a true denitrifier but N₂O production by NO₃⁻ respiring E. coli cultures does share similarities with the denitrification pathway of NO₃⁻ to nitrogen (N₂) via NO₂⁻, NO and N₂O. A summary of DNRA and NO generation and detoxification is shown in Fig. 1.

Fig. 1 DNRA and NO generation and detoxification by E. coli. Enzymes are displayed in red: Hcp, hybrid cluster protein; Hmp, flavohemoglobin; Nap, periplasmic nitrate reductase, NirB, NADH-dependent nitrite reductase; NorV, flavorubredoxin; NRA, nitrate reductase A; NrfA, periplasmic nitrite reductase; NRZ, nitrate reductase Z.

As a model organism, DNRA has been studied extensively in E. coli; however, comparatively less is known about the minor pathway leading to N₂O and how its generation differs between NO₃⁻ and NO₂⁻ respiring cultures. To gain a better mechanistic understanding, monitoring the key compounds and parameters of these processes is essential. Accurate and reliable analytical techniques are crucial for understanding cell biochemistry and pathway elucidation. This represents a challenge for analytical chemistry, requiring a combination of advanced analytical techniques.

Mass spectrometry and chromatographic techniques are widely applicable to the detection and quantification of a broad range of metabolites.¹¹ The tandem gas chromatography-mass spectrometry technique is considered the gold standard for the general analysis of volatile organic chemicals.¹² Despite this, these techniques are not readily applicable to rapid, online analysis either due to the need for sampling or for downstream chemical/physical processing before analysis can occur. Electrochemical sensors are widely used for monitoring pH, conductivity, dissolved O₂ and various other chemical species,¹³ including NO.¹⁴ Often such sensors are susceptible to cross-interferences from other species, changes in solution activity and long-term drift. For microbiological studies, the need for physical contact between the electrode and cell culture increases the risk of contamination, particularly in continuous cultures, and requires that the electrode is stable towards sterilisation.¹¹

Spectroscopic techniques can be readily applied for monitoring bioprocesses in situ and online, with no sampling. Vibrational spectroscopic techniques, such as Fourier Transform Infrared (FTIR) and Raman spectroscopies, show high specificity for different molecules due to characteristic spectral bands, making them potentially very valuable for metabolic studies. Additionally, vibrational spectroscopies can distinguish different isotopologues and isotopomers, allowing online monitoring of isotope labelling experiments.^15,16 Good sensitivities are observed in the condensed phase, but measuring headspace gases often suffers from low sensitivity, and special enhancement techniques are required such as Cavity Enhanced Raman Spectroscopy (CERS)^15–22 or long-path absorption White cells in FTIR spectroscopy.²³ Partial pressures in the headspace can be converted into concentrations in the solution via Henry's law. Quantum Cascade Laser (QCL) absorption spectroscopy has been applied to detect N₂O and other trace gases;^24–26 while sensitive, the limited tuning range of QCLs over a single IR absorption band limits the dynamic range due to band saturation effects. While FTIR spectroscopy has found some application in bioprocess monitoring, the broad absorption profile of water limits its application for monitoring metabolites at low concentrations in solution. In the gas-phase, the lack of an extended hydrogen network confines the absorption of water to certain spectral regions; molecules with absorption bands outside these regions can be readily detected, even in the presence of high levels of water vapour. Since Raman spectroscopy is comparatively insensitive to water, it is more readily applied to direct monitoring of the liquid-phase. However, fluorescence in complex media such as Lysogeny Broth (LB) can complicate the detection of the comparatively weak Raman light. Fluorescence can be avoided by moving to longer excitation wavelengths or by using media free of fluorescent components, such as M9 minimal media.²³ Vibrational spectroscopic tools have been previously applied to monitoring NO₃⁻ metabolism in bacteria; CERS has been used to follow N₂O and N₂ production in denitrifying organisms, with the use of ¹⁵NO₃⁻ to produce ¹⁵N₂ distinguishable from background ¹⁴N₂.^19,22 A robust CERS instrument has also been designed for field application to study the gas composition of soil samples.²¹

We report a combined approach for characterizing DNRA and N₂O production in anaerobic E. coli batch cultures using mostly non-invasive spectroscopic techniques. Sampling of the bacterial culture was only done for NO₂⁻ colorimetry and FTIR detection of ¹⁴NH₃ and ¹⁵NH₃ isotopomers. Headspace gas analysis was provided by the complementary techniques of FTIR and CERS, with CERS being a technique recently introduced by us in this Journal.¹⁷ FTIR allowed detection of CO₂, ethanol and N₂O while CERS enabled monitoring of the homonuclear diatomic molecules N₂, O₂ and H₂. Recently we introduced the capability of liquid culture analysis by Raman spectroscopy to monitor the microbial fermentation products of acetate and formate and the resulting in situ pH from phosphate signatures using a modified Henderson–Hasselbalch equation.²³ Here, we report on improvements that also allowed NO₃⁻ and glucose analysis during DNRA. With the use of ¹⁵N-labelling, we report on mechanistic insights into NO₃⁻ and NO₂⁻ reduction to NH₄⁺ and N₂O through interpreting the different ¹⁴N/¹⁵N-isotopomers produced. The aims of this report are to introduce and characterise a unique combination of advanced spectroscopic techniques with great potential for bioanalytical applications, and to introduce an interesting biochemical application, a ¹⁵N-isotope labelling study on N₂O production during DNRA by E. coli, with a focus on the differences observed between NO₃⁻ and NO₂⁻ reduction.

2. Experimental

Fig. 2 shows a scheme of our experimental setup. Since the previous iteration,²³ it was modified to include CERS for H₂, N₂ and O₂ detection with larger headspace and culture volumes to compensate for more frequent sampling. 250 mL of bacterial batch culture is contained in a round bottom flask with two side-arm ports and submerged in a 37 °C thermostated water bath. From the left side-arm, the bacterial suspension is circulated using a peristaltic pump (PP_(l), 4.5 L h⁻¹) for in situ OD₆₀₀ (optical density at 600 nm in a 1 cm cuvette) and Raman spectroscopy measurements. From the central-neck, the headspace (1425 mL volume) is cycled by a second peristaltic pump (PP_(g), 4.5 L h⁻¹) for gas-phase FTIR and CERS analysis. The right side-arm has a rubber septum enabling sampling of the liquid culture for further analysis. The CERS cavity is equipped with a capacitance pressure gauge (PG), N₂ inlet and vacuum line for purging O₂ to give anaerobic growth conditions (1 atm N₂) before starting experiments.

Fig. 2 Experimental setup for analysing the headspace by CERS and White cell FTIR spectroscopies and the liquid culture by Raman spectroscopy and in situ OD₆₀₀ measurements. DM, dichroic mirror; LP, laser pointer; MO, microscope objective; PD, photodiode; PG, pressure gauge; PP_(g), gas-phase peristaltic pump; PP_(l), liquid-phase peristaltic pump; SM, supermirror; WC, White cell.

Production of CO₂, ethanol and N₂O was quantified by gas-phase FTIR spectroscopy (Mattson Research Series, 0.4 cm⁻¹ spectral resolution, MCT detector) with a home-built multiple-pass absorption White cell.²³ The White cell pathlength was adjustable between 4–8 m, with 6 m used for this work. Spectra were recorded every 5 minutes. CO₂ partial pressures were obtained by integrating the ν₁ + 2ν₂ + ν₃ band (4920–5015 cm⁻¹, ν₀ = 4978 cm⁻¹) of the Fermi triad and comparing with a reference spectrum from the PNNL database.²⁷ N₂O partial pressures were obtained by integrating the 2ν₁ combination band from 2460–2580 cm⁻¹ and comparing the integral with simulated spectra from HITRAN 2012.²⁸ All four ¹⁴N/¹⁵N-isotopomers of N₂O could be distinguished, which enabled the ¹⁵N-isotope labelling studies. A multiplier equivalent to ethanol partial pressure was obtained by a least-squares fit of 1 ppmv ethanol and water reference spectra in the 2800–3100 cm⁻¹ region.²³ Using Henry's law, all partial pressures could be converted into concentrations in solution. Using the ideal gas law, we estimated that 10% of the CO₂ present in the sample was dissolved. Under our conditions, less than 1% of dissolved CO₂ was expected to be converted to carbonic acid and carbonates. 7% of N₂O and 99.7% of ethanol in the sample were also calculated to be dissolved.

