the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The isotopic signatures of nitrous oxide produced by eukaryotic and prokaryotic phototrophs
Abstract. Prokaryotic and eukaryotic microscopic phototrophs ('microalgae') can synthesize the potent greenhouse gas and ozone depleting pollutant nitrous oxide (N2O). However, we do not know how much microalgae contribute to aquatic N2O emissions because these organisms co-occur with prolific N2O producers like denitrifying and nitrifying bacteria. Here we demonstrate for the first time that microalgae produce distinct N2O isotopic signatures that will enable us to fill this knowledge gap. The eukaryotes Chlamydomonas reinhardtii and Chlorella vulgaris, and the prokaryote Microcystis aeruginosa synthesized N2O 265–755 nmol·g-DW-1·h-1 when in darkness and supplied with 10 mM nitrite (NO2-). The N2O isotopic composition (δ15N, δ18O, and site preference, SP) of each species was determined using a modified off-axis integrated-cavity-output spectroscopy analyser with an offline sample purification and homogenisation system. The SP values differed between eukaryotic and prokaryotic algae (25.8 ± 0.3 ‰ and 24.1 ± 0.2 ‰ for C. reinhardtii and C. vulgaris, respectively vs 2.1 ± 3.0 ‰ for M. aeruginosa), as did bulk isotope values. Both values differ from SP produced by denitrifiers. This first characterization of the N2O isotopic fingerprints of microscopic phototrophs suggests that SP-N2O could be used to untangle algal, bacterial, and fungal N2O production pathways. As the presence of microalgae could influence N2O dynamics in aquatic ecosystems, field monitoring is also needed to establish the occurrence and significance of microalgal N2O synthesis under relevant conditions.
- Preprint
(899 KB) - Metadata XML
-
Supplement
(167 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-2337', Anonymous Referee #1, 08 Sep 2025
-
AC1: 'Reply on RC1', Maxence Plouviez, 22 Sep 2025
We thank the Referee for praising the impact of the work and for his/her comments. Our answers to the comments can be found below.
Firstly, did the authors perform any kind of abiotic, illuminated control? Recent work has shown that sunlight can drive abiotic photochemical N2O production (Leon-Palmero et al., 2025), and it seems possible that this was occurring in the authors’ experiments.
We did not include an abiotic illuminated control during this study. We followed an established protocol for the production of N2O in laboratory microalgae cultures (Guieysse et al., 2013) and as we indicated Li 76, the cultures were kept in darkness. Consequently, it is unlikely that abiotic photochemical N2O production occurred in our samples. From our previous work (Guieysse et al., 2013; Plouviez et al., 2017) abiotic N2O production remained always low in our experimental setting.
However, we agree that under real setting (i.e. natural environment) this abiotic production must be taken into consideration. We will therefore modify section 3. Environmental implications to: 'In natural environments, N2O can be abiotically produced photochemically (Lean-Palmero et al., 2025). In addition, N2O can be produced and consumed by organisms….'
Secondly, the authors added 10 mM NaNO2 to their cultures, which is orders of magnitude higher than the amount of nitrite in natural aquatic environments. Did the authors do any kind of experiment, feeding the cultures lower levels of nitrite to ascertain if the organisms would still produce N2O under less nutrient-laden conditions?
We agree that the concentration of NaNO2 used is significantly higher than what would be expected in natural environments. As our focus was to identify the isotopic signature, we used a protocol known to trigger a strong N2O production in microalgae and cyanobacteria (Guieysse et al., 2013; Plouviez et al., 2017; Fabisik et al., 2023) in order to facilitate ease of detection (over relevance). It should also be noted that our prior work revealed a linear correlation between NaNO2 concentration (up to 12 mM) and N2O production in C. vulgaris, C. reinhardtii and M. aeruginosa, with N2O production being 3-5 fold lower at 3 mM than at 12 mM.
The authors provide the N2O site preference produced by each organism, but to incorporate this process into models, it is critical to also know the δ(15Nα) and δ(15Nβ) as well. What were the δ(15Nα) and δ(15Nβ) of the N2O produced by each organism, and what was the δ(15N) of the nitrite that they were supplied? This would allow us to calculate an isotope effect and thus incorporate this process into biogeochemical models.