The CERS setup has been described before with some modifications outlined below.^15–17,20 A 40 mW 636 nm single-mode cw-diode laser (HL63133DG) is coupled via a short-pass filter, a Faraday isolator and a mode matching lens into a linear optical cavity composed of two highly reflective mirrors (Newport SuperMirrors, R > 99.99%). If the laser wavelength matches the cavity length, an optical resonance builds up optical power inside the cavity by up to 3 orders of magnitude, enhancing the Raman signals. After the cavity, a dichroic mirror separates leftover excitation light from Raman signals which are coupled into a round-to-linear glass fibre bundle (7 × ∅ 105 μm) and transferred to the monochromator (Andor Shamrock SR163, 1200 l mm⁻¹ grating, DV420A-OE CCD). The 400–2500 cm⁻¹ spectral range at 6 cm⁻¹ resolution encompasses rotational S-branch lines of H₂, the ν₁/2ν₂ Fermi resonance of CO₂ and the vibrational fundamentals of O₂ and N₂. Part of the leftover excitation light is diverted back to the diode for optical feedback, locking the laser to the cavity. To normalize Raman signals, the N₂ peak is used as an internal standard since N₂ is not expected to change during bacterial activity. Raman intensities are converted to partial pressures using tabulated integrated peak areas.²⁰ CO₂ analysis by CERS was used to corroborate the FTIR analysis; however, CERS CO₂ data was not displayed in this study due to FTIR CO₂ detection being more sensitive. More details of the modified CERS setup are provided in the ESI.†

The bacterial suspension was circulated through a glass cuvette (1 cm path length) and the optical density OD₆₀₀ was recorded in situ by measuring the scattering of red laser pointer light with a photodiode. The transmitted intensity was calibrated with start and end-point OD₆₀₀ values externally measured using a UV-Vis spectrometer. The suspension was also circulated through a sealed borosilicate tube for recording liquid-phase Raman spectra using a home-built spectrometer.^29,30 A 532.2 nm, 20 mW laser (Lasos, GL3dT) and monochromator (Shamrock SR-750-A, 1200 l mm⁻¹ grating, DU420A-OE CCD) provided a spectral range from 830–1710 cm⁻¹ at about 0.8 cm⁻¹ resolution. Raman spectra were recorded every 5 minutes at 300 s integration time. No interfering fluorescence was noticeable in M9 minimal growth medium. The water bending vibration at 1630 cm⁻¹ was used to normalise decreasing Raman intensities as the turbidity of the bacterial suspension increased.²³ 0.1 M reference spectra of individual glucose, KNO₃, CH₃CO₂NH₄, HCO₂K, K₂HPO₄ and KH₂PO₄ solutions were recorded. As shown in Fig. 3, the 830–1200 cm⁻¹ region contains characteristic Raman peaks for HPO₄²⁻ (989 cm⁻¹), H₂PO₄⁻ (876 and 1076 cm⁻¹), NO₃⁻ (1049 cm⁻¹) and glucose (960–1180 cm⁻¹).³¹ Using a least-squares fitting routine, Raman spectra of the bacterial suspension in this region were fitted to the reference spectra, as well as a linear baseline. The returned multipliers of the reference spectra were then converted into concentrations via calibration plots. Noise analysis of background sample measurements (pure water) provided noise equivalent (1σ) detection limits of 0.6 mM NO₃⁻ and 1.9 mM glucose at 300 s integration time. With additional averaging to an integration time of 0.5 h (as was done with all time-dependent data displayed in this study), the limits improve to 0.25 mM for nitrate and 0.8 mM for glucose. The concentrations of the phosphate anions were used to calculate the pH in situ using a modified Henderson–Hasselbalch equation.^23,32 A least-squares fit determined acetate and formate concentrations in the 1310–1450 cm⁻¹ region to the sum of acetate (1414 cm⁻¹) and formate (1349 cm⁻¹) models and a linear baseline, as shown in the ESI.† At 300 s integration time, the noise equivalent (1σ) detection limits of acetate and formate were 2.6 mM and 1.5 mM, respectively. These limits improve to 1.1 mM and 0.6 mM with additional averaging to 0.5 h integration time. Although NO₂⁻ has a peak at 1326 cm⁻¹, the feature was too weak to be used in this study (1σ = 5.0 mM). Furthermore, NH₃/NH₄⁺ had no usable features within our spectral range.

Fig. 3 In black, an experimental Raman spectrum of M9 medium supplemented with 10 mM KNO₃ and 30 mM glucose. In red, the sum of the fitted NO₃⁻, glucose, HPO₄²⁻ (47 mM) and H₂PO₄⁻ (22 mM) models shown below the overlaid spectra.

E. coli (strain K-12 MG1655) was transferred from glycerol stock (maintained at −80 °C) and streaked on LB-agar plates. Plates were left to grow overnight at 37 °C. Before a measurement, 50 mL of sterile LB medium was inoculated with a single colony and incubated anaerobically in a sealed 50 mL centrifuge tube for 16 h (37 °C, 200 rpm) to a typical OD₆₀₀ of 1.2. From the starter culture, 20 mL was centrifuged, and the pellet resuspended into 20 mL of fresh M9 minimal medium. Our M9 medium formulation is given in the ESI;† but notably, it contains 30 mM glucose and 18 mM NH₄Cl. The M9 medium was supplemented with 10 mM K¹⁵NO₃ (10 mM, 98 atom % ¹⁵N, Sigma-Aldrich) and/or 5 mM KNO₂ (either ¹⁴N or ¹⁵N). A further 230 mL of M9 medium was prepared in the round bottom flask with two side-arms. The flask was pre-warmed and maintained at 37 °C using a thermostated water bath under rapid stirring to enable efficient gas transfer. The 20 mL M9 medium containing E. coli was added to the 230 mL M9 medium in the flask, giving a typical starting OD₆₀₀ of 0.1. The flask was then sealed and purged of O₂ by alternating between evacuating the headspace and refilling with N₂ at least five times. Experiments began once CERS measurements confirmed no O₂ remained.

During experiments, 1 mL of the bacterial culture was sampled every 40 min and centrifuged. The supernatant was analysed using a colorimetric method to determine NO₂⁻ concentration based on the Griess test.³³ Our M9 media began with a typical pH of 6.9 and ended between 5.0–5.5 due to organic acid excretion. With a pK_a of 9.25, NH₃ exists almost entirely as NH₄⁺ at acidic pH. For ¹⁴N/¹⁵N-analysis of NH₄⁺ samples, 2 mL 1 M NaOH was added to 0.6 mL of sample to release NH₃. The gas was analysed by a second FTIR setup (Bruker Alpha FTIR, 0.8 cm⁻¹ spectral resolution) with a home-built White cell (2.0 m pathlength). Spectra were recorded every 5 minutes with around 30 minutes needed before NH₃ concentration peaked in the headspace. The basified solution was rapidly stirred and the 2 L headspace in the closed system was cycled between the sample flask and White cell using a peristaltic pump. The ν₂ band is the strongest in the FTIR spectrum of NH₃ and can be used for ¹⁴N/¹⁵N-analysis.³⁴ At the end of bacterial activity, the suspension was centrifuged, washed and dried to record the dry biomass (typically around 200 mg when corrected for sampling). For comparison with the in situ spectroscopic pH measurements, the pH of start and end-point samples was recorded externally using a Mettler Toledo SevenMulti pH meter. See the ESI† for further experimental spectra and calibration plots for all aforementioned analytical techniques.

3. Results and discussion

3.1 FTIR spectroscopy of N₂O and its ¹⁴N/¹⁵N-isotopomers

N₂O has four ¹⁴N/¹⁵N-isotopomers, i.e., ¹⁴N₂O, the structural isomers ¹⁴N¹⁵NO and ¹⁵N¹⁴NO, and ¹⁵N₂O. N₂O is amenable to ¹⁵N-isotope labelling studies due to the low natural abundance of the ¹⁵N-isotope (0.37%). In the 2000–3000 cm⁻¹ spectral range, characteristic partially rotationally resolved bands of the N₂O isotopomers are available for FTIR analysis. Apart from ca. 2250–2400 cm⁻¹ which is saturated by CO₂, this region is free from significant spectral interferences. The HITRAN molecular database contains line lists for the three most abundant ¹⁴N/¹⁵N-isotopomers, excluding ¹⁵N₂O.²⁸ A survey of HITRAN and our experimental spectra has shown that the following vibrational bands are available for quantitative analysis, including band position of ¹⁴N₂O, integrated absorption cross-sections G and peak absorbances A_peak (defined as ln(I₀/I)) under our experimental conditions for 1 μbar (1 ppmv) at 6 m path length: the ν₃ fundamental near 2224 cm⁻¹ with G = 5.55 × 10⁻¹⁷ cm and A_peak ≈ 0.023 for rotational lines in the P- and R-branches, the 2ν₁ overtone near 2563 cm⁻¹ with G = 1.33 × 10⁻¹⁸ cm and A_peak ≈ 6 × 10⁻⁴ for rotational features in the P- and R-branches, and the ν₂ + ν₃ combination near 2798 cm⁻¹ with G = 9.0 × 10⁻²⁰ cm and A_peak ≈ 2.6 × 10⁻⁴ of its Q-branch. Characteristic spectral shifts allow distinction of the isotopomers, while their G and A_peak values remain essentially the same. For accurate quantitative results, A_peak should not exceed unity. The dynamic range of the ν₃ fundamental thus extends from trace levels up to ca. 45 μbar N₂O, the 2ν₁ overtone up to 1.7 mbar, and the ν₂ + ν₃ combination up to 3.8 mbar. This range can be extended by reducing the absorption pathlength of the White cell.

Fig. 4 shows the ν₃ fundamental with distinct P- and R-branch features, with ¹⁴N₂O having its origin near 2224 cm⁻¹. In a spectrum containing only ¹⁴N₂O, a least-squares fit to the reference spectrum in the region denoted ‘D’ in Fig. 4 returns a multiplier which corresponds to N₂O partial pressure. A simple integration over the ν₃ band would not be suitable because part of the R-branch is buried in ¹³CO₂ absorptions at higher wavenumbers. The region ‘D’ was selected because it has some of the strongest absorption features, it is very characteristic with partially resolved lines, and it is least affected by CO₂. With this fitting routine, noise analysis of blank samples provides a noise equivalent detection limit of 60 nbar (60 ppbv at 1 bar total pressure) at 6 m pathlength and 128 accumulations which take 2 min to acquire. Detection limits can be improved by more averaging or increasing the path length. Note that this is sufficient to detect the 330 ppbv ambient levels of N₂O for environmental analytical applications. The heavier isotopomers shift to lower wavenumbers, 2201 cm⁻¹ for ¹⁵N¹⁴NO, 2178 cm⁻¹ for ¹⁴N¹⁵NO, and 2155 cm⁻¹ for ¹⁵N₂O. Since the bands are overlapping, only a simultaneous fit to all four model spectra can yield individual isotopomer partial pressures. A fit in the entire 2100–2220 cm⁻¹ region, however, has serious problems with cross-correlations. After a careful analysis, a simultaneous fit only including the regions ‘A’ to ‘D’ in Fig. 4 returned multipliers which are not noticeably affected by cross-correlations. Each region was chosen so that an individual isotopomer has a maximum weight with the other isotopomers having as little weight as possible. This procedure yields reliable isotopic partial pressures up to a dynamic range of about 45 μbar per isotopomer.

Fig. 4 ν ₃ fundamental of N₂O isotopomers with partially resolved rotational P- and R-branches. Absorbances scaled to correspond to 1 μbar (1 ppmv) at 6 m path length. (a) Experimental FTIR spectra of ¹⁵N₂O (blue) and ¹⁴N₂O (brown). (b) Isotopomers (structural isomers) ¹⁴N¹⁵NO (red) and ¹⁵N¹⁴NO (green) calculated from the HITRAN database. A to D denotes spectral ranges used in the fit.

Fig. 5 shows the weaker absorption bands that are more suitable for N₂O analysis above 45 μbar. In isotopically pure samples, the 2ν₁ overtone near 2563 cm⁻¹ can be integrated from 2505–2613 cm⁻¹ to obtain ¹⁴N₂O partial pressure after comparison with a reference spectrum (Fig. 5a). For ¹⁵N₂O the shifted band near 2523 cm⁻¹ can be integrated from 2460–2580 cm⁻¹ (Fig. 5b). In samples with mixtures of isotopomers (Fig. 5c), the 2ν₁ bands overlap and require a more sophisticated simultaneous fit similar to the one described above for the ν₃ fundamental. Fortunately, this is not required as the ν₂ + ν₃ combination band (2798 cm⁻¹ for ¹⁴N₂O) has a sharp, characteristic Q-branch which remains well resolved and separated in isotopic mixtures. After comparison with reference spectra, simple integrations over the separate Q-branch peaks yield isotopic partial pressures in a mixture up to a dynamic range of about 3.8 mbar.

Fig. 5 Experimental FTIR spectra of N₂O overtone and combination bands for (a) 1 mbar ¹⁴N₂O, (b) 1 mbar ¹⁵N₂O and (c) a mixture of 2.8 mbar ¹⁵N₂O (49%), 1.1 mbar ¹⁴N¹⁵NO (20%), 1.2 mbar ¹⁵N¹⁴NO (22%) and 0.5 mbar ¹⁴N₂O (9%). The isotopomer mixture was recorded at 30 h during the anaerobic respiration of E. coli supplemented with 10 mM ¹⁵NO₃⁻ and 5 mM ¹⁴N-nitrite (see section 3.4).

3.2 Spectroscopic analysis of nitrate reduction by E. coli

Fig. 6 is a typical example of pH, OD₆₀₀ and number of moles (n) of electron acceptors and other metabolites measured during the reduction of 10 mM ¹⁵NO₃⁻ by anaerobic E. coli. Concentrations (mM) in solution were converted to n (mmol) by multiplying by the culture volume (0.25 L), as were partial pressures using the ideal gas law (V = 1.425 × 10⁻³ m³, T = 310 K) and correcting for the dissolved percentage calculated via Henry's law. All biological experiments were repeated in triplicate, and all repeats showed essentially the same behaviour. The time-dependent data displayed in this study is for a single representative experiment selected from the repeats. Phase A (0–6.5 h) lasted until all NO₃⁻ was reduced to NO₂⁻. Phase B (6.5–10 h) lasted until all NO₂⁻ was reduced to NH₄⁺ and N₂O. Phase C (>10 h) had no electron acceptors remaining so the bacteria utilised fermentative pathways solely. The ¹⁵N-label transferred to ¹⁵NH₄⁺ and ¹⁵N₂O with no trace of other N₂O isotopomers formed. This was consistent with other studies that found the N-atoms in N₂O both originate from NO₃⁻/NO₂⁻ and not other sources such as N₂ or NH₄⁺.^19,22,35 The externally measured start and end-point pH measurements showed good agreement with the time-dependent spectroscopically determined pH.

Fig. 6 Anaerobic E. coli growth in M9 medium supplemented with 10 mM ¹⁵NO₃⁻. A to C denotes three distinct phases: NO₃⁻ reduction (A), NO₂⁻ reduction (B) and NO₂⁻ depletion (C). (a) Time-dependent number of moles (n) of ¹⁵NO₃⁻, ¹⁵NO₂⁻, ¹⁵NH₄⁺ and ¹⁵N₂O (×10). (b) n of glucose, acetate and formate. (c) n of CO₂, ethanol (×5) and ¹⁴NH₄⁺. (d) Spectroscopically determined pH (open circles), externally measured pH (solid squares) and OD₆₀₀.

After a brief lag phase, exponential growth began at 3 h with a rapid increase in the OD₆₀₀. NO₃⁻ reduction to NO₂⁻ mirrored the growth curve with most of the NO₂⁻ produced excreted to prevent cytoplasmic toxification.³⁶E. coli expresses three NO₃⁻ reductases: the respiratory NO₃⁻ reductases A and Z (NRA and NRZ) and the periplasmic NO₃⁻ reductase (Nap).^37–39 NRA is the most active reductase at high NO₃⁻ levels (>2 mM).⁴⁰ Nap is induced by low NO₃⁻ levels, while NRZ is expressed at low levels constitutively and may function under stress-associated conditions.^40–42 NO₂⁻ peaked at 2.2 mmol, less than the initial 2.5 mmol NO₃⁻, as some NO₂⁻ was reduced alongside NO₃⁻ during A. 0.3 mmol ¹⁵NH₄⁺ and 1.6 μmol ¹⁵N₂O was produced, accounting for the total N-balance. Only 1% of the 0.3 mmol NO₂⁻ reduced in A was converted to N₂O instead of NH₄⁺. E. coli expresses two NO₂⁻ reductases: the NADH-dependent cytoplasmic NO₂⁻ reductase (NirB) and the membrane-bound periplasmic NO₂⁻ reductase (NrfA). NirB likely produced NH₄⁺ during A as it is active when NO₃⁻ is readily available, unlike NrfA.⁴³ Evidence also suggests NirB can generate NO.⁴ Anaerobically, NO is detoxified by reduction to N₂O, which is comparatively non-toxic and rapidly diffuses out of the cell. Flavorubredoxin (NorV),⁹ hybrid cluster protein (Hcp),¹⁰ NirB⁴⁴ and NrfA⁴⁵ have all been proposed to have NO detoxifying activity. Flavohemoglobin (Hmp) is primarily an NO oxidase but also acts as an NO reductase anaerobically.⁸

As E. coli does not possess any known N₂O reductases, further reduction to N₂ was not expected. However, there is some evidence that N₂ can be produced from high amounts of N₂O by a yet unknown mechanism.⁴⁶ To investigate whether under our conditions N₂ was produced, we repeated the experiment, but under an argon atmosphere instead of N₂. No trace of N₂ production was observed in the CERS spectra within our detection limit of ca. 0.2 mbar or 12 μmol N₂.