We agree that establishing the fractionation factors for nitrogen and oxygen during the reduction of NO2- to N2O would be useful – not just for biogeochemical models but also for elucidating the different biochemical pathways of reduction between eukaryotes and prokaryotes that our N2O isotope and isotopomer results suggest. A comparison of the ‘starting’ isotopic enrichment of NO2- v the ‘product’ enrichment reported in N2O would be a useful first step towards establishing such fractionation factors. However, we did not include this in our manuscript because the laboratory where these experiments were carried out at Massey University was closed and all the reagents used in the experiments thrown out. This is regrettable. However, we note that, because the same salt (with the same isotopic composition) was used across all experiments the uncertainty associated with this calculation will not alter the observed pattern of difference between the organisms nor the conclusions drawn from them.
We propose to use the range of d15N and d18O enrichment reported for NO2- salt solutions as possible end-member values to parameterise the potential range of fractionation factors for the different organismal N2O production pathways reported here. We will add these estimates (i.e. d15N-NO2 range of -16 to -61‰, and d18O-NO2 range of +6 to +14‰) to Table 1.
The authors point to other studies showing how phototrophs produce N2O from NO within the cell, but the vastly different site preferences of the eukaryotic and prokaryotic N2O suggest different mechanisms. Could the authors speculate on possible different reaction mechanisms for the two kinds of organisms, even though the intermediate (NO) may be the same?
The reviewer raised a good point. Proteins with similar functions (nitric oxide reductases) are involved in the reduction of NO into N2O. The eukaryotic and prokaryotic proteins are members of distinct families and consequently are structurally different (Hendriks et al., 2000). This could explain the differences between the cite preferences measured. While we prefer not to speculate as further experimental evidence would be needed, we will clarify that point in the manuscript (End of section 2.1).
‘Fabisik et al., (2023a) suggested a strong similarity between the biochemical pathways of N2O biosynthesis in the cyanobacterium M. aeruginosa and in the green microalgae C. reinhardtii. However, the vastly different site preferences between the eukaryotic and prokaryotic N2O measured (see Section 2.3 below) suggest different proteins are involved in the reduction of NO into N2O in eukaryotic and prokaryotic phototrophs. In eukaryotes, NOR belongs to the cytochrome P-450 family. In contrast prokaryotic NORs are related to the haem/copper cytochrome oxidases and these enzymes fall into two subclasses according to the electron donors used. Further research is therefore needed to confirm which protein catalyse the reduction of NO into N2O in cyanobacteria.’
Line-by-line comments:
Line 157: It seems possible that there may have also been photochemical N2O production in the authors’ experiments.
As mentioned above, this was unlikely in our experimental design, but we now acknowledge the photochemical production from natural ecosystems.
Line 174: What is “instrument-grade” N2?
This refers to a purity level of at least 99.99% N2. We will clarify that in the manuscript.
Line 233: What does “indicative” mean in this context?
The N2O mole fraction values reported were raw data and only used for sample processing purposes. We will clarify this is the revised manuscript.
Lines 283-284: Include the δ(15Nbulk) and δ(18O) from both gases in Table 3 to illustrate this.
We will update Table 3 to include all certified values from USGS51 and USGS52 as shown in the Table from the Supplement document.
Line 288: How does the uncertainty calculated this way compare to the standard deviation of replicate samples?
Typical values of the propagated uncertainty scale are around 1.2 ‰ (Table 1). The reproducibility for SP-N2O is around 0.4 ‰, with an accuracy of –0.3 ‰ (Figure 1). The analytical steps for the experiments included in the reproducibility assessment are identical to the analytical steps of the sample analysis for each sample measurement sequence. The propagated uncertainty is therefore a conservative uncertainty estimate.
Line 290: The term UREF_span2 should be multiplied by the correction factor, squared.
We will amend the manuscript and change the equation to
Utot = SQRT(Usam2 + UREF_a2 + UREF_b2 + UREF_span2 x FSPAN2+ Up-corr2 + UN2O-amount-corr2)
where FSPAN2 is the factor of the span correction.
Line 295: Not the standard error of the slope? Also, it would be highly useful to see a visual representation of these correction functions.
In addition to Figure 5 (i.e. N2O amount effect on SP-N2O values), we will include an additional Figure for the pressure effect. A preliminary version of the Figure is shown in the Supplement document, and we will include the following paragraph in the Appendix section.