Formate oxidation to CO₂ by the NO₃⁻-inducible formate dehydrogenase (FdhN) is a physiological source of electrons for NO₃⁻ reduction.³⁸ Other sources include NADH, lactate and glycerol.³⁷ 1.7 of the 2.5 mmol NO₃⁻ reduced was coupled to FdhN activity as CO₂ increased by such in A. The remaining 0.8 mmol NO₃⁻ was likely coupled to NADH oxidation.⁴⁷ As no formate was excreted in A, all formate produced by pyruvate formate lyase (PFL) must have been oxidised to CO₂. For each formate produced by PFL, one acetyl-CoA is formed which can be either directed into the anaerobic TCA cycle or converted to acetate (to produce ATP) or ethanol (to remove reducing equivalents). 1.7 mmol acetate and 0.05 mmol ethanol were excreted during A corresponding to 1.75 mmol formate, in good agreement with the 1.7 mmol CO₂ produced. Acetate must be excreted to prevent cytoplasmic acidification and caused the extracellular pH to decrease from 7.1 to 6.7. The minor amount of ethanol produced was due to reducing equivalents being coupled directly into reduction of NO₃⁻. Previous studies have found a similar repression of substrate-level NADH consuming pathways when electron acceptors are available.⁴⁸ Glucose decreased by 1.1 mmol owing to the production of CO₂, acetate, ethanol and biomass synthesis.

During phase B, NO₂⁻ was reimported into E. coli and reduced. From 6.5 to 10 h, 2.2 mmol ¹⁵NO₂⁻ was reduced almost linearly to 2.0 mmol ¹⁵NH₄⁺ and 0.1 mmol ¹⁵N₂O. 91% NO₂⁻ was reduced to NH₄⁺ and 9% to N₂O, a higher partitioning to N₂O than observed in A (1%). A higher partitioning to N₂O after NO₃⁻ was depleted is consistent with several studies of E. coli and Salmonella enterica that have implicated NRA as the enzyme that produces the majority of NO when NO₂⁻ is abundant and NO₃⁻ absent.^6,7,49,50 NrfA, which is induced by NO₂⁻ but repressed by NO₃⁻, may have also contributed towards the higher partitioning to N₂O in B as it has been proposed as a source of NO.^5,51 The radical NO has a distinct line-resolved absorption band centred at 1904 cm⁻¹ (for ¹⁴NO) and a favourable partitioning into the headspace.⁵² However, no intermediate ¹⁵NO gas was observed to accumulate, owing to its rapid detoxification to ¹⁵N₂O by E. coli. During B, a further 1.9 mmol CO₂ was produced and the pH dropped from 6.7 to 5.7 due to the excretion of 5.7 mmol formate and a further 7.5 mmol acetate. Due to the 3 : 1 stoichiometry of formate oxidation to CO₂ : NO₂⁻ reduction to NH₄⁺, 0.6 mmol NO₂⁻ was coupled to formate by NrfA.⁵³ The 5.7 mmol formate excreted during B would be plentiful to couple to the remaining 1.4 mmol NO₂⁻. However, NrfA is most active at low NO₂⁻ levels while NirB is most active at high NO₂⁻ levels for detoxification of excess NO₂⁻.^36,43 Thus, 1.4 mmol NO₂⁻ was likely reduced by NirB.

Phase C started with exponential growth ending as the OD₆₀₀ peaked at 1.7, due to the depletion of glucose and NO₂⁻. With no electron acceptors available, the bacteria funnelled reducing equivalents into ethanol as a further 0.7 mmol was made over the next 5 h. The remaining 5.7 mmol formate was slowly oxidised to CO₂ at a rate of 0.03 mmol h⁻¹. Under anaerobic conditions, the presence of formate induces formate hydrogenlyase (FHL) activity that disproportionates formate to CO₂ and H₂.⁵⁴ O₂ and NO₃⁻ repress FHL expression and instead induce the aerobic and the formate-NO₃⁻ respiratory chains. High formate concentrations can partially reverse the repression by NO₃⁻, but not by O₂.^55,56 However, CERS measurements detected no H₂ production during our 10 mM ¹⁵NO₃⁻ reduction experiments. During C, there was a slight decline in N₂O observed due to the gas adsorbing to tubing and glass surfaces.

Experiments were terminated after 2 days with 5 mmol formate still remaining. The dry biomass was typically around 200 mg. As E. coli can be approximated to be 48% carbon and 14% nitrogen by mass,⁵⁷ca. 8 mmol C and 2 mmol N in the biomass originated from the 7.5 mmol glucose (45 mmol C) and NH₄⁺, respectively. 44 out of the 45 mmol C from glucose can be accounted for in the biomass, 5 mmol CO₂, 5 mmol formate, 12 mmol acetate (24 mmol C) and 1 mmol ethanol (2 mmol C). During exponential growth, ¹⁴NH₄⁺ decreased from 4.5 to 3.0 mmol accounting for 1.5 out of the 2 mmol N in the biomass. The remaining 0.5 mmol N likely was taken from the excreted ¹⁵NH₄⁺. The 2.5 mmol ¹⁵N-label can be accounted for in the 2.0 mmol ¹⁵NH₄⁺, 0.1 mmol ¹⁵N₂O (0.2 mmol ¹⁵N) and ∼0.5 mmol ¹⁵NH₄⁺ used for biosynthesis.

3.3 Spectroscopic analysis of nitrite reduction by E. coli

To study the response to NO₂⁻ alone, anaerobic E. coli was supplemented with 5 mM ¹⁵NO₂⁻, as shown in Fig. 7. Phase A′ (0–9 h) corresponded to the reduction of NO₂⁻ to NH₄⁺ with concurrent N₂O production via NO. Phase B′ (9–15 h) was when the bacteria utilised fermentative pathways only, due to NO₂⁻ depletion. During the first 9 h of phase A′, 1.25 mmol ¹⁵NO₂⁻ was reduced almost exponentially to 1.15 mmol ¹⁵NH₄⁺ (90%) and 0.06 mmol ¹⁵N₂O (10%), mirroring the bacterial growth curve which increased to an OD₆₀₀ of 0.7. The 10% partitioning to N₂O here was consistent with the 9% observed during the NO₂⁻ reduction phase B in section 3.2. During A′, 1.9 mmol CO₂, 3.7 mmol acetate, 2.0 mmol formate and 0.3 mmol ethanol were produced as glucose decreased from 7.5 to 5.2 mmol. Excretion of acetate and formate caused the pH to decrease from 6.9 to 6.3. The sum of acetate and ethanol (4.0 mmol) showed good agreement with the sum of CO₂ and formate (3.9 mmol). Due to the 3 : 1 stoichiometry of formate oxidation : NO₂⁻ reduction and 1.9 mmol CO₂ being produced, 0.63 out of the initial 1.15 mmol NO₂⁻ reduced to NH₄⁺ was coupled to formate oxidation to CO₂. The remaining 0.52 mmol NO₂⁻ was likely reduced via coupling to NADH oxidation by NirB.