‘We determined the pressure correction using four gas mixtures with N2O mole fractions of 380 ppb, 1080 ppb, 2100 ppb and 3300 ppb. The effect of increasing cell pressure on N2O mole fractions and all measured isotope species was linear (Figure 6). However, the slope of that effect changed with the N2O mole fraction. Slopes of the pressure corrections for N2O and SP-N2O were determined using polynomial fits (Figure 6).’
Line 455 and elsewhere: The formatting of the tables is confusing and difficult to read.
We will re-format the Tables (e.g. change the orientation to landscape to extend the size of the columns) to improve readability.
New References
Hendriks, J., Oubrie, A., Castresana, J., Urbani, A., Gemeinhardt, S., Saraste, M.: Nitric oxide reductases in bacteria. Biochim Biophys Acta – Bioenerg, 1459, 266-273, 2000.
Leon-Palmero, E., Morales-Baquero, R., Thamdrup, B., Löscher, C., and Reche, I.: Sunlight drives the abiotic formation of nitrous oxide in fresh and marine waters, Science, 387, 1198–1203, https://doi.org/10.1126/science.adq0302, 2025.
-
AC1: 'Reply on RC1', Maxence Plouviez, 22 Sep 2025
-
RC2: 'Comment on egusphere-2025-2337', Anonymous Referee #2, 22 Sep 2025
Final comments on Plouviez et al. 2025
General comments
Plouviez et al. take on a significant problem within the biogeosciences – namely, that the N2O budget is not closed, and one of the main hurdles in closing the budget is accurately accounting for all sources and sinks of N2O since it can be produced / consumed by multiple biotic and abiotic pathways. A powerful tool in this space is measuring “Site Preference” (SP), which quantifies the relative ‘preference’ of 15N for the central (‘alpha’) or outer (‘beta’) site in the asymmetrical, linear N2O molecule. Therefore, many groups have been working to systematically measure the SP of all known sources and sinks of N2O in an effort to close the N2O budget, as well as identify sources of N2O that may be mitigated to prevent greenhouse gas emissions.
Plouviez et al. measure the N2O SP of eukaryotic and bacterial photosynthesizers, which have been shown to produce N2O outside of the metabolic pathways that N2O production has been typically attributed to (denitrification and nitrification, either by bacteria or fungi). Measuring this value is particularly important for understanding N2O cycling in the oceans, since denitrifiers, nitrifiers, and algae all coexist in complex microbial assemblages – therefore, finding potentially unique SP signatures for algae may help disentangle complex marine N2O cycling. They state that they describe a new method for the accurate laser-based analysis of N2O isotopes, which enables them to conduct novel SP measurements of algal N2O. They find significantly different SP signatures for the eukaryotic algae (C. reinhardtii and C. vulgaris) compared to the cyanobacteria (M. aeruginosa).
I have two main comments regarding this manuscript. The first is about the technical aspects of the measurement. The second is about the interpretation of the isotopic signatures.
For the first point, I cannot independently evaluate the quality of the data presented because it lacks key outputs that would enable independent calculation of this. I assume that authors are measuring the major isotopologues 14N14N16O (446), 14N15N16O (456), 15N14N16O (546) and 14N14N18O (448) and not the rarer clumped species (i.e. 14N15N18O), though this is never explicitly stated. SP is calculated as the relative difference between the 15N isotopologues (SP = d456 – d546) and the bulk nitrogen isotopic composition is the average of the alpha and beta sites (d15Nbulk = (d456 – d546)/2; see Kanterova et al. 2022 GCA, for example, for calculations) – therefore, understanding issues of sample bracketing, variations among samples, and etc. can be masked by reporting SP only. This is because variations in SP can be driving by variations in one isotopologue alone, since SP simply describes a relative difference in 456 and 546. In addition, as noted in Griffith 2018 GMT, several commercial manufacturers offer optical analyzers based on laser or FTIR spectroscopy that report results in various ways – as an isotopologue mole fraction and/or total mole fractions and/or in ‘traditional’ isotope delta values. Plouviez et al. do not report the equations used to convert from raw, instrument measurements to final delta values. They also do not report d456 and d546 (also denoted as d15N-alpha and d15N-beta), nor do they show that the calculated d15Nbulk from these values match the measured d15Nbulk values. They do report some necessary data in reporting “a new method” – i.e. Figure 1 and 5 – but, again. they do not report their full dataset and only their final calculated values. For example, Table 1 gives the averaged isotopic measurements across all replicates for each species, but the individual measurements behind each average are not in the main text or supplement. Therefore, it is difficult to independently evaluate the quality of their data. I would encourage the authors to publish a more complete dataset, as well as equations involved in converting from raw, instrument measurements to final, reported delta values. This could be amended to the existing supplemental.