Fig. 7 Anaerobic E. coli growth in M9 medium supplemented with 5 mM ¹⁵NO₂⁻. A′ and B′ denote two distinct phases: NO₂⁻ reduction (A′) and NO₂⁻ depletion (B′). (a) Time-dependent number of moles (n) of ¹⁵NO₂⁻, ¹⁵NH₄⁺ and ¹⁵N₂O (×10). (b) n of glucose, acetate and formate. (c) n of CO₂, H₂, ethanol and ¹⁴NH₄⁺. (d) Spectroscopically determined pH (open circles), externally measured pH (solid squares) and OD₆₀₀.

Phase B′ began at 9 h when ¹⁵NO₂⁻ was depleted. E. coli could only utilise fermentative pathways in the absence of NO₂⁻. The most notable difference between 10 mM ¹⁵NO₃⁻ reduction (discussed in section 3.2), and 5 mM ¹⁵NO₃⁻ reduction was H₂ production that occurred after NO₂⁻ depletion in Fig. 7. No H₂ production was observed during A′ as formate-dependent NO₂⁻ reduction likely made the intracellular formate unavailable for FHL induction. The presence of formate is required for FHL expression but it can be made unavailable by coupling to the reduction of electron acceptors. This inhibiting effect has been observed for NO₃⁻ and trimethylamine N-oxide respiring E. coli cultures and in both cases the effect could be partially relieved by adding exogenous formate.^56,58 When NO₂⁻ was depleted, 5.2 mmol glucose remained meaning further formate could be produced during B′ which may have triggered the induction of FHL. From 9–15 h, 6.0 mmol H₂ and a further 6.0 mmol CO₂ were produced from the disproportionation of formate. At 10 h, there was a peak of 2.1 mmol formate excreted. During B′, a further 5.3 mmol acetate and 3.5 mmol ethanol were produced. By 12 h, the pH dropped to 5.4 and then remained stable as 1.6 mmol acetate was produced and balanced by the reimport and disproportionation of 1.5 mmol formate. By 14 h, the OD₆₀₀ peaked at 1.5, just before the end of bacterial activity at 15 h due to the depletion of glucose and formate. 42.8 out of the 45 mmol C from glucose can be accounted for in the biomass (∼8 mmol C), 8.9 mmol CO₂, 9 mmol acetate (18 mmol C) and 3.8 mmol ethanol (7.6 mmol C). During exponential growth, ¹⁴NH₄⁺ decreased from 4.5 to 2.9 mmol as did ¹⁵NH₄⁺ from a peak value of 1.15 to 0.9 mmol accounting for 1.85 mmol out of the ∼2 mmol N in the biomass.

3.4 Simultaneous nitrate and nitrite reduction

In section 3.2, when E. coli was supplemented with NO₃⁻, there was a distinct hierarchy of metabolic pathways between phases A, B and C. NO₃⁻ reduction dominated in A, followed by NO₂⁻ reimport and reduction in B and finally fermentation in C. However, in A it was observed that some NO₂⁻ was simultaneously reduced alongside NO₃⁻ to NH₄⁺ and N₂O. To further investigate the overlap between the reductions of NO₃⁻ and NO₂⁻ in A, anaerobic E. coli was supplemented with 10 mM ¹⁵NO₃⁻ and 5 mM ¹⁴NO₂⁻ as shown in Fig. 8 and 9. Phase A (0–9 h) lasted until all ¹⁵NO₃⁻ was reduced to ¹⁵NO₂⁻. Phase B (9–30 h) corresponded to the reduction of NO₂⁻ to NH₄⁺ with concurrent N₂O production via NO. At 15.5 h, the OD₆₀₀ peaked and exponential growth of E. coli ended; thus, phase B1 (9–15.5 h) was NO₂⁻ reduction with glucose still present and phase B2 (15.5–30 h) was NO₂⁻ reduction during glucose depletion. Fig. 8 displays n of ¹⁵NO₃⁻, NO₂⁻, ¹⁴NH₄⁺, ¹⁵NH₄⁺ (×10) and N₂O isotopomers in A. The complete characterization of bacterial growth is given in Fig. 9.

Fig. 8 N₂O isotopomers produced in the NO₃⁻ reduction phase (A) of anaerobic E. coli growth in M9 medium supplemented with 10 mM ¹⁵NO₃⁻ and 5 mM ¹⁴NO₂⁻. (a) Time-dependent number of moles (n) of ¹⁵NO₃⁻, NO₂⁻, ¹⁴NH₄⁺ and ¹⁵NH₄⁺ (×10). (b) n of N₂O isotopomers produced. ¹⁴N¹⁵NO is omitted due to essentially having the same behaviour as ¹⁵N¹⁴NO.

Fig. 9 Anaerobic E. coli growth in M9 medium supplemented with 10 mM ¹⁵NO₃⁻ and 5 mM ¹⁴NO₂⁻. A to C denotes three distinct phases: NO₃⁻ reduction (A), NO₂⁻ reduction with glucose present (B1) and NO₂⁻ reduction with glucose depleted (B2). (a) Time-dependent number of moles (n) of ¹⁵NO₃⁻, NO₂⁻ (both ¹⁴N and ¹⁵N), ¹⁵NH₄⁺ and sum of all ¹⁴N/¹⁵N-isotopomers of N₂O (×10). (b) n of glucose, acetate and formate. (c) n of CO₂ and ¹⁴NH₄⁺. (d) Spectroscopically determined pH (open circles), externally measured pH (solid squares) and OD₆₀₀.

In phase A, 2.5 mmol ¹⁵NO₃⁻ was reduced and ca. 2.25 mmol ¹⁵NO₂⁻ was excreted. NO₂⁻ colorimetry cannot distinguish between ¹⁴NO₂⁻ and ¹⁵NO₂⁻, so NO₂⁻ was observed to increase from 1.25 to 3.5 mmol. During A, as in section 3.2, some NO₂⁻ was reduced alongside ¹⁵NO₃⁻ to 0.2 mmol ¹⁵NH₄⁺ and N₂O isotopomers. ¹⁴NO₂⁻ reduction to ¹⁴N₂O occurred immediately, with 2.2 μmol ¹⁴N₂O produced almost linearly by 9 h. This indicated that even before NO₃⁻ reduction began, some unknown enzymatic activity to reduce small quantities of NO₂⁻ to N₂O was immediately active. For the first 3 h, ¹⁵NO₃⁻ and NO₂⁻ measurements were virtually constant suggesting a lag in the expression of NRA. This lag was best indicated by the highly sensitive positional isomers ¹⁴N¹⁵NO and ¹⁵N¹⁴NO which were not detected until ¹⁵NO₂⁻ was made available by ¹⁵NO₃⁻ reduction starting from 3 h. ¹⁵N₂O production also began at 3 h, but much slower than the production of ¹⁴N₂O and the positional isomers, due to ¹⁴NO₂⁻ initially being more readily available than ¹⁵NO₂⁻. By the end of A, 1.5 μmol each of ¹⁴N¹⁵NO, ¹⁵N¹⁴NO and ¹⁵N₂O were produced alongside the 2.2 μmol ¹⁴N₂O, totalling 6.7 μmol. It is unknown if ¹⁴NO₂⁻ was also immediately reduced to ¹⁴NH₄⁺, due to the large background of 4.5 mmol ¹⁴NH₄⁺ in the growth medium. It can be assumed ca. 0.25 mmol NO₂⁻ was reduced during A based on the NO₂⁻ colorimetry measurements giving a partitioning of 5% NO₂⁻ reduced to N₂O, instead of NH₄⁺. This was a higher value than the 1% observed during A in section 3.2, indicating that the added ¹⁴NO₂⁻ led to more NO generation and detoxification to N₂O. During A, glucose decreased from 7.5 to 5.6 mmol due to the production of 2.1 mmol CO₂, 2.0 mmol acetate and biomass synthesis. The OD₆₀₀ began increasing indicating exponential bacterial growth while acetate excretion caused the pH to decrease from 7.0 to 6.5. No formate was excreted during A, as the n of CO₂ and acetate suggested all formate formed was oxidised to CO₂. No ethanol was detected during the entire 30 h experiment, likely due to the abundance of electron acceptors to couple reducing equivalents to.

Phase B1 began with ¹⁵NO₃⁻ depletion and ended at 15.5 h when glucose was depleted, coinciding with the OD₆₀₀ peaking at 1.2. The pH dropped further to 5.6 due to the excretion of 5.0 mmol formate and a further 6.0 mmol acetate. The sum of formate excreted and the further 1.3 mmol CO₂ produced was in good agreement with the amount of acetate excreted.