For the second point, I would: 1) Encourage the authors to comment more on the potential mechanism behind the large difference in SP values between the eukaryotic vs. bacterial algal strains; and 2) Have some clarifying questions regarding controls in their experimental systems. As the authors are likely aware of, in both eukaryotic and bacterial algae, it is thought that there are primarily two sources of N2O: flavodiiron proteins (FLV) and cytochrome p450s (CYP55). FLVs are used in pseudo-cyclic electron flow for Photosystem I (PSI) photoprotection, where electrons are put onto O2 instead of being used to generate NADPH. It has been shown that NO can be reduced instead of O2, generating N2O in the process (Burlacot et al 2020 PNAS). CYP55s are a broad class of enzymes involved in multiple metabolic pathways, including pigment biosynthesis and lipid metabolism – i.e., reactions not involved in the light reactions of photosynthesis. Hence, as noted by the authors, it has been shown in C. reinhardtii that FLV produce N2O in the light, while CYP55 produces N2O in the dark (Burlacot et al. 2020 PNAS). Due to similarities between the species, C. vulgaris should use a similar pathway, as noted by the authors. Prior work by some of the authors (Fabisik et al. 2023 Biogeosciences) performed a BLASTP search on M. aeruginosa and found hits for FLV and CYP55, suggesting that similar pathways exist in this strain as well.
Plouviez et al. perform all cell suspensions in the dark – this should isolate the CYP55 signal. The SP signals from C. reinhardtii and C. vulgaris are similar to that of the fungal nitric oxide reductase (Figure 2), which also belongs to the CYP55 family. However, the SP signals from M. aeruginosa are quite different and better match the bacterial nitric oxide reductase (Figure 2). One interpretation of their results is that CYP55 from eukaryotic and bacterial algae are quite different, and that is reflected in their N2O SP values – this appears to be the primary interpretation that the authors make, though they do not attribute it to the enzyme explicitly. Alternatively, in M. aeruginosa, since they note that the pathway has not been fully ‘elucidated,’ non-CYP55 sources of N2O may be possible. Potentially relevant, an enzyme called flavohemoglobin protein (FHP) has recently been measured for N2O SP (Wang et al. 2024 PNAS). FHP is similar to FLV as they both have flavins as a co-factor – diflavins like FLV have two, while FHP has a flavin and heme cofactor. Measured N2O SP values in Wang et al. 2024 PNAS of FHP from P. aeruginosa, A. baumannii and S. aureus are similar to those measured from M. aeruginosa in this paper (roughly 0 to 15‰ in Wang et al., depending on strain, compared to 2±7‰ for M. aeruginosa in this paper). The authors should also explicitly note if M. aueruginosa has a nitric oxide reductase (NOR) or not, since that would aid in interpretation of this unique signal. In addition, it may be potentially relevant that the standard deviation of their reported SP values from M. aeruginosa is much larger than that of C. reinhardtii and C. vulgaris, though they do not report the non-averaged data nor the 456 / 546 data, so this is difficult to interpret. That may also help better interpret why the SP values of M. aeruginosa are so different.
For this point, did the authors repeat their experiments in the light? Given the established light-dependent nature of N2O production in C. reinhardtii, if FLV does have a different N2O SP, one would expect to see a shift in the N2O SP of C. reinhardtii.