Phase B2 lasted until NO₂⁻ depletion at 30 h. From NO₂⁻ reduction, 1.6 mmol ¹⁵NH₄⁺ and 0.35 mmol N₂O were produced overall. The final composition of N₂O isotopomers was previously introduced in Fig. 5c. As the majority of N₂O production occurred in B when the NO₂⁻ composition was ca. 66% ¹⁵NO₂⁻ and 33% ¹⁴NO₂⁻, a near statistical mixture of N₂O isotopomers was formed of 0.17 mmol ¹⁵N₂O (49%), 0.08 mmol ¹⁵N¹⁴NO (22%), 0.07 mmol ¹⁴N¹⁵NO (20%) and 0.03 mmol ¹⁴N₂O (9%). For comparison, a perfect statistical mixture would have produced 44.4% ¹⁵N₂O, 22.2% ¹⁵N¹⁴NO, 22.2% ¹⁴N¹⁵NO and 11.1% ¹⁴N₂O. It is unknown whether the slight preference for ¹⁵N¹⁴NO over ¹⁴N¹⁵NO is significant or due to experimental uncertainty. The partitioning of the 3.5 mmol NO₂⁻ to N₂O in B was 20%, a much higher value than the 10% observed during A in section 3.2. This is consistent with previous studies that found that between 5–36% of NO₃⁻ is converted to N₂O by E. coli, depending on growth conditions.³⁵ During B2, CO₂ increased by a further 0.5 mmol while the pH remained constant at 5.6 due to no significant change in acetate and formate. 33 out of the 45 mmol C from glucose can be accounted for in the biomass (∼8 mmol C), 4 mmol CO₂, 8 mmol acetate (16 mmol C) and 5 mmol formate. The higher NO₂⁻ content may have had cytotoxic effects in E. coli resulting in other products that have not been accounted for in the C balance. During B2, the OD₆₀₀ dropped from 1.2 to 0.8 suggesting cell death or changes in cellular size and morphology, possibly due to the cytotoxicity of NO₂⁻ and NO. The 2.5 mmol ¹⁵N label was accounted for in the 1.6 mmol ¹⁵NH₄⁺, 0.17 mmol ¹⁵N₂O (0.34 mmol ¹⁵N), 0.08 mmol ¹⁵N¹⁴NO, 0.07 mmol ¹⁴N¹⁵NO and ca. 0.5 mmol ¹⁵NH₄⁺ assumed to have been used for biosynthesis. As ca. 2.0 mmol NH₄⁺ was needed for biosynthesis, it was assumed ca. 1.5 mmol was taken from ¹⁴NH₄⁺, which decreased overall from 4.5 to 4.0 mmol suggesting ca. 1.0 mmol ¹⁴NH₄⁺ produced from the reduction of the 1.25 mmol ¹⁴NO₂⁻. This was in good agreement with the 0.26 mmol ¹⁴NO₂⁻ reduced to N₂O isotopomers with 0.03 mmol ¹⁴N₂O (0.06 mmol ¹⁵N), 0.08 mmol ¹⁵N¹⁴NO and 0.07 mmol ¹⁴N¹⁵NO.

4. Conclusions

We have studied NO₃⁻ and NO₂⁻ reduction during DNRA by anaerobic E. coli batch cultures by a combination of advanced spectroscopic analytical techniques in conjunction with ¹⁵N-isotopic labelling. The online spectroscopic techniques described here are non-invasive, avoiding any contact with the bacterial suspension, and provide concentrations in real-time. We discussed in detail the spectroscopy, which spectral features are most useful for analysis, and data analysis and fitting routines for quantitative analysis. In situ analysis of the headspace is achieved using cavity-enhanced Raman (CERS) and long-path White cell FTIR spectroscopies alongside liquid-phase Raman spectroscopy. Gas phase CERS allows CO₂, H₂, N₂ and O₂ monitoring while White cell FTIR measures CO₂, ethanol and N₂O. The 6 m pathlength White cell affords trace gas detection of N₂O with a noise equivalent detection limit of 60 nbar or 60 ppbv in 1 atm (1σ noise equivalent, 128 scans corresponding to 120 s acquisition). This extremely high sensitivity could be utilised in situations where N₂O cannot be allowed to build up, e.g. in continuous culture studies. Quantitative analysis is discussed for all four ¹⁴N/¹⁵N-isotopomers, including the positional isomers ¹⁴N¹⁵NO and ¹⁵N¹⁴NO, a unique capability not available to other analytical techniques.

¹⁵N-isotopic labelling of NO₃⁻ identifies the sources of N-atoms in products of E. coli metabolism, in particular, it provides insight into the mechanism of N₂O production during mixed NO₃⁻ and NO₂⁻ reduction. This study is one of very few reporting quantitative analysis of N₂O production by E. coli under various conditions. The reductions of ¹⁵NO₃⁻, ¹⁵NO₂⁻, and mixed ¹⁵NO₃⁻ and ¹⁴NO₂⁻ to NH₄⁺ and N₂O have been discussed. In a major pathway, NO₃⁻ is reduced to NH₄⁺via NO₂⁻, with the bulk of NO₂⁻ reduction occurring after NO₃⁻ depletion. By isotopically labelling ¹⁵NO₃⁻, ¹⁵NH₄⁺ production is distinguished from background ¹⁴NH₄⁺ in the growth medium. In a minor pathway, NO₂⁻ is reduced to N₂O via the toxic radical NO. With excellent detection sensitivities, N₂O monitors trace NO₂⁻ reduction even when cells are predominantly reducing NO₃⁻; the analysis of N₂O isotopomers reveals that some enzymatic NO₂⁻ reduction activity occurs immediately for cultures supplemented with mixed ¹⁵NO₃⁻ and ¹⁴NO₂⁻. Optical density and pH measurements are discussed in context of acetate, formate and CO₂ production. H₂ production is repressed by NO₃⁻, but with NO₂⁻ only, CERS detects H₂ produced by formate hydrogenlyase after NO₂⁻ depletion.

In future work, we want to extend our spectroscopic approach to monitor different bacterial pathways, in particular, the relationship between fermentative and other respiratory pathways and to study nitrifying and denitrifying bacteria. These spectroscopic techniques are capable of detecting key species in the nitrogen cycle and with the ability to sensitively distinguish N₂O isotopomers they may be of great interest for helping better understand global N₂O budgets. Spectroscopic monitoring of bioprocesses has excellent potential to supplement or replace traditional techniques in analytical chemistry.

Author contributions

All authors have contributed equally to the conception of the project, the experimental work, the analysis and to the writing of the manuscript.

Conflicts of interest

The authors declare that they have no conflict of interest.

Acknowledgements

We are grateful to Dr Jim Reid and Profs Jeff Green and Robert Poole for help and discussions. We acknowledge the University of Sheffield and the EPSRC research council (DTP scholarship to GDM) for financial support of our research.