In addition, other enzymes besides NOR, FLV, and CYP55 can produce N2O. Currently, nitric oxide reductases (NOR), P450nor, cytochrome P460, cytochrome p450 (CYP55), cytochrome c554, flavodiiron proteins (FLVs) and flavohemoglobin proteins (FHPs) have been shown to produce N2O as a direct product of an enzymatic reaction (see Ferousi et al. 2020 Chem Rev, Kuypers et al 2018 Nat Rev Microbiol and Poole & Hughes 2000 Mol Microbiol for review). Did the authors attempt to check if their wild-type (WT) strains have genes encoding any of these potential enzymatic sources? This check can be doing through searches like BLASTP, qPCR, RNAseq, or other similar techniques for working on non-genetically tractable strains (i.e. strains where making clean deletions of a certain gene are difficult).
Finally, regarding experimental controls, it is established that N2O can be produced abiotically (i.e. ‘chemodenitrification’ Stanton et al. 2018 Geobiology), and this process is strongly pH-dependent, where acidic pHs produce nitric oxide radicals that can then be further reduced to N2O (i.e. Su et al 2019 ES&T). Did the authors control or check pH of their growth media? Though the media composition is given, there is no indication that the pH of the system was checked prior to incubation, or what the target pH of their media is. In addition, did the authors perform any no-cell controls, where the media was incubated with no cells? I may have missed this, but it does not appear that the authors did this. In addition, N2O can be formed readily from NO radicals, which makes it important to control for all sources of NO radicals, particularly in wild-type (WT) strains. Both bacteria and eukaryotes can create NO through a diverse set of nitric oxide synthases (Forstermann & Sessa 2011 Eur Heart J), and these NO radicals can spontaneously react to form N2O in the absence of oxygen. Did the authors check for these potential NO sources in their strains?
Overall, Plouviez et al. tackle an important problem in the biogeosciences – constraining the N2O SP of eukaryotic and bacterial photosynthesizers, which produce N2O outside of the metabolic pathways that N2O production has been typically attributed to (denitrification and nitrification, either by bacteria or fungi). Their work offers an important starting point for further, more detailed physiological work that will enable this measurement to be used to disentangle complex microbial communities of denitifiers, nitrifiers, and photoysnthesizers, helping the community close the N2O budget and disentangle complex marine N2O cycling.
Specific questions
The authors fine extremely depleted d15Nbulk values of about –100‰. This is outside of the range of their standards, and also of N2O in air (ranged from ~9 to 6‰ over the past 300 years; Park et al 2012 Nat Geosci). Did they use a very depleted source of nitrite? The d15N of the nitrite supplied should be included.
In Table 1, what does “F” mean in the footnotes? At first I thought it meant fraction consumed, but one value of F is 1200.
For Table 3, what are the d15N-alpha and d15N-beta values for the standards used?
For Figure 1, what are the d15N-alpha and d15N-beta values, not just the SP values? In addition, just to clarify, only USGS52 was measured over time, and not USGS51 as well? Related to this, I am slightly confused because Figure 5 shows only USGS51 and not USGS52, and Figure 6 suggests that both reference gases were measured regularly.
For Figure 2a, Wang et al. 2024 PNAS offers a more recent compilation of N2O SP measurements than Denk et al. 2017.
For Figure 2b, the authors are comparing their data vs. that from published denitfier data. In the text (line 113) and in the figure legend, which experimental denitrifer data are the authors comparing their data to? (In addition, the plot should specific ‘bacterial denitrifiers’ instead of just ‘denitrifiers’). Multiple groups have measured bacterial denitrifiers and there is a larger range of values than they show in their figure. For example, see Wang et al. 2024 PNAS or Toyoda et al 2017 Mass Spectrom Rev for recent compilations. In addition, unless the authors are using nitrite with the exact same d15N as that study, one would not expect the d15N-N2O and d18O-N2O to be the same. Instead, the relative fractionation (i.e. 15e or 18e) are comparable, not the bulk values. Therefore, the epsilons should be calculated and plotted instead.
For Figure 5, this is something where showing the full suite of data (d15N-alpha, d15N-beta, SP, d15N-N2O and d18O-N2O) would be helpful. The legend says that these are all measurements of USGS51, which should have a SP value of –1.67 at their target of 1000 ppb (1 ppm). However, at that pressure, the SP measured is between –25 and –30‰. Since this is showing “measurement bias,” am I to understand that the SP value being measured is between –26.67 and –31.67‰? In addition, what happened to the 5/10/2023 run? The authors do not talk about it in the figure legend or text. Was data from that run discarded? In addition, it would be helpful to show the “experimentally determined, linear correction function” (Line 276) to show how they correct for variation in cell pressure, and how that consistent or not consistent that correction was for all experiments.
Technical corrections
There’s a little floating “1)” in the upper left corner for Figure 3. Is this supposed to be there? And, the “2” in H2O and CO2 in the figure are not subscripted.
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
2,232 | 85 | 15 | 2,332 | 25 | 23 | 32 |
- HTML: 2,232
- PDF: 85
- XML: 15
- Total: 2,332
- Supplement: 25
- BibTeX: 23
- EndNote: 32
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
The manuscript titled “The isotopic signatures of nitrous oxide produced by eukaryotic and prokaryotic phototrophs” by Plouviez et al. aimed to quantify the isotopic content of nitrous oxide (N2O) produced by phototrophs (two eukaryotic and one prokaryotic). The authors achieved that goal by culturing the eukaryotes Chlamydomonas reinhardtii and Chlorella vulgaris, and the prokaryote Microcystis aeruginosa, then measuring the bulk δ(15N) and δ(18O) of the N2O produced by each culture as well as the site-specific δ(15N) in N2O (isotopomers). As part of this work, the authors developed a method for measuring N2O isotopomers more carefully with laser-based analysis by removing gases with matrix effects and controlling the amount of N2O seen by the laser system, both of which have been shown to have profound effects on the resulting isotopomer measurements.
The authors found that the eukaryotes, C. reinhardtii and C. vulgaris, both produce N2O with a site preference of ~25 ‰, while the prokaryote M. aeruginosa produced N2O with a site preference closer to 0 ‰. This highly useful result will facilitate source partitioning of N2O produced by phototrophs, because it provides two distinct endmembers for eukaroytic and prokaryotic sources, respectively. These results also complicate the interpretation of N2O site preferences in nature, since nitrification tends to also produce N2O with a high positive site preference (~30 ‰), and denitrification tends to produce N2O with a site preference around 0 ‰.
Altogether, this is a concise, impactful study, and I highly recommend its publication. I only have a few major concerns and some suggestions for improving the manuscript.
Firstly, did the authors perform any kind of abiotic, illuminated control? Recent work has shown that sunlight can drive abiotic photochemical N2O production (Leon-Palmero et al., 2025), and it seems possible that this was occurring in the authors’ experiments.
Secondly, the authors added 10 mM NaNO2 to their cultures, which is orders of magnitude higher than the amount of nitrite in natural aquatic environments. Did the authors do any kind of experiment, feeding the cultures lower levels of nitrite to ascertain if the organisms would still produce N2O under less nutrient-laden conditions?
The authors provide the N2O site preference produced by each organism, but to incorporate this process into models, it is critical to also know the δ(15Nα) and δ(15Nβ) as well. What were the δ(15Nα) and δ(15Nβ) of the N2O produced by each organism, and what was the δ(15N) of the nitrite that they were supplied? This would allow us to calculate an isotope effect and thus incorporate this process into biogeochemical models.
The authors point to other studies showing how phototrophs produce N2O from NO within the cell, but the vastly different site preferences of the eukaryotic and prokaryotic N2O suggest different mechanisms. Could the authors speculate on possible different reaction mechanisms for the two kinds of organisms, even though the intermediate (NO) may be the same?
Line-by-line comments:
Line 157: It seems possible that there may have also been photochemical N2O production in the authors’ experiments.
Line 174: What is “instrument-grade” N2?
Line 233: What does “indicative” mean in this context?
Lines 283-284: Include the δ(15Nbulk) and δ(18O) from both gases in Table 3 to illustrate this.
Line 288: How does the uncertainty calculated this way compare to the standard deviation of replicate samples?
Line 290: The term UREF_span2 should be multiplied by the correction factor, squared.
Line 295: Not the standard error of the slope? Also, it would be highly useful to see a visual representation of these correction functions.
Line 455 and elsewhere: The formatting of the tables is confusing and difficult to read.
References
Leon-Palmero, E., Morales-Baquero, R., Thamdrup, B., Löscher, C., and Reche, I.: Sunlight drives the abiotic formation of nitrous oxide in fresh and marine waters, Science, 387, 1198–1203, https://doi.org/10.1126/science.adq0302, 2025.