References

1. M. Tiso and A. N. Schechter, Nitrate reduction to nitrite, nitric oxide and ammonia by gut bacteria under physiological conditions, PLoS One, 2015, 10, e0127490,  DOI:10.1371/journal.pone.0119712.
2. V. Stewart, Nitrate- and nitrite-responsive sensors NarX and NarQ of proteobacteria, Biochem. Soc. Trans., 2003, 31, 1–10,  DOI:10.1042/bst0310001.
3. J. Green, M. D. Rolfe and L. J. Smith, Transcriptional regulation of bacterial virulence gene expression by molecular oxygen and nitric oxide, Virulence, 2014, 5, 794–809,  DOI:10.4161/viru.27794.
4. C. E. Vine and J. A. Cole, Nitrosative stress in Escherichia coli: reduction of nitric oxide, Biochem. Soc. Trans., 2011, 39, 213–215,  DOI:10.1042/BST0390213.
5. H. Corker and R. K. Poole, Nitric Oxide Formation by Escherichia coli: Dependence On Nitrite Reductase, The NO-Sensing Regulator Fnr, and Flavohemoglobin Hmp, J. Biol. Chem., 2003, 278, 31584–31592,  DOI:10.1074/jbc.M303282200.
6. M. S. Smith, Nitrous oxide production by Escherichia coli is correlated with nitrate reductase activity, Appl. Environ. Microbiol., 1983, 45, 1545–1547,  DOI:10.1128/AEM.45.5.1545-1547.1983.
7. R. Metheringham and J. A. Cole, A reassessment of the genetic determinants, the effect of growth conditions and the availability of an electron donor on the nitrosating activity of Escherichia coli K-12, Microbiol., 1997, 143, 2647–2656,  DOI:10.1099/00221287-143-8-2647.
8. S. O. Kim, Y. Orii, D. Lloyd, M. N. Hughes and R. K. Poole, Anoxic function for the Escherichia coli flavohaemoglobin (Hmp): reversible binding of nitric oxide and reduction to nitrous oxide, FEBS Lett., 1999, 445, 389–394,  DOI:10.1016/S0014-5793(99)00157-X.
9. A. M. Gardner, R. A. Helmick and P. R. Gardner, Flavorubredoxin, an Inducible Catalyst for Nitric Oxide Reduction and Detoxification in Escherichia coli, J. Biol. Chem., 2002, 277, 8172–8177,  DOI:10.1074/jbc.M110471200.
10. J. Wang, C. E. Vine, B. K. Balasiny, J. Rizk, C. L. Bradley, M. Tinajero-Trejo, R. K. Poole, L. L. Bergaust, L. R. Bakken and J. A. Coole, The role of the hybrid cluster protein, Hcp and its reductase, Hcr, in high affinity nitric oxide reduction that protects anaerobic cultures of Escherichia coli against nitrosative stress, Mol. Microbiol., 2016, 100, 877–892,  DOI:10.1111/mmi.13356.
11. N. M. Dixon and D. B. Kell, The control and measurement of ‘CO₂’ during fermentations, J. Microbiol. Methods, 1989, 10, 155–179,  DOI:10.1016/0167-7012(89)90048-1.
12. X. Chen, R. Hu, L. Hu, Y. Huang, W. Shi, Q. Wei and Z. Li, Portable Analytical Techniques for Monitoring Volatile Organic Chemicals in Biomanufacturing Processes: Recent Advances and Limitations, Front. Chem., 2020, 8, 837,  DOI:10.3389/fchem.2020.00837.
13. S. Sachse, A. Bockisch, U. Enseleit, F. Gerlach, K. Ahlborn, T. Kuhnke, U. Rother, E. Kielhorn, P. Neubauer, S. Junne and W. Vonau, On the use of electrochemical multi-sensors in biologically charged media, J. Sens. Sens. Syst., 2015, 4, 295–303,  DOI:10.5194/jsss-4-295-2015.
14. M. D. Brown and M. H. Schoenfisch, Electrochemical Nitric Oxide Sensors: Principles of Design and Characterization, Chem. Rev., 2019, 119, 11551–11575,  DOI:10.1021/acs.chemrev.8b00797.
15. T. W. Smith and M. Hippler, Cavity-Enhanced Raman Spectroscopy in the Biosciences: In situ, Multicomponent, and Isotope Selective Gas Measurements To Study Hydrogen Production and Consumption by Escherichia coli, Anal. Chem., 2017, 89, 2147–2154,  DOI:10.1021/acs.analchem.6b04924.
16. G. D. Metcalfe, S. Alahmari, T. W. Smith and M. Hippler, Cavity-Enhanced Raman and Helmholtz Resonator Photoacoustic Spectroscopy to Monitor the Mixed Sugar Metabolism of E. coli, Anal. Chem., 2019, 91, 13096–13104,  DOI:10.1021/acs.analchem.9b03284.
17. R. Salter, J. Chu and M. Hippler, Cavity-enhanced Raman spectroscopy with optical feedback cw diode lasers for gas phase analysis and spectroscopy, Analyst, 2012, 137, 4669–4676,  10.1039/c2an35722d.
18. R. Keiner, T. Frosch, T. Massad, S. Trumbore and J. Popp, Enhanced Raman multigas sensing - a novel tool for control and analysis of ¹³CO₂ labeling experiments in environmental research, Analyst, 2014, 139, 3879–3884,  10.1039/C3AN01971C.
19. R. Keiner, M. Hermann, K. Küsel, J. Popp and T. Frosch, Rapid monitoring of intermediate states and mass balance of nitrogen during denitrification by means of cavity enhanced Raman multi-gas sensing, Anal. Chim. Acta, 2015, 864, 39–47,  DOI:10.1016/j.aca.2015.02.007.
20. M. Hippler, Cavity-Enhanced Raman Spectroscopy of Natural Gas with Optical Feedback cw-Diode Lasers, Anal. Chem., 2015, 87, 7803–7809,  DOI:10.1021/acs.analchem.5b01462.
21. A. Sieburg, T. Jochum, S. Trumbore, J. Popp and T. Frosch, Onsite cavity enhanced Raman spectrometry for the investigation of gas exchange processes in the Earth's critical zone, Analyst, 2017, 142, 3360–3369,  10.1039/C7AN01149K.
22. A. Blohm, S. Kumar, A. Knebl, M. Herrmann, K. Küsel, J. Popp and T. Frosch, Activity and electron donor preference of two denitrifying bacterial strains identified by Raman gas spectroscopy, Anal. Bioanal. Chem., 2021 DOI:10.1007/s00216-021-03541-y.
23. G. D. Metcalfe, T. W. Smith and M. Hippler, On-line analysis and in situ pH monitoring of mixed acid fermentation by Escherichia coli using combined FTIR and Raman techniques, Anal. Bioanal. Chem., 2020, 412, 7303–7319,  DOI:10.1007/s00216-020-02865-5.
24. J. Mohn, B. Tuzson, A. Manninen, N. Yoshida, S. Toyoda, W. A. Brand and L. Emmenegger, Site selective real-time measurements of atmospheric N₂O isotopomers by laser spectroscopy, Atmos. Meas. Tech., 2012, 5, 1601–1609,  DOI:10.5194/amt-5-1601-2012.
25. P. Wunderlin, M. F. Lehmann, H. Siegrist, B. Tuzson, A. Joss, L. Emmenegger and J. Mohn, Isotope Signatures of N₂O in a Mixed Microbial Population System: Constraints of N₂O Producing Pathways in Wastewater Treatment, Environ. Sci. Technol., 2013, 47, 1339–1348,  DOI:10.1021/es303174x.
26. H. Moser, W. Pölz, J. P. Waclawek, J. Ofner and B. Lendl, Implementation of a quantum cascade laser-based gas sensor prototype for sub-ppmv H₂S measurements in a petrochemical process gas stream, Anal. Bioanal. Chem., 2017, 409, 729–739,  DOI:10.1007/s00216-016-9923-z.
27. S. W. Sharpe, T. J. Johnson, R. L. Sams, P. M. Chu, G. C. Rhoderick and P. A. Johnson, Gas-Phase Databases for Quantitative Infrared Spectroscopy, Appl. Spectrosc., 2004, 58, 1452–1461,  DOI:10.1366/0003702042641281.
28. L. S. Rothman, I. E. Gordon, Y. Babikov, A. Barbe, D. Chris Benner, P. F. Bernath, M. Birk, L. Bizzocchi, V. Boudon, L. R. Brown, A. Campargue, K. Chance, E. A. Cohen, L. H. Coudert, V. M. Devi, D. B. Drouin, A. Fayt, J.-M. Flaud, R. R. Gamache, J. J. Harrison, J.-M. Hartmann, C. Hill, J. T. Hodges, D. Jacquemart, A. Jolly, J. Lamouroux, R. J. Le Roy, G. Li, D. A. Long, O. M. Lyulin, C. J. Mackie, S. T. Massie, S. Mikhailenko, H. S. P. Müller, O. V. Naumenko, A. V. Nikitin, J. Orphal, V. Perevalov, A. Perrin, E. R. Polovtseva, C. Richard, M. A. H. Smith, E. Starikova, K. Sung, S. Tashkun, J. Tennyson, G. C. Toon, V. G. Tyuterev and G. Wagner, The HITRAN2012 molecular spectroscopic database, J. Quant. Spectrosc. Radiat. Transfer, 2013, 130, 4–50,  DOI:10.1016/j.jqsrt.2013.07.002.
29. C. Mohr, C. L. Spencer and M. Hippler, Inexpensive Raman Spectrometer for Undergraduate and Graduate Experiments and Research, J. Chem. Educ., 2010, 87, 326–330,  DOI:10.1021/ed800081t.
30. Y. Ryabenkova, N. Jadav, M. Conte, M. Hippler, N. Reeves-McLaren, P. D. Coates, P. Twigg and A. Paradkar, Mechanisms of Hydrogen-Bonded Complex Formation between Ibuprofen and Nanocrystalline Hydroxyapatite, Langmuir, 2017, 33, 2965–2976,  DOI:10.1021/acs.langmuir.6b04510.
31. M. D. Fontana, K. B. Mabrouk and T. H. Kauffmann, Raman spectroscopic sensors for inorganic salts, in Spectroscopic Properties of Inorganic and Organometallic Compounds: Techniques, Materials and Applications, Royal Society of Chemistry, Cambridge, U.K., 2013, vol. 44, pp. 40–67,  10.1039/9781849737791-00040.
32. M. Hippler and G. D. Metcalfe, Using activities to correct the Henderson-Hasselbalch equation, Bunsenmagazin, 2020, 22, 102–105,  DOI:10.26125/y7p7-an56.
33. M. B. Shinn, Colorimetric Method for Determination of Nitrate, Ind. Eng. Chem., Anal. Ed., 1941, 13, 33–35,  DOI:10.1021/i560089a010.
34. I. L. Marr, A. Kindness and M. S. Cresser, Measurement of ¹⁴N: ¹⁵N ratios by Fourier transform infrared spectrometry, Analyst, 1987, 112, 1491–1494,  10.1039/AN9871201491.
35. B. H. Bleakley and J. M. Tiedje, Nitrous Oxide Production by Organisms Other than Nitrifiers or Denitrifiers, Appl. Environ. Microbiol., 1982, 44, 1342–1348,  DOI:10.1128/AEM.44.6.1342-1348.1982.
36. T. M. Khlebodarova, N. A. Ree and V. A. Likhoshvai, On the control mechanisms of the nitrite level in Escherichia coli cells: the mathematical model, BMC Microbiol., 2016, 16, S7,  DOI:10.1186/s12866-015-0619-x.
37. V. Stewart, Nitrate respiration in relation to facultative metabolism in enterobacteria, Microbiol. Rev., 1988, 52, 190–232,  DOI:10.1128/mr.52.2.190-232.1988.
38. V. Bonnefoy and J. A. Demoss, Nitrate reductases in Escherichia coli, Antonie van Leeuwenhoek, 1994, 66, 47–56,  DOI:10.1007/BF00871632.
39. V. Stewart, Y. Lu and A. J. Darwin, Periplasmic nitrate reductase (NapABC enzyme) supports anaerobic respiration by Escherichia coli K-12, J. Bacteriol., 2002, 184, 1314–1323,  DOI:10.1128/JB.184.5.1314-1323.2002.
40. H. Wang, C. P. Tseng and R. P. Gunsalus, The napF and narG Nitrate Reductase Operons in Escherichia coli Are Differentially Expressed in Response to Submicromolar Concentrations of Nitrate but Not Nitrite, J. Bacteriol., 1999, 181, 5303–5308,  DOI:10.1128/JB.181.17.5303-5308.1999.
41. J. Cole, Nitrate reduction to ammonia by enteric bacteria: redundancy, or a strategy for survival during oxygen starvation?, FEMS Microbiol. Lett., 1996, 136, 1–11,  DOI:10.1016/0378-1097(95)00480-7.
42. L. Chang, L. I. Wei, J. P. Audia, R. A. Morton and H. E. Schellhorn, Expression of the Escherichia coli NRZ nitrate reductase is highly growth phase dependent and is controlled by RpoS, the alternative vegetative sigma factor, Mol. Microbiol., 2002, 34, 756–766,  DOI:10.1046/j.1365-2958.1999.01637.x.
43. H. Wang and R. P. Gunsalus, The nrfA and nirB Nitrite Reductase Operons in Escherichia coli Are Expressed Differently in Response to Nitrate than to Nitrite, Genet. Mol. Biol., 2000, 182, 5813–5822,  DOI:10.1128/jb.182.20.5813-5822.2000.
44. B. Weiss, Evidence for Mutagenesis by Nitric Oxide during Nitrate Metabolism in Escherichia coli, J. Bacteriol., 2006, 188, 829–833,  DOI:10.1128/JB.188.3.829-833.2006.
45. S. R. Poock, E. R. Leach, J. W. B. Moir, J. A. Cole and D. J. Richardson, Respiratory Detoxification of Nitric Oxide by the Cytochrome c Nitrite Reductase of Escherichia coli, J. Biol. Chem., 2002, 277, 23664–23669,  DOI:10.1074/jbc.M200731200.
46. M. Kaldorf, K. H. Linne von Berg, U. Meier, U. Servos and H. Bothe, The reduction of nitrous oxide to dinitrogen by Escherichia coli, Arch. Microbiol., 1993, 160, 432–439,  DOI:10.1007/BF00245303.
47. J. van der Plas, K. J. Hellingwerf, H. G. Seijen, J. R. Guest, J. H. Weiner and W. N. Konings, Identification and localization of enzymes of the fumarate reductase and nitrate respiration systems of Escherichia coli by crossed immunoelectrophoresis, J. Bacteriol., 1983, 153, 1027–1037,  DOI:10.1128/jb.153.2.1027-1037.1983.
48. W. J. Dobrogosz, Altered End-Product Patterns and Catabolite Repression in Escherichia coli, J. Bacteriol., 1966, 91, 2263–2269,  DOI:10.1128/JB.91.6.2263-2269.1966.
49. N. J. Gilberthorpe and R. K. Poole, Nitric oxide homeostasis in Salmonella typhimurium, J. Biol. Chem., 2008, 283, 11146–11154,  DOI:10.1074/jbc.M708019200.
50. C. E. Vine, S. K. Purewal and J. A. Cole, NsrR-dependent method for detecting nitric oxide accumulation in the Escherichia coli cytoplasm and enzymes involved in NO production, FEMS Microbiol. Lett., 2011, 325, 108–114,  DOI:10.1111/j.1574-6968.2011.02385.x.
51. L. Page, L. Griffiths and J. A. Cole, Different physiological roles of two independent pathways for nitrite reduction to ammonia by enteric bacteria, Arch. Microbiol., 1990, 154, 349–354,  DOI:10.1007/BF00276530.
52. R. H. Gillette and E. H. Eyster, The Fundamental Rotation-Vibration Band of Nitric Oxide, Phys. Rev., 1939, 56, 1113–1119,  DOI:10.1103/PhysRev.56.1113.
53. A. Abou-Jaoudé, M. Chippaux and M.-C. Pascal, Formate-Nitrite Reduction in Escherichia coli K-12. 1. Physiological Study of the System, Eur. J. Biochem., 1979, 95, 309–314,  DOI:10.1111/j.1432-1033.1979.tb12966.x.
54. J. S. McDowall, B. J. Murphy, M. Haumann, T. Palmer, F. A. Armstrong and F. Sargent, Bacterial formate hydrogenlyase complex, Proc. Natl. Acad. Sci. U. S. A., 2014, 111, 3948–3956,  DOI:10.1073/pnas.1407927111.
55. A. Pecher, F. Zinoni, C. Jatisatienr, R. Wirth, H. Hennecke and A. Böck, On the redox control of synthesis of anaerobically induced enzymes in enterobacteriaceae, Arch. Microbiol., 1983, 136, 131–136,  DOI:10.1007/BF00404787.
56. R. Rossmann, G. Sawers and A. Böck, Mechanism of regulation of the formate-hydrogenlyase pathway by oxygen, nitrate and pH: definition of the formate regulon, Mol. Microbiol., 1991, 5, 2807–2814,  DOI:10.1111/j.1365-2958.1991.tb01989.x.
57. R. Grosz and G. Stephanopoulos, Statistical mechanical estimation of the free energy of formation of E. coli biomass for use with macroscopic bioreactor balances, Biotechnol. Bioeng., 1983, 25, 2149–2163,  DOI:10.1002/bit.260250904.
58. H. Abaibou, G. Giordano and M.-A. Mandrand-Berthelot, Supression of Escherichia coli formate hydrogenlyase activity by trimethylamine N-oxide is due to drainage of the inducer formate, Microbiol., 1997, 143, 2657–2664,  DOI:10.1099/00221287-143-8-2657.

---

Footnote

† Electronic supplementary information (ESI) available: S.1. Key nitrate and nitrite reduction enzymes, S.2. M9 medium formulation, S.3. FTIR spectroscopy of CO₂ and ethanol, S.4. cavity enhanced Raman spectroscopy (Experimental details, spectral fitting procedures and calibration plots), S.5. liquid phase Raman spectroscopy (Experimental details of the home-built Raman spectrometer, spectral fitting procedures and calibration plots) and S.6. analysis of bacterial culture samples (nitrite colorimetry, ¹⁴N/¹⁵N-ammonium analysis). See DOI: 10.1039/d1an01261d

---